bVGcZiIhb VcY
heIc igVchedgi hiUYIZh dc
IcYIUb iIc dmIYZ
AlH MoraaC HakHmH
DarwHM CollDgD
UMHUDrsHty of CamArHCgD
A dissertation submitted for the degree of Doctor of Philosophy
at the University of Cambridge
February 2011
In loving mzmory of my GrvnyfvthzrsA Cyrus vny eztzr
Vwstrvxt
This dissertation reports on a detailed systematic study of the investigation into
using Indium Oxide based materials in next generation spin-transport electronic ap-
plications.
Initial studies focused on the optimisation of the electrical properties of Indium
Oxide (InGOH) and Tin(Sn)-doped Indium Oxide (ITO) thin films grown using DC
magnetron sputtering. The manipulation of various deposition parameters allowed the
electrical properties to be tuned effectively.
With the desire to create multi-functional spintronic devices, a dilute magnetic ox-
ide system is developed where the InGOH and ITO matrices are doped with low levels
of transition metals, in particular, Co. Using a number of characterisation techniques,
the origins of the magnetic response in these thin films is explored in great detail. In
particular, powerful probes such as x-ray and optical magnetic circular dichroism are
utilised. The major finding from these investigations is that the magnetism does not
necessarily emanate from the Co dopants alone. In fact, Co dopants give a strictly
paramagnetic response, suggesting that the magnetism observed may be a result of
polarised electrons in localised donor states in the InGOH and ITO hosts. Therefore, we
believe that the origins of magnetism in these films is related to a hybridisation and
charge transfer of electrons from a broad donor/defect-derived impurity band to a band
of unoccupied 3y states at the Fermi level. The emergence of a very weak magnetic
signal in pure ITO raises further questions as to the true origins of the ferromagnetic
behaviour and supports a defect-related mechanism.
To explore the suitability of ITO for a future in spintronics further, the perfor-
mance of some metal ferromagnet/oxide multilayered structures was investigated. The
investigations revealed a significant contribution to both the magnetic and magneto-
transport properties from a superparamagnetic component giving some insight into the
importance of the quality of interfaces between the metal ferromagnet/oxide layers and
heterostructures.
Using a three-dimensional focused-ion beam etching technique to fabricate submi-
cron spin-valve devices with ITO spacer layers, current-perpendicular-to-plane mag-
netoresistance measurements were carried out to estimate the spin diffusion length of
ITO at room temperature. In conjunction with a simplified Valet-Fert model, a spin
asymmetry ratio for Co of 0.55 and spin diffusion length of 6±1 nm in semiconducting
ITO at room temperature was estimated. These findings imply that spin information
can be conserved and transported through InGOH and ITO even up to and beyond room
temperature.
erefvxe
This dissertation is submitted for the degree of Doctor of Philosophy at the Uni-
versity of Cambridge. Except where specific references are made, this work is entirely
the result of my own work and includes nothing that is the outcome of work done in
collaboration. No part of this work has been, or is being submitted for any other quali-
fication at this, or any other university. This dissertation does not exceed 60,000 words.
Some of the significant results and interpretations derived in this work have been
published or submitted for publication and presented at international conferences, as
listed below. Original copies of these papers are reproduced in the appendix.
• “Estimation of the spin diffusion length of semiconducting Indium Tin Oxide
using Co/Indium Tin Oxide/Co spin valve junctions”
VC bC ]C gC ]vkimi, N. Banerjee, A. Aziz, J. W. A. Robinson, M. G. Blamire,
Applied Physics Letters NK, 102514, (2010).
• “Origin of magnetism in Cobalt-doped Indium Tin Oxide thin films”
VC bC ]C gC ]vkimi, F. Schoofs, R. Bali, N. A. Stelmashenko, M. G. Blamire,
S. Langridge, S. A. Cavill, G. van der Laan, S. S. Dhesi Physical Review B, (2010).
• “Probing the origins of magnetism in Dilute Magnetic Oxides”
VC bC ]C gC ]vkimi and S. S. Dhesi, Diamond Light Source Annual Report,
(2011).
• “Role of Superparamagnetism in [Co/ITO] multilayers”
VC bC ]C gC ]vkimi, M. Egilmez, J. W. A. Robinson, F. Schoofs, S. L. Sa-
honta, M. G. Blamire, submitted, (2011).
• “Donor band ferromagnetism in cobalt-doped indium oxide”
VC bC ]C gC ]vkimi, M. G. Blamire, S. M. Heald, M. Alshammari, M. Al-
Qahtani, D. S. Score, H. J. Blythe, A. M. Fox and G. A. Gehring, submitted,
(2011).
• “Indium Tin Oxide thin films for Spintronics”
VC bC ]C gC ]vkimi, N. Banerjee, A. Aziz, J. W. A. Robinson, F. Schoofs, M.
G. Blamire, S. A. Cavill, G. van der Laan, S. S. Dhesi, 17th Workshop on Oxide
Electronics, Poster Session, Awaji Island, Japan, (2010)
Collaborations with other members of the Device Materials Group have been fruit-
ful and I have been lucky enough to be part of some other peer-reviewed articles;
• “Strain dependent defect mediated ferromagnetism in Mn-doped and undoped
ZnO thin films”
F. Schoofs, T. Fix, VC bC ]C gC ]vkimi, S. S. Dhesi, G. van der Laan, S. A.
Cavill, S. Langridge, J. L. MacManus-Driscoll, M. G. Blamire, J. Appl. Phys.
108, 053911, (2010).
• “Origin of Magnetism in La and Fe doped SrTiOH−δ films”
M. Egilmez, G. W. Leung, VC bC ]C gC ]vkimi, M. G. Blamire, J. Appl. Phys.
108, 123912, (2010).
Vxknofileygements
The list of people who have supported me throughout these years is endless. I
convey my utmost appreciation to all those I speak of. In particular, I must thank my
supervisor Mark Blamire for his guidance throughout. Other members of the Device
Materials Group and beyond who deserve a special mention are Nadia Stelmashenko,
Jason Robinson, Mehmet Egilmez, Stuart Wimbush and Mary Vickers who have all
helped me with many scientific issues since I started working in the group. The office
of Room 221 has also seen some real characters passing through and I am happy to say
that I have made some dear friends during my time here - Rantej Bali, Sophie Harring-
ton, Jamie Young and Niladri Banerjee (along with Frank Schoofs who wishes he was
in Room 221) have been wonderful to both socialise and work with. I also thank the
Engineering and Physical Science Research Council and the Magnetism Group from
the Institute of Physics for their generous financial support.
I have spent many of my days here in Cambridge either on a football pitch or
Real tennis court. I therefore thank the Cambridge University Association Football
Club, Darwin College Football Club, Cambridge University Press Football Club and
the Cambridge University Real Tennis Club as well as the British Real Tennis Academy
for being such an excellent source of distraction.
A healthy portion of the work in this thesis has been performed in laboratories
elsewhere in the country. Our visits to Diamond Light Source were extremely valuable
and I thank all at Beamline IO6 for their devoted help, in particular Sarnjeet Dhesi.
My collaboration with the Magnetic Oxides group at the University of Sheffield under
the close supervision of Gillian Gehring has been extremely rewarding. I have spent a
great deal of time with them in Sheffield and thank them for always making me feel
so welcome and making our collaboration so fruitful. Gillian in particular has been a
wonderful source of knowledge and inspiration over the last one and a half years or so.
Sean Langridge has also been extremely kind in measuring a number of samples on the
SQUID in my absence at the Rutherford Appleton Laboratory.
I would also like to extend my gratitude to the Pearson family for being so good
to me during my time in Cambridge. In particular, I would like to offer my love and
appreciation to their wonderful daughter Laura.
This ramble would not be complete without me giving thanks to my wonderful
family. I have been away from them now for over 7 years and am grateful for their
patience in waiting for me to come home. That thought alone has been one of the
special things which has kept me going for all these years of study. My mother and
father have supported me throughout and I cannot thank them enough.
AMHR Hakimi
February 2011
Xontents
F Introyuxtion F
1.1 Setting the scene: Motivation for hpintronixs . . . . . . . . . . . . . . . 1
1.2 Dilute Magnetic Semiconductors . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Thesis Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
G Wvxkgrouny iheory vny [irst erinxiples M
2.1 Origins of Magnetism . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.1 Magnetic materials . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.2 Temperature dependence of magnetic susceptibility . . . . . . . 12
2.1.3 Ferromagnetism . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.1.4 Superparamagnetism . . . . . . . . . . . . . . . . . . . . . . . . 14
2.1.5 Competing interactions, frustration and randomness . . . . . . . 16
2.1.6 A quantum description of magnetism . . . . . . . . . . . . . . . 17
2.2 Existing theories for ferromagnetic origin in DMS and DMO systems . 18
2.2.1 Carrier-mediated exchange . . . . . . . . . . . . . . . . . . . . . 19
2.2.2 The Bound Magnetic Polaron Model . . . . . . . . . . . . . . . 20
2.3 Magnetoresistance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.3.1 Magnetoresistance of non-magnetic metals and ferromagnets . . 22
2.3.2 Anisotropic magnetoresistance . . . . . . . . . . . . . . . . . . . 23
H Zflperimentvl proxeyures G5
3.1 Principles of Sputtering . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.1.1 DC magnetron sputtering . . . . . . . . . . . . . . . . . . . . . 26
3.2 Substrates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2.1 Types of Substrates . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2.2 Substrate Preparation . . . . . . . . . . . . . . . . . . . . . . . 28
3.3 The MKII Sputter Deposition System . . . . . . . . . . . . . . . . . . . 28
3.4 Deposition parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.5 Thin Film Characterisation . . . . . . . . . . . . . . . . . . . . . . . . 29
3.5.1 X-Ray Diffraction studies . . . . . . . . . . . . . . . . . . . . . 29
3.5.2 Atomic Force Microscopy . . . . . . . . . . . . . . . . . . . . . . 31
3.5.3 Magnetometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.5.4 Electrical Measurements . . . . . . . . . . . . . . . . . . . . . . 34
7
3.5.5 Magnetic Circular Dichroism (MCD) . . . . . . . . . . . . . . . 36
I ihe Inyium dfliye system IF
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.2 The Indium Oxide system . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.3 Electrical Conduction mechanisms . . . . . . . . . . . . . . . . . . . . . 44
4.3.1 Free carrier concentration . . . . . . . . . . . . . . . . . . . . . 45
4.3.2 Carrier mobility . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.4 Previous Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.4.1 Effects of substrate temperature . . . . . . . . . . . . . . . . . . 48
4.4.2 Effects of oxygen partial pressure . . . . . . . . . . . . . . . . . 48
4.4.3 Post-growth annealing treatment . . . . . . . . . . . . . . . . . 49
4.4.4 Crystallographically oriented films . . . . . . . . . . . . . . . . . 49
4.5 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.6 Electrical properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.6.1 Brief discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.7 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 59
5 ihe fuest for v Yilute bvgnetix dfliye KH
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.2 Examination of ferromagnetism . . . . . . . . . . . . . . . . . . . . . . 64
5.2.1 Magnetometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.2.2 Magneto-optical effects . . . . . . . . . . . . . . . . . . . . . . . 64
5.2.3 The Anomalous Hall Effect . . . . . . . . . . . . . . . . . . . . 65
5.3 Previous work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.3.1 Ferromagnetism in InGOH and ITO thin films . . . . . . . . . . . 68
5.3.2 Other ferromagnetic TM-doped oxide systems . . . . . . . . . . 74
5.3.3 yE ferromagnetism . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.4 Problems associated with Dilute Magnetism . . . . . . . . . . . . . . . 76
5.5 Can a DMO be classed as a DMS? . . . . . . . . . . . . . . . . . . . . 77
5.6 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.6.1 Preliminary experiments . . . . . . . . . . . . . . . . . . . . . . 79
5.6.2 Dependence on extrinsic Sn doping concentration . . . . . . . . 80
5.6.3 Co-doped ITO(10) films . . . . . . . . . . . . . . . . . . . . . . 81
5.6.4 Dependence on oxygen partial pressure . . . . . . . . . . . . . . 84
5.6.5 Dependence on the growth temperature . . . . . . . . . . . . . . 90
5.6.6 Magnetic ITO . . . ? . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.6.7 Transport Measurements . . . . . . . . . . . . . . . . . . . . . . 91
5.6.8 Annealing tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.7 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 95
K mBrvy Vwsorption vny mvgnetix xirxulvr yixhroism xhvrvxterisvtion
of XoByopey Inyium iin dfliye thin films FEG
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
6.2 X-Ray Absorption Spectroscopy . . . . . . . . . . . . . . . . . . . . . . 103
6.3 Principles behind XMCD . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.3.1 Origin of the XMCD effect . . . . . . . . . . . . . . . . . . . . . 105
6.3.2 XMCD sum rules . . . . . . . . . . . . . . . . . . . . . . . . . . 106
6.3.3 Detection Methods . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.4 Previous Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.5 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6.5.1 In and Sn b GPH edges . . . . . . . . . . . . . . . . . . . . . . . . 113
6.5.2 Co aGPH edge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6.5.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
6.6 Influence of annealing : Studies at the Co aGPH edge . . . . . . . . . . . 122
6.7 Oxygen K -edge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.8 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 126
L dptixvl mvgnetix xirxulvr yixhroism xhvrvxterisvtion of XoByopey InB
yium dfliye thin films FHE
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
7.2 Principles of Magneto-optics . . . . . . . . . . . . . . . . . . . . . . . . 131
7.3 Previous results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
7.4 Experimental results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
7.4.1 Co-doped InGOH . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
7.4.2 Co-doped ITO(10) . . . . . . . . . . . . . . . . . . . . . . . . . 139
7.5 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 142
M XoDIid multilvyers FI5
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
8.1.1 The Giant Magnetoresistance effect . . . . . . . . . . . . . . . . 146
8.1.2 Importance of spin-dependent scattering . . . . . . . . . . . . . 149
8.1.3 Interlayer Exchange Coupling . . . . . . . . . . . . . . . . . . . 151
8.2 Previous Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
8.3 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
8.3.1 Magnetic Dead Layers . . . . . . . . . . . . . . . . . . . . . . . 153
8.3.2 Varying the number of repeat bilayers, c . . . . . . . . . . . . . 156
8.3.3 Varying the ITO spacer layer thickness, t Iid . . . . . . . . . . . 159
8.3.4 Considerations of the possible spatial distribution of superpara-
magnetic particles . . . . . . . . . . . . . . . . . . . . . . . . . . 165
8.3.5 Annealing effects on stack magnetoresistance . . . . . . . . . . . 167
8.4 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 167
N XuBwufferey pXoDIidr multilvyers FLG
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
9.2 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
9.2.1 Magnetotransport measurements and magnetometry . . . . . . 173
9.2.2 Structural properties . . . . . . . . . . . . . . . . . . . . . . . . 175
9.2.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
9.3 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 179
FE Zstimvting the hpin Yiffusion aength of hemixonyuxting Inyium iin
dfliye FMG
10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
10.2 Spin Valves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
10.2.1 Spin Valve structures . . . . . . . . . . . . . . . . . . . . . . . . 183
10.2.2 Spin accumulation effects . . . . . . . . . . . . . . . . . . . . . . 185
10.2.3 Controlling Spin Valve performance . . . . . . . . . . . . . . . . 185
10.3 Previous Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
10.4 Fabrication of the Spin valve structures . . . . . . . . . . . . . . . . . . 187
10.4.1 Patterning to micron scale tracks . . . . . . . . . . . . . . . . . 188
10.5 Focused ion beam processing of devices . . . . . . . . . . . . . . . . . . 191
10.5.1 The process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
10.5.2 Device fabrication . . . . . . . . . . . . . . . . . . . . . . . . . . 192
10.6 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
10.6.1 Cu(200nm)/Co(20nm)/ITO(t nm)/Co(5nm)/Cu(200nm) Spin valves197
10.6.2 Insertion of Cu buffer layers . . . . . . . . . . . . . . . . . . . . 204
10.6.3 Spin valves with DMO electrodes : Preliminary results . . . . . 205
10.7 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 208
FF hummvry vny Xonxlusions GFG
FG [uture Yirextions vny dutlook GFI
12.1 TEM study of the DMO . . . . . . . . . . . . . . . . . . . . . . . . . . 214
12.2 Photomagnetic effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
12.3 Metal/ITO interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
12.4 Functional vllBofliyz device . . . . . . . . . . . . . . . . . . . . . . . . . 216
V VwwrevivtionsA hymwols & ehysixvl xonstvnts GFM
W euwlishey vrtixles GGE
Xhvpter F
Introyuxtion
FCF hetting the sxeneO botivvtion for Sciateoaics
The concept of transparent conducting oxide films was born out of the discoveries made
by Badeker [1] over one hundred years ago. Since then, these materials have formed
integral components in a wide variety of modern day optoelectronic applications such
as liquid crystal displays and transparent electrodes of solar cells. In order to achieve
the simultaneous occurrence of high optical transmittance within the visible region to-
gether with good electrical conductivity, an electron degeneracy must be created within
the band gap, usually obtained via the introduction of non-stoichiometry and/or ap-
propriate dopants. Tin-doped Indium Oxide, commonly know as Indium Tin Oxide
(ITO), has been in practical use for most transparent optoelectronic applications for
many years now.
The relentless drive for greater information storage capacity in physically smaller
devices has pushed the size of components to scales where quantum mechanical effects
become important. To aid such an increase in information density, both the elec-
tron spin and charge are exploited in the field of spin transport electronics; hence the
portmanteau spintronixs. The basic idea is to combine the characteristics of existing
magnetic devices with semiconductor devices in order to realise a new generation of
devices that are much smaller, more energy efficient, non-volatile, and much faster than
presently available [2,3]. Spintronics may lead to devices such as spin-polarised LEDs,
spin-field effect transistors, and spin-based qubits for quantum computers. The key
requirement in the development of such devices is an efficient injection, transfer and
detection of spin-polarised current from a ferromagnetic material into a semiconduc-
tor. Due to the well known problem of a resistance mismatch at metal/semiconductor
interfaces, hindering an effective spin injection, much interest now concentrates on
the development of room-temperature ferromagnetic semiconductors [4]. A candidate
class of materials are the dilute magnetic semiconductors (DMS), in which magnetic
moments are incorporated into a semiconducting lattice.
1
Cb[pnel 1. Chnlido]ncih
Figure 1.1: Electronic configuration of the 3y -states and 4s-states of transition metal
elements (from right to left: V to Cu).
FCG Yilute bvgnetix hemixonyuxtors
A DMS is a compound with properties intermediate between a non-magnetic semicon-
ductor and a magnetic element, and is obtained by doping a non-magnetic semicon-
ductor with transition metal (TM) elements (elements with valence electrons corre-
sponding to the 4s orbital with a partially filled 3y shell). Figure 1.1 illustrates the
electronic configuration of these states for the TM elements. A DMS is characterised
by the random substitution of a fraction of the original host semiconducting atoms by
a homogeneous or “uniform” dilute concentration of magnetic impurity atoms. Since
one cannot normally get an ordered distribution of dopants, homogeneous or uniform
essentially means a statistically uniform dopant distribution wherein the number dis-
tribution of near neighbour dopant-dopant separation is not too broad. These mixed
crystal systems may essentially be considered as two interacting subsystems. The first
of these is the system of delocalised conduction and valence band electrons. The second
is the random, diluted system of localised magnetic moments associated with the mag-
netic atoms. A “true” DMS is one in which the magnetic dopant spins retain remanent
alignment in the presence of spin polarised carriers.
Figure 1.2 depicts the simple idea behind a DMS. The coupling between localised
magnetic moments leads to the existence of a number of magnetic phases. Mecha-
nisms that control the ferromagnetic ordering in these materials are therefore dictated
by the properties of the magnetic impurities. Coupling interactions between moments
will generally depend on the location of the dopants within the host lattice, the doping
properties of the impurities as well as on other dopants and defects that may be present
within the material.
A DMS should ideally offer the following [5];
• Ferromagnetic Curie temperature greater than 500 K
• A close connection between the ferromagnetism and the population of a spin-split
2
Cb[pnel 1. Chnlido]ncih
band of carriers
• A choice of p- or n-type doping
• High mobility and long spin diffusion length
• Magneto-optic effects including magnetic circular dichroism
• Anomalous Hall effect.
The discovery of carrier-mediated ferromagnetism in Mn-doped GaAs (MnxGaF−xAs
or GaMnAs) revamped the field of DMSs and created renewed interest in discovering
new and novel DMSs [6]. However, GaMnAs, and indeed other so-called DMS systems
(other TM-doped III-V semiconductors) are limited by their low ferromagnetic Curie
temperature (iX ≈ 170 K) and so their practical applications are limited. The quest
for new high-iX DMSs was given a breath of new life after pioneering theoretical pre-
dictions based on the Zener mean-field model of ferromagnetism [7] by Dietl zt vl. [8].
Here it was predicted that the i x of DMSs could be increased above room temper-
ature for p-type DMSs, and that ferromagnetism would be stable in DMSs based on
wide-bandgap semiconductors, i.e. ZnO and GaN.
Soon after, several oxide-based DMS (DMO) have been reported to be robust,
room temperature ferromagnets. In particular, Matsumoto zt vl. claimed intrin-
sic ferromagnetism in Co-doped TiOG thin films grown using pulsed laser deposi-
tion [9,10]. These results prompted extensive experimental work and theoretical studies
on other oxide systems such as ZnO (see chapter 5 and references within), SnOG [11–14],
(La,Sr)TiOH [15] and InGOH (see chapter 5 and references within), however, the most
heavily studied are those of Co- and Mn-doped ZnO and Co-doped TiOG. Although
a considerable amount of experimental data and corresponding mechanisms have been
accumulated over recent years, the origins of the ferromagnetic ordering in DMOs still
remain unclear. Early experiments relied on simple magnetic characterisation using
superconducting quantum interference device magnetometry to classify the validity of
the DMO. However, it is difficult to distinguish “intrinsic” ferromagnetism from that
which is “extrinsic”, such as ferromagnetic precipitations or secondary phases because
all the magnetisation signals are measured together, i.e. it is a volume sensitive mea-
surement. Therefore, magnetisation alone is not enough to verify intrinsic magnetism.
Thus, it is critical to distinguish those which are true DMOs from those which merely
show magnetic hysteresis.
One of the most troubling issues surrounding this area of research is the disparity
of results obtained by different laboratories across the world. It has become acceptable
to acknowledge the fact that the magnetic properties of these materials are strongly
dependent on the conditions in which they were prepared. As a result, the properties
are not even entirely reproducible from one growth to another, even in the same lab-
oratory. Therefore, the field is starving for a universal, unified theory which can shed
3
Cb[pnel 1. Chnlido]ncih
Element1: Non-magnetic Element 2: Non-magnetic Element 3: Magnetic
(a) (b) (c)
Figure 1.2: Illustration of the basic idea behind a dilute magnetic system: (a) a mag-
netic semiconductor exhibiting ferromagnetism due to carrier mediated alignment of a
periodic array of impurity transition metal dopants, (b) a non-magnetic semiconductor
which contains no magnetic ions, and (c) a dilute magnetic semiconductor where only
a small fraction of the host semiconductor sites are substituted by transition metal
dopants. Adapted from reference [6].
light on some of these unanswered questions.
FCH ihesis dverviefi
Given its tried and tested practical use, Indium oxide-based materials form the basis for
the developing field of transparent spintronics as well as the theme of this thesis. Hav-
ing introduced the dilute magnetic semiconductor and oxide family, this dissertation
proceeds further by reminding the reader of some of the basic principles of magnetism.
In addition, a brief history is given of the numerous theories which have been proposed
in recent years to aid the understanding of magnetism in the dilute magnetic semicon-
ductor and oxide systems. The experimental procedures and measurement techniques
used to characterise the materials are described in chapter 3.
Chapters 4 through to 10 contain a mixture of background and first principles and
experimental results. Because the nature of the work presented here is rather diverse,
it was decided that it would be better to introduce the important concepts in each
chapter as a prerequisite to the results. Therefore, chapter 4 is concerned with the ba-
sic introduction to the Indium Oxide system and a systematic study into the electrical
properties of the material as a function of various deposition parameters and extrin-
sic doping. Chapter 5 is an account of the struggle to develop a dilute magnetic oxide
system using Indium Oxide and ITO, and also touches upon the controversies surround-
ing the field. Following the development of an adequate DMO, powerful probes such
as synchrotron radiation x-ray (chapter 6) and optical (chapter 7) magnetic circular
dichroism techniques were used to try and better understand the origins of magnetism
4
Cb[pnel 1. Chnlido]ncih
in these materials. The last three chapters describe the gradual transition toward
integration of Indium Oxide based materials into spintronic devices. Ferromagnetic
metal/oxide multilayers (chapters 8 and 9) and spin valve junctions (chapter 10) show
the potential of Indium Tin Oxide based materials for future spintronic applications.
The thesis is summarily closed in chapter 11 before suggesting some potential avenues
that this area of research may wish to take in the near future (chapter 12).
5
Wiwliogrvphy
[1] K. Badeker, Ann. Phys. (Leipzig) GG, 749, (1907).
[2] S. Datta and B. Das, Appl. Phys. Lett. 56, 665, (1990).
[3] S. A. Wolf, D. D. Awschalom, R. A. Buhrman, J. Daughton, S. von Molnar, M.
L. Roukes, A. Y. Chtchelkanova, D. M. Treger Science, 294, 1488, (2001).
[4] G. Schmidt, D. Ferrand, L. W. Molenkamp, B. J. van Wees, Phys. Rev. B. 62,
R4790, (2000).
[5] J. M. D. Coey, Curr. Opin. Solid State and Materials Science 10, 8392, (2006).
[6] H. Ohno Science, 281, 951, (1998).
[7] C. Zener, Physics Review, 81, 440 (1951); 82, 403, (1995).
[8] T. Dietl, H. Ohno, F. Matsukura, J. Cibert, D. Ferrand, Science 287, 1019, (2000).
[9] Y. Matsumoto, M. Murakami, T. Shono, T. Hasegawa, T. Fukumura, M.
Kawasaki, P. Ahmet, T. Chikyow, S.-Y. Koshihara, H. Koinuma, Science 291,
854, (2001).
[10] Y. Matsumoto, M. Murakami, K. Hasegawa, T. Fukumura, M. Kawasaki, P. Ah-
met, K. Nakajima, T. Chikyow, H. Koinuma, Appl. Surf. Sci. 189, 344, (2002).
[11] N. H. Hong, J. Sakai, W. Prellier, A. Hassini, J. Phys.: Condens. Matter 17,
1697-1702, (2005).
[12] N. H. Hong, J. Sakai, N. T. Huong, N. Poirot, A. Ruyter, Phys. Rev. B 72, 045336,
(2005).
[13] S. B. Ogale, R. J. Choudhary, J. P. Buban, S. E. Lofland, S. R. Shinde, S. N.
Kale, V. N. Kulkarni, J. Higgins, C. Lanci, J. R. Simpson, N. D. Browning, S. Das
Sarma, H. D. Drew, R. L. Greene, T. Venkatesan, Phys. Rev. Lett. 91, 7, (2003).
[14] K. Gopinadhan, D. K. Pandya, S. C. Kshyap, S. Chaudhary, J. Appl. Phys. 99,
126106, (2006).
6
Bcbfcial[pbs
[15] S. X. Zhang, W. Yu, S. B. Ogale, S. R. Shinde, D. C. Kundaliya, W. K. Tse,
S. Y. Young, J. S. Higgins, L. G. Salamanca-Riba, M. Herrera, L. F. Fu, N. D.
Browning, R. L. Greene, T. Venkatesan, Phys. Rev. B 76, 085323, (2007).
7
Xhvpter G
Wvxkgrouny iheory vny [irst
erinxiples
ihis xhvptzr introyuxzs somz of thz funyvmzntvl physixs fihixh govzrn thz phznomznv
invzstigvtzy in this volumz of fiorkC ihz mvjority of thz physixvl xonxzpts zflplvinzy
in thz follofiing szxtions hvvz wzzn formulvtzy using rzfzrznxzs pFrA pGr vny pHrC borz
yztvilzy yzsxriptions of spzxifix zffzxts fiill wz yzsxriwzy in suwszquznt xhvptzrsC
8
Cb[pnel ,. B[]ealiohd Tbeils [hd Fclmn Plch]cpfem
GCF drigins of bvgnetism
Magnetism is a quantum mechanical phenomenon in which it is known that the mag-
netic properties arise from spin magnetic moments contributing to an overall molecular
field. There are three principal sources of a magnetic moment in a free atom: electron
spin, orbital angular momentum about the nucleus and change in orbital momentum
with an applied magnetic field [1]. An orbiting electron has a spin quantum number, s,
and angular momentum quantum number, l, which can be summed over all electrons
in an atom to calculate its total orbital moment, a, and spin angular moment, h, re-
spectively.
The Pauli exclusion principle states that no two fermions can simultaneously oc-
cupy the same quantum state. This becomes significant when considering the exchange
interactions between atoms and how electron orbitals overlap. It is also important to
consider Hund’s rules, which govern how useful energy interactions dictate the ground
state terms. Using these rules, one can calculate the values for J, the total angular
momentum, a and h for the ground states of the transition metal ions.
Therefore, to summarise, the universal origin of magnetism is almost always the
interplay between the electronic spin degree of freedom, the repulsive Coulomb inter-
actions between electrons, and the fermionic quantum statistics of electrons. However,
one universal theory of magnetism is yet to be agreed upon as it continues to be
amongst one of the most challenging of solid-state physics problems.
GCFCF bvgnetix mvterivls
The magnetic moment per unit volume of a material is defined as M⃗ and the macro-
scopic field within the material is defined as H⃗. The magnetic susceptibility, χ, de-
scribes the ability of a material to become magnetised and is defined as
χ =
M⃗
H⃗
O (2.1)
The permeability, µ, is a material property that defines the degree to which a material’s
magnetisation responds to a linear magnetic field. The induced magnetic field, B⃗, is
the net magnetic effect from an externally applied magnetic field. When in the presence
of a magnetic material, the external magnetic field is made up of contributions from
the current elements and the magnetic material such that
B⃗ = µEH⃗ + µEM⃗ = µEH⃗ + µEχH⃗O (2.2)
The relative permeability of the material, µr, is the ratio of a material’s permeability,
µ, to that of free space, µE (which is constant). It is known that µr is equal to 1 + χ.
Therefore, we now have
B⃗ = µEH⃗(1 + χ) = µrµEH⃗O (2.3)
9
Cb[pnel ,. B[]ealiohd Tbeils [hd Fclmn Plch]cpfem
(c)
m =0H
0
m H
0
ext
(b)(a)
Simple ferromagnet
Simple antiferromagnet
m
0
M
M
Figure 2.1: (a) Illustration of the typical dipole alignments in ferromagnetic (top)
and antiferromagnetic (bottom) materials using a linear array of localised moments;
(b) Schematic illustration of the process of domain rotation within an applied field
which gives a magnetic hysteresis; (c) Typical responses to an applied magnetic field
for a ferromagnetic (40 nm thick Co film, solid black line), paramagnetic (solid blue
line) and diamagnetic (solid red line) materials. As the external magnetic field (µEH)
is increased, the induced magnetic moment also increases. The induced magnetisation
eventually saturates at a point b s. Upon reversal of the external field the material
retains a certain permanent magnetisation termed the remanant magnetisation, b r.
The value of the reversed field when the sample is fully demagnetised is called the
coercive field, µE] x.
10
Cb[pnel ,. B[]ealiohd Tbeils [hd Fclmn Plch]cpfem
The values of µr and χ can be used to classify magnetic materials which can be
summarised in the following points:
• χ is small and negative, typically χ ≈ - 10−5. These materials are yivmvgnztix.
This occurs when an external magnetic field is applied to a material, causing an
induced electromotive force (e.m.f) to accelerate the material’s electrons. When
the field is constantly applied, the induced e.m.f returns to zero, but the current
remains. This is because the currents are sufficiently stable to screen the material
from the applied magnetic field and so the material has acquired a weak mag-
netisation which opposes the applied field. A very small diamagnetic moment is
induced in all materials, although it will only be observed with no other dipole
moments present in an externally applied magnetic field.
• χ is small and positive (χ ≈ 10−5 - 10−H). These materials are pvrvmvgnztix.
This is seen in materials in which a number of the atoms possess, in the ab-
sence of an applied field, a permanent magnetic dipole moment. However, these
moments are randomly orientated and so in the absence of an external field the
net magnetisation will be zero. Upon the application of an external field, the
moments begin to align with the field and the magnetisation is non-zero. There
is competition between the aligning effect of the field and thermal disorder. The
material will have, on average, greater alignment with the field than against it so
that magnetisation varies linearly with applied magnetic field. As temperature
increases, thermal disorder will also increase and so the paramagnetic suscepti-
bility decreases.
• χ ≫ 1 and positive (χ ≈ 50 - 10,000). These materials are fzrromvgnztix. The
dependence of χ on an applied magnetic field is non-linear in ferromagnetic ma-
terials. Ferromagnets also have a spontaneous magnetisation in the absence of an
external magnetic field (see section 2.1.3). As with paramagnets, magnetisation
occurs due to the alignment of dipoles at low temperatures [Fig. 2.1 (a)]. The
difference is that ferromagnets remain ordered to higher temperatures below a fer-
romagnetic Curie temperature (iX), indicating that there are stronger exchange
interactions present. Below iX the magnetic moments align and the material is
ferromagnetic, whilst above i x, thermal fluctuations destroy this alignment so
the material is paramagnetic.
• χ is small and positive, but with a different magnetic structure and temperature
dependence - these materials are classified as vntifzrromvgnztix. For an anti-
ferromagnet, exactly half of the dipoles are aligned in one direction and half in
the opposite [Fig. 2.1 (a)]. This means that, overall, there is no spontaneous
magnetisation. Similar to iX for ferromagnets, there is a critical temperature,
known as the Neel temperature, ic , above which the moments are disordered
11
Cb[pnel ,. B[]ealiohd Tbeils [hd Fclmn Plch]cpfem
Q- Q
q = + Q
q = 0
q = - Q
c1/c
q = - Q
q = 0
q = + Q
T T
(a) (b)
Figure 2.2: The Curie-Weiss law states that χ ∝ 1 / (i - θ) for i S θ. This can
be found in (a) for each of the three magnetic cases: θ = 0 (paramagnet), θ = Θ S 0
(ferromagnet) and θ = - Θ < 0 (antiferromagnet). Plotting 1/χ versus i gives results
in straight lines as shown in (b) with the intercept on the temperature axis yielding θ.
Adapted from reference [3].
paramagnetically and below which they are alternatively arranged parallel and
antiparallel.
GCFCG iempervture yepenyenxe of mvgnetix susxeptiwility
The change in the susceptibility χ with temperature can be described by the Curie-
Weiss law only for i S θ (i.e. for temperatures above the transition to magnetic order
in the paramagnetic state) and is given in Eqn. 2.4, where C is the Curie constant, i
is the absolute temperature and θ is the Curie-Weiss temperature. The relationship is
shown graphically in Fig. 2.2.
χ =
C
i − θ (2.4)
If;
• θ = 0, the the material is a paramagnet,
• θ S 0, the material is a ferromagnet and we would expect θ = iX ,
• θ < 0, the material is an antiferromagnet and we expect θ = - ic (see Fig. 2.2).
GCFCH [erromvgnetism
According to band theory, the magnetic moment is defined as the number of band
states per atom, per unit energy and per spin [4]. Figure 2.3 shows the density of
states (DOS) calculated for non-magnetic Cu and strongly ferromagnetic Co metals.
The shapes of the DOS in both scenarios are very similar, however, in the case of
12
Cb[pnel ,. B[]ealiohd Tbeils [hd Fclmn Plch]cpfem
E E- F (eV)
D
en
si
ty
o
f
st
at
es
(e
V
at
om
sp
in
)
-1
-1
-1
(a) (b)
Figure 2.3: Density of states of Cu (a) and Co (b) metals for the majority and minority
spins. The Fermi energy EF is set to zero. Cu is non-magnetic due to the absence of
exchange splitting. Co is a strong ferromagnet due to total occupancy of the majority
spin band. The dashed line in each figure indicates the different position of the Fermi
energy, EF - it separates the occupied from the unoccupied states. After reference [4].
Cu no exchange splitting is observed, as expected. The exchange splitting, ∆, can be
defined as the relative shift of spin-up and -down electron bands. The difference in
the number of spin-up and spin-down electrons gives rise to an observed spontaneous
magnetisation - ferromagnetism. Via consideration of the energy exchange involved in
a resultant spontaneous ferromagnetic signal we can define a condition that must be
met before this can occur. This is known as the htonzr xritzrion and is defined as
Ug(EF ) ≥ 1, (2.5)
where j is a measure of the Coulomb energy and g(EF ) is the density of states at the
Fermi level. The condition for the ferromagnetic instability requires that the Coulomb
effects are strong and the density of states at the Fermi energy is large.
In order to explain the fact that ferromagnetic materials do not remain saturated
when the applied field is removed, Weiss [5, 6] introduced the classical concept of
magnetic domains which form as a consequence of znzrgy minimisvtion. He proposed
that a ferromagnet is divided into regions called “domains”, in which the magnetisation
within the domain is almost saturated. The magnetisation in different domains is in
different directions, so that the magnetisation of a ferromagnetic specimen could be
small or even zero. The consequences of this development were [1]:
1. the permanent existence of atomic magnetic moments which make up the mag-
netic domains and are aligned due to a Weiss mean (or “molecular”) field [5],
2. in the demagnetised state, the atomic moments were aligned,
3. the magnetic domains were randomly aligned in the demagnetised state,
13
Cb[pnel ,. B[]ealiohd Tbeils [hd Fclmn Plch]cpfem
4. the magnetisation process involved the reorientation of the magnetic domains to
the applied magnetic field.
As point (4) states, b s is achieved by aligning the magnetisation of each domain
with the applied magnetic field [Fig. 2.1 (b)]. A direct consequence of ferromagnetic
domains is the existence of a magnetic hysteresis upon the application of an increasing
field to a magnetic sample. The process of magnetising a sample to saturation consists
of converting it to a single domain system, and can be described as a three-step process:
1. Low-field region - domains orientated favourably with respect to the applied
magnetic field may grow, by displacement of domain walls, at the expense of
those orientated unfavourably.
2. Intermediate-field region - atomic magnetic moments within an unfavourably
aligned domain overcome the anisotropy energy and rotate into a crystallographic
“easy” direction nearest to the direction of the applied field. This is called “do-
main rotation”.
3. High-field region - the domains aligned along the preferred easy axis close to
the field direction are gradually rotated so they are completely aligned with
the applied magnetic field. This occurs as the field is increased as is known as
“coherent domain rotation”.
Such processes are described by a ferromagnet’s hysteresis loop where magnetisation
is measured as a function of the applied field and is characterised by four material
and system dependent parameters: b s; b r; coercive field (µE] x); and saturation field
(µEHhAi ). The magnetisation process and the important parameters can be found in
Fig. 2.1 (c).
There is, however, another aspect of the magnetisation process which must be ac-
knowledged. Consider a ferromagnetic body exposed to an external magnetic field.
When the magnetisation inside a magnetic body meets the surface it suddenly has to
stop. Therefore, we have a divergence in the magnetisation and an equal but opposite
effect in the magnetic field. The result is a “demagnetising field”. In the scenario of
a thin film, if the magnetisation lies in the plane of the film, the only divergence of
magnetisation is at the ends, which we assume are an infinite distance away and there-
fore, there is no demagnetising field. If the field is perpendicular to the film, magnetic
poles are created on the top and bottom surfaces and give rise to a demagnetising field.
The effect is dependent upon the strength of the magnetisation and geometry of the
material.
GCFCI huperpvrvmvgnetism
Superparamagnetism is a rather special scenario. It is a size effect of ferromagnetism,
i.e. one cannot reduce the size of magnetic particles indefinitely and still retain useful
14
Cb[pnel ,. B[]ealiohd Tbeils [hd Fclmn Plch]cpfem
magnetic properties [2]. As already explained using Weiss’s theory, a ferromagnetic
sample divides itself into an array of magnetic domains in order to minimise its energy.
Below a certain size, moments are released from their constraints so we can think
of the particle as a single-domain structure, where the b r is no longer fixed in the
direction dictated by the particle shape or crystal anisotropy: instead, magnetisation
can randomly flip direction under the influence of thermal energy. The reason for the
small ferromagnetic particles becoming single-domain is simply due to the fact that the
energy xost of domain wall formation does not outweigh any svving in demagnetising
energy. A particle of volume k will need an activation energy equal to Kk, where
K is a constant which quantifies the energy density associated with the anisotropy,
to minimise its energy and thus flip its magnetisation. For very small particles, such
that Kk is small compared to the thermal energy, kBi, the magnetisation can be
easily flipped in this way by thermal fluctuations. If we now consider a distribution
of the small ferromagnetic particles throughout a non-magnetic matrix and assume
that the particles are far enough apart, then we can also assume that interparticle
interactions can be neglected, and for kBi ≫ Kk, the system will behave like a
typical paramagnet [3]. At high enough temperatures the moments on the particles
are able to fluctuate rapidly and the relaxation time (the time between spin flips), τ ,
of the moment on a particle is given by
τ = τEexp
(
Kk
kBi
)
, (2.6)
where τE is typically 10
−Ns. Another important parameter in the determination of
superparamagnetic effects is the measuring time, t, of a particular laboratory experi-
mental technique. The detection of superparamagnetism is time-dependent and hence
the relationship between τ and t is very important. If t≫τ in the absence of an applied
magnetic field then the magnetisation of the nanoparticles averages to zero and the
system is said to be in the superparamagnetic state. However, if t≪τ , the magnetisa-
tion will not flip during the measurement and will be equivalent to the instantaneous
magnetisation. The nanoparticles now appear to be “blocked”. A transition between
superparamagnetism and blocked states occurs when t=τ .
Similar to a paramagnetic material, the superparamagnetic particles can become
aligned and magnetised via the application of an applied external magnetic field, how-
ever, they have a magnetic susceptibility somewhere in between a paramagnetic and
ferromagnetic material. In the presence of a magnetic field, the magnetisation loop
has a characteristic “s”- shape [Fig. 2.4 (a)], and can be described well by a Brillouin
function;
M =Ms
[(
2J + 1
2J
)
coth
[
(2J + 1)y
2J
]
−
(
1
2J
)
coth
( y
2J
)]
(2.7)
where b s is the saturation magnetisation, y=gµBJ] /kBi (where g is the Lande g-
factor, µB is the Bohr magneton, ] is the applied magnetic field, kB is Boltzmann’s
15
Cb[pnel ,. B[]ealiohd Tbeils [hd Fclmn Plch]cpfem
T (K)
TB
FieldCooled
Zero-Field Cooled
(a)
(b)
Figure 2.4: (a) Simulated magnetic hysteresis loop for a superparamagnetic sample
using the Brillouin function (Eqn. 2.7), (b) Schematic of typical magnetisation ver-
sus temperature curves (field- and zero-field cooled) for a superparamagnetic material
showing the blocking temperature, iB, in the zero-field cooled curve only.
constant and i is temperature) and J is the total angular momentum. Important
features of the magnetisation loop for a superparamagnet are the distinct absence of
hysteresis (low coercive fields) and large saturation fields.
Superparamagnetism can be destroyed by cooling. This follows because the charac-
teristic fluctuation time for a particle’s moment varies exponentially with temperature,
so the magnetisation appears to switch sharply to a stable state as the temperature
is reduced. The temperature at which this occurs is called the blocking tempera-
ture (occurs when t=τ), iB, and depends linearly on the sample volume and on the
magnitude of the crystal-field anisotropy. iB is usually identified through field- and
zero-field cooling magnetometry [see Fig. 2.4 (b)]. Below iB the material loses its
preferred magnetisation direction in zero field (ZFC) and will begin to average to zero.
Through rearrangement of Eqn. 2.6 iB can be defined as
iB =
Kk
kBln
(
αt
τ0
) O (2.8)
Therefore, from Eqn. 2.6 we can see that iB is dependent on the applied field,
particle size as well as the size distribution.
GCFC5 Xompeting intervxtionsA frustrvtion vny rvnyomness
It is not always possible to satisfy all the interactions in a system and settle on a single
unique ground state. This results in a variety of similar low energy states distributed
around the system and means that energy minimisation is not achieved. As a result,
system is now said to show “frustration”. This usually occurs in situations where it
is not possible to orient a spin on a neighbouring lattice so that the other nearest
neighbours can achieve energy minimisation and the system contains a multiplicity of
equally unsatisfied states [3].
16
Cb[pnel ,. B[]ealiohd Tbeils [hd Fclmn Plch]cpfem
There are other consequences to energy imbalances within a magnetic system. hpin
glvsszs are defined as a random, mixed-interaction magnetic system characterised by a
random, yet cooperative, freezing of spins at a well defined temperature, below which
no long range magnetic ordering is present. Randomness within the system can be
introduced in a number of ways, namely via site- and bond-randomness. Equally
important to the inherent randomness of a spin glass is the presence of competing
interactions and random anisotropy which may lead to a locally varying easy-axis for
magnetisation. Competing interactions will lead to frustration so that there is a mul-
tidegenerate ground state. Another feature of spin glass materials is the xoopzrvtivz
frzzzing trvnsition which occurs at a certain temperature, i f . At high temperature the
behaviour of spins within the spin glass material is dominated by thermal fluctuations
(in fact, this is true for all magnetic materials) so that the spins can be considered
independent. As the temperature cools, the independent spins slow down and start
to cluster. Those not in the cluster will help to mediate interactions between clusters.
As the temperature nears i f the fluctuations in the clusters begins to slow down so
that interactions between clusters becomes more long range. At i f the system finds
one of its many ground states and freezes. Below i f the system is said to be “glassy”.
This is seen in a magnetisation versus temperature curve where there is a divergence
between the field cooled and zero-field cooled magnetic susceptibility below i f .
A signature of both spin glass and superparamagnetic materials is a sharp peak
close to either i f or iB in the real part of AC susceptibility where the position of the
peak should vary with the frequency of the alternating magnetic field.
GCFCK V quvntum yesxription of mvgnetism
Although Weiss developed a very successful phenomenological theory based on an in-
ternal molzxulvr field [5], it was not until the advent of quantum mechanics that the
origin of the interaction became clear. The interaction responsible for magnetic or-
der is the zflxhvngz intzrvxtion. This interaction results from the fact that electrons
are identical and obey the Pauli exclusion principle. Heisenberg showed that a quan-
tum mechanical exchange interaction between neighbouring electrons with overlapping
wavefunctions plays a decisive role in explaining ferromagnetism. If we consider two
localised electrons on neighbouring atoms with spins i and j, then the Hamiltonian
exchange between them is given by;
< H S= −2Jijsi · sj, (2.9)
where J ij is the exchange integral between the i
th and j th spins. However, when
considering a solid then it is necessary to sum the exchange over all the electrons
which can contribute to the energy, so that
< H S= −2
∑
ij
Jijsi · sjO (2.10)
17
Cb[pnel ,. B[]ealiohd Tbeils [hd Fclmn Plch]cpfem
In most cases, our interest lies only in the nearest-neighbour (n-n) interactions, which
simplifies the Heisenberg Hamiltonian
< H S= −2J
∑
n−n
si · sj, (2.11)
where it is assumed that the exchange integral J is the same for all nearest-neighbour
pairs. If J < 0, the lowest energy state results from antiparallel spins (antiferromag-
netism). J S 0 gives ferromagnetism due to the alignment of spin moments on adjacent
atoms. This type of exchange interaction is known as yirzxt zflxhvngz where electrons
on neighbouring magnetic atoms interact with each other.
In many transition metal oxides, magnetic interactions between the transition metal
ions are mediated by the conduction electrons and intermediate anions. These types
of magnetic coupling are known as inyirzxt zflxhvngz interactions, which can still be
described by the Heisenberg Hamiltonian. These interactions form the basis for the
theories used to describe the possible origins of magnetism in the dilute magnetic sys-
tems (see next section).
GCG Zflisting theories for ferromvgnetix origin in Ybh
vny Ybd systems
In light of the results mentioned in chapter 1 regarding DMS and DMO materials,
many have speculated as to the origins of the magnetic behaviour in these materials.
The mechanisms responsible for the origin of ferromagnetism in DMOs are far from
being well understood and it is this aspect which makes this area of research so excit-
ing. It was discussed briefly in chapter 1 that in a DMO a sizable portion of atoms are
randomly substituted by transition metal elements, giving rise to localised magnetic
moments in the oxide host matrix. The presence of magnetic ions affects the free carrier
behaviour through an sp-y exchange interaction between localised magnetic moments
and the spins of the itinerant carriers. However, as we shall see in chapter 5, this
simple picture of 3y cations with localised moments causing ferromagnetism is now
being strongly contested (see section 5.5). Since for dilute concentrations of magnetic
dopants the mean separation between neighbouring dopants is larger than that required
for any conventional mechanism of ferromagnetic exchange between neighbouring ions
to be operative, the occurrence of ferromagnetism in many cases is termed as “carrier”
mediated.
This section aims to describe the models that are most commonly used to describe
the magnetic interactions in dilute systems. Some, such as the double-exchange interac-
tion, exclusively promote ferromagnetism, whereas the others can result in ferromag-
netic or antiferromagnetic interactions, depending on the geometry, defect structure
and carrier concentration.
18
Cb[pnel ,. B[]ealiohd Tbeils [hd Fclmn Plch]cpfem
GCGCF XvrrierBmeyivtey eflxhvnge
ihe gKKn moyel
Early attempts to understand the magnetic behaviour in DMS and DMO systems are
based on models in which the local magnetic moments are assumed to interact with
each other via Rudeman-Kittel-Kasuya-Yosida (RKKY) interactions [7]. The RKKY
model describes the interaction between the delocalised conduction band electrons and
a lone, localised magnetic ion. Nearby conduction electrons are duely magnetised by
the magnetic ion. These act as an effective field and influence the polarisation of nearby
magnetic ions. This polarisation and sign of the exchange interaction, J zx, decays in
an oscillatory manner mediating either ferromagnetic or antiferromagnetic coupling
depending on the separation of the magnetic ions, r, and the density of electrons, in
the following manner;
Jzx(r) =
m∗kIF
h¯G
F (2kF r) (2.12)
where m∗ is the effective mass and kF is the Fermi wavevector of the electron gas. The
oscillating function follows Eqn. 2.13
F (x) =
x cos x− sin x
xI
O (2.13)
In general, this model is applicable when there is a high concentration of delocalised
carriers present in the host oxide.
ihe oenerBwvsey moyels
Interested by the interactions between y -shells in the transition metals, Zener explained
how direct coupling between y -shells is antiferromagnetic whereas indirect coupling via
conduction electrons is ferromagnetic [8]. These 3y spin and conduction electron in-
teractions were thought to be long-range.
The origin of ferromagnetism however is more complicated, and cannot be explained
by RKKY or Zener-like interactions alone. One of the prominent models is the Zener
model of carrier-mediated ferromagnetism developed by Dietl zt vl [9]. It is a mean-
field model in which there is a long range ferromagnetic exchange interaction between
TM (Mn) spins and p-type hole carriers. The model includes antiferromagnetic su-
perexchange interactions [10] and so goes a step further than simply considering spin
and free electrons. It also takes into account the anisotropy of the carrier-mediated
exchange interaction associated with the spin-orbit coupling in the host material. It
was this theory that predicted that Mn-doped p-type ZnO (and GaN) would have a
ferromagnetic Curie temperature above room temperature.
19
Cb[pnel ,. B[]ealiohd Tbeils [hd Fclmn Plch]cpfem
Figure 2.5: Representation of bound magnetic polarons. Cation sites are illustrated by
the small open circles. The unoccupied oxygen sites (oxygen vacancies) are represented
by the squares. Oxygen is not shown. Impurity ions within a BMP will tend to align
to minimise energy. BMPs overlap when their energies align. After Reference [13].
ihe Youwle Zflxhvnge boyel
This model is based on another model proposed by Zener to explain the experimentally
observed ferromagnetism in manganite materials [11]. The model is characterised by
the hopping of electrons between two neighbouring TM ions. In the DMO materials,
the 3y levels of the TM ions are split by the crystal field into higher energy doublet zg
and lower energy triplet tGg levels (10 electrons are spread over a total of five levels in
the y orbital). Spin-up and spin-down states are also split by the exchange splitting.
If neighbouring TM ions are aligned in parallel with each other, the electrons in the
partially filled 3y -orbitals of the TM ions are allowed to hop from one ion to the other
and stabilise the ferromagnetic ground state.
GCGCG ihe Wouny bvgnetix eolvron boyel
The model developed by Dietl zt vl. [9] predicts ferromagnetism arising from a ferro-
magnetic exchange provided by holes in the conduction band. In most samples however,
n-type electron doping is employed and so ferromagnetism is more likely to be medi-
ated by electron carriers.
The bound magnetic polaron (BMP) model is described by the exchange inter-
action between many localised TM ions spins via a much lower number of weakly
localised charge carriers. The localised holes of the polarons act on the TM impuri-
ties surrounding them, thus producing an effective magnetic field and aligning all the
spins. Therefore, a BMP can be defined as a localised hole centre containing magnetic
20
Cb[pnel ,. B[]ealiohd Tbeils [hd Fclmn Plch]cpfem
Figure 2.6: Schematic band structure of an oxide with different 3y TM impurities.
(a) The position of the 3y level for low Tx, when the splitting of the impurity band
is small. (b) and (c) show positions of the spin-down (b) and spin-up (c) 3y levels,
respectively, leading to high TX . After Reference [13].
impurity spins. Overlapping of neighbouring magnetic polarons leads to a long-range
ferromagnetic state, and tends to take place more readily if the temperature is de-
creased so that impurity spins in neighbouring BMPs are aligned. Coey zt vl. applied
this model to DMOs [12, 13] and proposed that ferromagnetic exchange coupling of
TM ions in n-type DMOs is mediated by the shallow donor electrons trapped by the
oxygen vacancies that tend to form BMPs within their hydrogenic orbits (Fig 2.5).
The general formula for the DMOs is;
(VF−xMx)(d ⋄ δ)n (2.14)
where V is a nonmagnetic cation, b is a magnetic TM cation, O is the oxygen, ⋄
represents a donor defect and n= 1 or 2. The exchange interaction ∆E zx between
the hydrogenic electron and TM cation can be written in terms of the s-y exchange
parameter J sy as;
∆Ezx = −Jsyh · s|ψ(r)|GΩ (2.15)
where h is the spin of the TM cation with a volume Ω = (4/3).rx
H, r x is the TM
cation radius, and s is the spin of donor electron. The wave function of the donor
electron in a 1s orbital is given by ψ(r) = (.rH
H)−FRG exp(-r/rH), where rH and r are
the hydrogenic orbital radius of the donor electron and the distance from the nucleus,
respectively.
Therefore, in order to increase the the exchange interaction and ferromagnetic
Curie temperature one needs to increase the donor electron density in the vicinity of
the TM cation. As the donor defect concentration, δ, increases, the hydrogenic orbitals
associated with the defects will begin to overlap to form a spin-split donor impurity
band (Fig. 2.6) [13]. At low concentrations, the electrons may remain localised because
of the influence of correlations and potential fluctuations in a narrow band, but above
21
Cb[pnel ,. B[]ealiohd Tbeils [hd Fclmn Plch]cpfem
a critical donor concentration the impurity band states would become delocalised. An
increasing density of defects would eventually lead to the hydrogenic orbitals associ-
ated with the randomly situated defects overlapping. The model implies that high
i x requires hybridisation and charge transfer from a donor-derived impurity band to
unoccupied 3y states at the Fermi level. This is shown in Fig. 2.6 (b) and (c) where a
high i x is only achieved when empty minority- or majority-spin “d” states lie at the
Fermi energy level fiithin the impurity band.
The BMP model works on the basis of direct exchange interactions which as has
been discussed, we know are antiferromagnetic. However, the interaction between
BMPs may be ferromagnetic for sufficiently large concentrations of magnetic impuri-
ties.
Courtesy of the BMP model applied to the DMO materials, the knowledge of de-
fects (oxygen vacancies in particular) in DMOs and their influence on the ferromagnetic
behaviour became very important.
GCH bvgnetoresistvnxe
Magnetoresistance (MR) is defined as a change in the electrical resistance of a substance
in the presence of a magnetic field. The signal response of a device is often characterised
by the percentage MR, as shown by Eqn. 2.16, where ∆g is the change in resistance in
an applied field (g(H) - g(0)) and g(0) is the resistance in the absence of an applied
field
Mg(%) =
g(H)−g(0)
g(0)
=
∆g
g(0)
O (2.16)
Many different forms of MR have been researched, using a variety of different ma-
terials and multilayer architectures. This section will concentrate on the “ordinary”
MR effect in metal, semiconducting and doped semiconducting thin films, leaving more
specific phenomena such as the giant magnetoresistance (GMR) effect seen in multi-
layer systems to later chapters of the thesis.
GCHCF bvgnetoresistvnxe of nonBmvgnetix metvls vny ferroB
mvgnets
When a metal carries an electrical current, the displacement of the electrons to different
parts of the Fermi surface is such that the scattering is minimised, i.e. the electrons
find the path of least dissipation to cross the sample. On application of a magnetic
field, the electrons are then forced to take another path which increases the scattering
rate and gives a positive MR (increasing g with increasing field). A negative MR
can be achieved through application of a magnetic field in the plane of the sample
but perpendicular to the current direction which results in reduced electron orbits and
22
Cb[pnel ,. B[]ealiohd Tbeils [hd Fclmn Plch]cpfem
hence scattering.
The situation in ferromagnets however is more complicated and can be visualised
by considering the “two-current model” proposed by Mott and involves electrons from
the s- and y -bands of the ferromagnet. The band structures and electron behaviour for
the s- and y bands are very different. The y -band is very narrow so that the density
of states at the Fermi energy level is large and the Stoner criterion is satisfied, and
the respective effective masses of the electrons follow my
∗ ≫ mz. The electrons in the
s-bands are nearly free so that ms
∗ ∼ mz. Therefore, the conductivity, σ, defined by
σ =
nse
Gτs
m∗s
+
nye
Gτy
m∗y
, (2.17)
will be dominated by the first term and conduction is mainly governed by the s elec-
trons. nsPy are the number of electrons in the s- and y -bands respectively, and τsPy
are the respective scattering times. The transition probability is therefore due to s
→ y electrons and is strongly temperature dependent. Generally at low temperatures
(below i x) there are few s → y transitions due to all unoccupied y -states being an-
tiparallel. At higher temperatures (above i x), all s electrons can make transitions and
there is a rise in the scattering rate. Once a magnetic field is applied the degree of spin
polarisation is increased which allows fewer s → y transitions, giving a negative MR.
GCHCG Vnisotropix mvgnetoresistvnxe
Magnetoresistance which depends strongly on the orientation of the magnetisation is
known as “anisotropic” magnetoresistance (AMR). Its origins are connected with the
spin-orbit interaction and its influence on s-y scattering which leads to anisotropic
scattering [3]. It is however, a rather small effect.
Investigating the MR of a system will therefore provide fundamental information
about the scattering mechanisms of carriers as well as giving insight into potential
technological applications. In particular, in regards to the origins of magnetism in
the DMO materials, any distinctive magnetotransport results might be used as a way
of identifying interactions, such as those between s-like conduction electrons and 3y
localised magnetic ions.
23
Wiwliogrvphy
[1] D. Jiles, Introduction to Magnetism and Magnetic Materials, Second Edition,
Chapman & Hall, (1998).
[2] R. C. O’Handley, Modern Magnetic Materials - Principles and Applications, John
Wiley & Sons, Inc. (1999).
[3] S. Blundell, Magnetism in Condensed Matter, Oxford University Press, Oxford,
p. 93, (2001).
[4] J. Stohr and H. C. Siegmann, Magnetism - From fundamentals to Nanoscale Dy-
namics, Springer, pg. 524, (2006)
[5] P. Weiss, Compt. Rend., 143, 1136, (1906).
[6] P. Weiss, J. Phys., 6, 661, (1907).
[7] M. A. Ruderman, C. Kittel, Phys. Rev. 96, 99, (1954) ; T. Kasuya, Prog. Theor.
Phys. 16, 45, (1956) ; K. Yosida, Phys. Rev. 106, 893, (1957).
[8] C. Zener Phys. Rev., 81, 440, (1951).
[9] T. Dietl, H. Ohno, F. Matsukura, J. Cibert, D. Ferrand, Science 287, 1019, (2000).
[10] P. W. Anderson Phys. Rev., 79, 350, (1950).
[11] C. Zener, Physics Review, 82, 403, (1995).
[12] J. M. D. Coey, J. Appl. Phys. 97, 10D313, (2005).
[13] J. M. D. Coey, M. Venkatesan, C. B. Fitzgerald, Nature, 4, (2005).
24
Xhvpter H
Zflperimentvl proxeyures
Vll struxturzs uszy in this fiork fizrz przpvrzy using DC mvgnztron sputtzring unyzr
ultrv high vvxuum xonyitionsC Chvrvxtzrisvtion fivs pzrformzy using v numwzr of tzxhB
niquzsC ihis xhvptzr yzsxriwzs thz tzxhniquzs uszy vny thz kzy prinxiplzs wzhiny zvxh of
thzmC hzvzrvl of thzsz vrz rzvisitzy in forthxoming xhvptzrs fihzrz furthzr zflplvnvtion
is rzquirzy.
25
Cb[pnel -. Erpelcgehn[f pli]edolem
HCF erinxiples of hputtering
Many factors must be considered and demands met when deciding upon a suitable film
deposition technique for a particular application. One must consider issues such as the
required substrate along with film coverage, residual or induced film strain, thickness
uniformity, sticking coefficients of the substrate and successful transfer of stoichiomet-
ric properties [1]. One of the major challenges in film deposition is acquiring the desired
phase and structure upon a substrate during deposition. The structural development
of the film requires the species arriving on the substrate to be able to migrate on the
growth surface. This surface mobility must be optimised allowing adatoms to reach
equilibrium sites. Therefore, the structure and properties of a thin film is critically
dependent on the film growth parameters.
The basic principle behind the sputtering process involves the ejection of species
from a solid target surface which has been subjected to energetic ion bombardment.
This provides the source of the vapour flux in sputter deposition, which eventually con-
denses onto a strategically positioned substrate surface. The energy of the sputtered
species depends on a number of factors, such as the atomic mass, the nature of the
target, the type of energy of the bombarding species, ztx. [1].
Sputtering requires a gas in order to maintain a discharge in front of the sputtering
target, which is held at a high negative potential, producing an electric field. The ion
source is usually a plasma generated by the impact of electrons within the noble gas
(predominantly argon (Ar) gas as well as mixtures containing oxygen in this work).
These positively charged gas ions from the discharge are accelerated in the electric field
and bombard the target surface to produce the sputtered flux. Eventually, the ejected
or sputtered atoms condense onto a substrate to form a thin film.
HCFCF YX mvgnetron sputtering
In this work, DC magnetron sputtering was used. For a full review of DC magnetron
sputtering, the reader can refer to reference [4] of this chapter. The action of the
electric field is to accelerate electrons (z∗) to collide with inert Ar atoms according to
the relationship; e∗+Ar→Ar++e. This yields positively charged Ar ions as well as free
electrons to form the glow discharge. Therefore, the electric field will determine the
momentum at impact of the plasma ions and also the momentum of the ejected mate-
rial. However, unlike a conventional DC sputtering target, the target has been mounted
upon a magnetron (stage containing SmCo permanent magnets) whose magnetic field
lines permeate through the target. The field lines serve to provide effective toroidal
confinement of secondary electrons generated by ion bombardment of the target sur-
face along a “race-track” by virtue of a Lorentz force zk⃗×W. The electron race track
is visible on the magnetron surface due to the higher sputtering rate along this track
[see figure 3.1 (a) - (c)]. The use of a magnetic backing and its effects on the motion of
26
Cb[pnel -. Erpelcgehn[f pli]edolem
Figure 3.1: Magnetron sputtering: (a) illustration of a standard rectangular dc mag-
netron target showing the magnetic field lines, an Ar plasma and an electron race track;
(b) enlargement of the electron race track; (c) photograph of a Cu target (35×55 mm).
Adapted from J. W. A. Robinson thesis [2].
the electrons within its vicinity enhances the ionisation of the plasma near the target
leading to a higher sputtering rate and improving its efficiency. It also means that
the plasma can be sustained at a lower pressure. The sputtered atoms are neutrally
charged and so are unaffected by the magnetic trap. Controlling the pressure of the
plasma and/or the power supplied to the magnetron enables the manipulation of the
kinetic energy of the Ar+ ions; this in turn controls the sputtering yield of the target
and the deposition rate on the substrate surface. There are a host of other sputtering
parameters that have an influence over the final outcome of the process such as the
substrate temperature, the oxygen/Ar gas pressure ratio, the total gas pressure, the
substrate-target distance, ztx.
HCG huwstrvtes
HCGCF iypes of huwstrvtes
The various thin film structures were grown on four different types of substrates; fused
amorphous silica glass, r - and x-plane sapphire (AlGOH) and single crystal silicon (Si)
substrates with a (100) crystallographic orientation and a 250 nm thermally grown
oxide on the surface (now termed Si (100)/SiOG substrates). Each chapter will specify
which of these has been used for the different structures investigated in this work. In
particular, silica glass is the typical substrate used for ITO thin films due to its excel-
lent optical transparency. Furthermore, ITO has been known to maintain favourable
properties even when amorphous and grown on silica glass. Silica glass substrates were
used for preliminary experiments where basic structural and electronic characterisation
of pure InGOG and ITO were carried out. However, it was found that there were some
27
Cb[pnel -. Erpelcgehn[f pli]edolem
odd magnetic contributions coming from the silica glass substrates, and we therefore
decided to switch to using r and x-AlGOH for all measurements where magnetic char-
acterisation was required.
HCGCG huwstrvte erepvrvtion
Substrates were typically cut using a diamond dicing saw. During this dicing process
the substrate was held in position under the saw by a transparent, acetone-soluble wax
to a glass microscope slide. Once cut, the chips were heated and removed from the glass
slide. The process leaves remains of various substances and dirt particles which are in
need of removal so to prevent the disruption of any nucleation and growth processes
or formation of impurities during thin film growth. Substrates were placed in a beaker
containing acetone and treated in an ultrasonic bath for 12 minutes. This process was
repeated three times before the substrates were finally soaked in ethanol and dried by
compressed air until optically clear. Prior to loading, the substrates were cleaned once
again.
HCH ihe bKII hputter Yeposition hystem
Thin films investigated in this work were prepared in the Mark II ultra-high vacuum
(UHV) sputter deposition system (Fig. 3.2). Once the substrate(s) had been loaded
into the system, it is roughed and then evacuated with a diffusion pump whilst baking
out overnight for 8 - 10 hours. This process removes material which has been absorbed
onto the chamber walls. The number of impurities within the chamber will be directly
proportional to the number of impurities incorporated into the growing films, and so
baking is of paramount importance. Typical base pressures achieved after baking and
over-night pumping are approximately 1 × 10−K Pa as measured by an inBsitu residual
gas analyser.
Two different flanges were used during the course of these investigations. Both
enabled three targets to be used simultaneously. Pure and doped InGOH and ITO
films were grown using a single target source with predetermined element composition.
Substrates were placed onto a Platinum (Pt) heater. The temperature calibration of
the Pt heater was performed using a pyrometer. All layered and stacked structures
were made using a rotating sample turntable which sits below the three magnetrons.
Note that when discussing growth temperatures in this thesis I am referring explicitly
to the temperature of the heater or sample stage. It is important to be aware that the
heater temperature is not the same as the substrate temperature or substrate surface
temperature. In fact, the difference can be as much as 120 - 150 ◦C [3], and will be
dependent on factors such as the type of substrate, the sputtering target, ztx.
For sputtering, argon and oxygen gases of purity better than 99.999 % were used.
28
Cb[pnel -. Erpelcgehn[f pli]edolem
Both argon and oxygen gases were bled into the gas reservoir one after the other. The
reservoir is closed off from the deposition chamber and allowed to mix for 20 minutes.
The reservoir is then roughed momentarily to create a turbulence to encourage further
mixing of the oxygen and argon gases before being injected into the main deposition
chamber.
HCI Yeposition pvrvmeters
The deposition rate of the sputtering target depends on a number of factors, namely;
rate of rotation of the sample stage as it moves below the target, the gas pressure,
the substrate-target distance, and linearly on the power supplied to the magnetron.
In this work, the deposition process was controlled using each of these parameters,
except for the gas pressure which was kept constant (1.50±0.01 Pa) for all film struc-
tures deposited in these investigations. All layered structures (grown using the rotating
turntable) had a constant substrate-target distance of 40 - 50 mm. The only additional
changes which were made to the rate of sputtering for these materials were done by
changing the width of the slit windows. Typically these were 7 mm wide and help to
control the substrate exposure time to the target [see Fig. 3.2 (c)]. The rate of rotation
was controlled using a LabView program [5] in conjunction with a stepper motor. For
pure and alloyed films grown on the Pt heater, the target-substrate distance varied
across the heater resulting in a concentration gradient along the heater.
Deposition rates for each target were calibrated in separate deposition runs, using
atomic force microscopy (AFM) (see section 3.5.2) in tapping mode from a step-edge
across a lift-off sample; this involved masking, with photoresist or permanent marker
pen, part of a substrate and depositing onto it. This created a step which was then
measured by the AFM. The rates were then back calculated using our knowledge of
deposition time, rotation speeds, ztx.
HC5 ihin [ilm Xhvrvxterisvtion
HC5CF mBgvy Yiffrvxtion stuyies
X-ray diffraction (XRD) was used for phase identification (θ-2θ scan) as well as low-
angle reflectivity (LAR) measurements to estimate film thicknesses and quality of mul-
tilayer structures. A Bruker D8 Advance was used for the θ-2θ scans and a Bruker axs
D8 Discover was used for the LAR measurements. The diffractometer used a CuKα
(λ = 1.5406 A˚) x-ray source. The peak positions in the θ-2θ scans were then later
verified using a diffraction PCPDFwin database. The vertical size of a crystal grain is
a measure of the size of a coherently diffracting domain. Scherrer derived an equation
29
Cb[pnel -. Erpelcgehn[f pli]edolem
1
3
4
2
Diffusion pump
Gas reservoir
Mechanical pump
Liquid nitrogen
inlet 
(a)
(b) (c)
Figure 3.2: Schematic of the MKII sputtering system. The black filled regions indicate
the flow of the mixture of sputtering gases. The grey regions accompanied by red
arrows indicate the pumping direction. (1) = inner-to-outer chamber valve, (2) =
main sputtering chamber, (3) digital pressure sensing Baratron, (4) mass spectrometer.
Adapted from J. W. A. Robinson thesis [2]; (b) a photograph of the tri-target sputtering
flange with targets installed; (c) a photograph of the sputtering shield with exposure
slits.
30
Cb[pnel -. Erpelcgehn[f pli]edolem
which relates the broadening of the diffraction peaks to the vertical size of a crystal [6],
yhkl =
ksλ
β cos θ
, (3.1)
where yhkl is the mean grain size perpendicular to the reflecting plane, k s is the Scher-
rer constant (somewhat arbitrary, but usually between 0.89 and 1, however, unity is
assumed here), λ is the wavelength of the chosen radiation (1.5406 A˚) and β is the
integral breadth of a reflection in radians at peak position 2θ identified from a stan-
dard θ-2θ scan. Using the Scherrer formula, one is able to calculate a value for the
average vertical grain size of the films and gain some insight in to the crystallographic
coherence throughout the film.
X-ray reflectivity is a powerful tool for investigating multilayered thin film struc-
tures and allows one to derive information on the interfaces, mass density of the layers,
and to estimate the corresponding stack thickness normal to the film surface with a
high degree of accuracy. Using this technique, x-rays are reflected from, and refracted
at, the interfaces which separate materials with different electron densities and there-
fore different refractive indices by virtue of electron - x-ray interaction. These scans
have a critical angle 2θx, which relates to the average electron density, and interference
Kiessig fringes which have periods relating to the total film thickness. As the scan-
ning angle increases beyond 2θx the x-rays penetrate deeper into the film resulting in
an exponential decrease in the recorded intensity. The film density and thickness can
generally be found using a modified Bragg law [7]. The thickness was extracted by
fitting the Kiessig fringe periods to a simulated scan using “Reflectivity” software.
HC5CG Vtomix [orxe bixrosxopy
The AFM Multiscope Nanoscope III (Digital Instruments) reveals topographic infor-
mation of the film, by scanning the surface with a microfabricated Si tip of nominal
radius 30 - 50 nm. The instrument was used in the tvpping moyz, wherein a piezo-
electric crystal is used to oscillate the tip over the film surface. The vertical resolution
of this instrument is around 0.10 nm. Information obtained from an AFM image can
provide us with insight into possible growth processes which may have dictated the
growth of the film.
HC5CH bvgnetometry
The magnetic properties were characterised using both DC and AC magnetometry.
In DC magnetometry the sample is magnetised by a constant magnetic field and the
magnetic moment of the sample is measured, producing a DC magnetisation curve. In
AC magnetic measurements, a small AC drive magnetic field is superimposed on top
31
Cb[pnel -. Erpelcgehn[f pli]edolem
0 1 2 3 4 5 6
-10
-5
0
5
10
15
-0.6 -0.4 -0.2 0.0 0.2 0.4 0.6
-100
-50
0
50
100
(b)
 
M
a
g
n
e
t
i
c
 
m
o
m
e
n
t
 
(
 
e
m
u
)
Position (cm)
(a)
  Raw data
M
a
g
n
e
t
i
c
 
m
o
m
e
n
t
 
(
 
e
m
u
)
0
H (T)
Diamagnetic
contribution
Figure 3.3: (a) Centering measurement over a scanning range of 6 cm to find the opti-
mum magnetisation signal within the SQUID sample space, (b) raw data for a typical
magnetisation loop at 2 K for a Co-doped ITO(10) thin film with linear components
arising from the diamagnetic background of the substrate (red line).
of the DC field, causing a time-dependent magnetic moment in the sample. Two differ-
ent methods of DC magnetometry were used: vibrating sample magnetometry (VSM)
and superconducting quantum interference device magnetometry (SQUID). The raw
magnetisation data from a sample will consist of two clear contributions. There is a
low field, saturating ferromagnetic contribution from the sample and a high field linear
dependency that is due to the diamagnetic contribution from the substrate [Fig. 3.3
(b)]. This may contain a paramagnetic contribution from unreacted components of the
film, which would also be linear in this high field range. To separate the ferromagnetic
contribution from the rest, we use a scan of the sample holder and blank substrate
(giving our linear diamagnetic background) and subtract it from the scan where a film
has been grown on top of a substrate. A source of error in this technique will be
the differences in volume of the substrates. The volume of the substrate is orders of
magnitude larger than that of the film and so we must be careful when performing the
background subtraction. One way to overcome this problem would be to measure the
same blank substrate that will be deposited upon later.
kiwrvting hvmple bvgnetometry
VSM measures the magnetic induction in a region where the sample is vibrated within
a magnetic field. The forced vibration induces an oscillatory change in the permeability
of the sample which is recorded as an alternating current signal by pick up coils. For
best results, it is advised that the entire sample must remain within the space bound
by the pick up coils. The maximum sensitivity of the VSM is approximately 2µemu.
An illustration of the VSM process can be found in Fig. 3.4.
32
Cb[pnel -. Erpelcgehn[f pli]edolem
Sampleflux
lines
Electromagnet
Sense coils
Computer
- td df  /1
+d df  / t2
Figure 3.4: An illustration of the VSM used to measure the magnetic properties of
some of the samples discussed in this thesis. Figure adapted from J. W. A. Robinson
thesis [2].
huperxonyuxting fuvntum Interferenxe Yevixe bvgnetometry
A superconducting quantum interference device magnetometer, or SQUID, is a very
sensitive device used for making magnetisation measurements and is typically 100 times
more sensitive than a VSM. A SQUID works on the principle of measuring a voltage
induced by the magnetic field from a sample in a field-sensing coil. The SQUID consists
of a superconducting ring with a small insulating layer called a Josephson junction. The
flux passing through the ring is quantised once the ring has become superconducting but
the Josephson junction enables the trapped flux within the ring to change by discrete
amounts. Without the Josephson junction, flux cannot enter the ring. Therefore, the
device can be used to measure the changes in flux quanta within the ring. The system
must be kept at liquid helium temperatures to operate. A least-squares fitting program
converts the variation in flux to a theoretical expression that gives a magnetisation
value.
Once in the system, the sample must be positioned so that the centre of the sample
corresponds to the centre of the field-sensing coils. A scan over a length of 6 cm is used
to identify where the sample is located [Fig. 3.3 (a)], after which it can be centred
either manually or automatically.
33
Cb[pnel -. Erpelcgehn[f pli]edolem
VX susxeptiwility
As explained above, the induced sample moment is time-dependent for AC measure-
ments such that the measurement yields information about magnetisation dynamics
which are not obtained in DC measurements. When the alternating magnetic field has
a very low frequency the measurement is most similar to DC magnetometry. However,
at much higher frequencies the AC moment of the sample does not follow along the
DC magnetisation curve due to dynamic effects in the sample - for this reason, AC
susceptibility is often know as the dynamic susceptibility. At these higher frequencies,
the magnetisation will in fact lag behind the drive field, creating a phase shift.
The AC susceptibility measurement yields two quantities: the magnitude of the
susceptibility, χ, and the phase-shift, ψ relative to the drive signal. The susceptibility
can be thought of in an alternative manner: an in-phase or real component, χ’, or
an out-of-phase or imaginary component, χ”. Using AC susceptibility measurements
superparamagnetism and spin glass behaviour may be probed where we expect to see
shifts in χ’ and χ” as a function of the alternating magnetic field frequency when mea-
sured as a function of temperature.
AC susceptibility measurements were performed using CRYOGENIC LIMITED
cryogen-free system multipurpose system in “VSM” mode where it was possible to
apply a small AC drive magnetic field over a DC field and measure the susceptibility
as a function of temperature.
HC5CI Zlextrixvl bevsurements
gesistivity erowe
Measurement of the sample resistance as a function of temperature was carried out
by four-point resistance measurements (in a colinear arrangement) using a custom-
built probe. This technique separates the voltage and current connections so that
the interfacial effects (from the connection of the contact points to the sample) are
circumvented and it is purely the resistance of the sample area of interest that is
measured [8]. Sample films are carefully inserted into the space between the contact
pogo pins and the insulating base platform. An adjustable head containing the contact
pin arrangement is then brought into contact with the film. This ensemble is then
placed into a liquid helium dewar, where the temperature of the sample is varied by
the height of the sample with respect to the level of helium. A minimum temperature
of 4.8 K was achieved resulting from almost complete submersion.
irvnsport mevsurements
Magnetotransport measurements were performed through standard four-point mea-
surements on a CRYOGENIC LIMITED cryogen-free system. In some cases gold
34
Cb[pnel -. Erpelcgehn[f pli]edolem
1 4
2 3V+
V
-
I+
I
-
2 3
1 4
VI
1 4
V
I
2 3
R V I= /v 2,4 1,3
R V I= /
h 1,4 2,3
Magneticfield
Figure 3.5: Schematic of the configurations used in a Hall measurement to determine
k H .
contacts were sputtered prior to the measurements in order to improve the electrical
contact between the wire bonds and the film surface. Magnetoresistance measurements
used a colinear contact configuration with an applied magnetic field applied either par-
allel or perpendicular to the applied current. The carrier concentration and mobility
were extracted using Hall measurements in a van der Pauw contact arrangement [8]
(see Fig. 3.5).
Charge carriers within an electric current flowing through a magnetic field in a
conductor will experience a transverse force which will tend to drift carriers away from
the original current direction, resulting in an excess surface electrical charge on one
side of the sample. This charge results in the Hall voltage (k H), and is measured as
the potential drop across the two sides of the sample. The resulting transverse force is
k H , expressed as;
kH =
IµEH
qns
, (3.2)
where I is the applied current through the sample, µE] is the applied magnetic field,
q is the charge on a single electron and ns is the sheet carrier density (cm
−G). Mobility
(µH) can be obtained once a value for the sheet resistance, gs, has been determined
through the relationship;
µH =
kH
gsIµEH
=
1
qnsgs
(3.3)
A value for gs can be obtained by considering two characteristic resistances: the
vertical, gv, and horizontal, gh, related by the van der Pauw equation;
e−.gTRgs + e−.ghRgs = 1 (3.4)
Measurement of k H consists of making a series of voltage measurements with a
constant current and a magnetic field perpendicular to the sample and measured by
applying a current through diagonal contacts (e.g. I FPH, measuring k GPI). Eight mea-
surements of k H , similar in nature are made (k GPI, k IPG, k FPH, k HPF) for both positive
35
Cb[pnel -. Erpelcgehn[f pli]edolem
and negative fields (see Fig. 3.5).
HC5C5 bvgnetix Xirxulvr Yixhroism (bXY)
dptixvl bXY (dBbXY)
Magneto-optical (MO) measurements presented in this thesis were made at the Univer-
sity of Sheffield. The experimental set-up is based upon the Sato [9, 10] transmission
method that allows the simultaneous measurement of two magneto-optical parameters,
MCD and Faraday rotation. The principles of the MO effects are explained in more
detail in chapter 7.
The sample is mounted on a disc with an aperture, in a cold finger cryostat which
allows measurements to be made in the range of 5 - 400 K. The sample space is located
between two electromagnetic poles which can achieve applied magnetic fields of up
to approximately 0.50 T, applied perpendicular to the sample. There is also another
option available, whereby the sample is not mounted using the cryostat. In this config-
uration it is possible to bring the electromagnetic poles much closer together, thereby
increasing the maximum applied magnetic field. Here the calibrated maximum applied
magnetic field was approximately 1.80 T.
As shown in Fig. 3.6, the setup can be altered between Kerr and Faraday geometry
which are used for samples that reflect and transmit light respectively. This is achieved
by simply moving the photoelastic modulator (PEM), analyser and photolultiplier tube
(PMT) to the position required in the arrangement shown. The setup uses a xenon
lamp which is useful over spectral ranges of 1.5 - 4.5 eV respectively.
The light propagates through a Spectro-275 spectrometer to produce monochro-
matic light. This has three diffraction gratings blazed at different wavelengths for
improved accuracy across a wide spectral range. We can select the diffraction grating
most appropriate for the wavelength of light required. However there are still unwanted
wavelengths and so a bandpass filter is used in the setup. Before it passes through the
sample, the light must be plane polarised. To do so, it is passed through a UV prism
polariser.
The light beam is focused to a size similar to that of the aperture which it is to
pass through onto the sample surface, and is achieved with a series of mirrors. The
mirrors must be optimised regularly to keep the maximum amount of light passing
through the sample. Any scattering will directly reduce the sensitivity of the experi-
ment. The role of the PEM is to produce a signal that is proportional to the rotation
and ellipticity, allowing them to be measured simultaneously. It consists of a birefrin-
gent crystal mounted on a piezo-vibrator of characteristic modulating frequency, f [12].
The birefringent crystal is vibrated, causing a periodic advancement and retardation
of the transversing light phase, which occurs parallel to the vibration direction. This
means that the light that emerges will alternate between LCP and RCP light with a
36
Cb[pnel -. Erpelcgehn[f pli]edolem
PEMcontroller
1 Lock in
amplifier
2 Lock in
amplifier
3 Lock in
amplifier
st
nd
rd
PC
PMT
Temperature
controller
Signal
Amplifier
Bias feedback
DC signal
PMT
PMT
PEM
PEM
Sample
(T=10-400K)
Polariser
Filter
Monochromator
Xenon
Lamp
Analyser
Analyser
Faraday geometry (Transmission)Kerr geometry (Reflection)
Figure 3.6: Magneto-Optic setup for measurements in either Faraday or Kerr geome-
try. Adapted from A. J. Behan thesis [11].
frequency f.
The light then passes through an analyser before it is focused on the PMT. Any
MCD or rotation exhibited by the sample will introduce oscillations in the light inten-
sity with a frequency f or 2f respectively. Therefore, if no MCD effect is present, the
intensity at the detector will remain constant.
The PMT will then convert the light intensity into an electrical signal. The signal
amplifier unit is initially used to amplify the signal and after, split it into the AC and
DC components respectively. Using a Keithley voltmeter, the DC voltage is read and
kept constant by controlling the gain of the PMT to maintain a constant sensitivity.
The AC component is measured using two lock-in amplifiers. Sato shows in ref. [9]
using complex theory how the measurement of the three intensities (I E, I f , and I Gf )
can then be translated to obtain values for the MCD and Faraday rotation.
The system is automated with instruments that are controlled from a PC using Lab-
VIEW. The measurements of the magneto-optical spectra are automatically repeated
to give an average and hence reduce the noise levels. The current of the electromagnet
can be swept to produce a magneto-optical hysteresis loop, or the wavelength swept to
produce a spectra.
37
Cb[pnel -. Erpelcgehn[f pli]edolem
Figure 3.7: Schematic representation of a typical synchrotron radiation facility. (i)
initial acceleration stages, (ii) the storage ring, (iii) insertion devices used to manipulate
and polarise the beam, and (iv) reaching the end station where the sample is located
and can be characterised.
mBgvy bXY (mbXY)
High intensity, tuned x-rays available at a synchrotron radiation facility has allowed
unparalleled access in the observation of magnetic phenomena. Synchrotron radiation
is emitted when charged particles are accelerated close to the speed of light and are
deflected by a magnetic field. The availability of its high brilliance and other supe-
rior properties compared to conventional light sources make synchrotron radiation an
extremely versatile tool. The XMCD measurements presented in this thesis were per-
formed at Beamline I06 at Diamond Light Source. The principles of XMCD effects are
explained in more detail in chapter 6.
Production of synchrotron radiation requires several components: a primary elec-
tron accelerating system, a synchrotron accelerator for further acceleration of electrons,
a storage ring to store the accelerated electrons, insertion devices in the storage ring
to produce the radiation, beamline systems equipped with monochromators to tune
and direct the radiation to an experimental end-station, and an undulator to provide
polarisation of the synchrotron source (see Fig. 3.7).
Electrons are initially emitted by a hot cathode through thermionic emission in
high vacuum and accelerated by an anode voltage before being accelerated again in the
microtron. The microtron is a linear accelerator where electrons pass through an elec-
38
Cb[pnel -. Erpelcgehn[f pli]edolem
tric field many times before being redirected on larger growing trajectories by a strong
magnetic field. The electrons then pass through to the synchrotron accelerator where
they are focused on their final trajectory and energy (approximately 3 GeV at Diamond
Light Source). The electrons then enter the storage ring through a transfer channel
where they gradually decay and are reinjected every 12 hours. In order to maximise the
density of electrons which have near total circular polarisation, undulators (a periodic
array of magnets) are often inserted in the storage ring. In an undulator, the electron
beam is periodically deflected by weak magnetic fields. The electrons emit radiation
at the wavelength of their periodic motion in the undulator. The spectral resolution
of the emitted radiation is proportional to the number of undulator periods and the
radiation wavelength can be changed by varying the magnetic field. The x-rays then
pass through an arrangement of slits, mirrors and monochromators to deliver light of
desired energy and polarisation to the beamline end station.
39
Wiwliogrvphy
[1] Z. Barber, Journal of Materials Chemistry, 16, 334-344, (2006).
[2] J. W. A. Robinson, PhD Thesis, Zero to Pi Oscillations in Ferromagnetic Joseph-
son Junctions, University of Cambridge, (2007).
[3] Z. Barber, Private communication.
[4] R. E. Somekh and Z. H. Barber, Journal of Physics E: Scientific Instruments,
21(11):1029-1033, (1988).
[5] Program created and coded by Gavin Burnell
[6] B. D. Cullity, Elements of X-Ray Diffraction, 2nd Edition MA: Addison Wesley,
Page 102, (1978).
[7] A. Segmuller, Thin Solid Films, 18(2):287-294, (1973).
[8] I. J. van der Pauw, Philip. Res. Repts., 13:1-9, (1958).
[9] K. Sato, Jpn. J. Appl. Phys. 20, No. 12, (1981).
[10] W.P. van Drent and T. Suzuki, J. Magn. Magn. Mater., 175, 53 (1997).
[11] A. J. Behan, PhD Thesis, Characterisation of Doped ZnO Thin Films for Spin-
tronic Applications, (2008).
[12] K.W. Hipps and G.A. Crosby, J. Physical Chemistry, 83, 555, (1979).
40
Xhvpter I
ihe Inyium dfliye system
ihis xhvptzr introyuxzs thz Inyium dfliyz systzmA vny hofi it xvn wz mvnipulvtzy to
owtvin Inyium iin dfliyzA thz primvry mvtzrivl systzm uszy for our invzstigvtionsC
erzvious fiork pzrformzy wy othzrs is yzsxriwzy vny rzlvtzy to thz finyings in this thzsisC
ihz suitvwility of thzsz mvtzrivls for suxxzssful trvnsfzr to thz fizly of ofliyz spintronixs
is xonsiyzrzy.
41
Cb[pnel .. Tbe Chdcog Ircde msmneg
ICF Introyuxtion
Transparent conducting oxide (TCO) thin films and techniques for their fabrication
have been investigated for many years because of their unique properties and extensive
application potential [1]. The mass interest in Indium Oxide (InGOH) thin films stems
from its dual ownership of high optical transparency together with electronic conduc-
tivity. In view of the high practical importance of InGOH thin films, a great deal of
research and development has been carried out on the electrical and optical properties.
Rupprecht [2] performed some of the earliest experiments on the InGOH semiconductor.
Inspired by this pioneering work, others later followed suit investigating the various
methods to produce InGOH-based materials such as direct current (DC) magnetron
sputtering, radio frequency (rf) sputtering, chemical vapour deposition (CVD), vac-
uum evaporation (VE), sol gel reaction, ztx, in an attempt to meet both economic and
technological needs. In particular, considerable interest has been devoted to the study
of Tin (Sn)-doped InGOH (ITO). This product of InGOH also holds very similar optical
and electrical properties as pure, undoped InGOH, and so is also of great technological
interest.
There is a fundamental interest in the conduction mechanisms within TCO ma-
terial systems. A great deal of work has been carried out in trying to correlate the
microstructural characteristics in an attempt to explain the physical reasons behind
conduction in InGOH and ITO-based thin films. One of the most important aspects of
working with these materials from a technological point of view is the balance between
the electrical and optical properties, i.e. achieving high optical transmission whilst
maintaining minimal sheet resistance. However, these two properties are not indepen-
dent, and optical transparency in the visible region is strongly affected by the electrical
properties of the film. Usually, good electrical properties can be achieved but at the
expense of transmission.
The electrical properties of the oxides depend heavily on the oxidation state of the
metal component, i.e. the stoichiometry of the oxide, and on the nature and quantity
of the impurities incorporated within the film. A perfectly stoichiometric oxide will be
insulating. Effective doping can be achieved provided that the dopant is of the same
size or smaller than the host ion it will be replacing, and that no secondary compounds
are formed between the dopant and host atoms.
The properties of ITO films are strongly dependent on the preparation techniques.
A huge archive of data collected on the various electrical and optical parameters of
ITO exist, none of which are identical. Because of this, understanding the limiting
physical phenomena in these materials is difficult. This chapter aims to introduce the
InGOH and ITO system and provide a relatively up-to-date picture of the work done on
the electrical and optical properties of InGOH and ITO based thin films and the various
methods used to make them. The conduction mechanisms which govern the electrical
properties will also be discussed along with the the results obtained with our materials.
42
Cb[pnel .. Tbe Chdcog Ircde msmneg
In2
In1
In2: sited
In1 : siteb
Oxygen
Vacant site
Figure 4.1: The crystal structure of InGOH.
ICG ihe Inyium dfliye system
InGOH exists in a modified cubic crystallographic phase (space group Ia
H) called the
wiflwyitz structure (also called x-type rare earth oxide structure) [3]. A conventional
unit cell consists of 16 formula units of InGOH where one fourth of the oxygen anions
located along the four <111S axes are missing. Indium cations are located in two
non-equivalent sites (see Figure 4.1), with 8 InH+ ions located in the centre of trig-
onally distorted oxygen octahedrons (w site). The remaining 24 InH+ ions located in
the centre of the more distorted octahedrons (y site) [4]. InGOH is reported to have a
lattice constant of 1.012 nm [5].
InGOH is a poor conducting material when stoichiometric. However, as deposited
InGOH generally lacks stoichiometry due to oxygen vacancies. At high concentrations
of oxygen vacancies, an oxygen vacancy impurity band forms and overlaps E x at the
bottom of the conduction band producing a degenerate semiconductor. Oxygen vacan-
cies act as doubly ionised donors, contributing two electrons to the conductivity. The
material composition can be represented as InGOH−x. As well as providing conduction
electrons, the oxygen vacancies also allow for OG− ion mobility. Generally, additional
electron carriers can be supplied by doping InGOH with four or more valence electrons.
In Sn-doped InGOH (ITO), Sn acts as a cationic dopant in the InGOH lattice and sub-
stitutes for In, which has a valence of three (InH+). This results in n-type doping of
the InGOH lattice by providing an electron to the conduction band. High electron con-
centrations (10FL- 10GE cm−H) can be achieved either with an oxygen deficiency (e.g.,
InGOH−x) or an excess of metal atoms (e.g. InG+yOH).
The Sn dopant essentially forms an interstitial bond with oxygen present in the
system and will exist as either SnO (valency +2) or SnOG (valency +4), the outcome
of which will have a direct influence on the overall conductivity of the ITO film. A
valency of +2 will result in a reduction in the conductivity of the ITO film as it will
create a hole, which subsequently behaves as a “carrier trap”, reducing the overall car-
43
Cb[pnel .. Tbe Chdcog Ircde msmneg
wh
Eg0
Eg
K K
wh
PureIn O2 3 Sn doping
’
Figure 4.2: The assumed parabolic band structure of pure, undoped InGOH (left) as
well as the effect of doping the lattice with Sn (right). Adapted from [7].
rier concentration. On the other hand, predominance of the SnOG state results in the
SnI+ acting as an n-type donor that releases electrons to the ITO conduction band,
thus improving ITO conductivity.
SnI+ ions also have another role to play in the electrical behaviour of ITO: they
act as scattering centres for electrons inside the InGOH lattice. This creates a trade-off
between electron carrier density and mobility when trying to optimise the conductivity
of an ITO film. The formal description of ITO can thus be written as InG−xSnxOH−Gx.
Optoelectronic properties of materials are dominated by the electronic struc-
ture near the bandgap. The understanding of the band structure of ITO is based on
the theory behind InGOH. In order to explain the conductivity mechanism of InGOH
when doped with Sn, Hamberg and Granqvist have assumed a simple parabolic band
structure (Fig. 4.2) with an intrinsic wide semiconductor bandgap of 3.75 eV [6]. Ad-
dition of Sn dopants results in the formation of donor states just below the conduction
band. As the doping levels increase, these donor states will eventually merge with the
conduction band. Once the material becomes degenerate, the mutual exchange and
Coulombic interactions shift the conduction band downwards and the valence band
upwards - effectively narrowing the band gap, E g, from E gE to E g
′
[7].
ICH Zlextrixvl Xonyuxtion mexhvnisms
Electrical conductivity (σ) depends on the concentration (nx) and mobility (µH) of a
free electron carrier according to the relationship:
σ = nxµHq, (4.1)
44
Cb[pnel .. Tbe Chdcog Ircde msmneg
or, in terms of resistivity, /,
/ =
1
nxµHq
, (4.2)
where q is the electron charge. Therefore, it is clear that to obtain high conductivity,
it is important to have a high carrier concentration and mobility simultaneously.
Most efforts to improve the conductivity of ITO thin films have focused on in-
creasing the number of free carriers via extrinsic doping. However, this technique
is self-limiting, as the process of increased doping will naturally impair the electron
mobility. Therefore, obtaining the highest conductivity is a trade-off between carrier
concentration and electron mobility. The two subsections below describe the limiting
factors to these two important conduction parameters. For a more detailed review the
reader is referred to reference [8].
ICHCF [ree xvrrier xonxentrvtion
Previously it was stated that a high carrier concentration can be achieved either by
introducing non-stoichiometry, i.e. creating oxygen vacancies, or an excess of metal
atoms, or extrinsic doping, e.g., with Sn. Frank and Kostlin have performed extensive
studies into the electrical properties of ITO films as a function of varying Sn concen-
tration [9]. Doping does not always result in a continual rise in carrier concentration
and improved electrical properties. The dopants can behave in a number of different
ways, e.g., they may occupy interstitial sites, form defects acting as carrier traps, or be
the cause of phonon or impurity scattering centres, leading to a fall in carrier mobility.
Frank and Kostlin listed four dominating lattice defects which they believe influence
the electrical properties of ITO films:
1. Impurity ions - Sn acts as a cationic dopant resulting in n-type doping by pro-
viding an electron to the conduction band.
2. czutrvl yzfzxts - Two neutral defects exist; those SnI+ ions remote from nearest
neighbour positions become loosely bound to an interstitial oxygen anion and
form a neutral compound (SnG
+Oi
G−). This interstitial neutral defect will disso-
ciate on annealing under reducing conditions, such that the OG− can drift out.
If the SnI+ ions are close enough to be bound to the three nearest neighbours
on regular anion sites as well as an additional interstitial oxygen ion, (SnGOI)i is
formed. This is usually the case at high Sn doping levels. A rise in Sn doping is
likely to result in an increase in the neutral scattering centre density.
3. Compouny nzutrvl yzfzxt - if there is a surplus of Sn as well as loosely bound
oxygen, the two neutral defects described above may combine.
4. dflygzn vvxvnxy - oxygen vacancies will act as doubly ionised donors. Deposition
within oxygen rich conditions will result in a film saturated with oxygen which
45
Cb[pnel .. Tbe Chdcog Ircde msmneg
Figure 4.3: Dependence of the free-electron density c z on the Sn concentration, x.
v indicates the region where Sn doping is low and the rise in carrier concentration
is linear, indicating that every Sn atom behaves as a donor. w represents the region
beyond 8% doping where the carrier concentration soon begins to fall off likely due
to Sn ions occupying nearest neighbour anion sites and forming neutral defects. After
reference [10].
for the purposes of improving the electrical properties is highly undesirable. In-
terstitial oxygen can diffuse through the ITO lattice reaching grain boundaries.
Exposing the film to higher temperatures has the potential to create free electrons
and dissolve interstitial oxygen to promote substitutional Sn.
It has been found that very high dopant concentrations (beyond a critical limit) result
in the observed ITO film carrier concentration being lower than expected. This implies
that a portion of Sn atoms are electrically inactive and is likely due to an increased
probability of the occupation of adjacent cation positions by two or more Sn atoms.
Figure 4.3 illustrates the dependence of the free carrier concentration on the level of
Sn doping [10]. The highest carrier concentration, which corresponds to the lowest
resistivity, occurs when the Sn doping level is approximately 10%. The solid solubility
of Sn in InGOH is approximately 8%. Beyond this range, a Sn oxide phase may form
and distort the lattice structure resulting in a decrease in the carrier concentration.
ICHCG Xvrrier mowility
From equation 4.1 we can see that to improve conductivity, we require an increase in
carrier concentration or mobility, or both. However, increasing carrier concentration
can lead to an increase in visible absorption and hence a reduction in transparency.
Therefore, increasing the mobility is extremely vital for ITO films in order to achieve
good electrical and optical properties.
The free carrier mobility, µH , can be defined as:
µH = qτRm
∗, (4.3)
46
Cb[pnel .. Tbe Chdcog Ircde msmneg
where q is the electron charge, τ is the average collision time of electrons and m∗ is
the effective electron mass in the conduction band. The mobility of the charge carriers
is strongly influenced by the disorder within the crystal structure. Mobility of the free
carriers is therefore strongly affected by the different scattering processes of free charge
carriers. These are summarised below.
1. Grvin wounyvry sxvttzring - these play a significant role in polycrystalline films.
Films of this nature tend to contain high densities of interface states which are
capable of trapping free carriers from the bulk of the grain [11]. The mobility
due to grain boundary scattering, µg, can be expressed as [12]:
µg = qD(2.m
∗kBi )−FRGexp
(
− qϕw
kBi
)
, (4.4)
where D is the grain size, ϕw is the grain boundary potential barrier, q is the
charge of a single electron, kB is Boltzmann constant, i is the absolute temper-
ature in Kelvin and m∗ is as before. From equation 4.4, it is evident that films
possessing larger grains and hence smaller grain boundary potential suffer less
from grain boundary scattering.
2. Ioniszy impurity sxvttzring - it has been suggested in some models of degenerate
semiconductors that the contribution from these impurities to the mobility of
free charge carriers can be described as follows [13]:
µi =
4e
3
(.
3
)
n
−GRH
i , (4.5)
where n i is the impurity concentration and the free electron carrier concentration
is usually assumed to be equal to the impurity concentration. Mobility therefore
decreases with the concentration of impurities.
3. avttixz sxvttzring - propagation of acoustic or optical waves through a lattice can
cause as electron-phonon scatter event inducing vibrations within the lattice.
The mobility due to this kind of scattering depends strongly on the temperature.
An increase in the temperature will result in more lattice vibration and therefore
a decrease in carrier mobility [14]:
µl =
(8.)FRGqh¯IC
3Eam∗5RG(kBi )HRG
, (4.6)
where C is the elastic constant, Ea is the shift in the edge of the conduction
band with temperature and the other constants are as before.
The total mobility, µtotvl, as a result of all the scattering sources can be represented by
the sum of the reciprocal values for µg, µi, µl.
1
µtotvl
=
1
µg
+
1
µi
+
1
µl
+ O O O+ O O O (4.7)
47
Cb[pnel .. Tbe Chdcog Ircde msmneg
ICI erevious lork
A large portion of the previous research has focused on the influence that the numerous
deposition techniques and conditions (table 4.1), post growth treatments [15–18] and
substrate effects [19, 20] may have on the electrical and optical properties of the ITO
films. Bel Hadj Tahar zt vl. provide a comprehensive review of the varying properties
of ITO when grown using a variety of deposition techniques [8].
ICICF Zffexts of suwstrvte tempervture
Lee performed an investigation into the behaviour of the carrier concentration and mo-
bility for reactive DC magnetron sputtered ITO thin films grown on alkaline-free glass
substrates at various process temperatures (70 - 380◦C) [38]. It was suggested that the
dependence of carrier concentration and mobility on the substrate temperature could
be divided into two distinct regions based on a shift in dominant scattering mecha-
nisms. In region 1 carrier concentration and mobility both increase with substrate
temperature as a result of increasing crystallinity and grain size; in region 2 carrier
concentration continues to rise but mobility falls due to ionised impurities which act
as scattering centres and trap the free electron carriers. Electron carriers have a much
clearer conduction path in well-ordered crystalline structures as opposed to a disordered
amorphous structure. The rise in substrate temperature also promotes favourable con-
ditions for Sn adatomic substitution.
ICICG Zffexts of oflygen pvrtivl pressure
It is widely accepted that excess oxygen present during the growth of ITO films is
detrimental to the electronic properties of the film. Many argue that this is largely
due to two main factors. The first one is connected to the fact that crystallinity wors-
ens compounded with a decreasing average grain size and hence a greater number of
grain boundaries. Also, adsorption of oxygen into the films could inhibit the surface
migration of In or Sn atoms leading to structural disorder. The second factor relates
to excess oxygen filling oxygen vacancies and thus deactivating the substituted SnI+
ions by forming Sn+O− complexes. Ultimately, this results in a decrease in carrier
concentration.
However, again we meet the compromise between the best electrical and optical
properties. High oxygen partial pressures yields films with superior optical trans-
parency but lower carrier concentration and therefore, reduced electrical quality.
48
Cb[pnel .. Tbe Chdcog Ircde msmneg
ICICH eostBgrofith vnnevling trevtment
The most common method of manipulating the as-deposited films is to use some form
of post-growth annealing treatment [15–18]. Changes observed with the annealing tem-
perature have been associated with transformations in the local ordering of the material
during re-crystallisation such as the reduction in structural distortion and increases in
mean crystallite size. Annealing treatment may also lead to the creation or annihi-
lation of oxygen vacancies by releasing trapped donor sites at grain boundaries and
dislocations as well as encouraging the out-diffusion of oxygen atoms from interstitial
positions.
ICICI Xrystvllogrvphixvlly orientey films
In a drive to improve the carrier mobility research has been carried out focusing on
developing preferentially oriented or epitaxial InGOH and ITO films on substrates such
as yttria-stablised zirconia (YSZ) [39]. It was thought that by improving the crystalline
quality, thus minimising large angle grain boundaries and defects, the conduction mech-
anisms of ITO could be understood better. Comparing the electrical properties of the
epitaxial ITO film with that of a polycrystalline film grown on glass, Kamei zt vl.
found no improvement and concluded that neither large angle grain boundaries nor
crystalline orientation were dominant electron scattering factors in ITO films. In fact,
they claim that the dominant factors are electron scattering events at point defects
such as ionised or neutral impurity scattering. Similarly, Fan zt vl. found that the
electrical properties of ITO films grown on single crystal AlGOH are not significantly
influenced by cystallographic orientation or the substrate material [40].
49
Cb[pnel .. Tbe Chdcog Ircde msmneg
T
ab
le
4.
1:
S
u
m
m
ar
y
of
th
e
el
ec
tr
ic
al
(/
=
re
si
st
iv
it
y,
n
x
=
ca
rr
ie
r
co
n
ce
n
tr
at
io
n
,
µ
H
=
el
ec
tr
on
ca
rr
ie
r
m
ob
il
it
y
)
an
d
op
ti
ca
l
p
ro
p
er
ti
es
of
IT
O
th
in
fi
lm
s
at
ro
om
te
m
p
er
at
u
re
gr
ow
n
u
si
n
g
a
va
ri
et
y
of
d
ep
os
it
io
n
te
ch
n
iq
u
es
.
D
C
M
S
=
D
C
m
ag
n
et
ro
n
sp
u
tt
er
in
g,
R
F
S
=
ra
d
io
fr
eq
u
en
cy
sp
u
tt
er
in
g,
P
L
D
=
p
u
ls
ed
la
se
r
d
ep
os
it
io
n
,
C
V
D
=
ch
em
ic
al
va
p
ou
r
d
ep
os
it
io
n
,
V
E
=
va
cu
u
m
ev
ap
or
at
io
n
.
q
ec
h
n
iq
u
e

n
x

H
p
u
b
st
ra
te
q
ra
n
sp
a
re
n
cy
o
ef
er
en
ce
(×
10
−I
Ω
cm
)
(×
10
G
E
cm
−H
)
(c
m
G
V
−F
s−
F
)
(%
)
D
C
M
S
7.
00
6.
00
25
.0
0
G
la
ss
..
.
[2
1]
D
C
M
S
3.
00
4.
00
40
.0
0
G
la
ss
S
80
[2
2]
D
C
M
S
1.
40
9.
00
46
.0
0
G
la
ss
S
90
[2
3]
D
C
M
S
37
0
7.
00
24
.5
0
G
la
ss
..
.
[2
4]
D
C
M
S
1.
60
8.
10
47
.0
0
..
.
..
.
[2
5]
R
F
S
4.
00
4.
60
18
.0
0
G
la
ss
≈
90
[2
6]
R
F
S
0.
68
36
27
.0
0
..
.
..
.
[2
7]
R
F
S
2.
55
15
.9
15
.4
..
.
..
.
[2
8]
R
F
S
13
.8
2.
60
17
.3
9
..
.
..
.
[2
9]
P
L
D
8.
00
10
-1
2
..
.
..
.
≈
90
[3
0]
P
L
D
3.
00
5.
00
40
.0
0
..
.
80
[3
1]
P
L
D
1.
30
10
35
.0
0
S
i,
G
la
ss
85
[3
2]
C
V
D
1.
60
-1
.8
0
9.
00
43
.0
0
..
.
..
.
[3
3]
S
ol
-g
el
12
1.
20
7.
00
..
.
..
.
[3
4]
S
ol
-g
el
3.
30
10
21
..
.
..
.
[3
5]
V
E
0.
70
20
-3
0
≈
10
..
.
..
.
[3
6]
V
E
0.
44
13
.8
10
3
..
.
..
.
[3
7]
V
E
6.
80
5.
70
16
..
.
..
.
[3
7]
50
Cb[pnel .. Tbe Chdcog Ircde msmneg
Table 4.2: Comparison of the room temperature electrical properties (resistivity, /,
carrier concentration, nx, and electron mobility, µH) of pure InGOH and ITO grown
using a heater temperature of 300◦C and an oxygen partial pressure of 0.15 Pa with
varying levels of extrinsic Sn doping.
pn concentration  nx H
(at. %) (Ω cm) (cm−H) (cmG V−F s−F)
0.0 3.00±0.32×10−F 7.90±0.36×10FM 2.65±0.28
5.0 1.45±0.09×10−H 1.83±0.15×10GE 23.58±0.14
10.0 2.33±0.08×10−I 8.84±0.08×10GE 25.22±0.14
IC5 Zflperimentvl gesults
InGOH and Sn-doped InGOH films were grown using DC magnetron sputtering from sin-
gle commercial targets onto silica glass substrates as described in chapter 3. The power
applied to the targets was kept constant at 20 W in both cases resulting in a deposition
rate of approximately 40 nm/min for each target. The gas pressure was maintained at
1.50±0.01 Pa with a gas flow rate of approximately 0.10±0.01 Pa/s. To investigate the
effects of temperature on the samples, heater temperatures were varied between room
temperature and 600◦C. When investigating the effects of oxygen partial pressure, an
oxygen partial pressure range of 0 - 0.60 Pa was used.
ICK Zlextrixvl properties
Yepenyenxe on hn xonxentrvtion
The Sn content in InGOH was varied in the range from 0 to 10 at%. Table 4.2 summarises
the values for resistivity (/), carrier concentration (nx) and electron mobility (µH) at
room temperature for thin films of pure InGOH and Sn-doped (5 and 10 at.%) InGOH
[now called ITO(5) and ITO(10)]. The electronic properties of the films improve with
increasing Sn doping. Pure InGOH samples were semicondcuting whereas those doped
with Sn showed metallic behaviour. The best electrical properties are found for ITO(10)
films. For this reason, the majority of the work presented in the rest of the thesis is
concerned with ITO(10).
Magnetoresistance (MR) measurements of a series of pure InGOH, ITO(5) and
ITO(10) grown using an oxygen partial pressure of 0.15 Pa and heater temperature
of 300◦C are shown in Fig. 4.4 (a). The samples doped with Sn show characteristics
typical of the ordinary MR effect for a semiconducting or metal thin film, indicating
weak localisation behaviour. Magnetoresistance is negative for all films suggesting
the reduction in spin disorder as the alignment increases with field. A plot of how
the magnitude of MR (at an applied magnetic field of 5 T) depends on the carrier
51
Cb[pnel .. Tbe Chdcog Ircde msmneg
Figure 4.4: (a) Magnetoresistance curves for pure InGOH (black), ITO(5) (red) and
ITO(10) (blue) thin films grown using an oxygen partial pressure of 0.15 Pa and a
heater temperature of 300◦C at 10 K. MR curves of ITO(5) and ITO(10) have been
multiplied by a factor of 4. An ITO(10) film grown on an unheated substrate is shown
in green [ITO(10)∗] and has also been multiplied by a factor of 4, (b) Dependence of
MR at 10 K on the carrier concentration under an applied magnetic field of 5 T.
concentration can be found in Fig. 4.4 (b). Samples with lower carrier concentration
possess a much larger negative MR, which is largest for non-Sn-doped samples. Two
electron waves may travel in opposite directions along the same closed path, being
elastically scattered by the same impurities. This will lead to a constructive interference
and an additional resistance. Upon application of a magnetic field, the two waves
will acquire a phase difference and the additional resistance present in the absence
of a magnetic field will decrease. Therefore, the addition of Sn may reduce the weak
localisation effect and the constructive interference. The observation of negative MR in
these samples may be related to the high density of oxygen vacancies. The suggestion
we make about a reduction in spin disorder with increasing field implies that the
samples may be also be magnetic (see section 5.6.6).
Zffexts of suwstrvte tempervture
Figure 4.5 shows the θ-2θ XRD diffractogram for approximately 300 nm thick ITO(10)
films grown on silica glass substrates at varying heater temperatures using an oxygen
partial pressure of 0.15 Pa. The films were highly polycrystalline showing diffraction
peaks from a number of reflection planes. It seems that there may be a shift in preferred
orientation from the ITO (400) to the ITO (222) for films grown above 300◦C. A rise
in the calculated peak intensity ratio of ITO (222) to ITO (400) [I(GGG)/I(IEE)] confirms
the change in orientation. For heater temperatures above 300◦C it is clear that the
ITO (222) peaks become more pronounced and sharper, suggesting an increase in the
average grain size.
52
Cb[pnel .. Tbe Chdcog Ircde msmneg
(a) (b)
ITO(222)
ITO(400)
ITO(440)
ITO(622)
Figure 4.5: (a) X-ray diffraction patterns for 300 nm thick ITO(10) films grown on
silica glass substrates at different heater temperatures. The oxygen partial pressure
during deposition was 0.15 Pa. Scans of the silica substrates are superimposed (dotted
red lines). (b) Variation in the out-of-plane lattice parameter, v, with heater temper-
ature during growth.
Any observed peak shifting may be related to the level of Sn incorporation or uni-
form (macro)strain effects, each of which may result in chemical and lattice disorder.
Figure 4.5 (b) shows a decreasing out-of-plane lattice parameter, v, as the heater tem-
perature during growth increases. This suggests that raising the heater temperature
results in greater Sn substitution for In within the lattice as the ionic radius for SnI+
(0.81 A˚) is smaller than that of InH+ (0.94 A˚).
Calculation of the average vertical grain size was made by applying the Scherrer
equation (Eqn. 3.1) to the ITO (222) peak. The results of these calculations are shown
in table 4.3 and clearly shows an increase in the average vertical grain size as the heater
temperature increases. At higher temperatures atoms within the film have excess en-
ergy which they use to diffuse throughout the material to form larger grains. Also, at
higher temperatures it is common for grains to coalesce.
AFM studies on these films show an obvious change in film morphology as the
substrate temperature increased (Fig. 4.6). It appears that the crystal size begins
to increase forming elongated crystalline outgrowths (heater temperature of 600◦C)
instead of smaller grains seen in the films grown at lower temperatures. Analysis of
these images using the AFM Nanoscope software allows us to make an estimate of
the average root mean square surface roughness of these films. The root mean square
surface roughness increased steadily from a minimum of 2.30±0.20 nm at 300◦C, to a
maximum at the highest substrate temperature of 3.30±0.15 nm over a 1×1µmG area
at 600◦C.
The heater temperature during growth was found to affect the electrical proper-
53
Cb[pnel .. Tbe Chdcog Ircde msmneg
Figure 4.6: 1×1 µmG AFM micrographs showing the film evolution of ITO(10) thin
films grown on silica glass substrates at varying heater temperatures, (a) 600◦C, (b)
300◦C and (c) unheated. A second order flattening procedure has been applied to these
images to reduce scanner bow.
Table 4.3: Summary of the variation of calculated vertical grain size for the ITO(10)
films grown at different heater temperatures. All films were deposited using an oxygen
partial pressure of 0.15 Pa.
eeater temperature cilm thickness sertical grain size
(◦C) (nm) (nm)
300 290±11 27±2
400 280±11 38±2
500 295±10 41±2
600 290±11 49±2
ties of the ITO films. A selection of typical resistivity versus temperature curves for
ITO films grown at varying substrate temperatures are shown in Fig. 4.7. ITO films
grown on unheated substrates (labelled 25◦C) show a negative temperature coefficient
of resistance (TCR), i.e. semiconducting characteristics over the temperature range of
5 - 290 K. All films grown with a substrate temperature over 100◦C exhibit a positive
TCR, i.e. metallic characteristics.
Figure 4.8 shows the variation of /, nx, and µH , as a function of the substrate
temperature for the ITO films deposited using an oxygen partial pressure of 0.15 Pa.
The resistivity of the ITO films decreased from 13.83×10−I to 1.65×10−I Ωcm as the
heater temperature increased from 25 to 600◦C. In these samples, nx rose substan-
tially when the heater temperature was increased from 25 to 100◦C, after which the
change was very small. The increase in nx with heater temperature may be due to an
increase in diffusion of Sn atoms from interstitial locations and grain boundaries into
the In cation sites. The decrease in resistivity with increasing heater temperature can
be explained by the fact that the crystallite size increases with increasing substrate
temperature (table 4.3), thus reducing the grain boundary scattering. This is reflected
by the rise in µH .
The magnetoresistance response of an unheated ITO(10) thin film is shown in Fig.
54
Cb[pnel .. Tbe Chdcog Ircde msmneg
0 50 100 150 200 250
1.5
2.0
2.5
20
22
24
 Unheated
 300
o
C
 400
o
C
 500
o
C
 600
o
C
 
 
 
(
 
c
m
)
T (K)
Figure 4.7: Resistivity (/) versus temperature curves for ITO(10) films grown at
varying heater temperatures using an oxygen partial pressure of 0.15 Pa.
Figure 4.8: Variation of room temperature resistivity (/), carrier density (nx), and
Hall mobility (µH) as a function of heater temperature for the ITO(10) films deposited
using an oxygen partial pressure of 0.15 Pa.
55
Cb[pnel .. Tbe Chdcog Ircde msmneg
(a) (b)
ITO(222)
ITO(400)
ITO(440)
ITO(622)
Figure 4.9: (a) X-ray diffraction patterns for 300 nm thick ITO films grown on
silica glass at different oxygen partial pressures. The heater temperature was 300◦C.
(b) Variation in the out-of-plane lattice parameter, v, as a function of oxygen partial
pressure.
4.4 (a) (green). MR becomes positive rather than more negative and is typical of an
ordinary MR effect due to a Lorentz force. This is likely to be related to the changes
in film microstructure as we now have a material which is richer in defects and grain
boundaries. This may result in changes to the phonon periodicity which will naturally
influence the scattering mechanisms contributing to the MR.
Zffexts of oflygen pvrtivl pressure
Figure 4.9 shows the θ-2θ XRD diffractogram for approximately 300 nm thick ITO(10)
films grown on silica glass substrates at varying oxygen partial pressures (0 - 0.60 Pa)
using a constant heater temperature of 300◦C. With decreasing levels of oxygen, the
ITO films become increasingly polycrystalline and the ITO (222) peak becomes very
well pronounced. The out-of-plane lattice parameter is significantly reduced for films
grown in increasingly oxygen deficient conditions and suggests a rise in the level of Sn
substitution for In.
Table 4.4 summarises some of the properties of films grown at different oxygen par-
tial pressures. The vertical grain size of the films decreases as oxygen partial pressure
increases. Excess oxygen may migrate towards grain boundaries and interstitial sites,
inhibiting the movement of grain boundaries and hence grain growth.
The oxygen pressure was also found to affect the electrical properties of the ITO
films. Typical resistivity versus temperature curves for ITO(10) films grown at varying
oxygen partial pressures are plotted in Fig. 4.10. It is clear that the film resistivity
increases with the addition of more oxygen during the growth process. All films grown
56
Cb[pnel .. Tbe Chdcog Ircde msmneg
Table 4.4: Summary of the variation of calculated vertical grain size for the ITO(10)
films grown at different oxygen partial pressures. All films were deposited using a
heater temperature of 300◦C.
lxygen partial pressure cilm thickness sertical grain size
(Pa) (nm) (nm)
0.000 290±10 45±2
0.075 270±10 44±2
0.150 285±10 32±2
0.300 280±10 30±2
0.600 290±10 26±2
in oxygen partial pressure ≤ 0.15 Pa showed a positive TCR and were metallic, how-
ever, those grown at much higher levels of oxygen possessed a negative TCR and were
semiconducting.
Figure 4.11 illustrates the variation of /, nx and µH as a function of oxygen pres-
sure for the ITO(10) films grown using a heater temperature of 300◦C. As already
mentioned above, resistivity is highest for films grown with the highest levels of oxygen
partial pressure. This decrease in resistivity with decreasing oxygen pressure can be
explained by the number of oxygen vacancies in the ITO films. The oxygen vacancies
create free electrons in the films as one oxygen vacancy is capable of creating two ex-
tra electrons. The increase in the number of oxygen vacancies leads to an increase in
carrier density and a consequent increase in conductivity. Growing the films in higher
oxygen pressure leads to the filling of oxygen vacancies and hence a fall in the car-
rier concentration. Therefore, we can conclude that the carrier concentration in our
ITO(10) films is strongly influenced by the concentration of oxygen vacancies as well
as the level of Sn doping.
Investigations into the effects of oxygen partial pressure on pure InGOH and ITO(5)
were also carried out, and the results are summarised in table 4.5. Like the case of
the ITO(10), electrical transport properties of these films deteriorated as the partial
pressure of oxygen used during the growth process increased.
ICKCF Wrief yisxussion
From the XRD and AFM analysis, it has been shown that the grains become larger
as the substrate growth temperature increases or oxygen partial pressure levels fall.
Free electron carriers conduct faster in well-ordered crystalline structures, rather than
in those with more grain boundaries which would act as blocking barriers in carrier
transfer.
57
Cb[pnel .. Tbe Chdcog Ircde msmneg
0 50 100 150 200 250
1.0
1.2
1.4
1.6
1.8
500
1000
1500
2000
2500
 
 
 
 
T (K)
 
(
 
c
m
)
 0 Pa       0.075 Pa
 0.15 Pa  0.30 Pa
 0.60 Pa
Figure 4.10: Resistivity (/) versus temperature curves for ITO(10) films grown at
varying oxygen partial pressures using a heater temperature of 300◦C.
Figure 4.11: Variation of room temperature resistivity (/), carrier density (nx), and
Hall mobility (µH) as a function of oxygen deposition partial pressure for the films
grown at 300◦C.
58
Cb[pnel .. Tbe Chdcog Ircde msmneg
If grain boundary scattering is assumed to be the predominant scattering mecha-
nism we find that / can be expressed by combining Eqns. 4.2 and 4.4;
/ =
1
cµHq
=
1
(ne)eD(2.m∗ki )−FRGexp
(− zϕb
ki
)
O
(4.8)
Therefore, / is inversely proportional to the grain size, D. We have seen that the average
vertical grain size increases with increasing substrate temperature and subsequently /
decreases, however, it does not follow an inverse relationship. As for the films grown
with varying oxygen partial pressure, increasing oxygen levels led to large increases
in /. It is likely that in the regions with excess oxygen, the oxygen may do any of
the following: (i) incorporate into the bixbyite lattice; (ii) become segragted to grain
boundaries; or (iii) become complexed with Sn ions, forming Sn+O−. Excess oxygen
can therefore lead to two types of electron scattering centres : those which form SnO
complexes and distort the lattice, creating neutral scattering centres, and those segre-
gated to become charge carrier traps, therefore acting as negatively charged scattering
centres.
ICL hummvry vny Xonxlusions
Transparent InGOH and Sn-doped InGOH films with n-type conductivity were success-
fully made using DC magnetron sputtering. The results show that the progressive in-
crease in the level of Sn doping leads to increasing carrier concentration and improved
electrical properties. Variation of growth parameters such as the heater temperature
and oxygen partial pressure allowed the tuning of the electrical properties: in general,
higher heater temperatures and lower oxygen partial pressures lead to more crystalline
films with greater levels of electrically active substitutional Sn, and as a result, much
improved electrical characteristics. Experiments where the oxygen partial pressure
was varied highlighted the role that oxygen vacancies play in determining the electri-
cal properties of the films. Excess oxygen is seen to be detrimental to the conductive
nature of the films, and is believed to be the cause of a rise in the density of electron
scattering centres.
With this level of understanding for the InGOH and ITO systems, the goal is to now
use these materials for the development of a dilute magnetic oxide material by doping
with a variety of transition metal ions. The next 3 chapters will tell this story.
59
Cb[pnel .. Tbe Chdcog Ircde msmneg
T
ab
le
4.
5:
S
u
m
m
ar
y
of
th
e
ro
om
te
m
p
er
at
u
re
tr
an
sp
or
t
p
ro
p
er
ti
es
(r
es
is
ti
v
it
y,
/
,
ca
rr
ie
r
co
n
ce
n
tr
at
io
n
,
n
x
,
an
d
el
ec
tr
on
m
ob
il
it
y,
µ
H
)
of
In
G
O
H
an
d
IT
O
(5
)
th
in
fi
lm
s
gr
ow
n
u
n
d
er
va
ri
ou
s
ox
y
ge
n
p
ar
ti
al
p
re
ss
u
re
s
u
si
n
g
a
h
ea
te
r
te
m
p
er
at
u
re
of
30
0
◦ C
.
M
at
er
ia
l
l
x
y
ge
n
p
ar
ti
al
p
re
ss
u
re

n
x

H
(P
a)
(Ω
cm
)
(c
m
−H
)
(c
m
G
V
−F
s−
F
)
In
G
O
H
0.
00
0
2.
71
±0
.1
5×
10
−H
6.
87
±0
.1
5×
10
F
N
33
.5
0±
0.
12
In
G
O
H
0.
07
5
5.
55
±0
.1
1×
10
−H
4.
54
±0
.1
7×
10
F
N
24
.7
9±
0.
16
In
G
O
H
0.
15
0
3.
00
±0
.3
2×
10
−F
7.
86
±0
.2
6×
10
F
M
5.
65
±0
.2
0
In
G
O
H
0.
30
0
2.
41
±0
.0
8×
10
F
5.
18
±0
.4
7×
10
F
L
4.
99
±0
.2
6
In
G
O
H
0.
60
0
5.
85
±0
.0
7×
10
F
1.
43
±0
.5
4×
10
F
K
3.
03
±0
.3
1
IT
O
(5
)
0.
00
0
2.
53
±0
.0
9×
10
−I
7.
00
±0
.0
8×
10
G
E
35
.2
3±
0.
15
IT
O
(5
)
0.
15
0
1.
45
±0
.1
1×
10
−H
1.
83
±0
.0
8×
10
G
E
23
.5
8±
0.
14
IT
O
(5
)
0.
30
0
2.
54
±0
.1
7×
10
−G
1.
39
±0
.1
2×
10
F
N
17
.5
9±
0.
24
60
Wiwliogrvphy
[1] K. L. Chopra, S. Major, D. K. Pandya, Thin Solid Films 102, 1, (1983).
[2] G. Rupprecht, Z. Phys. 139, 504, (1954).
[3] F. S. Galasso, Structure and properties of inorganic solids. Oxford New,York,
Toronto, Sydney, Branschweig, Pergamon Press. pp. 90-102, (1970).
[4] N. Nadaud, N. Lequeux, M. Nanot, J. Jov, T. Roisnel, J. Solid State Chem. 135,
140-148, (1998).
[5] P.Nath, R.Bunshah, Thin solid films 69, 63, (1980).
[6] I. Hamberg, C. Granqvist, Journal of Applied Physics, 60, R123, (1986).
[7] L. Gupta, A. Mansingh, P. K. Srivastava, Thin Solid Films, 176, 33-44, (1989).
[8] R. Bel Hadj Tahar, T. Ban, Y. Ohya, Y. Takahashi, J. Appl. Phys. 83, 5, (1998).
[9] G. Frank, H. Kostlin, Appl. Phys. A27, 197-206, (1982).
[10] H. Kostlin, R. Jost, W. Lems, Phys. Stat. Sol. (a)29, 87-93, (1975).
[11] H. Hartnagel, A. Dawar, A. Jain, C. Jagadish, Institute of Physics Publishing,
(1995).
[12] J. Seto, Journal of Applied Physics, 46, 5247, (1975).
[13] V. Johnson, K. Lark Horovitz, Physical Review, 71, 374, (1947).
[14] K. Ramaiah, V. Raja, A. Bhatnagar, R. Tomlinson, R. Pilkington, A. Hill, S.
Chang, Y. Su, F. Juang, Semiconductor Science and Technology, 15, 676, (2000).
[15] S-S. Kim, S-Y. Choi, C-G. Park, H-W. Jin, Thin Solid Films, 347, 155-160, (1997).
[16] Y. Hoshi, Y. Kiyomura, Thin Solid Films, 411, 36-41(2002).
[17] C. Guillen, J. Herrero, Thin Solid Films, 510, 260-264, (2006).
[18] N. Mori, S. Ooki, N. Masubuchi, A. Tanaka, M. Kogoma, T. Ito, Thin Solid Films,
411, 6, (2002).
61
Bcbfcial[pbs
[19] G. M. Zhen, S. Hi-Gang, R. Job, L. Pa-Shen, W. R. Fahrner, Chinese Physical
Letters, Vol. 22, No. 5, 1228, (2005).
[20] L. J. Meng, M. P. D. Santos, Thin Solid Films, 322, 56, (1998).
[21] S. Takayama, A. Tanaka, T. Sugawara, T. Himuro, Jpn. J. App. Phys. 41, 619-621,
(2002).
[22] R. P. Howson, I. Safi, G. W. Hall, N. Danson, NIMB, 121, 96-101, (1997).
[23] H-C. Lee, J-Y. Soo, Y-W. Choi, D-W. Lee, Vacuum, 72, 269-276, (2004).
[24] N. Kikuchi, E. Kusano, H. Nanto, A. Kinbara, H. Hosono, Vacuum 59, 492-499,
(2000).
[25] C. H. L. Weijtens, J. Electrochem. Soc. 138, 3432, (1991).
[26] G. Franz, B. Lange, S. Sotier, J. Vac. Sci. Tehnol. A, 19(5), (2001).
[27] S. Ray, R. Banerjee, N.Basu, A. K. Batabayal, A. K. Barua, J. Appl. Phys. 54,
3497, (1983).
[28] H. Nanto, T. Minami, S. Orito, S. Takata, J. Appl. Phys. 63, 2711, (1988).
[29] L. J. Meng, M. P. D. Santos, Thin Solid Films, 289, 65, (1996).
[30] H. Kim, J. S. Horwitz, W. H. Kim, Z Kafafi, D. B.Chrisey, Materials Research
Society Symposium Proceedings, 780, 21-32, (2003).
[31] C. Coutal, A. Azema, J.C. Rostau, Thin Solid Films, 288, 248-253, (1996).
[32] F. O. Adurodija, H. Izumi, T. Ishihara, H. Yoshioka, H. Matsui, M. Motoyama,
Jpn. J. App. Phys. 38, 2710-2716, (1999).
[33] L. A. Ryabova, V. S. Salun, I. A. Serbinov, Thin Solid Films 92, 327, (1982).
[34] K. Nishio, T. Sei, T. Tsushiya, J. Mater. Sci. 31, 1761, (1996).
[35] Y. Takahashi, H. Hayashi, Y. Ohya, Mater. Res. Soc. Symp. Proc. 271, 401, (1992).
[36] P. Nath, R. F. Bunshah, B. M. Basol, O. M. Staffsud, Thin Solid Films 72, 463,
(1980).
[37] I. A. Rauf, J. Appl. Phys. 79, 4057, (1996).
[38] H-C. Lee, Applied Surface Science, 252, 2647-2656, (2006).
[39] M. Kamei, Y. Shigesato, S. Takaki, Thin Solid Films 259, 38, (1995).
[40] J. C. C. Fan, F. J. Bachner, J. Electrochem. Soc. 122, 1719, (1975).
62
Xhvpter 5
ihe fuest for v Yilute bvgnetix
dfliye
ihis xhvptzr wzgins wy yisxussing przvious fiork in thz fizly of yilutz mvgnztix ofliyzsC
Common prowlzms vssoxivtzy fiith thz xhvrvxtzrisvtion of thzsz mvtzrivls fihixh xonB
tinuz to puzzlz thz fiiyzr sxizntifix xommunity vrz vlso introyuxzyC ihz xhvptzr xloszs
fiith mvgnztix vny mvgnztotrvnsport xhvrvxtzrisvtion mzvsurzmznts mvyz on CoByopzy
InGdH vny Iid filmsC
63
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
5CF Introyuxtion
Chapter 1 has already expressed the principal goal in the field of DMS and DMO ma-
terials for implementation within the emerging field of spintronic device applications:
the capability to exhibit substantial carrier spin polarisation at and above room tem-
perature.
The large electronegativity of oxygen in DMOs is expected to produce strong sp-y
exchange interactions between band carriers and localised spins [1]. As a result, the
wide bandgap oxides are considered to be well suited to the cause as they are capable
of providing a multi-functional foundation for next generation spintronic devices. The
combination of high optical transparency, electrical conductivity and potential ferro-
magnetic behaviour opens up many new attractive avenues for the future.
This chapter introduces the experimental techniques used to investigate the mag-
netic behaviour of DMS and DMO systems. An extensive review on DMO materials,
in particular InGOH and ITO, is given, and a discussion on the more popular systems
studied such as ZnO and TiOG is provided. A dilute magnetic oxide system based on
InGOH and ITO is then developed and described.
5CG Zflvminvtion of ferromvgnetism
5CGCF bvgnetometry
Early experiments relied heavily on SQUID magnetometry to classify the origins of
magnetism in DMOs. However, it is difficult to distinguish “intrinsic” ferromagnetism
from that which is “extrinsic”, such as ferromagnetic precipitations or secondary phases
because all the magnetisation signals are measured together, i.e. it is a volume sensitive
measurement. Therefore, magnetisation alone is not enough to verify the formation of
a “true” DMO.
5CGCG bvgnetoBoptixvl effexts
As a consequence of a coupling interaction between s- and p-carriers in the host ox-
ide/semiconductor with the y -electrons of the TM ions, photoexcited charge carriers
in the host oxide show a sizeable magneto-optical effect. MCD is the difference in the
absorption coefficients for two circularly polarised waves of light and is proportional to
the energy derivative of the absorption coefficient. Therefore, MCD is largest around
the onset of absorption. This gives a clear indication of the extent to which the states
involved in the transition at that particular energy are influenced by the magnetism.
More on this matter will follow in chapter 7.
64
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
5CGCH ihe Vnomvlous ]vll Zffext
For an introduction to the Hall effect, the reader is referred to reference [2] of this
chapter, which is the original report of Edwin H. Hall’s momentous discovery. In addi-
tion to the ordinary Hall effect (OHE) that is present in all semiconductors and metals,
there is an additional voltage present in magnetic materials which is proportional to the
magnetisation, called the “Anomalous Hall effect” (AHE). It was found that the Hall
resistivity /xy in an applied perpendicular magnetic field µE] z is qualitatively differ-
ent in ferromagnetic and non-magnetic conductors. In the latter, /xy increases linearly
with µE] z as expected from the Lorentz force; in ferromagnets, however, /xy initially
increases steeply in weak µE] z, but saturates at a large value that is nearly indepen-
dent of µE] z. The point of saturation is roughly proportional to the magnetisation
b z. Therefore, /xy in ferromagnetic materials can be expressed as
/xy = gEµEHz +gsMz, (5.1)
where the first term denotes the OHE and the second the contribution from the spon-
taneous magnetisation (AHE). This additional term to the electron group motion has
been attributed to the asymmetric carrier scattering by magnetic impurities in the
presence of a spin-orbit interaction [3] [Fig. 5.1 (b)]. Imagine that each carrier has
an associated orbital angular momentum about the magnetic scattering centre, the
sign of which depends on whether the carrier passes on one side or the other. The
spin-orbit coupling between the orbital angular momentum of the carrier and spin an-
gular momentum of the magnetic impurity means that the electron carriers will find it
energetically favourable to pass on one side rather than the other, resulting in the asym-
metry in scattering. Because of this term, it was believed that ferromagnetic ordering
will generate the AHE, and so it therefore became sought after as evidence for the
direct manifestation of intrinsic ferromagnetism in DMO materials. An encouraging
sign for detecting a genuine intrinsic ferromagnetic signal using AHE measurements
is the quantitative scaling with the magnetisation loops. This has been reported in
Co:TiOG [3–5] [see Fig. 5.1 (c)] and Co:ZnO [6] systems.
Observing the AHE alone is not adequate to conclude a material is an intrinsic
DMO. The AHE response has also been detected in samples where the presence of
metallic Co clusters has been confirmed [7, 8]. They conclude that the weak AHE
originates from the spin-interaction in close proximity to the clustered Co, in complete
contrast to Toyosaki’s proposal described and illustrated in Fig. 5.11 (a) and (b). Sur-
prisingly, the AHE effect appears in the non-magnetic material ZnEONM5CuEOEF5O [9].
Figure 5.2 (a) - (c) summarises the effects described in sections 5.2.1 to 5.2.3 in try-
ing to determine the nature of magnetism in the DMO systems along with the general
phenomenological scheme behind each characterisation technique [10].
With the difficulty in gaining access to MCD-centred techniques such as optical
65
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
(a) (b) (c)
Figure 5.1: Sources of extrinsic and intrinsic ferromagnetism in DMO materials pro-
posed in reference [3]: (a) ferromagnetic secondary phase (red arrows) segregates caus-
ing ferromagnetic magnetisation, regardless of the non-ferromagnetic ordering of the
incorporated magnetic dopants (green arrows); (b) an exchange interaction between
magnetic dopants and charge carriers (blue arrows) results in ferromagnetic order-
ing. Spin-orbit interaction between the two leads to ferromagnetic spin polarisation of
charge carriers, generating an AHE (light blue curves); (c) quantitative scaling of the
magnetic field dependence of MCD (red curve), resistivity (blue curve) and magneti-
sation (green curve) for Co:TiOG samples at room temperature with varying carrier
concentrations. After references [3, 4].
66
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
Figure 5.2: Schematic diagrams of intrinsic and extrinsic sources of ferromagnetism
for (a) magnetisation, (b) magneto-optical effects, and (c) magneto-transport mea-
surements. The left hand side and central diagrams represent intrinsic and extrinsic
ferromagnetism respectively. The diagrams on the right show the phenomenological
scheme for each measurement. After reference [10].
67
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
and x-ray MCD, many researchers must make do with what they have. In this volume
of work we have managed to use both forms of MCD characterisation (see chapters 6
and 7), as well the more widely accessible techniques such as x-ray diffraction, SQUID
magnetometry, transport measurements, ztx.
5CH erevious fiork
5CHCF [erromvgnetism in InOdP vny Iid thin films
Compared to ZnO and TiOG, InGOH and ITO are relatively unexplored materials for
dilute magnetic oxide materials. However, a host of authors have published reports
on the observation of room temperature ferromagnetism in a variety of TM-doped and
undoped InGOH-based materials; table 5.1 summarises a selection of these results.
One of the most important practical goals of this research area is to produce a
magnetic oxide system with tunvwlz ferromagnetic behaviour, capable of exhibiting
independently controlled carrier and magnetic doping along with a high Curie temper-
ature. Early work by Philip and co-workers identified Cr-doped InGOH as a material
system where electrical and magnetic behaviour, from ferromagnetic metal-like to ferro-
magnetic semiconducting to paramagnetic insulator, can be controllably tuned through
the defect concentration [11]. They proposed that both the electronic and magnetic
properties can be enhanced by making the films more oxygen deficient. By altering the
oxygen partial pressure during the growth process, the investigators were able to tune
the defect concentration and hence the carrier concentration. The fall in nx results in
both electrical and magnetic phase transitions. The authors estimate a critical con-
centration of oxygen vacancies of 6.0×10GG cm−H and equivalent carrier concentration
of 2.0×10FN cm−H to observe ferromagnetism.
Kim zt vlC investigated the effects of carrier concentration on the magnetic and
magnetotransport behaviour of Cr-doped ITO(10) films. They found that the anoma-
lous Hall effect only appeared in the presence of a very high carrier concentration
(1.32×10GF cm−H) and again this pointed towards a carrier mediated ferromagnetic re-
sponse [12].
Using density-functional calculations, Raebiger zt vl. showed that long range fer-
romagnetic Cr-Cr coupling can exist in Cr-doped InGOH films, via the inclusion of
extrinsic dopants in the form of Sn, which will provide free electrons [13] (see Fig.
5.3). Essentially, the authors present an idea where the ferromagnetism can be turned
“on” and “off” simply by tuning the carrier concentration via external Sn doping.
In the absence of free carriers, Cr-doped InGOH is paramagnetic, whereas in the de-
generate doping regime (using Sn), Cr-doped InGOH is ferromagnetic. The onset of
ferromagnetism occurs as the free-electron concentration reaches half of the Cr dopant
concentration.
68
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
Figure 5.3: Ferromagnetic stabilisation energy, ∆Fb , at room temperature as a func-
tion of Cr-Cr distance, r. The black line denotes ∆Fb in the absence of free carriers
(nz=0), while green, red and blue lines correspond to electron concentrations of nz=
10GE, 10GF and 2×10GFcm−H respectively. The red arrow indicates the shift in ∆Fb after
exceeding the limit for nz. After reference [13].
Jiang zt vl. showed experimental evidence consistent with the predictions made
by Raebiger, whereby the ferromagnetism of Mn-doped InGOH films show reversible
behaviour, which can be switched between on and off by controlling the carrier den-
sity [14]. In addition to this, increasing the carrier concentration through additional
Sn doping led to marked increases in ferromagnetic response in Cr-doped InGOH films.
Panguluri zt vlC recently opened up another dimension to the discussions sur-
rounding the origin of magnetism in these oxide materials, particularly in those which
are pure and undoped [15]. As-deposited samples of pure, undoped InGOH were non-
magnetic, but those doped with Cr were weakly ferromagnetic at room temperature.
After a vacuum annealing treatment, both sets of samples are ferromagnetic, showing
the need for a high concentration of oxygen vacancies to mediate magnetism. To in-
vestigate the coupling interaction between the charge carriers and the ferromagnetic
moment in InGOH and Cr-doped InGOH films, point contact Andreev reflection was used.
This revealed a spin polarisation of approximately 50% at 1.3 K in both pure InGOH
and the samples doped with Cr. In addition to this, a model was described which pre-
dicts that room temperature ferromagnetism should emerge in undoped InGOH above a
critical carrier concentration of 4.0×10FN cm−H. The model also predicts a carrier con-
centration dependent iX , as was suggested by Philip zt vl [11]. These findings raised
doubts over the necessity to have transition metal dopants to produce carrier-mediated
ferromagnetism.
Contrary to the findings described above where increasing carrier concentration gen-
erally leads to improved magnetic properties, Wang zt vl. have shown that in Mn-doped
InGOH, increasing the Sn level leads to a degradation of the magnetic response [16]. In
69
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
the InFONE−xMnEOFSnxOH samples, increasing fl from 0 to 0.06 resulted in a suppression
of the coercive field. Furthermore, the magnetisation loops resembled something more
like a paramagnetic material. It was suggested that the introduction of Sn into the
system caused a larger amount of magnetic ions to couple antiferromagnetically. This
may be connected to the fact that the Sn and TM dopant ions are both competing for
the In lattice sites.
Codoping TM-doped InGOH with Cu has also been investigated [17, 18]. In par-
ticular, the magnetic properties of (Cu/Fe)-codoped InGOH (IFCO), are not directly
related to the Cu doping. Using detailed analysis of anomalous Hall effect measure-
ments, Yu zt vl. ascribe the ferromagnetic properties of their IFCO samples to an
exchange coupling between magnetic ions which is mediated by mobile electrons rather
than localised ones, i.e. much like the behaviour expected for a typical ferromagnet
where the anomalous Hall coefficient is proportional to the carrier density. In contrast,
Brize zt vl. have reported worsened magnetic properties of (Cu,Mn)-doped InGOH thin
films, suggesting that doping with Cu promotes the formation of undesirable precipi-
tates which degrade the magnetic properties [19].
Stankiewicz zt vl. and more recently Subias zt vl., have performed extensive stud-
ies into the Co-doped InGOH and ITO(10) systems by studying the local structural
environment of Co through x-ray absorption studies at the Co K -edge in conjunc-
tion with x-ray photoelectron spectroscopy [21, 22] (Fig. 5.4). It was concluded that
different contributions of oxidised (CoG+) and metallic (CoE) cobalt to the observed
ferromagnetism are found depending on the host oxide and the Co doping content.
The threshold was estimated to be less than 7.0 at.% Co, after which the Co doping is
no longer within the dilute limit (Co substitutes for In lattice sites) and metal Co clus-
ters form. Magnetisation characterisation of these samples showed a lack of saturation
at low temperatures even up to magnetic fields of 5 T, clearly indicating that a large
portion of the CoG+ ions are giving a paramagnetic response, and not contributing to
the ferromagnetism directly.
There have been a number of surprising results where room temperature ferromag-
netism has been observed in pure, undoped oxide materials (see section 5.2.3): InGOH
is one of these. We have described above that Panguluri zt vl. have even estimated the
extent of spin polarisation in pure InGOH. Hong zt vl. reported that room temperature
ferromagnetism is present in a host of undoped oxides [31]. The controversy surround-
ing these kinds of reports will be discussed further later, where the sources of error in
measuring nanomagnetism will be also be raised.
Some of these results have caused former supporters and pioneers of this area to
turn on their previous ideas and look at the situation from another angle. J. M. D. Coey
in particular is certainly one who has recently begun to change his attitude towards
these materials (see section 5.4). Recently, Venkatesan zt vl. reported measurements
on Fe- and Mn-doped ITO(10) films where it is claimed that the ferromagnetism in the
Fe-doped samples is due to secondary ferromagnetic FeHOI grains, but are unable to
70
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
(a) (b)
Figure 5.4: (a) X-ray photoemission spectra at the Co 2p core-level for varying levels
of Co doping in ITO(10) samples. Increasing the level of Co leads to the formation
of metallic Co clusters, implied by the extra features appearing in the spectra, and
indicated by the arrows. (b) X-ray absorption near edge structure analysis at the Co
K -edge for ITO(10) and InGOH films for varying levels of Co content. Again, low doping
levels reveal spectra which resemble divalent CoO, whereas, high levels of doping lead
to features emerging in the spectra which would suggest Co metal segregation. After
reference [22].
71
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
identify the source of the magnetism in those doped with Mn [15]. The authors con-
clude by claiming that these materials cannot be regarded as a DMS in the accepted
sense.
It is clear with the varying results described above that there is still much to un-
derstand regarding the origins of magnetism in these materials.
72
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
T
ab
le
5.
1:
S
u
m
m
ar
y
of
a
se
le
ct
io
n
of
T
M
-d
op
ed
In
G
O
H
an
d
IT
O
th
in
fi
lm
s.
D
C
M
S
=
D
C
m
ag
n
et
ro
n
sp
u
tt
er
in
g,
R
F
S
=
ra
d
io
fr
eq
u
en
cy
sp
u
tt
er
in
g,
P
L
D
=
p
u
ls
ed
la
se
r
d
ep
os
it
io
n
,
R
E
=
re
ac
ti
ve
th
er
m
al
ev
ap
or
at
io
n
.
p
y
st
em
q
M
co
n
te
n
t
p
u
b
st
ra
te
d
ro
w
th
te
ch
n
iq
u
e
T
X
M
a
g
n
et
is
a
ti
o
n
fe
at
u
re
s
o
ef
er
en
ce
(a
t.
%
)
(K
)
(3
00
K
)
C
r:
In
G
O
H
2.
0
(0
00
1)
sa
p
p
h
ir
e
R
F
S
..
.
0.
50
em
u
/c
m
−H
[1
5]
C
r:
IT
O
5.
0
S
i/
S
iO
G
P
L
D
..
.
0.
70
em
u
/c
m
−H
[1
2]
C
r:
In
G
O
H
2.
0
S
i/
S
iO
G
R
E
..
.
1.
5
µ
B
/C
r
[1
1]
C
r:
In
G
O
H
3.
0
x
-s
ap
p
h
ir
e
P
L
D
..
.
0.
95
µ
B
/C
r
[1
4]
C
r:
IT
O
3.
0
x
-s
ap
p
h
ir
e
P
L
D
..
.
2.
10
µ
B
/C
r
[1
4]
C
r:
IT
O
2.
.0
(0
00
1)
sa
p
p
h
ir
e
S
ol
-g
el
..
.
0.
22
µ
B
/C
r
[2
3]
N
i:
In
G
O
H
5.
3
M
gO
P
L
D
S
40
0
0.
40
-0
.7
0
µ
B
/N
i
[2
4]
N
i:
In
G
O
H
5.
3
sa
p
p
h
ir
e
P
L
D
S
40
0
0.
70
µ
B
/N
i
[2
4]
N
i:
In
G
O
H
6.
0
..
.
S
ol
-g
el
..
.
0.
06
µ
B
/N
i
[2
5]
N
i:
IT
O
6.
0
..
.
S
ol
-g
el
..
.
0.
03
µ
B
/N
i
[2
5]
M
n
:I
T
O
5.
0
(0
00
1)
sa
p
p
h
ir
e
R
E
≫
30
0
0.
80
µ
B
/M
n
[2
0]
M
n
:I
T
O
5.
0
S
i/
S
iO
G
R
E
≫
30
0
0.
08
-0
.1
2
µ
B
/M
n
[2
0]
M
n
:I
T
O
5.
0
S
il
ic
a
gl
as
s
R
F
S
S
40
0
8.
0
em
u
/c
m
−H
[2
6]
M
n
:I
T
O
3.
2
S
il
ic
a
gl
as
s
S
ol
-g
el
≥6
00
0.
39
µ
B
/M
n
[2
7]
M
n
:I
n
G
O
H
5.
0
x
-s
ap
p
h
ir
e
P
L
D
..
.
0.
10
µ
B
/M
n
[1
4]
M
n
:I
T
O
5.
0
x
-s
ap
p
h
ir
e
P
L
D
..
.
0.
18
µ
B
/M
n
[1
4]
C
o:
IT
O
8.
0
S
il
ic
a
gl
as
s
D
C
M
S
≤4
0
0.
26
µ
B
/C
o
[2
9]
C
o:
IT
O
30
.0
F
u
se
d
q
u
ar
tz
D
C
M
S
..
.
1.
10
µ
B
/C
o→
m
et
al
cl
u
st
er
s
[3
0]
C
o:
In
G
O
H
3.
0
F
u
se
d
q
u
ar
tz
D
C
M
S
..
.
0.
06
µ
B
/C
o
[2
2]
C
o:
IT
O
4.
0
F
u
se
d
q
u
ar
tz
D
C
M
S
..
.
0.
02
µ
B
/C
o
[2
2]
C
o:
IT
O
12
.0
F
u
se
d
q
u
ar
tz
D
C
M
S
..
.
1.
05
µ
B
/C
o→
m
et
al
cl
u
st
er
s
[2
2]
73
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
5CHCG dther ferromvgnetix ibByopey ofliye systems
It is not possible, nor entirely necessary for a long discussion on the other ferromagnetic
DMOs which have been reported. Due to the rapidly growing number of reports in this
field, some references have been overlooked. Instead, the reader is referred to other,
more comprehensive reviews, and the references within, on the DMO family [10,33–36].
TM:ZnO has been adopted as the prototypical model DMO system (analogous to
the GaMnAs DMS system) and has received the majority of attention from researchers
globally. The problem has been approached both experimentally and theoretically.
Pan zt vl. provided a detailed overview of the recent progress in the TM-doped ZnO
system, highlighting the importance of influential factors such as growth techniques,
substrate, strain and post-annealing effects and the role that codoping can play [36].
Here, we focus on those systems which have been doped using Co (Co:ZnO in partic-
ular), as this is the dopant which has been studied in detail in this thesis.
A myriad of studies are available on the Co-doped ZnO system where the investiga-
tors strive to determine the exact role of the TM dopants. Is it that they are uniformly
distributed and substituting for host matrix ions, or perhaps, are they forming metal
or secondary oxide phase clusters or lying in interstitial positions? In most cases where
extrinsic ferromagnetism has been claimed as the source of magnetism in Co:ZnO, the
investigators report clustering of Co [38]. However, antiferromagnetic CoO has also
been reported to result in weak ferromagnetism in Al-doped Co:ZnO [39]. Here, Dietl
zt vl. argue that the magnetic behaviour in their films are a result of uncompensated
spins at CoO nanocrystal surfaces formed via spinodal decomposition, and also from
Co precipitates exchange biased by the surrounding Co-rich Co:ZnO lattice. Chang zt
vl. attribute the lack of ferromagnetism in their Co:ZnO samples to the lack of free
electron carriers and direct Co-Co interactions [40]. Through first-principle calcula-
tions, they conclude that interactions between substitutional Co atoms (at Zn lattice
sites) only induces antiferromagnetic coupling, similar to that described above by Di-
etl and also later suggested through XMCD studies [41]. The authors suggest that in
addition to free charge carriers, an exchange interaction between substitutional and
interstitial Co through cluster formation is required for high i x ferromagnetism. So,
from these two reports, it is implied that magnetism in Co:ZnO can only come about
via an extrinsic source.
Once again, from first-principle calculations of the oxygen surfaces in Co:ZnO,
Sanchez zt vl. show that the spin polarisation induced by the substitutional Co atoms
is enhanced depending on the proximity to the surface [42]. It is also predicted that
the oxygen atoms are spin polarised only at the surface, and not the bulk, and that this
surface stability is reinforced by Co. This polarised oxygen surface may also explain
the spontaneous magnetisation reported in ZnO (and indeed others) structures in the
absence of TM dopants [43].
To exercise the feasibility that uncompensated spins at the surface of CoO nanoclus-
ters embedded in ZnO are a source of magnetism, Hanafin zt vl. used a combination of
74
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
density-functional theory and Monte Carlo simulations to show that these finite sized
particles (of various shapes) cannot be considered as the origin of ferromagnetism in
Co:ZnO [44].
However, with each of the scenarios described above, there are many simplifica-
tions, assumptions and parameterisations of the possible magnetic interactions within
the Co:ZnO system. These models will neglect important parameters such as grain
boundaries, interfaces with the substrate (Co clusters have been reported at the in-
terface between film and substrate in Co:ZnO [8]), as well as intrinsic defects, each of
which can alter the local magnetic order.
Experimentally, we must consider factors such as the growth conditions, sample
morphology and microstructure in addition to the spatial distribution of the Co atoms
when determining the magnetic properties. In particular, there is now an emerging view
that samples of Co:ZnO with high structural quality (and uniform Co distribution as
some report above) will not result in long-range ferromagnetism at high temperature,
and that the magnetism may be related to structural [45] or point-defects (e.g. oxygen
vacancies, ztx,.) [46].
As with the Co:ZnO scenario, the Co:TiOG system is intensely debated. After initial
reports of intrinsic ferromagnetism in this system [3,8,47,48], the general consensus has
now become that ferromagnetism in Co:TiOG is a result of metal Co clustering [49–52].
5CHCH d M ferromvgnetism
Perhaps one of the most interesting, if not the most controversial, wrinkles in the field
of high temperature DMOs, is that oxygen vacancy defects in closed-shell yE transi-
tion metal oxides can result in room temperature ferromagnetism. This phenomenon
has now been termed “yE ferromagnetism”. The idea was born after experiments
by Venkatesan zt vl. who had prepared pure, undoped HfOG by PLD and observed
anisotropic ferromagnetic behaviour up to 500 K [53]. In bulk form, these oxides are
diamagnetic, however, in the thin film form, they produce a modest magnetic response,
likely to be due to the creation of defects or oxygen vacancies on transformation to the
thin film form. A theory was put forward which postulates the hybridisation of these
oxygen vacancy defect states with unoccupied Hf-derived 5y states in the conduction
band, forming an impurity band. Mixing and transference of electrons between the
impurity band and the unoccupied Hf 5y states was thought to be enough to polarise
the impurity band and lead to ferromagnetism. Later predictions by the same group
proposed that a finite magnetisation per unit area of film as a function of dopant
concentration for ZnO and SnOG host oxides, doped with a variety of TM dopants,
could be extrapolated to zero [54]. Therefore, room temperature ferromagnetism can
occur in the absence of TM dopants in other closed-shell transition metal oxides. This
highlighted the apparent importance of defects as a critical driver of ferromagnetism in
these systems. Following this work, room temperature ferromagnetism was observed
75
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
in other undoped oxides such as TiOG and InGOH [55].
A report which directly challenges the findings of Hong zt vlC, is that of Berardan
zt vl., where they show that pure InGOH is diamagnetic, as would be expected for a y
FE
element [56]. In fact, Berardan goes a step further to say that their findings imply that
all magnetic transitions which occur in the TM-doped InGOH thin films are due to the
presence of small amounts of secondary phases.
These results make us reconsider the role that 3y element doping has in introduc-
ing ferromagnetism in the oxide hosts. Does the doping genuinely introduce ferromag-
netism? Or does it simply enhance the magnetism that already exists in the oxide
hosts when in thin film form? (see section 5.4).
Rather embarrassing, yet highly relevant, results which have come out of other re-
search groups attempting to repeat and build on the results of HfOG, have discovered
that their films only become magnetic once handled with stainless steel tweezers [57].
These troubling aspects of the data compounded with the distinct lack of reproducibil-
ity has cast much doubt over their validity.
5CI erowlems vssoxivtey fiith Yilute bvgnetism
Given the fact that there is still no unified theory or model to explain the phenomena
observed when non-magnetic oxides become ferromagnetic in thin film form (but not
in bulk) at room temperature after being doped (or not at all) with only a few percent
of TM dopants, the field is naturally met with great sceptism.
We must not overlook some of the technical issues accompanied with measuring
dilute magnetism. For one, typical magnetic signal levels are generally of the order
of 10−K emu. Therefore, for magnetic characterisation, SQUID magnetometry and
not VSM should be used, as its sensitivity levels are generally 100 times better. The
presence of small measuring artifacts, which are not so important when measuring
traditional bulk ferromagnets, can completely modify the results, leading to surprising
yet incorrect results. Thin film heterogeneity cannot be ignored. There is no reason to
believe that the profile of the dopant element is uniform over the entire film without
good elemental profile analysis such as Rutherford backscattering. Already a case
has been referenced where clustered material was found at the interface between film
and substrate [8]. It is difficult to rule out the possibility of the 3y dopant not being
uniformly diluted in the oxide lattice, instead tending to form secondary phases such as
metallic clusters. Coey reminded us in his recent article that the “Vwsznxz of zviyznxz
is not zviyznxz of vwsznxz” [58]. It was discussed in section 5.2 that phenomena such as
the AHE and MCD can take place in the presence of non-uniform dopant distributions
where the dopants may be fully or partially clustered. One could then take a view
arguing that even if some secondary phase clusters are present why worry if there is
still a chance of the properties being application-worthy? The main reason is simply the
76
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
lack of reproducibility in extrinsic states that lead to these phenomena. It is far more
sensible to pursue achieving a statistically homogenous dopant dispersion whereby the
properties can be influenced by manipulating the carrier density, ztx.
However, secondary phases cannot be a general explanation for all that we see in
the DMOs for the following reasons:
• The moment per cation at low concentrations can exceed that of any known
ferromagnetic phase [59];
• An unusual anisotropy of the saturation magnetisation, unlike that of known
ferromagnetic phases [53, 60].
Contamination is also a major concern and must be treated extremely seriously. Typ-
ically we are only dealing with 10 - 30 µg of material on top of a substrate which is
approximately 50 mg. Therefore, only the slightest speck of contamination could be
sufficient to produce a spurious magnetic moment in a SQUID magnetometer. The
substrate holder must also be cleaned thoroughly. The general protocol is to always
measure the substrate holder and a blank substrate before any measurement of the film
and make sure that there is no trace of ferromagnetism. Garcia zt vl. give a compre-
hensive review of the possible sources of experimental errors in nanoscale magnetism,
and give examples of magnetic signals which have emanated from Kapton tape, iron
based tools (tweezers, ztx), silver paste amongst others [61].
Above all, the reproducibility of the magnetisation from one sample to the next can
in some cases be very poor. Small changes in the deposition environment may lead to
drastic changes in the magnetisation.
The issues described above are all symptoms of the limited information that we can
get from simple magnetometry. If, for some laboratories, this is to be the sole source
of characterisation, then the work must be carried out carefully and in extremely clean
conditions.
5C5 Xvn v Ybd we xlvssey vs v YbhT
With all of the above taken into account, Coey zt vl. [62] recently produced a report
whereby the right to classify a DMO as a DMS in the strictest sense is questioned. They
coin the irreproducibility in magnetic behaviour as “phantom ferromagnetism”, and
suggest a new model for ferromagnetism in DMO materials based on the spin-split band
(Stoner model) theory rather than a Heisenberg exchange coupling of localised spins.
The model has been named the “charge-transfer model”. In fact, the model suggests
that the spins of 3y dopant electrons plays no significant role in the ferromagnetic order
in DMOs (contrary to DMS models based on the Heisenberg exchange interaction).
The basic idea is that a narrow, structured local density of states c (E ) is associated
with defects, but the Fermi level (which may lie above or below a mobility edge) will
77
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
M
n+ M ( +1)n +
e-
(a) (b) (c)
Figure 5.5: A representation of the charge-transfer ferromagmetism model showing the
charge transfer involved with a ferromagnetic instability. (a) Defect related density of
states is linked to an electron reservoir. (b) By electron transfer to or from the defects,
the Fermi level can be shifted to a position where spontaneous ferromagnetic splitting
of the defect related density of states occurs. Adapted from reference [62].
not normally coincide with a peak in c (E). However, if there is a local charge reservoir,
such as a dopant cation coexisting in two different charge states or a charge-transfer
complex at the surface, then it may be possible for electron transfer to raise the Fermi
level to a peak in the local density of states, provided that the energy gain from spin
splitting exceeds the energy cost of the charge transfer. This leads to a Stoner splitting
of c (E ) (Fig. 5.5). In the case of TM-doped oxides, the charge transfer in doped
oxides is facilitated by mixed-valence TM ions. Note that in this model the enabling
feature of the TM ions is not that they carry a localised magnetic moment, but it
is their ability to feed electrons into the local density of states (or to accept charge
from it) that explains the dopant cations ability to promote ferromagnetism, i.e. the
3y dopants act as a charge reservoir from which electrons are transferred to and from
local defect states. This general theory is able to explain a wide range of experimental
observations, but needs solid experimental proof. Furthermore, spontaneous Stoner
ferromagnetism can arise in percolating defect-rich regions, such as the nanoparticle
surface, and is therefore not uniform, and instead is restricted to some defect-rich
regions.
A selection of the different defects and segregations that may form are represented
in Fig. 5.6. Defects may be distributed uniformly throughout the material [Fig. 5.6
(a)] or form via spinodal decomposition [Fig. 5.6 (b)]. They may also be associated
with the surface or interface of a thin film [Fig. 5.6 (c)] or with grain boundaries [Fig.
5.6 (d)]. In this way, only a small fraction of the total volume of the sample need be
involved.
Through consideration of the total electronic energy involving the Stoner integral
in the defect band, the defect bandwidth, the total number of electrons (including the
defect band and reservoir) and the on-site Coulomb energy for the multi-valent TM
78
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
Figure 5.6: Possible distributions of defects: (a) random distribution of point defects
at the percolation threshold; (b) spinodal decomposition; (c) interface defects; and (d)
grain boundary defects. After reference [63].
dopants within the charge reservoir, Coey zt vl. have been able to devise a magnetic
phase diagram for the charge transfer ferromagnetism model.
It is also noted that the magnetic moment measured is not limited by the magnitude
of the charge transfer. It reflects the occupancy of the defect band, which becomes spin-
split as electrons are transferred to or from the charge reservoir. In fact, the moment
may exceed the number of transferred electrons. The charge reservoir simply serves to
push the energy balance of the defect band in favour of Stoner splitting. This model is
not only limited to the DMOs, but can also apply to yE ferromagnets as well, provided
there is a charge reservoir.
5CK Zflperimentvl gesults
5CKCF ereliminvry eflperiments
The reader must by now understand the difficulties experienced when working with
DMO materials. The struggle to develop my own dilute magnetic oxide system began
by first using a range of different transition metal dopants. In this work in particular,
those included Cr, Mn, Cu and Co. All films grown using Cu doping were diamagnetic.
Those ITO films doped with Cr and Mn were more promising, however, the success
rate for obtaining a ferromagnetic signal was at best 1 in 8. In addition, the Cr and
Mn-doped samples were only characterised using a VSM rather than a more sensitive
SQUID magnetometer. Altering the oxygen partial pressure during the growth process
as well as post-growth annealing did not improve the situation in the slightest. As we
will describe below, those samples doped with dilute amounts of Co were found to be
ferromagnetic and with a much better success rate: it was therefore decided that these
would be studied in greater detail, the results of which will follow in this chapter and
subsequent chapters 6 and 7. Due to the relatively low signals recorded we assume
that there are minimal demagnetising effects present in these samples.
79
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
Table 5.2: Summary of the magnetic and transport characteristics of 5.4 at.% Co-doped
InGOH thins films with Sn doping levels of 0, 5, and 10 at.% at room temperature.
Samples were grown using an oxygen partial pressure of 0.075 Pa in an Ar:OG mix and
heater temperature of 300◦C.
pn concentration nx Ms EHx
(at. %) (×10GE cm−H) (emu/cmH) (mT)
0.0 4.60±0.21×10FN 0.48±0.06 10.0±0.25
5.0 3.36±0.14×10GE 0.67±0.10 7.76±0.25
10.0 6.17±0.10×10GE 1.11±0.07 8.14±0.25
5CKCG Yepenyenxe on efltrinsix hn yoping xonxentrvtion
The magnetisation was strongly influenced by the level of extrinsic Sn doping intro-
duced into the InGOH matrix. Table 5.2 summarises the influence that external Sn dop-
ing had on InGOH, ITO(5) and ITO(10) films each doped with 5.4 at.% Co. Increasing
levels of Sn doping leads to a rise in the bulk magnetisation of the film, contrary to
the findings in reference [16]. The rise in magnetisation was accompanied by a steady
increase in the carrier concentration. This suggests that there may be some corre-
lation between the carrier concentration and magnetisation in these films and more
importantly, that the magnetisation is stronger with increasing levels of extrinsic car-
rier doping. The films doped with Sn are metallic whereas those without Sn exhibit
semiconducting properties. However, in both such cases, the magnetisation increases
sharply upon cooling. In the case of the Co-doped InGOH this seems rather counter-
intuitive as the carrier concentration will decrease slightly with decreasing temperature.
So why does the magnetisation continue to rise in these semiconducting films? One
possible explanation may be as follows: the carrier density at room temperature is the
density of ionised donor states; at low temperatures the donors will become neutral
and these localised states can also contribute to magnetism. The temperature depen-
dent data shows us that the localised states are actually more efficient at generating a
magnetic moment than the delocalised states. The magnetisation versus temperature
scans of the Co-doped films with 0, 5 and 10 at.% Sn follow similar paths with no
sign of a superparamagnetic blocking temperature. The zero-field cooled (ZFC) and
field-cooled (FC) magnetisation versus temperature curves for a 5.4 at.% Co-doped
InGOH thin film with an applied magnetic field of 20 mT is shown in Fig. 5.7.
From here onwards in this chapter, the data presented will focus on measurements
performed on Co-doped ITO(10) thin films as these provided us with the largest fer-
romagnetic response.
80
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
0 50 100 150 200 250 300
0.2
0.4
0.6
0.8
1.0
 ZFC
 FC
 
 
N
o
r
m
a
l
i
s
e
d
 
M
a
g
n
e
t
i
s
a
t
i
o
n
T (K)
Figure 5.7: ZFC-FC normalised magnetisation versus temperature scans for a 5.4
at.% Co-doped InGOH film. The applied field was 20 mT.
5CKCH XoByopey Iid(FE) films
A series of Co-doped ITO(10) (InG−xSnx−F(Cox)OH−Gx with fl = 0.016-0.054) thin films
(≈ 250 nm thick) were co-deposited (two targets, one of pure Co and the other a
commercially made ITO(10) target) on r -plane sapphire substrates by DC magnetron
sputtering in a UHV system. A power of 20 and 25 W was applied to the ITO(10)
and Co targets respectively. The heater was maintained at a temperature of 300◦C
during deposition. The system base pressure was better than 1 × 10−K Pa. Films were
sputtered in 1.50±0.01 Pa of an Ar-OG (95:5) mix (0.075 Pa OG) with a gas flow rate
of approximately 0.10±0.01 Pa/s.
htruxturvl eroperties
Figure 5.8 shows the θ-2θ XRD scans of Co-ITO films grown on r -plane sapphire for a
range of Co doping levels. All peaks displayed on the diffractogram could be indexed
assuming the same cubic bixbyite structure as pure ITO. Peaks were sharp, suggesting
that the films are crystalline. At first glance, no detectable peaks corresponding to any
secondary Co oxides or Co metal were found. This is often the case in DMOs which
appear to be phase-pure and of high crystalline quality. To verify these findings, the
integration time per point was raised to 100 seconds - a route also taken by Opel zt
vl. [64] for Co:ZnO, where they increased the integration time further to 400 seconds
and only then detected reflection peaks corresponding to the (111) reflection peak of fcc
Co metal as well as ZnCoGOI (400) and CoHOI (400). Still, with this adjustment and
within the detection limit of the instrument, we were unable to identify any diffraction
81
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
Figure 5.8: (a) θ-2θ XRD diffractograms focusing on the ITO (222) and ITO (400)
peaks for Co-ITO films of varying Co concentrations, demonstrating the shift in the
peak position and intensity upon Co doping. Inset: lattice parameter, a (A˚), as a
function of Co doping. The dotted line serves as a guide to the eye.
peaks which would suggest the presence of metallic Co clusters or any other secondary
Co oxide phases.
A shift in the peaks (predominantly the highly pronounced ITO (222) diffraction
peak) towards higher values of 2θ with increasing level of Co doping is observed and
suggests a gradual decrease in the out-of-plane lattice parameter, v (inset, Fig. 5.8).
This is evidence that the smaller CoG+ ions (ionic radius 0.74 A˚) substitute for the
larger InH+ ions (ionic radius 0.94 A˚) at In sites rather than forming a defect phase.
We know that this characterisation technique has its limitations in terms of de-
tecting dilute levels of transition metal dopants. A more detailed study involving the
use of high-resolution cross-sectional transmission electron microscopy would certainly
help and give a more representative picture of the microstructure. Co is renowned
for forming clusters and in the past has been shown to accumulate close to the sub-
strate/film interface [8] (see Fig. 5.9), and so caution must be exercised.
bvgnetometry
In-plane magnetisation (b ) versus field (µEH) measurements showed clear evidence for
ferromagnetism in each of the samples. The diamagnetic background of the substrate
and sample holder has been subtracted from all the magnetisation data shown here.
Figures 5.10 - 5.12 (a) show the magnetisation loops for fl = 0.016, 0.03 and 0.054
respectively over a temperature range of 2 - 300 K. A noticeable difference in the shape
of the magnetisation curves between the 1.6 at.% doped sample and those of the 3.0
and 5.4 at.% Co-doped samples is apparent. The 1.6 at.% doped sample appears to
82
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
Figure 5.9: TEM images of a Co:TiOG film at different magnifications showing the
formation of Co clusters close to the substrate interface with the film [8].
have magnetisation curves which are more “square” than those of the other doping lev-
els, i.e. a larger remanent magnetisation. This is confirmed by a simple calculation of
the ratio between the saturation and remanent magnetisation (mr/ms). The average
mr/ms for the 1.6 at.% Co-doped sample was 0.41, compared to values of approxi-
mately 0.20 in the other samples with higher doping levels (calculated by averaging
over all temperatures). Interestingly, it is at 2 K in the 1.6 at.% doped sample that
mr/ms is at its minimum value (≈ 0.31). It is only at temperatures ≥ 10 K that the
magnetisation curves become more square. The magnetisation saturation fields for the
samples are typically between 0.20 - 0.45 T, the largest values arising at 2 K.
Figures 5.10 - 5.12 (b) provide an expanded view of the low-field data to demon-
strate the nature of the coercive field (µE] x) and how it changes between 2 and 300
K. Hysteretic behaviour is observed, consistent with ferromagnetism, with values for
µE] x ranging from 8 - 75 mT across the Co doping range.
Figures 5.10 - 5.12 (c) and 5.13 (c) summarise the behaviour of µE] x as a function of
the temperature. For each doping level µE] x decreases steadily with temperature, and
in the case where we have 5.4 at.% Co-doped ITO [Fig. 5.12 (c)], µE] x falls according
to an inverse relationship to the temperature. Fig. 5.13 (c) in particular gives a clearer
picture of how µE] x behaves as a function of temperature for each of the doping levels.
By displaying the normalised µE] x (normalised to the maximum µE] x(2 K) for each
data set) we can see that in the changes in µE] x over the temperature range become
more drastic as the level of Co doping increases.
The magnetisation loops at 300 K for the 1.6, 3.0 and 5.4 at.% Co-doped samples
are shown in Fig. 5.13 (a). The magnetisation rises almost linearly with increasing
levels of Co content. This implies that the Co dopants are acting independently of
each other. Had there been an exponential rise with Co content, it would have implied
ferromagnetic clustering, whereas a negative parabolic trend would suggest antiferro-
83
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
magnetic clusters.
Figures 5.10 - 5.12 (d) show the normalised FC magnetisation of the 1.6, 3.0
and 5.4 at. % Co-doped ITO films as a function of temperature in a constant applied
magnetic field of 20 mT in the 5 - 300 K temperature range. In each case, the sample
magnetisation decreases rapidly particularly between 2 - 40 K. This implies that the
sample possesses a strong paramagnetic component in addition to a weak ferromag-
netic component, as seen in other DMO systems [27]. Initially we attempted to fit
the data using a purely paramagnetic relationship, however, this proved unsuccessful.
Assuming that we have the coexistence of magnetic phases within this sample, we were
able to model the magnetic susceptibility, χ, using a modified Curie-Weiss relationship
in combination with a standard mean-field model for a weak itinerant ferromagnet [65]
using the relation:
χ = [1− α]
(
χE +
C
i− θ
)
+ α
√
1− i
G
iGX
, (5.2)
where α is magnetic phase fraction (the ferromagnetic fraction), χE is a constant non-
paramagnetic contribution to the magnetic susceptibility, C is the Curie constant (=
nµG/3kB, where kB is Boltzmann’s constant, n number of magnetic ions and µ the
magnetic moment of the ion), θ is the Curie-Weiss temperature and describes the mag-
netic interactions between the paramagnetic spins, i is the measurement temperature,
and iX is the Curie temperature of the ferromagnet. The values for α are found to
be 17.8, 17.2 and 17.6% for the 1.6, 3.0 and 5.4 at.% Co-doped ITO films respectively,
confirming that the paramagnetic contribution is dominant in these samples. The fit
estimates values for iX of approximately 500 K in each case. There was no sign of a
blocking temperature, thus suggesting the absence of secondary Co metal nanoclusters
within the bulk of the film. However, it has been suggested that a magnetisation versus
temperature measurement is not necessarily the most rigorous method to expose a su-
perparamagnetic blocking temperature (iB). It is believed that iB can sometimes be
suppressed by the ever present superexchange coupling between the Co particles [66].
5CKCI Yepenyenxe on oflygen pvrtivl pressure
As one would expect, changes in the oxygen partial pressure during the growth process
influenced the magnetic properties of the samples. It was found that samples grown in
more oxygen deficient atmospheres produced a greater magnetic response than those
grown in environments which were richer in oxygen. The majority of films which
were metallic (generally those grown in oxygen partial pressures of ≤ 0.15 Pa) showed
some level of ferromagnetic response, whereas those which were grown and possessed
semiconducting properties were found to be diamagnetic. Figure 5.14 shows the mag-
84
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
(d)
Figure 5.10: (a) Magnetic hysteresis loops using a SQUID magnetometer for a 1.6
at.% Co-doped ITO film at varying temperatures. (b) Magnified image showing the
large change in coercive field between 2 K and 300 K. (c) Variation in coercive field as
a function of temperature; the dotted line is a guide to the eye. (d) Normalized field-
cooled magnetization for a 1.6 at.% Co-doped ITO film as a function of temperature
in an applied magnetic field of 20 mT; the solid line fits the experimental data to the
magnetic dual-phase model described in Eqn. 5.2.
85
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
(d)
Figure 5.11: (a) Magnetic hysteresis loops using a SQUID magnetometer for a 3.0
at.% Co-doped ITO film at varying temperatures (b) Magnified image showing the
large change in coercive field between 2 K and 300 K (c) Variation in coercive field as
a function of temperature; the dotted line is a guide to the eye. (d) Normalized field-
cooled magnetization for a 3.0 at.% Co-doped ITO film as a function of temperature
in an applied magnetic field of 20 mT; the solid line fits the experimental data to the
magnetic dual-phase model described in Eqn. 5.2.
86
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
Figure 5.12: (a) Magnetic hysteresis loops using a SQUID magnetometer for a 5.4
at.% Co-doped ITO film at varying temperatures. (b) Magnified image showing the
large change in coercive field between 2 K and 300 K. (c) Variation in coercive field as
a function of temperature; the dotted line is a guide to the eye. (d) Normalized field-
cooled magnetization for a 5.4 at.% Co-doped ITO film as a function of temperature
in an applied magnetic field of 20 mT; the solid line fits the experimental data to the
magnetic dual-phase model described in Eqn. 5.2.
87
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
-0.4 -0.2 0.0 0.2 0.4
-1.0
-0.5
0.0
0.5
1.0
0 50 100 150 200 250 300
0.0
0.2
0.4
0.6
0.8
1.0
-0.10 -0.05 0.00 0.05 0.10
-0.5
0.0
0.5
0
H (T)
(c)(b)
 1.6 at.%
 3.0 at.%
 5.4 at.%
0
H (T)
M
a
g
n
e
t
i
s
a
t
i
o
n
 
(
e
m
u
/
c
m
3
)
(a)
M
a
g
n
e
t
i
s
a
t
i
o
n
 
(
e
m
u
/
c
m
3
)
 
 
N
o
r
m
a
l
i
s
e
d
 
0
H
C
T (K)
Figure 5.13: (a) Magnetisation versus applied field loops at 300 K, (b) expanded view
of the low-field region for the magnetisation curves shown in (a) at 300 K, and (c)
normalised coercive field, ] x, as a function of temperature for 1.6 (open black circles),
3.0 (open red diamonds) and 5.4 at.% (open blue triangles) Co-doped ITO films.
88
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
10.2 10 cm
20 -3
7.2 10 cm
20 -3
2.1 10 cm
20 -3
Figure 5.14: Room temperature magnetisation versus applied field loops for 5.4 at.%
Co-doped ITO films grown using a heater temperature of 300◦C under various oxygen
partial pressures. Corresponding carrier concentrations are also given.
netisation loops (using a VSM) along with the corresponding carrier concentrations at
room temperature for 5.4 at.% Co-doped ITO films grown under a number of different
oxygen partial pressures. It clearly shows the reduction in the overall magnetisation
of the film as the oxygen partial pressure rises. As we know from the previous chapter
from our work on the ITO system, there are a number of film properties which are
influenced by the oxygen partial pressure during the growth stage. The first obvious
reason for the reduction in the strength of the magnetic signal is likely to be due to
the decreasing carrier concentration with increasing levels of oxygen. As stated above,
those films which were semiconducting were found to be diamagnetic. Earlier we dis-
cussed predictions by Raebiger zt vl. where it is said that InGOH will only show room
temperature ferromagnetism in the company of extrinsic Sn and TM doping, provided
the condition of the concentration of free carriers reaches half that of the Cr dopant
concentration (however, we have shown here that Co-doped InGOH films without extrin-
sic Sn doping are ferromagnetic at room temperature) [13]. In the case of Raebiger’s
predictions, this carrier concentration was equivalent to 3.0×10GE cm−H. The carrier
concentration of the Co-doped ITO films showing semiconducting behaviour was typ-
ically 0.9 - 2.0×10FN cm−H. Therefore, we can only estimate that somewhere between
the region of 2.0×10FN cm−H - 2.0×10GE cm−H, there is a transition from a diamag-
netic semiconductor to a metallic-like oxide ferromagnet for our Co-doped ITO films.
Oxygen vacancies are perceived by many to have a strong role to play in the origins
of ferromagnetism in these materials, and so the rising levels of oxygen will naturally
disrupt their influence as a result of the filling of oxygen vacancies.
No detailed structural analysis was performed on these samples, however, in chap-
89
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
ter 4 it was found that increasing levels of oxygen during the growth process led to a
reduction in the vertical grain size of the films. This is likely to lead to more point
defects in the films. Some believe that defects can improve the strength of the mag-
netic response in the DMO materials, and have shown that moving towards epitaxy
can suppress the ferromagnetism [67]. Perhaps in the case here however, the magnetic
response benefits from improved crystallinity.
Increasing levels of oxygen during the growth process will surely increase the chances
of forming Co oxide clusters and the likelihood of a reduction in the magnetic response
of the samples. The majority of the likely candidates for Co oxide clustering e.g. CoHOI
and CoGOH, generally have ferromagnetic Curie temperatures of below 40 K and so, if
present, will not show any ferromagnetic response at room temperature.
5CKC5 Yepenyenxe on the grofith tempervture
The majority of the results above describe data collected for films using relatively high
heater temperatures (generally ≥ 300◦C), however, the scenario whereby the films were
grown unheated was also investigated. These films exhibited typical semiconducting
characteristics with a room temperature resistivity of the order of 10−G Ωcm. The
magnetic response of the films could be modelled using a Brillouin function, i.e. that
of a paramagnetic or superparamagnetic material. This suggests that some heating is
required during the deposition process in order to promote a ferromagnetic response
from the films, and may be linked to factors such as the microstructure, density of
oxygen vacancies, ztx. In addition, it could be that the Co atoms are not successful in
substituting for the In atoms under these growth conditions.
5CKCK bvgnetix Iid C C C T
The possibility of pure, undoped ITO being ferromagnetic was also investigated very
briefly. Two identical pure ITO(10) samples, of thickness 300 nm, were grown un-
der identical conditions to those described above (see section 5.6.3) for the Co-doped
samples and measured using SQUID magnetometry at 2 and 290 K. Rather surpris-
ingly a small magnetic moment, albeit very weak, at both 2 and 290 K is observed
(see Fig. 5.15). There is a minimal change in magnetisation between 2 and 290 K.
The magnetisation at room temperature (≈ 0.22 emu/cmH) is comparable to, yet lower
than that of the 1.6 at.% Co-doped ITO sample at 290 K (0.35 emu/cmH). Others
who have observed room temperature ferromagnetism in undoped InGOH report values
of 0.30 emu/cmH (200 nm thick films) [15], 0.70 emu/cmH (600 nm thick films) [31]
and even as high as 1.40 emu/cmH (100 nm thick films) [69]. To most onlookers, the
obvious explanation for these results would be contamination. There are not too many
experimental cross-checks which are readily available to confirm whether these results
90
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
-0.4 -0.2 0.0 0.2 0.4
-0.3
-0.2
-0.1
0.0
0.1
0.2
0.3
 2 K
 290 K
M
a
g
n
e
t
i
s
a
t
i
o
n
 
(
e
m
u
/
c
m
3
)
0
H (T)
Figure 5.15: Magnetisation curves for a pure, undoped ITO(10) film grown under
indentical conditions to the Co-doped ITO samples at 2 and 290 K.
are valid or not. In section 5.3.1 we discussed the results of Panguluri zt vl. where it
was shown that pure InGOH has a sizeable spin polarisation [15]. Another option would
be to study the oxygen K -edge in detail using XMCD, but again, this is not an easy
task, due to low signal levels (the moment per oxygen atom is likely to be very small)
and the difficulty in being granted beam time.
As Coey zt vl. [62] and others have suggested, the role of the transition metal dop-
ing is questionable. Hong zt vl. [31] report that there is a strong thickness dependence
on the magnetisation in their undoped oxide films suggesting that the magnetism is
likely to be oxygen vacancy or defect related. Perhaps we do need to rejudge the role
that TM doping has on introducing ferromagnetism in oxides: does the doping really
introduce ferromagnetism? Or, in fact, does it just enhance the magnetism that al-
ready exists in the oxide hosts when in thin film form? Certainly in our case it seems
that the addition of Co does enhance the magnetism we come to observe.
5CKCL irvnsport bevsurements
The carrier concentration is these films at 10 and 290 K was typically in the 10GE - 10GF
cm−H range, with electron mobilities between 8 - 10 cmGV−Fs−F. The anomalous Hall
effect was not observed in any of the films even at 10 K. This suggests that the degree
of spin polarisation in these materials may be relatively low.
Figure 5.16 shows a typical in-plane MR curve for a Co-doped (5.4 at.%) ITO film
at 10 K. A negative MR is observed for field up to ±5 T, with a maximum of -0.17%.
91
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
-4 -2 0 2 4
-0.16
-0.12
-0.08
-0.04
0.00
M
R
 
(
%
)
0
H (T)
Figure 5.16: Magnetoresistance curve at 10 K for a 5.4 at.% Co-doped ITO(10)
sample.
The MR curves are extremely similar to those shown in the pure ITO(10) samples
shown in Fig. 4.4 implying that the addition of Co has little effect on the magnetore-
sistance behaviour. The negative MR can be attributed to electron scattering by spin
disorder. Looking back at the hysteresis loops for this sample [Fig. 5.13 (a)] we recall a
small remanent magnetisation, indicating that the moments are not well ordered in the
remanent state. Therefore, as the magnetic field increases, these magnetic spins will
become increasingly aligned, thereby reducing the spin disorder scattering [68]. Hence,
the resistance is expected to become smaller with increasing applied field, resulting in
the observed negative MR. No MR was recorded at 290 K.
As we discussed in section 2.3.1, magnetotransport measurements can be used as
an indicator as to whether a material is undertaking the role of an efficient spin po-
lariser. The positive contribution is most frequently quoted as being due to the field
induced spin-splitting of the conduction band and the population change this causes.
The negative MR is attributed to a reduction in spin scattering with increasing mag-
netic field. Since the overall MR seen is determined by a competition between the
negative and positive mechanisms, this means that in the Co-doped ITO(10) (and in
fact even in those with no Sn and only 5 at.% Sn) weak localisation effects and spin-
disorder scattering dominate the s-y exchange interaction in our films.
5CKCM Vnnevling tests
The effects of annealing on the magnetisation and magnetotransport behaviour of Co-
doped ITO films have also been investigated. Here we focus solely on 5.4 at.% Co-doped
films which have been grown in the same conditions as those described in section 5.6.3.
92
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
M
(em
u
/cm
)
s
3
Figure 5.17: Summary of the transport properties (carrier concentration, nx, and
mobility, µH) and magnetisation (b s) at room temperature for 5.4 at.% Co-doped
ITO films as a function of the heater temperature, iAnn (
◦C). Note that 300◦C is the
as-deposited condition. The grey shaded area signifies the region where iAnnS600
◦C
and we suspect the presence of ferromagnetic Co clusters. The data points encased
within the dotted oval signifies that this sample was later used for further experiments
using XMCD (see chapter 6).
Films were annealed in UHV conditions within the same chamber in which they were
deposited. Films were annealed at different heater temperatures (iAnn) for a fixed
time of 2 hours.
θ-2θ XRD studies revealed no secondary Co phases. This does not rule out the
possibility of these phases being present. The dilute levels of these materials makes it
extremely difficult to identify any parasitic phases.
Figure 5.17 shows the dependence of the film carrier concentration, mobility and
magnetisation at room temperature on iAnn. Between 300≤iAnn(◦C)≤600, the carrier
concentration and magnetisation show a clear correlation, as they follow a similar path
as a function of the annealing temperature. After the initial annealing stage at 400◦C,
magnetisation seems to saturate, however, rises suddenly for iAnnS600
◦C (grey shaded
area) reaching a maximum magnetisation of 3.03 emu/cmH at iAnn=900
◦C. This is ac-
companied by a drop in the carrier concentration, but is not large enough outside the
error range to be considered too significant. There was no suggestion of the presence
of superparamagnetic particles from the magnetisation loops. The coercive field pro-
gressively increased with annealing temperature, starting at approximately 7.5 mT for
the as-deposited sample and rising to 32 mT for the sample annealed at 900◦C.
ZFC-FCMagnetisation versus temperature curves for the sample annealed at 700◦C,
93
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
0 50 100 150 200 250 300
2.2
2.4
2.6
2.8
3.0
3.2
3.4
3.6
M
 
(
e
m
u
/
c
m
3
)
 ZFC
 FC
 
 
T (K)
10 mT
Figure 5.18: ZFC-FC magnetisation versus temperature curves for a 5.4 at.% Co-
doped ITO sample annealed at 700◦C for 2 hours. The applied field was 10 mT.
taken using a SQUID magnetometer, are shown in Fig. 5.18. There is a clear separa-
tion between the ZFC and FC curves, which coalesce at approximately 272 K. This is
the value for iB which can now be used to estimate the cluster size using Eqn. 2.8.
Assuming spherical shaped clusters, the calculation yields a particle radius of 3.36 nm.
Note that the ZFC curve is very broad and may even indicate multiple values for iB.
This suggests that there may be a broad distribution of particle sizes.
Therefore, in the case of the samples annealed at iAnnS600
◦C, we believe that
the source of magnetism is extrinsic in the form of ferromagnetic, metallic Co clusters.
In particular, the sample annealed at 700◦C shows clear signs of secondary metallic Co
segregation, and in fact is investigated further in chapter 6 using XMCD.
No significant changes in the magnetoresistive response were seen when comparing
the as-deposited and annealed samples. A negative MR was recorded, and is displayed
as a function of temperature for the sample annealed at 700◦C in Fig. 5.19. No MR
was observed at 290 K.
From these annealing experiments it would not be too out of place to presume that
there is some correlation between carrier concentration and the magnetisation. Sam-
ples annealed between 300≤iAnn(◦C)≤600 are good evidence for this. Exposing the
samples to 700◦C (and above) for prolonged periods only leads to undesirable effects
in terms of achieving intrinsic ferromagnetic properties. The AHE was not observed in
any of these samples.
94
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
-4 -2 0 2 4
-0.16
-0.14
-0.12
-0.10
-0.08
-0.06
-0.04
-0.02
0.00
 
 
M
R
 
(
%
)
0
H (T)
 10 K
 100 K
 200 K
Figure 5.19: Magnetoresistance as a function of temperature for a 5.4 at.% Co-doped
ITO film annealed at 700◦C for 2 hours.
5CL hummvry vny Xonxlusions
This chapter has highlighted the key aspects of dilute magnetic oxide materials and
brought the controversies and experimental difficulties that overshadow the topic to
the readers’ attention.
It was found that only those films doped with Co exhibited reproducible ferromag-
netic behaviour beyond room temperature. Ferromagnetism was found in Co-doped
InGOH and 5 and 10 at.% Sn-doped InGOH, however the results presented are predomi-
nantly concerned with the Co-doped 10 at.% Sn-doped InGOH films. The ferromagnetic
response was found to increase with increasing levels of Sn doping which was accom-
panied by a rise in the carrier concentration.
X-ray diffraction studies reveal that the out-of-plane lattice parameter decreases
upon Co doping, suggesting that the Co ions are substituting for In ions and not form-
ing metal Co clusters. This is supported by the absence of any secondary Co metal or
Co oxide phases in the θ-2θ x-ray scan. However, it is made clear that this alone is not
strong enough evidence to suggest that the samples are completely Co cluster free.
Magnetometry revealed that the ITO(10) films doped with 1.6, 3.0 and 5.4 at.%
Co were all ferromagnetic beyond room temperature, with no sign of a superparam-
agnetic blocking temperature from magnetisation versus temperature scans. iX was
estimated to be approximately 500 K in each case from the dual phase fitting. Samples
grown in oxygen deficient environments were found to be more magnetic than those
grown using higher levels of oxygen. The rising magnetic response was accompanied
95
Cb[pnel /. Tbe Qoemn fil [ Dcfone M[ahenc] Ircde
by higher carrier concentrations, i.e. only those films which had metallic-like conduct-
ing properties exhibited ferromagnetism, generally with carrier concentrations ≥10GE
cm−H. Films grown at room temperature were found to be superparamagnetic even
when deposited in highly oxygen deficient atmospheres, suggesting that some level of
heating is required during the growth process.
Rather surprisingly, a reproducible weak ferromagnetic response was measured
above room temperature for a 300 nm thick pure (non-TM doped) ITO(10) film. This
result brings the role that transition metal dopants play in the observation of ferro-
magnetism into major dispute, and supports a defect-related origin for magnetism.
Annealing a set of 5.4 at.% Co-doped ITO(10) films under UHV conditions at
various annealing temperatures revealed that both the carrier concentration and mag-
netisation increase in tandem up to an annealing temperature of 600◦C. It is believed
that further annealing beyond 600◦C leads to Co metal segregation and the forma-
tion of metal Co clusters. This is accompanied by a sharp rise in the magnetisation
which continues as the annealing temperature is increased further. The presence of
superparamagnetic particles is supported by the observation of a superparamagnetic
blocking temperature from magnetisation versus temperature measurements.
The anomalous Hall effect was not observed in any of these samples reported in this
chapter. This suggests that the Co:ITO system is likely to be a weak spin polariser.
This is supported by very low values for magnetoresistance even at temperatures as
low as 10 K. No magnetoresistance observed at room temperature.
Some of the data gathered here may be considered solid enough evidence to suggest
that the magnetism observed in these films could be correlated to the carrier concen-
tration, therefore pointing towards a carrier-mediated mechanism for ferromagnetism.
However, the observation of room temperature ferromagnetism from the pure ITO
samples is confusing. For this reason, I refrain from expressing the magnetisation in
terms of Bohr magnetons per Co atom (µB/Co), which is often the standard method
of expressing the magnetisation. It implies that the magnetisation is solely a conse-
quence of the transition metal doping, which from our results, may or may not be the
case. With this in mind, one way to try and make head-way of these issues is to try
and understand the role of the transition metal dopants. As discussed, XRD, mag-
netometry and magnetotransport characterisation alone does not suffice when trying
to elucidate the physical mechanisms behind the magnetism in these systems - more
powerful techniques are called for to give greater detail on the dopant distribution and
chemical states. In the following two chapters, characterisation of Co-doped InGOH and
ITO thin films are carried out using two different forms of magnetic circular dichroism:
first, using x-rays (chapter 6), and second, using an optical light source (chapter 7).
96
Wiwliogrvphy
[1] T. Mizokawa, T. Nambu, A. Fujimori, T. Fukumura, M. Kawasaki, Phys. Rev.
B65, 085209, (2001).
[2] E. H. Hall, Amer. J. Math. 2, 287, (1879).
[3] H. Toyosaki, T. Fukumura, Y. Yamada, K. Nakajima, T. Chikyow, T. Hasegawa,
H. Koinuma, M. Kawasaki, Nature Mater. 3, 221, (2004).
[4] H. Toyosaki, T. Fukumura, Y. Yamada, M. Kawasaki, Appl. Phys. Lett. 86,
182503, (2005).
[5] K. Ueno, T. Fukumura, H. Toyosaki, M. Nakano, M. Kawasaki, Appl. Phys. Lett.
90,072103, (2007).
[6] H. S. Hsu, C. P. Lin, H. Chou, J. C. A. Huang, Appl. Phys. Lett. 93, 142507,
(2008).
[7] S. X. Zhang, W. Yu, S. B. Ogale, S. R. Shinde, D. C. Kundaliya, W. K. Tse,
S. Y. Young, J. S. Higgins, L. G. Salamanca-Riba, M. Herrera, L. F. Fu, N. D.
Browning, R. L. Greene, T. Venkatesan, Phys. Rev. B 76, 085323, (2007).
[8] S. R. Shinde, S. B. Ogale, J. S. Higgins, H. Zheng, A. J. Millis, V. N. Kulkarni, R.
Ramesh, R. L. Greene, and T. Venkatesan, Phys. Rev. Lett. 92, 166601, (2004).
[9] Q. Xu, H. Schmidt, S. Zhou, K. Potzger, M. Helm, H. Hochmuth, M. Lorenz, A.
Setzer, P. esquinazi, C. Meinecke, M. Grundmann, Appl. Phys. Lett. 92, 082508,
(2008).
[10] T. Fukumura, H. Toyosaki, Y. Yamada, Semicond. Sci. Technol. 20, S103S111,
(2005).
[11] J. Philip, A. Punnoose, B.I. Kim, K. M. Reddy, S. Layne, J. O. Holmes, B.
Satapati, P. R. Leclair, T. S. Santos, J. S. Moodera, Nature Mater. 5, (2006).
[12] H. S. Kim, S. H. Ji, H. Kim, S. K. Hong, D. Kim, Y. E. Ihm, W. K. Choo, Solid
State. Comm. 137, 41-43, (2006).
[13] H. Raebiger, S. Lany, A. Zunger, Phys. Rev. Lett. 101, 027203, (2008).
97
Bcbfcial[pbs
[14] F. X. Jiang, X. H. Xu, J. Zhang, X. C. Fan, H. S. Wu, G. A. Gehring, Appl. Phys.
Lett. 96, 052503, (2010).
[15] R. P. Panguluri, P. Kharel, C. Subdakar, R. Naik, R. Suranarayanan, V. M. Naik,
A. G. Petukhov, B. Nadgorny, G. Lawes, Phys. Rev. B 79, 165208, (2009).
[16] X. L. Wang, G. Peleckis, S. X. Dou, R. S. Liu, J. G. Zhu, J. Appl. Phys. 101,
09H121, (2007).
[17] J. He, S. Xu, Y. K. Yoo, Q. Xue, H. C. Lee, S. Cheng, X. D. Xiang, I. Takeuchi,
Appl. Phys. Lett. 86, 052503, (2005).
[18] Z. G. Yu, J. He, S. Xu, Q. Xue, O. M. J. van’t Erve, B. T. Jonker, M. A. Marcus,
Y. K. Yoo, S. Cheng, X. D. Xiang, Phys. Rev. B 74, 165321, (2006).
[19] V. Brize, J. Sakai, N. H. Hong, Physica B, 379-382, (2007).
[20] J. Philip, N. Theodoropoulou, G. Berera, J. Moodera, B. Satpati, Appl. Phys.
Lett. 85, 5, (2004).
[21] J. Stankiewicz, F. Villuendas, J. Bartomle, Phys. Rev. B 75, 235308, (2007).
[22] G. Subias, J. Stankiewicz, F. Villuendas, M. P. Lozano, J. Garcia, Phys. Rev. B
79, 094118, (2009).
[23] P. Kharel, C. Sudakar, M. B. Sahana, G. Lawes, R. Suryanarayanan, R. Naik, V.
M. Naik, J. Appl. Phys. 101, 09H117, (2007).
[24] N. H. Hong, J. Sakai, N. T. Huong, V. Brize, Appl. Phys. Lett. 87, 102505, (2005).
[25] G. Peleckis, X. L. Wang, S. X. Dou, Appl. Phys. Lett. 89, 022501, (2006).
[26] T. Nakamura, K. Tanabe, K. Tsureishi, K. Tachibana, J. Appl. Phys. 101, 09H105,
(2007).
[27] K. M. Reddy, J. Hays, S. Kundu, L. K. Dua, P. K. Biswas, C. Wang, V. Shut-
thanandan, M. H. Englehard, X. Mathew, A. Punnoose, J. Mater. Sci: Mater.
Electron 18:1197-1201, (2007).
[28] G. Peleckis, X. L. Wang, S. X. Dou, Appl. Phys. Lett. 88, 132507, (2006).
[29] J. M. Baik, Y. Shoon, T. W. Kang, J. L. Lee, Jap. J. Appl. Phys., (2008).
[30] J. Stankiewicz, F. Villuendas, J. Bartolome, J. Sese, J. Magn. Magn. Mater. 310,
2084-2086, (2007).
[31] N. H. Hong, J. Sakai, N. Poirot, V. Brize, Phys. Rev. B 73, 132404, (2006).
[32] M. Venkatesan, R. D. Gunning, P. Stamenov, J. M. D. Coey, Appl. Phys. Lett.
103, 07D135, (2008).
98
Bcbfcial[pbs
[33] S. J. Pearton, C. Abernathy, M. E. Overberg, G. T. Thaler, D. P. Norton, N.
Theodoropoulou, A. F. Hebard, Y. D. Park, F. Ren, J. Kim, L. A. Boatner, J.
Appl. Phys. 93, 1, (2003).
[34] J. M. D. Coey, J. Appl. Phys. 97, 10D313, (2005).
[35] S. A. Chambers, Surf. Sci. Rep. 61, 345-381, (2006).
[36] F. Pan, C. Song, X. J. Liu, Y. C. Yang, F. Zeng, Mat. Sci. Eng. R 62, 1-35, (2008).
[37] M. Venkatesan, P. Stamenov, L. S. Dorneles, R. D. Gunning, B. Bernoux, J. M.
D. Coey, Appl. Phys. Lett. 90, 242508, (2007).
[38] Z. W. Zhao, B. K. Tay, J. S. Chen, J. F. Hu, B. C. Lim, G. P. Li, Appl. Phys.
Lett. 90, 152502, (2007).
[39] T. Deitl, T. Andrearczyk, A. Lipinska, M. Kiecana, M. Ttay, Y. Wu, Phys. Rev.
B 76, 155312, (2007).
[40] G. S. Chang, E. Z. Kurmaev, D. W. Boukhvalov, L. D. Finkelstein, S. Colid, T.
M. Pedersen, A. Moewes, A. Dinia, Phys. Rev. B 75, 195215, (2007).
[41] M. Kobayashi, Y. Ishida, J. l. Hwang, T. Mizokawa, A. Fujimori, K. Mamiya, J.
Okamoto, Y. Takeda, T. Okane, Y. Saitoh, Y. Muramatsu, A. Tanaka, H. Saeki,
H. Tabata, T. Kawai, Phys. Rev. B 72, 201201(R), (2005).
[42] N. Sanchez, S. Gallego, M. C. Munoz, Phys. Rev. Lett. 101, 067206, (2008).
[43] F. Schoofs, T. Fix, A. M. H. R. Hakimi, S. S. Dhesi, G. van der Laan, S. A.
Cavill, S. Langridge, J. L. MacManus-Driscoll, M. G. Blamire, J. Appl. Phys. 108,
053911, (2010).
[44] R. Hanafin, T. Archer, S. Sanvito, Phys. Rev. B 81, 054441, (2010).
[45] T. C. Kaspar, T. Droubay, S. M. Heald, P. Nachimuthu, C. M. Wang, V. Shut-
thanandan, C. A. Johnson, D. R. Gamelin, S. A. Chambers, New J. Phys. 10,
055010, (2008).
[46] C. D. Pemmaraju, R. Hanafin, T. Archer, H. B. Braun, S. Sanvito, Phys. Rev. B
78, 054428, (2008).
[47] Y. Matsumoto, M. Murakami, T. Shono, T. Hasegawa, T. Fukumura, M.
Kawasaki, P. Ahmet, T. Chikyow, S.-Y. Koshihara, H. Koinuma, Science 291,
854, (2001).
[48] Y. Matsumoto, M. Murakami, K. Hasegawa, T. Fukumura, M. Kawasaki, P. Ah-
met, K. Nakajima, T. Chikyow, H. Koinuma, Appl. Surf. Sci. 189, 344, (2002).
99
Bcbfcial[pbs
[49] J. Y. Kim, J. H. Park, B. G. Park, H. J. Noh, S. J. Oh, J. S. Yang, D. H. Kim,
S. D. Bu, T. W. Noh, H. J. Lin, H. H. Hsieh, C. T. Chen, Phys. Rev. Lett. 90, 1,
(2003).
[50] D. H. Kim, J. S. Yang, Y. S. Kim, T. W. Noh, S. D. Bu, S. I. Baik, Y. W. Kim,
Y. D. Park, S. J. Pearton, J. Y. Kim, J. H. Park, H. J. Lin, C. T. Chen, Y. J.
Song, Phys. Rev. B 71, 014440, (2005).
[51] G. C. Han, P. Luo, Z. B. Guo, F.Un Nahar, M. Tay, Y. H. Wu, S. J. Wang, Thin
Solid Films 505, 137-140, (2006).
[52] Y. J. Lee, M. P. de Jong, R. Jansen, Appl. Phys. Lett. 96, 082506, (2010).
[53] M. Venkatesan, C. B. Fitzgerald, J. M. D. Coey, Nature (London) 430, 630, (2004).
[54] J. M. D. Coey, M. Venkatesan, C. B. Fitzgerald, Nature, 4, (2005).
[55] N. H. Hong, J. Sakai, N. Poirot, V. Brize, Phys. Rev. B 73, 132404, (2005).
[56] D. Berardan, E. Guilmeau, D. Pelloquin, J. Magn. Magn. Mater. 320, 9830989,
(2008).
[57] D. W. Abraham, M. M. Frank, S. Guha, Appl. Phys. Lett. 87, 252502, (2005).
[58] J. M. D. Coey, Curr. Opin. Solid State and Materials Science 10, 8392, (2006).
[59] S. B. Ogale, R. J. Choudhary, J. P. Buban, S. E. Lofland, S. R. Shinde, S. N.
Kale, V. N. Kulkarni, J. Higgins, C. Lanci, J. R. Simpson, N. D. Browning, S. Das
Sarma, H. D. Drew, R. L. Greene, T. Venkatesan, Phys. Rev. Lett. 91, 7, (2003)
[60] J. M. D. Coey, M. Venkatesan, C. B. Fitzgerald, L. S. Dorneles, P. Stamenov, J.
G. Lunney, J. Magn. Magn. Mater 290, 1405-7, (2005).
[61] M. A. Garcia, E. Fernandez Pinel, J. de la Venta, A. Quesada, V. Bouzas, J. F.
Fernandez, J. J. Romero, M. S. Martin Gonzalez, J. L. Costa-Kramer, J. Appl.
Phys. 105, 013925, (2009).
[62] J. M. D. Coey, K. Wongsaprom, J. Alaria, M. Venkastesan, J. Phys. D: Appl.
Phys. 41, 134012, (2008).
[63] J. M. D. Coey, P. Stamenov, R. D. Gunning, M. Venkatesan, K. Paul, New J.
Physics 12, 053025, (2010).
[64] M. Opel, K.-W. Nielsen, S. Bauer, S. T. B. Goennenwein, J. C. Cezar, D.
Schmeisser, J. Simon, W. Mader, and R. Gross, Eur. Phys. J. B 63, 437, (2008).
[65] E. P. Wohlfarth, J. Appl. Phys. 39, 2, (1968).
[66] G. A. Gehring, Private discussion
100
Bcbfcial[pbs
[67] T. C. Kaspar, S. M. Heald, C. M. Wang, J. D. Bryan, T. Droubay, V. Shut-
thanandan, S. Thevuthasan, D. E. McCready, A. J. Kellock, D. R. Gamelin, S.
A. Chambers, Phys. Rev. Lett. 95, 217203, (2005).
[68] Y. Shapira and R. L. Krautz, Phys. Rev. B 10, 4781, (1974).
[69] C. Y. Park, S. G. Yoon, Y. H. Jo, S. C. Shin, Appl. Phys. Lett. 95, 122502, (2009).
101
Xhvpter K
mBrvy Vwsorption vny mvgnetix
xirxulvr yixhroism xhvrvxterisvtion
of XoByopey Inyium iin dfliye thin
films
ihis xhvptzr introyuxzs thz xonxzpts wzhiny flBrvy vwsorption vny flBrvy mvgnztix xirB
xulvr yixhroism vny yzsxriwzs somz of thz rzsults owtvinzy wy othzrs fiho hvvz vlso
invzstigvtzy yilutz mvgnztix ofliyz systzms using thzsz tzxhniquzsC dur rzsults owtvinzy
using Wzvmlinz IEK vt Divmony aight hourxz vrz thzn przszntzy xonxzntrvting mvinly
on vn vsByzpositzy JCI vtC% CoByopzy Iid thin filmC
102
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
KCF Introyuxtion
In chapter 5 we have established that the origin of magnetism in dilute magnetic ox-
ide systems is still an area which puzzles much of the scientific community. An ideal
intrinsic DMO exhibits ferromagnetism as a result of a coupling interaction between
localised moments at the substituted sites mediated by charge carriers of the host ox-
ide. The need for more rigorous characterisation of the DMO materials have also been
described in chapter 5.
Here we perform element specific measurements on Co-doped ITO(10) films using
x-ray magnetic circular dichroism (XMCD). The majority of previous element spe-
cific XMCD measurements at the Co aGPH edges have been performed on ZnF−xCoxO
thin films [1–5]. Our results here concentrate on the 5.4 at.% Co-doped ITO(10) film
characterised in chapter 5 and are the first set of results which describe the origins
of magnetism in TM-doped ITO using advanced element specific techniques such as
XMCD.
KCG mBgvy Vwsorption hpextrosxopy
X-ray Absorption Spectroscopy (XAS) is a technique used to probe the electronic
configuration of a specific element within a material. Scanning the photon energy over
the chosen element’s absorption edge results in a sharp increase in the x-ray absorption.
Electrons from an initial core state are promoted to the lowest unfilled state with
allowed symmetry, just above the Fermi level. After absorption, the atom relaxes
by one of two channels. The first involves the Auger process, where an intermediate
electron recombines with the core hole and the energy of the excitation is carried
away by the ejection of a secondary electron. The other channel is fluorescent photon
emission, the inverse of the absorption process, where the excited electron recombines
with the core hole and emits a photon.
Figure 6.1 shows the a-edge absorption spectra of Fe, Co and Ni in metal and
oxide form. This is a 2p → 3y transition with 2pK3yc and 2p53yc+F initial and final
states. a-edge absorption studies are best suited to probe magnetic properties which
are governed by the 3y valence electrons (as is the case for transition metals) for the
following reasons;
1. The dipole transition excites the 2p electron into the localised 3y shell, which
has the magnetic moment.
2. The absorption spectrum is dominated by dipole transitions from the core 2p
level to the empty 3y states which have a large Coloumb interaction between the
levels and therefore gives information on the local electronic structure.
3. Spectra can give information about the oxidation state and symmetry of the 3y
103
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
Figure 6.1: a-edge x-ray absorption edge spectra of Fe, Co and Ni in the form of
the elemental metals and as oxides. The two main structures are called the aH (lower
energy) and aG absorption edges. From reference [6].
transition metal ions.
4. The 2p and 3y core levels have high spin-orbit coupling resulting in very intense
spectral peaks.
5. The 2pHRG and 2pFRG spectral parts are clearly separated by the core-hole spin-
orbit interaction.
6. Core-hole lifetime broadening is small, resulting in sharp “multiplet” structures.
Metal spectra generally show two broad peaks which reflect the width of the empty y -
band states. The oxides on the other hand exhibit a fine structure, often referred to as
a “multiplet” structure. These empty oxide states are more localised than metal states.
A strong coupling interaction between the spin and orbital momentum of different 3y
valence electrons in the electronic ground state gives rise to a multitude of degenerate
final states. All final states which are resolvable show up as multiplets in the absorption
spectra.
KCH erinxiples wehiny mbXY
XMCD, the difference in core-level absorption between right (/+) and left (/−) cir-
cularly polarised x-rays (/XbXY = /− − /+), is an element and site specific probe of
magnetism [8]. Since the absorption of polarised light leads to spin selectivity, (/+)
or (/−) light preferentially probes the spin up and down states above the Fermi level.
104
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
Further, the XMCD sum rules [9] (see 6.3.1) can be applied to separate the contribu-
tion of spin and orbital contributions to the magnetic moment [10]. XMCD therefore
provides a direct probe of the nature of ferromagnetism in DMOs, in particular the
role that the TM ions play.
KCHCF drigin of the mbXY effext
The microscopic origin of the XMCD effect is explained by electronic transitions from
spin-orbit split initial 2p states to exchange-split 3y final states (Fig. 6.2). A model,
termed the “two step” model, proposed by Stohr and Wu [11] explains the origin of
XMCD at the aGPH edges of 3y transition metals. Firstly, the interaction of circularly
polarised x-rays with the p shell electrons leads to the excitation of spin-polarised
electrons with a spin and/or orbital momentum from a localised atomic inner shell. In
the second step, the 3y shell serves as the detector of the spin or orbital momentum of
the excited electron. For transition metals, the 3y final states are exchange split and
there is an imbalance in the number of available unoccupied 3y spin-up and spin-down
states. Therefore, the absorption of /+ and /− is different. In the case of non-magnetic
materials, the 3y band will not be exchange-split (i.e. equal population of spin-up
and spin-down) so that the transition intensities are identical for /+ and /−. In this
scenario there will be no XMCD effect. Therefore, we can conclude that to be able to
detect an XMCD effect at all, it is essential that the measured sample must have both
strong spin-orbit interactions (to clearly separate the 2pHRG and 2pFRG spectral parts)
and spin polarisation (y -band splitting).
The XMCD spectrum is calculated as the difference between the two measured
XAS (/+ and /−), i.e. /XbXY = /− − /+. Figure 6.3 (a) shows the typical absorption
spectra for left and right circularly polarised light at the Co aGPH edge for a Co thin
film [10]. It shows the Co 2p→3y transitions which results in two observable peaks
(aH and aG) due to the spin-orbit splitting of p-orbitals; aH corresponding to 2pHRG→3y
transition, and the aG corresponding to 2pFRG→3y transition. The XMCD spectrum
is opposite in sign at the aH and aG edge because of the opposite sign of the spin-
orbit coupling in the 2p states: 1+s for 2pHRG and 1-s for 2pFRG (the orbital angular
momentum number for a p orbital is equal to 1 and the total angular momentum
number, j is a combination of orbital and spin angular momenta, j = l + s). The
dichroism is given in Fig. 6.3 (b) which shows up with opposite signs at the two Co
absorption edges.
105
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
Electron
excitation
Circularly
polarised
X-rays
Fermi
level
L3
Unoccupied
3 levelsd
Occupied
3 levelsd
(2 )p
3/2
L2(2 )p3/2
EF
bandd
Figure 6.2: Schematic representation of the spin-dependent absorption process in the
“two-step” model.
KCHCG mbXY sum rules
The known spin polarisation of the 2p→3y transitions can be used as a probe to quan-
titatively measure the spin polarisation of the density of unoccupied valence states,
which is directly linked with the spin magnetic moment. Due to the opposite sign at
the aH and aG edges, the contributions from spin and orbital magnetic moments define
an orthonormal basis for a measured XMCD spectrum. It can therefore be used to de-
compose a signal into its constituent spin (µh) and orbital (µa) components, described
by the sum rules, and can be written as;
µa = −
4
∫
aH
/XbXYUω
3
∫
aH+aG
(/− + /+)Uω
· nh
ppol
= −4q
3r
· nh
ppol
(6.1)
and
µh = −
6
∫
aH
/XbXYUω − 4
∫
aH+aG
/XbXYUω∫
aH+aG
(/− + /+)Uω
· nh
ppol
+7 < iz S= −6p− 4q
r
· nh
ppol
+7 < iz S
(6.2)
where nh is the number of holes in the 3y band, and ppol is the degree of the polariza-
tion of the light (roughly equal to unity at Diamond Light Source synchrotron facility),
<i zS is the expectation value of the magnetic dipole operator (assumed to be zero in
the bulk of cubic crystals), and,
106
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
(b)
(c)
ρ+
ρ-
ρ+ - ρ-
∫ ρ+ - ρ-
ρ+ + ρ-
∫ ρ+ + ρ-
L3
L2
(a)
Figure 6.3: aGPH-edge XAS and XMCD spectra for Co under an applied magnetic field
of 0.05 T. (a) the XAS absorption spectra, (b) and (c) show the XMCD and summed
XAS spectra and their integrations (dotted lines) calculated from the spectra shown
in (a). p and q shown in (b) and the r shown in (c) are the three integrals required
for the sum rule analysis. After reference [10].
107
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
p =
∫
aH
/XbXYUω, q =
∫
aH+aG
/XbXYUω, r =
∫
aH+aG
(/− + /+)Uω (6.3)
which are clearly defined in Figs. 6.3 (b) and (c).
For a more comprehensive review of the XMCD processes, the reader is directed
towards more detailed reviews [10,12,13].
KCHCH Yetextion bethoys
The XAS measurements can be made in a variety of ways. For a detailed review of these
methods, the reader is referred to reference [14] (see Fig. 6.4). The results presented in
this thesis use the total electron (TEY) and fluorescence yield (FY) methods. In the
TEY method, the absorption spectra is obtained by monitoring all excited photoelec-
trons created during the absorption process. The electron yield is typically measured
directly using a channeltron electron multiplier, or indirectly as the photocurrent of
electrons flowing to the sample via the ground (the drain current). Since electrons
are only emitted from the surface of the sample this technique is a surface sensitive
technique only, with an escape or probing depth of the order of less than 10 nm and
provides information representative of the near-surface region [14]. Therefore, for thick
films, the spectra may not be representative of the bulk behaviour of the material. The
FY mode can probe depths of the order of hundreds of nm [14]. When photoelectrons
vacate the sample, positively charged holes are left behind. These holes are subse-
quently filled by Auger electrons from outer shells and x-ray fluorescence decay. The
fluorescence signal is detected using a photodiode detector located close to the sample
surface. TEY and FY modes of measurements are recorded simultaneously.
KCI erevious lork
As stated earlier, the majority of previous element specific XMCD measurements at
the Co aGPH edges have been performed on ZnF−xCoxO [1–5]. Kobayashi zt vlC [2] were
the first to investigate the role of the Co dopants in Co-doped ZnO concluding the
coexistence of two magnetic phases emanating from the Co ions: the first, a very small
fraction of ferromagnetic Co, and the second, strongly antiferromagnetically coupled
Co. The magnetic field dependence of the spin (b spin), orbital (b orwitvl) and total
(b tot) magnetic moments of Co (estimated using the XMCD sum rules) indicate that
there are both paramagnetic (signal which increases linearly with the applied field) and
ferromagnetic (signal which persists at zero applied field) contributions to the overall
XMCD spectrum (Fig. 6.5). The authors also conclude after some careful analytical
subtractions that the XMCD line shape of both the paramagnetic and ferromagnetic
components show similar multiplet structures, very different to that of pure Co metal,
108
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
I 0
I PEY
I FY
I TEY I R
I T
e -
e -
g
0
g
0
g
0
g
f
Figure 6.4: Experimental geometries showing five different methods for measuring
XAS and XMCD spectra, where I E is the incident radiation, I i is the transmitted
flux, Ig the reflected flux, I iEn the total electron yield, I Fn the fluorescence yield and
I eEn the partial electron yield. Adapted from reference [14].
therefore indicating that only a fraction of the doped transition metal ions are ferrro-
magnetic.
Gacic zt vlC performed a more detailed investigation into 5 at.% Co-doped
ZnO and undoped ZnO [3]. As in the investigations performed by Kobayashi zt vlC,
multiplet-like spectral features were observed in the XAS and XMCD spectra, however
the dependence on temperature was very different. In the case of Gacic zt vl. the
XMCD signal and moment calculated via the application of the sum rules reduced
drastically as a function of increasing temperature [Fig. 6.6 (b)] resembling behaviour
typical of a paramagnet and suggested a weak interaction between the Co ions. The
field dependence of the calculated moment (ms) gives no indication of ferromagnetic Co
either as they see a perfectly linear dependence with no sign of remanent magnetisation.
These results therefore contradict the findings of Kobayashi zt vl. with no evidence
for any antiferromagnetic interaction between the dispersed Co ions. Measurements
at the Zn aGPH edge confirmed that no magnetic moment was present on the Zn ions.
However, these results were only made in TEY mode, and therefore only describe the
properties at the very near-surface region.
Barla zt vl. furthered this work by performing XAS and XMCD measurements in
both TEY and FY modes, therefore probing the bulk of the sample, in ZnEONECoEOFEO
and ZnEOL5CoEOG5O thin films [4]. They confirm the results of Gacic zt vl. showing
the purely paramagnetic nature of the Co dopants down to 2 K with no evidence of
ferromagnetic metallic clusters. From these findings, both Gacic and Barla exclude
transition metal Co dopants as the sole source of room temperature magnetism, and
instead suggest a defect- or oxygen-related mechanism, similar to that suggested for
109
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
Figure 6.5: Magnetic field and temperature dependences of the Co aGPH XMCD spectra
of ZnEON5CoEOE5O. (a) XMCD spectra under different magnetic fields at 20 K. Average
magnetic moments b spin, b orw, and b tot as functions of magnetic field (b) and tem-
perature (c), estimated using the XMCD sum rules, (d) Magnetization curves measured
by a SQUID magnetometer. After reference [2].
Figure 6.6: (a) XAS (top panel) and XMCD (bottom panel) spectra at the Co aGPH
edges of a 5% Co doped ZnO thin film sample, (b) Field-dependent Co XMCDmagnetic
moments for three different temperatures. The inset shows the temperature dependent
Co XMCD magnetisation and a corresponding Curie-Weiss law fit. After reference [3].
110
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
Figure 6.7: (a) XMCD magnetisation loops measured at the Co aH edge for
ZnEON5CoEOE5O (red), ZnEOL5CoEOG5O (blue), (b) XMCD spectra measured at the Co K
edge for ZnEON5CoEOE5O (red), ZnEOL5CoEOG5O (blue), and Co metallic clusters (black) em-
bedded in an alumina matrix. FY spectra were recorded at T = 10 K and a 6 T applied
field. After reference [5].
undoped, ferromagnetic oxides like HfOG [15].
The most comprehensive study of transition-metal doped ZnO using XAS and
XMCD characterisation techniques was performed by Tietze zt vl. [1]. In addition
to the focus at the Co aGPH edge, measurements were also performed at the oxygen
K -edge. Once again, only paramagnetic Co was found both in TEY and FY modes,
and in all cases only spectra consistent with a multiplet description of Co yL configu-
ration in tetrahedral coordination were observed. The measurements at the O K -edge
showed no sign of magnetisation. Having not found any element-specific signature of
ferromagnetism, Tietze zt vl. suggest oxygen vacancies as a possible mediator for fer-
romagnetism in these materials.
Each of the experimental results described above show that within the dilute limit
of Co in ZnO, the Co dopants successfully substitute for Zn lattice sites. Theoretical
simulations performed to fit many of the XMCD spectra at the Co aGPH edge suggested
that Co is tetrahedrally coordinated within the matrix in the 3yL state.
Rode zt vl. investigated the ferromagnetic Co-doped ZnO system where low (5
%) and high (25 %) Co doping was used [5]. Figure 6.7 summarises the main findings
of the report. Here we see that XMCD magnetisation loops at the Co aGPH edge reveal
that in the lower doping regime, the magnetisation behaviour is paramagnetic (com-
pounded by the multiplet XAS and XMCD line structure which implies ionic Co),
whereas in the higher doping regime, it is clearly ferromagnetic with magnetisation
persisting at remanence [Fig. 6.7 (a)]. In fact, the magnetisation loop in the highly
doped regime also gives a paramagnetic component at higher fields (above 1 T) as is
seen in the lower doping regime. This was confirmed by simply performing XAS and
XMCD measurements at low (remanence) and high fields. Not only did the XMCD
signal persist at remanence, but it also gave a different line shape, one that resembled
that of pure Co metal, confirming that the origin of magnetism in the highly doped
111
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
Figure 6.8: (a) Co aGPH edge XMCD spectra recorded in TEY mode at 10 K with an
applied magnetic field of 3 T for films grown at 10−L (red) and 10−I (blue) mbar. The
dotted black line shows a reference of metallic cobalt, (b) XMCD spectra in TEY mode
for a sample grown at 10−K mbar as a function of temperature with an applied magnetic
field of up to 3 T. The inset simply provides an expanded view of the Co aH edge as a
function of temperature, (c) XMCD magnetisation loop measured in TEY mode at Co
aH edge for a high oxygen pressure film (left) and for a low oxygen pressure film (right)
recorded for an angle between the applied magnetic field and the sample surface of 90◦
and 20◦ respectively. The applied magnetic field and the photon wavevector are both
along the black arrow. After reference [7].
samples was metallic Co clustering. Taking the investigation a step further, XAS and
XMCD measurements were performed at 10 K in a 6 T field at the Co K -edge, there-
fore probing the 1s→4p transitions, in FY mode. In each scenario, both low and high
doping regimes, two clear dichroic peaks were observed, one corresponding to CoG+
ions and the other to metallic Co [see Fig. 6.7 (b)]. This was a hugely significant result
and proved the coexistence of magnetic phases, even in the low doping regime sample.
Extensive investigations into the Co-doped (La, Sr)TiOH (LSTO) system via XAS
and XMCD measurements have been performed by Copie zt vl. [7]. Two samples with
identical Co doping level (2 %) were grown in different oxygen partial pressures: one
high (10−I mbar) and one low (10−L mbar). Like Rode zt vl. they too claim evidence
of mixed phases of ionic CoG+ and metallic Co clustering, particularly in the sample
grown in low oxygen pressure. This is seen clearly in Fig. 6.8 (a) (solid red line) where
the sharp multiplet line features are beginning to disappear once the oxygen partial
pressure is reduced, and the XMCD spectrum starts to resemble that of pure metallic
Co (dotted black line). Further evidence for metallic Co clustering was found when
analysing the temperature dependence of the XMCD line structure [Fig. 6.8 (b)]. Here
the authors conclude that the progressive vanishing of the fine multiplet structure is
indicative of poor ionic behaviour.
112
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
KC5 Zflperimentvl gesults
In order to probe the magnetic nature of Co-doped ITO in greater detail, XAS and
XMCD measurements were performed at the In and Sn b GPH, Co aGPH- and oxygen
K -edges on an as-deposited 5.4 at.% Co-doped ITO(10) sample (described earlier in
chapter 5). The sample surface orientation was set at 45◦ to the incident beam. The
magnetic field, in all XMCD measurements here, was applied in a direction parallel to
the incident x-ray beam. The XAS spectra were recorded in both TEY and FY modes
in order to measure the surface and bulk components, respectively.
KC5CF In vny hn M O;P eyges
In order to investigate the possibility of magnetic polarisation at the In and Sn lattice
sites, XAS and XMCD spectra were recorded at the In and Sn b GPH edges. Both In
H+
and SnI+ have an electronic configuration of 4yFE and we do not expect any magnetic
moment to be present at all. Figures 6.9 (a) and 6.9 (b) show the XAS and XMCD
spectra measured in TEY mode at the In and Snb G edges respectively at 2 K under an
applied magnetic field of 4 T. No evidence of dichroic behavior at the In or Snb G edges
can be seen within the detection limits of the instrument. This excludes the presence of
any detectable ferromagnetic polarisation of the In and Sn s and y states. The same be-
havior was observed at both edges at 300 K also under an applied magnetic field of 4 T.
KC5CG Xo LO;P eyge
The XAS and XMCD spectra measured at 2 K in a 4 T field at the Co aGPH edge
using TEY are shown in Fig. 6.10. The absorption edges appear due to a Co 2p−→3y
transition. This transition provides clues on the density of unoccupied Co 3y states,
therefore probing the 3y magnetism of Co. A large XMCD signal is present at the
aH edge, indicating a significant magnetic moment at the Co
G+ sites in the applied
magnetic field. The XMCD asymmetries [(/+−/−)/(/++/−)] at 2 K were estimated
as 5.5% and 14.2% respectively for TEY and FY spectra. TEY is generally lower than
FY and is consistent with the fact that there is a lower volume fraction of Co dopants
and magnetic coupling near the surface region of the film. The pronounced multiplet
structure at the Co aGPH edges is typical of Co in a localised state. By comparing the
XAS and XMCD spectra to those of metallic Co [10] (see Fig. 6.11), we conclude that
Co clusters are absent in the surface region. The multiplet features observed are very
similar to those reported in other Co-doped oxide systems [1–5,7,16,17]. Furthermore,
the fine structure of the XMCD spectra appears to be maintained as the temperature
approaches room temperature and is shown in Fig. 6.10 (c). This is contrary to that
observed in Co-doped LSTO where it was suggested that the gradual loss in fine struc-
113
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
652 656 660 664 668 672 676
-6.0x10
-4
-3.0x10
-4
0.0
3.0x10
-4
6.0x10
-4
0.620
0.625
0.630
0.635
0.640
712 716 720 724
-1.0x10
-4
0.0
1.0x10
-4
0.618
0.624
0.630
0.636
Sn M
2
(b)
 
 
 
X
A
S
 
&
 
X
M
C
D
 
i
n
t
e
n
s
i
t
y
 
(
a
r
b
.
 
u
n
i
t
s
)
(a)
In M
2
 
 
Photon energy (eV)
 
 
 
 
  
 
Figure 6.9: XAS and XMCD spectra in TEY mode at 2 K under an applied field of 4
T at the (a) In b G edge and (b) Sn b G edges, along with the raw XMCD data (solid
blue circles) and a smoothed extrapolation (solid blue lines).
ture could be interpreted as a signature of poor ionic behavior and could indicate the
presence of metallic Co clusters [7].
XAS recorded using FY mode [inset, Fig. 6.10 (a)] also exhibits pronounced
multiplet features implying that the bulk of the film is also free from clustering. Fig.
6.12 (a) and (b) display the XAS and XMCD spectra for a 5.4 at.% Co-doped ITO
film measured in FY mode at both 2 and 300 K. Multiplet features are visible in both
cases, however, perhaps more so at 300 K. Even at 2 K, the XMCD spectrum does
not coincide with that of pure metallic Co. In addition, the fact that the aH peak
becomes increasingly multiplet-like with temperature implies that the samples are not
superparamagnetic in any way. The observation of multiplet features in the FY XMCD
spectra is hugely significant as it suggests that the bulk of the film is free from metal
Co inclusions.
To corroborate the experimental results, the XAS and XMCD spectra were cal-
culated (by Geerit van der Laan of Diamond Light Source) for the octahedral site
symmetry of the Co ion (assuming Co substitutes for InH+). Figure 6.10 (b) shows
the calculated results, which are in excellent agreement with the experimental data.
The spectra show the electric-dipole allowed transitions between the ground-state Co
3yL t5Ggz
G
g (
Ii Fg) and the final-state 2p
53yM configuration in octahedral symmetry with
crystal field parameter 10D q = 0.7 eV split by first-order spin-orbit interaction and
in the presence of a magnetic exchange field of gµB] = 1 meV, where g is the spec-
troscopic splitting factor (set as 2.0023 for free electrons), µB is the Bohr magneton
114
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
0.00
0.05
0.10
0.15
0.20
768 772 776 780 784 788 792
-0.10
-0.05
0.00
0.05
0.10
0.15
0.20
772 776 780
0.12
0.16
0.20
0.24
772 776 780
-0.10
-0.08
-0.06
-0.04
-0.02
0.00
0.02
 
 
X
M
C
D
 
(
a
r
b
.
 
u
n
i
t
s
)
 
I
n
t
e
n
s
i
t
y
 
(
a
r
b
.
 
u
n
i
t
s
)
  
 
(c)
(b)
(a)
 
 XAS ( ) 2
 XAS  Calculated
 XMCD ( )
 XMCD Calculated
  
 
 
 
X
A
S
 
(
a
r
b
.
 
u
n
i
t
s
)
  
 
 
Photon Energy (eV)
 2 K
 100 K
 200 K
 300 K
Photon Energy (eV)
Figure 6.10: Co aGPH edge for a 5.4 at. % Co-doped ITO film at 2 K under an applied
magnetic field of 4 T; (a) TEY XAS for right (/+) and left (/−) circularly polarised
light. Inset: FY XAS spectra for right (/+) and left (/−) circularly polarised light
under the same conditions, (b) average experimental TEY XAS and the calculated
XAS spectra along with the experimental XMCD spectrum and calculated XMCD
spectrum for octahedral CoG+ (10D q = 0.70 eV), (c) TEY XMCD spectra at the Co aG
edge of a 5.4 at. % Co-doped ITO film as a function of temperature under an applied
magnetic field of 4 T.
115
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
770 780 790 800 810
-1.00
-0.75
-0.50
-0.25
0.00
0.25
772 776 780 784
-1.0
-0.5
0.0
N
o
r
m
a
l
i
s
e
d
 
X
M
C
D
 
i
n
t
e
n
s
i
t
y
 
(
a
r
b
.
 
u
n
i
t
s
)
Photon energy (eV)
3
 Co foil
 Co:ITO
Photon energy (eV)
L
Figure 6.11: Comparison of the TEY XMCD spectra of a Co metal foil (open violet
circles) and a 5.4 at.% Co-doped sample (closed black circles) at 2 K under an applied
magnetic field of 4 T. Inset: an expanded view of the Co aH edge for each sample.
Figure 6.12: XAS and XMCD in FY mode at the Co aGPH edge for a 5.4 at. %
Co-doped ITO film at (a) 2 K and (b) 300 K under an applied magnetic field of 4 T.
116
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
and ] is the applied magnetic field. The wave functions of ground and final states
were calculated in intermediate coupling using Cowan’s Hartree-Fock (HF) code with
relativistic correction following the method described in Ref. [18]. The Slater and
spin-orbit parameters were as tabulated in Refs. [18, 19]. Interatomic screening and
mixing was taken into account by reducing the atomic values of the Slater integrals
[ k(3y,3y), [ k(2p,3y), and Gk(2p,3y) to 70%, 80%, and 65%, respectively. The 2p
spin-orbit interaction was scaled to 97% of the HF value. The calculated line spectra
were broadened by a Lorentzian of Γ = 0.10 (0.40) eV for the aH (aG) edge to account
for intrinsic line width broadening and a Gaussian of σ = 0.20 eV for the instrumental
broadening.
Element-specific magnetisation curves were recorded using XMCD in TEY and
FY mode at 2, 100, 200 and 300 K by sweeping the magnetic field and maintaining
the incident photon energy at 777.3 eV. Figure 6.13 (b) shows the magnetisation curve
measured in FY mode at 2 K. Note that this measurement does not give an abso-
lute value for the moment on an individual CoG+ ion, but provides an indication of
the field dependence of the Co moment in 5.4 at.% Co-doped ITO. Similar results
were obtained in TEY mode, confirming that we observe no difference in the magnetic
properties of the surface and throughout the bulk of the sample. The curve can be
modelled well with a Brillouin function (Eqn. 6.4) to describe paramagnetic behavior,
where y=gµBJ] /kBi and ms is the saturation magnetisation. The curve shows no
ferromagnetic or remanent behavior for the CoG+ ions at any of the given temperatures.
The fit yields a total angular momentum, J, of 0.761 and an effective total (recall that
J = a + h ) magnetic moment ptotvl = g [J (J+1)]
FRG of 2.32µB/Co.
MXbXY = ms
[(
2J + 1
2J
)
coth
[
(2J + 1)y
2J
]
−
(
1
2J
)
coth
( y
2J
)]
(6.4)
Theoretically, h = 3/2, a = 3 and therefore J = 9/2 for CoG+ in the 3yL configuration.
This yields a calculated theoretical total moment of ptotvl = 6.63µB. The value we find
experimentally is much lower, however is within reasonable agreement with the spin-
only value (3.87µB, calculated using pspin = 2[h (h+1)]
FRG) expected for paramagnetic
Co ions in the upper spin state [20]. Therefore, it is clear that the spin-only values are
in much better agreement with experiment than the values calculated using the total
angular momentum. This phenomenon is known as “quenching” of the orbital angular
momentum (a is reduced to zero), and is a result of the electric field (crystal field)
generated by the surrounding ions in the solid.
A more complete plot of the XMCD magnetisation curves is shown in Fig. 6.14,
where data is plotted at each of the temperature intervals along with their respective
Brillouin fits. The plot supports the data presented in Fig. 6.13 (c) where the XMCD
magnetisation decreases as the temperature increases.
Fig 6.13 (a) shows the respective XAS and XMCD spectra in FY mode at the Co
aGPH edge at 300 K in the remanent state (0 T). The sample was first magnetised at 4 T
117
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
0.00
0.05
0.10
0.15
0.20
0.25
768 772 776 780 784 788 792 796 800
-0.04
0.00
0.04
0 50 100 150 200 250 300
0.00
0.02
0.04
0.06
0.08
0.10
-4 -2 0 2 4
-0.08
-0.04
0.00
0.04
0.08
 Photon energy (eV)
 
 
I
n
t
e
n
s
i
t
y
 
(
a
r
b
.
 
u
n
i
t
s
)
 
XMCD (
 
 
 
 
 T (K)
 
(b)
 Experimental
 Curie-Weiss fit
(c)
 Experimental
 Brillouin fit
X
M
C
D
 
(
a
r
b
.
 
u
n
i
t
s
)
 
 
 
0
H (T)
X
M
C
D
 
(
a
r
b
.
 
u
n
i
t
s
)
(a)
Figure 6.13: (a) XAS and XMCD spectra at the Co aGPH edge in FY mode at remanence
(0 T) and 300 K. The XMCD is shown in the lower panel (solid blue circle with solid
blue line), (b) Magnetisation curve of a 5.4 at. % Co-doped ITO measured at the Co
aGPH edge in FY mode at 2 K (solid black circles), along with the fit to a Brillouin
function (solid red line), and (c) temperature dependence of the FY XMCD signal at
an applied magnetic field of 4 T (solid black circles), along with the fit to a modified
Curie-Weiss law (solid red line).
118
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
-4 -2 0 2 4
-0.08
-0.04
0.00
0.04
0.08
 2K
 100K
 200K
 300K
 
 
 
0
H (T)
X
M
C
D
 
(
a
r
b
.
 
u
n
i
t
s
)
Figure 6.14: XMCD magnetisation curve for a 5.4 at. % Co-doped ITO measured
at the Co aGPH edge in FY mode at 2, 100, 200 and 300 K respectively, along with
corresponding fits to a Brillouin function (solid and dotted lines).
before being ramped down to 0 T for measuring. No XMCD signal could be detected
which is consistent with the XMCD magnetisation curve discussed above, where no
remanent XMCD effect is expected. The same was seen in the TEY measurements.
Therefore, we can conclude that the specific contribution of Co 3y ions to the total
bulk ferromagnetism measured in 5.4 at.% Co-doped ITO is purely paramagnetic, i.e.
we can assume that the Co ions are not exchange-coupled with each other.
The temperature dependence of the XMCD measured in FY for the CoG+ ions in
an applied field of 4 T is shown in Fig. 6.13 (c), and follows a modified Curie-Weiss
law. From these findings, we conclude that there is strong evidence that the Co ions
are not directly responsible for the ferromagnetism observed in the film.
Copie zt vlC also observed a similar “s-like” shaped XMCD magnetisation curve,
much like the data presented in Fig. 6.13 (b) [7]. They, however, observe some
anisotropy and claim that this may indicate ferromagnetic behaviour regardless of
the fact that there is no hysteresis. Together with the metallic-like Co XMCD spec-
tra, they suggest a non-negligible contribution to the magnetic signal arising from Co
clusters in a metallic state. Rode zt vlC also measured a closed XMCD magnetisation
loop in their ZnEON5CoEOE5O film, but analysis at the Co K -edge revealed the coexis-
tence of two magnetic phases: a paramagnetic phase associated with ionic Co and an
extrinsic ferromagnetic phase associated with Co metal clusters [5]. We believe that
the preservation of the multiplet line structure in the XMCD spectra and paramagnetic
dependence of the XMCD signal over the entire temperature range up to 300 K in our
sample is good evidence to suggest that within the detection limits of our instruments,
119
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
metallic and ferromagnetic or superparamagnetic Co clusters are not present within
this system.
Quantitative information regarding µa and µh components to the total magnetic
moment (µXbXY) were calculated by applying the sum rules [9, 10], where we have
included the magnetic dipole term <µiS. Their dependence on temperature shown in
Figs. 6.15 (a) and (b). We have assumed a pure 3yL configuration for the Co atoms,
which sets the number of holes in the 3y shell to 3. Values for µXbXY are taken at
an applied magnetic field of 4 T in TEY mode. The magnitude of µa and µh are
relatively high in comparison to some reports with comparable levels of Co doping in
ZnO [3, 4]. This could be due to the different crystal field effects on the CoG+ ions in
the ITO matrix. The value for µXbXY at 2 K (2.32 µB/Co) calculated from the sum
rules is in excellent agreement with that estimated using the Brillouin fit in Fig. 6.13
(b). In addition to this, the ratio µa/µh of the experimental data (0.31) is in excellent
agreement with that extracted from the theoretical calculations (0.29). This is further
evidence for Co being in a 2+ state and substituting for In in the ITO matrix.
The data recorded in FY mode are qualitatively similar. However, due to self
absorption effects which affect the relative intensity of aGPH peak, along with the dif-
ferences in the matrix elements and selections rules for TEY and FY, application of
the XMCD sum rules is meaningless here. They do however, qualitatively confirm the
absence of Co clusters throughout the bulk of the film and the decreasing magnetic
moment with increasing temperature.
KC5CH hummvry
We have studied the origin of magnetic ordering in the Co-doped ITO system. The
line shape of both the XAS and XMCD spectra we obtain show clear multiplet features
over a wide range of temperatures suggesting that we have Co existing in an ionic state.
The absence of dichroism at remanence at the Co aGPH edge and the modelling of the
XMCD magnetisation loops using a Brillouin function points towards paramagnetic
behaviour thus excluding their significant contribution to the ferromagnetism. This is
supported further by the clear difference in comparison to the SQUID magnetometry
measurements. In contrast to the XMCD magnetisation loop at 2 K, the magnetic
hysteresis loop [see Fig. 5.13 (a)] shows a curve with saturation field of approximately
0.40 - 0.50 T with a large coercive field. The XMCD magnetisation curve only shows
signs of saturating beyond our measurement range of 4 T. This clearly shows that the
two techniques are probing different phenomena and at the same time complementing
each other in trying to determine the origins of the ferromagnetism in our samples.
These results suggest that the ferromagnetism we observe in Co-doped ITO is not
related to the 3y electronic states. This is rather contradictory to theoretical models
used to explain the magnetic properties in the DMO systems. Like those before, we be-
120
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
mS
(m
B
/C
o
)
mL
M
a
g
n
e
ti
c
m
o
m
e
n
t
Figure 6.15: (a) Temperature dependence of calculated total magnetic moment
(µh+µa), µXbXY, using normalised TEY XMCD spectrum and the application of
the XMCD sum rules, (b) Temperature dependence of the orbital (µa) and spin (µh)
magnetic moments moments. The dotted lines are a guide to the eye.
121
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
lieve that these findings may point towards the likelihood that the bulk ferromagnetism
observed from the SQUID magnetometry measurements may be oxygen-mediated, e.g.,
oxygen vacancies or oxygen-induced defects. Oxygen K -edge XAS data would certainly
be capable of providing useful information on whether the features could be signatures
of oxygen related lattice defects. Certain atomic defects in structures, such as oxygen
vacancies with trapped electrons or holes, may contain low-lying triplet states that are
capable of overlapping to form an impurity band [22]. These magnetic defects may
interact through long-range interactions mediated by the impurity band to create a
ferromagnetic ground state. Therefore, further studies are required to definitively ob-
serve dichroism at oxygen sites at remanence (see section 6.7). This would confirm the
presence of a spontaneous ferromagnetic contribution.
KCK Influenxe of vnnevling O htuyies vt the Xo L2,3
eyge
Annealing of the Co-doped ITO films significantly changed the magnetic properties,
and was described previously in chapter 5. Differences were also observed in the line
shapes of the XAS and XMCD line spectra as shown in Fig. 6.16 where we present the
TEY XAS and XMCD spectra for for a pure Co foil (top), and as-deposited (middle)
and annealed (in UHV for 2 hours at 700◦C) (bottom) 5.4 at.% Co-doped ITO films
at 2 K under an applied magnetic field of 4 T. After annealing at 700◦C the multiplet
features at the aH peak are less pronounced. Figure 6.16 (b) shows the corresponding
XMCD spectra. The annealed sample looks to have features which are in between that
of metallic and ionic Co. This suggests that we may have a mixture of metallic and
ionic Co present at the near-surface region of the sample after annealing.
Confirmation of the presence of metallic Co was obtained from the XAS and
XMCD spectra measured in FY mode and is shown in Fig. 6.17. Comparison of
the XMCD spectra for the as-deposited and annealed samples clearly shows that the
multiplet structures have been almost completely smeared out after annealing. The
line shape of the spectrum now resembles that of metallic Co and therefore suggests
that the bulk of the film contains metallic Co. This is consistent with the observation
of a superparamagnetic iB shown in Fig. 5.18. The arrows in Fig. 6.17 (a) indicate
the emergence of two new features in the XAS spectrum which only appear after the
annealing treatment. These new features may indicate different hybridisation of Co
states to oxygen or In after the annealing process.
To probe the origins of the magnetism in the annealed sample further, XAS and
XMCD measurements were performed at remanence. The multiplet fine structure were
seen in both TEY and FY XAS spectra, however, no XMCD signal was detected in
TEY. A very weak dichroic effect was observed in the FY spectra but was plagued
122
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
(b)(a)
Figure 6.16: TEY XAS (a) and XMCD (b) spectra for Co metal (top), as-deposited
(middle) and annealed (bottom) 5.4 at.% Co-doped ITO at 2 K in an applied magnetic
field of 4 T.
(a)
(b)L3
Figure 6.17: (a) Normalised FY average XAS and XMCD spectra for an as-deposited
and annealed 5.4 at.% Co-doped ITO sample at 2 K in an applied magnetic field of 4
T; (b) an expanded view of the Co aH peak for the each sample.
123
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
Table 6.1: Summary of the spin (µh) and orbital (µa) components to the total magnetic
moment (µXbXY) at 2 K in an applied magnetic field of 4 T calculated using the sum
rules.
Sample Number of holes µh µa µXbXY
(µB/Co) (µB/Co) (µB/Co)
Co foil 2.49 1.55±0.11 0.14±0.04 1.69±0.12
As-deposited 3.00 1.76±0.09 0.56±0.03 2.32±0.09
Annealed 3.00 1.72±0.12 0.41±0.04 2.13±0.13
with noise so that it was difficult to determine whether the multiplet features were
present or not. The observation of a remanent signal is indicative of a ferromagnetic
component. In the as-deposited samples, it was shown that there was no dichroism in
TEY or FY at remanence confirming the paramagnetic nature of the CoG+ ions (recall
Figs. 6.13 (b) and 6.14).
Applying the XMCD sum rules to the TEY spectra at 2 K and 4 T for the annealed
sample gives a µXbXY of 2.13±0.10 µB/Co which is lower than that for the as-deposited
sample (2.32 µB/Co). Table 6.1 summarises the results for each sample using the sum
rule analysis. For the Co foil the number of holes used for our calculations was 2.49
(as used in ref. [10]), but in the case of the Co-doped samples, the number of holes
was set to 3. Our value for the total moment for the Co foil is in excellent agreement
with that of the theoretical value for Co metal (1.72 µB). Within the error estimations
it is evident that the as-deposited and annealed samples have a similar spin moment,
but there is a greater reduction in the orbital moment for the annealed sample. The
quenching in the orbital moment may be due to a lowering in the degeneracy of the
states arising from the change in crystal field around the Co sites due to the clustering
of the Co dopants. As we mentioned in section 6.5.2, the estimated total moments from
the sum rules are lower than that expected for CoG+ in the upper spin state. Rode zt
vl. also observed similar reductions in the sum rule calculated magnetic moments [5].
They suggest that the origins of the reduction are difficult to explain but that it may be
due to factors such as the lack of magnetic saturation, antiferromagnetic interactions
between nearest neighbor CoG+ ions, and even the presence of Co metal clusters which
have a lower Co magnetic moment (1.72 µB).
The rise in magnetisation upon UHV annealing can therefore be largely ascribed
to the creation of metallic Co clusters.
124
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
KCL dflygen K Beyge
Nefedov zt vlC showed that from XMCD magnetisation curves in Co-doped TiOG that
the O K -edge showed clear hysteresis with switching fields identical to those recorded
at the Co aGPH edge [23]. This result implied that the oxygen atoms in close proximity
to the Co atoms become spin polarised, even in samples where Co clustering was not
detected. This result suggested that oxygen polarisation can play an important role
in the existence of long range magnetic order in TM-doped oxide systems. Thakur zt
vlC have also observed an appreciable XMCD signal in the O 2p region in undoped
MoOG thin films which also exhibit room temperature ferromagnetic properties [24].
This was attributed to a strong hybridisation between the O 2p - Mo 4y levels and
thus a charge transfer from the O 2p orbitals to the Mo 4y orbitals.
The measurement sample (20 - 30 averages) for our O K -edge studies was very
large in order to account for the fact that dichroic effects were expected to be very low
compared to that at the Co aGPH edge. Earlier we showed the paramagnetic behaviour
at the Co a-edge, and therefore the absence of Co 3y polarisation at remanence [Fig.
6.13 (a)]. Figure 6.18 shows the XAS and XMCD spectra in TEY mode for a 5.4 at.%
Co-doped ITO film at 2 K after applied magnetic fields of ±4 T at the O K -edge.
A very pronounced peak in the XAS spectra at approximately 529.60 eV is observed,
extremely similar to that seen by Abbamonte zt vlC in LaG−xBaxCuOI [25]. A sizeable
XMCD signal, with line shape features similar to others who report dichroism at the
O K -edge is produced.
XMCD requires the breaking of time-reversal symmetry and the presence of a
spin-orbit interaction. The XMCD signal is due to electric-dipole and, in some cases,
electric-quadrupole transitions between a deep core state and an empty conduction
state. However, in some cases it is possible that one can observe another effect, largely
due to the interference between electric-dipole and electric-quadrupole transitions. This
gives a dichroic signal with a very similar line shape to that of an XMCD dichroism,
however, its intensity is much smaller in comparison. One way to determine whether
these effects are magnetic in origin is to see whether the dichroic signal “flips” when
the applied magnetic field has been reversed (but the polarisation of the x-rays remain
constant). Also, by using a normal incidence geometry, we can exclude non-XMCD
effects such as linear dichroic contribution, ztx., in order to determine whether the
features we see are truly magnetic dichroic effects or not. As can be seen from Fig.
6.18, the XMCD signal flips on reversal of the applied magnetic field.
We believe that the magnetic polarisation of the O 2p levels is a result of the
intense hybridisation between the Co and O orbitals. As in the case of Co-doped TiOG,
the oxygen atoms that are in close proximity of the Co atoms become polarised upon
application of a magnetic field [23]. The XMCD magnetisation loops in TEY mode at 2
K for positive (/+) and negative (/−) polarisation are shown in Fig. 6.19. The slope of
125
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
0.35
0.40
0.45
0.50
0.55
0.60
0.65
 
 
 
X
A
S
 
i
n
t
e
n
s
i
t
y
 
(
a
r
b
.
 
u
n
i
t
s
)
528 532 536 540 544 548
-0.3
-0.2
-0.1
0.0
0.1
0.2
 
 + 4T
 -  4T
X
M
C
D
 
i
n
t
e
n
s
i
t
y
 
(
a
r
b
.
 
u
n
i
t
s
)
Photon energy (eV)
 20
Figure 6.18: TEY XAS and XMCD spectra at the O K -edge at 2 K under applied
magnetic fields of 4 and -4 T respectively.
the magnetisation curve flips upon reversal of the polarisation as expected. It is clear
that the dependence of the XMCD signal at the O K -edge is paramagnetic, as is the
case for the Co. This is further evidence that the Co and O orbitals are overlapping.
KCM hummvry vny Xonxlusions
In summary, we have used XAS and XMCD characterisation at the In b GPH, Sn b GPH,
Co aGPH and O K -edges in Co-doped ITO to study the origin of the magnetism over a
range of temperatures and applied magnetic fields. Element specific hysteresis loops
recorded at the Co aGPH edge indicate a clear paramagnetic contribution to the overall
magnetic moment of the Co-doped ITO film and confirm the presence of non-coupled
CoG+ ions substituting for the In ions. Multiplet features in the Co aGPH edge XAS mea-
sured in both TEY and FY mode from 2 to 300 K indicate that the Co atoms are in
a divalent state, therefore non-metallic in nature throughout the entire film, and thus
substituting for the In ions. Therefore, within the detection limits of the techniques,
and from all the information we have through our exhaustive characterisation efforts,
we believe that there is no contribution to the observed ferromagnetism solely from the
Co dopants. No XMCD could be detected at the In or Sn b G edges, thus excluding
the presence of a large magnetic polarisation of the In and Sn sublattices. Therefore
126
Cb[pnel 0. X-l[s Abmilpncih [hd g[ahenc] ]cl]of[l dc]blicmg
]b[l[]nelcm[ncih if Ci-diped Chdcog Tch Ircde nbch fcfgm
-4 -2 0 2 4
-0.015
-0.010
-0.005
0.000
0.005
0.010
0.015
 Linear fit
 Linear fit
 
 
N
o
r
m
a
l
i
z
e
d
 
T
E
Y
 
X
M
C
D
 
(
a
r
b
.
u
i
n
t
s
)
0
H (T)
Figure 6.19: TEY XMCD magnetisation curve at the O K -edge in positive (/+) and
negative (/−) polarization at 2 K.
the ferromagnetic moment observed in the Co-doped ITO films cannot be ascribed to
the Co and In sublattices.
Upon UHV annealing, XAS and XMCD in both TEY and FY modes display fea-
tures which suggest a mixed phase of both CoG+ ions and metallic Co. This confirms
the results from annealing in chapter 5 where it was believed that Co metal segregation
had taken place leading to an enhanced ferromagnetic response.
Magnetic dichroism was also observed at the O K -edge at 2 K, indicative of spin
polarisation of oxygen atoms in the ITO host matrix. Not enough detailed experiments
were performed to investigate whether there was any remanent magnetisation on the
oxygen atoms, due to the experimental difficulty to confirm the presence of a genuine
dichroic signal in the first place. However, for now the magnetic dichroism at the O
K -edge can be attributed to the strong hybridisation between Co 3y and oxygen 2p
orbitals.
So, what is the source of magnetism in these films if no remanent magnetisation is
measured at the In, Sn, Co or O absorption edges? Through the process of elimination,
what remains is atomic defects (either TM- or O-induced) such as interstitial defects,
oxygen vacancies, ztx.
On this evidence, the Co dopants alone are not responsible for the ferromagnetic
response in these films. The only way this issue can be resolved would be to perform
XMCD measurements at the Co K -edge, probing deeper energy levels, as was carried
out by Rode zt vlC [5].
In the next chapter however, we use optical MCD which can provide useful infor-
mation on which energy states in the solid may be involved in the magnetism.
127
Wiwliogrvphy
[1] T. Tietze, M. Gacic, G. Schutz, G. Jakob, S. Bruck, E. Goering, New J. Phys. 10,
055009, (2008)
[2] M. Kobayashi, Y. Ishida, J. l. Hwang, T. Mizokawa, A. Fujimori, K. Mamiya, J.
Okamoto, Y. Takeda, T. Okane, Y. Saitoh, Y. Muramatsu, A. Tanaka, H. Saeki,
H. Tabata, T. Kawai, Phys. Rev. B 72, 201201(R), (2005)
[3] M. Gacic, G. Jakob, C. Herbort, H. Adrian, T. Tietze, S. Brck, E. Goering, Phys.
Rev. B 75, 205206, (2007).
[4] A. Barla, G. Schmerber, E. Beaurepaire, A. Dinia, H. Bieber, S. Colis, F. Scheurer,
J.-P. Kappler, P. Imperia, F. Nolting, F. Wilhelm, A. Rogalev, D. Mller, J. J. Grob,
Phys. Rev. B 76, 125201, (2007)
[5] K. Rode, R. Mattana, A. Anane, V. Cros, E. Jacquet, J.-P. Contour, F. Petroff,
A. Fert, M.-A. Arrio, Ph. Sainctavit, P. Bencok, F. Wilhelm, N. B. Brookes, A.
Rogalev, Appl. Phys. Lett. 92, 012509, (2008).
[6] See: http://ssrl.slac.stanford.edu/stohr/xmcd.htm
[7] O. Copie, K. Rode, R. Mattana, M. Bibes, V. Cros, G. Herranz, A. Anane, R.
Ranchal, E. Jacquet, K. Bouzehouane, M.-A. Arrio, P. Bencok, N. B. Brookes, F.
Petroff, A. Barthlmy, J. Phys.: Condens. Matter 21, 406001, (2009).
[8] G. Schutz, W. Wagner, W. Wilhelm, P. Kienle, R. Zeller, R. Frahm, G. Materlik,
Phys. Rev. Lett. 58, 737, (1987).
[9] B. T. Thole, P. Carra, F. Sette, G. van der Laan, Phys. Rev. Lett. 68, 1943, (1992)
[10] C. T. Chen, Y. U. Idzerda, H.-J. Lin, N. V. Smith, G. Meigs, E. Chaban, G. H.
Ho, E. Pellegrin, F. Sette, Phys. Rev. Lett. 75, 152, (1995).
[11] J.Stohr and Y.Wu, “New Directions in Research with Third-Generation Soft X-
Ray Synchrotron Radiation Sources”, NATO ASI Series, Series E: Applied Sci-
ences, Vol. 254, Eds. A.S.Schlachter and F.J.Wuilleumier, pg. 221-250, (1994).
[12] J. Stohr, J. Electron Spectrosc. Relat. Phenom. 75, 253, (1995)
[13] J. Stohr, J. Magn. Magn. Mater. 200, 470-497, (1999)
128
Bcbfcial[pbs
[14] Y. U. Idzerda, C. T. Chen, H. J. Lin, G. Meigs, G. H. Ho, C. C. Kao, Nucl. Instr.
and Meth. in Phys. Res. A 347, 134-141, (1994)
[15] M. Venkatesan, C. B. Fitzgerald, and J. M. D. Coey, Nature, 430, 630, (2004)
[16] N. R. S. Farley, K. W. Edmonds, A. A. Freeman, G. van der Laan, C. R. Staddon,
D. H. Gregory, and B. L. Gallagher, New J. Phys. 10, 055012, (2008).
[17] G. van der Laan, E. Arenholz, R. V. Chopdekar, Y. Suzuki, Phys. Rev. B 77,
064407, (2008).
[18] G. van der Laan and B. T. Thole, Phys. Rev. B 43, 13401, (1991).
[19] G. van der Laan and I. W. Kirkman, J. Phys.: Condens. Matter 4, 4189, (1992)
[20] S. Blundell, Magnetism in Condensed Matter, Oxford University Press, Oxford,
p. 49, (2001).
[21] P. Gambardella, S. S. Dhesi, S. Gardonio, C. Grazioli, P. Ohresser, C. Carbone,
Phys. Rev. Lett. 88, 047202, (2002).
[22] S. Krishnamurthy, C. McGuinness, L. S. Dorneles, M. Venkatesan, J. M. D. Coey,
J. G. Lunney, C. H. Patterson, K. E. Smith, T. Learmonth, P.-A. Glans, T.
Schmitt, and J.-H. Guo, J. Appl. Phys. 99, 08M111, (2006).
[23] A. Nefedov, N. Akdogan, H. Zabel, R. I. Khaibullin, L. Tagirov, Appl. Phys. Lett.
89, 182509, (2006)
[24] P. Thakur, J. C. Cezar, N. B. Brookes, R. J. Choudary, R. Prakash, D. M. Phase,
K. H. Chae, R. Kumar, Appl. Phys. Lett. 94, 062501, (2009)
[25] P. Abbamonte, A. Rusydi S. Smadici, G. D. Gu, G. A. Sawatky, D. L. Feng, Nat.
Phys. 1, (2005)
129
Xhvpter L
dptixvl mvgnetix xirxulvr yixhroism
xhvrvxterisvtion of XoByopey
Inyium dfliye thin films
ihz rzsults in this xhvptzr fizrz pzrformzy in xollvworvtion fiith thz bvgnztix dfliyzs
group vt thz jnivzrsity of hhzffizlyC ihz mvjority of thz fiork yzsxriwzy is xonxzrnzy
fiith CoByopzy InGdH filmsA wut vlso givzs przliminvry rzsults from Iid(FE) films vlso
yopzy fiith CoC lz przsznt furthzr zviyznxz thvt thz fzrromvgnztism in thzsz mvtzrivls
is not solzly from thz Co yopvntsA rvthzr it mvy wz rzlvtzy to loxvliszy yonor stvtzs
vssoxivtzy fiith oflygzn vvxvnxizsC
130
Cb[pnel 1. Ipnc][f g[ahenc] ]cl]of[l dc]blicmg ]b[l[]nelcm[ncih if
Ci-diped Chdcog Ircde nbch fcfgm
LCF Introyuxtion
In chapter 6 we have made it clear that magnetic circular dichroism is one of the most
powerful tools available to unravel the mysteries surrounding dilute magnetic oxides.
Magneto-optics, MO, investigates the way light is modified when it passes through
a magnetic crystal. The nature of the material’s magnetic states can be investigated
by measuring the properties of circularly polarised light transmitted through the mate-
rial. MO spectroscopy is a direct method for evaluating the s,p-y exchange interactions
within DMS materials [1]. The s,p-y exchange interactions induce a spin-dependent
modification of the host semiconductor band structure in the presence of an applied
magnetic field, that is, the Zeeman splitting [recall Fig. 5. 2 (b)]. A strong MO signal
implies a strong s,p-y exchange interaction. Having reviewed the previous work on
other DMOs using this technique, we then present our own MO data.
Here we present optical MCD (O-MCD) measurements on Co-doped (doping levels
1.0 - 5.0 at.%) InGOH and ITO(10) samples grown using DC magnetron co-sputtering
under the same conditions described in chapter 5. However, the films were grown upon
x-plane sapphire due to the undesirable birefringent properties of r -plane sapphire. In
addition, optical absorption spectra of the Co-doped InGOH films are also given.
LCG erinxiples of bvgnetoBoptixs
If we consider light passing through a material of thickness t, the material will have a
refractive index, n˜ , that is made up of real and imaginary parts.
n˜ = nE + ik (7.1)
Linearly polarised light is made up of equal parts of left (LCP) and right circularly
polarised (RCP) light, as shown in Fig. 7.1 (a). It can be decomposed into these two
states and passed through a magnetic medium. This would mean Eqn. 7.1 now being
written as;
n˜± = nE± + ik± (7.2)
where the LCP light is equivalent to the negative and the RCP light the positive
terms. RCP and LCP light cause charges in a material to rotate in opposite senses.
Each polarisation therefore produces a contribution to the orbital angular momentum
having an opposite sign. If the system is then exposed to a magnetic field this gives
rise to spin-polarisation along the magnetic field direction. The spin-orbit interaction
then leads to an energy contribution for the two circular polarisations having the same
magnitude but opposite sign and therefore, RCP and LCP light will propagate through
the medium with different speeds. An MO effect is seen when, upon recombination,
the refractive index of LCP and RCP are not equal, i.e. where n˜+ ̸= n˜−, and there
is a phase-lag between the two light beams. This is shown in Fig. 7.1 (b) where the
131
Cb[pnel 1. Ipnc][f g[ahenc] ]cl]of[l dc]blicmg ]b[l[]nelcm[ncih if
Ci-diped Chdcog Ircde nbch fcfgm
LCP RCP
E y
E x
E x
q
j
(a)
(b)
Figure 7.1: (a) Linear light made up of equal parts of LCP and RCP. (b) If these
are unequal after being incident upon a sample in a magnetic field, the transmitted
or reflected light will be elliptically polarised and rotated. Adapted from A. J. Behan
thesis [2].
light is elliptically polarised by a rotation of angle θ and change in intensity. This
occurs as a result of the relative changes in phase and amplitude (due to a difference
in attenuation rates for LCP and RCP light) of LCP and RCP light. Therefore, the
presence of a spin-orbit coupling is of paramount importance for the orbital moment
to reflect information about the magnetism of a system.
The ellipticity and rotation of circularly polarised light results in MO effects such
as the Faraday effect and MCD [3]. The Faraday effect is a measurement of the dif-
ference in refractive indices for LCP and RCP light: it is the rotation of the plane
of polarisation for polarised light as it propagates through a medium in the direction
of an applied magnetic field. The effect is dependent on a weighted average over all
transitions in the crystal, so that it is finite in spectral regions when the crystal is not
absorbing.
The MCD is the difference in intensity for left and right circularly polarised light
at a frequency ω. Since it is only dependent upon transitions at ω, it is only non-zero
if the crystal is absorbing. This however, makes it very useful when determining the
nature of a magnetic state. For this reason, it is the MCD that we are most interested
in. The Faraday effect can therefore be used as a check that any MCD observed is
not just because a material is absorbing strongly at its band edge. Both of these MO
quantities are obtained when measuring in transmission mode. Measurements made in
reflection give another MO effect called the Kerr effect, but is outside the scope of this
thesis.
The quantum selection rules that dictate the electronic dipole transitions (∆l =
±1) leads to two possible origins of an optical MCD effect. The first is that we have a
132
Cb[pnel 1. Ipnc][f g[ahenc] ]cl]of[l dc]blicmg ]b[l[]nelcm[ncih if
Ci-diped Chdcog Ircde nbch fcfgm
spin-split valence or conduction band so that the absorption coefficients for LCP and
RCP light are different. The second is a difference in the strength of the transition
intensity for LCP and RCP light which will occur is there is a difference in orbital state
population.
The XMCD and O-MCD techniques are different in terms of their physics but are
complementary to each other in trying to explain the origins of magnetism in these
materials. Whilst XMCD is element specific and can therefore give information on the
magnetic nature of individual elements, it cannot tell us anything about which energy
states are involved in the magnetism. O-MCD is primarily concerned with the energy
states of the solid and the band gap and can therefore provide clues as which energy
states are important in producing a magnetic signal. Correlating absorption spectra
with that of the MCD as a function of energy is an excellent way of doing this, as
we know that MCD will only occur if an absorption takes place. However, as we will
explain later, we are still left not knowing where these states are located within the film.
LCH erevious results
The previous work presented here focuses on those investigations carried out on ZnO-
based DMOs. To my knowledge, no such studies regarding magneto-optical effects in
TM-doped InGOH are available. The major work carried out on the TM-doped ZnO
system has been performed by Ando zt vl. [5,6] and the Magnetic Oxides group at the
University of Sheffield [7, 8].
The first detailed study of magneto-optical effects in TM-doped ZnO was performed
by Ando zt vl. where a large array of TM dopants (Sc, Ti, V, Cr, Mn, Co, Ni and
Cu) were used to dope ZnO, grown using PLD [6]. Undoped ZnO was found to give a
weak and positive diamagnetic signal near the ZnO band gap energy of 3.4 eV. Samples
doped with Sc, Ti, V and Cr showed no enhancement in the MCD signal compared to
that of pure ZnO. However, those samples doped with Mn, Co, Ni and Cu all showed
sizeable MCD effects at both 5 and 270 K. The ZnF−xCoEOEFKO sample gave the largest
MCD signal [see Fig. 7.2 (a)]. In addition, the Co-doped sample shows clear MCD
structures near to a photon energy of 2.0 eV which are clear signatures of the intraionic
y -y∗ of CoG+ symmetry sites in tetrahedral coordination.
The temperature dependence of the Co-doped sample near 3.4 eV [Fig. 7.2 (b)]
shows that the large negative signal at 5 K decreases with increasing temperature
and that the MCD becomes positive at higher temperatures, however, much smaller.
Evidence that this temperature dependence is induced by the transition metal comes
from the fact that the undoped ZnO MCD signal is almost completely independent of
temperature. Ando interprets the MCD signal at the band edge as being a consequence
of conduction and valence band splitting. A sizable sp-y exchange interaction should
lead to a pronounced MCD signal. The strength of the sp-y exchange interaction is
133
Cb[pnel 1. Ipnc][f g[ahenc] ]cl]of[l dc]blicmg ]b[l[]nelcm[ncih if
Ci-diped Chdcog Ircde nbch fcfgm
Figure 7.2: (a) MCD spectra of ZnF−xTMxO (TM=Sc, Ti, V, Cr, Mn, Co, Ni, and
Cu) films at 5 K, (b) MCD spectral shape of ZnF−xCoEOEFKO under a magnetic field of
10 kOe at various temperatures. After reference [6].
explained as being dependent upon the number of 3y electrons present.
A similar study was carried out by Neal zt vl. where the ZnO system was doped
with a variety of TM dopants (Co, Mn, Ti, V) [7]. The MCD spectra along with
absorption spectra, and Faraday spectrum for the Ti-doped sample, are shown in Fig.
7.3 (a). Each show an edge at the band gap in the absorption spectra along with
a corresponding MCD signal. The ferromagnetic nature of the band edge MCD is
confirmed by the observation of open MCD hysteresis loops for the Co-, V-, and Ti-
doped samples at room temperature (see Fig. 7.3 (d) for Co-doped sample). As was
observed by Ando zt vl. the Co-doped ZnO sample shows a negative excitonic feature
at low temperatures [see Fig. 7.3 (b)] which weakens rapidly as the temperature is
increased, due to its paramagnetic nature and the fact that the number of carriers
increases leading to exciton screening.
134
Cb[pnel 1. Ipnc][f g[ahenc] ]cl]of[l dc]blicmg ]b[l[]nelcm[ncih if
Ci-diped Chdcog Ircde nbch fcfgm
Figure 7.3: (a) Room temperature MCD measurements for ZnO doped samples with
a variety of TM dopants, together with the absorption spectra and Faraday rotation in
the case of Ti-doped ZnO, (b) Temperature dependence of the MCD of ZnEONMCoEOEGO,
and, (c) and (d) MCD hysteresis loops obtained at room temperature and an energy
of 3.4 eV for a pure ZnO film (c) and a Co-doped ZnO film (d). After reference [7].
135
Cb[pnel 1. Ipnc][f g[ahenc] ]cl]of[l dc]blicmg ]b[l[]nelcm[ncih if
Ci-diped Chdcog Ircde nbch fcfgm
LCI Zflperimentvl results
LCICF XoByopey InOdP
Film thickness ranged from 210 - 230 nm to ensure a sizable MCD signal close to the
band edge. If the sample is too thin the light will pass through with little interaction,
whilst if it is too thick, not enough light will be transmitted above the band edge for
a signal to be measured.
bvgnetix vny bvgnetotrvnsport properties
Each of the samples were found to be magnetic at both low and above room temper-
ature. The films were all n-type as expected since they had been grown in an oxygen
deficient environment. The resistivities of all the films rose rapidly as the tempera-
ture was lowered, indicating semiconducting behaviour, and obeyed the variable range
hopping law at temperatures below 50 K. The transport measurements taken at room
temperature are summarised in Fig. 7.4. The carrier concentration is reduced for films
with 1.0<Co (at.%)<4.0 in comparison with the pure InGOH (4.53 × 10FN cm−H), as we
would expect if the Co dopants are in the CoG+ state and acting as acceptors. Within
this doping range the carrier concentration is relatively constant with minimal varia-
tion in the saturation magnetisation. At 5 at.% doping, the carrier concentration has
increased and has almost returned to the value of pure InGOH; this is accompanied by
a sudden rise in the saturation magnetisation. The mobility dropped monotonically
with doping between 1.0 at.% and 5.0 at.%; from 3.34 cmG V−F s−F to 0.95 cmG V−F
s−F.
The total magnetisation that can arise from spin-polarised free electrons is nxµB.
This is of the order of 0.10 emu/cmH for carrier densities of approximately 10FN cm−H,
which is more than a factor of 10 smaller than the observed values given in Fig. 7.4.
We thus deduce that the ferromagnetism cannot be attributed to the free electrons.
Given that we have argued above that the ferromagnetism is not due to the CoG+ ions
either (chapter 6), we therefore need to look at the possible contribution from localised
donor states. These are best analysed by optical methods.
Vwsorption properties
Figure 7.5 (a) presents the optical absorption coefficient for Co-doped InGOH with
doping levels of 1.0, 3.0 and 5.0 at.%. The band edge of the InGOH is apparent, together
with a shoulder from approximately 2.8 eV down to 2.5 eV. This shoulder is not
observed in undoped films, and we therefore attribute it to donor states that have been
partially ionised through the compensating effect of the CoG+ acceptors substituting
for InH+. The donor levels are observable in absorption after doping with Co since
136
Cb[pnel 1. Ipnc][f g[ahenc] ]cl]of[l dc]blicmg ]b[l[]nelcm[ncih if
Ci-diped Chdcog Ircde nbch fcfgm
Figure 7.4: Summary of the magnetisation and transport behaviour of Co-doped
InGOH thin films at 290 K.
transitions from the valence band are now possible to these partially empty states.
The most likely origin of these donor levels is oxygen vacancy states which have been
calculated to lie ∼ 0.9 eV below the band edge [9].
Figure 7.5 (b) shows the “Tauc” plots for 0.0 - 5.0 at.% Co-doped InGOH samples.
The Tauc plot is a method used to determine the optical band gap energy, E g, having
first determined the absorption coefficient, α. Using the relationship α∼√h, − Eg
and plotting (αh,)G against the photon energy, one can estimate E g. It is clear that
there is a gradual decrease in E g as the concentration of Co doping increases. The
inset of Fig. 7.5 (b) summarises the dependence of E g, extracted from the absorption
measurements, as a consequence of the Co doping. The pure InGOH has a considerably
larger E g than those which have been doped with Co. After the initial doping stage
where just 1.0 at.% Co has been added, E g falls rapidly. Subsequent doping however
results in a monotonic decrease in E g. This is further evidence for substitution of Co
G+
ions at InH+ sites instead of forming metallic clusters. There is a notable increase in the
slope of the absorption edge between the pure, undoped InGOH and Co-doped InGOH
samples and suggests that a now much heavier band is contributing to the absorption
as a result of the Co doping.
dBbXY xhvrvxterisvtion
Strong evidence that these donor states are related to the ferromagnetism was ob-
tained by O-MCD spectroscopy. The magneto-optical measurements were carried out
using photon energies between 1.5 and 4.2 eV. Low temperature measurements were
conducted within a cryostat with a temperature cooling range of 10 - 300 K. The mag-
137
Cb[pnel 1. Ipnc][f g[ahenc] ]cl]of[l dc]blicmg ]b[l[]nelcm[ncih if
Ci-diped Chdcog Ircde nbch fcfgm
0 1 2 3 4 5
0
20
40
60
80
100
120
0.0 1.0 2.0 3.0 4.0 5.0
3.80
3.85
3.90
3.95
2.0 2.5 3.0 3.5 4.0 4.5
0
5
10
15
20
 
 
(
h
)
2
1
0
1
0
 
(
e
V
/
c
m
2
)
Energy (eV)
 Pure In
2
O
3
 1.0 at.%
 2.0 at.%
 3.0 at.%
 4.0 at.%
 5.0 at.%
E
g
 
Co dopant level (at.%)
E
g
 
(
e
V
)
(b)
 
 
 
(
1
0
4
 
c
m
-
1
)
Energy (eV)
 1.0 at.%
 3.0 at.%
 5.0 at.%
(a)
Figure 7.5: (a) Effect of Co doping on the optical absorption spectra for Co-doped
InGOH thin films at room temperature. (b) Tauc plots for Co-doped InGOH films. The
arrow indicates the direction of the shift in the band gap energy, E g. Inset: dependence
of E g as a function of the Co doping level estimated using the Tauc plots.
netic field applied was 1.8 T at room temperature but was reduced to 0.5 T when using
the cryostat.
Figure 7.6 shows the effect of Co-doping on the room temperature O-MCD spectra
for a set of Co-doped InGOH films. The data presented here are the spectra of the films
obtained after the MCD of a blank substrate has been subtracted. A clear MCD signal
emerges, peaking at the band edge for each of the samples, which gains intensity with
increasing magnetisation (refer to Fig. 7.4). The observation of an MCD signal at the
onset of absorption is good evidence for a spin-split conduction band. The intensity of
the MCD is determined both by the strength of the absorption at a particular energy
and the extent to which that state is magnetically polarised. The observation that the
MCD is rising for energies above 2.5 eV demonstrates that these transitions which gave
rise to the shoulder on the absorption edge involve spin-polarised states. No equivalent
MCD signal was observed for pure InGOH films.
The origin of the MCD signal was explored in more detail by investigating its tem-
perature dependence. Figure 7.7 portrays the MCD signal for the 5.0 at.% Co-doped
sample from 10 K to 300 K. The ratio of the magnitude of the MCD at 10 K and 300 K
agrees very well with the reduction of the magnetisation between 5 K and 300 K. The
fact that the MCD and the magnetisation have the same temperature dependence is
strong evidence to support the contention that the observed magnetisation originates
from the same donor states as those involved in the optical transitions.
CoG+ ions in octahedral coordination give rise to a broad MCD signal [10] near 2.0
eV which we believe is too weak to be detected. The observed signal in this energy
range does not have the correct dependence on cobalt concentration or temperature to
138
Cb[pnel 1. Ipnc][f g[ahenc] ]cl]of[l dc]blicmg ]b[l[]nelcm[ncih if
Ci-diped Chdcog Ircde nbch fcfgm
2.0 2.5 3.0
-100
0
100
200
300
400
1.0 2.0 3.0 4.0 5.0
160
200
240
280
320
 1.0 at.%   2.0 at.%   3.0 at.%   4.0 at.%   5.0 at.%
M
C
D
 
(
D
e
g
r
e
e
s
/
c
m
)
Energy (eV)
 
 
 
 
M
C
D
 
(
D
e
g
r
e
e
s
/
c
m
)
Co doping level (at.%), x
Figure 7.6: O-MCD spectra for the entire Co doping range (1.0-5.0 at.%) at room
temperature for Co-doped InGOH thin films. Inset: MCD signal at 3.28 eV plotted as
a function of the Co doping level.
separate it from the noise. However the shape of the MCD spectrum at 300 K does
rule out a sizeable contribution from metallic cobalt [11, 12] or CoO [13] and there is
no temperature dependence to indicate CoHOI and CoGOH where the transition tem-
peratures are below 40 K [14].
We believe that the transition metal ions are acting as acceptors and ionising a
sufficient fraction of the donors to produce a partially occupied narrow donor band,
much like in the sketches shown in chapter 2 in reference to Coey’s work. Therefore,
the inevitable hybridisation of the broad donor band with the 3y band is the origin of
the magnetism in our Co-doped InGOH films. However, we are unable to determine the
location of these donor states - they may be distributed evenly through the material or
concentrated at the grain boundaries as has been suggested by other studies [15, 16].
LCICG XoByopey Iid(FE)
Ferromagnetic 1.0 - 5.0 at.% Co-doped ITO(10) thin films were investigated in much
the same way as the Co-doped InGOH samples described above in 7.4.1. The films were
metallic over the entire temperature range with carrier concentrations typically in the
range of 10GE cm−H.
Figure 7.8 (a) shows the room temperature O-MCD spectra as a function of Co
139
Cb[pnel 1. Ipnc][f g[ahenc] ]cl]of[l dc]blicmg ]b[l[]nelcm[ncih if
Ci-diped Chdcog Ircde nbch fcfgm
2.0 2.5 3.0
0
200
400
 
 
M
C
D
 
(
D
e
g
r
e
e
s
/
c
m
)
Energy (eV)
 10 K
 50 K
 100 K
 300 K
Figure 7.7: Temperature dependence of the O-MCD for a 5.0 at.% Co doped InGOH
sample.
doping from 1.0 - 5.0 at.% in Co-doped ITO(10) films. All spectra have been corrected
for by subtracting the measured signal from a x-plane sapphire substrate. A clear
MCD signal emerges, peaking at the band edge for each of the samples, which mono-
tonically gains intensity with increasing levels of Co doping [Fig. 7.8 (b)]. Notice that
the onset of MCD and therefore absorption occurs at a slightly higher energy, 2.75 eV
instead of 2.50 eV is the case of the Co-doped InGOH films. This suggests that the band
gap energy has increased or that the energy level of the donor states has shifted upon
doping with Sn. One of the major reasons for shifting of the band gap can be related
to changes in the lattice constant. Compression of the lattice will widen the bands and
therefore will widen the gaps. The changing lattice will overide the influence of the
Sn donor band that will form just below the conduction band, which would normally
reduce the band gap (recall section 4.2). The MCD signal appeared to be correlated
to the magnetisation as is clear from a comparison between Figs 7.8 (b) and (c).
Figure 7.8 (d) portrays the temperature dependence of the MCD signal as a func-
tion of temperature for a 5.0 at.% Co-doped ITO(10) film. The sample behaved in a
very similar way to the 5.0 at.% Co-doped InGOH sample described earlier, i.e. decreas-
ing temperature led to a continual rise in MCD intensity.
Magnetisation in samples containing additional Sn doping is generally larger than
those without Sn doping (section 5.6.2) and we now know that the MCD scales with
the strength of the total film magnetisation. The larger values for magnetisation and
MCD in the Sn doped samples confirms this.
140
Cb[pnel 1. Ipnc][f g[ahenc] ]cl]of[l dc]blicmg ]b[l[]nelcm[ncih if
Ci-diped Chdcog Ircde nbch fcfgm
1.0 2.0 3.0 4.0 5.0
5
6
7
8
9
10
1.0 2.0 3.0 4.0 5.0
330
360
390
420
450
480
510
2.0 2.5 3.0 3.5
-100
0
100
200
300
400
500
 1.0 at.%
 2.0 at.%   
 3.0 at.%
 4.0 at.%
 5.0 at.%
 
 
M
C
D
 
(
D
e
g
r
e
e
s
/
c
m
)
Energy (eV)
 
 
Co doping level (at.%)
M
s
 
(
e
m
u
 
c
m
-
3
)
 
 
 
 
M
C
D
 
(
D
e
g
r
e
e
s
/
c
m
)
Co doping level (at.%)
2.0 2.5 3.0
-100
0
100
200
300
400
(b)
(a)
 
 
 10 K
 50 K
 100 K
 300 K
M
C
D
 
(
D
e
g
r
e
e
s
/
c
m
)
Energy (eV)
(d)
(c)
Figure 7.8: (a) Room temperature O-MCD spectra for 1.0-5.0 at.% Co-doped ITO(10)
films; Dependence of room temperature O-MCD signal at 3.28 eV (b) and magneti-
sation, b s, (c) on the Co doping level; (d) Temperature dependence of the O-MCD
spectra for a 5.0 at.% Co-doped ITO(10) film.
141
Cb[pnel 1. Ipnc][f g[ahenc] ]cl]of[l dc]blicmg ]b[l[]nelcm[ncih if
Ci-diped Chdcog Ircde nbch fcfgm
LC5 hummvry vny Xonxlusions
In summary, we have shown that the ferromagnetism observed in semiconducting Co-
doped InGOH films is not related to the formation of secondary metallic Co clusters,
similar to the findings described in chapter 6 using XMCD. Taken together, our data
show that the magnetism may be due to polarised electrons in localised donor states,
most likely related to oxygen vacancy defect states. Therefore, we can think of the
origins of magnetism in these films in terms of Coey’s band pictures where we have a
hybridisation and charge transfer of electrons from a broad donor/defect-derived im-
purity band to a band of unoccupied 3y states at the Fermi level. The evidence for
this comes from the fact that the optical MCD shows a strong dependence on the bulk
magnetisation both as a function of temperature and as a function of the Co doping.
This suggests that the states that we see optically in the absorption spectra are in
fact the states that are giving rise to the magnetisation. We believe that the same
localised donor states are also present in the Co-doped ITO samples. The presence of
Sn however will have an influence on the donor state energy levels and is evident from
the shift in band gap energy and magnitude of the magnetisation and MCD compared
to samples without additional Sn doing. We expect that similar mechanisms might be
important in other DMOs.
142
Wiwliogrvphy
[1] K. Ando, Springer Series in Solid-State Science, Vol. 128, edited by S. Sugano and
N. Kojima, p. 211, (Springer, Berlin, 2000).
[2] A. J. Behan, PhD Thesis, Characterisation of Doped ZnO Thin Films for Spin-
tronic Applications, University of Sheffield, (2008).
[3] F. J. Kahn and P. S. Pershan, Phys. Rev. 186, 3, (1969).
[4] K. Ando, Appl. Phys. Lett. 82, 1, (2003).
[5] K. Ando, H. Saito, Z. Jin, T. Fukumura, M. Kawsaki, Y. Matsumoto, H. Koinuma,
Appl. Phys. Lett. 78, 18, (2001).
[6] K. Ando, H. Saito, Z. Jin, T. Fukumura, M. Kawsaki, Y. Matsumoto, H. Koinuma,
J. Appl. Lett. 89, 11, (2001).
[7] J. R. Neal, A. J. Behan, R. M. Ibrahim, H. J. Blythe, M. Ziese, A. M. Fox, G. A.
Gehring, Phys. Rev. Lett. 96, 197208, (2006).
[8] A.J. Behan, J.R. Neal, R.M. Ibrahim, A. Mokhtari, M. Ziese, H.J. Blythe, A.M.
Fox, G.A. Gehring, J. Magn. Magn. Mater., 310, 21582160, (2007).
[9] 26. L. M. Tang, L. L. Wang, D. Wang, J. Z. Liu, K. Q. Chen, J. Appl Phys. 107,
083704, (2010).
[10] I. N. Douglas, Phys. Stat. Sol. (b) 73, 641, (1976).
[11] C. Clavero, A. Cebollada, G. Armelles, Y. Huttel, J. Arbiol, F. Peiro, A. Cornet,
Phys. Rev. B 72, 02441, (2005).
[12] D. Score et al J. Phys.: Conf. Ser. 200, 062024, (2010).
[13] M. A. White, S. T. Ochsenbein, D. R. Gamelin, Chem. Mater. 20, 7107-7116,
(2008).
[14] Y. Ichiyanagi, J. Magn. Magn. Mater., 272-276, e1245-e1246, (2004); S. A.
Makhlouf, J. Magn. Magn. Mater., 246, 184-190, (2002).
[15] J. M. D. Coey, P. Stamenov, R. D. Gunning, M. Venkatesan, K. Paul, New Journal
of Physics 12, 053025, (2010).
143
Bcbfcial[pbs
[16] B. B. Straumal, A. A. Mazilkin, S. G. Protasova, A. A. Myatiev, P. B. Straumal,
G. Schutz, P. A. van Aken, E. Goering, B. Baretzky, Phys. Rev. B. 79, 205206,
(2009).
144
Xhvpter M
XoDIid multilvyers
Givnt mvgnztorzsistvnxz in multilvyzrzy CoDIid thinBfilms xvn oxxur zvzn vt room
tzmpzrvturzA wut thz xruxivl rolz plvyzy wy supzrpvrvmvgnztix rzgions vt thz CoDIid
intzrfvxz to thz mvgnztotrvnsport propzrtizs hvs rzmvinzy unxlzvrC Wy grofiing v szB
rizs of pCoDIidr×c multilvyzrs fiith wilvyzr rzpzvt numwzrA cA vvrying from G up to IEA
vny Iid thixknzss in thz FCGE B ICGE nm rvngz fiz hvvz yiszntvnglzy thz vvrious mvgB
nztix xontriwutions fihixh most strongly influznxz thz physixvl propzrtizs of this systzmC
ihz rzsults shofi thvt Iid is v promising mvtzrivl for intzgrvtion fiithin ofliyzBwvszy
spintronix yzvixzsC
145
Cb[pnel 2. Ci/CTI gofncf[selm
Figure 8.1: Magnetoresistance (a) and magnetisation saturation field (b) of sputter
deposited Si (111)/Cr (10 nm)/[Fe (2 nm)/Cr (tXr nm)]c/Cr (5nm) (c = 20 or 30)
multilayers measured at 4.2 K. After reference [3].
MCF Introyuxtion
Exchange coupling of ferromagnetic (FM) metals through a non-magnetic (NM) spacer
in layered structures is of considerable interest, both from a phenomenological and tech-
nological viewpoint. The most striking discoveries in these systems are the observa-
tion of giant magnetoresistance (GMR) [1], antiferromagnetic (AF) interlayer exchange
coupling [2] as well as oscillations in the GMR ratio as a function of NM spacer layer
thickness [3]. The following sections aim to introduce the GMR effect and its origins
as well as some of the most influential factors which can affect the behaviour of these
FM/NM layered systems.
MCFCF ihe Givnt bvgnetoresistvnxe effext
In 1988 Baibich zt vl. observed magnetoresistance changes as large as 50% at low tem-
peratures in Fe/Cr multilayer structures [1]. The Fe layers were antiferromagnetically
coupled across the Cr spacer layers, and the large change in resistance was found to
be due to the change in the relative orientation of the ferromagnetic layers from an-
tiparallel to parallel alignment as the field was increased. In order to achieve a large
resistance change it was important for the Fe layers to be antiparallel in the absence
of an applied field. In addition, it was found that the interlayer coupling oscillated
between ferromagnetic and antiferrromagnetic depending on the thickness of the Cr
interlayer [3–5] (see Fig. 8.1). The effect has since been seen in a number of other
FM/NM systems such as Co/Cu and Co/Ru [5,6].
The GMR effect in magnetic multilayers is the dependence of the resistance on
the relative angles between the magnetisation direction of successive FM layers. The
effect exploits the spin-dependence of the electrical conduction in FM metals and spin-
146
Cb[pnel 2. Ci/CTI gofncf[selm
R
R
R
R
Carrier
spin
RAP = R R+
2
R R
R R
RP = R R2
R R+
(a) (b)
FM
NM
“+”
“-”
Spin “+”Spin
Spin “-”Spin
Parallel Antiparallel
Figure 8.2: (Top panel) Electron transport in FM/NM multilayers for parallel (a) and
antiparallel (b) alignments of magnetisation. Typical trajectories of spin “+” and spin
“-” electrons are indicated by dashed lines; (Bottom panel) Effective resistor scheme
that gives the parallel and antiparallel resistances of the multilayer.
dependent scattering at the interfaces. As explained in chapter 2, the essence of an
MR response is the change in electrical resistance as a reaction to an applied magnetic
field. In multilayered systems, the magnetisation of the respective FM layers will be
altered upon application of a magnetic field. The GMR ratio can therefore be defined
as
∆g
g
=
gAe −ge
ge
(8.1)
where ge and gAe are the resistances in the parallel and antiparallel states respec-
tively.
The GMR effect is best described by the two current model, based on a model of
the electrical conduction in ferromagnetic metals by Mott [7] which was introduced
in section 2.3.1. Here, the conductance of ferromagnetic materials can be viewed as
the sum of separate contributions from electrons with opposite spin directions. There
is a dissimilar scattering rate for the two electron channels due to a difference in the
available energy states at the Fermi level. Figure 8.2 illustrates the origin of the GMR
effect for a periodic FM/NM multilayer, depicting typical trajectories of diffusively
scattered electrons with opposite spin directions.
In Fig. 8.2 (a), electrons for which in the parallel state the scattering probability is
smallest are called the spin “+” electrons. In the parallel state the spin “+” contribu-
147
Cb[pnel 2. Ci/CTI gofncf[selm
V
I
VI
(a)CIP (b) CPP
Figure 8.3: Schematic diagram of the CIP (a) and CPP (b) geometries of GMR.
tion to the conductance is much larger than the spin “-” contribution. In contrast, in
the antiparallel state [Fig. 8.2 (b)], both spin “+” and spin “-” scatter strongly within
the FM layers and results in a larger antiparallel resistance. Therefore, each layer acts
as a spin-selective valve (see chapter 10), i.e. its magnetisation determines which spin
polarisation (“+” or “-”) it will transmit most easily .
Within a simple model for the GMR effect, ge and gAe in the antiparallel and
parallel configurations can be found from the resistor schemes illustrated in the bottom
panel of Fig. 8.2. This series resistor model leads to a GMR ratio equal to
∆g
g
=
(g↓ −g↑)G
4g↓g↑
O (8.2)
A number of semi-classical and quantum-mechanical approaches to explaining the
GMR effect have been made by a host of authors [9, 10], but are beyond the scope of
this thesis.
It is possible to measure the GMR in either the Current-In-Plane (CIP) or Current-
Perpendicular-to-Plane (CPP) configuration, as shown in Figure 8.3. The CIP mea-
surement requires no patterning and is therefore easier to perform on thin films (this
technique was used in this chapter to measure the GMR properties of Co/ITO multi-
layers) and is characterised by the same drop in electrical resistance of the thin film
sample when a magnetic field is applied. However, for the electrons to sample all of
the different layers the requirement is that the mean free path of the electrons is longer
than the sum of the layer thicknesses. The CPP configuration gives easier access to
many of the fundamental parameters of the system [11], but is more complex to achieve
experimentally (see chapter 10). The critical length scale is usually the spin-diffusion
length. As we shall discuss later in chapter 10, CPP GMR measurements are governed
by spin accumulation effects, however, this is not the case in the CPP geometry since
the current flow is parallel to the layers and instead depends on the differential mobil-
ities of the spin-up and spin-down electrons. The CIP geometry has the disadvantage
of a lower GMR relative to the CPP measurement.
148
Cb[pnel 2. Ci/CTI gofncf[selm
Electron Impuirty
(a)Spin-dependent scattering
Scattering off a magnetic
impurity with spin change of 1
Scattering by a spin wave with
spin 1
Spin orbit scattering
(b) Spin-flip scattering
Figure 8.4: Different types of scattering in magnetic multilayers, (a) spin-dependent
scattering, and (b) spin-flip scattering.
MCFCG Importvnxe of spinByepenyent sxvttering
The GMR effect is based upon spin-dependent scattering. We are mainly concerned
with elastic (energy conserving) scattering. In each scattering event only the direc-
tion of propagation of electrons changes. It is essential to distinguish between spin-
dependent scattering which causes the GMR and spin-flip scattering which is detri-
mental to the GMR. The two major forms of scattering are shown in Fig. 8.4. In
the case of spin-dependent scattering the orientation of the electron spin is conserved
in each scattering event but the probabilities of scattering for electrons with up and
down spin projections are different. On the other hand, when an electron undergoes a
spin-flip scattering, its spin orientation changes from up to down (or vice versa) and,
at the same time, the spin of the scattering centre changes so that the total spin is
conserved.
There are a number of crucial factors that lead to the spin-dependent scattering
all of which derive from the fact that ferromagnetic materials have a spin-split band
structure (recall section 2.1.3). The sp and y bands are strongly hybridised where
the y band is only partially filled and therefore has a number of available unoccupied
states. This gives the s electrons (which dominate the conduction in 3y TMs) options
to scatter into the unoccupied states of the part-filled y band, thus reducing the mean
free path of the electrons. This explains why the the conductivity of noble metals is
considerably higher compared to ferromagnetic transition metals.
Another consequence of the spin-split y bands is a difference in the density of states
at the Fermi energy level for spin-up and spin-down electrons and therefore a different
scattering probability for each. Spin-dependent scattering which originates from the
relative shift in the spin-up and spin-down bands takes place at the interface between
149
Cb[pnel 2. Ci/CTI gofncf[selm
Lowerferromagnet, t
Upper ferromagnet, t L
n n+1
FM2
FM1
Non-magnetic spacer, t
NM
s
Upper ferromagnet, tFM1
Lower ferromagnet, t
FM2
Non-magnetic spacer, t
NM
(a)
(b)
Figure 8.5: Two of the commonly invoked models for biquadratic coupling: (a) Thick-
ness fluctuations and biquadratic coupling. For a trilayer where the spacer layer thick-
ness varies periodically between n and n+1, the bold arrows show the local variation in
the direction of the magnetisation. n-layer thick regions are taken to be anitferromag-
netic and the magnetisations rotate away from each other. n+1 layer thick regions are
considered as ferromagnetically coupled regions and the magnetisations align towards
each other; (b) the loose spin model.
the FM and NM layers. This interfacial spin-dependent scattering arises from the band
structure, for example, if there is a matching of the electronic potentials in one spin
direction and not in the other. Therefore, for optimum interfacial spin-dependent scat-
tering we require a good match between the bands of the magnetic layers and those
of the spacer layer in one spin channel and as large as possible mismatch in the other
spin channel.
There are several sources of spin-flip scattering (spin-independent scattering). Of-
ten these scattering events are results of microstructural imperfections such as stray
impurities, vacancies, defects, roughness and dislocations. Also shown in Fig. 8.4 (b)
are spin-flip processes which take place due to scattering with spin-waves and collisions
with impurities where a spin-orbit interaction may take place. As we shall discuss in
detail later, interdiffusion at interfaces where the elements are miscible can complicate
the situation further. Not only will this have repercussions on the scattering processes,
but also on the interlayer exchange coupling (see section 8.1.3). Since all these pro-
cesses mix the up and down spin channels (“spin mixing”), they are detrimental to the
GMR.
150
Cb[pnel 2. Ci/CTI gofncf[selm
MCFCH Interlvyer Zflxhvnge Xoupling
Interlayer exchange coupling (IEC) is concerned with the alignment of ferromagnetic
layers that are separated by a spacer layer, much like the schematics shown in Fig. 8.2.
The interest in IEC was revamped after observations made by Parkin zt vl. [3] where
such coupling was found to oscillate with the spacer layer thickness. The two main
forms of coupling are known as the “bilinear” and “biquadratic” coupling.
Phenomenologically, the IEC energy per unit area (E FPG) between two FM layers
(FMF and FMG) separated by a spacer of thickness tcb , can be written in the following
form;
EFPG(tcb) = −JF(tcb) cosαFPG − JG(tcb) cosG αFPG (8.3)
where α refers to the relative magnetisation orientations of FMF and FMG, and J F
and J G are the bilinear and biquadratic coupling constants. If we consider a trilayer
structure, when the two FM layers of thickness tFb are aligned antiferromagnetically,
the coupling between the layers can be determined according to the equation;
xJF = −µEHhAiMFb tFb (8.4)
in which x = 1 or 2 and is a constant which depends on the order of magnetic re-
orientation, b Fb is the total magnetisation and µE] hAi is the saturation field. In
the case where repeated bilayers are used, e.g. a multilayer stack, J F is further halved
to take into account that two surfaces of each FM layer are being coupled.
Two general approaches have been taken to explain bilinear IEC; the first, an ex-
tension of the RKKY coupling model [12], and the second, based on the consideration
of quantum confinement [13] arising from the periodic modulations of the superlattice
structures. Detailed reviews of the bilinear IEC effect can be found in reference [14].
Biquadratic coupling is a more complex matter. Discrepancies between experimen-
tal and theoretical values for J G have led to attempts of explaining biquadratic coupling
by extrinsic effects. Slonczewski [15], and later Demokritov zt vl. [16], looked to solve
the problem by taking into account microstructural factors such as spacer layer fluc-
tuations, formulating rough interface models. Considering the hypothetical surfaces
shown in Fig. 8.5 (a), with a terrace of width a within the spacer layer, both models
conclude that J G is directly proportional to a, i.e. J G increases with a up to a limit
(obviously, beyond a critical limit, a will be so wide that we have a non-terraced con-
tinuous layer) .
Slonczweski later proposed another model which better explained the unexpected
strong temperature dependence of the biquadratic coupling [17], where he considered
the case in which magnetic impurities were present in the spacer or at its interfaces,
as shown in Fig. 8.5 (b), where a small arrow of spin h is in the NM layer. Such
impurities interact with the two FM layers on both sides of the NM and mediate the
interaction of the two FM layers. There is no direct exchange coupling, only indirect
exchange coupling (RKKY-type) mediated by the spacer layer electrons.
Another form of coupling which will compete with the IEC is that due to the
151
Cb[pnel 2. Ci/CTI gofncf[selm
presence of pinholes, or the break in the spacer layer giving direct exchange coupling
between the two FM layers. Since there is direct contact between the FM layers, pin-
hole coupling is ferromagnetic.
MCG erevious lork
The contributions towards the total GMR signal are not always straightforward, and
there are a number of FM/NM multilayer systems where the GMR signal exhibits a
very strong non-saturating behaviour explained by the coexistence of FM and SPM
regions [18]. In this scenario, the SPM regions are magnetically decoupled from the
FM regions of the magnetic layers. Typically, the magnitude of the total GMR is lower
than in systems which are free from the SPM contribution. These systems can be mod-
elled well using the Wiser-Hickey model [19,20], usually prescribed to granular systems,
whereby the total GMR signal constitutes three separate components which depend
on the path of the spin-dependent scattering event of each electron. In conventional
GMR, where there is no contribution from SPM particles, scattering will typically take
place between FMF→NM→FMG. However, the presence of SPM particles will result
in an electron scattering path of FMF⇀>NM⇀>SPM and the emergence of a high-field
contribution to the total GMR. The quality of the multilayers has been shown to influ-
ence the SPM term in the total GMR signal: lower surface roughness of the substrate
has been found to decrease the SPM contribution [21].
The relevant pioneering work for the phenomena observed here have been briefly
explained in earlier sections, however, are all concerned with metal based multilayers.
Here we are interested in FM metal/oxide interfaces. Chen zt vl. have produced work
on Fe- and Co/ITO multilayers [22, 23]. In particular, oscillations in the GMR as a
function of ITO spacer thickness are observed and the origins of the GMR response is
explained as having its origins from ferromagnetic and interfacial superparamagnetic
contributions. However, the work does not go into any great detail about the presence
of superparamagnetic particles at the interface. The results presented in this chapter
aim to shed some light on these issues.
MCH Zflperimentvl gesults
All [Co/ITO]×c multilayers (ITO(10) composition) were prepared at room temperature
on unheated 5×10 mmG area r -plane sapphire substrates by DC magnetron sputtering
in a diffusion pumped UHV system. Prior to film growth, the system’s base pressure
was better than 1×10−K Pa. Films were sputtered in 1.50±0.01 Pa of Ar with each
layer grown in sequence and without breaking vacuum by rotating the substrates below
152
Cb[pnel 2. Ci/CTI gofncf[selm
Figure 8.6: Illustration of a Co/ITO multilayer, where tIid is the ITO layer thickness,
tγ is the thickness of the magnetically dead layer at the Co/ITO interface, tfs is the
thickness of the magnetically dead layer at the Co/r -plane sapphire interface, twx is
the equivalent bottom Co layer thickness (= tXo - tfs - tγ) and tXo is the equivalent
thickness of subsequent Co layers.
stationary Co and ITO targets (as described in sections 3.3 and 3.4). The gas flow was
approximately 0.10±0.01 Pa/s. The deposition rates for the ITO and Co targets were
0.02 and 0.04 nm/W/pass at 1 rpm calibrated in 1.50±0.01 Pa respectively. A power
of 10 and 15 W was applied to the ITO and Co targets.
Because the lattice parameter for ITO is 1.012 nm, the minimum ITO thickness
(tIid) used was 1.20 nm so that the risks of discontinuities and fluctuations, the for-
mation of magnetic pinholes and/or the presence of loose spins could be minimised.
Stacks where the bilayer repeat number was varied (c = 2 - 40) had an ITO thickness
of tIid = 3.0 nm. In those stacks where c = 10, the ITO thickness was systematically
varied between each sample in the (1.20 - 4.20±0.10) nm range. The thickness of the
Co layer remained constant for all multilayers at (1.20±0.1) nm. In Fig. 8.6 an illus-
tration of a [Co/ITO]×c multilayer used in these experiments is shown.
MCHCF bvgnetix Yevy avyers
Interfacial intermixing between the Co and ITO layers was investigated. For FM/oxide
systems, an equivalent thickness of the FM (in this case Co) can be assumed whereby
the FM atoms are surrounded by a host of O atoms, resulting in the FM atoms be-
coming oxidised. This can happen for one of two reasons; (1) due to oxygen diffusion
153
Cb[pnel 2. Ci/CTI gofncf[selm
0.0 0.5 1.0 1.5 2.0 2.5
0
1
2
3
4
Co on r-Al
2
O
3
Co on 3nm ITO
 
 
M
S
 
(
 
1
0
-
5
 
e
m
u
/
c
m
2
)
t
Co
 (nm)
Figure 8.7: Moment per unit area versus Co thickness deposited onto r -plane sapphire
substrates (closed circles) and on a 3.0 nm layer of ITO (open circles). Solid lines linear
least-square fits to the experimental data used to extrapolate the dead layer thicknesses.
into the Co layer, and (2) due to stray Co atoms diffusing into the oxide layer. We
define this region in our Co/ITO system as the magnetically dead layer (MDL) and
more precisely can be thought of as a length scale over which there is little or no mag-
netic exchange. The quality of the interfaces is therefore important as we would expect
increasing levels of discontinuity in the Co or ITO layers to increase the thickness of
the MDL. Potentially, the MDL can be detrimental since their magnetic and electronic
properties will differ from those of the metallic Co atoms in a ferromagnetic and metal-
lic crystal. Investigation into the possibility of an MDL being present and its potential
scale was carried out by measuring the magnetisation, b s, per unit surface area as
a function of Co layer thickness, tXo (nm), on a number of samples where a 3.0 nm
layer of ITO had been deposited onto r -plane sapphire substrates. All samples were
capped with 2.0 nm of gold to prevent further oxidation of the Co layer. The MDL
thickness per interface, tγ (see Fig. 8.6), was extrapolated as the point where the linear
fit intersected the fl -axis (b s = 0) and was found to be approximately (0.17±0.15) nm
(Fig. 8.7) similar to that found in Co/ZrOG [24] and Co/MgO multilayers [25].
The behavior of the first Co layer deposited at the surface of the r -plane sapphire
substrate has also been considered. We have found through systematic variation of
the Co layer thickness (0.2 - 5.0 nm) onto r -plane sapphire substrates that an MDL
of (0.22±0.15) nm [(tfs (Fig. 8.6)] exists at the Co-r -plane sapphire interface (Fig.
8.7). This implies that the first Co layer is mostly redundant as almost half of it is
magnetically inert due to the MDL formed at both the substrate and ITO interfaces.
154
Cb[pnel 2. Ci/CTI gofncf[selm
0 1 2 3 4 5
0 1 2 3 4 5
 
 
 
 
 
 
L
n
 
I
n
t
e
n
s
i
t
y
 
(
a
r
b
.
 
u
n
i
t
s
)
 
 
N = 10
(e)
(f)
(d)
(b)
(c)
N = 3
N = 20
N = 5
N = 2
 
 
 
N=1
(a)
  
 
Incident angle (Degs)
 
 
 
 
Figure 8.8: Low angle reflectivity for [Co/ITO]c multilayers with varying c between
1 to 20. Solid black lines represent measured data, and the solid red circles are the
corresponding simulations.
155
Cb[pnel 2. Ci/CTI gofncf[selm
0 5 10 15 20 25 30 35 40
0.00
0.06
0.12
0.18
0.24
0.30
290 K
 
 
G
M
R
 
(
%
)
Number of multilayers, N
10 K
Figure 8.9: Absolute GMR dependence on the number of bilayer repeats, c, in
[Co(1.20nm)/ITO(3.00nm)]c multilayers at 10 K (closed circles) and 290 K (open
circles). The solid lines correspond to fits made using the Fuchs-Sondheimer model.
MCHCG kvrying the numwer of repevt wilvyersA N
htruxturvl properties
The number of bilayers, c, in the multilayers is varied to understand the influence of
interfacial features on the overall multilayer performance. Generally c is kept rather
large in other FM/NM multilayers, typically cS20, and has been reported to play a
significant role in achieving large values for GMR [3,26].
Figures 8.8 (a) - 8.8 (f) show the low-angle x-ray reflectivity (LAR) measurements
for the build up of c from 1 to 20 with comparative simulation fits. The main assump-
tion made when attempting to simulate the experimental data was that the quality of
the layers improved as they moved further away from the substrate. The typical ranges
used for the roughness parameters from start to finish for the fittings were 0.40 - 0.10
nm for Co and 0.60 - 0.20 nm for the ITO layers. There is excellent agreement between
the experimental and simulated data up to c = 10 after which the experimental data
starts to deviate from the simulation. After c = 20, the total stack thickness is verging
on the limit of obtaining good low angle data and may also be nearing the maximum
acceptable roughness. c = 40 (not shown) reveals no discernable Kiessig fringes, but
does show two multilayer Bragg peak reflections.
bvgnetotrvnsport properties
The thickness (equivalent c ) dependence of our MR data was fitted using the Fuchs-
Sondheimer model [27,28] following the method applied in reference [29]. All boundary
conditions are assumed to be diffusive, such that the measured effective resistivity, /∗,
156
Cb[pnel 2. Ci/CTI gofncf[selm
-0.50 -0.25 0.00 0.25 0.50
-1.0
-0.5
0.0
0.5
1.0
 a = 10
 a = 500
 
 
M
/
M
S
0
H (T)
Figure 8.10: Sensitivity of the shape of the magnetisation at 290 K on the Langevin
parameter, v, in Eq. 8.7.
corresponds to
/∗ = /
[
1 +
3ℓ
8t
]
, (8.5)
where t is the thickness of the film which leads to a bulk scattering resistivity / and
a mean free path ℓ. The first term relates to bulk scattering, while the second term
corresponds to additional scattering at surfaces. According to this equation, resistivity
of any given film will show finite size effects with characteristic bulk resistivity and
mean free path. The GMR ratio can then be defined as
GMg =
/∗E − /∗H
/∗E
, (8.6)
where /∗H and /
∗
E are resistivities in an in-plane applied magnetic field (µEH) and zero
magnetic field respectively. Eqn. 8.6 has been used to fit the thickness dependence
of the Co/ITO stacks. Fig. 8.9 shows the thickness dependence of the multilayers
with corresponding fits to Eqn. 8.6. The bulk resistivities in an applied magnetic field
and zero field were determined from straight line fits of /∗t against t using Eqn. 8.5.
The free fit parameters in Eqn. 8.6 (once Eqn. 8.5 has been substituted accordingly)
were ℓE and ℓH . We obtained values of 17.0 nm and 4.7 nm for ℓE at 10 and 290 K
respectively. The corresponding ℓH values were 25.1 nm and 31.1 nm at 10 and 290
K. Notice that the values obtained here agree well with ones reported elsewhere for
Co/Cu multilayers [29]. The temperature dependence of the values for the mean free
path seem to be unrealistic; in a metallic system it is reasonable to expect an increas-
ing mean free path with decreasing temperature. However, this not the case in our
multilayers. Considering that similar analysis by Marrows zt vl. on Co/Cu multilayers
resulted with similar unphysical temperature dependence in the mean free path of the
multilayers, we assume that these unphysical values are a result of the limitations in
157
Cb[pnel 2. Ci/CTI gofncf[selm
the Fuchs-Sondheimer model. The most interesting result of the Fuchs-Sondheimer
model fitting is the prediction of finite size effects in GMR for our multilayers.
GMR achieved in similar Co/ITO multilayers was considerably higher than the
GMR we obtain here [23]. We believe that this is due to high levels of spin-independent
interfacial scattering between the Co and ITO layers in our samples. We expect that
the interfacial region between Co and ITO is expected to have a frustrated magnetic
structure which is possibly superparamagnetic or spin glass type. Large scattering
effects makes it very difficult to interpret the resistance data and so analysis of the
magnetic data may be a more fruitful avenue to gain information on the different mag-
netic contributions to the physical properties of the multilayers (see section below).
bvgnetix boyelling
It is well documented that ultra thin films of Co exhibit superparamagnetism [30].
Since in our films tXo is approximately (1.20±0.10) nm, we expect a SPM contribution
to the overall magnetisation.
The magnetic properties of the [Co/ITO] stacks could be modelled assuming we
have a two-phase system with SPM and FM components contributing to the magnetic
and magnetotransport properties. Since magnetisation measurements are volume sen-
sitive, the volume fraction of each phase can be extracted.
The Langevin function describes the field dependence of the magnetisation of both
ferromagnets and superparamagnets [31, 32]:
M(H) =Mh[xoth(HRaheb)− ahebRµEH], (8.7)
where b s is the saturation magnetisation as before, i is the absolute temperature
and v is the Langevin parameter. The value of the Langevin parameter will define
the shape of the field dependence of the magnetisation as illustrated in Fig. 8.10. For
vheb=10, the magnetisation curve has a relatively low magnetisation saturation field
similar to those observed in ferromagnetic materials, while larger values for vheb , such
as 500, produce a non-saturating magnetisation curve similar to those observed in su-
perparamagnetic materials. Therefore, one can write an expression which can describe
the magnetisation of a system which contains ferromagnetic and superparamagnetic
phases:
M(H) = fMh[xoth(HRaFb)− aFbRµEH]
+ (1− f)Mh[xoth(HRaheb)− ahebRµEH], (8.8)
Here aFb and aheb are the Langevin parameters for the ferromagnetic and super-
paramagnetic phases, f is the ferromagnetic volume fraction and b s, i and µEH are
as before. In Eqn. 8.8 a value of f = 1 means the sample contains only a ferromagnetic
phase, while f = 0 corresponds to a case where the superparamagnetic phase is the
158
Cb[pnel 2. Ci/CTI gofncf[selm
Figure 8.11: Normalised magnetisation curves for a selection of [Co/ITO]c ML stacks
where (a) c = 1, (b) c = 3, (c) c=5, (d) c = 20, and (e) low-field region of a stack
where c = 20, along with model fits (solid green line).
lone phase in the system. To investigate the possibility of a SPM contribution in our
samples, normalised b -µE] loops were fit to Eqn. 8.8.
Figs. 8.11 (a) - 8.11 (c) show the magnetisation curves for a selection of [Co/ITO]×c
multilayers along with fits using the model described above. The fits are excellent at
both low and high magnetic fields which is consistent with the presence of a contribu-
tion to the magnetisation and GMR from SPM particles. aFb is varied between 8-20
while aheb was fixed at 320, and f was used as the fitting parameter. The values of
f for all stacks is given in Figure 8.12. It is interesting that the ferromagnetic volume
fraction increases rapidly when the number of stacks is increased to 10 and almost
reaches saturation when cS10. This could be due to an improved interface between
the Co and ITO for a larger number of stacks, where the quality of the bilayers im-
proves as the distance from the substrate interface increases and is consistent with the
assumptions made to fit the LAR data. This has also been observed in Co/Cu multi-
layers [6]. It is expected that the SPM phase forms at the interface of the film with
the substrate or the FM/ITO interface. Therefore, it is reasonable to assume that the
SPM volume fraction is related to the roughness and quality of the Co/ITO interfaces.
MCHCH kvrying the Iid spvxer lvyer thixknessA t ITO
[Co(1.20nm)/ITO(tIid)]FE multilayers were all semiconducting over the temperature
range of 5 - 300 K. The observation of semiconducting behaviour for the entire stack
159
Cb[pnel 2. Ci/CTI gofncf[selm
F
M
v
o
lu
m
e
fr
ac
ti
o
n
,
(%
)
f,
Figure 8.12: Ferromagnetic volume fraction, f , as a function of the number of bilayer
repeats, c.
suggests that a high degree of electron scattering between the layers and the higher
resistivity of the ITO layers plays a dominant role in determining the transport prop-
erties of these multilayers.
The temperature evolution of the GMR for a [Co(1.20nm)/ITO(1.65nm)]FE multi-
layer is shown in Fig. 8.13. As shown in the inset of Fig. 8.13, pronounced hysteresis is
seen in the GMR and indeed in the the magnetic hysteresis loops (see later) at 5 - 10 K.
As the temperature begins to rise, the GMR switching field and magnetic coercive field
decrease in tandem, as expected. A similar dependence on temperature is observed in
all [Co(1.20nm)/ITO(tIid)]FE multilayers. The magnitude of the GMR signal also de-
creases as the temperature increases as we would expect, and is also shown in the main
panel of Fig. 8.13 (solid black circles with black line).
To rule out any contribution from AMR, measurements were made having rotated
the in-plane field by 90◦ with respect to the current. The resulting curves for field
parallel or perpendicular to the current were identical, thus eliminating a significant
AMR contribution. The absence of an AMR contribution is not what we would have
expected, regardless of the very low GMR values. We believe that one of the major
origins in the low GMR signal observed is due to magnetic spin frustrations as a result
of a fierce competition between respective magnetic phases. It is possible that these
magnetic frustrations are also responsible for the lack of an AMR contribution to the
total GMR. The frustrations may act to suppress or eliminate spin-orbit coupling in-
teractions between neighbouring Co layers. Alternatively, the absence of a detectable
AMR signal could be related to the lack of clearly defined current paths and so rotating
the field or current with respect to each other would not be expected to have any effect.
160
Cb[pnel 2. Ci/CTI gofncf[selm
Figure 8.13: (a) GMR (closed circles) and resistance (line) as a function of increasing
temperature for a [Co(1.20nm)/ITO(1.25nm)]FE multilayer. Arrows indicate data axes.
Inset (a) shows the GMR of the same stack at 5 (closed circles) and 80 K (closed
diamonds). The GMR at 80 K has been multiplied by a factor of 2.5. Solid lines serve
to guide the eye.
Figures 8.14 (a) - 8.14 (c) illustrate the oscillatory nature of magnetisation sat-
uration field (µE] hAi−b) at 290 K and the magnitude of GMR at both 10 K and
290 K. µE] hAi−b was estimated directly from the magnetisation loops and defined
as the point at which the measured magnetisation signal deviates more than the as-
signed error in magnetisation. The oscillatory nature of the saturation magnetisation
and GMR can be modelled by considering oscillatory exchange coupling due to RKKY
interactions across a NM spacer thickness th;
J(ts) =
JE
tGs
sin
(
2.ts
Λ
+Ψ
)
tsRa
sinh(tsa)
(8.9)
where J E is the interlayer exchange coupling strength (erg/cm
G), Λ and Ψ are the period
and phase of the oscillation and a is an attenuation length [33]. Table 8.1 summarises
the fitting parameters that resulted from the fits for the magnetisation saturation field
and GMR at both 10 and 290 K. Two clear oscillations were identified: the first at 1.65
nm and the second at approximately 3.0 nm. After the second oscillation peak, GMR
becomes very low even at 10 K and begins to fall steadily with increasing tIid. The
oscillations seen in µE] hAi−b and GMR, both at 10 K and 290 K, coincide. A further
weak oscillation is seen however in the µE] hAi−b at approximately 4.0 nm, indicating
an oscillation period of Λ ≈ 1.40 nm. This value is in reasonably good agreement with
that estimated using Eqn. 8.9 (see table 8.1). The discrepancy between the values for
J E are due to differences in the saturation fields between the b -µE] and GMR curves.
From µE] hAi−b we can estimate the AF interlayer exchange coupling energy, JAF ,
161
Cb[pnel 2. Ci/CTI gofncf[selm
0.20
0.30
0.40
0.50
0.60
0.70
1.2 1.8 2.4 3.0 3.6 4.2
0.08
0.12
0.16
0.20
0.4
0.6
0.8
1.0
H
S
A
T
-
M
 
(
T
)
 
 
 
G
M
R
 
(
%
)
10 K
290 K
290 K
 
 
G
M
R
 
(
%
)
t
ITO 
(nm)
 
  
Figure 8.14: Dependence of the magnetisation saturation field, µE] hAi−b , at 290 K
(a), and (b) GMR at 10 and 290 K, as a function of the ITO spacer layer thickness for
[Co/ITO]FE multilayers. The solid lines are fits to Eqn. 8.9.
162
Cb[pnel 2. Ci/CTI gofncf[selm
Table 8.1: Summary of the fitting results and parameters using the RKKY equation
described in Eqn. 8.9 to model the results shown in Fig. 8.14. ∗ represents the fitting
results at 290 K obtained from the magnetisation data.
qemperature TE  	 L p
(K) (erg/cmG) (nm) (nm) (nm) (arb. units)
10 -1.57 1.43 4.23 0.78 0.28
290 -0.36 1.46 4.15 0.88 0.11
290∗ -1.75 1.40 4.05 1.19 0.52
-1.2 -0.8 -0.4 0.0 0.4 0.8 1.2
-1.0
-0.5
0.0
0.5
1.0
 10 K
 290 K
 
 
N
o
r
m
a
l
i
s
e
d
 
M
a
g
n
e
t
i
s
a
t
i
o
n
0
H (T)
Figure 8.15: Normalised magnetisation versus applied field for a multilayer stack with
c = 10 and t Iid = 1.20 nm at 10 (solid black circles) and 290 K (open red circles).
between the Co layers using Eqn. 8.4. The magnitude of JAF follows the same path as
µE] hAi−b as t Iid is varied, with the maximum JAF corresponding to -0.26 erg cm−G
at 290 K for [Co(1.20nm)/ITO(1.65nm)]FE at the peak of the first AF oscillation which
is in good agreement with the value obtained from the fittings using Eqn. 8.9 (-0.36
erg cm−G), and approximately 1.7 times larger than that reported in Co/Cu [5].
Large magnetisation saturation fields, compared to pure thin film Co, combined
with low magnetic remanence and coercive field observed in the [Co(1.20nm)/ITO(tIid)]FE
multilayers, suggest that there is a non-negligible contribution to the magnetotransport
properties of the stacked structures from stray or loose SPM Co particles which are
present in the ITO spacer layer. A comparison of the GMR and magnetisation curves
shows that µE] hAi−Gbg S µE] hAi−b . We attribute this discrepancy in saturation
fields to large interfacial scattering of conduction electrons and the difficulty for mo-
163
Cb[pnel 2. Ci/CTI gofncf[selm
0 50 100 150 200 250 300
0.2
0.3
0.4
0.5
50 100 150 200 250 300
(b)
T (K)
 
 
 ZFC
 FC (0.01T)
M
a
g
n
e
t
i
c
 
m
o
m
e
n
t
 
(
 
1
0
-
3
 
e
m
u
)
T (K)
(a)
 
 
A
C
 
S
u
s
c
e
p
t
i
b
i
l
i
t
y
,
 
A
C
(
a
r
b
.
 
u
n
i
t
s
)
Figure 8.16: (a) Magnetisation versus temperature scan in the field cooled
(FC) (open circles) and zero-field cooled (ZFC) (closed circles) modes for a
[Co(1.20nm)/ITO(1.25nm)]FE stack in an applied field of 0.01 T, and (b) Real com-
ponent of the AC susceptibility, χAX , as a function of temperature for the same
[Co(1.20nm)/ITO(1.25nm)]FE stack. The AC field of 0.787 mT was applied parallel
to the film plane at a frequency of 376 Hz. The measurements were taken in a 0.01 T
DC field.
ments to align in a parallel fashion within respective FM layers due to a high degree
of magnetic frustration.
Figure 8.15 compares the normalised b -µE] loops for a multilayer with c =
10 and t Iid = 1.20 nm at 10 and 290 K. It has already been shown that there is a
certain SPM contribution to these b -µE] loops at higher temperatures. At low tem-
peratures, the b -µE] shows a clear hysteresis with a coercive field of approximately
0.20 T. Therefore, we expect to see some SPM signature in the magnetisation versus
temperature scans. Confirmation of the presence of SPM Co through the ITO spacer
was carried out via standard ZFC-FC magnetisation versus temperature scans from 5
- 300 K in an applied magnetic field of 0.01 T using a SQUID [Fig. 8.16 (a)]. It is
clear that a iB is present at approximately 140 K.
Further evidence for the presence of SPM particles was gathered from measure-
ments of the real component of the AC susceptibility as a function of AC frequency
[Fig. 8.16 (b)]. The peak observed indicates a iB of approximately 210 K. The dif-
ference in values for iB are due to the fact that iB for superparamagnetic particles
depends on the time scale of the experimental technique, and it is therefore expected
to be different in DC (SQUID) and AC measurements. iB may also depend on the
shape and particle size distribution. Using the information from the temperature de-
pendence of magnetisation and AC susceptibility we can apply Eqn. 2.8 to estimate
the Co particle radius (using KXo = 4.9 × 10K ergs cm−H and measurement time t = 10
s). Assuming spherical particles, this yields a value of approximately 1.30 nm. Other
multilayers with different ITO spacer thicknesses also produced similar results.
164
Cb[pnel 2. Ci/CTI gofncf[selm
SPM
SPM
NM(ITO)
FM (Co)
FM (Co)
FM FM FM SPM
FMSPM
SPM SPM
Decoupling region
Figure 8.17: Schematic illustration of the SPM isolated island model. The sold-line
and dashed-line straight arrows indicate the directions of the magnetisations lying in
the plane of the FM Co layer. The significance of the shorter dashed arrows refers
to the fact that the magnetisations in successive layers are not aligned in antiparal-
lel fashion. The multi-arrow stars within the SPM regions represent the temporary
fluctuations of the orientation of their magnetic moments. Wavy arrows indicate the
possible trajectories of electrons between FM and SPM regions that can contribute to
the GMR. Adapted from [35].
The bulk characterisation features for identifying a magnetically frustrated system
such as a spin glass (recall section 2.1.5) or superparamagnet are very similar, however,
we are confident that the phenomena observed in these multilayers are of superparam-
agnetic origin. Spin glasses are usually alloyed materials, e.g. CuF−xMnx, whereas in
the case here, we have ultrathin Co layers separated by ITO. At most we are likely to
have discontinuous Co layers forming islands or stray Co particles dispersed within the
ITO spacer (see section 8.3.4). Also, spin glasses will usually show a cooperative phase
transition [34], while superparamagnets show a gradual blocking of the superparamag-
netic particles. The freezing temperature of spin glasses is rarely as large as the peak
points in Fig 8.16 which serves as further evidence for superparamagnetism. The best
way to distinguish between the two would be to perform muon spin rotation, relaxation
and resonance measurements, where a spin glass will give very specific relaxation times.
MCHCI Xonsiyervtions of the possiwle spvtivl yistriwution of suB
perpvrvmvgnetix pvrtixles
The results above unambiguously confirm the presence of SPM regions in these multi-
layers, which are believed to be decoupled from the FM parts of the magnetic layers.
However, it is not easy to determine the location of the magnetically separated SPM
regions. Based on models by Bakonyi zt vl. [35] it is believed that two possible scenarios
can be proposed to explain the SPM contributions to both the magnetic and magneto-
transport behaviour in these multilayers: the isolvtzy islvny moyzl and the intzrfvxivl
looszBspin moyzl (analogous to the model described by Slonczewski [15,17]).
165
Cb[pnel 2. Ci/CTI gofncf[selm
SPM
SPM SPM
SPM SPM
SPM
FM FM
NM(ITO)
FM SPM
SPM SPM SPM SPM SPM SPM FM FMFM FM
FM SPM FM SPM
FM (Co)
FM (Co)
Decoupling region
Low-field High-field
(a) (b) (c)
Figure 8.18: Schematic illustration of the loose-spin model, where SPM regions are
present at the interfaces between FM and NM layers at low (a), intermediate (b) and
high (c) applied magnetic fields. The features can all be defined as in Fig. 8.16.
The star-like arrow arrangements indicate the random temporal fluctuations in SPM
particle magnetisation which gradually align with higher magnetic fields (c). Solid
wavy arrows indicate electron paths which will contribute to the magnetoresistance,
whereas those which are dotted represent ineffective paths. Adapted from [35].
The possibility of intermixing between the Co and ITO layers has been proven in
section 8.3.1. In the extreme case, this may lead to the formation of small magnetic
islands separated by non-magnetic material from the FM part of the magnetic layer.
These separating regions may be alloys of Co-ITO and may be non-magnetic. These
are referred to as decoupling regions between the FM and SPM parts of the FM layer.
Due to their probable nanoscale size, the isolated islands may exhibit SPM behaviour.
A schematic illustration of the dynamics involved within this model scenario is depicted
in Fig. 8.17. Note that this is an over-simplified picture, and that possible fluctuations
in FM and NM layer thicknesses have been neglected, which is likely to not be the
case.
The other possibility is to assume the presence of loosz spins located between FM
and NM layers. This is depicted in Fig. 8.18 where each of the possible electron paths
leading to spin-dependent scattering are specified. At low applied magnetic fields,
where there is partial antiparallel alignment of the weakly coupled or non-coupled
FM layers, the resulting GMR curve will be a result of FM→SPM (or SPM→FM)
and FM→FM transitions, whereas at higher fields, where the FM layers and SPM re-
gions are strongly aligned (above the magnetic saturation field), only FM→SPM and
SPM→FM transitions remain effective. The intermixing between the Co and ITO
layers could potentially result in the formation of such interfacial loose-spin regions,
adding to the magnetic frustration throughout the FM Co layers, and thus the overall
low values for GMR.
Therefore, it is hard to convince oneself that the layers in these multilayer struc-
tures are strictly continuous. If the interfacial layers were continuous, it is difficult to
understand how the surface spins can be decoupled from the rest of the magnetic layer.
166
Cb[pnel 2. Ci/CTI gofncf[selm
MCHC5 Vnnevling effexts on stvxk mvgnetoresistvnxe
We have found that the electrical properties of ITO are very sensitive to the temper-
ature at which it is grown (chapter 4). Our studies show that an increasing growth
temperature results in ITO films with lower resistivity (due to increasing carrier con-
centration and mobility), and is in agreement with others who have found similar
trends [36]. In an attempt to improve the electrical conductivity of the ITO such as to
reduce the resistivity mismatch between Co and ITO, a [Co(1.20nm)/ITO(3.00nm)]GE
stack was annealed using a heater temperature of 300 ◦C under UHV conditions for
2 hours. Total stack resistivity at 10 K was reduced from 7.2×10−IΩcm for the as-
deposited stack to 2.1×10−IΩcm upon annealing, yet still follows the path of a semi-
conducting system over a range of temperatures. However, as we anticipated, there has
been significant interdiffusion between the Co and ITO layers, as confirmed by LAR
measurements before and after annealing treatment.
The LAR spectra shown in the inset of Fig. 8.19 (a) show reduced clarity in the
Keissig fringes and major reduction in the Bragg peak intensities. In fact, the second
Bragg peak has disappeared in the case of the annealed sample. This clearly indicates
interdiffusion as a reduction in Bragg peak intensity would normally signify a change
in density. The density of Co (8.79 g cm−H) is larger than that of ITO (7.11 g cm−H),
and upon mixing, is likely to have increased the density of ITO. The improved con-
ductive properties are likely to be due to both the improved crystalline quality of the
ITO layers as well as the probable creation of metallic Co conducting paths which run
through the ITO layers.
Annealing resulted in a significant reduction in the GMR from 0.25% to 0.05% [Fig.
8.19 (b)]. The reduction in GMR is likely to be due to the destruction of the interfaces
upon annealing. The total saturation magnetisation also fell upon annealing, and again
is likely to be linked to the breakdown of the interfaces
MCI hummvry vny Xonxlusions
We have systematically studied the transport and magnetic properties of Co/ITO mul-
tilayers as a function of the number of Co/ITO repeats and semiconducting ITO spacer
layer thickness. We have shown that alternating Co and ITO layers can produce a gi-
ant magnetoresistance effect which persists above room temperature. The magnetic
hysteresis loops imply a significant contribution arising from superparamagnetic parti-
cles along with a ferromagnetic phase. Using a simple magnetic model where we have
included the contribution made by superparamagnetic particles, we were able to sep-
arate the different magnetic phase constituents. The occurrence of multiple magnetic
contributions in our multilayers suggests that there is a high degree of magnetic frus-
tration throughout the ultrathin Co layers. As the thickness of the non-magnetic ITO
167
Cb[pnel 2. Ci/CTI gofncf[selm
G
Figure 8.19: (a) LAR measurements for an as-deposited (black) and annealed (red)
(iAnn = 300
◦C) [Co(1.20nm)/ITO(3.00nm)]GE stack. (b) corresponding GMR curves
for the same stacks at 10 K.
is varied between 1.20 - 4.20 nm both the magnetisation saturation field and GMR
oscillate with the ITO thickness implying RKKY interaction.
The quality of the Co/ITO interfaces and layers is still unknown. Due to high
levels of spin disorder within the thin Co layers, it is highly likely that there is a signif-
icant amount of spin-independent/flipping scattering at the Co/ITO interfaces. This
results in the interfacial spin-dependent scattering averaging to zero, thus giving little
or no contribution to the GMR. High resolution cross-section TEM would certainly
shed some light on the situation. The possibility of magnetically isolated SPM islands
within the magnetic layer plane and SPM loose-spins present at the interfaces between
the FM and NM layers has been discussed, however, locating these regions is difficult.
168
Wiwliogrvphy
[1] M. N. Baibich, J. M. Broto, A. Fert, F. Nguyen Van Dau, F. Petroff, P. Etienne,
G. Creuzet, A. Friederich, J. Chazelas, Phys. Rev. Lett. 61, 2472, (1988).
[2] P. Grunberg, R. Schreiber, Y. Pang, M. O. Brodsky, and H. Sowers, Phys. Rev.
Lett. 57, 2442, (1986).
[3] S. S. P. Parkin, N. More, K. P. Roche, Phys. Rev. Lett. 64, 19, 2304, (1990).
[4] S.S.P. Parkin, Phys. Rev. Lett., 67, 3598, (1991).
[5] S.S.P. Parkin, R. Bhadra, K. P. Roche, Phys. Rev. Lett. 66, 2152 (1991).
[6] A. Paul, T. Damm, D. E. Burgler, S. Stein, H. Kohlstedt, P. Grunberg, J. Phys.
: Condens. Matter, 15, 2471-2491, (2003).
[7] N. F. Mott, Proc. R. Soc. A 153, 699, (1936).
[8] A. Fert, P. Grunberg, A. Barthelemy, J. Magn. Magn. Mater. 140, 1, (1995).
[9] R. E. Camley, J. Barnas, Phys. Rev. Lett. 63, 664-667, (1989).
[10] P. M. Levy, S. F. Zhang, A. Fert, Phys. Rev. Lett. 65, 1643-1646, (1990).
[11] T. Valet and A. Fert, Phys. Rev. B 48, 7099, (1993).
[12] P. Bruno, C. Chappert, Phys. Rev. Lett. 67, 1602-1605, 2592, (1991); P. Bruno,
C. Chappert, Phys. Rev. B. 46, 261-270, (1992).
[13] P. Lang, L. Nordstrm, R. Zeller, P.H. Dederichs, Phys. Rev. Lett. 71, 1927-1930,
(1993); L. Nordstrm, P. Lang, R. Zeller, P.H. Dederichs, Phys. Rev. B. 50, 13058-
13061, (1994).
[14] P. Bruno, J. Magn. Magn. Mater. 121, 248-252, (1993); P. Bruno, J. Phys.: Con-
dens. Matter. 11, 9403-9419, (1999).
[15] J. C. Slonczewski, Phys. Rev. Lett. 67, 3172-3175, (1991).
[16] S. Demokritov, E. Tsymbal, P. Grunberg, W. Zinn, I. K. Schuller, Phys. Rev. B
49, 720-023, (1994).
169
Bcbfcial[pbs
[17] J. C. Slonczewski, J. Appl. Phys. 73, 5957-5962, (1993).
[18] I. Bakonyi, L. Pter, Z. Rolik, K. Kiss-Szab, Z. Kupay, J. Tth, L. F. Kiss, J. Pdr,
Phys. Rev. B, 054427, (2004).
[19] N. Wiser, J. Magn. Magn. Mater. 159, 119 (1996).
[20] B.J. Hickey, M.A. Howson, S.O. Musa, and N. Wiser, Phys. Rev. B 51, 667 (1995).
[21] A. Cziraki, L. Peter, V. Weihnacht, J. Toth, E. Simon, J. Padar, L. Pogany, C.M.
Schneider, T. Gemming, K. Wetzig, G. Tichy, I. Bakonyi, J. Nanosci. Nanotechnol.
6, 2000 (2006).
[22] T. Yu, X. Li, D. Li, S. Hao, L. Wang, Z. Zhang, G. Wu, X. Zhang, Q. Li, P. Chen,
J. Magn. and Mag. Mat., 320, 2185-2189, (2008).
[23] L. Lu, G. Lu, Z. Zhang, C. Gao, T. Yu, P. Chen, Solid. State. Comm. 149, 47-48,
(2009).
[24] O. Proux, J. S. Micha, J. R. Regnard, A. Traverse, B. Dieny, F. Ernult, P.
Bayle-Guillemaud, J. LHazemann, L. Giacomoni, J. Phys.: Condens. Matter 15,
72377252, (2003).
[25] K. Oguz, P. Jivrajka, M. Venkatesan, G. Feng, J. M. D. Coey, J. Appl. Phys. 103,
07B526, (2008).
[26] H. Kano, K. Kagawa, A. Suzuki, A. Okabe, K. Hayashi, K. Aso, Appl. Phys. Lett.
63, 2839, (1993).
[27] K. Fuchs, Proc. Camb. Phil. Soc. 34 100, (1938).
[28] E. H. Sondheimer, Adv. Phys. 1, 1, (1952).
[29] C. H. Marrows, M. Perez, B. J. Hickey, J. Phys.: Condens. Matter 18, 243, (2006).
[30] J. Xu, M. A. Howson, B. J. Hickey, D. Greig, E. Kolb, P. Veillet, N. Wiser, Phys.
Rev. B. 55, 1, (1997).
[31] D. C. Jiles, J. Magn. and Mag. Mat., 242-245, 116-124 (2002).
[32] M. Mordjaoui, B. Boudjema, M. Chabane, R. Daira, J. Computer Sci. 3(6): 399-
405 (2007).
[33] J. Dubowik, B. Szymanski, F. Stobiecki, K. Roll, Phys. Stat. Sol. (a), 196, 86,
(2003).
[34] S. Blundell, Magnetism in Condensed Matter, Oxford University Press, Oxford,
p. 93, (2001).
170
Bcbfcial[pbs
[35] I. Bakonyi, J. Toth, L. F. Kiss, E. Toth-Kadar, L. Peter, A. Dinia, J. Magn. Magn.
Mater. 269, 156-167, (2004); I. Bakonyi, L. Peter, V. Weihnacht, J. Toth, L. F.
Kiss, C. M. Schneider, J. Optoelectron. Adv. Mater. 7, 589, (2005).
[36] H. Kim, C. M. Gilmore, A. Pique, J. S. Horowitz, H. Mattoussi, H. Murata, Z. H.
Kafafi, D. B. Chrisey, J. Appl. Phys. 86, 11, (1999).
171
Xhvpter N
XuBwufferey pXoDIidr multilvyers
lz nofi invzstigvtz thz influznxz of inszrting vn vyyitionvl F nm thixk Cu wuffzr lvyzr
vt zvxh CoDIid intzrfvxz fiithin thz multilvyzrs yzsxriwzy in xhvptzr MC In pvrtixulvr
fiz look vt hofi this xvn vffzxt thz mvgnztotrvnsport wzhvviour of thz multilvyzrsC Gbg
riszs immzyivtzly upon thz inszrtion of Cu wuffzr lvyzrs vny is vttriwutzy to intzrfvxivl
yoping vt thz CoDCu intzrfvxzC Vnnzvling rzsults in v lvrgz inxrzvsz in GbgA wzlizvzy
to wz v rzsult of grvyuvl Cu yiffusion into thz Iid lvyzrs fihixh forms v grvyzy intzrB
fvxz wztfizzn vyjvxznt Cu vny Iid lvyzrsC [urthzr vnnzvling vt progrzssivzly highzr
tzmpzrvturzs rzsultzy in v grvyuvl yzxlinz in Gbg vs thz intzrfvxzs wrzvk yofinC
172
Cb[pnel 3. Co-boffeled [Ci/CTI] gofncf[selm
NCF Introyuxtion
Following on from the results discussed in chapter 8, a new component is now added
to the [Co/ITO] multilayer structure; Cu buffer layers have been inserted at each and
every Co/ITO interface with a view to prevent the undesirable diffusion of magnetic
impurities into the NM spacer layer. Co/Cu multilayers and interfaces are one of the
most extensively studied systems and so are considered compatible for the purposes of
this investigation. They have been reported to form high quality interfaces with little
evidence of intermixing. However, this is very much dependent on the preparation
technique used to make the multilayers. Already in chapter 8, the models described
for the SPM contribution to GMR (in section 8.3.4) were originally based on Co/Cu
multilayers which had been deposited using an electrodeposition process [1].
NCG Zflperimentvl gesults
A series of [Co/Cu/ITO/Cu]GE films were prepared on unheated r -plane sapphire sub-
strates by DC magnetron sputtering under identical conditions (chamber base pres-
sure, sputtering pressure, target power, etc.) to those described in chapter 8 (ITO(10)
composition). The deposition rate for the Cu target was 0.10 nm/W/pass at 1 rpm
calibrated at 1.50±0.01 Pa. The respective layer thicknesses were 1.30±0.10 nm for
Co, 3.00±0.10 nm for ITO and 1.00±0.10 nm for the additional Cu buffer layers. UHV
annealing took place in the same system in which the samples were prepared at various
heater temperatures (iAnn). Each annealing stage lasted 1 hour.
NCGCF bvgnetotrvnsport mevsurements vny mvgnetometry
Figure 9.1 shows a selection of GMR curves for as-deposited and annealed multilayers
with and without additional Cu buffers. The insertion of a 1 nm Cu buffer layer at each
Co/ITO interface led to an immediate increase in GMR in the as-deposited samples,
from 0.25% to 0.64% at 10 K and 0.11% to 0.44% at room temperature (Fig. 9.2,
dotted encasing and arrow). Annealing these structures in a UHV environment led to
a steady rise in the GMR signal at both 10 and 290 K. A maximum GMR of 2.18% at
10 K was measured after using a heater temperature of 300◦C. Further annealing led
to a gradual decline in GMR. These results are summarised in Fig. 9.2.
Similar results were obtained in Co/Cu multilayers after annealing, however, a
sharper fall in GMR was observed for iAnn S 400
◦C until GMR was completely de-
stroyed when annealed above 700◦C [2]. Khalyapin zt vl. however, observed an im-
mediate drop in GMR in their Co/Cu multilayers if annealed at iAnn = 100
◦C for
just 15 minutes [3]. Annealing for a longer period of 1.5 hours resulted in a complete
disappearance of the GMR. My preliminary results on Co/Cu multilayers also show an
173
Cb[pnel 3. Co-boffeled [Ci/CTI] gofncf[selm
-2 -1 0 1 2 3 4 5
0.0
0.5
1.0
1.5
2.0
2.5
 
0
H (T)
 
M
R
 
(
%
)
 As-dep [Co/ITO]
20
 As-dep [Co/Cu/ITO/Cu]
20
  150
o
C 
 300
o
C 
 450
o
C 
 
 
Figure 9.1: Magnetoresistance curves for an as-deposited [Co/ITO]GE (from chapter
8) and as-deposited [Co/Cu/ITO/Cu]GE multilayers at 10 K and corresponding mag-
netoresistance curves for Cu-buffered multilayers which have been annealed at various
iAnn.
immediate fall in GMR upon annealing for 1 hour using iAnn = 100
◦C.
Any contributions from AMR were dismissed having rotated the in-plane field by 90◦
with respect to the current as was performed for the [Co/ITO] multilayers in chapter 8.
Magnetometry measurements show that the absolute moment of the Cu-buffered
multilayers increases gradually upon annealing up to a heater temperature of 300◦C,
after which it saturates. This suggests that the magnetic properties of the Co layers
do not degrade upon annealing. A test on a capped 5 nm film of pure Co was carried
out where we simply measured the magnetisation before and after UHV annealing of
a film grown under identical conditions to those described for the multilayers. Simi-
larly, the 5 nm Co film was annealed at three different heater temperatures (150, 300
and 450◦C). After the first annealing stage, the saturation magnetisation increased by
approximately 14%, after which it saturated even after two further anneals. In the
[Co/ITO] multilayer stacks where no Cu buffer was present (chapter 8), annealing led
to a reduction in the saturation magnetisation, likely to be due to the promotion of
intermixing as a result of oxygen diffusion from the ITO towards the Co layer. In the
scenario where Cu is present, we believe that the Cu buffer layer acts as a barrier to
prevent oxygen from reaching the Co layer and vice versa.
174
Cb[pnel 3. Co-boffeled [Ci/CTI] gofncf[selm
TAnn
Figure 9.2: Magnetoresistance as a function of iAnn in [Co/Cu/ITO/Cu]GE multilayers
at both 10 K (solid black circles) and 290 K (open red circles). The shaded region shows
the magnetoresistance measured in [Co/ITO]GE multilayers for an as-deposited sample
and after annealing at iAnn = 300
◦C (solid blue circles). The arrow shows the increase
in magnetoresistance upon addition of the Cu buffers.
NCGCG htruxturvl properties
The dramatic decrease in GMR upon annealing of the [Co/ITO]GE multilayers (chap-
ter 8) was attributed to the subsequent breakdown of Co/ITO interfaces, confirmed
by the observation of suppressed Bragg reflection peaks and Keissig fringes by LAR
measurements. This was considered sufficient evidence to suggest that there was some
level of interdiffusion between Co and ITO layers resulting in a more granular system.
θ-2θ scans [Fig. 9.3 (a)] for the Cu-buffered multilayers revealed that the as-
deposited stacks were amorphous, however, on annealing, they became increasingly
crystalline demonstrated by the emergence of a strong ITO (222) peak which becomes
stronger as the annealing temperature increases [Figs. 9.3 (b) and (c)]. There were no
Bragg peak reflections which were identified as belonging to Cu or Co after any of the
annealing stages. The average lattice parameter of the ITO estimated at each stage
of annealing was 1.010 nm, which is relatively small compared to the typical value
of 1.012 nm for ITO. This suggests that ITO grown on top of Cu and Co is highly
strained. A systematic shift towards lower incident angles of the ITO (222) would
have implied a systematic change in lattice parameter, providing clues regarding the
possible diffusion between Cu and ITO. Unfortunately, no information regarding this
issue could be obtained from our scans.
Figure 9.4 shows the LAR pattern obtained for an as-deposited and annealed
[Co/Cu/ITO/Cu]GE multilayer. In the as-deposited samples, there are no signs of Kies-
175
Cb[pnel 3. Co-boffeled [Ci/CTI] gofncf[selm
S1 S2(a)
(b) (c)
Figure 9.3: Effects of heater temperature on [Co/Cu/ITO/Cu]GE multilayers, (a) θ-2θ
diffractograms, (b) expanded view of the θ-2θ diffractograms at the ITO (222) reflection
peak, and, (c) ITO (222) reflection peak intensity normalised to the first substrate peak
(S1).
176
Cb[pnel 3. Co-boffeled [Ci/CTI] gofncf[selm
0.5 1.0 1.5 2.0 2.5 3.0 3.5
1.0 1.5 2.0
 As-deposited
 300
o
C
 
L
n
 
i
n
t
e
n
s
i
t
y
 
(
a
r
b
.
 
u
n
i
t
s
)
 Incident angle (Degs)
(a)
 Incident angle (Degs)
 
 
L
n
 
i
n
t
e
n
s
i
t
y
 
(
a
r
b
.
 
u
n
i
t
s
)
 
 
 
 
Figure 9.4: Low-angle reflectivity scans for an as-deposited (black circles) and an-
nealed (300◦C) (red circles) [Co/Cu/ITO/Cu]GE multilayer. Inset: simulated data of
the [Co/Cu/ITO/Cu]GE multilayer.
sig fringes, however, there are pronounced first, second and third order Bragg reflection
peaks, a good indicator of a well-defined multilayer structure. It is not until the third
annealing stage (300◦C) that Keissig fringes become clear and the LAR pattern ob-
served resembles the simulated pattern (Fig. 9.4, inset). Further annealing resulted in
the gradual disappearance of the Keissig fringes.
NCGCH Yisxussion
To investigate what may be happening upon annealing in these multilayers, a series of
Cu-doped (3.3 - 16.0 at.%) ITO(10) and undoped ITO(10) thin films were grown. These
films were grown under the same conditions as the multilayers, i.e. on unheated r -plane
sapphire substrates. The room temperature transport properties of these films are
summarised in Fig. 9.5; the shaded regions show the changes in the transport properties
having been annealed in UHV using a heater temperature of 300◦C. Surprisingly, the
transport properties of ITO deteriorated upon Cu doping in the room temperature as-
grown samples where we see a rise in total film resistivity, a fall in carrier mobility and
minimal changes in carrier concentration with increasing Cu doping levels. All as-grown
Cu-doped ITO films were semiconducting, as determined from standard resistance
versus temperature measurements, however, the pure, undoped ITO sample showed
metallic behavior. In addition, the as-grown Cu-doped ITO films were all amorphous,
177
Cb[pnel 3. Co-boffeled [Ci/CTI] gofncf[selm
Figure 9.5: Summary of the transport properties [carrier concentration, nx, (black cir-
cles), carrier mobility, µH , (red circles) film resistivity, /, (blue circles)] of as-deposited
Cu-doped ITO(10) films grown using dc magnetron sputtering at room temperature.
The shaded regions represent the values after annealing using iAnn = 300
◦C (data
points now solid squares with the same colour scheme).
whereas the pure ITO sample showed a very weak ITO (222) peak. Upon annealing, the
Cu-doped ITO films become more crystalline, showing the ITO (222) peak, however,
there were still no signs of any Cu peaks which implies that the Cu and ITO may have
now mixed. Therefore, it is highly likely that the Cu and ITO mix upon annealing in
the multilayers at heater temperatures of at least 300◦C. The electrical properties of
both the pure ITO and 8.0 at.% Cu-doped ITO improve upon annealing (solid squares
inside the shaded region of Fig.9.5).
In the as-deposited state, the rise in GMR of the multilayers may be related to
interfacial-doping of the Co interface by the additional Cu buffer layer [4, 5]. The
structural quality of the stack, in particular, the roughness, plays an important role
in determining the continuity across the layers. Layer fluctuations provide pathways
for interdiffusion or penetration of stray atoms into unwanted regions. It has been
shown in Co/Cu-based spin valves and multilayers that Cu impurities within the Co
layer close to the interface can lead to a rise in GMR. Conversely, Co impurities in the
Cu spacer layer strongly reduce the GMR signal if they are located close to the inter-
face [5]. Therefore, these effects are penetration depth dependent, and so the quality
of the layers in our stacks is of vital importance.
It may be possible that the mixing between the Cu and ITO layers leads to a
graded interface so that an ultra-thin layer of Cu-doped ITO exists at the Cu/ITO
interface. This results in improved electrical properties and the likelihood of reduced
bulk scattering of electrons due to a decrease in the resistivity mismatch between
178
Cb[pnel 3. Co-boffeled [Ci/CTI] gofncf[selm
As-deposited
Earlyanneal Increase in
TAnn
Sharpening of
interfaces
Graded interfaces
Increase in
TAnn
Complete mixing of
Cu and ITO layers
Total interfacial
breakdown
Increase in
TAnn
Co
Co
Cu
Cu
ITO
Figure 9.6: Illustrated interpretation of the influence of Cu buffer layers at the Co/ITO
interface and the subsequent effects of annealing.
Cu, Cu-ITO and ITO. In addition, the annealing process results in greater crystalline
quality in each layer which will undoubtedly result in greater electron transport.
From the LAR measurements, we see that the Cu-buffered multilayers have pro-
nounced Keissig fringes only after a particular annealing temperature, which coincides
with the maximum value recorded for GMR. This result is similar to Hecker zt vl.
who argue that the initial rise in GMR in their Co/Cu multilayers correlates with the
sharpening of the interfaces [2]. Thereafter, with further annealing, the features in
the LAR measurements begin to deteriorate, as does the magnetoresistance due to the
breakdown of interfaces. If the Cu/ITO interfaces have now mixed completely, it is
highly likely that the Co atoms will begin to migrate towards the neighboring oxide
when annealed. The Co may either remain as stray, isolated Co atoms or become
clusters within the oxide layer, interacting superparamagnetically, therefore leading to
spin-independent scattering, and thus a reduction in GMR. Fig. 9.6 provides an illus-
trative summary of these points.
NCH hummvry vny Xonxlusions
The combination of interfacial-doping at the Co/Cu interface and the creation of the
graded Cu/Cu-ITO interface along with improved magnetic properties of the Co layers
upon annealing leads to enhanced GMR signal in Cu-buffered [Co/ITO]GE multilayers.
Further annealing leads to a breakdown of the interfaces and a reduction in the mea-
sured GMR signal.
This work still has a great deal of unanswered questions particularly regarding the
179
Cb[pnel 3. Co-boffeled [Ci/CTI] gofncf[selm
state of the interfaces at each annealing stage. Therefore, the conclusions we make
here are speculative and based only on the limited knowledge we have available to us
at the current time.
180
Wiwliogrvphy
[1] I. Bakonyi, J. Toth, L. F. Kiss, E. Toth-Kadar, L. Peter, A. Dinia, J. Magn. Magn.
Mater. 269, 156-167, (2004); I. Bakonyi, L. Peter, V. Weihnacht, J. Toth, L. F.
Kiss, C. M. Schneider, J. Optoelectron. Adv. Mater. 7, 589, (2005).
[2] M. Hecker, J. Thomas, D. Tietjen, S. Baunack, C. M. Schneider, A. Qiu, N.
Cramer, R. E. Camely, Z. Celinski, J. Phys. D: Appl. Phys. 36, 564, (2003).
[3] D. L. Khalyapin, P. D. Kim, J. Kim, I. A. Tupanov, A. Y. Beten’kova, G. V.
Bondarenko, T. N. Isaeva, I. Kim, Physics of the Solid State, 52, 1787, 9, (2010).
[4] M. Perez, C. H. Marrows, B. J. Hickey, Europhys. Lett., 5I, 262-268, (2001)
[5] C. H. Marrows and B. J. Hickey, Phys. Rev. B, KH, 220405(R), (2001)
181
Xhvpter FE
Zstimvting the hpin Yiffusion
aength of hemixonyuxting Inyium
iin dfliye
ihis xhvptzr is xonxzrnzy fiith zflploring thz fzvsiwility of intzgrvting Iid fiithin spinB
tronix yzvixzsC lz wzgin wy introyuxing szmixonyuxting Iid thin films into xonvznB
tionvl spin vvlvz struxturzs vny proviyz vn ovzrvizfi of thz stzps involvzy to fvwrixvtz
xurrzntBpzrpznyixulvrBtoBplvnz nvnopillvr junxtionsC [rom this fiz hvvz zstimvtzy thz
room tzmpzrvturz spin yiffusion lzngth of szmixonyuxting IidC ihz nzflt stvgz involvzs
thz inszrtion of tfio thin Cu wuffzr lvyzrs on zithzr siyz of thz Iid spvxzr lvyzr in
thz spin vvlvz rzsulting in v mvrkzy improvzmznt in thz GbgC [invllyA fiz vttzmpt to
suwstitutz thz mztvllix fzrromvgnzt in thz spin vvlvz struxturz fiith v yilutz mvgnztix
ofliyzC
182
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
FECF Introyuxtion
In FM/NM multilayers and spin valves (SV) with FM/NM/FM structure, the magni-
tude of the GMR strongly depends on the spin scattering asymmetries of the FM layer
and FM/NM interfaces as well as the length over which an electron spin information
is conserved throughout the NM spacer layer [1]. The electron spin diffusion length
of the NM spacer, λsyl, in thin films is a fundamental measure of spin conservation
and ultimately determines the performance of magnetic sensor-based devices. Valet
and Fert produced a phenomenological model (VF model) which describes CPP-GMR
from Boltzmann’s equations [1]. As explained in the VF model of CPP-GMR, spin po-
larised currents flowing perpendicularly to the layers induce spin accumulation effects
and the final result is that the length scale governing the thickness dependence is λsyl.
Finding an all-purpose NM semiconducting oxide to build spintronic devices is
the first and most difficult challenge. The material should have high electron carrier
mobility and bipolar transport properties. In addition, it should be structurally and
chemically compatible with the neighbouring layers. As we have shown in chapters 5
through 7, further doping of ITO with various transition metals has now been widely
studied with reports of ferromagnetism above room temperature. However, in addi-
tion to ferromagnetism at room temperature, practical implementation of FM ITO
into spintronic devices requires minimal interfacial spin scattering and long λsyl. The
observation of oscillatory GMR in [Co/ITO]FE multilayers in chapter 8 at room temper-
ature and spin injection in CoFe-ITO based junctions [2] indicates a coherent transfer
of spin information through thin layers of ITO; however, the length scale over which
efficient spin information can be transmitted through ITO is still unclear with little
understanding of the spin dynamics at the FM/ITO interface. Provided that the spin
polarisation is maintained across the ITO layer, a spin valve effect is expected: po-
larised carriers injected from one electrode reach the other electrode and the resistance
of the heterostructure depends on the relative orientations of the electrode magnetisa-
tion.
FECG hpin kvlves
FECGCF hpin kvlve struxtures
The fields required to rotate the magnetic moments from parallel to anti-parallel (or
vice-versa) in chapters 8 and 9 are relatively large and unsuited for use in magnetic
memory applications. Dieny zt vl. provided a solution to this dilemma [3]. Instead of
relying on interlayer exchange coupling to achieve parallel/antiparallel magnetisation
configurations, Dieny zt vl. employed a structure where two FM layers were yzxouplzy
from each other via the use of a relatively thick NM spacer layer. In this scenario,
the saturation fields could effectively be reduced to that of the individual FM layers.
183
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
Substrate
FM(free)
FM (fixed)
AF (pinning)
Capping
NM Spacer
Buffer
(a)
(b)
(c)
Figure 10.1: (a) Schematic illustration of a typical spin valve structure. The core
components are those of the fixed and free ferromagnetic layers either side of a non-
magnetic spacer. Those layers outlined with dotted lines are additional components
to the structure which are often utilised, however, are not used in this study; (Right)
Simplified representation of the spin accumulation at an interface between a ferromag-
netic metal and a non magnetic layer assuming no interfacial resistances, (b) Incoming
and outgoing spin-up and spin-down currents (c) Splitting of the chemical potentials,
EFup and EFyown, in the interface region (spin accumulation). The arrows symbolise
the spin flips induced by this out of equilibrium distribution. These spin-flips govern
the progressive depolarisation of the current. asf
F and asf
c are the spin diffusion
lengths of the ferromagnetic and non-magnetic layers, respectively. (b) and (c) taken
from reference [5].
This structure is illustrated in Fig. 10.1 (a) and is known as a simple spin-valve (SV).
Differential scattering of electrons in FM layers gives rise to a GMR signal in these
heterostructures in much the same way as the multilayers.
There are a number of variations on the SV structure. If the two FM layers are
made of different materials (a pseudo-spin-valve) the different coercive fields can be
exploited: the softer FM layer can switch its magnetisation with a relatively small
applied field, with the magnetisation of the harder material unaffected. This gives rise
to the antiparallel state with higher resistance. In this case the spacer layer is relatively
thick, and is used to decouple the ferromagnetic layers to prevent them from switching
at the same field. A detailed review of CPP SVs can be found in reference [4].
184
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
FECGCG hpin vxxumulvtion effexts
Spin accumulation effects govern the propagation of a spin-polarised current through
any succession of magnetic and nonmagnetic materials. Figures 10.1 (b) and (c) pro-
vide a simplified illustration of the principles behind spin accumulation which occurs
when electrons cross a FM/NM interface. In Fig. 10.1 (b) the incoming electrons
are predominantly of the spin up direction whereas the outgoing electrons are carried
equally by both up and down spins. As a result, there is an accumulation of spin up
electrons at the interface which diffuses on both sides of the FM/NM interface to a
distance of the order of the λsyl. This process can also be described in terms of the
splitting of spin up and spin down Fermi energies [see Fig. 10.1 (c)]. The imbalance
of electron distribution in the spin accumulation zone generates spin-flips and provides
the mechanism for the adjustment between the incoming and outgoing spin currents.
The spin polarisation of current decays exponentially as it travels through the spin
accumulation zone.
To summarise, there is a broad zone of spin accumulation which extends on both
sides of the ferromagnetic layer to the distance of λsyl and in which the current is
progressively depolarised by the spin-flips generated by the spin accumulation.
FECGCH Xontrolling hpin kvlve performvnxe
To review all the work performed on SV structures in recent years would be outside the
scope of this thesis. The intense research surrounding this area is naturally boosted
by the huge prospects and potential of employing these structures in magnetic storage
technology and miniaturised devices. Below is a summary of some of the important
parameters which can influence the GMR performance of the spin valve structures.
If the magnetic layers are made thinner, it has been reported that the GMR de-
creases [6] and is found for multilayer and spin valve structures.
A discussion on multilayer or stacked structure systems would never be complete
without considering the quality of the microstructure at the interfaces. As we have
already discussed in chapter 8, rough samples are generally accompanied by the pres-
ence of interfacial mixing. This is particularly relevant for sputtered films, where
grain boundaries may act as transport channels for diffusion across different layers. In
general, the scenario where there is interfacial intermixing leads to a situation where
momentum scattering (spin-independent scattering) dominates leading to a reduction
in device performance.
An ideal spin valve structure will have FM layers which are totally decoupled from
each other, but due to microstructural defects and imperfections in the FM layers or
at the interfaces, this is never achieved in practice. Magnetic bridges or pinholes result
as a consequence of these imperfections leading to FM coupling between the layers.
Stray fields emanating from domain walls of a FM layer can act as local barriers to
185
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
Figure 10.2: An SEM micrograph of a Py/Cu/Py spin valve (inset shows a finite
element simulation of the equipotential lines across the device). The right hand panel
shows the compiled GMR results of all the devices with a Valet-Fert model fit (solid
line). The inset shows typical CPP GMR curves of a selection of spin valves with
differing Cu spacer layer thickness. After reference [7].
the wall motions in the adjacent FM layers, resulting in coupling between FM layers
across spacer layers.
FECH erevious lork
The use of CPP measurements on sub-micron nanopillar spin valves to estimate the
spin diffusion length of a non-magnetic spacer layer material has been done in the
past. Through measurements of NiMEFeGE(Py)/Cu/Py CPP spin valves, Leung zt vl.
used this technique to estimate the spin diffusion length of Cu at 77 K to be 600±245
nm [7] (see Fig. 10.2).
Investigations into the spin diffusion lengths of oxide materials are fairly scarce.
Recently Pallecchi zt vl. have estimated the spin diffusion length of CuGO to be approx-
imately 40 nm at 10 K [8]. The investigators tried to solve this problem in two ways
using two different trilayer spin valve structures: an LaEOLSrEOHMnOH(LSMO)/CuGO/Co
trilayer, and an all-oxide LSMO/CuGO/LSMO trilayer. Measurements involving the
all-oxide structure proved unsuccessful as it was difficult to attribute the magnetore-
sistance features to genuine spin transport across the CuGO layer. The estimation for
the spin diffusion length was made using the LSMO/CuGO/Co trilayers (using CuGO
thicknesses of 5, 50, 75 and 100 nm). Figure 10.3 shows typical magnetoresistance
curves obtained for a LSMO/CuGO(50 nm)/Co trilayer over a range of temperatures.
A rather interesting observation made by the authors is that the spin diffusion length
of CuGO is relatively insensitive to the temperature and claim that this is more to
do with the fact that the LSMO has a reduced spin polarisation efficiency at higher
temperatures as it approaches its ferromagnetic Curie temperature.
186
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
Figure 10.3: Upper panel: resistance versus temperature across a LSMO/CuGO/Co
trilayer with CuGO thickness of 50 nm. The other panels show the resistance versus
magnetic field across the trilayer at different temperatures. Arrows indicate the direc-
tion of the field sweep. The dotted circled area indicates the contribution to the GMR
from AMR. After reference [8].
Ghosh zt vl. have used time-resolved optical techniques to explore electron spin
dynamics in bulk and epilayer samples of n-type ZnO as a function of temperature and
magnetic field [9]. In the bulk sample, the spin coherence time extends to very long
times at low temperatures (approximately 20 ns). In the epilayers, the spin coherence
time increases with carrier concentration, but is much smaller than in the bulk. How-
ever, spin precession in the epilayers persists until room temperature (see Fig 10.4).
FECI [vwrixvtion of the hpin vvlve struxtures
A series of Cu(200nm)/Co(20nm)/ITO(t nm)/Co(5nm)/Cu(200nm) SV structures were
prepared on unheated Si (100)/SiOG substrates in a UHV system by DC magnetron
sputtering in 1.5 Pa of Ar. The system’s base pressure before depositing was better
than 1×10−K Pa. Substrates with an area of 6×11 mmG were mounted on a circular
sample table below stationary Cu, Co, and ITO targets. The total ITO thickness, t,
was linearly varied between each sample in the 0 to 20 nm range by controlling the
rotation speed of a substrate as it passed below the ITO target, as described in section
3.3.
There are two distinct paths that can be taken towards fabrication of nanometre-
scale structures for the study of spin electronic properties: the usual top-down approach
187
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
Figure 10.4: Temperature dependence of time-resolved Faraday rotation of the spin
precession on a ZnO sample at three temperatures for an applied field of 0.8 T. After
reference [9].
uses lithographic patterning techniques to define device structures by the removal of
unwanted material and to mask the subsequent growth of different layers; the alter-
native bottom-up route requires directed additive assembly of subunits. The devices
reported in this thesis have have been fabricated using a top-down approach, pattern-
ing thin film heterostructures using novel 3D focused ion beam (FIB) lithography. All
the layers are deposited in a single UHV cycle, therefore achieving excellent interface
cleanliness.
FECICF evtterning to mixron sxvle trvxks
dptixvl lithogrvphy
Optical lithography or photolithography is a process used in microfabrication to selec-
tively remove parts of a thin film or the bulk of a substrate. It uses light to transfer a
geometric pattern from a photo-mask to a light-sensitive chemical photoresist on the
substrate. A series of chemical treatments and developing then engraves the exposure
pattern into the material underneath the photoresist. A positive “DQN” photoresist
deemed appropriate for the desired feature sizes (4 µm) was chosen. The resist is spun
onto a sample and subsequently baked-out on a hot-plate to remove the solvent and to
set the resin. The process of exposure to high intensity ultra-violet (UV) light results
in the resist undergoing a complex chemical change. The exposed regions now become
extremely soluble in alkaline solutions, compared to the non-exposed parts. Regions
that were not subjected to high intensity UV light will remain stable so that only a
fraction of the surface is dissolved. This leaves behind a photoresist pattern which
resembles the mask.
The standard lithographic technique used for this work is outlined below and
illustrated in Fig. 10.5.
188
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
(1) (2)
(3)
(4)
(5)
(6)
(7)
Substrate
Film
Photoresist
Mask
Figure 10.5: Schematic illustration of the mask pattern transfer by optical lithography
to define 4 µm tracks. (1) film or layered stacks structure has been deposited onto a
substrate, (2) photoresist is spun onto the substrate surface, (3) initial edge bead
pattern is engraved into the resist via exposure to ultraviolet light through a mask
pattern, (4) the resist is developed leaving behind a photoresist pattern on the surface of
the substrate for more intricate patterning, (5) photoresist again exposed to ultraviolet
light through a mask where smaller features (4 µm scale) are engraved into the resist, (6)
photoresist is developed for a second time, leaving behind the final photoresist pattern
on the surface, (7) Ar ion beam milling removes the material which still remains around
the final pattern and the resist is removed with acetone.
189
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
1. the sample was cleaned by soaking in acetone and then isopropanol for 2 minutes
and dried using a nitrogen air gun.
2. the sample was then placed on a small platform in a spinner. Photoresist was
applied using a pipette to the surface of the sample, whilst actively trying to
avoid any air bubbles becoming trapped (this would affect the photoresist surface
quality and cause many problems in the subsequent patterning stages).
3. the sample was spun using a program set to 6000 rpm for 30 seconds. Photoresist
of thickness ∼ 3 µm settles around the central region of the sample. Thicker edge-
beads are found around the perimeter of the sample.
4. the sample was then transferred to a hot-plate to bake for 2 minutes allowing the
solvent to evaporate and the photoresist to harden.
5. using an edge-bead removal mask, the sample was then exposed to high intensity
UV light for 90 seconds.
6. the photoresist was developed in an alkaline KOH:HGO solution with a 3:2 ratio
for approximately 40 seconds and subsequently rinsed in water.
7. using a high resolution mask with the final pattern the samples were exposed to
high intensity UV light for a second time for 8 seconds to define the device area.
8. stage 6 was repeated but with interval inspection under a light microscope while
taking paying particular attention to the 4 µm tracks: if edge-sharp the pattern
was good; if the track edges were rough, faded or absent, the photoresist was
removed with acetone and the entire process started again.
The mask aligner used operates in the contact mode, such that the mask is pressed
against the photoresist. Therefore, it is imperative that the mask itself and the sub-
strate are clean, i.e. any dust or excessive photoresist on the surface must be removed
otherwise defects are generated in the photoresist, or, if the surface is not uniform, the
undulating surface will not result in even contact from the mask, leaving behind poorly
defined features.
bvsk yesign
Two masks were used in the photolithography process, the first for the removal of
unwanted photoresist edge-beads, and the second containing the high resolution device
area pattern itself. Figure 10.6 (a) shows an illustration of the final mask pattern used.
The pattern permits the fabrication of 14 devices on one sample [Fig. 10.6 (b)].
190
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
Figure 10.6: (a) CAM39 mask design used for these experiments, (b) an expanded
view of the device area.
Ion milling
Ion milling is a non-selective, anisotropic process which uses high energy ions to bom-
bard the sample surface and erode away unwanted material from a surface. In this
case, it is used to remove unwanted material around the resist pattern. The resist
itself is also milled away, however, its erosion rate is considerably lower than that of
metal. The instrument used for these experiments was a broad-beam Ar ion miller.
The typical base pressure of the system was 1×10−N Torr. Electrons are emitted from
an ion gun ionising a neutral gas intake of Ar to Ar+. This ionised gas is accelerated
across a potential difference of 500 V towards the surface of the sample with a pressure
of 4×10−I mBar. More precisely, the input gas is actually a mixture of Ar with 2 %OG.
The role of oxygen is essentially to improve the etch rate of the photoresist, which
results in the etch profile being more uniform, and to also aid the reduction in the
redeposition of photoresist products. The average mill time used was approximately
23 minutes and was enough to effectively remove all the unwanted material around the
masked photoresist pattern. After the milling was complete the remaining photoresist
was removed by ultrasonically cleaning the sample in acetone for 20 seconds and finally
dried with a jet of nitrogen gas.
FEC5 [oxusey ion wevm proxessing of yevixes
FEC5CF ihe proxess
The focused ion beam (FIB) system used in these experiments was an FEI200 series
FIB XP 2.25 which utilises Gallium ions (Ga+) ejected from a liquid Ga source, with
a minimum spot size of approximately 10 nm [10]. The ion beam is generated from a
liquid-metal (maintained in a high-vacuum of 1×10−5 Pa) source by the application of a
large electric field. The large electric field causes the emission of positively charged ions
from a liquid gallium cone, which is formed on the tip of a hot tungsten filament [11].
The ion beam energy is typically between 10 and 50 keV, with beam currents varying
191
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
between 1 pA and 10 nA. It is possible to alter the beam current, allowing both a fine
beam for high-resolution imaging on sensitive areas and a high energy beam for fast
and the less delicate stages of milling. Typical beam currents used in this work were 11
pA for high resolution cleaning, and 150 pA for trench milling and larger cuts. Figure
10.7 (a) shows an illustration of the focused ion beam column.
Once the sample is mounted and loaded within the chamber (pumped down to a
high vacuum pressure of < 10−H Pa), the Ga source is switched on - this now allows
images to be taken (using secondary electrons). The first stages of the operation, before
any milling takes place, is to align the beam, i.e. focusing and stigmatism correction.
The physical process of the ion beam milling effect is similar to that described in
the previous section. It is a site-specific technique and so allows us to define very small
features (here we are fabricating devices of the order of hundreds of nanometers). The
Ga+ primary ion beam hits the sample surface and sputters a small amount of material,
which leaves the surface as either neutral atoms (enabling milling) or secondary ions
(enabling imaging) [see Fig. 10.7 (b)]. Milling and imaging (particularly the imaging
process which exposes larger areas to the Ga beam) can result in a number of unde-
sirable effects such as electron charging, displacement of atoms at the surface, and the
implantation of Ga+ [11]. Ion implantation can be a great problem for ferromagnets -
over milling the ferromagnet, in our case Co, with a Ga+ beam is known to severely
reduce ferromagnetic Curie temperature and the spin asymmetry [13]. Therefore, it is
important to optimise the processing of these devices when using a FIB microscope,
otherwise we risk compromising the device performance.
FEC5CG Yevixe fvwrixvtion
The three-dimensional technique for creating CPP devices used for these studies was
developed by C. Bell zt vl. for studying the device transport properties of supercon-
ductor/ferromagnet structures [14, 15]. In the present work this specific technique is
extended to produce nanoscale spin valve junctions for the characterisation of the CPP
properties.
Within the chamber of the FIB microscope, a sample holder is screwed onto a
sample carriage. The carriage has three degrees of freedom (z, y, and θ). In order to
facilitate both top-cuts and side-cuts of the 4 µm tracks, a custom made 45◦ wedge
holder (designed by D. J. Kang) was used, as illustrated in Fig. 10.8 (a) - (c).
The process of fabricating a three-dimensional nanopillar step device is outlined
in Fig. 10.9 (a) - (d). The first stage uses a high beam current of 150 pA milling
away the 4 µm track so that its width is reduced to approximately 500 - 700 nm range
using an angle of θ = 45◦ and ϕ = 0◦. The milling is monitored closely using an
end-point detection measurement. During the track thinning step, sputtered material
192
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
LiquidGa source
Aperture control
Steering quadrapole
Blanking plates
Blanking apertures
1 lens
st
Detection ocupole
2 lens
nd
Sample platform
(a)
Sputtered material, atoms, ions
secondary electrons
Sample surface
Ga beamGa beam+ +
Large current density Finer area millingSmall current density Large area milling
(b)
Figure 10.7: (a) Illustration of the ion beam column, adapted from reference J. W. A.
Robinson thesis [12], (b) Imaging (left) and milling (right) illustrations for a focused
ion beam microscope, adapted from reference [11].
193
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
q = 0
O
z
y
q
45
q = 45
O
O
32
O
D
f
=180 O
0 - 90
O
1
2
3
Ga
+
f
(a) (b) (c)
Figure 10.8: Schematic of the sample holder used in FIB; (a) the sample holder with
two axes of rotation, where 0◦≤ϕ≤360◦ and 0◦≤90◦, (b) milling of a sample where
θ=45◦ and ϕ=0◦, (c) side-cuts are made with θ≈32◦. Blue numbers 1-3 indicate that
it is possible to mount and load 3 samples at a time for processing.
is redeposited onto the sidewalls. This redeposited material will change the electrical
transport properties of the device, providing alternative current pathways. The side-
walls of the narrowed track are therefore cleaned with a beam current of 11 pA, using
the standard cleaning tool which repeatedly steps in a single pixel wide line. The clean-
ing process will also remove excessive gallium that was implanted during the narrowing
stage, where a higher beam current was used. The final track width after the cleaning
treatment is then ≤ 500 nm (Fig. 10.10). The sample was then rotated to ϕ = 180◦,
µ = 40◦ so that the isolated side-cuts could be made using a beam current of 11 pA,
giving the final device step structure. Beam drift was a problem at this stage and had
to be monitored carefully. Generally, before making any of the side-cuts, the process
would be stalled for approximately 1 minute until satisfied that the beam position had
stabilised. The side cuts in the device force the current vertically through the barrier,
hence giving us the CPP configuration (Fig. 10.11).
The effect on the magnetic properties of thin films when patterned into nanoscale
features of varying aspect ratio and shape is of course critical technologically. Tsymbal
has given a theoretical approach to the influence that the aspect ratio has on the mag-
netostatic coupling [16]. The magnetoresistance obtained in small and narrow shapes
is expected to be enhanced. Already we have discussed the findings of Kaminsky zt
vl. [13] who found that direct Ga implantation can modify magnetic thin films, reduc-
ing the coercivity even to the extent that the ferromagnetism at room temperature can
be destroyed. In contrast, Ozkaya zt vl. found that the coercive field of Py increased
upon Ga+ implantation, which they attribute to magnetostrictive stresses induced in
the film, as well as a slight change in the composition of the Ni:Fe ratio [17]. However,
Xiong zt vl. conducted very similar studies using Py, and found that objects in soft
magnetic materials can be fabricated down to as low as 100 nm with no significant
194
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
4 mm
d (nm)
200-500nm
1mm
100-300 nm
45
O
32
O
Dw=180
O
Ga beam+
(a)
(b)
(d)
(c)
0.5mm
0.5mm
0.5mm
0.5mm
Figure 10.9: Three-dimensional FIB procedure for nanopillar device fabrication; (a)
the initial micron scale track, (b) the 4 µm wide track is narrowed with a 150 pA
Ga+ beam followed by sidewall cleaning with 11 pA beam current, (c) sample stage
is rotated so that the first and second (d) side cuts can be made. Adapted from
references [12, 14].
195
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
1 mm 0.5 mm
(a) (b)
Figure 10.10: (a) Track narrowed with 150 pA (b) after further cleaning and narrowing
with an 11 pA beam.
5nm
Cu
Cu
20 nm
Co (top)
Co (bottom)
ITO 0-20 n
m
Figure 10.11: Nanopillar device structure design used for these experiments. To the
right of the structure is an expanded view of the respective layers in the “active” region
of the spin valve. The dotted arrows indicate the general principle of CPP-GMR mea-
surements where the current is forced to flow perpendicularly through an isolated active
region composing of two ferromagnetic layers separated by a non-magnetic spacer, via
two thick Cu electrodes.
effect on the magnetic properties [18]. The FIB is also known to increase the tem-
perature locally when milling material. Therefore, we must take great care over the
exposure and milling times so as to not over-expose a particular region for a prolonged
period of time. It has been quantified previously that large doses of the order of 10FK
ions/cmG are required before significant removal of material begins [17]. We expect
that within the time scales used for fabrication in this work, milling only damages the
top most Cu electrode layer.
Wu zt vl. have simulated typical nanopillar device structures using a FEMLAB
finite element simulation technique to optimise the geometry of the CPP devices [19].
The simulation models the current flow and current density of a conducting metal with
a specific geometry. The current flow in the optimal nanopillar geometry is shown in
Fig. 10.12 (b) and shows that it is perpendicular to the plane of the magnetic active
region. Fig. 10.12 (c) also shows that the nanopillars are subject to resistive heating
196
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
Figure 10.12: Finite element simulation of the optimized nanopillar showing (a) mesh
map, (b) current flow and (c) resistive heating. After reference [19].
and subsequent damage if the applied current is too high. Magnetic and surface mor-
phology characterisation of the structures also suggested that using 200 nm thick Cu
electrode leads is the optimum (as used here in these experiments).
FECK Zflperimentvl gesults
FECKCF Xu(GEEnm)DXo(GEnm)DIid(t nm)DXo(5nm)DXu(GEEnm)
hpin vvlves
Figure 10.13 (a) shows a low-angle reflectivity scan of an active region stack, consist-
ing of Co(20nm)/ITO(3nm)/Co(5nm). Like in the case of the [Co/ITO] multilayers
investigated in chapter 8, the low-angle reflectivity data suggests Co/ITO interfaces of
reasonable quality.
The magnetic properties of the samples were characterised in the VSM, from which
the in-plane magnetisation as a function of applied magnetic field was determined. In
Fig. 10.13 (b), we plot the magnetisation loop for an unpatterned film with and ITO
thickness, t, of t = 2 nm at room temperature. It is clear that there is no complete
antiparallel state but the discernable changes in gradient close to switching do suggest
a certain degree of misalignment which is dependent on the magnetic field. Generally
it is difficult to discern the different switching fields given that the area over which we
are examining is so large (an unpatterned film) and effects such as large magnetostatic
coupling and stray fields will be dominant.
The electrical resistance of the devices were measured as a function of an ap-
plied magnetic field (µEH) directed along a films’ magnetic easy-axis and parallel to
the Co/ITO interface. Measurements were performed at room temperature and 77 K
using a four-terminal lock-in technique with an AC amplitude of ±100 µA. The current
propagated through the nanopillar from the free 5 nm Co layer first. In Fig. 10.14,
197
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
(a) (b)
Figure 10.13: (a) Low-angle x-ray reflectivity measurement of a SV structure with t
= 3 nm along with a simulated fit, and, (b) Room temperature magnetisation versus
applied magnetic field for a SV structure where t = 2 nm.
we plot the typical GMR response (0.16%) of a device with t = 2 nm at room tem-
perature (main panel). Due to the relatively large cross sectional areas of the devices,
we expect an inhomogeneous magnetisation within the Co layers at low fields, which
was later verified with three dimensional OBJECT ORIENTED MICROMAGNETIC
FRAMEWORK (OOMMF) [20] simulations [stages (ii) and (iv) in Fig. 10.14]. The
simulations were carried out assuming a nanopillar device with a cross-sectional area
of 600×350 nmG (identical to that estimated for the actual nanopillar during the FIB
processing) which has been divided into 5×5×2 nmH cells. The saturation magnetisa-
tion and uniaxial anisotropy for Co were set to 1400 emu cm−H and 4.9×10K ergs cm−H,
respectively, and the damping coefficient was set to 0.5 as is usually done to obtain
rapid convergence [21,22]. The resulting simulated GMR loop for a SV structure with
t = 2 nm is plotted on top of the experimental data for easy comparison.
We observe a close resemblance between the experimental and simulated GMR
curves, with minimal differences in switching fields. The simulated GMR curve shows
larger magnetostatic coupling at zero field compared to the experimental data. A pos-
sible explanation for this could be due to the uncertainties in measuring the device
dimensions, which, if overestimated, can lead to significant changes in the degree of
magnetostatic coupling present over the device area. Of particular interest is the mag-
netic configuration of the layers at low fields, since they can reveal the formation of any
non-uniform magnetisation regions which may be annihilated at higher fields. These
are labeled (ii) and (iv) on the GMR curves and illustrate the magnetic configuration
of the two FM layers with respect to each other at an applied magnetic field of ±26
mT. Blue and red coloured regions indicate opposite orientations in the magnetisation.
Here we can see that the top Co layer is beginning to switch and shows regions of non-
uniform magnetisation, whereas the bottom Co layer remains fairly uniform, resulting
in the angle between the magnetisation of the two layers being below 180◦. This devi-
ation from the strictly antiparallel state ultimately results in a lower GMR and could
198
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
(ii)
Co(5)
Co (20)
- 26 mT
Co (5)
Co (20)
(iv) + 26 mT
Co (20)
Co (5) Co (5)
Co (20)
(ii)
(i)(iii)
(iii) (i)
(iv)
s
G
Figure 10.14: Normalised experimental and simulated GMR of a
Co(5nm)/ITO(2nm)/Co(20nm) SV structure at room temperature. Labels (i)-
(iv) represent specific stages of the GMR loop with respective micromagnetic
OOMMF simulations. Stages (i) and (iii) represent the parallel state and therefore
low resistance state of the Co layers. Stages (ii) and (iv) show the simulated magnetic
configurations of the two Co layers at µEH = ±26 mT at the onset of the antiparallel
state.
199
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
-200 -100 0 100 200
0.00
0.04
0.08
0.12
0.16
 
 
 77 K
 290 K
0
H (mT)
G
M
R
 
(
%
)
Figure 10.15: GMR curves for a spin valve structure with t= 2 nm at 77 and 290 K.
be a source of uncertainty in our measurements. Larger devices also contribute toward
considerable magnetostatic coupling as evident from the orientation of the two layers
at zero applied field in Fig. 10.14. The average magnetisation of the top and bottom
Co layers are in between 0◦ and 180◦, thus giving rise to an intermediate resistance
state. This magnetostatic coupling combined with the non-uniform magnetisation of
the FM layers affects sharp switching behaviour at low fields and will influence the
shape of the GMR signal.
GMR increased as the temperature decreased from room temperature to 77 K. Fur-
thermore, switching fields have increased with decreasing temperature resulting from
the higher coercive fields at low temperatures. Figure 10.15 shows the rise in GMR
from 0.16% to 0.18% for a SV structure where t = 2 nm.
Figure 10.16 shows the evolution of the GMR response at room temperature for
a selection of devices as the thickness of the ITO spacer layer was increased from 0
to 7 nm. We can see that there is a strong magnetostatic coupling contribution to
the GMR signal in all of the devices, but in particular, those with thinner ITO spacer
layers. The short, flat region at the point of switching suggests the occurrence of the
antiparallel alignment of the FM magnetisations.
AMR is considered a major concern for the majority of spin injection-based exper-
iments [23]. If there was a contribution to the total GMR arising from AMR, then
the most likely source would be from the FM Co layers. In order to eliminate the
possibility of an AMR contribution from the Co to the total GMR signal measured,
SV structures were fabricated with t = 0 nm. Measurements revealed no indication of
a GMR or AMR response (Fig. 10.16 (a) - open black circles).
The switching fields in the GMR curves show an interesting variation with the ITO
spacer thickness (Fig. 10.17). There is a slight increase in switching field from 1 to
200
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
Figure 10.16: Evolution of the GMR response at 290 K as a function of ITO spacer
thickness, t, where (a) t = 1 nm, (b) t = 2 nm, (c) t = 3 nm and (d) t = 7 nm. The
open black circles in (a) shows the GMR response for t = 0 nm.
2 nm, after which, we see a gradual fall. This matches the trend seen in the coercive
field values extrapolated from the magnetisation loops. Differences are likely to arise
due to the physical difference between what is actually being probed. Magnetometry
is a volume sensitive measurement and is carried out on unpatterned samples, whereas
the magnetotransport properties were performed only after the micromachining had
been carried out. Once patterned into a nanoscale structure, it is highly likely that
the properties of the materials are going to change to some extent, both from the FIB
process (exposure to doses of Ga+ ions etc) itself and due to the influence of size effects.
No GMR was recorded in the devices for t ≥ 10 nm, indicating that λsylIid for
semiconducting ITO is less than 10 nm at room temperature. In Fig. 10.18, we plot
the ITO thickness dependence of the area product with change in resistance (V∆g)
of all the devices measured at room temperature where it is clearly shown that V∆g
decays exponentially with t. We compare this data to a simplified CPP-VF model [1]
from which we estimate λsyl
Iid. This method also provides quantitative information
regarding the spin asymmetry ratio, α, of bulk Co. We have assumed infinitely thick
Co layers (assuming an upper limit of λsyl
Xo ≈ 50 nm at room temperature [24]) and
for simplicity, did not include any Co/ITO interfacial resistances. t was expected to
vary between t < λsyl
Iid and t ≈ λsylIid. The VF equation was solved without any
assumptions on t, and read as
201
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
0 5 10 15 20
0
1
2
3
4
5
 Magnetometry
 Magnetotransport
S
w
i
t
c
h
i
n
g
 
f
i
e
l
d
s
 
(
m
T
)
t (nm)
Figure 10.17: Switching fields extrapolated from magnetometry (solid black circles)
and magnetotransport measurements (open red circles) at room temperature.
V∆g =
4αGXo/Iidλ
Iid
syl[
1 +
/ITOλ
ITO
sBl
/∗Coλ
Co
sBl
]
sinh
(
a
λITOsBl
)
+
G/ITOλ
ITO
sBl
/∗Coλ
Co
sBl
cosh
(
a
λCosBl
) (10.1)
where Co /Xo
∗ = /Xo/ (1-αG) represents the normalised resistivity of Co [1] and a
is the length of the active region Co/ITO/Co. The resistivity values employed were
/Xo = 63 nΩm and /Iid = 10 µΩm (measured from 200 nm thick films deposited in
identical conditions). This resulted in αXo = 0.55. This value for αXo required to fit the
data is relatively small compared to those values found in the literature [7]. Using Eqn.
10.1, fits assuming different values for λsyl
Iid (5 - 7 nm) have been performed. For low
values of t the simulated curve with λsyl
Iid = 7 nm fits well while at higher values of
t the curve corresponding to λsyl
Iid = 5 nm serves to describe the data better. This
leads us to propose that λsyl
Iid at room temperature is 6±1 nm.
From LAR measurements, we are confident that the quality of the Co/ITO inter-
faces are reasonable. However, we cannot completely exclude the possibility of inter-
diffusion at the interfaces which can result in additional electron scattering centers and
should be accounted for in the calculations for λsyl
Iid. In fact, we know from the inves-
tigations described in chapter 8 that a magnetic dead layer may exist at the interface
between Co and ITO. The introduction of impurities within the NM ITO spacer will
shorten the mean free path of the charge carriers as well as shorten the spin-flip time
via spin-orbit scattering. This will ultimately lead to λsyl diminishing rapidly.
In chapter 4 it is shown that the temperature at which ITO is grown can affect
its morphology and structure, in particular the average grain size. θ-2θ x-ray studies
202
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
1 2 3 4 5 6 7
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
 Experimental
 
sdl
ITO
 = 5 nm
 
sdl
ITO
 = 6 nm
 
sdl
ITO
 = 7 nm
 
 
 
A
R
 
(
f
m
2
)
 t (nm)
Figure 10.18: Compiled results for a series of Co/ITO/Co SV nanopillar devices at
room temperature. The solid lines represent fits using the VF model.
show that ITO is amorphous on top of layers of Cu and Co, and so a smaller grain
size and greater density of defects and grain boundaries is highly likely. This, for now,
is unavoidable when using the sputtering set-up here. The application of heat during
the deposition process is likely to cause many complications (interdiffusion, increased
roughness profiles of initial Cu and Co layers, ztx). Perhaps the rather modest GMR
and estimation for the spin diffusion length of ITO is indicative of the critical role that
interdiffusion of magnetic impurities in the NM ITO spacer layer, as well as the ITO
microstructure, may have in spin depolarisation.
A further drawback is possibly related to the injection from a FM metal electrode
(Co). When using FM transition metal electrodes the problem of resistance mismatch,
which is responsible for severe depolarisation at the interface, has to be faced [25]
leading to a resistance build-up at the interfaces. The result of this interface resis-
tance is that the spin accumulation acts as a “bottle neck” for carrier injection, as a
consequence of the fact that the spin carriers are also charge carriers [8]. Here, the
interfacial resistance between Co and ITO has also been neglected. We justify this
assumption on the basis that all the sample structures were prepared simultaneously
in situ. However, the large resistivity mismatch between Co and ITO is ultimately an
issue and the model needs to be developed further to account for this.
Furthermore, due to the formation of regions where there is an inhomogeneous dis-
tribution of the magnetisation within the fixed FM Co layer, the size of the devices is
likely to have had a significant influence on the size of the GMR measured and λsyl
Iid
estimated.
203
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
-160 -80 0 80 160 240
0.0
0.1
0.2
0.3
0.4
0.5
 No Cu
 Cu buffer
 
 
G
M
R
 
(
%
)
0
H (mT)
Figure 10.19: Room temperature GMR curves for a spin valve where t = 2 nm with
(blue) and without (black) a 1 nm Cu buffer layer inserted at each Co/ITO interface.
FECKCG Insertion of Xu wuffer lvyers
A series of Cu(200nm)/Co(20nm)/Cu(1nm)/ITO(2nm)/Cu(1nm)/Co(5nm)/Cu(200nm)
SV structures were prepared in much the same way as the spin valves discussed above.
Figure 10.19 shows the GMR curve for an as-deposited Cu buffered SV with t = 2
nm at room temperature along with the corresponding SV without the 1 nm Cu buffer
layer inserted (from section 10.6.1). It is clear that the addition of the Cu buffer layer
results in a significant increase in the room temperature GMR from approximately
0.16% in the case of the SV without the Cu buffer to 0.47% once the Cu buffer had
been added. This is concurrent with the findings for the multilayers in chapter 9 in the
as-deposited state where an immediate improvement to the GMR was observed once a
Cu buffer layer had been inserted at each and every Co/ITO interface. This behaviour
was attributed to possible interfacial doping at the Co/Cu interface and the prevention
of stray Co scattering centres diffusing into the ITO spacer layer.
The stability of the spin valve devices was tested by monitoring its behavior over
a range of applied DC bias currents. A DC bias current was switched on and held
constant for 20 minutes and then turned off back to zero, after which the measurement
commenced. The spin valve structures were stable even up to DC bias currents of
±5 mA. In fact, the only changes observed were in the switching fields. Fig. 10.20
shows the linear dependence of the GMR switching field as a function of applied DC
current. These values were extrapolated from minor loop measurements of the device
between ±5 mA. The behaviour of the switching field with current can be explained
by considering an induced Oersted field on application of the current. The Oersted
field should cancel each other out within the nanopillar junction. However, sometimes
a remanent component may remain causing a shift in the switching fields [26].
204
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
-6 -4 -2 0 2 4 6
20
24
28
32
36
 Experimental
 Linear fit
M
i
n
o
r
 
l
o
o
p
 
s
w
i
t
c
h
i
n
g
 
f
i
e
l
d
 
(
m
T
)
I
Appl
 (mA)
Figure 10.20: Linear dependence of the switching field as a function of the DC bias
current.
With the intention of repeating what had been observed in the Cu buffered multi-
layer stacks (chapter 9), we also made an attempt to anneal these spin valve structures.
Annealing was carried out via two avenues; (1) prior to the photolithography stage and
(2) after the FIB patterning process, neither of which proved to be successful. Anneal-
ing before the photolithography stage resulted in poor mask contact due to the rising
levels of surface roughness of the top Cu electrodes. Annealing post-photolithography
and machining led to the extreme oxidation of the nanopillar structure so that the
device resistance was now far too high to measure.
FECKCH hpin vvlves fiith Ybd elextroyes O ereliminvry results
It has been made very clear that the interest in spin injection through semiconducting
materials stems from the fact that they are extremely versatile in terms of doping,
microstructure, optical tuning, bipolar transport and above all, due to their ease-of-
integration with conventional electronics.
Some of the potential problems associated with the Co/ITO/Co SV nanopillars
have been discussed in section 10.6.1. An alternative option could be the replacement
of the FM metal Co electrodes with a FM oxide like the Co-doped ITO materials
developed in earlier chapters. Immediate advantages of doing this are the likelihood
of favourable band alignment and low resistivity mismatch. Two routes were taken to
investigate the possibility of carrying this out:
1. In situ shadow masking to make Co(5 at.%):ITO(200)/ITO(2)/Co(5)/Au(200)
SVs on r -plane sapphire substrates (all thicknesses are in nm); (set A).
205
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
I+
I
-
V
+
V-
I H
Co:ITO
r-planesapphire
ITO
Co
Figure 10.21: g versus µEH curves at 10 K for set A SV structures (with junction area
of approximately 80 µmG) for the entire device (a), bottom (c) and top (d) electrodes
respectively measured using the configuration shown in the inset of (b). Resistance
versus temperature variation is shown in (b) also measured using the same configuration
shown in the inset. All measurements were made using I ‖ µE].
206
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
2. FIB machined nanopillars of Cu(200)/Co(5 at.%):ITO(200)/ITO(2)/Co(5)/Cu(200)
SVs onto SiOG/Si substrates (all thicknesses are in nm); (set B).
Both routes involved using growth procedures very similar to those described in earlier
sections of the thesis.
The compatibility of the growth conditions of the different elements is a key issue.
From chapter 5 (section 5.6.5) we know that to obtain room temperature ferromag-
netism in Co:ITO we require heater temperatures ≥ 300◦C. Each of the other layers
can be grown at room temperature. However, we also know that the crystalline quality
and electrical properties of ITO are heavily affected by its growth temperature. To
avoid interdiffusion between the two initial ITO-based layers no heat was applied after
the deposition of the initial FM oxide electrode, i.e. the FM oxide was always the
designated bottom FM electrode (first layer to be deposited).
Figure 10.21 shows representative g versus µEH measurements of a sample from set
A with the applied current (I ) flowing in a direction parallel to the applied magnetic
field (I ‖ µE] ) at 10 K along with device resistance as a function of temperature. The
device resistance shows semiconducting-like properties [Fig. 10.21 (b)] as a function of
temperature suggesting that the current is flowing through the entire device.
Emerging from the reversible negative magnetoresistance, there is a clear sign of
spin valve hysteresis with two maxima symmetric either side of zero field [Fig. 10.21
(a)]. In order to better explore the origins of this magnetoreistance, similar measure-
ments were carried out on the upper and lower electrodes at 10 K [Fig. 10.21 (c) and
(d)]. The bottom Co:ITO electrode shows a weak negative MR with no hysteresis as
we have seen in previous chapters, but the top Co electrode exhibits evident hystere-
sis. The hysteresis displayed in Fig. 10.21 (a) unfortunately cannot be unambiguously
attributed to charge carriers through the ITO maintaining their spin polarisation. The
switching fields of the curves showing hysteresis [Figs. 10.21 (a) and (d)] are identical
and we therefore attribute the origins of this behaviour to a series contribution from
top and bottom electrodes.
The results obtained from the FIB devices (set B) were also not so fruitful. The first
major issue was the growth of Co:ITO on top of Cu which may cause many undesirable
microstructural problems. The second was the FIB machining of thick oxide layers.
This is likely to cause significant changes in the material properties. We have not
addressed this issue in detail here, however has been investigated for LSMO films [27].
Depending on the radiation dosage, both the electrical and magnetic properties are at
risk of becoming affected. Note that this will also apply to the ITO spacer layer used in
the Co/ITO/Co SVs described in section 10.6.1. Figure 10.22 shows a typical g versus
µEH curve at 10 K for a FIB machined SV. The changes in g as a function of applied
magnetic field are very different to that of the larger devices from set A. Here we see
a shift from negative to positive GMR at applied magnetic fields greater than 60 mT.
Negative MR is an intrinsic property of the Co:ITO films grown when heated. The
emergence of a positive contribution may be due to the damage caused by exposure to
207
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
5nm
Cu
200 nm
Co
Co:ITO
ITO 0-20
nm
IT
Cu
Figure 10.22: CPP g versus µEH measurement of a set B nanopillar device (with
junction area of approximately 600 × 400 nmG) at 10 K.
Ga ions during the FIB process. In addition, the size of the devices is likely to have
a significant influence on the characteristic g versus µEH behaviour: the CPP- pillar
GMR contribution will be much larger in the smaller devices. In the larger devices it
is likely that we are only taking a measurement of the in-plane MR of the individual
FM electrodes and not a genuine CPP effect.
FECL hummvry vny Xonxlusions
Sub-micron spin valve devices with a range of semiconducting ITO spacer thickness
have been fabricated and current-perpendicular to plane GMR measurements have
been performed. Using a simplified VF model an estimate of the spin diffusion length
of semiconducting ITO at room temperature of 6±1 nm has been made. The spin dif-
fusion length of ITO may be improved by optimising its structural quality and physical
properties. The addition of a 1 nm Cu buffer layer at each Co/ITO interface led to a
significant rise in the room temperature GMR signal.
This work has taken a step toward realising transparent spintronics with the demon-
stration that spin polarised information can be injected and detected through ITO.
However, once spin transport in ITO is fully explored and its limits identified, a full-
crystalline oxide spin-injection device will be the ultimate target.
Our preliminary experiments where the metal FM Co electrodes were replaced by
208
Cb[pnel 10. Emncg[ncha nbe Mpch Dcffomcih Fehanb if Megc]ihdo]ncha
Chdcog Tch Ircde
the Co:ITO DMO were largely unsuccessful. The development of such devices are still
in their infancy with many hurdles yet to be met, let alone overcome. The FIB process
needs to optimised, especially parameters such as the Ga ion exposure time and dosage.
209
Wiwliogrvphy
[1] T. Valet and A. Fert, Phys. Rev. B 48, 7099, (1993)
[2] Q. Wen, Y. Song, Q. Yang, H. Zhang, Rare Met. Mater. Eng. 25, 536, (2006).
[3] B. Dieny, V.S. Speriosu, S.S.P. Parkin, B.A. Gurney, D.R. Wilhoit and D. Mauri,
Phys. Rev. B 43, 1297-1300, (1991).
[4] J. Bass and W. P. Pratt Jr., J. Magn. Magn. Mater. 200, 274, (1999).
[5] C. Chappert, A. Fert, F. Nguyen Van Dau, Nature Mater. 6, 813, (2007).
[6] D. J. Kubinski and H. Holloway, J. Appl. Phys. 79, 7395, (1996).
[7] C. W. Leung, C. Bell, G. Burnell, M. G. Blamire, Phys. Rev. B 72, 212409, (2005).
[8] I. Pallecchi, L. Pellegrino, N. Banerjee, M. Cantoni, A. Gadaleta, A. S. Siri, D.
Marre, Phys. Rev. B 81, 165311, (2010).
[9] S. Ghosh, V. Sih, W. H. Lau, D. D. Awschalom, S.Y. Bae, S. Wang, S. Vaidya,
G. Chapline, Appl. Phys. Lett. 86, 232507, (2005).
[10] FEI Company, [oxuszy Ion Wzvm lorkstvtion gzfzrznxz Guiyz, (1996).
[11] S. Reyntjens and R. Puers, Journal Of Micromechanics And Microengineering,
11(4):287-300, (2001).
[12] J. W. A. Robinson, PhD Thesis, Zero to Pi Oscillations in Ferromagnetic Joseph-
son Junctions, University of Cambridge, (2007).
[13] W. M. Kaminsky, G. A. C. Jones, N. K. Patel, W. E. Booij, M. G. Blamire, S. M.
Gardiner, Y. B. Xu, J. A. C. Bland, Appl. Phys. Lett. 78, 1589, (2001)
[14] C. Bell, PhD Thesis, Nanoscale Josephson Junctions, University of Cambridge,
(2004).
[15] C. Bell, G. Burnell, D-J. Kang, R. H. Hadfield, M. J. Kappers, M. G. Blamire,
Nanotechnology, 14(6):630-632, (2003).
[16] E. Y. Tsymbal, Appl. Phys. Lett. 77, 17, (2000)
210
Bcbfcial[pbs
[17] D. Ozkaya, R. M. Langford, W. L. Chan, A. K. Petford-Long, J. Appl. Phys. 91,
9937, (2002).
[18] G. Xiong, D. A. Allwood, M. D. Cooke, R. P. Cowburn, Appl. Phys. Lett. 79,
3461, (2001).
[19] M. C. Wu, A. Aziz, J. D. S. Witt, M. C. Hickey, M. Ali, C. H. Marrows, B. J.
Hickey, M. G. Blamire, Nanotech. 19, 485305, (2008).
[20] See, http://math.nist.gov/oommf, Version 1.2.0.4 of the software was used.
[21] S. H. Lim, S. H. Han, K. H. Shin, and H. J. Kim, J. Magn. Magn. Mater. 223,
192, (2001)
[22] S. Bedanta, E. Kentzinger, O. Petracic, W. Kleemann, U. Rcker, T. Brckel, A.
Paul, S. Cardoso, and P. P. Freitas, Phys. Rev. B 74, 054426, (2006).
[23] Y. Ji, A. Hoffmann, J. S. Jiang, S. D. Bader, Appl. Phys. Lett. 85, 6218, (2004).
[24] L. Piraux, S. Dubois, A. Fert, and L. Belliard, Eur. Phys. J. B 4, 413, (1998).
[25] A. Fert, H. Jaffres, Phys. Rev. B 64, 184420, (2001).
[26] N. Banerjee, A. Aziz, M. Ali, J. W. A. Robinson, B. J. Hickey, M. G. Blamire,
Phys. Rev. B. 82, 224402, (2010).
[27] I. Pallecchi, L. Pellegrino, E. Bellingeri, A. S. Siri, D. Marre, G. C. Gazzadi, J.
Magn. Magn. Mater. 320, 1945, (2008).
211
Xhvpter FF
hummvry vny Xonxlusions
This dissertation provides a detailed and systematic study into the investigation of
using InGOH and ITO based materials for next generation spintronic applications. This
was approached by first developing the InGOH and ITO oxides and later a functional
dilute magnetic oxide system. Thereafter, ITO was incorporated into device-like struc-
tures and tested for its efficiency as both a medium for spin polarised information to
move through, but also as a spin injector.
The major findings of this comprehensive study are summarised in the following
points:
1. A DMO has been developed using transparent and conducting Co-doped InGOH
and ITO thin films with a ferromagnetic Curie temperature in excess of room
temperature (approximately 500 K).
2. XMCD and O-MCD exploration indicates that the Co dopants alone are not
the sole source of ferromagnetism in these materials. In fact, the Co dopants in
dilute form give a strictly paramagnetic response suggesting that the magnetism
observed may be a result of polarised electrons in localised donor states within the
host InGOH and ITO oxide, most likely related to oxygen vacancy defect states.
Therefore, the origins of magnetism in these films is related to a hybridisation
and charge transfer of electrons from a broad donor/defect-derived impurity band
to a band of unoccupied 3y states at the Fermi level. However, we are unable
to determine the location of these donor states - they may be distributed evenly
through the material or concentrated at the grain boundaries.
3. A unified theory capable of explaining the magnetic behaviour of the DMO sys-
tems still eludes us. It appears from our experimental results and those of others
that different doped magnetic oxides may have different underlying mechanisms
leading to the observed ferromagnetism. There may even be several competing
ferromagnetic mechanisms at play in these DMO materials.
4. Co/ITO multilayers exhibited low GMR persisting to room temperature. Both
the magnetoresistance and magnetisation saturation field oscillate with the ITO
212
Cb[pnel 11. Mogg[ls [hd Cih]fomcihm
spacer layer thickness consistent with RKKY interaction throughout the struc-
ture. From our magnetic modelling we believe that a significant portion of the
magnetic and magnetotransport properties were a result of a superparamagnetic
contribution. These superparamagnetic regions may be due to intermixing at
the Co/ITO interfaces leading to stray Co particles within the ITO spacer layers.
This highlights the importance of the quality of the interfaces between the layers
in multilayer stacks and metal-oxide heterostructures.
5. Insertion of a 1 nm Cu buffer layer at each Co/ITO interface leads to an imme-
diate rise in GMR. This is likely to be due to interfacial-doping at the Co/Cu
interface. Upon annealing, GMR was found to increase as it is believed that a
graded interface between the Cu and ITO has formed, hence reducing the burden
of a high Fermi surface mismatch. Further annealing at higher temperatures led
to a reduction in GMR, possibly due to enhanced atom mobility and interfacial
mixing.
6. Further evidence of efficient spin transport through ITO at room temperature
was provided by the estimation of a spin diffusion length in ITO of 6±1 nm using
CPP-GMR measurements of Co/ITO/Co nanopillar junctions.
Overall, this research shows that ITO is a candidate material for transparent spintron-
ics. This new level of understanding opens up the possibility of creating an all-oxide
device.
213
Xhvpter FG
[uture Yirextions vny dutlook
FGCF iZb stuyy of the Ybd
There is a need for more detailed atomic characterisation on the dopant distribution
and chemical states in the DMO systems. More extensive use of techniques such as
high resolution TEM combined with element specific sensitive detection, e.g. electron
energy loss spectroscopy, would be ideal and would complement characterisation us-
ing the synchrotron based techniques. High-resolution cross-sectional TEM on these
samples would improve our knowledge of the material, and possibly help to put the
argument about the possible presence of Co metal clustering to bed once and for all.
These studies must take into account the observations made on various length scales,
because even in the cases of clustered dopant samples, local TEM may look deceptively
good. In chapter 5, preferential Co clustering close to the substrate/film interface was
discussed. This would certainly be an area of interest. A combination of these tech-
niques with the synchrotron and magneto-optical based methods could bring far greater
clarity on the issues of importance in the field.
FGCG ehotomvgnetix effext
Recently, photoinduced magnetisation in various materials has been the focus of much
attention [1–3], where ferromagnetism is induced or enhanced by the interaction with
the incident light.
I have carried out some preliminary measurements investigating the photoconduc-
tive properties of pure InGOH and ITO(10) thin films, i.e. the influence of light of various
wavelengths on the electrical properties of the material. It was found that those films
which were more insulating were more susceptible to changes in conductivity when
exposed to different wavelengths of light. Figure 13.1 shows the conductivity gain (%)
as a function of the wavelength of light for pure InGOH which had been grown at 300
◦C
and in relatively high levels of oxygen (0.20 Pa). It is clear that in each case, when
the sample is exposed to light, the conductive properties of the sample improve and is
214
Cb[pnel 1,. Fonole Dcle]ncihm [hd Ionfiie
0 50 100 150 200
0
50
100
150
200
250
300
350
C
o
n
d
u
c
t
i
v
i
t
y
 
G
a
i
n
 
(
%
)
Time (s)
Figure 12.1: Photoconductive properties for pure InGOH thin films grown at 300
◦C and
0.20 Pa of oxygen. The conductivity gain (%) is given as a function of the wavelength
of light [Red (650 nm), Green (510 nm), Blue (475 nm) and ultraviolet (400 nm)].
likely to be related to photoexcitation of electron carriers. The largest effects are seen
in those films exposed to ultraviolet light.
An interesting advance could be to investigate the influence that this may have on
the magnetic properties of the films. Evidence described in earlier chapters suggested
that there may be a correlation between the carrier concentration and the ferromagnetic
response in our samples. Therefore, such an experiment may yield some interesting
results, in particular, using the Co-doped InGOH samples as these are far more insu-
lating than those which have additional Sn doping. The measurements would involve
monitoring the carrier concentration and ferromagnetic response before and after ex-
posure to light. However, one problem which is evident from Fig. 13.1 is the time for
the electrical resistance to return back to its original value after light exposure (relax-
ation time). Therefore, when performing these measurements it is first necessary to
determine the relaxation time of the films for the particular wavelengths of light. This
investigation may lead to a material with dynamic “on/off” control of the ferromag-
netism.
FGCH betvlDIid interfvxes
This work has highlighted the importance of the interface between metal and oxide
surfaces. The major issue of the presence of isolated SPM regions in the vicinity of
the Co/ITO interface has been discussed but have not been located. More detailed
215
Cb[pnel 1,. Fonole Dcle]ncihm [hd Ionfiie
structural characterisation using high resolution TEM would certainly prove useful
and would also provide information on the quality of the interfaces and the level of
interdiffusion. Neutron diffraction studies may also be able to shed some light on the
magnetic structure of the multilayers as we argue that the magnetic frustration is a
result of multiple magnetic contributions. Perhaps the presence of Cu buffer layers
not only has an influence on the structural properties of the multilayers but also the
magnetic properties. If this is the case, then neutron diffraction studies may help to
explain the behaviour in these samples.
FGCI [unxtionvl all-oxide yevixe
The development of an all-oxide device which exploits both the conducting and mag-
netic nature of the Co:ITO DMO is still very much in its infancy. One of the major
areas which is still in need of some serious development is that of the FIB processing
of the oxide itself. A worthwhile future investigation should use the work of Pallecchi
zt vlC [4] as a guideline for the systematic study of the effects of FIB radiation dosage
on the electrical and magnetic properties of the pure and TM-doped ITO films.
Perhaps if there are still problems arising in obtaining genuine spin valve behaviour
in these all-oxide devices, then the first question to be asked could be whether the
Co:ITO is an efficient spin polariser at all. Therefore, it would be useful to try and
quantify the magnitude of the spin polarisation of the Co:ITO DMO materials using
Point Contact Andreev Reflection, similar to the study by Panguluri zt vl. [5]. If this
can be quantified reliably then it may still be worth pursuing this material for future
development.
216
Wiwliogrvphy
[1] J. Fernandez-Rossier, C. Piermarocchi, P. Chen, A. H. MacDonald, L. J. Sham,
Phys. Rev. Lett., 93, 127201, (2004).
[2] S. Mishra, G. S. Tripathi, S. Satpathy, Phys. Rev. B. 77, 125216, (2008).
[3] J. S. Bettinger, R. V. Chopdekar, Y. Suzuki, Appl. Phys. Lett., 94, 072505, (2009).
[4] I. Pallecchi, L. Pellegrino, E. Bellingeri, A. S. Siri, D. Marre, G. C. Gazzadi, J.
Magn. Magn. Mater. 320, 1945, (2008).
[5] R. P. Panguluri, P. Kharel, C. Subdakar, R. Naik, R. Suranarayanan, V. M. Naik,
A. G. Petukhov, B. Nadgorny, G. Lawes, Phys. Rev. B 79, 165208, (2009).
217
Vppenyifl V
VwwrevivtionsA hymwols & ehysixvl
xonstvnts
Vwwrevivtions
VFb Vtomix forxe mixrosxopy
VHE Vnomvlous Hvll effext
Vbg Vnisotropix mvgnetoresistvnxe
Wbe Wouny bvgnetix eolvron
XIe Xurrent inBplvne
Xee Xurrent perpenyixulvrBtoBtheBplvne
Ybd Yilute mvgnetix ofliye
Ybh Yilute mvgnetix semixonyuxtor
FIW Foxusey Ion Wevm
Fb Ferromvgnet
Fn Fluoresxenxe yiely
Gbg Givnt mvgnetoresistvnxe
Iid Inyium iin dfliye
aVg aofiBvngle mBrvy reflextivity
bXY bvgnetix xirxulvr yixhroism
bYa bvgnetix yevy lvyer
bg bvgnetoresistvnxe
cb conBmvgnetix
dHE dryinvry Hvll effext
dBbXY dptixvl mvgnetix xirxulvr yixhroism
ddbbF dwjext drientey bixromvgnetix Frvmefiork
eaY eulsey lvser yeposition
gKKn guyemvnBKittelBKvsuyvBnosiyv
heb huperpvrvmvgnetix
hfjIY huperxonyuxting quvntum interferenxe yevixe
hk hpin vvlve
iXd irvnspvrent xonyuxting ofliye
iXg iempervture xoeffixient of resistvnxe
iEn iotvl elextron yiely
ib irvnsition metvl
jHk jltrv high vvxuum
kF kvletBFert
khb kiwrvting svmple mvgnetometer
mVh mBrvy vwsorption spextrosxopy
mbXY mBrvy mvgnetix xirxulvr yixhroism
mgY mBrvy yiffrvxtion
218
Appehdcr A. Abblepc[ncihm& Msgbifm & Pbsmc][f ]ihmn[hnm
Xonstvnts
A Vrev
g avnye g fvxtor
C Xurie xonstvnt
H Vppliey mvgnetix fiely
H c Xoerxive fiely
I Xurrent
m∗ Effextive mvss of v xhvrge xvrrier
ms hvturvtion mvgnetisvtion
mr gemvnent mvgnetisvtion
N cumwer of wilvyer repevts
nc Xvrrier xonxentrvtion
ns hheet xvrrier xonxentrvtion
q xhvrge
R gesistvnxe
Rs hheet resistvnxe
T iempervture
TC Ferromvgnetix Xurie tempervture
TB huperpvrvmvgnetix wloxking tempervture
VH Hvll voltvge
V 0 dflygen vvxvnxy
χ bvgnetix susxeptiwility
β hpin vsymmetry rvtio
λs hpin yiffusion length
µH Xvrrier mowility
µL drwitvl moment
µS hpin moment
n˜ gefrvxtive inyefl
ϕb Grvin wounyvry potentivl wvrrier
/ gesistivity
/xy Hvll resistivity
/+ D /− gightDaeft xirxulvrly polvrisey lightDmBrvys
σ Xonyuxtivity
θ XurieBleiss tempervture
ehysixvl xonstvnts
Boltzmann constant kB 1×P8 × 1M−GH g h−F
Bohr magneton B VKOT × 1M−GI g q−F
blectron charge q 1KSMO × 1MFN C
blectron mass mz VK11 × 1M−HF kg
blectron volt es 1KSMO × 1M−FN g
mermeability of vacuum E 4 × 1M−L e m−F
mlanck constant h SKSP × 1M−HI g s
oeduced mlanck constant 3h 1KMR × 1M−HI g s
219
Vppenyifl W
euwlishey vrtixles
220