as“ . . .
5.3.1..“ .1.

1 .
« at». ‘ y7udfl a!
‘ ii.

i.

(p.
,_

..
(4‘ rd... 93%.;

"\

.fi 3

.. .~‘ ,5: .O

33%.! _
.. ‘ . Ari a.

I“

 

III
III
‘III!

TH? "5‘75
2

s’HUiqflfi

University

 

This is to certify that the

' Li afiA RY dissertation entitled
Michigan State STRUCTURAL STUDIES OF GMR MAGNETIC MULTILAYERS

AND SPIN-VALVES GROWN BY DC SPUTTERING

 

 

presented by

Hong Geng

has been accepted towards fulfillment
of the requirements for

Ph.D. degree in Materials Science
and Engineering

//M

A
Major professor Y /

 

 

Date [/ZZ/aj

MS U is an Affirmative Action/Equal Opportunity Institution 0-12771

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJCIRC/DateDuo.p65-p.15

STRUCTURAL STUDIES OF GMR MAGNETIC
MULTILAYERS AND SPIN-VALVES GROWN BY DC
SPUTTERIN G

By

Hong Geng

A DISSERTATION

Submitted to
Michigan State University
in partial fiJlfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Chemical Engineering and Materials Science

2002

ABSTRACT

STRUCTURAL STUDIES OF GMR MAGNETIC MULTILAYERS AND SPIN-
VALVES GROWN BY DC SPUTTERING

By

Hong Geng

The structures and growth characteristics of sputtered GMR magnetic multilayers
and spin-valves have been characterized using conventional and high resolution
transmission electron microscopy (CTEM and HRTEM). The first part of this
dissertation focused on the characterization of polycrystalline Cu/Py (permalloym),
Ag/Py, and Cu/CoZr spin-valves. For the Cu/Py spin-valves, structures with different Py
thicknesses sputtered at different temperatures were examined. All of the spin-valves
displayed polycrystalline structures with columnar grains. The grains grew on close
packed planes ({111} in the bcc Nb contacts and {110} in the fcc Cu, F eMn, and Py spin-
valve layers) taking up a near Kurdjumov-Sachs {1]1}fcc//{1l0}bcc;<110>fcc//<111>bcc
orientation relationship. FFT analysis and HRTEM image simulations indicate that in
some of the columnar grains, the Cu, FeMn, and Py regions take up non-equilibrium bcc
structures, regardless of differences in layer thicknesses and sputtering temperatures. For
the Ag/Py spin-valves, in addition to the expected polycrystalline morphology, in one
instance an epitaxial single crystal Ag layer was discovered when the spin-valve was
grown directly on a Si substrate. HREDS has shown that this single crystal (epitaxial)
layer is likely composed of Ag and In in a 4:1 atomic ratio. In all of the other Ag/Py
spin-valve samples, polycrystalline morphologies were consistently observed regardless

of the substrate. For the Cu/CoZr spin-valves, CTEM observations of two spin-valves

revealed that both samples are polycrystalline with columnar grains similar to the
morphologies observed in Cu/Py and Ag/Py spin-valves.

The second part of this dissertation is focused on determining the correct
parameters for growing a high quality Cu buffer layer on an epitaxial Nb contact. CTEM
observations demonstrate that the quality of epitaxial growth of Cu buffer layers depends
strongly on deposition temperatures and post annealing. High quality epitaxial Cu buffer
layers were obtained when the Cu layer was grown at low temperatures (100°C and
150°C) without post annealing. Diffraction patterns of multilayers show that twin
variants appear in epitaxial Cu and Py layers, consistent with electron backscatter
patterns (EBSP) results from thick films. llRTEM images reveal that the Cu buffer layer
grown at 150°C appears as twin variants with two stacking sequences of {Hi} fcc
planes. A three-layer periodic contrast observed in the Py layer can be rationalized based

on overlapped twins along [110] with {11 1} twin planes.

TO MY PARENTS,

MY HUSBAND,

AND MY DA UGHTER.

ACKNOWLEDGMENTS

I would like to express my sincere appreciation to my major advisor Professor
Martin A. Crimp for all his guidance, encouragement, and immeasurable support
throughout my doctoral studies at Michigan State University. Without his support and
patience, this work would not have been possible.

I would also like to sincerely thank my guidance committee members, Professor
Jack Bass, Professor William P. Pratt, Professor Thomas R. Bieler, and Professor David
Grummon for their helpful insights, instruction, and valuable suggestions throughout the
entire work.

The achievements of this work are also based on the help from the senior
colleagues in our research group. A special thank goes to Dr. Reza Loloee for specimen
fabrication and information exchange that contributed greatly to my understandings.
Thanks are due to Dr. John W. Heckman for sharing experiments and characterization
techniques.

Many thanks go to Mr. Robert J. Pcionek for the precious time he committed to
proof reading of my dissertation.

I also thank Drs. J. P. Mansfield and C. Wauchope of the University of Michigan
EMAL for access to the J EOL-4000EX.

My deepest thanks are extended to my husband, Zhe Li, my parents, Dezhang
Geng and Qingfen Chen, and my daughter, Erin for their generous love, understanding,

and encouragement, which have supported me in my pursuit of a Ph.D. degree from

Michigan State University. I also want to extend my thanks to my husband’s family, for
caring about me.
Finally, I sincerely appreciate the financial support provided by the MSU CF MR

and the US NSF under grants DMR 98-20135 and MRSEC DMR 98-09688.

vi

TABLE OF CONTENTS

LIST OF TABLES. . . . . . . . . . x
LIST OF FIGURES. . . . . . . . . . x1
CHAPTER I INTRODUCTION. . l
1.1 THE GMR EFFECT 2
1.1.1 Origin of the GMR Effect 3
1.1.2 GMR Measurements 4

1.2 GMR MAGNETIC MULTILAYERS AND SPIN-VALVES 8
1.3 CORRELATION BETWEEN GMR AND MULTILAYER STRUCTURES 9
1.3.1 Interfacial or Layer Roughness 12

1.3.2 Layer Quality 12

1.3.3 Crystal Orientation or Growth Direction 15

1.3.4 Layer Thickness 17

1.3.5 Growth and Deposition Conditions 17

1.4 CHARACTERIZATION TECHNIQUES 18

1.5 OBJECTIVES OF THIS RESEARCH 19
CHAPTER II EXPERIMENTAL PROCEDURES. . . . . 21
2.1 GMR MULTILAYERS AND FABRICATION 21
2.1.1 Polycrystalline Multilayers/Spin-Valves And Fabrication 21

2.1.1.1 Exchange-biased Spin-Valve (EBSV) Samples 24

2.1.1.2 Hybrid Spin-Valve Sample 25

2.1.2 Epitaxial Multilayers/Spin-Valves And Fabrication 26

2.2 CROSS-SECTIONAL TEM SAMPLE PREPARATION 26
Form a composite slab 28

Grind and polish slabs to ~100 pm thick 28

Dimple slabs to 10~2O pm thick 30

Ion-mill samples to electron transparency 31

2.3 CHARACTERIZATION AND IMAGE ANALYSIS TECHNIQUES 32
2.3.1 Conventional Transmission Electron Microscopy (CTEM) 32

2.3.2 High-resolution Transmission Electron Microscopy (HRTEM) 33

2.3.3 Fast Fourier Transform (FFT) Analysis 36
CHAPTER III STRUCTURAL STUDIES OF POLYCRYSTALLINE Cu/Py GMR
SPIN-VALVES. . . . . . . . 40

3.1 INTRODUCTION 40
3.2 CTEM OBSERVATIONS 41
3.2.1 Spin-valves with Thick Py Layers 41

3.2.2 Spin-valves with Thin Py Layers 45

3.3 HRTEM AND FFT ANALYSIS 45
3.3.1 Spin-valves with Thin Py Layers 45

3.3.2 Spin-valves with Thick Py Layers 62

vii

3.4 DISCUSSION 70

3.5 SUMMARY 76
CHAPTER IV STRUCTURAL STUDIES OF POLYCRYSTALLINE Ag/Py GMR
SPIN-VALVES. . . . . . . . 78
4.1 INTRODUCTION 78
4.2 CTEM OBSERVATIONS 79
4.2.1 Spin-Valve Grown on the Nb Contact 79
4.2.2 Spin-Valves Grown directly on the Si Substrate 79
4.3 HRTEM AND F FT ANALYSIS 82
4.3.1 Spin-Valve Grown on the Nb Contact 82
4.3.2 Spin-Valve Grown directly on the Si Substrate 89
4.4 HREDS RESULTS 96
4.5 SUMMARY 101
CHAPTER V STRUCTURAL STUDIES OF POLYCRYSTALLINE Cu/CoZr
GMR SPIN-VALVES. . . . . . . 103
5.1 INTRODUCTION 103
5.2 CTEM OBSERVATIONS 103
5.2.1 Nb//Cu/FeMn/CoZr/Cu/CoZr/Cu//Nb EBSV (1043-7) 103
5.2.2 Nb//Cu/(Py/Cu/CoZr/Cu)x10/Cu//Nb Hybrid Spin-Valve (1042-3) 105
5.3 SUMMARY 108
CHAPTER VI STRUCTURAL STUDIES OF EPITAXIAL GMR MULTILAYERS
AND SPIN-VALVES GROWN BY DC SPUTTERING. . 109
6.1 INTRODUCTION 109
6.2 PRELIMINARY STUDIES ON EPITAXIAL SAMPLES 110
6.3 ORIENTATION RELATIONSHIPS BETWEEN LAYERS DETERMINED BY
EBSP [33] 112
6.4 CTEM OBSERVATIONS 114
6.4.1 The Initial Series of Epitaxial Samples 1 14
Nb//Cu/Py/FeMn//Nb Multilayer (31-2) 1 14
Nb//Cu/Py/Cu/Py/FeMn//Au/Nb/Au Spin-valve (27-3) 116
Nb//Cu/Co/(Cu/Co)x20//Au/Nb/Au Multilayer (28-3) 116
6.4.2 A New Series of Epitaxial Samples 120
6.4.2.1 Epitaxial samples grown at high deposition temperature 121
Nb//Cu/Py Multilayer (41-3) 121
6.4.2.2 Epitaxial samples grown at low deposition temperature 121
Nb//Cu/Py//Nb Multilayer (43-1) 123
Nb//Cu/Py//Nb Multilayer (43-2) 123
6.4.3 Epitaxial Samples of Nb//Py/Cu//Nb Multilayers (43-3 and 43-4) 126
6.5 HRTEM AND FFT ANALYSIS 129
6.5.1 Epitaxial Nb Grown on Sapphire Substrates 129
6.5.2 Nb//Cu/Py/FeMn//Nb Multilayer (31-2) 135
6.5.3 Nb//Cu/Co/(Cu/Co)20//Au/Nb/Au Multilayer (28-3) 139

6.5.4 Nb//Cu/Py//Nb Multilayer (43-2) 139

viii

6.6 DISCUSSION

6.6.1 The Effect of Growth Temperature and Post Deposition Annealing

6.6.2 Overlapping Twins in Py
6.7 SUMMARY

CHAPTER VII SUMMARY AND CONCLUTIONS.

6.1 SUMMARY AND CONCLUSIONS

6.2 FUTURE STUDIES
6.2.1 Analytical Analysis
6.2.2 Epitaxial Growth

REFERENCES .

148
148
149
153

157
157
160
161
162

164

LIST OF TABLES

Table 2.1 Polycrystalline Spin-Valve Samples and Sputtering Temperatures ............. 22
Table 2.2 Epitaxial Samples and Growth Temperatures ...................................... 27
Table 2.3 Unit Cell Parameters for Layer Materials .......................................... 37

Table 3.1 Projected atomic spacings for the <111> and <1 10> zone axes for layer
components in fcc and bcc forms ............................................. 57

LIST OF FIGURES

Figure 1.1 Schematic representation of the mechanism of spin-dependent scattering for
two magnetic states: (a) parallel (P) state with the F -layer magnetizations parallel to each
other; (b) anti—parallel (AP) state with the F-layer magnetizations anti-parallel to each
other .................................................................................................... 5

Figure 1.2 A representation of the anti-parallel (AP) and parallel (P) configurations of the
ferromagnetic (F) layer moments and the magnetic field dependence of the GMR for a
Co/Cu multilayer sample at the first antiferromagnetic coupling peak ........................ 6

Figure 1.3 Schematic diagrams of the (a) CIP and (b) CPP configurations and
corresponding examples of the GMR measurements in a (Co/Ag) magnetic multilayer.
CPP-MR=45% > CIP-MR=9°/o [15] ............................................................... 7

Figure 1.4 The first discovery of the giant magnetoresistance phenomena in single crystal
Fe/Cr superlattice. The maximum GMR was observed with (Fe 30A /Cr 9 A)6o
superlattice [3] ....................................................................................... 10

Figure 1.5 Schematic representation of (a) an EBSV structure and (b) the
magnetoresistance response of a spin-valve [N iFe24/Ag20/N iF e24/FeMn3]nm (N iFe-denoted
as Py) [34] ............................................................................................ 11

Figure 1.6 The quantitative roughness of each individual Cr layer inside a series of
superlattices [Fe(3nm)/Cr(l.2nm)]N with N=20, 40, and 60. Roughness of each
individual bilayer as a function of the bilayer index M for N = 20 (o), N = 40 (D), and N
= 60 (O). 0 represents the roughness. Note that the roughness increases with the bilayer
index [37] ............................................................................................. 13

Figure 1.7 Cr mapping taken on the [Fe(30A)/Cr(l2A)]x2O multilayer grown at (a) 4
mTorr and (b) 10 mTorr showing the layer roughness increased with sputtering pressure
[38] .................................................................................................... 14

Figure 1.8 Cross-section transmission electron micrographs of (a) Si//[Cu/(Co/Cu)16/Cu]
with Cu buffer layer and (b) Si//[Fe/(Co/Cu)16/Cu] with Fe buffer layer, which show
considerable waviness and buckling of the layers in (a) and comparatively smooth and
flat layers in (b). (c) An illustration of the relationship between GMR and Co/Cu
superlattices [42]. Note that the Co/Cu grown on the Fe buffer layer has a much larger
magnetoresistance, indicating high-quality structure results high GMR value ............. 17

Figure 2.1 Top view of duplex designed CIP and CPP measurement configuration ...... 23

Figure 2.2 Schematic representation of the cross-section TEM samples preparation. . ....29

xi

Figure 2.3 A schematic representation of sinx(u) versus u without damping of the higher

spatial frequencies. The first zero position of sinx curve corresponds to Scherzer
resolution [69] ....................................................................................... 35

Figure 2.4 A schematic representation of T(u) versus u modified by the damping
envelope (dash line) which ends up the information limit [69] ............................... 38

Figure 3.1 (a) Bright field and (b) dark field cross-sectional CTEM images of the
[Nb25.0//(Cu/FeMn/Py/Cu/Py)//Nb250]nm spin-valve with 30 nm Py layers sputtered at
65°C. The darkfield image was taken using an arbitrary diffraction direction ............. 42

Figure 3 .2 A cross-sectional brightfield CTEM image of the
[Nb250//(Cu/FeMn/Py/Cu/Py)//Nb10]nm spin-valve with 30 nm Py layers sputtered at
-65°C .................................................................................................. 43

Figure 3 .3 A cross-sectional brightfield CTEM image of the
[Nb250//(Cu/FeMn/Py/Cu/Py)//Nb10]nm spin-valve with 24 nm Py layers sputtered at
22°C ................................................................................................... 44

Figure 3.4 A cross-sectional Brightfield CTEM image of the
(Nb250//Cu/FeMn/Py/Cu/Py//Nb10),,m spin-valve with 3 nm Py layers sputtered at -
56°C ................................................................................................... 46

Figure 3.5 A cross-sectional Brightfield CTEM image of the
(sz50//Cu/FeMn/Py/Cu/Py//Nb10)nm spin-valve with 3 nm Py layers sputtered at -
33°C ................................................................................................... 47

Figure 3.6 cross-sectional Brightfield CTEM image of the
(sz50//Cu/FeMn/Py/Cu/Py//NbIo)nm spin-valve with 3 nm Py layers sputtered at
18°C ................................................................................................... 48

Figure 3.7 A cross-sectional HRTEM image of the [Nb//Culo/FeMng/Py3/Cuzo/Py3//Nb]nm
spin-valve with 3 nm Py layers sputtered at -56°C with nominal layer positions indicated.
The first Cu layer and FeMn layer are weakly visible, while the other individual layers
are difficult to distinguish due to the small differences in the average atomic numbers of
the multilayer components. (SF -— stacking fault) .............................................. 49

Figure 3.8 A HRTEM micrograph showing a stacking fault lying in the second Cu layer
of the spin-valve with 3 nm Py layers sputtered at -56°C. The Burgers circuit results a
net displacement of a/4[ T2T] ..................................................................... 51

Figure 3.9 HRTEM images (left) and corresponding FFT patterns (right) of the top
(upper) and bottom Nb contacts of the Cu/Py spin-valve with 3 nm Py layers sputtered at

-56°C. These patterns were used as a calibration for the patterns displayed in figure
3.10 ................................................................................................... 52

xii

Figure 3.10 HRTEM images and F FT patterns of various layers of the Cu/Py spin-valve
with 3 nm Py layers sputtered at -56°C. All of the component layers display their
equilibrium fcc (Cu, FeMn) or L12 (Py) <110> zone patterns ................................ 54

Figure 3.11 Cross-sectional HRTEM image of the Cu/Py spin-valve with 3 nm Py layers
sputtered at -33°C. Some of the columnar grains of the Cu, F eMn and Py layers display
<111> symmetry characterized by 60° interplanar angles (inset column) .................. 55

Figure 3.12 (A) and (B) Bloch wave HRTEM simulations of copper in the fcc form
described in Table 3.1. The images simulate those viewed on a JEOL-4000 EX (400
KeV, Cs 1.1mm, beam- convergence semi-angle 0.72 mrad, objective aperture 19 nm’l,
focus spread 9 nm, Scherzer defocus, 52.08 nm.) All images were printed using identical
brightness and contrast ranges. (A) fcc Cu viewed along a <110> zone axis with
specimen thickness and focus below Gaussian illustrated. (B) fcc Cu viewed along a
<111> axis using identical operating conditions. (to be continued) ......................... 59

Figure 3.12 (C) The non-equilibrium bcc Cu viewed along a <110> axis. (D) The non-
equilibrium bcc Cu viewed along a <111> axis. Although the microscope should allow
the pass of fcc<m> and bcc<no> image information, in practice these images (especially
fCC<11 1)) are likely to be unresolved due to the very low amplitude (: 0.07 of the average
<110> axis in the simulated images) of the passed frequencies in this range. ((Images and
analysis are provided by J.W. Heckman.) ....................................................... 60

Figure 3.13 HRTEM images and corresponding fast Fourier transform diffractograms of
various layers of the Cu/Py spin-valve with 3 nm Py layers sputtered at -33°C showing
<111> zone patterns in the Cu, FeMn and Py layers ........................................... 61

Figure 3. 14 A cross-sectional HRTEM image of the

[Nbl/Culo/FeMng/Py3/Cu2o/Py3/be]nm spin-valve with 3 nm Py layers sputtered at 18°C
with nominal layer positions indicated. The individual layers of the spin-valve are
difficult to distinguish due to the small differences in the average atomic numbers of the
multilayer components. ............................................................................ 63

Figure 3.15 HRTEM images and FFT patterns of various layers of the Cu/Py spin-valve
with 3 nm Py layers sputtered at 18°C. All of the component layers display their
equilibrium bcc (Nb) <1 11> and fcc (Cu, FeMn) or L12 (Py) <110> zone patterns ....... 64

Figure 3.16 A cross-sectional HRTEM image of the spin-valve with 30 nm Py layers
sputtered at -65°C (left) and selected areas from the F eMn, Cu and Py layers (right). The
nominal layer locations are given in the image ................................................. 65

Figure 3.17 A {111} coherent twin boundary in the second Py layer in the Cu/Py spin-
valve with 30 nm Py layers sputtered at -65°C ................................................. 67

xiii

Figure 3.18 HRTEM images and FF T diffractograms of various layers of the Cu/Py spin-
valve with 30 nm Py layers sputtered at 65°C showing six-fold symmetry and lattice
spacings consistent with bcc structures ........................................................... 68

Figure 3.19 HRTEM images and FFT diffractograms of selected Cu layers within the
Cu/Py spin-valves (a) with 30 nm Py layers sputtered at -65°C and (b) with 24 nm Py
layers sputtered at 22°C, indicating that both the spin-valves can have the non-
equilibrium bcc structures characterized by 60° interplanar angles in <111> zone
patterns ................................................................................................ 69

Figure 3.20 Bloch wave HRTEM simulations (using the same microscope conditions as
in figure 3.12) of the effects of specimen tilt on the resulting image for a representative
crystal viewed at various foci near Scherzer defocus. (A) fcc (2 6 nm thick) Cu viewed
along a <110> axis with the specimen tilted, progressively, parallel to the {111} growth
planes. (B) bcc (2 7 nm thick) Nb viewed along a <111> axis with the crystal tilted,
progressively, parallel to the {110} growth planes. Although there is a distortion in the
shape of the atom column images, the overall arrangement and lattice spacing remains
unambiguous through at least 3.5 degrees of tilt for Cu and 1.4 degrees for Nb. (Images
and analysis are provided by J .W. Heckman.) ................................................... 73

Figure 4. 1 A cross-sectional brightfield CTEM image of the
[Nb250//(Ag/Py/Ag/Py/FeMn)//Nb250]nm spin-valve with 3 nm Py layers grown on Nb.
Note that the individual layers within the spin-valve can be distinguished .................. 80

Figure 4.2 A cross-sectional brightfield CTEM image of the Ag/Py/Ag/Py/FeMn spin-
valve with 3 nm Py layers grown directly on Si substrate showing the first Ag layer grew
as a single crystal .................................................................................... 81

Figure 4.3 A cross-sectional brightfield CTEM image of a second
Ag/Py/Ag/Py/FeMn/Nb multilayer with 3 nm Py layers grown directly on the Si substrate
showing polycrystalline morphology with columnar grains for all the layers ............... 83

Figure 4.4 Cross-sectional bright field CTEM images of the Ag/Py/Ag/Py/Nb multilayers
with (a) 6 nm Py layers and (b) 24 nm Py layers, grown directly on Si substrate showing
polycrystalline morphology for all the layers ................................................... 84

Figure 4.5 A cross-sectional HRTEM image of the first (Agi0/Py3/z‘3.gzo/Py3/FeMng)mm
spin-valve, from an area between the Nb contacts. The individual layers are clearly
distinguished due to the large differences in the average atomic numbers of the multilayer
components .......................................................................................... 85

Figure 4.6 HRTEM images and FFT diffractograms of various layers of the
[Nb250//(Aglo/Py3/Ag20/Py3/FeMn3)//Nb25o]nm spin-valve grown on Nb contact. All of the
component layers display their equilibrium fcc (Ag, FeMn) or L1; (Py) <110> zone
patterns ................................................................................................ 88

xiv

Figure 4.7 A cross-sectional HRTEM image of the (Ag;o/Py3/Ag2o/Py3/FeMng)nm spin-
valve grown directly on Si. The first Ag layer displays a single orientation throughout
the entire region of view (and throughout the entire thin area of the TEM sample) ....... 90

Figure 4.8 HRTEM images and FFT diffractograms of Si and Ag layers of the
(Aglo/Py3/Ag20/Py3/FeMng)nm spin-valve grown directly on Si. The FFT from first Ag
layer is inconsistent with an fcc equilibrium structure ......................................... 91

Figure 4.9 A cross-sectional HRTEM image of the (Aglo/Py24/Ag20/Py24/Nbl0)nm
multilayer grown directly on Si. The interfaces between the layers are sharp and flat,
indicating less interfacial roughness and interdiffusion between the spin-valve
components compared with the one grown on Nb ............................................. 93

Figure 4.10 HRTEM images and FFT diffractograms of various layers of the
[(Aglo/Py24/Ag20/Py24)//Nb10]nm spin-valve grown directly on Si. All of the component
layers display their equilibrium fcc (Ag, FeMn) or L12 (Py) <110> and bcc (Nb) <1 11>
zone patterns ......................................................................................... 95

Figure 4.11 A cross-sectional brightfield STEM image of the initial Ag/Py spin-valve
grown directly on the Si substrate. Points A-D indicate the loci of origin of the X-ray
spectra acquired ..................................................................................... 97

Figure 4.12 EDS X-ray spectra collected from the Si substrate and the first three layers of
the Ag/Py spin-valve illustrated in Figure 4.11. (A) Spectrum collected from the silicon
substrate. (B) Spectrum from the first Ag layer. (C) First Ag layer spectrum with the Ag
L series contribution stripped. (D) Spectrum from the first (unpinned) Py layer. (E)
Spectrum from the second Ag layer .............................................................. 98

Figure 4.13 Simulated EDS spectra of simple stoicheometric ratios of Silver and Indium.
The Ag421n ratio is the best integer match with the experimental spectrum ............... 100

Figure 5.1 (a) Brightfield and (b) darkfield cross-sectional CTEM images of the
(Nb250//Cu1O/FeMng/CoZr2o/Cu2o/CoZr2o/Cu10//Nb250)nm EBSV sputtered on Si substrate
showing polycrystalline morphology with columnar grains ................................. 104

Figure 5.2 (a) Brightfield and (b) darkfield cross-sectional CTEM images of the
(Nb250//Cu10/(Py24/Cu20/CoZr2o/Cu20)x10/Cu10//Nb250)nm hybrid spin-valve sputtered on
Si substrate showing polycrystalline morphology with columnar grains .................. 106

Figure 6.1 Stereographic projections of A1203, Nb, and Cu films. The epitaxial

relationship between the Nb films and A1203 substrates was (110)Nb || (1120)Al2O3 with
[111]Nb || [0001]Al2O3, while the epitaxial relationship between the bcc Nb and fee Cu
variants was the Nishiyama-Wasserman (N -W) relationship ................................ 113

XV

Figure 6.2 (a) Brightfield and (b) darkfield cross-sectional CTEM images of the epitaxial
(Nb200//Cu.oo/Pyloo/FeMnso/lemo)nm multilayer ............................................. 1 15

Figure 6.3 (a) Brightfield and (b) darkfield cross-sectional CTEM images of the
polycrystalline (Nb250//Culo/FeMng/Py3o/CU2o/PY3o/be250)nm spin-valve .................. 1 17

Figure 6.4 (a) Brightfield and (b) darkfield cross-sectional CTEM images of the epitaxial
(N b250//Cu20/Py24/Cu20/ Py24/FeMng//Au1 5/Nb100/Au15)nm spin-valve ....................... 1 18

Figure 6.5 (a) Brightfield and (b) darkfield cross-sectional CTEM images of the epitaxial
(Nb250//Cu2o/Col.5/(Cu2/Col,5)x20//Au15/Nb100/Au15)nm multilayer ......................... l 19

Figure 6.6 A cross-sectional brightfield CTEM image of the epitaxial
(Nb200//Cu100/Py200)nm multilayer with Cu deposited at 500°C showing the poor epitaxial
growth of Cu and Py layers ....................................................................... 122

Figure 6.7 A cross-sectional brightfield CTEM image of the epitaxial
(Nb2oo//Cu20/Py20//Nb20)nm multilayer with Cu deposited at 100°C and rest of the layers
deposited at room temperature. It shows the high-quality epitaxial growth of Cu and Py
layers ................................................................................................ 124

Figure 6.8 (a) A cross-sectional brightfield CTEM image of the epitaxial
(Nb2oo//Cu2o/Py20//Nb20)nm multilayer with Cu deposited at 150°C and the rest of the
layers deposited at room temperature. It shows high-quality epitaxial growth of Cu and
Py layers. (b) The diffraction pattern taken from Nb-Cu-Py shows a portion of the
<110> twin reflections of Cu and Py ............................................................ 125

Figure 6.9 (a) Brightfield and (b) darkfield cross-sectional CTEM images of the epitaxial
(Nb200//CU2o/Py2o//Nb20)nm multilayer with Cu deposited at 150°C showing the high-
quality epitaxial growth of Cu and Py layers. The sample was tilted about 5° parallel to
the growth planes. (c) The diffraction pattern is taken from Nb-Cu-Py layers which
shows the full <1 10> twin pattern of Cu and Py after tilting ................................ 127

Figure 6.10 (a) Brightfield and (b) darkfield cross-sectional CTEM images of the
epitaxial (Nb200//Cu2o/Py20//Nb20)nm multilayer with Cu deposited at 150°C. The inset
diffraction pattern shows the sample was oriented to the (1 T00) zone of sapphire
substrate ............................................................................................. 128

Figure 6.11 A cross-sectional brightfield CTEM image of the epitaxial
(Nb200//Py20/Cu20//Nb2o)nm multilayer with Py deposited at 350°C on Nb showing poor
epitaxial growth of the Py and Cu layers when Py grew on Nb first instead of Cu ...... 130

Figure 6.12 A cross-sectional brightfield CTEM image of the epitaxial
(Nb200//Py2o/CU20//Nb20)nm multilayer with Py deposited at 150°C on Nb showing the

xvi

polycrystalline growth of the Py and Cu layers when Py grew on the Nb first instead of
Cu .................................................................................................... 131

Figure 6.13 Cross-sectional HRTEM image of the Nb-Al2O3 interface and corresponding
F F T patterns of the Nb and A1203 in the Cu/Py/FeMn multilayer, showing epitaxial single

crystal (1 I0) Nb has grown epitaxially on the (l 120) A1203 substrate ................... 132

Figure 6.14 A cross-sectional HRTEM image of the Nb-Sapphire interface and
corresponding FFT patterns of the Nb and sapphire in the Cu/Py/FeMn multilayer,
showing epitaxial single crystal (1T0) Nb has grown epitaxially on (1120) sapphire
substrate ............................................................................................. 134

Figure 6.15 A cross-sectional HRTEM image of the Cu-Nb interface showing multiple
small twin variants near the nucleation interface in the Cu layer of the Cu/Py/FeMn
multilayer. Note that the regions labeled as E are oriented along (111)fcc || (110)],cc while
those labeled N are not ............................................................................ 136

Figure 6.16 A cross-sectional HRTEM image of the Cu-Py interface in the Cu/Py/FeMn
multilayer ........................................................................................... 1 3 7

Figure 6.17 Selected HRTEM images and corresponding FFT diffractograms of Cu and
Py showing <110> zones with growth planes of {1 l 1} ...................................... 138

Figure 6.18 A cross-sectional HRTEM image of the Cu-Co bilayers in the (Cu/Co)x20
multilayer. Note that the observed Cu-Co periodicity is consistent with the nominal layer
thicknesses .......................................................................................... 140

Figure 6.19 Selected HRTEM images and corresponding FFT diffractograms of Cu and
Co showing <110> zones (slightly off axis) with growth planes of {100} ................ 141

Figure 6.20 A cross-sectional HRTEM image of the (Nb250//Cu2o/Py20//Nb2o)nm multilayer
with Cu deposited at 150°C and rest of the layers deposited at room
temperature ......................................................................................... 142

Figure 6.21 An enlarged HRTEM image and corresponding FFT diffractogram of the
bottom Nb showing <111> zone axis of bee Nb .............................................. 144

Figure 6.22 The HRTEM images from selected areas (a) from left side (b) from right side
(c) twin boundary and corresponding FFT diffractograms of the Cu buffer layer showing
<110> zones with growth planes of {11 1} ..................................................... 145

Figure 6.23 Selected HRTEM images and corresponding FFT diffractograms of Py
showing (a) <110> zone with growth planes of {111} (b) a periodic lattice contrast with
the interval of three {111} planes. The additional reflections at 1/3 and 2/3 of the primary
(111) reflection spacings are indicated by arrows ............................................. 147

xvii

Figure 6.24 (a) A cross-sectional HRTEM image showing a periodic contrast with the
three-layer interval and (b) its corresponding Fourier transform of the image in the
massively transformed y grain [99] .............................................................. 151

Figure 6.25 (a) Enlargement of the framed part of figure 5.24. Simulated images of the
overlapping twins (b) for the total thickness of 4 nm at Af= —60 nm and (c) for the total
thickness of 16 nm at Af= —75 nm [99] ........................................................ 152

Figure 6.26 Schematic illustrations of the overlapping twin model to explain the periodic
contrast with the three-layer interval. (a) A projection of atoms along the [1 T0] direction
of the twin-related Py single crystal. (b) Two Py crystals superimposed to each other
showing the atom positions of both crystals become coincident with each other at every
third plane as indicated by arrows ............................................................... 154

Figure 6.27 (a) Schematic illustration of the expected electron diffraction pattern from

superimposed overlapping twin model in y-TiAl [99] (b) The F FT diffractogram of the
epitaxial Py showing overlapping twin reflections ............................................ 155

xviii

CHAPTER I

INTRODUCTION

Giant magnetoresistance (GMR) is a phenomenon in which the resistance of
synthetic multilayers changes dramatically in response to an external magnetic field. The
GMR effect has been the subject of intense investigation due to its already existing and
potential technological applications for magnetic recording devices, magnetoresistive
sensors, and magnetic random access memories. This effect is most usually observed in
magnetic multilayered structures, where ferromagnetic layers (Fe, Co, NiFe) are closely
separated by a non-magnetic layer of a noble metal (Cu, Ag, Au). GMR in magnetic
multilayers has been found to be very sensitive to the layer structure and orientation, as
well as to the interfacial roughness at the micro and atomic level. Therefore, it is critical
to have a fundamental understanding of the details of the layer structures to facilitate the
development of magnetic materials associated with GMR for field-sensing applications.
This dissertation is an attempt to investigate the microstructures of magnetic multilayers
using conventional and high-resolution transmission electron microscopy (CTEM and
HRTEM) complemented with image simulations and fast Fourier transform (FFT)
analysis. These techniques will be used to further characterize crystallographic epitaxy
and texture, layer quality, interfacial roughness, defects, and any changes of

microstructure, which can be correlated with the GMR effect.

1.1 THE GMR EFFECT

The GMR effect is the large relative change in resistance with applied external
magnetic field in ultra-thin magnetic multilayers composed of alternating ferromagnetic
and non-magnetic metallic layers. This effect is due to the quantum mechanical
phenomena of spin-dependent scattering, which is a function of the relative orientation of
the conduction electron spins and the magnetizations of the ferromagnetic materials.

Early studies of unusual magnetoresistive effects were reported by Pratt et al. in
1987 [1] with Ag/M (M=Fe, Ni, Co) layered structures and by Ve’lu et al. in 1988 [2]
with (Au/Co)n multilayers. However, the first evidence of the GMR effect was
discovered independently by two European research groups in the late 19805: Baibich et
al. [3] with single crystal Fe/Cr multilayers grown by molecular beam epitaxy (MBE) and
Binasch et al. [4] with MBE-grown antiferromagnetically coupled single crystal Fe/Cr/F e
sandwiches. While working with large magnetic fields and thin layers of magnetic
materials, both of these groups noticed very large resistance changes when these
materials were subjected to external magnetic fields. Since these discoveries of the GMR
effect, a new chapter has been opened up in understanding the fundamental science
underlying the phenomenon of magnetism and magnetic materials. This progress also
stimulated an interest in deve10ping applications ranging from automotive sensors to
information storage technology.

The most promising application of GMR in magnetic multilayers has been
recording heads in magnetic storage devices, enabling greater storage densities and
smaller head sizes [5,7]. IBM was the first company to announce GMR magnetic

recording read heads in a family of disk drive products in 1997. These drives are capable

of reading extremely small magnetic bits at an areal density of 2.69 gigabytes per square
inch. By the year 2001 , densities of 10 gigabytes per square inch have been achieved [6].
Another important application of GMR is for magnetic field sensors. In 1994, Non-
Volatile Electronics introduced the world’s first products using GMR materials, a line of
GMR magnetic field sensors that can be used for wheel speed and position sensing,
magnetic media and current sensing [7]. Other important applications under study are the
use of GMR materials for very low field sensing and magnetoresistive random access
memory (MRAM), which is an integrated combination of non—volatile thin film magnetic
storage and semiconductor support circuits for high density nonvolatile random access
memory. GMR materials show promise to ameliorate one of the most difficult problems
which has faced MRAM technology — that of a small signal level, leading to relatively
long read access times for memory applications [8].

Conventional magnetoresistance materials such as PermalloyTM (NiFe) exhibit
resistance changes at room temperature of just a few percent in magnetic fields of a few
oersteds [9], whereas GMR multilayers show sensitivities five times greater than
permalloy under the same field conditions [10,11]. Today, the vast majority of GMR
studies are devoted to researching magnetic multilayered structures that show large

sensitivities in low external fields at room temperature.

1.1.1 Origin of the GMR Effect:
The origin of the GMR effect has been theorized to be the spin-dependent
scattering of conduction electrons at ferromagnetic(F)/non-magnetic(N) interfaces and in

bulk F layers [3,12]. It is well known that electrons can exist in two quantum states: spin

up if their spin is parallel to the magnetic field and spin down if their spin is anti-parallel
to the magnetic field. Electron scattering at the PM interfaces depends on whether the
electron spin is parallel or anti-parallel to the magnetic moment of the magnetic layers.
Figure 1.1 illustrates this mechanism for two magnetic states: (a) parallel state (P) with F
layer magnetizations parallel to each other and (b) anti-parallel (AP) state with F layer
magnetizations anti-parallel to each other. The resistance of the magnetic multilayers can
be altered by changing the orientations of the magnetic moments of the F layers to be
either parallel or anti-parallel to each other. It is observed that the resistance of the
structure is much higher when the magnetic moments of the adjacent magnetic layers are
aligned anti-parallel than when they are parallel. Switching from the antiparallel to the
parallel configuration can be achieved by applying a magnetic field. This effect is called

giant magnetoresistance (GMR), and commonly expressed as a percentage of AR/Rsal

[13] as shown in figure 1.2.

1.1.2 GMR Measurements

The GMR effect is usually measured with a geometry of electric current flowing
in the layer planes (CIP) of the multilayer [3,14], as shown in figure 1.3 (a). With a CIP
measurement, it is difficult to separate the bulk and interface contributions to GMR,
because the current paths are randomly distributed throughout the interfaces and bulk
materials. Measurement of GMR with a geometry of the current flowing perpendicular to
the layer planes (CPP) through magnetic multilayers was pioneered by a group at MSU
[15] and is shown in figure 1.3 (b). CPP is an advancement over CIP because it allows

quantitative determinations of the parameters describing the spin-orientation dependent

Spin upA Spin up‘

Spin down' Spin down'

   

(a) P state (b) AP state
Electron scattering occurred

Figure 1.1 Schematic representation of the mechanism of spin-dependent scattering for
two magnetic states: (a) parallel (P) state with the F-layer magnetizations parallel to each
other; (b) anti-parallel (AP) state with the F—layer magnetizations anti-parallel to each
other.

 

 

   

H 2 Hm (P state)

 

 
 

 

 

 

 

 

 

 

4+ l J: i i H»
-15 -10 -s 0 s 10 15
H (KOe)

AR/Rsat = MR°/o = 100 x [R(AP) - R(P)]/R(P)

Figure 1.2 A representation of the anti-parallel (AP) and parallel (P) configurations of the
ferromagnetic (F) layer moments and the magnetic field dependence of the GMR for a
Co/Cu multilayer sample at the first antiferromagnetic coupling peak.

 

h)
CD

+1 +1: -U -1 .53.

,— ____________

MR(H)% (CIP)

   

UI
v'vvaTv'.

 

 

.l....l ...i....i..-.i.
-0.5 -0.25 0 0.25 0.5

 

(a) CIP configuration and CIP-MR in a (Co/Ag) magnetic multilayer

 

   

 

 

 

133,
+1 +1? 5 H0

E- E
E3 60:
e: 403
E :
“2‘ 20_

OE I I I I I

3.7 -I -0.5 ‘—0.25 0 0.25 0.5‘

H(k0e)

(b) CPP configuration and CPP-MR in a (Co/Ag) magnetic multilayer

Figure 1.3 Schematic diagrams of the (a) CIP and (b) CPP configurations and
corresponding examples of the GMR measurements in a (Co/Ag) magnetic multilayer.
CPP-MR=45% > CIP-MR=9% [15].

electron scattering in the F layers and at the PM interfaces. In CPP measurements,
superconducting Nb contacts are used as current and voltage leads to ensure uniform
current density through the multilayer from the top and bottom of the multilayers. A
primary difference between measurements in the two configurations is that the
magnetoresistance (MR) is usually greater in a CPP sample than that in a corresponding

CIP sample, i.e. CPP MR > CIP MR.

1.2 GMR MAGNETIC MULTILAYERS AND SPIN-VALVES

The GMR effect was first observed in single crystal Fe/Cr multilayers grown by
MBE [3] and in polycrystalline Fe/Cr structures grown by the much simpler technique of
magnetron sputtering [16]. These types of the multilayers consist of alternating
ferromagnetic (Fe) and non-magnetic (Cr) layers. Initially, with a low external field, the
magnetic moments of the Fe layers are aligned in an anti-parallel (AP) state to minimize
the energy. When a saturating magnetic field is applied, the magnetic moments of the Fe
layers are aligned along the field direction (parallel magnetic moments). The maximum
GMR effect for this arrangement was observed with single crystal (001)Fe/(001)Cr
multilayers of 30 A and 9 A thicknesses respectively in a magnetic field at 20kOe as
shown in figure 1.4; similar results were also observed in polycrystalline Fe/Cr structures.
These experiments also revealed that GMR can be observed in a wide variety of magnetic
systems such as Co/Cu [17,18], Fe/Cu [l9], NiFe/Cu [18,20], and NiFe/Ag [21]
multilayers, and the largest magnetoresistance effect has been observed in Co/Cu based

structures [22].

Due to the requirement of a high magnetic field to align the magnetic moments
parallel to each other (for example: 20 kOe for single crystal Fe/Cr multilayers as shown
in figure 1.4), an alternative GMR arrangement, known as the exchange-biased spin-
valve (EBSV) configuration, was invented by Dieny et al. [23]. This configuration is
desirable because of its high sensitivity in a low magnetic field. Due to this feature, these
structures have generated a high level of interest as existing and potential candidates for
use in magnetic recording heads of the next generation of hard disk drives in computers.

The typical EBSV structure is in the form of F/N/F/AF [24], which consists of
two ferromagnetic (F) layers separated by a non-magnetic (N) layer. An additional anti-
ferromagnetic (AF) material is provided either on the top or on the bottom of the trilayer
to couple to the adjacent F layer and pin the moment of this layer in a fixed direction.
The other F layer is free to rotate in small fields, as shown in Fig. 1.5 [9,10,23,25-32].
This arrangement allows the magnetic moments to be easily switched between parallel
and anti-parallel configurations. Another type of spin-valve, the hybrid spin-valve, was
also studied. Hybrid spin-valves differ from EBSVs, in that the anti-ferromagnetic
materials are no longer needed. This type of spin-valve uses either different thickness of
the same type of ferromagnetic layer or different ferromagnetic materials to enable the
magnetization of both ferromagnetic layers to be rotated independently [33]. This

dissertation will mainly focus on spin-valve structures.

1.3 CORRELATION BETWEEN GMR AND MULTILAYER STRUCTURES
With the size of magnetic devices becoming smaller and smaller (typical

dimensions of the single layer thickness are in the nanometer range), there is a greater

 

 

 

 

 

 

(Fe 30A/Cr 12A)35
(Fe 30A/Cr 9A)“
0.5. l
:v HS
«in -30 -20 -10 0 10 20 30 40 ’
Magnetic Field (kG)

Figure 1.4 The first discovery of the giant magnetoresistance phenomena in single crystal
Fe/Cr superlattice. The maximum GMR was observed with (Fe 30A /Cr 9 A)6o
superlattice [3].

10

.533 Ferromagnetic layer (free layer)
Non-magnetic layer (spacer)
Ferromagnetic layer (pinned layer)

A ntiferromagnetic layer (pinning layer)

 

(a)

 

 

 

 

 

 

15.6 -

A
15.2 —
14.8 -— AR

MR % (CPP)
14.4 -
7
14.0 ‘
Parallel state Anti-Parallel state _
13_6 l l l l l
-60 -40 -20 0 20 40 60
Magnetic field (0e)
(b)

Figure 1.5 Schematic representation of (a) an EBSV structure and (b) the
magnetoresistance response of a spin-valve [NiFe24/Ag2o/NiFe24/FeMn3]nm (NiFe-denoted
as Py) [34].

interest in establishing the correlation between the structures of magnetic
multilayers/spin-valves and the GMR effect. For example, it has been found that the
GMR effect varies with layer thickness [29]. A broad range of studies on
microstructures, such as layer interfacial roughness, layer quality, crystal orientation, and

layer thickness, has shown these factors to have important relations with the GMR effect.

1.3.1 Interfacial or Layer Roughness

Since the discovery of the GMR effect in magnetic multilayers or spin-valves,
numerous observations have indicated that interfacial roughness plays a significant role
in the GMR effect [31,35-41]. Some studies conclude that in (Fe/Cr) and (Co/Cu)
multilayers, the interfacial roughness produces very strong spin-dependent scattering of
conduction electrons and therefore tends to increase GMR [40]. In contrast, in (N iF e/Cu)
multilayers or spin-valves, the interfacial spin-dependent scattering appears to be not as
critical as the bulk scattering, and that an increase in the interfacial roughness was
observed to lead to a systematic decrease in GMR [31,36]. Recently, a quantitative
structural study of Fe/Cr superlattices showed that the interfacial or layer roughness is
correlated and increases cumulatively with the number of the bilayers [37] as shown in
figure 1.6 and with sputtering pressure [38] and as shown in figure 1.7 (a) and (b); the
GMR were also increased with both factors. Reports from different groups on the
correlation between the interfacial roughness and GMR effect are not conclusive and

require further theoretical and experimental investigation [39-41].

1.3.2 Layer Quality

Another variable influencing the GMR effect is the crystallographic quality of the

 

6 (nm)

 

 

 

 

M

Figure 1.6 The quantitative roughness of each individual Cr layer inside a series of
superlattices [Fe(3nm)/Cr(1.2nm)]N with the number of the bilayers N = 20, 40, and 60.
Roughness of each individual bilayer as a function of the bilayer index M for N = 20 (o),

N = 40 (D), and N = 60 (O). 0 represents the roughness. Note that the roughness
increases with the bilayer index [3 7].

112nm

 

Figure 1.7 Cr mapping taken on [Fe(30A)/Cr(12A)]x20 multilayer grown at (a) 4 mTorr
and (b) 10 mTorr showing the layer roughness increased with sputtering pressure [3 8].

overall multilayered structure and in particular of the interfaces. Some studies have
reported that high-quality Co/Cu multilayer structures grown on Fe buffer layers result in
high GMR, as shown in figure 1.8 [42]. Other studies state that Fe(001) and Cr(001)
have similar bulk crystallographic structures (bcc), with lattice parameters of 2.88 and
2.87 A. This good crystallographic matching allows for the epitaxial growth of (Fe/Cr)
multilayers, and in turn leads to a larger GMR amplitude, although Fe and Cr tend to
interdiffuse to one or two monolayers at the interfaces [43]. Lower GMR has been
observed in Fe/Cu systems due to the different bulk crystal structures of these two

elements (bcc vs. fcc) [44].

1.3.3 Crystal Orientation or Growth Direction

Using x-ray diffraction techniques, Nakatani et al. [45,46] found that the preferred
orientation of the multilayer changes with the thickness of the non-magnetic material in
certain magnetic multilayers. They reported that the highest GMR ratios are obtained
when the orientation is <100> in NiFe/Cu multilayers grown by ion beam sputtering on a
Si (100) substrate. Pardavi-Horvath [13] has also been established that certain
orientations of the non-magnetic layers are related to the observed GMR effect. Jungblut
et al. [47] found that the exchange-biasing field and coercive field strongly depend on the
crystal orientation for highly oriented MBE grown NiFe/FeMn exchange-biased spin-

valves.

l.3.4 Layer Thickness

The dependence of the GMR effect on layer thickness in various multilayer

15

 

 

 
  
   

40 - buffer/cap layer
P l Fe/Fe ..
30 .. 2 F e/Cu ‘
AR/R (%) _ 3 Cu/Fe .[

4 Cu/Cu

  

20

10

 

 

 

H (kOe)

Figure 1.8 Cross-section transmission electron micrographs of(a) Si//[Cu/(Co/Cu)16/Cu]
with Cu buffer layer and (b) Si//[Fe/(Co/Cu)1(,/Cu] with Fe buffer layer, which show
considerable waviness and buckling of the layers in (a) and comparatively smooth and
flat layers in (b). (c) An illustration of the relationship between GMR and Co/Cu
superlattices [42]. Note that the Co/Cu grown on the Fe buffer layer has a much larger
magnetoresistance, indicating high-quality structure results high GMR value.

systems has been studied extensively. Baibich et a1. [4] found that the GMR of
antiferromagnetically coupled Fe/Cr multilayers decreases with increasing the non-
magnetic Cr thickness. The variation of GMR with the magnetic layer thickness was also
studied. Dieny et al. [29] demonstrated that the magnetoresistance amplitude always
shows a maximum when the thickness of the magnetic layers is varied from a monolayer
thickness to a few tens of angstroms (A). It has also been shown that grain or column

size increases with increasing layer thickness [29].

1.3.5 Growth and Deposition Conditions

The variation of the GMR amplitude with deposition temperature has been
investigated by various groups [18,26,27,48,49]. Generally the GMR effect decreases
monotonically with increasing temperature. Experiments by W.F. Egelhoff Jr. et al. [50]
have shown that the GMR effect in Co/Cu/Co spin-valves was 14% for a growth
temperature of 150 K, 10% for a growth temperature of 300 K, and 0% for a growth
temperature of 450 K.

As discussed previously, the GMR effect is critically affected by interfacial
roughness. Further studies have shown that roughness depends strongly on growth
conditions and deposition parameters [51]. This topic is still not yet well understood,

mainly due to the non-thermodynamic growth conditions of most deposition techniques.

1.4 CHARACTERIZATION TECHNIQUES
There are a number of characterization techniques that have been applied to study

the microstructures of magnetic multilayers and spin-valves. X-ray diffraction (XRD) is

most commonly used to determine structural information averaged over multilayer
structures [35,45,46]. Reflection high-energy electron diffraction (RHEED) is utilized
for determining the growth features of magnetic multilayers [52,53]. More recently,
electron backscattered patterns (EBSP) have begun to be used for measuring
crystallographic orientations of the magnetic materials [54]. Additional details about the
surface roughness morphology is available from atomic force microscopy (AFM) [55]
and scanning tunneling microscopy (STM) [56]. Other techniques, such as low-energy
electron diffraction (LEED) [57], high-resolution energy dispersive ‘x-ray spectroscopy
(HREDS) [58], and X-ray absorption fine structure (XRAFS) spectroscopy [59], have
also been performed.

Cross-sectional transmission electron microscopy including conventional and
high-resolution transmission electron microscopy (CTEM and HRTEM) was the primary
tool used to investigate the structural information in this thesis [60-62]. The main
advantage of using cross-sections instead of planar sections is that it allows investigation
of interface structures between the layers, or any changes in orientation and new phases
that may be present. CTEM images can provide useful information about overall layered
microstructures and has been used to characterize microstructural features of the
multilayers, such as the uniformity of the layers, grain or crystallite size, the orientation
and texture of the grains, and the propagating roughness of the layers as a function of
growth. Additional details about microstructural features at the atomic level are available
from high-resolution transmission electron microscopy (HRTEM). When complemented
with image simulations and fast Fourier transform (FFT) image analysis techniques,

HRTEM has the ability to extract accurate structural information concerning crystal

18

structures and interplanar spacings [63]. To get additional analytical information energy
dispersive x-ray spectroscopy (EDXS) and electron energy loss spectroscopy (EELS)

have also been used [58].

1.5 OBJECTIVES OF THIS RESEARCH

It is known that the GMR effect in magnetic multilayers and spin-valves depends
on structure at the micro and atomic level, where interfacial or layer roughness, layer
quality, and layer orientation are important factors. Therefore, the initial objective of this
study was to characterize the interfacial structures of polycrystalline spin-valves using
cross-sectional CTEM and HRTEM.

Polycrystalline Cu/Py and Ag/Py multilayers and spin-valves were first studied.
In the Cu/Py spin-valves, non-equilibrium bcc structures appeared in addition to the
equilibrium fcc Cu, FeMn, and L12 Py structures in the layers. In one of the Ag/Py spin-
valves, the first Ag layer grew as an epitaxial single crystal although the remaining layers
displayed the expected polycrystalline morphology. It was found that indium
contamination existed in this sample when the spin-valve was grown directly on the Si
substrate.

Because of the non-equilibrium phases that were found in polycrystalline Cu/Py
and Ag/Py multilayers and spin-valves, and because of the difficulties in imaging
interfaces in polycrystalline samples, the initial objective of this study was modified to
focus on the interfacial structures in epitaxial multilayers which diminishes the effects of
structural defects associated with grain boundaries. However, as the program progressed,

it became clear that growing such samples was not a trivial matter. Thus, the final

objective of this study was to assess the structures of samples grown to be epitaxial.
Through an iterative process, success has now been achieved in growing smooth epitaxial

samples using a sputter deposition technique.

20

CHAPTER II

EXPERIMENTAL PROCEDURES

2.1 GMR MATERIALS AND FABRICATION:

The magnetic multilayers and spin-valves described in this study were fabricated
by Dr. Reza Loloee of the department of Physics and Astronomy at Michigan State
University. Details of the fabrication system and sputtering procedures have been
described elsewhere [33]. Two sets of multilayer/spin-valve samples were produced by
dc. magnetron sputtering: one set included as-grown polycrystalline multilayers and
spin-valves sputtered onto Si substrates; the other set included attempts to grow

epitaxially multilayers and spin-valves on sapphire (A1203) substrates.

2.1.1 Polycrystalline Multilayers/Spin-Valves And Fabrication

All of the polycrystalline multilayer/spin-valve samples were sputtered onto the Si
substrates in the temperature range of —67°C to +56°C as shown in table 2.1. Figure 2.1
shows the duplex design [15] used for sputtering polycrystalline samples that allows both
current-in-plane (CIP) and current-perpendicular-to-plane (CPP) measurements
simultaneously. Both Nb contact layers and spin-valves were produced by sequential d.c.
magnetron sputtering onto a p-type (001) Si substrate, while masking as appropriate for
the different layers. Spin-valve multilayer structures were sandwiched between 1.1 mm
wide, 250 nm thick, crossed Nb strips. The strips overlapped in a small area of the spin-

2

valve ~ 1.2 mm . When cooled in liquid helium at 4.2 K, these thick Nb strips become

superconducting and form equipotential surfaces above and below the multilayer,

21

Table 2.1 Polycrystalline Spin-Valve Samples and Sputtering Temperatures

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

No. Sample Composition and Thickness (nm) S??? (131)“ g SV Type
761-4 Nb250//Cu1o/FeMng/Py3/Cu20/Py3/bezso -33
763-1 1 Nb250//Cu1o/FeMng/Py30/Cu20/Py3o//Nb250 6.5
826- l 2 Nb250//Culo/FeMng/Py3/Cu20/Py3//Nb10 18
826-13 Nb250//Cu. o/FeMng/Py24/Cu20/Py24//Nb1o 22
881-la Nb250//Cu1o/FeMng/Py30/Cu2o/PY3d/Nblo -65
881-2a Nb250//Culo/FeMng/Py3/Cu20le3//Nbm ~56 EBSV
719-1 Nb250//Aglo/Py3/Ag2o/Py3/FeMng//Nb250 -67
719-1 Aglo/Py3/Ag20/Py3/FeMn3 -67
990-8b Ag] o/Py3/Ag20/ Py3//N b 1 0 -14
990-8a Aglo/Py6/Ag2o/Py6//N b 1 o -14
990-7b Agm/Py24/Ag20/Py24//Nb10 -31
1042-3 Nb250//Cu.o/(Py24/Cu20/CoZr20/Cu20)x10/Cu10//Nb250 8 Hg“
1043-7 Nb250//Cu1o/FeMng/CoZr2o/Cu20/CoZr20/Cu10//Nb250 -7 EBSV

 

 

 

 

22

 

 

Top Nb Contact

CPP Area, A=1.2mm2

~12 mm

 

 

~12 mm

Figure 2.1 Top view of the duplex designed CIP and CPP measurement configuration.

23

which ensure a uniform current distribution through the layers for CPP resistance
measurements [64]. The sputtering took place using a chilled substrate stage which tends
to increase a few degrees in temperature during the sputtering. The direction of
magnetization of the pinned ferromagnetic layer was fixed by annealing the samples in an
external magnetic field at about 170°C for about 10 to 15 minutes. The following section
will introduce two kinds of polycrystalline spin-valve samples: exchange-biased spin-

valves (EBSVs) and hybrid spin-valve.

2.1.1.1 Exchange-Biased Spin-Valve (EBSV) Samples

Three EBSV systems, Cu/NiFe, Ag/NiFe, and Cu/CoZr were investigated. These
structures are based on either ferromagnetic PermalloyTM (Py = Ni..,,Fex with x z 0.16) or
CoZr ferromagnetic layers separated by either a Cu or a Ag non-magnetic layer. At the
end of the alternating non-magnetic and magnetic layers, an anti-ferromagnetic FeMn
layer was sputtered to be used as an exchange bias material for pinning the magnetic
moments of one of the two ferromagnetic (either Py or CoZr) layers in a fixed direction.
The magnetic moment of the other ferromagnetic layer remained free to rotate under the
influence of an external applied magnetic field. This arrangement allows the
magnetizations of the ferromagnetic layers to easily switch between parallel and anti-
parallel configurations. If the thicknesses of both magnetic layers are identical, the
EBSV structure is referred to as a symmetrical EBSV.

The first set of samples included six Cu/Py EBSVs with configurations of
[Nb250//Cu1o/FeMng/Pyx/Cu20/Py,,.//Nb250]nm (where x varies from 3 nm to 30 nm) layer

arrangements, which were deposited by dc. magnetron sputtering at different

24

temperatures onto p-type (100) Si substrates with natural thin oxide coatings at different
temperatures. Three spin-valve samples with tpy = 3 nm were deposited at substrate
temperatures of approximately -56°C, -33°C and 18°C. The other three spin-valve
samples with tpy = 24 or 30 nm were deposited at substrate temperatures of approximately
-65°C, 65°C and 22°C.

The second set of polycrystalline spin-valve samples examined in this study
consisted of Ag/Py EBSVs. The first sample studied was in the form of
[Aglo/Py3/Ag20/Py3/FeMng] sandwiched between crossed Nb contacts and in adjacent
uncrossed areas. Three additional Ag/Py spin-valve samples in the form of
[Aglo/Py3/Ag20/Py3/FeMng/be20]nm and [Ag]O/Pyx/Ag20/Pyx//Nb10]nm with x = 6 nm or 24
nm were fabricated under the same conditions as the first sample, but sputtered directly
onto an area of approximately 1 cm2 of Si substrate.

The third set consisted of one CoZr based symmetrical EBSV in the form of

[Cu]()/FeMng/C0Zr2O/CU2o/CoZr20/Cu10]nm between crossed Nb contacts.

2.1.1.2 Hybrid Spin-Valve Sample

One hybrid spin-valve structure, based on different ferromagnetic materials Py
and CoZr separated by a non-magnetic Cu layer, was studied. The reason for using two
different magnetic layers is to allow independent rotation of the direction of the
magnetizations in these layers. This CoZr based hybrid spin-valve was in the form of
[Nb250//Culo/(Py24/Cu20/C0Zr20/Cu20)x10/Cu10/be250]nm. The Py layer rotates first and

easier than CoZr in this sample.

25

2.1.2 Epitaxial Multilayers/Spin-Valves And Fabrication
The substrates for epitaxial GMR multilayer fabrication were (1120) oriented

sapphire (A1203) single crystal wafers that had been pre-deposition annealed at 1200°C.

Single-crystal (1 I0) Nb contact layers were sputter deposited on these annealed sapphire
substrates at a substrate temperature of 750°C, and annealed at 950°C for about 10
minutes [54]. Cu buffer layers were then deposited onto the Nb contact layers at
temperatures ranging from room temperature to 500°C; some of these buffer layers were
then annealed at 350°C. Subsequent Co or Py or FeMn films were deposited onto the Cu
buffer layer at various temperatures; and the rest of the multilayers were deposited at
80~100°C. The top Nb layers were applied at room temperature. All films were grown
with a 0.325 Pa Ar chamber pressure (base pressure ~ l.3><10'6 Pa). The deposition rate
was 0.3 nm/s for Nb, Co and Py; 0.4 nm/s for Cu as determined by Loloee [33]. Table

2.2 shows the epitaxial samples in this study.

2.2 CROSS-SECTIONAL TEM SAMPLE PREPARATION:

The technique used for cross-section TEM thin foil preparation was the slab-on-
ring method developed by Romano et al. [65,66] and Shaapur and Park [67], and
modified by Howell [68] and Geng, et al. [63]. In this method, a composite slab was
formed by using GatanTM G-l epoxy to join two pieces of the multilayer face-to-face.
Cross sections of this slab were made normal to the layer plane. The individual pieces
were then epoxied onto molybdenum rings and pre-thinned using a dimpler and low
angle ion milled to electron transparency. Figure 2.2 shows an overview of the cross-

section TEM sample preparation sequence. The detailed procedure to form a cross-

26

Table 2.2 Epitaxial Samples and Growth Temperatures

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Cu (buffer layeg Py Co
N Sample Composition and Thickness .
0' (nm) Growth T Annealing Growth Growth
(°C) T (°C) T (°C) T (°C)
20-1 Nb250//Cu20/(Co(,/Cu6)x6//Nb.00 150 350 N/A 80~100
20-3 Nb250//Cu20/(C06/CU6)X6//NbIOU 450 --- N/A 80"] 00
20-4 Nb350//CU30/(CO(,/CLl(,)X6//Nbl00 450 --- N/A (131530)
27.3 Nbg50//CUgo/Py34/CUgo/Py24/F6Mngl/AUl5/ 350 "- 80~I 00 N/A
NblOO/AUIS
27-4 Nb250//CUgo/Py24/CU20/Py24/FCMl'lg//AU15/ 450 “_ 80~I 00 N/A
NbIOO/AUIS
28-3 Nb250//CU20/C01 5/(CU2/CO] 5) 350 --- “_ 400
X20//AU|5/Nb100/AU|5 (Ist CO)
31-2 Nb250//CUloo/Pyloo/FCMII50//meo 150 350 350 N/A
37-3 Nb250//CU20/P)’24/CUzo/Py24/FCMII3//AUl5/ 150 350 80~I 00 N/A
meo/ Allis
40-3 Nb250//CU30/F€Mng/Py6/CU24/Py34//AUl5/ 350 "_ --.. N/A
meo/AUIS
41-2 Nb200//Cum/me 500 --- N/A N/A
4 I -3 Nb200//CU l oo/Py200 500 --- 80~I 00 N/A
43-] Nb200//Cu20/Py20//Nb20 100 --- RT N/A
43-2 meo/lCUgo/PyM/[Nbgo 150 --- RT N/A
43-3 Nb200/ny20/Cu20//Nb20 RT --- 350 N/A
43-4 Nb200//Py20/CU20//Nb30 RT --- 150 N/A

 

 

 

 

 

 

27

 

section TEM sample was as follows:

Form a composite slab:

1.

8.

Section multilayer sample on substrate into about 2.2 mm wide slabs with a
StruersTM Accuton-S diamond saw operated at 3000-rpm wheel speed with a 5
[rm/sec feed speed.

Clean slabs in acetone, followed by a rinse with ethanol.

. Glue two slabs of the multilayer face-to-face with a thin layer of GatanTM G-l

epoxy.
Clamp multilayer sample with gentle pressure in a spring-loaded TeflonTM lined
clamp.

Cure epoxy on a hot plate for about 10 minutes at 130°C, then cool it down.
Mount the resulting stack on a 3 x 1 inch micro slide with CrystalbondTM
mounting wax.

Slice the stack perpendicular to the growth plane into 800~1000 um thin slabs
with a diamond saw using Struers Accuton-S operated under the same conditions
as in step 1. This step prepares the cross-section interfaces of the slabs to be

polished.

Dissolve CrystalbondTM mounting wax in acetone and rinse the slabs by ethanol.

Grind and polish slabs to ~100 pm thick:

9.

Mount a composite slab onto a 19 mm diameter sapphire flat with CrystalbondTM
mounting wax and attach it into a sapphire flat holder that fits into the polishing

block. (The use of the polishing block grinder allowed the slabs to be uniformly

polished to a thickness of 100 um while maintaining parallel top and bottom

28

     

Score
and Break

 

Glue on
Mo-ring

Low-Angle Dimpling
Ion milling

Figure 2.2 Schematic representation of the cross-section TEM samples preparation.

29

surfaces.)

10. Grind and polish one side of a slab with a series of diamond lapping films (15, 9,
6, 3, l, 0.5 mm AlliedTM Diamond Lapping Film).

1 1. Separate the slab from the sapphire flat using acetone; rinse with ethanol.

12. Measure sample assembly thickness using a micrometer to determine the amount
of material needed to be removed by grinding.

13. Glue polished side of the slab to a 3 mm in diameter molybdenum support ring
with GatanTM G-l epoxy and cure epoxy on a hot plate for 10 minutes at 130°C
using an 8 mm thick piece of TeflonTM to prevent attachment of the slab and ring
to the hot plate itself.

14. Mount the slab and Mo ring to the center of a sapphire flat with a thin layer of
CrystalbondTM mounting wax.

15. Measure the total thickness of sample assembly plus Mo-ring and sapphire flat.

l6. Grind and polish the unpolished side of sample assembly to ~100 um excluding
the ~30 mm thickness of the Mo-ring with a series of diamond lapping films (same
sequence as used in step 10).

Dimple slabs to 10~20 pm thick:

TM Group DSOOi Dimpler platen using

17. Mount and center the assembly on a VCR
an eccentric platen assembly.

18. Dimple the top surface of the Si based samples using a 15.5 mm stainless steel
dimpling wheel with 1 pm diamond slurry in reservoir; or for the sapphire based

samples, dimple using a 1 pm embedded-diamond lapping wheel with VCRTM

lapping vehicle as the lubricant until obtaining a center thickness of about 10 to

30

20 11m.
19. Final polish with colloidal diamond (0.05 nm) to a specular (mirror like) surface.
20. Dismount the sample assembly from sapphire flat with acetone and rinse in
ethanol and air-dried.
Note: The final dimpling depth of the cross-section was monitored by removing both the
sample and platen and placing the assembly in a transmitted light microscope. At a
thickness less than 10 pm, the silicon substrate will transmit red light. The thickness of
the sapphire substrate is difficult to monitor since the sapphire itself is transparent.
However it was found that the samples would have a tendency to crack before a thickness
of 10 um was reached due to a lack of dampening control on the dimpler. It was
determined that the best way to monitor the depth was to calculate the depth using the
approximation, d = rz/D, relating the diameter of dimple (Zr) and wheel diameter (D) to
dimple depth (d). The dimple diameter was measured under the microscope using a
calibrated reticule. Based on the amount of material removed the estimated thinnest
portions of the discs prior to ion milling were approximately 20 um.
Ion-mill samples to electron transparency:
21. Mount the sample assembly on a Gatan DuoPostTM support and ion mill with a
Gatan M691 PIPSTM ion mill.
22. Initially ion mill the center of sandwich interface at i3°~ i4° beam incident angle
with 5 keV Ar+ while turning at 3 rpm until perforation.
23. Final ion mill at i2° or less until samples were thin enough for HRTEM

observations.

31

2.3 CHARACTERIZATION AND IMAGE ANALYSIS TECHNIQUES:
Cross-section transmission electron microscopy (TEM) has been widely
recognized as one of the most powerful techniques to perform analysis of the layers and
interfaces of magnetic multilayers and spin-valves. The combination of a local
diffraction pattern with the cross-sectional image provides unique and direct information
on the local multilayer structures. With the advancement of high-resolution TEM, direct
atomic structural imaging can be achieved with proper specimen preparation techniques

and computer image simulations.

2.3.1 Conventional Transmission Electron Microscopy (CTEM)

Conventional transmission electron microscopy (CTEM) was used to characterize
the overall morphology of the cross-sectional thin foils. Both bright field and axial dark
field imaging were carried out in conjunction with selected area electron diffraction
(SAED) to characterize overall microstructure features, such as the uniformity of the
layers, grain or crystallite size, the orientation and texture of the grains, and the
propagating roughness of the layers as a function of growth. All samples analyzed were
tilted to a low index zone of the Si or sapphire substrates in order to orient the substrate
plane and the growth plane of layers edge on.

The cross-section TEM thin foils were also characterized by CTEM to determine
if the samples were thin enough for high-resolution transmission electron microscopy
(HRTEM) study.

In this study, CTEM observations were performed using a Hitachi H-800 operated

at 200 keV with a double-tilt specimen stage. Additional CTEM images were recorded at

32

the University of Michigan Electron Microbeam Analysis Laboratory on a J EOL-4000EX
microscope operated at 400 keV with a double-tilt top entry stage. Images and SAED
patterns were recorded on KodakTM Electron Microscope film 4489 (200 keV) and SO-

163 (400 keV).

2.3.2 High-resolution Transmission Electron Microscopy (HRTEM)

The cross-sectioned samples which were determined to be suitable for high-
resolution TEM study were further examined using a JEOL-4000EX operated at 400 keV
at the University of Michigan Electron Microbeam Analysis Laboratory. HRTEM was
used to directly image the details of the individual layers at the atomic level, such as
crystal lattice, interfaces, and defects.

In general, HRTEM imaging requires: (1) a low index zone; in this study, the

polycrystalline and epitaxial samples, were tilted to low index zones normal to the (001)

Si and (1120) A1203 substrate planes respectively, to orient the layers edge-on: (2) thin
specimens (thickness < 50 A); when the specimen is too thick (> 50 A), the weak-phase
object approximation is not satisfied.

HRTEM images are essentially phase-contrast images. Often, the contrast in such
images is described by the contrast transfer function (CTF), which is a theoretical
parameter that models the performance of the microscope objective lens. This function is
applied to the Fourier transform of the wave function after it exits the sample. After
application of the CTF, the Fourier transform is again applied, resulting in the sample
image. The CTF is described as,

T(u) = 2A(u) sin x(u)

33

where T(u) is the transfer function, A(u) is the aperture function, u is a reciprocal-lattice
vector (frequency spacings). and x(u) is the phase shift of the electron that occurs in the
sample. The best resolution corresponds to the first zero position of T(u) vs. sinx curve,
which results in optimized balancing of the effect of spherical aberration against a
particular negative value of defocus, as shown in figure 2.3. In the resulting wide
plateau, all frequencies are transferred in a nearly equal manner. This particular defocus
is called Scherzer defocus. Therefore, to optimize the contrast, all HRTEM imaging was

carried out near the Scherzer defocus condition, which is described as

AfScherzer: C A

S

where Cs is the spherical aberration and it is the wavelength. When operating under
Scherzer condition with a properly oriented thin crystal, the image can be directly
interpreted in terms of the projected potential corresponding to the positions of the atom
columns.

The JEOL-4000EX has a 0.17 nm point-to-point Scherzer resolution. Within this
limit only the <001> and <111> bcc Nb, and the <011> fcc Cu, Ag, Py, and FeMn zone
axes are directly interpretable; the values above this limit are shown in bold in Table 2.3.
In the hcp structure of A1203 the <0001> and <01T0> axes have spacings above the
Scherzer limit as well.

At different conditions of defocus, the frequencies will be transferred differently,
resulting in different plots of sinx as a function of u. However, these plots will be
damped out by the envelope damping function which includes the spatial coherence
spread of the electron source and chromatic aberration effects, as shown in figure 2.4.

Therefore, above a certain frequency, no information will be passed, resulting in a new

34

':

 

 

 

 

 

 

 

U

Scherzer Resolution

 

 

 

U

 

 

 

Figure 2.3 A schematic representation of sinx(u) versus 11 without damping of the higher
spatial frequencies. The first zero position of sinx curve corresponds to Scherzer

resolution [69].

35

Table 2.3 Unit Cell Parameters for Layer Materials

 

 

 

 

 

 

 

 

 

 

 

 

 

 

C stal Lam“ s acin
Materials ”y Parameter p g <001> <011> <111>
Structure (A)

(A)
pt. to pt. 2.34 1.66 2.70

Nb bcc 3.31
line to line 1.66 1.66 2.34
pt. to pt. 2.21 1.47

Cu fcc 3.61
line to line 1.81 1.28
pt. to pt. 2.50 1.67

Ag fcc 4.09
line to line 2.05 1.45
. pt. to pt. 2.17 1.45

Py L12 3.55
line to line 1.78 1.26
pt. to pt. 2.22 1.48

FeMn fcc 3.63
line to line 1.82 [.28
pt. to pt. 2.17 1.45

Co fcc 3.54
line to line 1.78 1.26

 

 

 

 

 

 

Note: the Bold numbers represent the values above the Scherzer limit (1.7 A

 

0.

while values in Italics represent spacings below the information limit (1.4 A).

* Bulk PermalloyTM (Py) displays the fcc derivative L12 structure. However, because of the fairly small
differences in the atomic scattering factors and because the Py used in this study is significantly off-
stoichrometric, the superlattice reflections are very weak and do not appear on the FFT diffractograms. For
the same reasons, the L12 ordering is not evident in the HRTEM images. Thus, for the purposes of this
work, Py will be treated as fcc.

36

 

 

 

Information Limit

 

Figure 2.4 A schematic representation of T(u) versus 11 modified by the damping
envelope (dash line) which ends up the information limit. [69]

37

resolution limit which is called the information limit. If the defocusing condition is
properly modified, resolution limits can be improved down to the information limit. The

available defocusing values are given by

Afn I A/Seherzer J27:
where Afgchmc, is Scherzer defocus and n is an integer. When n=1, this relation
represents the Scherzer defocus condition. When n is larger than 1, good contrast
imaging is also possible by making use of “passbands” in the transfer function, which
allows higher spatial frequencies to be included in the image and permits access to finer
detail of the structure. However, the images so obtained may not give a true image of the
specimen because information at certain frequencies, such as defects, could be lost. In
this case, the image cannot be interpreted directly. Rather, image simulations, FFT
diffractograms, and image processing are necessary to complement the analysis of
HRTEM images. The JEOL-4000EX has an information limit of 0.14 nm; values below

this limit are shown in italics in Table 2.3.

2.3.3 Fast Fourier Transform (F FT) Analysis

It is difficult to obtain electron diffraction patterns directly from areas within
grains in the individual spin-valve layers due to their small width normal to the beam.
Therefore, fast Fourier transform (FFT) image analysis has been employed to assist the
interpretation of the HRTEM images. The FFT diffractograms of the experimental
HRTEM images reveal equivalent plane spacing and angular information from those
selected areas of the specimen.

In this study, FFT image analysis was accomplished by capturing selected area

38

of constant magnification HRTEM prints using a flat bed scanner, then processing and
analyzing the images using the NIH Image [70] and/or Adobe PhotoShop®. The Si, Nb,
and A1203 reflections were used as internal calibration standards to determine the
effective camera lengths (for a given micrograph negative) for interpreting the FFT
diffractograms. Based on these calibrations, the individual layers were then accurately
characterized from the diffractograms by measuring the lattice parameters and interplanar
angles of the layer phases.

HRTEM image simulations were carried out using the EMS software package
[71] by J.W. Heckman of the department of Materials Science and Mechanics at
Michigan State University, details have been described in [63]. These simulations have
been carried out for changes of thickness and defocus, based on the fixed microscope
parameters of: accelerate voltage 400 KeV, spherical aberration C, 1.1 mm, beam
convergence semi-angle 0.72 mrad, objective aperture 19 nm", focus spread 9 nm,

Scherzer defocus 52.08 nm.

39

CHAPTER III

STRUCTURAL STUDIES OF POLYCRYSTALLINE Cu/Py GMR SPIN-VALVES

3.1 INTRODUCTION

The first set of the polycrystalline exchange-biased spin-valves examined in the
present study had the form of FegoMnso/Py/Cu/Py. Six Cu/Py spin-valves sputtered at
different temperatures have been studied. Three of these spin-valves are in the form of
[Nb250//Culo/FeMng/Py3/Cu2o/Py3//Nbx] with thin (3 nm) Py layers, while the others are of
form [Nb250//Cu1o/FeMng/Py24-30/Cu20/Py24-30//Nbx] (where x is 10 or 250 nm) with thick
(24 nm and 30 nm) Py layers. The top and. bottom Nb layers serve as superconducting
contacts for perpendicular resistance measurements [64]. All of the spin-valves were
produced by dc. magnetron sputtering onto p-type (100) Si substrates with natural thin
oxide coatings. The sputtering took place using a chilled substrate stage. With this stage,
the sample temperature tends to increase a few degrees during the sputtering. The initial
temperatures were approximately 18°C, -33°C, and -56°C for the three spin-valves with
thin Py layers and 22°C, 65°C, and -65°C for those with thick Py layers.

In this study, CTEM and HRTEM have been performed on cross-sections of the
six Cu/Py spin-valves. CTEM allows characterization of microstructural features such as
crystallographic epitaxy and texture, layer quality, and interfacial roughness. HRTEM
with associated quantitative image analysis techniques including image simulations and
fast Fourier transform techniques, allows extraction of structural information at the

atomic scale [72,73].

40

3.2 CTEM OBSERVATIONS
3.2.1 Spin Valves with Thick Py Layers

CTEM was used to characterize the overall structure of the spin valve multilayers.
Figure 3.1 shows a cross-sectional brightfield/darkfield image pair of the spin valve with
30 nm Py layers sputtered at 65°C. In both images, the Si substrate, the top and bottom
Nb contacts, and the spin valve (SV) multilayer can be distinguished. The Nb contacts
and the spin valve both appear to be continuous, polycrystalline, and are uniform in
thickness. Unfortunately, because of the low mass and diffraction contrasts between
FeMn (Z = 25.5), Py (Z = 27.7), and Cu (Z = 29), the individual layers within the spin
valve are not readily resolved.

The dark field image (figure 3.1 (b)), which was taken using an arbitrary
diffraction direction, helps to establish some structural characteristics. Polycrystalline
grains have a columnar structure with a wide range in crystal size perpendicular to the
growth direction, with observable widths S 90 nm. The individual grains generally have
grown in columns, starting in the bottom Nb contact. In many cases, epitaxy within these
grains has continued into, and sometimes through, the spin valves. In other cases, abrupt
changes in image intensity (due to changes in diffraction contrast) at the Nb/SV interfaces
indicate significant changes in orientation and loss of strict epitaxy during the growth
process. In most areas, the grains within the spin valve appear epitaxial.

The overall structures of the spin valves with 30 nm Py layers sputtered at -65°C
and with 24 nm Py layers sputtered at 22°C as shown in figure 3.2 and 3.3 also show
excellent contrast between both polycrystalline Nb contacts and the spin valves, and weak

contrast between the individual spin valve layers. The morphology is again columnar.

41

I irection

(1))

4 rowt-h
/’ Direct-ion
y

 

Figure 3.1 (a) Bright field and (b) dark field cross-sectional CTEM images of the
[Nb250//(Cu/FeMn/Py/Cu/Py)//Nb25o]nm spin valve with 30 nm Py layers sputtered at
65°C. The darkfield image was taken using an arbitrary diffraction direction.

42

Growth

 

Figure 3.2 A cross-sectional brightfield CTEM image of the
[Nb250//(Cu/FeMn/Py/Cu/Py)//Nbl0],,m spin valve with 30 nm Py layers sputtered at
-65°C.

43

 

- 3‘“ 1 .
m sfi ' ‘wu‘a "

Figure 3.3 A cross-sectional brightfield CTEM image of the
[Nb250//(Cu/FeMn/Py/Cu/Py)//Nblo]nm spin valve with 24 nm Py layers sputtered at 22°C.

44

3.2.2 Spin Valves with Thin Py Layers

Figure 3.4 shows a cross-sectional CTEM micrograph of the spin valve with 3 nm
Py layers sputtered at -56°C. The overall structure is similar to that in the spin valves
with thick Py layers. The Nb contacts and the spin valve appear continuous and are
uniform in thickness. The grains are generally columnar with a significant degree of
epitaxy, with in-plane grain diameters similar to those in the thick Py spin valves (less
than 80 nm). Again, due to the degree of epitaxy and the low mass contrast, it is difficult
to distinguish the individual spin valve layers although some evidence of layering can be
seen, most likely due to the Fresnel effects. Similar overall structures were also observed
in the spin valves with 3 nm Py layers sputtered at -33°C and 18°C as shown in figure 3.5

and figure 3.6.

3.3 HRTEM AND FFT ANALYSIS
3.3.1 Spin Valves with Thin Py Layers

A cross-sectional HRTEM image of the (Cu1o/FeMng/PngCu2o/Py3)nm spin valve
with 3 nm Py layers sputtered at -56°C is shown in figure 3.7. On this scale, it is apparent
that the interfaces between the Nb contacts and the spin valve are wavy, displaying some
interfacial roughness. The first Cu layer and FeMn layer are now weakly visible. The
other individual layers are not readily distinguishable due to weak mass contrast.
Consequently, only the nominal locations of the individual spin valve layers are indicated.
Both the Nb contacts and the spin valve appear polycrystalline with columnar grains,
consistent with the CTEM observations. While most of the grains display lattice fringes

running perpendicular to the growth direction, only a fraction of the cross-section is

45

; £2; ,_ I .. v. .: v a
50nm 1"” " V “’13-

lath. W”

 

Figure 3.4 A cross-sectional brightfield CTEM image of the
[Nb250//Cu/FeMn/Py/Cu/Py/be10]nm spin valve with 3 nm Py layers sputtered at -56°C.

46

     

Growth
Direction

Figure 3.5 A cross-sectional brightfield CTEM image of the
[Nb250//Cu/FeMn/Py/Cu/Py//Nb10],,m spin valve with 3 nm Py layers sputtered at -33°C.

47

Growth
Direction

 

 

Figure 3.6 A cross-sectional brightfield CTEM image of the
[Nb250//Cu/FeMn/Py/Cu/Py//Nbno]nm spin valve with 3 nm Py layers sputtered at 18°C.

48

1' \. 1
5 mm
<::>

 

Figure 3.7 A cross-sectional HRTEM image of the [Nb//Cu1o/FeMng/Py3/Cu20/Py3//Nb]nm
spin valve with 3 nm Py layers sputtered at -56°C with nominal layer positions indicated.
The first Cu layer and FeMn layer are weakly visible, while the other individual layers
are difficult to distinguish due to the small differences in the average atomic numbers of
the multilayer components. (SF — stacking fault)

49

oriented to zones favorable for HRTEM structural imaging. In addition to lattice image
contrast, significant numbers of Moire’ fringes appear on the image due to overlapping of
grains in the cross-sectional TEM sample. Numerous dislocations, twins and stacking
faults can also be seen. An HRTEM micrograph of a typical stacking fault in the second
Cu layer is shown in figure 3.8. It is clear that there are no extra planes inserted parallel
to the defect plane, indicating that the defect was not created from the condensation of
vacancies, but instead formed directly during growth or as a consequence of dislocation
slip. The Burgers circuit drawn in figure 3.8 reveals a net displacement of a/4[I2I]
perpendicular to the electron beam direction, which is consistent with a perfect lattice
translation of a/2[T10] in an fcc lattice, with a component of a/4[T01] in the beam
direction.

To more fully assess the structures seen in the HRTEM images, fast Fourier
transform (FFT) diffractograms were taken from selected areas within the nominal layer
locations. These diffractograms allowed accurate analysis of the lattice parameters and
interplanar angles of the different spin valve layers. An example of this type of analysis
is shown in figures 3.9 and 3.10 for the spin valve with 3 nm Py layers sputtered at -56°C.
To calibrate the FFTs, the Nb contacts have been used as internal standards. Figure 3.9
shows selected areas from the Nb contacts and their corresponding F FT diffractograms,
which effectively give the diffraction patterns from these areas. The FFT patterns of the
top and bottom Nb contacts show that these grains are imaged in <111> orientations,
characterized by the 60° interplanar angles. The characteristic diffraction spots are {110}
type (the first reflections allowed for bcc materials and the first diffraction vectors which

lie in the {111} zone). In both the top and bottom Nb contacts, the growth directions are

50

V, "

:55'0
fr.

. biéa/nizalil
I. &¢“‘=—o
’1‘ o O n a ('7.

”Os. 1' :j

 

”O : 'G'wt
Om.
'94I::
Om.

Figure 3.8 A HRTEM micrograph showing a stacking fault lying in the second Cu layer
of the spin valve with 3 nm Py layers sputtered at -56°C. The Burgers circuit results a net
displacement of a/4[ i 2 i ].

 

 

51

0‘ '
.......|.§O
,0. _..

Growth Direction—é

 

 

Figure 3.9 HRTEM images (left) and corresponding FFT patterns (right) of the top
(upper) and bottom Nb contacts of the Cu/Py spin valve with 3 nm Py layers sputtered at
-56°C. These patterns were used as a calibration for the patterns displayed in figure 3.10.

52

close to <110>, indicating the growth and subsequent epitaxy on the {110} (closest
packed) bcc planes. The equilibrium {110} interplanar spacing for bulk Nb is 0.234 nm
(aNb=0.331nm). From this value the spacing of the {110} spots has been used to calibrate
an effective camera length for the subsequent FFTs.

The layers within a large number of columnar grains from the layers of the spin
valve grown at -56°C with 3 nm Py layers have been analyzed, examples of which are
shown in figure 3.10. Beginning with the first Cu layer, the measured angles are 55°, 55°,
and 70°, indicating that the HRTEM image corresponds with a <110> fcc zone. The
interplanar spacing of the {111} planes is measured to be 0.21 nm, which corresponds to
an fcc lattice parameter of 0.359 nm. This is very close to the bulk value of 0.361 nm for
fee Cu. A consistent analysis is found for the second Cu layer. In both cases, the growth
direction is near <111>. Using the same approach, the FeMn and the Py layers were also
found to display fcc and L12 structures respectively, with <111> growth directions (the
superlattice reflections from the Py L12 structure are extremely weak). This behavior
suggests that the closest packed {111} planes in the fcc Cu, FeMn and L12 Py layers are
epitaxial with the closest packed {110} planes in the bee Nb contacts. The {111}
interplanar spacing are measured to be 0.210 nm for FeMn and 0.204 nm for Py, which
correspond to lattice parameters of 0.364 nm (armn = 0.363 nm) and 0.353 nm (apy =
0.355 nm).

Similar imaging and analysis was carried out on the spin valve with 3 nm Py
layers sputtered at -33°C. A cross-sectional HRTEM image of this material (figure 3.1 1)
again shows that the interfaces between the spin-valve and Nb contacts are wavy, and that

the individual component layers within the spin-valve cannot be clearly identified. Once

53

Growth Direction ——-—>
' . .. W ' I‘. ?..

 

Figure 3.10 HRTEM images and FFT patterns of various layers of the Cu/Py spin valve
with 3 nm Py layers sputtered at -56°C. All of the component layers display their
equilibrium fcc (Cu, FeMn) or L12 (Py) <1 10> zone patterns.

54

 

Figure 3.11 Cross-sectional HRTEM image of the Cu/Py spin valve with 3 nm Py layers
sputtered at -33°C. Some of the columnar grains of the Cu, FeMn and Py layers display
<111> symmetry characterized by 60° interplanar angles (inset column).

55

again full structural images cannot be resolved in all of the grains, but lattice fringes with
consistent spacing can be seen running perpendicular to the growth direction in the
bottom Nb contact, the spin valve, and the top Nb contact. This suggests epitaxial growth
throughout the structure. Some defects can also be seen in the images, such as
dislocations, twins and stacking faults.

In grains oriented favorably for structural imaging, some columnar grains are
found to display interplanar angles of 55°and 70° (indicative of <110> cubic zones)
within the spin valve, while other columns display interplanar angles of 60° (indicative of
<111> cubic zones). While the observation of the <110> symmetry is expected, and
consistent with the expectation of fee and L12 structures for the Cu, FeMn and Py layers,
the observation of the structural images displaying <111> cubic symmetry (as shown in
inset column of figure 3.11) is very surprising. All of the spin valve components have the
fee or the fcc derivative L12 structure in their bulk states. However, due to instrument
resolution limitations, it should be difficult, if not impossible, to resolve the atomic
structure of the <111> orientation in these materials. The JEOL-4000EX used in this
study has a Scherzer resolution of 0.17 nm with an information limit of approximately
0.14 nm; obtaining images at resolutions near these limits can be very difficult. Table 3.1
shows the projected atom column spacing (the limiting spacing to be resolved) of the
materials examined in this study for the <111> and <110> zone axes. The table shows
these spacings, assuming the equilibrium structures, along with the spacing expected if
they are assumed to take up non-equilibrium bcc structures (the non-equilibrium spacings
have been calculated based on a simple hard sphere atomic modeling approach to

determine the atomic radii). This table illustrates that for the fcc structures, the projected

56

Table 3.] Projected atomic spacings for the <111> and <1 10> zone axes for

layer components in fee and bcc forms.

 

Layer

l'cc

bcc

Minimum Column to Column Spacing (11m)

 

 

 

 

 

 

Components a0 (nm) a, (mu) <11 1>fcc <110>fcc <1 11>bcc <110>bcc
FeMn 0.363 0.296 0.148 0.222 0.242 0.148
Py 0.355 0.290 0.145 0.217 0.237 0.145
Cu 0.36] 0.295 0.147 0.221 0.241 0.148
Nb 0.331 0.270 0.166

 

 

 

 

 

 

 

57

 

point to point spacings of the atomic columns are well above the resolution limits of the
instrument for the <1 10> zones, but well below the Scherzer limit, and very close to the
information limit for the <111> zones. Conversely, if bcc structures are considered, the
<111> zones can be readily imaged, while the <110> zones will be very difficult to
resolve. Thus, the observation of <111> symmetry in the HRTEM images suggests
crystal structures other than fcc, while observations of <110> symmetry in the images
indicates non-bcc structures. This concept is further supported by the HRTEM image
simulations. Figure 3.12 shows series of HRTEM image simulations comparing the
image characteristics of the <111> and <110> zones for both fee and bcc structures for
the range of specimen thickness’ and image foci encountered experimentally. The
amplitudes of the <1 1 l> fee and <1 10> bcc structure images are so low as to render them
effectively unresolvable in practice.

Figure 3.13 shows FFT diffractograms taken from selected areas within the
nominal layer locations in this spin valve that display this <111> symmetry. Using the
Nb for calibration, the measured interplanar angles for the Cu, FeMn and Py layers again
indicate a <111> zone is being imaged. The interplanar spacing determined from the
FFTs (0.21 nm) appears to correspond with the spacing of {111} planes in fee Cu
(dm=0.208 nm), FeMn (dm=0.210 nm) and L12 Py (dm=0.205 nm). However, these
reflections cannot appear in <111> zone patterns. Thus, the observed structures cannot
be the equilibrium fcc or L12 structures. If it is assumed that these columns of the spin-
valve layers have instead taken on the bee structure, the measured interplanar spacings
(based on the hard sphere model) are consistent with those expected for {110} bcc

reflections. Furthermore, while it is not possible to resolve <111> zones in fcc Cu,

58

1.48
2.22
2.96

4.44
5.18
5.94
6.66

 

IIB
II
II
II
II
II
II
II
.-

 

Defocus (nml

Figure 3.12 Bloch wave HRTEM simulations of copper in the fee and bcc forms
described in Table 3.1. The images simulate those viewed on a JEOL-4000 EX (400
KeV, CS 1.1mm, beam convergence semi-angle 0.72 mrad, objective aperture 19 nm‘l,
focus spread 9 nm, Scherzer defocus, 52.08 um.) All images were printed using identical
brightness and contrast ranges. (A) fcc Cu viewed along a <110> zone axis with
specimen thickness and focus below Gaussian illustrated. (B) fcc Cu viewed along a

<111> axis using identical operating conditions. (continued on next page)

59

—
° 59 EEE-===flflfl
“8 =_E_E-===flfll
‘ ’7 EEE--==""fll-Ti
I

4 ‘3 IEl-I-EH-I
“2 I...

5.90

.U‘
(A)
d

0"

O
(D
O

A;
O

S” F- .“ 9° I" N _. .4 o

to <1 _. m w ‘ - - -

_. w c.) A m 32’ :j a; g
b

9”
o
o

01
O

     

Defocus (nm)

Figure 3.12 (C) The non-equilibrium bcc Cu viewed along a <110> axis. (D) The non-
equilibrium bcc Cu viewed along a <111> axis. Although the microscope should allow
the pass of fcc<m> and bCC<1|0> image information, in practice these images (especially
fcc<m>) are likely to be unresolved due to the very low amplitude (2 0.07 of the average
<1 10> axis in the simulated images) of the passed frequencies in this range. (Images and

analysis are provided by J .W. Heckman.)

60

O

U

Growth Direction é.)

 

Figure 3.13 HRTEM images and corresponding fast Fourier transform diffractograms of
various layers of the Cu/Py spin valve with 3 nm Py layers sputtered at -33°C showing
<1 1 l> zone patterns in the Cu, FeMn and Py layers.

61

FeMn, and Py, bcc forms of those materials can be resolved in <1 11> orientations (Table
3.1). Thus, this analysis is consistent with the Cu, FeMn and Py layers taking up non-
equilibrium bcc structures in this spin valve. Approximately six grain columns were
examined in studying this spin-valve. Of these, two were found to display the bee
structure, while the rest displayed the equilibrium fcc structure. When either bcc or fcc
structures were identified, the structure was maintained across the entire spin valve within
that particular grain column.

A cross-sectional HRTEM image and similar analysis of the spin valves with 3
nm Py layers sputtered at 18°C as shown in figure 3.14 and figure 3.15 also revealed
wavy interfaces between the polycrystalline Nb contacts and the spin valve, and weak
contrast between the individual layers. The FFT analysis indicated that both Nb contacts
and the spin valve appear in the equilibrium structures (bcc for Nb and fcc for spin-
valves) with the close packed {111} planes in the fcc Cu, FeMn and Py layers epitaxial
with the closest packed {110} planes in the bcc Nb. No bcc structures were found for this
spin-valve. But since only four grain columns could be characterized in each sample, it is
possible that non-equilibrium structures are present, but were not found in the small
number of observations.

3.3.2 Spin Valves with Thick Py Layers

In general, the spin valves with thicker Py layers all displayed similar structures,
including the existence of non-equilibrium structures in some columnar grains. A cross-
sectional HRTEM image of the spin valve with 30 nm Py layers sputtered at -65°C (fig.
3.16) again shows excellent contrast between the polycrystalline Nb contacts and the spin

valve, but weak contrast between the individual layers. A high density of planar defects

62

.‘ "uh”

V

. ‘1
H
,9.
i:

'5'?

2,-

r
J .

. . ...._-....M
- awn

:udngéi (1

w '
,u...
op.
: -

 

Figure 3.14 A cross-sectional HRTEM image of the

[Nb//Cu.o/FeMng/Py3/Cu2()/Py3//’Nb]nm spin valve with 3 nm Py layers sputtered at 18°C
with nominal layer positions indicated. The individual layers of the spin valve are
difficult to distinguish due to the small differences in the average atomic numbers of the

multilayer components.

63

 

Figure 3.15 HRTEM images and FF
with 3 nm Py layers sputtered at 18°C. All of the component layers display their
equilibrium bcc (Nb) <1 1 l> and fee (Cu, FeMn) or L12 (Py) <110> zone patterns.

64

‘V
‘3 ,
“1" C -:
§ 1:

1‘ a.

)I"

44.

v

«a

£15.“

:3

‘1 53'4“- ; '

 

Figure 3.16 A cross-sectional HRTEM image of the spin valve with 30 nm Py layers
sputtered at -65°C (left) and selected areas (indicated by arrows) from the FeMn, Cu and
Py layers (right). The nominal layer locations are given in the image.

65

such as stacking faults and twins are evident in some columnar grains (in fact, these
appear only in the fcc grains). A typical twin with a {ll 1} coherent twin boundary from
the second Py layer is shown in figure 3.17. As might be expected, some twin interfaces
are parallel with the growth plane. However, twins are also observed on other {111}
planes inclined to the growth direction. Characterization of the Nb contacts shows that,
as expected, the Nb has grown with the bee structure on the closest packed {1 10} planes
(<1 10> wire texture).

As noted, HRTEM imaging and FFT analysis revealed both equilibrium fcc and
non-equilibrium bcc structures in all three spin valves with thick Py layers. As was the
case in the thinner spin valves, entire grain columns displayed the same structure across
the spin valve. Figure 3.18 shows selected areas and their corresponding diffractograms
from the spin valve with 30 nm Py layers sputtered at 65°C. The resolvable zones of this
columnar grain in the Cu, FeMn, and Py layers are all characterized by 60° interplanar
angles, indicating <111> zone axes. As explained earlier, the atomic structure for Cu,
FeMn and Py cannot be readily resolved along <111> zones. Likewise, the interplanar
spacing of 0.21 nm is not consistent with allowed reflections in fcc and L12 <111> zone
patterns. Therefore, the spin valve appears to have some non-equilibrium bcc columnar
grains. These grains grow on the {110} closest packed planes throughout the columnar

grains. Likewise, in the spin valve with 30 nm Py layers sputtered at -65°C and with 24

nm Py layers sputtered at 22°C, resolvable <111> zones and corresponding bcc FFT
diffractograms are found in some of the columnar grains (figure 3.19). This indicates that

both of the spin valves can have the non-equilibrium bcc structures.

66

 

Figure 3.17 A {111} coherent twin boundary in the second Py layer in the Cu/Py spin
valve with 30 nm Py layers sputtered at -65°C.

67

Growth Direction ——)

 

Figure 3.18 HRTEM images and FFT diffractograms of various layers of the Cu/Py spin
valve with 30 nm Py layers sputtered at 65°C showing six-fold symmetry and lattice
spacings consistent with bcc structures.

68

”Growth Direction -——>

 

Figure 3.19 HRTEM images and FFT diffractograms of selected Cu layers within the
Cu/Py spin valves (a) with 30 nm Py layers sputtered at -65°C and (b) with 24 nm Py
layers sputtered at 22°C, indicating that both the spin valves can have the non-equilibrium
bcc structures characterized by 60° interplanar angles in <1 11> zone patterns.

69

3.4 DISCUSSION

In all of the spin valves examined, growth appears to have occurred on the closest
packed atomic planes; {110} planes in bee and {111} planes in fcc. Such growth is
expected since the close packed planes are typically characterized by low surface
energies. However, the finding of non-equilibrium bcc forms of the spin valve
components Cu, FeMn, and Py in some grain columns is somewhat surprising, and
requires further discussion. Initially, it might be suspected that these findings are a
consequence of chemical variations within the layers, which might lead to changes in
crystal structure. However, high-resolution energy dispersive x-ray spectroscopy studies
we have performed on similarly grown Ag/Py spin valves using a field emission gun
STEM show that the layers are well defined chemically, with little or no interfacial
mixing beyond the instrumental resolution of ~ 0.5 nm, and chemistries representative of
the intended values in the interiors of the layers, away from the boundaries [58]. As the
spin valves in the present study were all sputtered under similar conditions, differing only
in the substitution of Cu for Ag, it is unlikely that chemical variations within the layers
are responsible for the layers taking up non-equilibrium forms.

The HRTEM image simulations and FFT analysis clearly show that the Cu, FeMn,
and Py in the spin valve with 3 nm Py layers sputtered at -33°C, and in all of three spin
valves with thick Py layers, display some columnar grains that are not in their equilibrium
fcc forms. The supposition that non-equilibrium structures occur in some of the spin
valves follows from the identification of 60° interplanar angles in the HRTEM images
and FFT analysis characteristic of cubic <111> zone axes, which cannot be resolved in

the fcc structures. Furthermore, it should be recalled that these zone axes with three fold

70

symmetry (actually six in the FFTs due to :g reflections) are identified in the cross-
sectional orientation and the growth planes must lie close to perpendicular to this
orientation. However, if the materials were fee, the growth planes could not be {111}
planes, as the angles between {111} planes in cubic materials are 70.5° and 109.5°. This
is unlikely and inconsistent with the observed characteristic growth when fcc phases are
present.

While it is clear that the layers noted above are not in their equilibrium forms, the
suggestion that they have taken up the bcc crystal structure is thus far based only on the
fact that the analysis is consistent with an assumed bcc structure. Thus, the conclusion
that the layers are truly taking up bcc structures needs further justification. The key is
again based on the observation of the three-fold symmetry of the HRTEM images in these
layers. Such three fold symmetry can only be displayed by cubic (which display four 3
fold axes), hexagonal (truly 6-fold symmetry), and rhombohedral crystal structures.

However, if these layers were taking up a hexagonal structure, the growth
direction perpendicular to the observed 3-fold axes would be of the form {XXXO}.
These planes of the [0001] zone in general have low packing densities, with a maximum
packing factors of 0.556 on the {1120} planes and 0.481 on the {1010} planes (as
opposed to the maximum possible of 0.908 for {111} planes in fee and {0001} planes in
hcp). It is thermodynamically unlikely that growth would occur on such low-density
planes. Similarly, if the layers took up a rhombohedral structure, the growth planes
would have to be low-density irrational planes.

In contrast, if the non-equilibrium layers take up bcc structures, the growth can

occur on {110} planes that are high density (packing factor of 0.833) and perpendicular to

71

the observed 3-fold zone axes. This situation is consistent with the observations noted
above.

Epitaxial strains associated with the layer interfaces are the likely cause of the
layers taking up non-equilibrium structures. The interfacial characteristics of
{III}fcci{IIO}bcc interfaces have been studied extensively, beginning with the well
known Nishiyama-Wasserrnan [74,75] and Kurdjumov-Sachs [76] relationships which
characterize bcc:fcc epitaxy as {1 11}t‘cc//{11O}bcc;<101>fcc//<001>bcc and
{11lifcc//{110}bcc;<l10>fcc//<111>bcc respectively. This analysis was furthered by Hall,
Aaronson, and Kinsman [77], and Rigsbee and Aaronson [78] who found that a wide
range of rotations around the {111}»fccz{110}bcc normals could occur depending on the
ratio of fcc to bee lattice parameters. In all cases, only small regions of good matching at
the interface can be achieved, separated by interfacial misfit dislocations and structural
ledges. Consequently, the energies associated with the interfaces will be a function of the
ratio of fcc and bcc lattice parameters. In the present study, it would appear that the
fcc/bcc interfaces are close to the Kurdjumov-Sachs orientation as the <110> fcc zones
are imaged simultaneously with the <111> bcc zones in the same columns. However,
slight deviations from the actual zone axis [79] can be present and still result in HRTEM
structural images as in figure 3.20. Thus, in practice, it can only be concluded that the
observed interfaces are within ~ 4° of the Kurdjumov-Sachs orientation. As the
Nishiyama-Wasserrnan orientation is actually only 5.27° from the Kurdjumov-Sachs
orientation, it is impractical to distinguish these using HRTEM.

It is expected (and found) that both top and bottom Nb contacts have taken up

their bulk bcc forms. As a result, both the first Cu layer and second Py layer will have

72

 

a
a
a

Specimen tilt (degrees)

9
o

  
  

N
_A
r
l"
9

9°
01
i
l

b
0
01
O

60
Defocus (nm)

Figure 3.20 Bloch wave HRTEM simulations (using the same microscope conditions as
in figure 3.12) of the effects of specimen tilt on the resulting image for a representative
crystal viewed at various foci near Scherzer defocus. (A) fcc (: 6 nm thick) Cu viewed
along a <110> axis with the specimen tilted, progressively, parallel to the {111} growth
planes. (B) bcc (z 7 nm thick) Nb viewed along a <111> axis with the crystal tilted,
progressively, parallel to the {110} growth planes. Although there is a distortion in the
shape of the atom column images, the overall arrangement and lattice spacing remains
unambiguous through at least 3.5 degrees of tilt for Cu and 1.4 degrees for Nb. (Images
and analysis are provided by J .W. Heckman.)

73

misfit strains associated with their Nb interfaces if they take up their bulk fee (and L12)
crystal structures. These interfacial strains may be decreased if these layers instead take
up non-equilibrium bcc structures. However, the decrease in strain energy is offset by an
increase in the volume free energy associated with the non-equilibrium structure, and as
the layer thickness increases, the volume free energy becomes dominant. In the case of
Cu, it has been shown using full-potential muffin-tin orbital calculations [80] that the bcc
structure has an energy only about 0.01 eV/atom greater than the equilibrium fcc form.
Thus, the balance between bee and fcc Cu might easily be shifted by interfacial strain.
Mitchell, et al. [81] have used the critical thickness (IC for the loss of coherency, which is
the thickness at which the coherency of the interface begins to be disrupted by the
requirement of interfacial dislocations, as a criterion for the point at which the crystal
structure of a thin layer will revert to its equilibrium structure. This critical thickness has
been given as:
dc/b=(1/or€o)[ln(dC/b) + 0.5]

where b is the Burgers vector of the interfacial dislocations, so is the misfit strain, and or is
a geometric constant approximately equal to 471. For Nb/Cu multilayers, Mitchell et a1.
[80] have determined this thickness to be only a few atomic layers, depending on the
value of on. While this thickness was used as a general parameter for the point where the
breakdown of the coherent Nbbcc/CubCC would occur, it only considers the point at which
dislocations will begin to be incorporated into the interface and does not consider the
relative energies of the Nbbcc/CubCC and NbbCC/CUfcc interfaces. Thus, this parameter can
give only a general indication of the point at which the interface loses coherency, which

might lead to the transition from the non-equilibrium bcc to the equilibrium fcc structure

74

for the Cu. However, it does suggest that the thicknesses of the bcc Cu observed in the
present study are well beyond that at which coherency should exist.

While the observation of bee Cu may be surprising, it is not unprecedented. The
first observation of bee Cu was reported in 1973 by Goodman, Brenner and Low [82] for
an Fe—1.4at%Cu alloy where coherent Cu precipitates up to 5.0 nm in diameter were
identified in the bee Fe matrix using field-ion microscopy. Since then, a number of other
findings of fine-scale bcc Cu have been reported. Using reflection high energy electron
diffraction (RHEED) Heinrich e! a]. [83] identified (001) oriented bcc Cu epitaxially
grown on (001) bcc Fe. In Fe(001)/Cu(001)/Fe(001) trilayers, bcc Cu with thickness’
between 6 and 12 monolayers (~0.9-l.8 nm) were grown. In Fe/Cu multilayers (60
bilayers) with 1.5 nm thick Fe layers Pizzini et al. [44] found Cu in the bee form in layers
up to 1.3 nm thick using x-ray adsorption spectroscopy (XAS). More recently, HRTEM
has been used to characterize bcc Cu in Cu/Nb multilayers [81, 84]. In these studies, the
bcc Cu was found to grow psudomorphically on the bcc Nb for Cu thickness up to 1.2
nm. For thicker layers, the Cu took up an fcc structure with a Kurdjumov-Sachs
relationship with the Nb layers.

In contrast to the existence of bee Cu at very fine scales that has been established,
the scale observed in the present study is much larger. The two Cu bcc layers in the spin-
valves are 10 and 20 nm thick, approximately an order of magnitude greater than
previously observed. When the bee F eMn and Py layers are considered in the spin valves,
the total thickness of all of the non-equilibrium grain columns reaches approximately 44
nm in the spin-valve with 3 nm Py layers and 98 nm in the spin valves with thick Py

layers. At present, it is not clear why the bcc structure exists to such large thicknesses in

75

these spin-valves.

Low target temperatures during sputtering might be suspected to contribute to the
formation of non-equilibrium phases. However, these temperature variations do not seem
to have influenced the formation of the bee phases in the present study. Sputtering on
targets with temperatures ranging from -65°C to +22°C have all resulted in some of the
grain columns taking up non-equilibrium bcc structures. Mitchell et al. [81,84] suggests
that the transformation from bcc to fcc in thin layers of Cu occurs martensitically, not
diffusely. A low sputtering temperature would not be expected to restrict this
transformation. It is worth noting that in the Nb/Cu multilayers [81,84] the bee Cu was
found to have a lattice parameter of 0.328i0.007 nm, which is very close to the bulk
value of bcc Nb of 0.331 nm. In the present study, the value of the lattice parameter of
the bee Cu measured from the FFTs is approximately 0.29 nm, close to calculated values
of 0.287 and 0.296 nm [85,86]. This suggests that significantly more relaxation of the
bee Cu has taken place in the present study than in the Nb/Cu multilayers examined by
Mitchell etal. [81,84].

Further, it is interesting to note that the bee structures investigated here display a
significant degree of stability. Neither the energy imparted by ion milling during thin foil
preparation nor electron beam radiation during CTEM or HRTEM examination has

resulted in rearrangement to the equilibrium fee structure.

3.5 SUMMARY
Structures of Cu/Py GMR spin valves with different Py thickness sputtered at

different temperatures have been characterized using HRTEM and CTEM complemented

76

with HRTEM image simulations and FFT analysis techniques. All of the spin valves
displayed polycrystalline structures with columnar grains. The grains have grown on the
close packed planes ({111} planes in the bee Nb contacts and {1 10} planes in the fcc Cu,
FeMn, and Py spin-valve layers) taking up a near Kurdjumov-Sachs
{III}fCC//{IIO}bCC;<IIO>fCC//<III>bCC orientation relationship. The FFT analysis and
HRTEM image simulations indicate that in some of the columnar grains, the Cu, FeMn,
and Py take up non-equilibrium bcc structures, regardless of different layer thicknesses
and sputtering temperatures. In these cases, the bcc Cu, FeMn, and Py layers grew on

{110} planes and were epitaxial with the Nb contacts.

77

CHAPTER IV

STRUCTURAL STUDIES OF POLYCRYSTALLINE Ag/Py GMR SPIN-VALVES

4.1 INTRODUCTION

In this chapter, results of the characterization of polycrystalline GMR spin-valves in
the form of Ag/Py/Ag/Py/FeMn, sputtered either on to a Nb superconducting contact or
directly on to a Si substrate, are presented.

The first sample of [Ag]o/Py_~,/Ag20/Py3/FeMng]nm (with 3 nm Py layers) was
produced by sequential dc magnetron sputtering onto a p-type (001) Si substrate with a
duplex design as shown in figure 2.1. This sample was examined between the crossed Nb
contacts and also on the adjacent Si substrate to investigate the effect of substrate material
on the subsequent spin-valve layer growth. CTEM and HRTEM observations found that
the first Ag (buffer) layer of the spin-valve grown directly on the Si substrate displayed an
epitaxial single crystal, in contrast to the polycrystalline form seen between the Nb
contacts. To further characterize this observed single crystal and to determine if this phase
was unique to this particular sample or if the epitaxial phase is induced by the substrate,
additional HRTEM and CTEM studies were performed on this first sample and other
subsequently fabricated specimens. In addition, HREDS was used to qualitatively
determine the chemical composition of the first Ag layer and the other individual layers of
this spin-valve. The EDS spectra collected from the first Ag layer indicated the presence of
a significant amount of indium, which may have been inadvertently sputtered on the Si

substrate during the bottom Nb contact layer being deposited.

78

Three additional samples were fabricated under the same conditions as this first
sample, but sputtered directly on an approximately 1 cm2 area of Si substrate in the form
[Ag]o/Py3/Ag20/Py3/FeMn3//Nb20]nm and [Ag]0/Py,,./Ag20/Py,//Nb10]nm with x = 6 nm or 24
nm. Further TEM investigations were performed on these structures to determine whether

this phenomenon was reproducible or unique to the initial sample.

4.2 CTEM OBSERVATIONS
4.2.1 Spin-Valve Grown on the Nb Contact

Figure 4.1 shows a cross-sectional brightfield CTEM image of the
(Aglo/Py3/Ag20/Py3/FeMng)nm spin-valve, with 3 nm Py layers, grown on a Nb
superconducting contact. The figure clearly illustrates the Si substrate, both Nb contacts
and the spin-valve. Note that, unlike the Cu/Py spin-valves, the individual layers within the
spin-valve are resolvable due to the large differences in the average atomic numbers of the
layer components (ZAg = 47, Zpy = 27.7, and ZycMn = 25.5). Both the Nb contacts and the
spin-valve were observed to be polycrystalline with columnar grains, oriented with the long
axis in the growth direction. The grains had widths ranging up to 50 nm normal to the
growth direction. This structure is consistent with that observed for Cu/Py spin-valves

sputtered on Nb contacts [63].

4.2.2 Spin-Valves Grown Directly on the Si Substrate
Figure 4.2 shows a cross-sectional CTEM image taken from an area of the initial
(Ag1o/Py3/Ag2o/Py3/FeMng)nm spin-valve with 3 nm Py layers grown directly on the Si

substrate. As can be noted from the image, a significantly different morphology was

79

 

Figure 4. l A cross-sectional brightfield CTEM image of the
[‘Nb25o//(Ag/Py/Ag/Py/FeMn)//Nb250]nm spin-valve with 3 nm Py layers grown on Nb. Note
that the individual layers within the spin-valve can be distinguished.

80

   
 

Growth
Direction

 

Figure 4.2 A cross-sectional brightfield C TEM image of the Ag/Py/Ag/Py/FeMn spin-valve
with 3 nm Py layers grown directly on Si substrate showing the first Ag layer grew as a
single crystal.

81

observed. The first Ag layer appeared as a continuous single crystal throughout the entire
viewable area of the TEM sample, with the rest of the layers being polycrystalline and
morphologically similar to the spin-valve grown on the Nb contact.

However, images of three similar spin-valve structures subsequently prepared on Si
substrates do not show any layers with a similar single-grain growth pattern. An overall
structure of another (Aglo/Py3/Ag2o/Py3/FeMng)nm spin-valve (with 3 nm Py layers), grown
directly on Si substrate, under the same sputtering conditions, was characterized using
CTEM and is shown in figure 4.3. This spin-valve was polycrystalline throughout and
morphologically similar to that seen in spin-valve layers grown between the Nb contacts in
the first specimen. The individual layers within the spin-valves are, again, clearly visible
(due to the contrast mechanisms described above) and appear continuous with uniform
thicknesses. Additionally, a similar polycrystalline morphology was also observed in two
other spin-valves of (Ag:o/Py/Ag20/Py/FeMng)nm with 6 nm and 24 nm Py layers grown
directly on Si substrates as shown in figure 4.4 (a) and (b). The epitaxial single crystal

morphology was not found in either of these spin-valve samples.

4.3 HRTEM AND FFT ANALYSIS
4.3.1 Spin-Valve Grown on the Nb Contact

A cross-sectional HRTEM image of the (AglO/Py3/Ag20/Py3/FeMng)nm spin-valve
with 3 nm Py layers grown on the Nb contact is shown in figure 4.5. The sample was
imaged with the Si substrate tilted to the <110> zone axis, which puts the (001) substrate
plane and the spin-valve edge-on. The individual spin-valve layers can be distinguished

clearly, even in sample areas thin enough to yield good HRTEM images, due to the above

82

     
 

Growth
Direction

Figure 4.3 A cross-sectional brightfield CTEM image of a second Ag/Py/Ag/Py/FeMn/Nb
multilayer with 3 nm Py layers grown directly on the Si substrate showing polycrystalline
morphology with columnar grains for all the layers.

Growth
Direction

 

 
     

Growth
Direction

Figure 4.4 Cross-sectional bright field CTEM images of the Ag/Py/Ag/Py/Nb multilayers
with (a) 6 nm Py layers and (b) 24 nm Py layers, grown directly on Si substrate showing
polycrystalline morphology for all the layers.

84

G
H

e ,
i
:3"

MI

@991

1101

 

Figure 4.5 A cross-sectional HRTEM image of the first (Ag;o/Py3/Ag2n/Pyg/FeMng)nm spin-
valve, from an area between the Nb contacts. The individual layers are clearly distinguished
due to the large differences in the average atomic numbers of the multilayer components.

85

mentioned range in Z of the layer materials. The observed period of the
[Agm/Py3/Ag2o/Py3/FeMng]nm spin-valve layers in the image is consistent with that given by
the nominal sputtered thickness (indicated in figure 4.5). As expected, only some of the
grain columns are properly oriented for structural lattice imaging, because both the Nb
contacts and the spin-valve exhibit a columnar polycrystalline morphology. However,
continuous lattice fringes are observed, running perpendicular to the growth direction, in
almost all of these columnar grains. These fringes run from the substrate to the top of the
structure. This lattice fringe pattern is evidence of growth on close packed planes and of a
highly textured structure. In grain columns suitably oriented for structural imaging, the
lattice structure maintains an epitaxial morphology across the spin-valve layer interfaces.
In addition to the lattice image contrast, numerous Moiré fringes appear in the image, due
to overlapping of different grain columns in the plane of observation of the cross-sectional
TEM samples.

It is difficult to obtain electron diffraction patterns directly from areas within grains
in the individual spin-valve layers due to their small width normal to the beam. Therefore,
fast Fourier transform (F F T) image analysis has been used to complement the interpretation
of the HRTEM images. The fast Fourier transforms, taking the form of optical
diffractograms of selected areas of the experimental HRTEM images, reveal equivalent
plane spacing and angular information to electron diffraction patterns from those areas of
the specimen. In this study, Nb or Si reflections were used as internal calibration standards
to determine the effective camera lengths for interpreting the FFT diffractograms [63].
Based on these calibrations, the interplanar spacings and angles of the individual spin-valve

layers were determined.

86

Portions of HRTEM images and FFT diffractograms for the
(Ag1o/Py3/Ag2o/Py3/FeMng)nm spin-valve with 3 nm Py layers grown on Nb are shown in
figure 4.6. In this case, the diffractogram of the Nb contact was first calibrated. The
bottom Nb is characterized by 60° interplanar angles, indicating a <111> zone axis of bee
Nb. The {110} interplanar spacings from this layer were used as a standard to calculate an
effective camera length for measuring subsequent diffractograms. In this bottom Nb
contact the growth direction is close to <110>, indicating the growth has occurred on the
closest packed bcc {110} planes. Using this Nb calibration, selected areas from the
individual spin-valve layers have been analyzed as follows. Beginning with the first Ag
layer, the resolvable zone is characterized by 55° and 70° interplanar angles. This indicates
that this layer is imaged along a <011> zone axis and is found to display the fee structure.
The first reflections in the FFT diffractogram correspond to a real space interplanar spacing
of 0.24 nm, which corresponds to the {111} planes of an fcc cell with a lattice parameter of
0.42 nm. This is very close to the bulk value of 0.41 nm for fee Ag. A similar analysis
applied to the second Ag layer also reveals it to be fee. In both cases, the growth direction
is near <111>, suggesting the close packed planes in the fcc Ag layers are epitaxial with the
{01 l} closest packed planes in the bee Nb. Subsequent analysis of the Py and FeMn layers
also confirms these layers to be fee structure1 with <111> growth directions. The {111}
interplanar spacings were measured to be 0.20 nm for Py and 0.21 for FeMn, which

correspond to fcc lattice parameters of 0.35 nm (apy = 0.355nm in the Py bulk state) and

 

' Bulk PermalloyTM (Py) displays the fcc derivative L12 structure. However, because of the fairly small
differences in the atomic scattering factors and because the Py used in this study is significantly off-
stoichrometric, the superlattice reflections are very weak and do not appear on the F FT diffractograms. For
the same reasons, the L12 ordering is not evident in the HRTEM images. Thus, for the purposes of this work,
Py will be treated as fcc.

87

 

Figure 4.6 HRTEM images and FFT diffractograms of various layers of the
[Nb250//(Ag1o/Py3/Ag2o/Py3/FeMng)//Nb250],,m spin-valve grown on Nb contact. All of the
component layers display their equilibrium fcc (Ag, FeMn) or L12 (Py) <1 10> zone patterns.

0.36 nm (at-2Mn = 0.363 nm in the FeMn bulk state) respectively.

4.3.2 Spin-Valve Grown Directly on the Si Substrate

A cross-sectional HRTEM image of the first (AglO/Py3/Ag20/Py3/FeMng)nm spin-
valve (with 3 nm Py layers) from an area grown directly on the Si substrate is shown in
figure 4.7. It is evident that the crystal structure of the first Ag layer is quite different than
that seen in those grown on the Nb contact (cf figure 4.5). A thin amorphous region
(probably SiO2) is seen between the first Ag layer and the Si substrate. In this case the first
Ag layer displays a single orientation throughout the entire region pictured (and throughout
the entire thin area of the TEM sample). Epitaxial growth of (1 l 1) Ag on Si (001) has been
reported and a mechanism for accommodating the lattice mismatch has been proposed,
although the published single crystal domains were considerably smaller than those
presented here [87]. However, this single, epitaxial layer also displays a striking contrast
variation in the planes running parallel to the growth direction with a contrast period of
every third atomic plane. All of the Ag non-magnetic layers in the present study had a
nominal 10 nm thickness, as determined with a quartz crystal thickness monitor. However,
this epitaxial layer appears approximately 45% thicker than both the nominal thickness
sputtered and the Ag layer thickness found between the spin-valve and Nb contacts (figure
4.5). The rest of the spin-valve grown on this single crystal layer is polycrystalline and
appears similar in morphology to the portion grown on the Nb contact.

The enlarged areas of the HRTEM images of both Ag layers and Si substrate, along
with their corresponding FFT diffractograms are shown in figure 4.8. In this case (without

the Nb contacts), the reflections from the Si substrate were used as internal calibration

89

 

\\\§ 1
\ \\

a _\
a, \\§\

 

Figure 4.7 A cross-sectional HRTEM image of the (Aglo/Pyg/Ag20/Py3/FeMng)nm spin-valve
grown directly on Si. The first Ag layer displays a single orientation throughout the entire
region of view (and throughout the entire thin area of the TEM sample).

90

 

 

Cubic

 

Figure 4.8 HRTEM images and FFT diffractograms of Si and Ag layers of the
(Ag;o/Pyg/Ag2o/Py3/FeMng)nm spin-valve grown directly on Si. The F FT from first Ag layer
is inconsistent with an fcc equilibrium structure.

91

standards to determine the effective camera lengths for interpreting the F FT diffractograms
of the Ag layer. Consistent with the contrast modulation noted from the HRTEM images,
additional reflections at 1/3 and 2/3 of the primary reflection spacings are apparent in the
directions perpendicular to the growth direction in the first Ag layer. The projected zone is
characterized by 90° interplanar angles. If the additional 1/3 and 2/3 reflection spots were
disregarded, the interplanar spacings of all four of the primary reflections would equal 0.23
nm. This spacing is consistent with the {111} interplanar spacing of 0.235 nm in bulk fcc
Ag. However, the {111} planes cannot lie at 90° to one another. Thus, it appears this first
layer is not an equilibrium fee structure. Previous HRTEM studies of Cu/Py spin-valves
showed that in some of the columnar grains, the Cu, Py and FeMn layers had been found
taking up a non-equilibrium bcc structure [88]. In present situation, if the Ag had assumed
a similar non-equilibrium bcc structure, the measured interplanar spacings (based on a hard
sphere atomic radii calculated from the equilibrium fee structure) would be consistent with
those expected for {110} reflections (aAgbcc = 0.334 nm thus (11 10b“ (Ag) = 0.236 nm). These
can easily be resolved in <100> orientation. However, packing density difference between
the fcc and bcc structures (0.68 versus 0.74), alone, does not nearly account for the
observed difference in the thickness of the first layer.

For comparison, a cross-sectional HRTEM image of the
(Aglo/Py3/Ag20/Py3/FeMng)nm spin-valve with 24 nm Py layers grown directly on the Si
substrate is shown in figure 4.9. Overall, the spin-valve has a morphology similar to the
one grown on the Nb contact. The Si substrate, the spin-valve and top Nb layer can be
distinguished clearly and the latter two appear to be polycrystalline, which is consistent

with observed by CTEM study. The individual spin-valve layers are readily apparent and

 

Figure 4.9 A cross-sectional HRTEM image of the (Ag;o/Py24/Ag20/Py24/Nb10)nm multilayer
grown directly on Si. The interfaces between the layers are sharp and flat, indicating less
interfacial roughness and interdiffusion between the spin-valve components compared with
the one grown on Nb.

93

their widths consistent with the nominal sputtered thicknesses. The selected areas of the
HRTEM images and FFT diffractograms of the individual components are shown in figure
4.10. In this case (without the Nb contacts), again, the reflections from the Si substrate
were used as internal calibration standards to determine the effective camera lengths for
interpreting the FFT diffractograms of the rest spin-valve layers. As expected, the Si
substrate is a single crystal grown on (100) plane. Using Si calibration, selected areas from
the individual spin-valve layers have been analyzed as follows. The resolvable zone of the
first Ag layer is characterized by 55° and 70° interplanar angles, indicating that this layer is
imaged along a <011> zone axis and is found to display the fee structure. The first
reflections in the FFT diffractogram correspond to a real space interplanar spacing of 0.24
nm, which corresponds to the {111} planes of an fcc cell with a lattice parameter of 0.42
nm, which is very close to the bulk value of 0.41 nm for fcc Ag. A similar analysis applied
to the second Ag layer also reveals it to be fcc. In both cases, the growth direction is near
<1 11>, suggesting the close packed planes in the fcc Ag layers are epitaxial with the {100}
substrate planes in the cubic Si. Subsequent analysis of both Py layers also confirms these
layers to be fcc structure with <111> growth directions. The {111} interplanar spacings
were measured to be 0.20 nm, which correspond to fcc lattice parameter of 0.35 nm
(apy=0.355nm in the Py bulk state). The top Nb is characterized by 60° interplanar angles,
indicating a <1 11> zone axis of bee Nb. The {110} interplanar spacings were measured to
be 0.23, which correspond to bcc lattice parameter of 0.33 nm (aNb = 0.331 nm in the Nb
bulk state). The growth direction of this top Nb contact is close to <110>, indicating the
growth has occurred on the closest packed bcc {110} planes. Therefore the FFT analysis

indicated that the Ag, Py and Nb layers display their equilibrium fcc and bcc

94

 

Figure 4.10 HRTEM images and FFT diffractograms of various layers of the
[(Aglo/Py24/Ag2o/Py24)//Nb10]nm spin-valve grown directly on Si. All of the component
layers display their equilibrium fcc (Ag) or L12 (Py) <110> and bcc (Nb) <111> zone
patterns.

95

structures with <111> growth directions for fcc and <110> growth directions for bcc

respectively.

4.4 HREDS RESULTSZ (By J.W. Heckman)

After analysis by HRTEM, the first spin-valve sample (grown directly on the Si)
was analyzed in the STEM. A brightfield STEM image of the area of the spin-valve
analyzed by HREDS is presented in figure 4.11. The lighter areas in the micrograph
(labeled) are due to an ablation of material and damage at each 0f the analysis points. X-
ray spectra collected from the four points spanning the spin-valve are shown in figure 4.12.
Figure 4.12 (A) is the spectrum from point A in figure 4.11, which is a reference spectrum
that shows mainly a Si K family with a trace of O K and C K radiation. The spectrum from
point B (figure 4.12 (B)) was captured within the single crystal first Ag layer. The edge of
the beam ablated region for this analysis point is approximately 3 nm from the Ag layer /Py
layer interface. The spectrum consists mainly of a series of L family peaks in the range of
2.5 to 4.5 keV and shows just a trace of Ni K“ and Cu KCl radiation. While this is the
general range of values expected for the Ag L family, the relative intensities do not match
those expected. If the contribution of the Ag L family is manually stripped from the
spectrum (figure 4.12 (C)), the resulting series of peaks matches the distribution and
intensities of Indium L peaks. Figure 4.12 (D) shows the spectrum obtained by positioning
the probe directly on the first Py layer. The expected Fe and Ni peaks, with their intensities
roughly consistent with the Py stoicheometry, dominate the spectrum. There is a

persistence of the L family of lines from the Ag-In region, which is likely due to beam

 

Z In order to have a better understanding of the structure of Ag/Py spin-valves. it is necessary to include
HREDS analysis that was performed by John W. Heckman.

96

 

Figure 4.11 A cross-sectional brightfield STEM image of the initial Ag/Py spin-valve
grown directly on the Si substrate. Points A-D indicate the loci of origin of the X-ray
spectra acquired.

97

 

S' '
3000 - 1k“ A

2000:

1000

 

 

 

 

3000’

3000’
2000

2000

 

 

 

 

 

 

1000 InLB‘

1000

 

 

 

 

 

 

 

-Ray Counts

X
Z.
r-
E

1000

 

 

4000 » E

3000

/\g14n
2000

 

1000

 

AgL02
AgT-rl

0.00 1.00 2.00 3.00 40—0 5.00 6.00 700 8.00 900 1000
keV

Figure 4.12 EDS X-ray spectra collected from the Si substrate and the first three layers of
the Ag/Py spin-valve illustrated in Figure 4.11. (A) Spectrum collected from the silicon
substrate. (B) Spectrum from the first Ag layer. (C) First Ag layer spectrum with the Ag L
series contribution stripped. (D) Spectrum from the first (unpinned) Py layer. (E) Spectrum
from the second Ag layer.

 

 

 

98

spreading, or beam overlap, of the layer boundary. In figure 4.12 (E), the beam was
positioned on the polycrystalline (second) Ag layer. The spectrum obtained shows only the
expected silver and residual O K and C K peaks with no evidence of indium.

The x-ray spectra of several discrete ratios of Agzln were modeled to elucidate the
stoicheometry in this region and the effect of altering this ratio on the resultant X-ray
spectra can be seen in figure 4.13. The Ag4zln ratio most closely matches the experimental
data shown in figure 4.12 (B).

The observed differences in thickness of the first Ag layer when grown on Si versus
Nb can almost be explained as follows: Not only is there an additional 20 atomic percent In
in this layer, but based on the HRTEM images of this layer, it is reasonable to expect that
the Ag-In phase is not a close packed structure. Thus, the observed dilatation of
approximately 45% is not unreasonable. At present it has not been possible to reconcile the
structure observed using HRTEM with any known Ag-In structures. According to the
equilibrium phase diagram [89], compositions in the range of Ag4ln should lie in a two-
phase field between the FCC Ag solid solution and the hexagonal 5, phase. Neither of these
phases are consistent with the HRTEM images and FFTs of the observed phase, indicating
that this layer is non-equilibrium in nature. As such, the source of the observed 3-fold
contrast modulation in the single crystal structure of this layer is not clear.

The source of the In in the first Ag layer has not been unequivocally identified.
However, indium is used as a solder in the sputtering target assemblies. Thus, it is likely
that the indium contamination was a result of a small amount of excess In solder in the
sputtering system. It is curious that the contamination only affected the Ag grown directly

on the Si, but not on Nb. Thus suggests that the In may have contaminated the substrate

99

 

Agsln

 

 

 

 

 

Agzln
Agln
A9
Ag4|n A95ln

 

 

 

 

 

 

 

 

.1 2.1} ,0

2.00 3:00 4.00 5.00 3.00 4200 5.00 300 4.00 5.00
keV

 

 

Figure 4.13 Simulated EDS spectra of simple stoicheometric ratios of Silver and Indium.
The Ag4:In ratio is the best integer match with the experimental spectrum.

100

during the sputtering of the bottom Nb layer. Unfortunately, sample limitations
precluded EDS examination of the Si/Nb interface for the presence of In. It is important
to note that no form of In contamination was observed in any of the other sputtered
samples examined neither in this study nor in any other spin-valve studies to date.

Despite the accidental nature of how this Ag-In phase was produced, the fact that
it grows epitaxially is intriguing. However, if this phase can be consistently grown, it
may provide a basis for further epitaxial growth of additional structures that cannot be
epitaxially grown directly on 81. In particular it may be possible to grow additional
epitaxial layers in the Ag/Py spin-valve stack using this Ag4ln epitaxial layer as a buffer.
As it is likely that this phase will be non-magnetic yet electrically conducting such an

epitaxial buffer layer should not affect the magnetic nature of the spin-valve.

4.5 SUMMARY

The results here, in agreement with previous work, indicate that Ag/Py spin-
valves show polycrystalline equilibrium structures throughout the device. Even at cross
section thicknesses thin enough to allow HRTEM imaging, the Ag layers provide good
contrast between adjacent layers. There appears to be no major differences between the
spin-valves sputtered on the Nb contacts and those grown directly on the Si substrate. In
one instance, when the spin-valve is grown directly on the Si substrate, the first Ag layer
displayed an epitaxial single grain. HREDS has shown that this single crystal (epitaxial)
layer is likely composed of Ag and In in a 4:1 atomic ratio. The In contamination may be
due to the inadvertent sputtering or evaporation of target solder in the sputtering system

prior to the Nb contact deposition. Indium contamination was only observed in this one

101

sample. In all other spin-valve samples examined to date, polycrystalline morphologies
have been consistently observed regardless of substrate. If reproducible, epitaxial Ag4ln

may be useful in growing epitaxial spin-valves or other structures on Si substrates.

CHAPTER V

STRUCTURAL STUDIES OF POLYCRYSTALLINE Cu/CoZr
GMR SPIN-VALVES

5.1 INTRODUCTION

In the previous two chapters, the details of the structural studies were focused on
ferromagnetic Py based polycrystalline GMR spin-valves including Cu/Py and Ag/Py
spin-valves. In this chapter, instead of a study of ferromagnetic Py based spin valves,
ferromagnetic CoZr based spin valves of two different types will be investigated. One is
a symmetrical exchange biased spin valve (EBSV) in the form of FeMn/CoZr/Cu/CoZr
and the other is a hybrid spin valve with the nominal form of Py/Cu/CoZr/Cu. Both spin

valves were characterized using CTEM.

5.2 CTEM OBSERVATIONS
Nb//Cu/FeMn/CoZr/Cu/CoZr/Cu/INb EBSV (1043-7)

The symmetrical exchange biased spin valve in the nominal form of
(Nb250//Cu.o/FeMng/CoZr2o/Cu20/CoZr2o/CuIo//Nb250)nm was sputtered on a Si substrate.
The cross-sectional brightfield and darkfield CTEM images of this Cu/CoZr spin valve
are shown in figure 5.1 (a) and (b). In both images, the Si substrate, both Nb contacts,
and spin valve multilayers, can be distinguished clearly. Unlike for the Cu/Py spin
valves, the individual layers within the spin valve are also distinguishable. However,
these layers are not clearly imaged as in the Ag/Py spin valves, due to the low mass and
diffraction contrast between FeMn (Z = 25.5), CoZr (Z = 27.3), and Cu (Z = 29). All of

the layers are continuous. smooth, and are uniform in thickness. The interfaces between

103

Stackin Fault
g Moire Fringes (a)

is" ‘ .,. Dislocations

.7; /'

 

Figure 5.1 (a) Brightfield and (b) darkficld cross-sectional CTEM images of the
(Nb250//Cu1o/FeMng/CoZr20/Cu2o/CoZr2U/Cu10//Nb250)nm EBSV sputtered on Si substrate
showing polycrystalline morphology with columnar grains.

104

the layers are straight and sharp when compared to the Cu/Py polycrystalline spin valves.
This is interesting, because the Z differences in these two spin valves are similar (ZCu =
29, Zpy = 27.7, and ZycMn = 25.5 for Cu/Py spin valves). This might suggest less
interrnixing or lower roughness in the Cu/CoZr. Both the Nb contacts and the spin valve
were observed to be polycrystalline with columnar grains, whose long axis is oriented
along the growth direction. Some dislocations, stacking faults, and Moire fringes are
observed in these columnar grains as indicated in figure 5.1 (a). The darkfield image
more clearly illustrates that the sample grows mostly in a column morphology with grain
widths ranging up to 50 nm normal to the growth direction. In most cases, epitaxy within
these columns starts from the bottom Nb contact and continues into, and sometimes
through the spin valve, all the way up to the top Nb contact. In some cases, at the Nb and
spin valve interfaces, sudden changes in image intensity, due to changes in diffraction
contrast, indicate significant changes in grain orientation and loss of strict epitaxy during
the growth process. This morphology is consistent with that observed for polycrystalline

Cu/Py and Ag/Py spin valves.

Nb/lCu/(Py/Cu/CoZr/Cu)x10/Cu/fN b Hybrid Spin Valve (1042-3)

The cross-sectional brightfield and darkfield CTEM images of the
(Nb250//Cu.o/(Py24/Cu20/CoZr20/Cu2o)x l 0/Cu10//Nb250)nm hybrid spin valve sputtered on Si
substrate is shown in figure 5.2 (a) and (b). Again, the Si substrate, both Nb contacts, and
the spin valve can be seen clearly. The individual layers within the spin valve are rather

difficult to distinguish due to the small differences in the average atomic numbers of the

105

Nb

IP“.

0" sv: Cu/(Py/Cu/CoZr/Cu)x10/Cu

A - I.

l

 

0.2 m

Figure 5.2 (a) Brightfield and (b) darkfield cross-sectional CTEM images of the
(Nb250//Culo/(Py24/Cu20/CoZr20/Cu2o)x l 0/Cu10//Nb250)nm hybrid spin valve sputtered on Si
substrate showing polycrystalline morphology with columnar grains.

106

layer components (Zcu = 29, Zpy = 27.7, and Zed, = 27.3). The total thickness of the spin
valve is consistent with its nominal thickness of 860 nm. It is evident that both Nb
contacts and the spin valve are polycrystalline with columnar grains. Significant defects,
such as stacking faults, Moiré fringes, and dislocations, result in substantial contrast in
the columns. All the layers appear to be continuous with uniform thicknesses. It can also
be seen clearly in the darkfield image that the diffraction contrast within the spin valve is
consistent, with 10 quadrilayers of relative uniform thicknesses. In general, the
quadrilayers are continuous and smooth. However, a number of bumps appear in some of
the columns starting from the 5‘h quadrilayer and propagating all the way up to the top of
the Nb contact, with increasing curvature towards the top. The source of these bump
formations is not clear at this moment. HRTEM and analytical analysis are needed to
further characterize these structures. Interestingly, the columns do not appear to get
wider, high in the stack, unlike other previous cases (for example see figure 3.1).
Possibly, for this hybrid spin valve, the widening occurs early in the stack, and reaches a
stable point where all the columns are crystallographically/energetically the same. That
is, the widening observed before (and possibly present in the first or second layer here)
likely occurs from different columns growing preferentially at the expanse of others.
Although the total thickness of this spin valve (860 nm) is much larger than all the ones
previously studied (< 98 nm), epitaxy within the individual columns continues into and

through the spin valve in most cases.

107

5.3 SUMMARY

The results here, in agreement with previous polycrystalline Cu/Py and Ag/Py
CTEM observations, indicate that Cu/CoZr spin valves are polycrystalline with columnar
grains for both EBSV and hybrid spin valve. Some defects, including dislocations,
stacking faults, and Moire fringes, are observed in both spin valves. However, in
comparison to some previous polycrystalline exchange biased spin valves, the columns in

the hybrid spin valve appear to grow straight up with uniform width.

108

CHAPTER VI
STRUCTURAL STUDIES OF EPITAXIAL GMR MULTILAYERS AND SPIN-
VALVES GROWN BY DC SPUTTERING

6.1 INTRODUCTION

Initial structural characterization of GMR materials for magnetic devices
concentrated on polycrystalline materials grown by dc magnetron sputtering, as discussed
in last three chapters. Most of these multilayer incorporated Nb superconducting
contacts, required for GMR measurements with the current perpendicular to the layer
planes (CPP) geometry. Because of the non-equilibrium phases that were found in
polycrystalline Cu/Py and Ag/Py multilayers and spin-valves, and because of the
difficulties in imaging interfaces in polycrystalline samples, it is of interest to grow
epitaxial magnetic multilayers and spin-valves instead of polycrystalline layers, to reduce
structural defects associated with grain boundaries. However, as the program progressed,
it became clear that growing such samples was not a trivial matter. Thus, the study
became focused on assessing the structures of samples grown to be epitaxial. From a
fundamental point of view, epitaxially grown samples should be useful for assessing the
basic GMR parameters with respect to the film crystal structures. In general, molecular
beam epitaxy (MBE) is widely used for growing epitaxial films. However sputter
deposition offers higher growth rates and lower costs. In the present study, the group at
MSU developed parameters to grow epitaxial CPP samples using sputter deposition.
Through an iterative process, success has now been achieved in growing smooth epitaxial

samples using this technique.

109

6.2 PRELIMINARY STUDIES ON EPITAXIAL SAMPLES
A preliminary study of epitaxial growth of GMR multilayers by magnetron

sputtering was carried out by R. Loloee [33]. This study focused on determining the

optimum growth techniques for epitaxial films on (1120) sapphire substrates by sputter
deposition, has shown partial success in growth of epitaxial Nb(110), Cu(l 1 1), Co(111),
Py(l 11) films. That study characterized the films using atomic force microscopy (AFM)
and electron backscatter patterns (EBSPs) [33]. AFM was used to characterize the
surface topography while EBSPs were used to determine crystal structures, orientations,
and growth epitaxy. The EBSP results showed that Nb films grow as high-quality
epitaxial single crystals with (1 10)ch orientations. The EBSP crystallographic
characterization of Cu films revealed two in-plane twin variants corresponding to two
different stacking sequences of {111} fcc planes. Subsequent growth of Co and Py films
resulted in similar morphology. The orientation relationship between Nb and Cu was
found to be the Nishiyama-Wasserman relationship. Although EBSPs can provide useful
orientation information, it requires that films have a minimum effective film thickness of
approximately 200 nm in order to be readily characterized. This film thickness is much
greater than the thickness required for GMR devices (20 nm). Therefore, the details of
the microstructure features of epitaxial magnetic multilayers and spin-valves need to be
further characterized by high resolution microscopy such as CTEM and HRTEM.

A preliminary set of epitaxial C o/Cu multilayers was examined by J.W. Heckman
[90]. In this study, the structures and composition of the epitaxial Nb//Cu/(Co/Cu)x6
multilayers were analyzed using CTEM, HRTEM, and HREDS. One of the conclusions

reached was that when the Nb contact layer is deposited at high temperature and post

110

deposition annealed, a uniform epitaxially grown single crystal layer is formed with a
high defect density as compared to the polycrystalline Nb layers, which are produced at
room temperature. Another conclusion was that the structure of the Cu buffer layers of
all the epitaxial samples deviates significantly from the nominal structure of the
underlying Nb contact layer. Two different structural deviations were observed. For Cu
deposition at low temperatures with post deposition annealing, the structure was
continuous along a lateral direction (parallel to the base layer), but displayed a zone of
composition variation at the interface of the Nb layer. For deposition at higher
temperatures, formation of islands in the Cu buffer layer was observed. In the general
case, the volume between islands was absent of any buffer material, and on some of the
islands, there appear voids in the Cu buffer layer.

Successful methods for growing high quality epitaxial Nb contact layers on
sapphire substrates were perfected by Loloee’s research and confirmed by analysis and
characterization performed by Heckman. However, attempts at growing high quality
epitaxial layers of Cu, Py, and Co were not completely successful, particularly for the
growth of the Cu buffer layer on the Nb contact layer which is the foundation for growing
subsequent epitaxial films in a preferred orientation. Therefore, besides focusing on
characterization of epitaxial multilayers and spin-valves, another objective of this study
was to determine the correct parameters for growing a high quality Cu buffer layer on the

epitaxial Nb contact.

111

6.3 ORIENTATION RELATIONSHIPS BETWEEN LAYERS DETERMINED BY
EBSP [33]

EBSP is a straightforward approach for characterizing the orientation of the
sputtered films. Loloee’s research using EBSP determined the orientation relationship
among the sapphire (A1203), Nb, and Cu layers. Although this information was obtained
on samples with considerably thicker films than are typically used on an actual GMR
device, the data were very useful as a guide for epitaxial relationships examined in
further CTEM and HRTEM studies.

EBSP and AFM revealed that Nb films grow as high-quality epitaxial single
crystals with (110) bcc single crystals. It has been Shown that bcc Nb(l 10) planes can be
grown epitaxially on sapphire (A1203) (1 120) substrates [33]. EBSP results also revealed
that Cu films grown on the epitaxial Nb contacts display two 23 twin variants
corresponding to different stacking orders of {111} fcc planes, i.e. ABCABC versus
ACBACB. The two different orientation variants are also observed in Py and Co films.

The EBSP derived stereographic projections of the epitaxial Cu film and the
epitaxial Nb under layer as well as A1203 substrate are shown in figure 6.1. It can be seen
that the epitaxial relationship between Nb and A1203 was (110)bchb || (1 120 )Al203 with
[1I1]Nb u [0001]/11203 and [112]Nb ll [lTOO]Al203. The subsequent epitaxial
relationship between the bcc Nb and fee Cu variants was found to be (110)bchb ll
(111).~,,Cu, with [001],,,Nh in [0T1 ],~,,Cu and [1i0],,,Nb n [2ii],~,ccu, which is the
well-known Nishiyama-Wasserman (N-W) relationship [91]. This relationship is
different from that reported in section III of this dissertation on polycrystalline samples

sputtered at room temperature, which used HRTEM imaging to show that the Nb/Cu

112

duty—0:28 05-29 ungowmmkrsfimbsflz 05 mg,» 35C? :0

out“ 98 92 can 05 50253 mimcosfle 353on 2: 223 SON—«A Sci __ n2: _ 2 £3, MON—<A cm; 3 __ £75: 3 mg, 8:333.
MON? 98 mEE £2 05 :83qu E52832»: Riggs 2:. .mEE :0 was .32 SON—ace mo mcosoomoa oiafiwoocoum mo Rewr—

 

 

25¢

 

 

:3.

113

orientation relationship was close to the Kurdjumov-Sachs (K-S) relationship (110)bCCNb

ll (111)1ccCu with 11 '1’ 1 11..ch ll [0T11rccCu.

6.4 CTEM OBSERVATIONS
6.4.1 The Initial Series of Epitaxial Samples
Nb/lCu/Py/FeMn/le Multilayer (31-2)

The first epitaxial sample was grown in the nominal form of
(Nb2so//Cu100/Pyloo/FeMnso/Nbioo)nm. For this sample, the Cu bUffer layer was grown at
150°C and post-annealed at 350°C; the Py layer was then grown at 350°C; no heat was
applied during the growth of the rest of layers (FeMn and top Nb layers). A cross-
sectional brightfield/darkfield image-pair of this epitaxial Cu/Py multilayer is shown in
figure 6.2 (a) and (b). The sapphire substrate, both Nb contacts, and the spin-valve
multilayer can be distinguished clearly. The individual layers are well defined and
consistent with their nominal thicknesses. It can be seen that both bottom Nb and the
spin-valve display epitaxial morphology which shows a single crystal Nb Spanning the
image, while the Cu, Py, and FeMn regions appear as large grains, up to 1 pm wide.
Significant grooving appears between these grains, suggesting island growth of the Cu
region with subsequent disruption of the Py and FeMn layers. In the darkfield image,
different contrast is evident in the Cu, Py, and FeMn layers, which indicates orientation
variants exist in these layers. In the first Cu buffer layer, it can be seen that there are
some voids and discontinuities, suggesting poor epitaxial Cu layer grown on the Nb. A
different morphology can be seen in comparison to that of the cross-sectional

brightfield/darkfield image-pair of a polycrystalline Cu/Py spin-valve as shown

114

 

Figure 6.2 (a) Brightfield and (b) darkfield cross-sectional CTEM images of the epitaxial
(Nb20(.//Cu1m/Pyloo/FeMn50/7’Nb.00),,m multilayer.

115

in figure 6.3 (a) and (b). In this polycrystalline sample, the Si substrate, the Nb contacts,
and the spin-valve can all be clearly distinguished and appear continuous and uniformly

thick with columnar growth.

NbI/Cu/Py/Cu/Py/FeMnl/Au/Nb/Au Spin-valve (27-3)

A cross-sectional brightfield and darkfield CTEM image-pair of an epitaxial
[N0250//Cu20/Py24/Cu20/Py24/FeMn3//Au15/Nb100/Au.5],1m spin-valve is shown in figure 6.4
(a) and (b). The Cu buffer layer was grown at 450°C with no post annealing; the first Py
layer was grown at a low temperature ranging from 80°C to 100°C; the remaining layers
were grown at room temperature. The sapphire substrate, the two Nb contacts, both of
the Au layers, and the spin-valve layers can be seen clearly. The individual Cu, Py, and
FeMn spin-valve layers are relatively distinguishable and consistent with their nominal
thicknesses. Similar with the first epitaxial Cu/Py/FeMn multilayer sample, a single
crystal of Nb spans the image, while the Cu, Py, and FeMn layers appear as single
crystals as well. The Au and top Nb contacts appear polycrystalline as expected. Only
minor grooving is observed in this sample, but some voids and discontinuities appear in

the Cu buffer layer, suggesting poor epitaxial Cu buffer layer grown on Nb.

Nb//Cu/Co/(Cu/Co)x20//Au/Nb/Au Multilayer (28-3)
A cross-sectional brightfield and darkfield CTEM image-pair of an epitaxial
[N0250//Cu20/Co._5/(Cu2/Col,5)x20//Au15/Nb.oo/Au15]nm multilayer is shown in Figure 6.5.

For this sample, the Cu buffer layer was grown at 350°C with no post annealing; the first

116

I irection

(b)

/

 

Figure 6.3 (a) Brightfield and (b) darkfield cross-sectional CTEM images of the
polycrystalline (Nb250//Culo/FeMng/Py30/CU20/Py3olbe250),,m spin-valve.

117

(a)

  
     

Growth
Direction

Sapphire

 
 

0.1!1m

I" .
.413. ‘\ T

Figure 6.4 (a) Brightfield and (b) darkfield cross-sectional CTEM images of the epitaxial
(Nb250//Cu20/Py24/CU20/Py21/FeMng/lAu] 5/me/Au15)nm spin-valve.

118

(a)

   
 

Growth
Direction

    

Sapphire

 

in .. J ‘7. j"._ b

Figure 6.5 (a) Brightfield and (b) darkfield cross-sectional CTEM images of the epitaxial
(Nblfitlll/CUZo/COI,5/(CU2/C0l.5)><20/s/AU15/Nb100/Au15)nm multilayer.

119

Co layer was grown at 400°C; no heat was applied during the growth of the remaining
layers. The sapphire substrate, the Nb contact layers, both Au layers, and the Cu/Co
bilayers can be readily distinguished. A weak contrast modulation in the 20 Cu and Co
bilayers is also evident (see the inset of enlarged image in Fig. 6.5). The bottom single
crystal Nb shows diffraction contrast consistent with a large number of dislocations. The
Cu and Co layers appear single crystal as well. The Au and Nb contacts appear as
polycrystalline as expected. A small amount of grooving also appears between these

grains. Again, some voids and discontinuity are displayed in the [Cu buffer layer.

6.4.2 A New Series of Epitaxial Samples:

In the previous studied epitaxial samples (31-2, 27-3, and 28-3), the Cu buffer
layer was grown either at low temperature (150°C) then post-annealed at a relatively high
temperature (350°C) or grown at a relatively high temperature (350°C or 450°C) with no
annealing. Poor epitaxial morphology was observed in all these samples, especially for
the Cu buffer layer grown on Nb. Loloee’s AFM studies of similar samples have shown
that the surface topography is strongly affected by the temperature and surface quality of
the substrates. AFM images and line trace surface analyses show relatively smooth
surfaces on 150 nm thick Cu films deposited at a substrate temperature of 475°C, and on
75 nm thick Cu films deposited at 150°C. All of the Cu films that were post-annealed

displayed rougher surface. Therefore, based on these AFM studies and CTEM
Observations, it was decided that a new series of samples was needed to fully determine

the effect of temperature on epitaxial growth.

120

To obtain good epitaxial Cu films grown on Nb, four new samples were
fabricated, two with a high deposition temperature (500°C), and two with low deposition

temperatures (100°C and 150°C). None of the samples were subjected to post-annealing.

6.4.2.1 Epitaxial samples grown at high deposition temperature
Nb/lCu/Py Multilayer (41-3)

A cross-sectional brightfield CTEM image of the (Nb200//Cu100/Py200)nm
multilayer is shown in figure 6.6. For this sample, the Cu bUffer layer was grown at
500°C and received no post-deposition annealing, while the Py layer was grown over a
low temperature range from 80°C to 100°C. Similar with the first epitaxial sample (31-
2), the Nb contact layer appears as a single crystal and spans the image, while the Cu and
Py regions appear as large grains. Significant irregular grooving and a discontinuous
morphology appear between these Cu and Py grains. The image also shows that the Cu
buffer layer and Py layer were broken into islands, suggesting pin holes may occur in
these layers and cause the potential for electrical shorts which should be avoided in
magnetic devices. In some areas, the Cu buffer layer appears to be absent, with the Py
layer in direct contact with the Nb layer. Some voids and dislocations are displayed in
the Cu buffer layer. In the epitaxial Py layer, dark mottled contrast appears throughout
the layers. This contrast is consistent with the effects of a very high density of

dislocations and stacking fault fringes.

121

Figure

 
 
 
 
 

High defect densities
(most likely dislocations
and faults)

 
 
 
  

Growth
Direction

  

6.6 A cross-sectional brightfield CTEM image ' of the epitaxial

(Nb2oo//Culoo/Py2oo)nm multilayer with Cu deposited at 500°C showing the poor epitaxial
growth of Cu and Py layers.

122

6.4.2.2 Epitaxial samples grown at low deposition temperature
Nb/lCu/Py/le Multilayer (43-1)

Homogeneous and smooth epitaxial Cu layers grown on Nb were found in the
multilayers when the Cu films were deposited at low temperatures (100°C and 150°C),
and the rest of the layers were deposited at room temperatures. A cross-sectional
brightfield CTEM image of the epitaxial (Nb200//Cu20/Py2o/be20)nm with Cu deposited at
100°C is shown in figure 6.7. The sapphire substrate, both Nb contacts, Cu, and Py
layers can be distinguished clearly. It can be seen that all the layers are continuous,
smooth and uniform in thickness, and no grooves or voids appear in the Cu buffer layer.
The bottom Nb, Cu, and Py layers display epitaxial growth. Some dislocations are
observed throughout the entire view of the cross-sections. The top Nb layer is

polycrystalline as expected.

Nb/lCu/Py/le Multilayer (43-2)

A cross-sectional brightfield and darkfield CTEM image-pair of an epitaxial
(Nb200//Cu20/Py20/be20)nm multilayer with Cu deposited at 150°C is shown in figure 6.8.
The sample was oriented with electron beam parallel to sapphire (0001) zone axis. It is
evident that all the layers are continuous, smooth, and uniform in thickness. There are no
grooves or voids in the Cu buffer layer. The bottom Nb, Cu, and Py layers display
epitaxial growth throughout the entire view of the cross-sections. The top Nb layer is
polycrystalline as expected. The inset shows the diffraction pattern of the entire

multilayer which contains the Nb, Cu, and Py reflections. Only faint Cu and/or Py twin

123

Dislocations

Growth
Direction

 

Figure 6.7 A cross-sectional brightfield CTEM image of the epitaxial
(Nb2oo//Cu20/Py20//Nb2o)nm multilayer with Cu deposited at 100°C and rest of the layers
deposited at room temperature. It shows the high-quality epitaxial growth of Cu and Py
layers.

124

 
 
 
 
   

Growth
Direction

a tile.

  

t" .v as.

211“,
. lll('u.l'_\ I',720

.IIOW (11.1} ‘

0021'“,-

-- l
I I l(‘u.l‘y ' . v. _ I
. . 111m,

002011.13 ‘ Q

0
o ' 11
111W,

 

Figure 6.8 (a) A cross-sectional brightfield CTEM image of the epitaxial
(Nb200//Cu2o/Py2ol/Nb2o).1m multilayer with Cu deposited at 150°C and the rest of the
layers deposited at room temperature. It shows high-quality epitaxial growth of Cu and
Py layers. (b) The diffraction pattern taken from Nb-Cu-Py shows a portion of the
<1 10> twin reflections of Cu and Py.

125

reflections appear in the pattern. From EBSP studies, it is known that there is an
orientation relationship between Nb and Cu with 526° between the (1 10)Nb and (111)cu
directions (see figure 6.1). Therefore, to show the full <110> twin pattern of the Cu, the
sample was tilted about 5° parallel to the growth planes. A cross-sectional brightfield
and darkfield CTEM image-pair and corresponding diffraction pattern of the sample after
tilting is shown in figure 6.9. It is now very clear that there are two different orientations
in the Cu and Py layers. In the darkfield image, the two variants are more evident. Both
the Cu(111) and Py(111) planes appear to be epitaxially grown on the single crystal
Nb(110) layer. The diffraction pattern in figure 6.9 (c) shows two complete sets of the
<110> twin pattern. This figure indicates that the Cu and Py layers were grown as two
variants, consistent with the EBSP results.

A cross-sectional brightfield and darkfield CTEM image-pair of the same
sample taken in the (1 100) zone axis of the sapphire is shown in figure 6.10. Again, all
the layers can be distinguished clearly and display continuous and smooth structure and
are uniform in thickness. No grooves or voids appeared in the Cu buffer layer. The inset
diffraction pattern of the entire multilayer shows that (110)bchb is epitaxially grown on
(11§O)hcpA1203 substrate, while (1 1 l)fccCU and (111)fcch are epitaxially grown on

(110)bchb. No twin reflections were observed in this diffraction pattern.

6.4.3 Epitaxial Samples of Nb/le/Cu/le Multilayers (43-3 and 43-4)
Two epitaxial samples were sputtered by growing Py on top of the epitaxial Nb

contact instead of a Cu buffer layer: one was deposited at 150°C, and the other was

  
   

Growth
Direction

Sapphire

Figure 6.9 (a) Brightfield and (b) darkfield cross-sectional CTEM images of the epitaxial
(Nb20(,//Cuzo/Pyzo/bezo)nm multilayer with Cu deposited at 150°C showing the high-
quality epitaxial growth of Cu and Py layers. The sample was tilted about 5° parallel to
the growth planes. (c) The diffraction pattern is taken from Nb-Cu-Py layers which
shows the full <110> twin pattern of Cu and Py after tilting.

127

(a)

 
      
     

Growth
Direction

Growth

Sapphire, J, I’

0.1 pm

 

Figure 6.10 (a) Brightfield and (b) darkfield cross-sectional CTEM images of the
epitaxial (Nb200//Cuzo/Pyzo//Nb20)nm multilayer with Cu deposited at 150°C. The inset
diffraction pattern shows the sample was oriented to the (ITOO) zone of sapphire

substrate.

128

deposited at 350°C. Neither of these samples were post annealed. Figure 6.1] shows a
cross-sectional brightfield CTEM image of (Nb200//Py20/Cu20//Nb20)nm multilayer with Py
deposited on Nb at 350°C. It is evident that irregular and rough grooves appear in the Py,
Cu, and top Nb layers. It is difficult to distinguish the individual Py, Cu, and top Nb
layers. A cross-sectional brightfield CTEM image of (l\lb200//Py20/Cu20//Nbgo)nm
multilayer with Py deposited on Nb at 150°C as shown in figure 6.12. The sapphire
substrate, the top and bottom Nb contacts, the Py and Cu layers can all be clearly
distinguished. The individual layers are well defined and consistent with their nominal
thicknesses. However the Py and Cu layers do not appear to have epitaxial morphology
as expected. Instead both layers are polycrystalline with large columnar grains
throughout the entire view of the cross-sections. Considerable grooving appears between

these grains.

6.5 HRTEM STUDIES AND F F T ANALYSIS
6.5.1 Epitaxial Nb Grown on Sapphire Substrates

The (110) Nb contacts were the first successfully grown epitaxial layers on the
(1 150) sapphire (A1203) substrates. As shown in the stereographic projections of A1203
and Nb (see figure 6.1), the (110)Nb and (11§O)A1203 planes are parallel with [1T1]Nb

H [0001]A1203 and the [1T§]Nb || [1I00]A1203 directions. Both of these directions
offer a potential imaging axis, and cross-sections can be cut along either orientation.
A cross-sectional HRTEM image of the Nb-A1203 interface in the Cu/Py/FeMn

multilayer is shown in figure 6.13. The Cu buffer layer was grown at 150°C and post-

     

Growth
Direction

Sapphire
HE:

Figure 6.11 A cross-sectional brightfield CTEM image of the epitaxial
(Nb200//Py20/Cuzo//Nb20)nm multilayer with Py deposited at 350°C on Nb showing poor
epitaxial growth of the Py and Cu layers when Py grew on Nb first instead of Cu.

130

Growth
Direction

 

Figure 6.12 A cross-sectional brightfield CTEM image of the epitaxial
(Nb200//Py20/Cu20//Nb20)nm multilayer with Py deposited at 150°C on Nb showing the
polycrystalline growth of the Py and Cu layers when Py grew on the Nb first instead of
Cu.

131

.8823 MON?
SE C 2: :o gflxfifio 858m was £2 3 M S Rambo BwEm 353:5 was/02m .5wa38 :Eorfizmxzo 05 E MON?

can 92 2: mo 28qu WE wiveommobg 98 ooflbfim MON—<52 05 c8 owns: 2955 620303-380 3.0 DSwE

 

132

deposition annealed at 350°C; the Py layer was then grown at 350°C; and no heating
applied during the growth of the rest of the layers. The cross-section was oriented with
the electron beam parallel to the [0001] A1203 direction. It can be seen that the Nb layer
displays a single crystal morphology, consistent with the CTEM observations. The Nb-
A1203 interface is structurally coherent, sharp, and defect free. The lattice planes transfer
continuously from A1203 to Nb as indicated in the figure with no obvious epitaxy
dislocations. The corresponding FFT diffractograms (see right column of figure 6.13)

show a <111> zone in Nb aligned along the [0001] zone of the A1203, and show that the
single crystal Nb has grown with the (110) planes epitaxial to the (1120) A1203
substrate, consistent with the EBSP results.

A cross-sectional HRTEM image of the Nb-Ale3 interface oriented with the
electron beam parallel to the [1 T00] A1203 direction is shown in figure 6.14. The lattice
image also shows that the interface is atomically sharp and flat. The corresponding F FT
analysis shows a <112> zone in Nb aligned along the [1 TOO] zone of the A1203, and
shows that the single crystal Nb has grown with the (110) planes epitaxially to the
(1120) A1203 substrate, also consistent with the EBSP results. The (110) Nb planes and
(1 120) A1203 planes lie parallel to the interface as indicated in the image. Only the (110)
Nb planes with a spacing of 0.233 nm are resolvable due to the resolution limits of the
microscopy. The (222) lattice planes, which have a spacing of 0.095 nm, is below the

information limit of the JEOL-4000EX microscope used.

133

“1 Q‘s ‘~

\‘n‘: {1‘5 \Yat

 

Figure 6.14 A cross-sectional HRTEM image of the Nb-Sapphire interface and
corresponding FFT patterns of the Nb and sapphire in the Cu/Py/FeMn multilayer,
showing epitaxial single crystal (110) Nb has grown epitaxially on (1120) sapphire
substrate.

134

6.5.2 Nb/lCu/Py/F eMnI/N b Multilayer (31-2)

A cross-sectional lattice image of the Cu-Nb interface in the Cu/Py/FeMn
multilayer is shown in figure 6.15. The Cu buffer layer was grown at 150°C and post-
annealed at 350°C, the Py layer was grown at 350°C, with no applied heat during the
growth of the rest of layers. There are multiple small twins in the Cu region growing
from the nucleation interface. As labeled in the image, some of these twins are in the

(111)1-ccCu||(110)bchb epitaxial orientations (labeled as E), while others are not (labeled

as N). Instead the volumes labeled N are oriented with the ~(3T1)fccCu||(110)bchb. It is
possible that these non-epitaxial orientations occur as a result of strain accommodation.
The small twins exist with thicknesses up to ~12 nm, beyond which large single (1 1 l)fcc
variants exist over the remaining breadth of the 100 nm Cu layer.

A cross-sectional HRTEM image of the Cu-Py interface in the Cu/Py/FeMn
multilayer is shown in figure 6.16. The interface between the Cu and Py is just barely
distinguishable. The interface between the Cu and Py is relatively sharp and flat. Both
layers display epitaxial morphology. The lattice planes grow continuously from Cu layer
to Py layer. Selected HRTEM images and corresponding FFT diffractograms of Cu and
Py are shown in figure 6.17. Both diffraction patterns are characterized by 55° and 70°
interplanar angles, indicating <110> zone axes of fee Cu and L12 Py. In both cases, the
growth direction is near <1 1 1>, suggesting the close packed planes are epitaxial in the fcc
Cu and Py. The {111} interplanar spacings were measured to be 0.21 nm for Cu and
0.20 nm for Py, which correspond to fcc lattice parameters of 0.36 nm (acu = 0.361 nm in

the Cu bulk state) and 0.35 nm (apy = 0.355 nm in the Py bulk state), respectively.

135

_h~._t)>»

E
8
5.3.

.0
:3

 

Figure 6.15 A cross-sectional HRTEM image of the Cu-Nb interface showing multiple
small twin variants near the nucleation interface in the Cu layer of the Cu/Py/FeMn
multilayer. Note that the regions labeled as E are oriented along (111)fcc || (110)l,cc while
those labeled N are not.

136

 

Figure 6.16 A cross-sectional HRTEM image of the Cu-Py interface in the Cu/Py/FeMn
multilayer. The bars indicate the Cu-Py interface.

137

 

arms;- .. . ..
Figure 6.17 Selected HRTEM images and corresponding FFT diffractograms of Cu and
Py showing <110> zones with growth planes of {1 1 l }.

138

6.5.3 Nb//Cu/Co/(Cu/Co)zo//Au/Nb/Au Multilayer (28-3)

In the (Cu/Co)><20 multilayer, high quality single crystal (1 T0) Nb also grew
epitaxially on the (1120) A1203 substrate. Again the interface is found to be sharp, flat,
and defect free. A cross-sectional HRTEM image of the Cu-Co bilayers in the
(Cu/Co)x20 multilayer is shown in figure 6.18. Somewhat surprisingly, a contrast
modulation between the Cu and Co layers is observed. Because Cu and Co have similar
atomic scattering factors (as their atomic numbers differ only by 2), this contrast is not
expected from differential scattering and the source of the contrast is not clear. However,
the modulation period is consistent with the nominal thicknesses of the Cu (2.0 nm) and
Co (1.5 nm) layers. The enlarged HRTEM images and corresponding FF T
diffractograms shown in figure 6.19 indicate that both the Cu and Co are fee. This is
consistent with EBSP results [92] and prior studies on Co films [93], which reported that
Co takes on an fcc structure rather than the bulk equilibrium hcp structure for thin films.
However, the HRTEM results here show that the growth planes of both the Cu and Co in

this (Cu/Co)x20 multilayer are {100} rather than the close packed {111} planes.

6.5.4 Nb/lCu/Py/IN b Multilayer (43-2)

A cross-sectional HRTEM image of the (Nb250//Cu20/Py20//Nb20)nm multilayer
with Cu deposited at 150°C and rest of the layers deposited at room temperature, is
shown in figure 6.20. The bottom Nb, the Cu buffer layer, and the Py layer all display a
single orientation throughout the entire area of the imaged cross-section sample. The top
Nb displays a polycrystalline morphology with columnar grains. The interface between

Cu and bottom Nb is relatively sharp, flat, and smooth. Although the interface between

139

 

Figure 6.18 A cross- -sectional HRTEM 1mage of the Cu- Co bilayers 1n the (Cu/Co)x20
multilayer. Note that the observed Cu-Co periodicity lS consistent with the nominal layer
thicknesses.

140

  

 

 

Growth [Direction .. __

119‘“ ‘ . .1 ”.—-—— 0"
Figure 6.19 Selected HRT EM images and corresponding FFT diffractograms of Cu and
Co showing <110> zones (slightly off axis) with growth planes of {100}.

141

 

Figure 6.20 A cross-sectional HRTEM image ofthe (Nbg50//CUgo/Py2()//Nb20)nm multilayer
with Cu deposited at 150°C and rest ofthe layers deposited at room temperature.

142

the Cu and Py is not as evident as the Cu-Nb interface, it is distinguishable. It can be
seen that the lattice planes grow continuously from the bottom Nb contact to the Cu
buffer layer and up to the Py layer. The Cu layer displays a high quality epitaxial
morphology with the Nb contact layer. It is evident from HRTEM observations that the
Cu displays two orientation variants separated by a narrow slab which is perpendicular to
the growth planes and starts from the bottom of Cu layer through into the Py layer. This
narrow slab is an incoherent twin boundary between two single crystals of Cu. In the Py
layer, there are some dark contrast variations running perpendicular to the growth
direction. These areas occupied more than 50 percent of total observed Py cross-
sectional areas and averaged about 10 nm in width. Given an approximate foil thickness
of 10 nm, it is not surprising that these fringes appear in about 50% of the viewable cross-
section.

For FFT analysis in this sample, the Nb reflections were used as internal
calibration standards to determine the effective camera lengths for interpreting the
subsequent FFT diffractograms. The enlarged HRTEM image and corresponding FFT
diffractogram of bottom Nb is shown in figure 6.21. The interplanar angles are
characterized by 60°, indicating a <111> zone axis of bcc Nb. The growth direction is
close to <110>, indicating the growth has occurred on the closest packed bcc {110}
planes. The {110} interplanar spacings were used as a standard to calculate an effective
camera length for measuring subsequent diffractograms. Using this Nb calibration,
selected areas from the Cu and Py layers have been analyzed as following. Figure 6.22
shows the HRTEM images from selected areas (labeled and indicated in figure 6.20) and

corresponding FFT diffractograms of the Cu buffer layer. Figure 6.22 (a), (b), and (c)

143

 

Figure 6.21 An enlarged HRTEM image and corresponding FFT diffractogram of the
bottom Nb showing <1 1 l> zone axis of bcc Nb.

144

 

Figure 6.22 The HRTEM images from selected areas (a) from left side (b) from right side
(c) twin boundary and corresponding FFT diffractograms of the Cu buffer layer showing
<110> zones with growth planes of {111}.

145

correspond to the left side, the right side, and the twin boundary area of Cu. In figure
6.22 (a) and (b), both diffraction patterns were characterized by 55° and 70° interplanar
angles, indicating <1 10> zone axes of fcc Cu. In both cases, the growth direction is near
<111>, suggesting the close packed planes are epitaxial in the fcc Cu. The {111}
interplanar spacings were measured to be 0.21 nm, which correspond to fcc lattice
parameters of 0.36 nm (acu = 0.361 nm for the Cu bulk state). In figure 6.22 (c) which
corresponds to the area around the twin boundary between the two single grains of Cu,
the FFT diffractogram shows the twin reflections with two common spots in growth
direction. The interplanar spacing of these two common spots is calculated to be 0.21
nm, which corresponds to the {111} planes of fcc Cu. The resolvable zones of these twin
spot patterns are characterized by 55° and 70° interplanar angles, indicating two <110>
zone axes of fee Cu with two common {111} reflections. These patterns show that the
(11]) close packed planes in the fcc Cu are epitaxial growth with the {110} closest
packed planes in the bee Nb and display two 23 twin variants corresponding to different
stacking orders of {111} fcc planes, i.e. ABCABC versus ACBACB (as indicated in
figure 6.20), consistent with the EBSP results from the thick films.

The HRTEM images from selected areas (indicated in figure 6.20) and the
corresponding FFT diffractograms of Py layer are shown in figure 6.23. In some areas,
the Py layer displayed epitaxial growth with their equilibrium phases as shown in figure
6.23 (a). FFT analysis shows that the diffractograms of these areas are <110> zone
patterns and the growth planes are {111}, indicating the close packed {111} planes in the

L1; Py are epitaxial growth with the {111} close packed planes in the fcc Cu. However

146

L 12(1'CC)

 

Figure 6.23 Selected HRTEM images and corresponding FFT diffractograms of Py
showing (a) <110> zone with grth planes of {111} (b) a periodic lattice contrast with
the interval of three {1 1 1} planes. The additional reflections at 1/3 and 2/3 of the primary
(111) reflection spacings are indicated by arrows.

147

in some other areas as shown in figure 6.23 (b), a periodic lattice contrast with the three-
layer interval running perpendicular to the growth direction is observed. The FF T
diffractogram obtained from the periodic contrast region shows additional reflections (as
indicated by arrows) at 1/3 and 2/3 of the primary (111) reflection spacings which are in
the growth direction, and is consistent with the contrast modulation noted from HRTEM
image. If the additional 1/3 and 2/3 reflection spots were ignored, the primary reflections
are characterized as two <110> fcc patterns with two common {111} reflections similar

with the one observed in Cu twin diffractogram.

6.6 DISCUSSION
6.6.1 The Effect of Growth Temperature and Post Deposition Annealing

The characterization results of the epitaxial magnetic multilayers and spin-valves
demonstrated that the quality of the epitaxial growth of the Cu buffer layer is very
temperature dependent. For example, in the first series of epitaxial samples (31-2, 27-3,
and 28-3) poor epitaxial morphology was observed in all the samples when the Cu buffer
layer was grown either at a low temperature (150°C) then post-annealed at a relatively
high temperature (350°C), or when the buffer layer was grown at a relatively high
temperature (350°C or 450°C) with no annealing. Loloee’s AFM studies of similar
samples have also shown that the surface topography is strongly affected by the
temperature and surface quality of the substrates. In these studies, AFM images and line
trace surface analyses show relatively smooth surfaces on 150 nm thick Cu films
deposited at a substrate temperature of 475°C, and on 75 nm thick Cu films deposited at

150°C. However, all of the Cu films that were post-annealed displayed a rougher

148

surface. Therefore, based on this AF M study and the CTEM observations of the first
series of epitaxial samples, a new series of samples was fabricated in order to fully
determine the effect of temperature on epitaxial growth. Two samples were grown at a
high deposition temperature (500°C) and two were grown at low deposition temperatures
(100°C and 150°C). None of the samples was subjected to post-annealing. Following
CTEM observations showed poor epitaxial morphology when Cu buffer layers were
grown at 500°C. High quality epitaxial Cu buffer layers were obtained when Cu was
grown at low temperatures (100°C and 150°C) without post annealing. Although the
layers are smooth and continuous, significant contrast from faults and dislocations are
observed. Ideally, one would like to reduce these. However, the results presented here
indicate that annealing these layers to reduce the defects formed in sputtering is
problematic, with post-sputtered annealing samples showing agglomeration. However,
all of these post-sputtering annealing treatments occurred after the Cu was sputtered, but
before any additional layers were laid down on the Cu. It might be possible to reduce the

defects by annealing after capping layers are applied.

6.6.2 Overlapping Twins in Py

The three-layer periodic contrast observed in the epitaxial Py layer is somewhat
surprising, although this kind of the contrast is not unusual and has been found and
studied in other materials by many groups. Bender et al. [93] reported seeing triple
periodicity along the [111] direction in very high dose phosphorus ion—implanted (011) Si

wafers. This phenomenon was explained using the model of overlapped twins by

149

simulating the high resolution electron microscopic images with the real space method.
Hutchison, et al. [95] and Vanhellemont, et al. [96] also observed the similar twinning
structures in Si crystals by different techniques. A similar model to explain the same
phenomenon in Au films was proposed by Hashimoto et a1. [97]. Wu et al. observed
triple periodicity in GaAs crystalline film grown on Si substrates by the MBE method
[98]. They explained this periodicity by an overlapped twins model which was
confirmed by image matching between observed HRTEM images and simulations using
the multi-slice method. Abe et al. [99] reported three-fold periodicity of {111} in
massively transformed y-TiAl alloys and concluded that the contrast is caused by
overlapping twin related crystals along the incident beam direction. An example of this
three-layer periodic contrast in y-TiAl alloys [99] is shown in Figure 6.24. It can be seen
clearly from the HRTEM image shown in figure 6.24 (a) that the contrast corresponds to
the periodic contrast of intervals of three close-packed {111}, planes. The periodicity
gives rise to extra reflections at positions of 1/3 and 2/3 of the primary {111}Y spot as
indicated. by the arrows in the FFT diffractogram in the inset of figure 6.24 (b). Figure
6.25 shows an enlargement of the framed part in figure 6.24 and simulated images of the
overlapping twins. The simulations confirm that the contrast is due to double diffraction
caused by overlapping twin-related crystals along the incident beam direction.

In this study, the HRTEM images and corresponding FFT diffractograms of Py
show similar contrast and periodicity as those mentioned above. To explain this
phenomenon in more detail, a schematic illustration of the overlapping twin model of Py
is shown in figure 6.26. The projections of atoms of Py (NiFe) single crystal are along

the [110] direction for twins of two stacking orders of {111} fcc planes, i.e. ABCABC

150

II I lllllll % Illl‘ll 5.]

d

II’I:

'~'l"-~‘ .
-c~. ‘nOv.-r

\-ur.o‘l I-v - '
. want». - “pun... ..

 

Figure 6.24 (a) A cross-sectional HRTEM image showing a periodic contrast with the
three—layer interval and (b) its corresponding Fourier transform of the image in the
massively transformed 7 grain [99].

151

6*.§&~-~¢.m.pk
fi-O.‘D.--r'

.§m€~.§a‘

l‘
l
v
1
O

l
‘
l
'

 

Figure 6.25 (a) Enlargement of the framed part of figure 5.24. Simulated images of the
overlapping twins (b) for the total thickness of 4 nm at Af= —60 nm and (c) for the total
thickness of 16 nm at Af= —75 nm [99].

152

versus ACBACB. The black and white dots indicate atoms of Fe and Ni respectively.
When the atomic arrangements of these two planes are superimposed on each other as
shown in figure 6.26 (b), it can be seen that the atom positions of both crystals become
coincident with each other at every third plane as indicated by the arrows. The resulting
periodicity is in good agreement with the HRTEM image of Py in figure 6.23. For this

overlapping twin situation, Abe et al. [99] depicted the corresponding diffraction pattern
as shown in figure 6.27 (a) which consists of reflections with the [110] zone axis of the
matrix and its twin. The pattern is in good agreement with the FFT diffractogram of Py
in figure 6.27 (b). The additional two reflections indicated by arrows are due to double
diffraction. For example, the 1/3 spot can be the result of (002) diffraction of the matrix
double diffracting with the (III) reflection of the twin. Therefore based on these
comparisons, we can conclude that the three-layer periodic contrasts observed in epitaxial

Py layer results from overlapped twins.

6.7 SUMMARY

The microstructure of the epitaxial magnetic multilayers and spin-valves using
sputter deposition were characterized by C TEM and HRTEM.

The CTEM observations suggest that the quality of good epitaxial growth of Cu
buffer layers depends strongly on deposition temperatures and post annealing. Poor
epitaxial morphology was observed when Cu was grown either at low temperatures with
post annealing or at high temperatures without annealing. The high quality of epitaxial

Cu buffer layer was obtained when Cu was grown at low temperatures (100°C and

153

(a)

 

r. o
C O O C‘ O 0
0 O O O O " 0 C
o 1 o o v c. o o o r.» o v
Q C ) C . O ) C '2 I 0 O 3 . O
o o o o o o c o o o o o o o c- o o o o o o
oooooorvoc»o->ot.- oooooouoooooo
oocoooooooeoc-or; «oooooooooooooo
oooocooooooooocoo oooocoooooooOoooo
cooocroooooooooeoc-C Booooooaooooooocoo
oooooocoooeononorfifl Cnooooonoooooooooo
ooeoooc-ocoeooorosA Aaoooooooooooooooo
cocooococoeofioaoac Boot-oaoooooooooooo
o o c o n o o o c o r o - o a o cB Cr. o c. o 1. o n o o o o o o o o o o
o o o o n o o o a o -- o . o c o A A o o o .2 o z- o o o o o o o o o o
O O o o o O C o o o =: o o o DC B.) 9 ~;- o o 1. o o o O o o o 0 o 9
o o o o e o > o o u» o o o B Co o .-.- o o , o o o c» o o o o o o
o o o o e o o o o o o c o o o 91‘ x o c o o o . o o o o o o o o o o
o o o o u o o o a) o '- o . . o .- .= o v o 9 o 1;,- o o o o o a o o
o o o o r‘ o o o o o : o ,; v o o o c o o o o o :1 o o
0 O ’.' . ’.‘ C C‘ C C C 5' C? O O O '3 . O C 0 O O
1 O O > O O v 0 O O u C v O
o o 1 o r \ / n o o o o r
(‘3 C O 3 O . ) O '3
0 . \- O O O
(V
W
J. C" O

o «3 o c “o N ooooooon oo o..- LOO-51'! o o o o
2 o c o 5 o= 1...“in moot. :oooco on'-'-o c o 1; o :»
—>\:»ot;ooooooooooooooo-“ox:
aoc-noel-ow'ownooeaoooooooooorooo
-:-oooo~<woo<~)oon¢ooonooor~ooor~ooo;o:
(b) —>ooso-~ooooooocoooo

O’CI‘ '0 0“) -. 0') ”COG ’3.“ 0.."3 MON C3." 5').

o uxoooomotioooocooooon-Joc»o
——> o o ~21 o u o o o 3 o o o n o a o o
o .- o or.“ «o on -.>o or no o:_~ oo o«: oo o‘ Ao o o
T-O“. o oooueoowz-oooooooooor- o o o
-———I>» o . o .. o o o x; o u o «9 o a o . o » o a
‘ o ~ 0 c on no oc- rz-o oo no o 3o -. o o ~
r o o - co ov~ 0o ow to o o -; o
-————I>«; o c o 0 o w o o o : o ..
o ». o' ..o -. ‘-o g ,,
o . no o. o
o o

Figure 6.26 Schematic illustrations of the overlapping twin model to explain the periodic
contrast with the three-layer interval. (a) A projection of atoms along the [1 T0] direction
of the twin-related Py single crystal. (b) Two Py crystals superimposed to each other
showing the atom positions of both crystals become coincident with each other at every
third plane as indicated by arrows.

154

1:1m11
X
010‘th x, 0%011

 

‘ . .0031
Q x‘r O , . ~ ‘IIT
X”. ----- O --------- K ‘ f. . :7:
o x C ‘__‘-"- HIT,
o x o '
x . x
[101] zone axis

(a)

Figure 6.27 (a) Schematic illustration of the expected electron diffraction pattern from
superimposed overlapping twin model in y-TiAl [99] (b) The FFT diffractogram of the
epitaxial Py showing overlapping twin reflections.

155

150°C) without post annealing. The diffraction patterns show that twin variants appear in
epitaxial Cu and Py layers. consistent with EBSP results on thick films.

The HRTEM images revealed that the Cu buffer layer grown at 150°C appear as
twin variants with two stacking sequences of {111} fcc planes, which is in agreement
with CTEM observations. The three-layer periodic contrast observed in Py can be

explained by overlapped twins along [1 10] direction with {1 1 1} twin planes.

156

CHAPTER VII

SUMMARY AND CONCLUSIONS

In this chapter, the structural studies of polycrystalline and epitaxial GMR
magnetic multilayers and spin-valves will be summarized and the key experimental
results from this research work will be drawn together. Furthermore, some areas for
further research are outlined, including analytical analysis and growth of full epitaxial

GMR multilayers and associated MR evaluation studies.

7.1 SUMMARY AND CONCLUSIONS
In order for the GMR effect in magnetic multilayers and spin-valves to be utilized
for existing and potential technological applications in sensors and magnetic devices, a
complete understanding of the relationships between processing, structure, and behavior
is necessary. Previous works [5,62] have found that the GMR effect depends on structure
at the micro and atomic level, including interfacial or layer roughness, layer quality, and
layer orientation. Therefore, to gain a full understanding of the GMR effect, the initial
motivation of this study was to characterize the interfacial structures of polycrystalline
spin-valves using cross-sectional CTEM and HRTEM.

The first key experimental result was the finding of non—equilibrium bcc
structures in the Cu/Py spin-valves in addition to the equilibrium fcc structures of Cu,
FeMn, and L12 Py layers, as described in detail in Chapter III. In the Cu/Py spin-valves,
structures with different Py thicknesses sputtered at different temperatures were

characterized using CTEM and HRTEM complemented with HRTEM image simulations

157

and FFT analysis techniques. All of the spin valves displayed polycrystalline structures
with columnar grains. The grains grew on close packed planes ({111} planes in the bcc
Nb contacts and {110} planes in the fcc Cu, FeMn, and Py spin-valve layers) taking up a
near Kurdjumov-Sachs {lll}rcc//{l 10}bcc;<l 10>fcc//<111>bcc orientation relationship.
FFT analysis and HRTEM image simulations indicate that in some of the columnar
grains, the Cu, FeMn, and Py regions take up non-equilibrium bcc structures, regardless
of a difference in layer thicknesses and sputtering temperatures. In this non-equilibrium
case, the bcc Cu, F eMn, and Py layers grew on the {110} closest packed planes and were
epitaxial with the Nb contacts.

Chapter IV discussed interesting results in polycrystalline Ag/Py spin-valves. In
one instance, when the Ag/Py spin-valve was grown directly on the Si substrate, the first
Ag layer grew as an epitaxial single crystal although the remaining layers displayed the
expected polycrystalline morphology. HREDS has shown that this single crystal
(epitaxial) layer is likely composed of Ag and In in a 4:1 atomic ratio. The In
contamination may be due to the inadvertent sputtering or evaporation of target solder in
the sputtering system prior to the Nb contact deposition. Indium contamination was only
observed in this one sample. If reproducible, this epitaxial Ag4ln may be useful as a seed
layer for growth of different epitaxial spin valves or other structures on Si substrates. In
all of the other Ag/Py spin-valve samples examined to date, polycrystalline morphologies
have been consistently observed regardless of substrate.

In agreement with polycrystalline Cu/Py and Ag/Py CTEM observations,
Cu/CoZr spin valves are polycrystalline with columnar grains for both exchange biased

spin valve (EBSV) and hybrid spin valve. Some defects. including dislocations, stacking

158

faults, and Moire’ fringes, are observed in both spin valves. However, in comparison to
some previous polycrystalline EBSVs. the columns in the hybrid spin valve appear to
grow straight up with uniform width.

Because of the various non-equilibrium phases that were found in polycrystalline
Cu/Py and Ag/Py spin-valves, and because of the difficulties in imaging interfaces in
polycrystalline samples, the initial objective of this study was modified to focus on the
interfacial structures in epitaxial multilayers to diminish the effect of structural defects
associated with grain boundaries. By minimizing such defects, assessment of the
interfacial structures should be facilitated. Furthermore, elimination of these defects
should allow more fundamental CPP studies to be carried out. However, as the program
progressed, it became clear that growing such samples was not a trivial matter. Thus, the
final objective of this study was to assess structures of epitaxially grown samples using a
sputter deposition technique.

The major achievement of the epitaxial samples study was the successful growth
of a high quality Cu buffer layer on the epitaxial Nb contact, as detailed in Chapter VI.
In this study, the CTEM observations suggest that the quality of good epitaxial growth of
Cu buffer layers depends strongly on deposition temperatures and post annealing. Poor
epitaxial morphology was observed when Cu was grown either at low temperatures with
post annealing or at high temperatures without annealing. High quality epitaxial Cu
buffer layers were obtained when the Cu layer was grown at low temperatures (100°C
and 150°C) without post annealing. The corresponding diffraction patterns show that
twin variants appear in epitaxial Cu and Py layers, consistent with EBSP results from

thick films. The HRTEM images revealed that the Cu buffer layer grown at 150°C

159

appears as twin variants with two stacking sequences of {111} fcc planes, which is in
agreement with CTEM observations. The three-layer periodic contrast observed in Py
can be explained by overlapped twins along [110] direction with {111} twin planes.
Those areas occupied more than 50 percent of total observed Py cross-sectional areas.
These fine twin variants may not in themselves affect CPP measurements. However,
they will be separated by incoherent twin boundaries in the growth direction which might
result in some electron scattering.

With the extensive structural characterization of polycrystalline Cu/Py, Ag/Py,
and Cu/CoZr, and epitaxial GMR magnetic multilayers and spin valves, new insights into
the layer structures have been achieved. In particular, the parameters for growing a high
quality epitaxial Cu and Py layers on Nb contact layers (required for GMR measurements
with CPP geometry), which is the foundation for growing subsequent epitaxial films in a
preferred orientation, have been established. This success confirms the validity of the
sputter deposition technique for growing quality epitaxial thin films at high rates and
presents a greater opportunity for designers and manufactures to realize further
applications of epitaxial GMR devices.

F urtherrnore, the unique combination of this research with Loloee’s study [33] on
magnetic and magneto-transport properties of these materials has led to a more
comprehensive understanding of the relationship between microstructures and magnetic
properties associated with the GMR effect. One of the advantages of using epitaxial
samples over polycrystalline ones is that they afford the capability of studying the
magnetic anisotropy (in particular, magnetocrystalline anisotropy) and determining the

direction of easy magnetic axis in the ferromagnetic layers. In particular for spin valves,

160

knowing the direction of easy axis and magnetocrystalline anisotropy can lead to stronger
pinning between the antiferromagnetic and ferromagnetic layers by application of an
external magnetic field along the direction of easy axis during the pinning process. Also,
knowing the direction of easy axis in both hybrid and exchanged biased epitaxial spin-
valve samples would allow establishing true parallel and antiparallel states. Another
important advantage of epitaxial samples is that the grain boundary effect is greatly
reduced. For polycrystalline samples, even in a small unit area, because of the large
number of grains and corresponding grain boundaries, there will be many magnetic
domains, which will have an effect on electron scattering for CPP measurements.
Whereas for epitaxial samples, only large twins and single crystals are present. CPP
measurements that use the optical-lithography-based sample preparation technique [100]
provide a cross-sectional pillar of 2x2 pm area, which is even smaller than the average
twin variant size. In this case, grain boundaries may not exist at all in the cross-section
areas. Therefore, there would be no electron scattering caused by grain boundaries (as
compared to the polycrystalline case) and the magnetic domains would extend across the
width of the pillars.

These research results should further enhance the knowledge of the mechanisms
of magnetic multilayer structures and establish a theoretical basis for GMR application in

the fields of sensors and information storage devices.

7.2 FUTURE STUDIES

Analytical Analysis:

161

In this research study, a number of electron microscopy techniques were used to
assess the layer structures and growth characteristics of GMR magnetic multilayers and
spin valves. However, no one technique can provide a comprehensive measurement of
all of the important characteristics of multilayer structures. Scanning transmission
electron microscopy (STEM) Z-contrast imaging combination with analytical techniques
as microdiffraction, electron energy loss spectroscopy (EELS), and high resolution
energy dispersive x-ray spectroscopy (HREDS) is a powerful tool to assess the
nanoanalytical results from the multilayer samples. This technique can provide chemical
information of individual components at atomic resolution to determine if there is
interrnixing or interdiffusion occurring at the interfaces. In particular, Z-contrast with
EELS should be able to achieve characterization of the interfaces, allowing assessment of
the degree of interfacial mixing or roughness.

During the period of this study, access was not available to this type of analytical
instrumentation. However, in the near future a new advanced analytical TEM/STEM
with Z-contrast and EELS capabilities will be installed on the MSU campus. This new
capability should assist in further characterization of the chemical nature of these
materials.

Epitaxial Growth:

The successful growth of the epitaxial films by the sputter deposition technique
has opened up remarkable possibilities for making epitaxial GMR devices at higher
growth rates and lower costs. It may also provide unique electronic or magnetic
properties in comparison to existing devices made by polycrystalline materials by

eliminating defects that scatter electrons. This thesis demonstrated the feasibility of

162

growing high quality epitaxial layers. It should not be difficult to grow full epitaxial spin
valves and multilayers from this foundation. Even in the samples that did not have good
epitaxial Cu buffer layers, the remaining layers typically epitaxed well. Furthermore, the
results presented here should provide a foundation for determining the fabrication
variables which are necessary to improve the quality of epitaxial growth of GMR

magnetic multilayers.

163

REFERENCES

[l] H. Sato, P.A. Schroeder, J .M. Slanghter, W.P. Pratt, Jr., and W. Abdul-Razzaq,

Superlattices Microstruct. 4, 45 (1987).
[2] E. Vélu, C. Dupas, D. Renard, J .P. Renard, and J. Seider, Phys. Rev. B37, 668 (1988).
[3] MN. Baibich, J .M. Broto, A. Fert, F. Nguyen Van Dau, F. Petroff, P. Etienne,

G. Creuzet, A. Friederich, and J. Chazelas, Phys. Rev. Lett. 61, 2472 (1988).
[4] G. Binasch, P. Grtinberg, F. Saurenbach, and W. Zinn, Phys. Rev. B39, 4828 (1989).
[5] B. Dieny, J. of Magn. Magn. Mater. 136, 335-359 (1994).
[6] S.S.P. Parkin, IBMJ Res. Dev. 42, no. 1, 3 (1998).
[7] CH. Smith and R. W. Schneider, Proceedings Sensors Expo Boston, 139 (1997).
[8] J.M. Daughton, J. of Magn. Magn. Mater. 192, 334 (1999).
[9] A. Barthelemy, Phys. World. Nov., 35 (1994).
[10] B. Dieny, V.S.Speriosu, S.S.P. Parkin, B.A. Gurney, D.R. Wilhoit, and D. Mauri,

Phys. Rev. B43, 1297 (1991).
[1 1] B.Dieny, V.S. Speriosu, B.A. Gurney, S.S.P. Parkin, D.R. Wilhoit, K.P. Roche,

S. Metin, D.T. Peterson, and S. Nadimi, J. Magn. Magn. Mater. 93, 101 (1991).

[12] A. Barthelemy, A.Fert, M.N. Baibich, P. Etienne, S. Lequien, and R. Cabanol,

J. Appl. Phys. 67, 5908 (1990).
[13] M. Pardavi-Horvath, Magnetic Multilayers, 355 (1994).
[14] PM. Levy, Solid State Physics 47, 367 (1994).
[15] WP Pratt, Jr., S.F. Lee, J .M. Slanghter, R. Loloee, P.A. Schroeder, and J. Bass,

Phys. Rev. Lett. 66, 3060 (1990).

164

[l6] S.S.P. Parkin, N. More, and KP. Roche, Phys. Rev. Lett. 64, 2304 (1990).
[17] DH. Mosca, F. Petroff, A. Fert, P.A. Schroeder. W.P. Pratt, and R. Loloee,
J. Magn. Magn. Mater. 94, L1 (1991).
[18] S.S.P. Parkin, R. Bhadra, and KP. Roche, Phys. Rev. Lett. 66, 2152 (1991).
[19] F. Petroff, A. Barthelemy, D.H. Mosca, D.K. Lottis, A. Fert, P.A.Schroeder,
W.P. Pratt, R. Loloee, and S. Lequien, Phys. Rev. B44, 5355 (1991).
[20] S.S.P. Parkin, App]. Phy. Lett. 60, 512 (1992).
[21] B. Rodmacq, G. Palumbo, and Ph. Gerard, J. Magn. Magn. Mater. 118, L1 1 (1993).
[22] T. Kai, K. Shiiki, J. Magn. Magn. Mater. 195, 537 (1999).
[23] B. Dieny, V.S.Speriosu, S. Metin, S.S.P. Parkin, B.A. Gurney, P. Baumgart, and
DR. Wilhoit, J. App]. Phys. 69, 4774 (1991).
[24] VS. Speriosu, B. Dieny, .l.P. Nozieres, B.A. Gurney, P. Humbert, H. Lefakis,
P. Baumgart, and DR. Wilhoit, Digests INT ERMA G '92, St. Louis, paper CA-04.
[25] B. Dieny, V.S. Speriosu, J.P. Nozieres, B.A. Gurney, A. Vedyayev, and
N. Ryzhanova, in: Magnetism and Structure in Systems of Reduced Dimension,
eds. R. Farrow et al., NA T0 ASI Ser. B: Physics v. 309, (Plenum Press, NY 1993).
[26] B. Dieny, V.S. Speriosu, and S. Metin, Europhys. Lett. 15, 227 (1991).
[27] B. Dieny, P. Humbert, V.S. Speriosu, and BA. Gurney, Phys. Rev. B45, 806 (1992).
[28] B. Dieny, Europhys. Lett. 17, 261 (1992).
[29] VS. Speriosu, B. Dieny, P. Humbert, B.A. Gurney, and H. Lefakis, Phys. Rev. B44,
5358(1991)
[30] B. Dieny, V.S. Speriosu, .l.P. Nozieres, and BA. Gurney, App]. Phys. Lett. 61, 211 l

(1991).

165

[31] TC. Huang, J .P. Nozieres, V.S. Speriosu, B.A. Gurney, and H. Lefakis, Appl. Phys.
Lett. 62, 1478 (1993).

[32] S.S.P. Parkin, Phys. Rev. Lett. 71, 1641 (1993).

[33] R. Loloee, Ph.D. Dissertation, Michigan State University (2000).

[34] WP Pratt Jr., S.D. Steenwyk, S.Y. Hsu, W.-C. Chiang, A.C. Schaefer, R. Loloee,
and J. Bass, IEEE Transactions on Magn. 33, 3505 (1997).

[35] Y. Bruynseraede, K. Temst, R. Schad, C.D. Potter, P. Belien, G. Verbanck,
G. Gladyszewski, J. Bamas, M. Baert, V.V. Metlushko, M.J. Van Bael,
V.V. Moshchalkov, Thin Solid Film 275, l (1996).

[36] VS. Speriosu, J .P. Nozieres, B.A. Gurney, B. Dieny. T.C. Huang. and H. Lefakis,
Phys. Rev. B47, 11579 (1993).

[37] MC. Cyrille, S. Kim, M.E. Gomez, J. Santamaria, Kannan M. Krishnan, and
Ivan K. Schuller, Phys. Rev. B62, 3361 (2000).

[38] MC. Cyrille, S. Kim, M.E. Gomez, J. Santamaria, C. Leighton, Kannan M.
Krishnan, and Ivan K. Schuller, Phys. Rev. B62, 15079 (2000).

[39] F. Petroff, A. Barthelemy, A. F ert. P. Etienne, and S. Lequien, J. Magn. Magn.
Mater. 93, 95 (1991).

[40] EB. Fullerton, D.M. Kelly, J. Guimpel, I.K. Schuller, and Y. Bruynseraede, Phys.
Rev. Lett. 68, 859 (1992).

[41] Y. Obi, K. Takanashi, Y. Mitami, N. Tsuda. and H. F ujimori, J. Magn. Magn.
Mater. 104-107, 1747 (1992).

[42] S.S.P. Parkin, Appl. Phys. Lett. 58(23), 2710 (1991)

[43] F. Petroff, A. Barthelemy, A. Fert, P. Etienne, and S. Lequien, J. Magn. Magn.

166

Mater. 104-107, 1712 (1992).

[44] S. Pizzini, F. Baudelet, D. Chandesris, A. Fontaine, H. Magnan, J .M. George,
F. Petroff, A. Barthelemy, A. Fert, R. Loloee and PA. Schroeder, Phys. Rev. B46,
1253(1992)

[45] R. Nakatani, T. Dei, and Y. Sugita, Jpn. J. Appl. Phys. 31, L1417 (1992).

[46] R. Nakatani, T. Dei, and Y. Sugita, MMM’92, EQ04.

[47] R. Jungblut, R. Coehoom, M.T. Johnson, J. aan de Stegge, and A. Reinders, .1. App].
Phys. 75(10), 6659 (1994).

[48] T. Shinjo, in: Magnetism and Structure in Systems of Reduced Dimension, eds. R.
Farrow et al., NA T0 AS] Ser. B: Physics 309, 323 (Plenum Press, NY 1993).

[49] A. Vedyayev, B. Dieny, N. Ryzhanova, J .B. Genin, and C. Cowache, Europhys. Lett.
25, 465 (1994).

[50] W.F. Egelhoff Jr., T. Ha, R.D.K. Misra, C.J. Powell, M.D. Stiles, RD. McMichael,
C.-L. Lin, J .M. Sivertsen, J .11. Judy, J. App]. Phys. 79, 282 (1996).

[51] ME. Gomez, J. Santamaria, M.C. Cyrille, E.C. Nelson, Kannan M. Krishnan, and
Ivan K. Schuller, (submitted to Phys. Rev. B). i

[52] B. Heinrich, M. From, J .F.Cochran, L.X. Liao, Z. Celinski, C.M. Schneider, and
K. Myrtle, Mat. Res. Soc. Symp. Proc. 313, 119 (1993).

[53] Z. Celinski and B. Heinrich, J Magn. Magn. Mater. 99, L25 (1991).

[54] R. Loloee, M.A. Crimp, W. Zhu, and WP. Pratt Jr., Mat. Res. Soc. Symp. Proc. 528,
203 (1998).

[55] R.D.K. Misra, T. Ha, Y. Kadmon, C. J. Powell, M.D. Stiles, R.D. McMichael, and

W.F. Egelhoff, Jr., Mat. Res. Soc. Symp. Proc. 384, 373 (1995).

167

[56] M. Chladek. V. Valvoda, C. Domer. and W. Ernst, J. Magn. Magn. Mater. 172, 209
(1997)

[57] J .P. Deville, A. Barbier, C. Boeglin, and B. Carriere, Mat. Res. Soc. Symp. Proc.
313,519(1993)

[58] H. Geng, J .W. Heckman, R. Loloee, W.P. Pratt, Jr., J. Bass, and MA. Crimp, Mat.
Sci. and Eng. B86, 245 (2001).

[59] K. Sumiyama, T. Hihara, S.A. Makhlouf, K. Wakoh, M. Sakurai, Y. Xu, T.J. Konno,
S.Yamamuro, and K. Suzuki, Mat. Sci. and Eng. A217/218, 340 (1996).

[60] FM. Ross, K.M. Krishnan, N. Thangaraj, R.F.C. Farrow, R.F. Marks, A. Cebollada,
S.S.P. Parkin, M.F. Toney, M. Huffman, C.A. Paz De Araujo, L.D. McMillan,
J. Cuchiaro, M.C. Scott, C. Echer, F. Ponce, M.A. O’Keefe, and EC. Nelson, MRS
.Buflefin,l7(1996)

[61] P. Galtier and T. Valet, Mat. Res. Soc. Symp. Proc. 475, 131 (1997).

[62] P. Bayle—Guillemaud, A.K. Petford-Long, T.C. Anthony, J .A. Brug, IEEE Trans. 0n
Magn. 32, No. 5, 4627 (1996).

[63] H. Geng, J .W. Heckman, W.F. Pratt, Jr., J. Bass, F .J . Espinosa, S.D. Conradson,
D. Lederrnan, and MA. Crimp, J. App]. Phys. 86, 4166 (1999).

[64] SD. Steenwyk, S.Y. Hsu, R. Loloee, J. Bass, and WP. Pratt, Jr., J. Appl. Phys. 81
(8), 4011 (1997).

[65] A.Romano, J. VanHellemont, H. Bender, and J .R. Morante, Ultramicroscopy, 31,
183(1989)

[66] A.Romano, J. VanHellemont, and H. Bender, Mat. Res. Soc. Symp. Proc., 199, 167

(1990).

168

[67] F. Shaapur and M. Park, Mat. Res. Soc. Symp. Proc., 199, 177 (1990).

[68] D.A. Howell, Master Thesis, Michigan State University (1995).

[69] DB. Williams and CB. Carter, Transmission Electron Microscopy: A textbook for
materials science, Plenum Press, NY (1996).

[70] W.S. Rasband and BS. Bright, Microbeam Anal. Soc. J. 4, 137 (1995).

[71] P. Stadelman, Ultramicroscopy 21, 131 (1987).

[72] K.L. Merkle, Ultramicroscopy 37, 130 (1991).

[73] DJ. Smith, W.J. DeRuijter, M.R. McCartney, and J .K. Weiss, Ultramicroscopy 52,
591(1993)

[74] Z. Nishiyama, Sci. Rept. Tohoku Univ. 23, 368 (1934).

[75] G. Wasserman, Arch. Eisenhuttenwesen 16, 647 (1933).

[76] G.V. Kurdjumov and G. Sachs, Z. Physik 64, 325 (1939).

[77] MG. Hall, H.I. Aaronson and KR. Kinsman, Surface Science 31, 257 (1972).

[78] J.M. Rigsbee and HI. Aaronson, Acta Metal]. 27, 351 (1979).

[79] DJ. Smith, W.O. Saxton, M.A. O’Keefe, G.J. Wood and W.M. Stobbs.,
Ultramicrosocpy 11, 263 (1983).

[80] AT. Paxton, M. Methfessel and HM. Potatoglou, Phys. Rev. B4l(12), 8127 (1990).

[81] T.E. Mitchell, Y. C. Lu, A. J. Griffen Jr., M. Nastasi, and H. Kung, J. Am. Ceram.
Soc. 80(7), 1673 (1997).

[82] SR. Goodman, S.S. Brenner, and J .R. Low, Jr., Met. Trans. 4, 2363 (1973).

[83] Heinrich, et al., In science and technology of nanostructured Materials, G.C .
Hadjipanayis and GA. Prinz. eds. Plenum, NY. pl 5 (1991).

[84] H. Kung, Y-C. Lu, A.J. Griffen, Jr., M. Nastasi, T.E. Mitchell, and J .D. Embury,

169

App]. Phys. Lett. 71 (15) 2103 (1997).

[85] PM. Marcus, V.L. Moruzzi, Z.Q. Wang, Y.S. Li, and F. Jona, Mat. Res. Soc. Symp.
Proc., 83, 21 (1987).

[86] C. Schmidt, F. Ernst, M.W. F innis, and V. Vitek, Phys, Rev. Lett, 75, 2160 (1995)

[87] F. K. LeGoues, M. Liehre, M. Renier, W. Krakow, Phil. Mag. B. 57 (2), 179 (1988).

[88] J. Kudmovsky, V. Drchal, I. Turek, C. Blaas, P. Weinberger, and P. Bruno,
J. Mater. 52 (7), 29 (2000).

[89] M. R. Baren, in: r. B. Massalski (Ed), Binary/llloy Phase Diagrams: 2"(1 ed., ASM
International, Materials Park, OH, pp. 47-48 (1990).

[90] J .W. Heckman, Master Thesis, Michigan State University (2001).

[91] Z. Nishiyama, Sci. Rept. Tohoku Univ, 23, 368 (1934).

[92] R. Loloee, W.P. Pratt, Jr., and MA. Crimp, Phil. Mag. A, 81 (2), 261 (2001).

[93] A.E.M. De Veirman, F.J.G. Hakkens, and AG. Dirks, Ultramicroscopy, 51, 306
(1993)

[94] H. Bender, A. De Veirman, J. Van Landuyt, and S. Amelinckx, App]. Phys. A, 39, 83
(1986)

[95] J .L. Hutchison, G.R. Booker, and MS. Abrahams, Inst. Phys. Conf. Ser., 60, 139
(1981)

[96] J. Vanhellemont, J. Van Landuyt, C. Claeys, G. Declerck, H.H. Babbar,
D.N. Nichols, and C. Anagnostopoulos, Inst. Phys. Conf. Ser., 76, 195 (1985).

[97] H. Hashimoto, H. Endoh, M. Tomita, N. Ajika, M. Kuwalara, Y. Hata,
Y. Tubokawa, T. Honda, Y. Arada, S. Sakurai, T. Etoh, and Y. Yokota, J. Electron.

Microsc. Tech, 3, 5 (1986).

170

[98] X.J. Wu, F.H. Li, and H. Hashimoto, Phil. Mag. B, 63 (4), 931 (1991).

[99] E. Abe, S. Kajiwara, T. Kumagai, and M. Nakamura, Phil. Mag. A, 75 (4), 975
(1997)

[100] R.D. Slater, J.A. Caballero, R. Loloee, and WP. Pratt, Jr., J. App]. Phys. 90 (10),

5242(2001)

171

111:111111111171