.5 ~
.1 1|:
.3 17'...“

{z
I S).
t.

z 1... .\..
.1. a... .x t
2:;

t: .
‘ a“?

1! .1
h:

kart...
1».

£5...

 

 

 

\

, Vr .
.Hfié..i 3 : . . ¢.h.....‘.
3 r. in . IL...” . ‘ .‘4 . 3..

.. ‘

 

Tryffrt

5.4“

C)

Co
Q.)

This is to certify that the
thesis entitled

Quantitative and Analytical High-Resolution
Transmission Electron Microscopy Study

of Epitaxial Co/Cu GMR Multilayers

presented by

John W. Heckman, Jr.

has been accepted towards fulfillment
of the requirements for

Master's degree in Materials Science

and Engineering

Major professor

Datefl/oé Z: 7002

0-7 639

MS U is an Affirmative Action/Equal Opportunity Institution

 

LIBRARY
Michigan State
University

 

 

 

 

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 c:/ClRC/DateDue.p65vp. 15

QUANTITATIVE AND ANALYTICAL HIGH-RESOLUTION TRANSMISSION
ELECTRON MICROSCOPY STUDY OF EPITAXIAL C0/CU GMR MULTILAYERS

By

John W. Heckman, Jr.

ATHESIS

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

MASTER OF SCIENCE

Department of Materials Science and Mechanics
2002

ABSTRACT

QUANTITATIVE AND ANALYTICAL HIGH-RESOLUTION TRANSMISSION
ELECTRON MICROSCOPY STUDY OF EPITAXIAL Co/Cu GMR MULTILAYERS

By

John W. Heckman, Jr.

Co/Cu multilayers were grown on Nb layers by sputter deposition as part of
an effort to develop a system for producing epitaxial multilayer (ML) devices for
perpendicular to plane, giant magnetoresistance measurements (CPP-GMR).
Epitaxial MLs and a polycrystalline ML were analyzed using conventional,
analytical and high-resolution transmission electron microscopy. The Nb base
layer observed was a single defect-dense crystal on the sapphire substrate.
Apparently amorphous interfaces and non—equilibrium lattice structures were
seen in one epitaxial ML. This sample also showed 3 overall Cu mass deficit.
MLs In the other epitaxial specimens were broken into "island-like" regions
associated with Cu seed-layer discontinuity. These epitaxial MLs had strata
showing a crystal defect density like the Nb base and Co/Cu interlayer contrast.
The polycrystalline sample was small grained and textured in the [111] growth
direction. Co/Cu interlayer contrast in the polycrystalline ML and the epitaxial

island MLs arises from stacking changes in Co.

ACKNOWLEDGMENTS

I firstly thank Dr. M.A Crimp, major professor, for inviting me to join his group
and to pursue this program. Thanks also to the others on my committee Dr. D.
Mason, and J. Bass. I also thank Drs. R. Loloee, and WP. Pratt, Jr., for
providing me with the sample multilayers and the GMR data. A special thanks
also goes to Hong Geng, fellow graduate student, for essential help in preparing
specimens of thin metal films on sapphire.

I also thank Drs. J.F. Mansfield and C. Wauchope of the University of
Michigan EMAL for access to the JEOL 4000EX and 201OF (supported by grant
number NSF DMR 987177).

This work has been supported by the Michigan State University Center for
Fundamental Materials Research and by the National Science Foundation,
through the Materials Research Science and Engineering Center MRSEC grant

number NSF DMR 9809688.

iii

TABLE OF CONTENTS

LIST OF TABLES ................................................................................. vii
LIST OF FIGURES ............................................................................... viii
CHAPTER 1 INTRODUCTION ................................................................ 1
1.1 Overview ................................................................................. 1
1.2 Giant Magnetoresistance .............................................................. 2
1.3 Oscillatory Exchange Coupling and GMR ML devices ........................ 4
1.4 The Co/Cu ML system ................................................................. 6
1.5 Fabrication Considerations in Co/Cu GMR MLs ................................. 8
1.6 Epitaxial Co/Cu GMR MLs .......................................................... 13
1.7 Characterization of Magnetic MLs ................................................. 15
1.8 Relevance of Previous Work to This Study ..................................... 19
CHAPTER 2 EXPERIMENTAL PROCEDURES AND‘MATERIALS ................. 21
2.1 Magnetron Sputtering and ML Device Fabrication ............................ 21
2.2 Co/Cu GMR ML Devices ............................................................. 21
2.3 CIP GMMR Measurements .......................................................... 26
2.4 TEM Sample Preparation ............................................................ 27
2.5 TEM Experimental Methods ......................................................... 30
2.6 TEM Image Analysis .................................................................. 33
CHAPTER 3 EXPERIMENTAL RESULTS ................................................ 37
3.1 Sample 20-1 (epitaxial) ............................................................... 37'
3.2 Sample 20-2 (polycrystalline) ...................................................... 56
3.3 Sample 20-3 (epitaxial) ............................................................... 74

iv

3.4 Sample 20-4 (epitaxial) ............................................................ 87

3.5 Samples 91-3 and 19-4 (epitaxial bilayers) .................................... 100
CHAPTER 4 DISCUSSION .................................................................. 116
4.1 Growth of the Nb Base layer ....................................................... 116
4.2 Cu Buffer Layers ........................................................................ 119
4.3 Co/Cu Multilayers ..................................................................... 123
CHAPTER 5. CONCLUSIONS AND RECOMMENDATIONS ...................... 128
LIST OF REFERENCES ...................................................................... 132

LIST OF TABLES
Table 1.1 Physical Characteristics of the multilayer materials ........................ 16

Table 2.1 Co/Cu bilayer and GMR multilayer fabrication temperatures ............ 24

vi

LIST OF FIGURES

Figure 1.1. Schematics of the GMR effect and mechanism. A) The
magnetoresistance of 3 Fe/Cr superlattices at 4.2K. The GMR effect was greater
for thinner Cr (nonmagnetic) layers. The authors current along the same [110]
axis as the layers were grown (current in plane or CIP) [2]. B) A schematic of
the GMR mechanism. Electron trajectories are represented by the

straight segments, scattering by an abrupt change in direction [1] .................. 3

Figure 1.2. Oscillatory exchange coupling and its effect on ferromangetic layer
orientation. A) The effect of NM layer thickness on GMR [23]. B) The effect of
the coupling on alternating FM and NM layers, with and without an external
magnetic field ....................................................................................... 4

Figure 1.3. The relationship between number of bilayers and GMR observed
with the current in the multilayer plane (redrawn from [14]) ........................... 9

Figure 1.4. Proposed grain structures of Co/Cu MLs. A) Grown directly on a Fe
seed layer with a thick Cu capping layer. B) Grown on a 35 nm Cu base layer
From [30] .......................................................................................... 10

Figure 1.5. The immiscibility of copper and cobalt. A) The binary phase diagram
for Cu and Co [36]. B) The calculated free energy of mixing for Co and Cu [37].

........................................................................................................ 12
Figure 2.1 Schematic cross-section diagrams of the test specimens. A) The
Co/Cu GMR ML pattern. B) A preliminary Co/Cu bilayer sputtered on a Nb

substrate ........................................................................................... 22

Figure 2.2 Schematics of ML stacking. A) Fcc Cu over bcc Nb in the
Kurdjumow-Sachs orientation. B) The non-relaxed incoherent spacing between
Co and Cu in the fcc crystal structure ...................................................... 25

Figure 2.3 The general CIP MR response in Cos nm lCue nm [13] .................... 26

Figure 2.4. Schematic of the main steps in specimen cross-section preparation.
........................................................................................................ 29

Figure 2.5 Objective apertures used in TEM imaging in the JEOL 4000EX. A)
The CTEM aperture, a 7.2 nm‘1 diameter. B) The Scherzer aperture, 20.8 nm“.
The diffraction pattern is Si, B=<110> showing 36 transmitted beams (including
the direct spot) .................................................................................... 30

vii

Figure 2.6 The Sapphire/Nb boundary. The transform of this area shows that
there is no detectable difference in the Nb {110} and sapphire {1 0T0} spacings.

Thus Nb lattice images were used to calibrate the HRTEM images ................ 35

Figure 2.7 Resolvable lattice planes and their interplaner angles for equilibrium ,
structures of the elements used in the ML fabrication .................................. 36

Figure 3.1. CTEM images of sample 20-1. A) Brightfield minimum contrast
focus image, normal to the sapphire (1 0T0). B) (200) Darkfield on the same
axis. C) Over-focused brightfield image showing Fresnel contrast parallel to the
layer interfaces ................................................................................... 37

Figure 3.2. Multibeam (HRTEM) approximately Scherzer focus images and
average density plots along MLs. A) Sample 20-1. B) Sample 20-2. C.
Sample 20-4 ....................................................................................... 37

Figure 3.3. HRTEM image of a selected region of sample 20-1. The growth
direction is about 30° off the depicted vertical. The three regions selected for
lattice image analysis are indicated ......................................................... 41

Figure 3.4 An HRTEM image of the upper layers in sample 20-1. This figure is
oriented with the growth direction running vertically. The top dark region is the
‘ capping Nb layer. The insets are areas that were tested by F FT analysis ....... 42

Figure 3.5 Selected HRTEM images A-F and their FFTs from the upper layers of
sample 20-1 ....................................................................................... 43 .

Figure 3.6 HRTEM image of the middle layers in sample 20-1. This figure is
oriented with the growth direction running vertically ..................................... 46

Figure 3.7. Selected HRTEM images A-F and their FFTs from the middle layers
of sample 20-1 .................................................................................... 48

Figure 3.8. HRTEM image of the bottom layers in sample 20-1. This figure is
oriented with the growth direction running vertically. The lowermost area, F, is
the bottom Nb layer ............................................................................. 49

Figure 3.9. Selected HRTEM images A-F and their FFTs from the bottom layers
of sample 20-1 .................................................................................... 50

Figure 3.10. EDS spectra from a cross-section the 20-1 ML structure. A) The
spectrum formed by averaging three whole-ML spectra. B) DTSA simulation of a
Co/Cu sample with the same volume fractions as a circular sample of the 20-1
ML motif ............................................................................................. 54

viii

Figure 3.11. EDS spectra from a cross-section the 20-1 ML structure. A) The
spectrum formed by averaging 18 Cu layer spectra. B) The spectrum formed by
averaging 18 Co layer spectra ................................................................ 55

Figure 3.12. CTEM images of sample 20-2. A) Single beam brightfield image of
a cross section of the 20-2 ML. B) The darkfield image created using the (OTC)
Nb spot. C) A single-beam brightfield image taken 11 pm below minimum
contrast ............................................................................................. 57

Figure 3.13. Low magnification HRTEM image of the cross-sectioned 20-2
sample. The ML area (ML) and lattice analysis area (inset) are shown ........... 59

Figure 3.14. HRTEM of the ML 20-2. Four regions were selected for lattice
analysis of the metal layers. Based on crystal diffraction contrast there appears to
be 5 rather than 6 bilayers ..................................................................... 60

Figure 3.15. HRTEM of the upper region of sample 20-2. The image
corresponds to the inset (1) from figure 3.14. The upper half of the region is the
Nb capping layer ................................................................................. 62

Figure 3.16. HRTEM of the middle regionof sample 20-2. The image
corresponds to the inset (2) from figure 3.14 .............................................. 63

Figure 3.17. High magnification HRTEM areas and their corresponding FFT
transforms from Figures 3.15 (A-C) and 3.16 (D-F) ..................................... 64

Figure 3.18. HRTEM of the first Co and adjacent Cu layers from sample 20-2.
The area corresponds to the inset (3) from figure 3.14 ................................. 67

Figure 3.19. High magnification HRTEM areas and their corresponding FFT
transforms from Figures 3.18 (A-C) and 3.16 (D-F) ..................................... 68

Figure 3.20. HRTEM of the Cu buffer layer and adjacent Nb base layer from
sample 202 The area corresponds to the inset (4) from figure 3.14 .............. 70

Figure 3.21. High magnification HRTEM areas and their corresponding FFT
transforms from Figures 3.19 (AC) and 3.16 (D-F) ..................................... 71

Figure 3.22. EDS spectrum from a cross-section of the 20-2 ML structure ...... 72

Figure 3.23. CTEM images of sample 20-3. A) Single beam brightfield image of
a cross section of the 20-3 ML. B) The darkfield image obtained with the Nb
(0T0) reflection. C. A single-beam brightfield image focused 11 um below the
minimum contrast setting. Arrows 1-4 highlight a prominently layered region, an
anomalously thick buffer region, a region apparently sans buffer layer and an
apparent hole, respectively ................................................................... 74

ix

Figure 3.24. CTEM image of the cross-sectioned 20—3 sample. A) An area
within an island. B) The area around an apparent hole .............................. 77

Figure 3.25. HRTEM image of an island area of 20-3 selected for structural
analysis. Areas A-F are representative areas of the upper and lower Nb layers
and the first and last Co/Cu bilayers ........................................................ 78

Figure 3.26. High magnification HRTEM areas and their corresponding FFTs
from areas A-F in Figure 3.25 ................................................................ 79

Figure 3.27. High magnification HRTEM areas and their corresponding FFTs
from areas G-J in Figure 3.25 ................................................................. 81

Figure 3.28. A HRTEM image of a Cu buffer layer hole and surrounding area
from sample 20-3 ................................................................................ 83

Figure 3.29. High magnification HRTEM images and their corresponding FFTs.
from areas around the hole shown in Figure 3.28 ....................................... 84

Figure 3.30. CTEM images of sample 20-4. A) Single beam brightfield image of
a cross section of the 20-4 ML. B) The darkfield image obtained with the Nb
(0T0) reflection. C) A single-beam brightfield image focused 11pm below the
minimum contrast setting. Arrows 1-3 highlight: A region apparently sans buffer
layer, an apparent hole, and a prominently layered region, respectively .......... 88

Figure 3.31. Low magnification HRTEM image of the cross-sectioned 204
sample. The area within a typical island chosen for structural imaging is outlined.

........................................................................................................ 90
Figure 3.32. HRTEM image of an island area of 20-4 selected for structural
analysis. Areas AM are representative areas throughout Co/Cu bilayers ....... 91
Figure 3.33. High magnification HRTEM areas and their corresponding FFTs
from areas A-F in Figure 3.32 ................................................................. 92
Figure 3.34. High magnification HRTEM areas and their corresponding FFTs
from areas G-M in Figure 3.32 ................................................................ 95
Figure 3.35. An EDS spectrum collected from the whole ML cross-section of
sample 20-4. The collection time and beam parameters were essentially the
same as those used previously ............................................................... 96

Figure 3.36. An EDS spectrum from the top amorphous layer covering the ML 96

Figure 3.37. EDS spectra from adjoining Cu and Co layers near the top of the
ML. A) ACu layer and B) 3 Co layer ...................................................... 98

Figure 3.38. An EDS spectrum from the Cu buffer layer of sample 20-4 .......... 99

Figure 3.39. Brightfield CTEM cross-section images of samples 19-3 and 19-4.
A) Sample 19-3. B) Sample 19-4. C. An island structure from sample 194
showing three regions analyzed via HRTEM ............................................. 102

Figure 3.40. HRTEM overview of the "inset 1" area of Figure 3.39C. Six
representative areas were selected for lattice image analysis ....................... 104

Figure 3.41. High magnification HRTEM areas and their corresponding FFTs
from areas A-F in Figure 3.40 ................................................................ 105

Figure 3.42. HRTEM overview of the "inset 2" area of Figure 3.39C. The three
inset areas and their F FTs are also shown ............................................... 108

Figure 3.43. HRTEM overview of the "inset 3" area of Figure 3.39C. Six
representative areas were selected for lattice image analysis ....................... 110

Figure 3.44. High magnification HRTEM areas and their corresponding FFTs
from areas A-F in Figure 3. 43 ................................................................. 111

Figure 3. 45. HREDS spectrum from the Co layer on an island from sample 19-4
....................................................................................................... 113

Figure 3.46. An HREDS spectrum from the thicker (A) and thin (B) middle areas
of the Cu layer on a island from sample 19-4 ........................................... 113

Figure 3.47. An HREDS spectrum from the "gap" area in an island from sample
19-4 ................................................................................................. 114

Figure 3.48. An HREDS spectrum from the Culle interface at the base of an
island from sample 19-4 ........................................................................ 114

Figure 4.1. Binary phase diagrams. A) Co-Nb. B) Cu-Nb. Below 800°C there
should be no solubility of Nb in Cu .......................................................... 120

Figure 4.2 Atomic force microscope images of 20 nm Cu layers grown on the

surface of 250 nm epitaxial Nb layers. A) Deposited at 350°C. B) Deposited at
150°C and annealed at 350°C for 10 min. From [47] .................................. 122

xi

CHAPTER 1 INTRODUCTION
1.1 Overview:

In the last decade great progress has been made in the application of thin-film
multilayer technology. Nowhere has this development been more apparent than
in the area of thin-film giant magnetoresistance (GMR) devices. The operation of
these devices is based on the property that certain layered structures, containing
a ferromagnetic/nonmagnetic/ferromagnetic layer motif, display a great change in
electrical resistance upon the application of an external magnetic field. Magnetic
random access memory (MRAM), GMR read heads, and magnetic position
sensors are commercial examples of the application of this technology.
Improvement and development of this technology will depend on the refinement
of the knowledge of the effects of the physical characteristics of the magnetic and
non-magnetic materials used on the resultant magnetic properties. For example,
an early discovery towards the facile production of these devices was that GMR
devices readily be made in high throughput sputtering devices. Already a wealth
of information exists describing the effects of elemental choice, crystal orientation
and size within the films, interfacial roughness and intermixing on the GMR
response. Developing fabrication protocols for single crystal epitaxial structures
is currently an important area of investigation. In this thesis, the results of
experiments aimed at the development of such protocols are evaluated by
transmission electron microscopy (TEM) techniques. For preliminary work, a well
characterized and technologically applicable ColCu multilayer was chosen.

Examples were sputtered under varying temperatures and annealing treatments

to develop epitaxial model systems for GMR studies. The focus of this
investigation was the crystallographic and compositional characterization of
these examples, at the atomic level, and their comparison to similar
polycrystalline preparations.

1.2 Giant Magnetoresistance:

One of the factors influencing electrical resistance in metals is the magnetic
condition of the medium. Giant magnetoresistance (GMR) refers to an
anomalously high resistance to electrical current passage through some
materials, as a consequence of their magnetic order [1]. This resistance can be

greatly reduced through the application of an external magnetic field (Figure 1a).

The GMR effect was first observed by the independent groups of A. Fert [2], in
France, and P. Gruenberg [3], in Germany. These groups noticed a drastic
reduction of electrical resistance in the plane of ultra-thin Fe/Cr multilayers (MLs)
in the presence of an externally applied magnetic field. This effect was far
greater than the angularly dependent change seen in a single magnetic layer
(anisotropic magnetoresistance effect). Under the application of the external
field, the magnetic moments of the thin ferromagnetic layers change from their
minimum energy configuration, having their magnetic moments aligned
antiparallel to each other (i.e. antiferromagnetically coupled), to the parallel
condition. The MR effect stems from the spin dependence of the electron
conduction in the ferromagnetic layers. In the two-current model for conduction

in ferromagnets, two current channels + and - (from spins of +1/2 and -112 on

an absolute axis) are considered. These correspond to T and I, the majority and

minority spin directions in the ferromagnetic layers,

 

  
 

(Fe 303:0 izilas

IF. 303/092)...

 

 

f
—.

as
.4.
1" l
o

“s

L l A

V

a so a
Magnetic field lkGl

 

 

 

 

MNMM NMM

ulli-

Parallel configuration Antiparallel configuration

 

 

 

Figure 1.1. Schematics of the GMR effect and mechanism. A) The
magnetoresistance of 3 Fe/Cr superlattices at 4.2K. The GMR effect was greater
for thinner Cr (nonmagnetic) layers. The authors applied the current long the
same [110] axis as the layers were grown (current in plane or CIP) [2]. B) A
schematic of the GMR mechanism. Electron trajectories are represented by the
straight segments, scattering by an abrupt change in direction [1].

The resistively in the ferromagnetic material can be expressed as a function of

the resitivities of each electron channel:

 

p: prI+pT~I(pT+pI)

pr+pl+4pri “I

The orientations of the magnetism in the ferromagnetic layers either promotes or

inhibits the mobility of the electron through the layer stack (Figure 1 b).

The final resistance in the parallel state ML rID can be written as:

 

(2)

 

r = * ‘
A. 4 <3)
The percent change in resistance between the two states is the GMR:

GMR=5£rix100 (4)

re
This is the form generally reported in the literature [1].

1.3 Oscillatory Exchange Coupling and GMR in multilayer devices:

In MLs, consisting of alternating layers of non-magnetic (NM) and
ferromagnetic (FM) materials, the degree of exchange coupling in many systems
oscillates with variation in NM layer thickness, (Figure 2a), rather then decaying
monotonically with NM layer thickness [3]. The exchange coupling (the ability of
one magnetic domain to affect the alignment of another) in these layers promotes

a concomitant thickness dependent GMR effect [4, 5].

    

 

I'v‘r'vvvvlvvvv'vj

.T:4.2K
OT=300K1

4
11.-

-i

IUIAA‘

MR Ratio (%)

I‘LL.

 

L-llll“

 

 

 

 

 

 

 

 

 

 

Figure 1.2. Oscillatory exchange coupling and its effect on ferromagnetic layer
orientation. A) The effect of NM layer thickness on GMR [23]. B) The effect of
the coupling on alternating FM and NM layers, with and without an external
magnetic field.

Thus, the magnetizations of alternating ferromagnetic layers will exist in the
maximum antiferromagnetic (AF) configuration only if the interleaving non-
magnetic layers are of certain thicknesses. The magnitude of the
antiferromagnetic coupling is also dependent on (among other attributes) the
thickness of these NM layers and a decreasing effect can be seen as a damped

oscillating response in the GMR of the ML. This effect is seen in FM/NM MLs for

a number of elemental systems. Although the GMR effect appears to be the

highest for relatively thin NM layers (ca. 1 nm), the GMR effect is evident even in
Co/Cu MLs having NM layer thicknesses as great as 50nm [6].

\Nith the ability to switch the magnetic order with an externally applied
field, materials exhibiting the GMR effect have progressed from the realm of
experimental novelties to engineering materials [e.g.7,8]. The sensitivity of a
GMR device to an external field can be improved in at least two ways. Highly
sensitive structures, known as exchange biased spin valves, can be made by
pinning the orientation of one ferromagnetic layer with an antiferromagnetic
material while leaving another to rotate in the applied field. These are currently
being used in sensor applications such as read-head devices for high-density
disk drives and non—volatile, magnetic memory devices (MRAM) [8]. Second, by
using a soft-magnetic adjacent layer (e.g. VitrovacTM 6025, a C0 based metallic
glass) the low field sensitivity of a simple ML, such as the Co/Cu motif, can be
greatly enhanced [9]. The various applications of thin-film GMR devices for
engineering materials share the common requirement of a precisely fabricated

ML metallic structure.
1.4 The Co/Cu multilayer system:

Early in the exploration of the GMR effect, Parkin, et al. [10] tested a number of
ML superlattices for oscillatory exchange coupling that promotes the GMR effect.
The system that possessed the highest GMR values was a configuration of
alternating Co (ferromagnetic) and Cu (non-magnetic) layers [11]. This system
has yielded the highest room-temperature magnetoresistance (65% at 295 K and

120% at 4.2 K) of any such superlattice and is still of research interest [12,13].

The Co/Cu system does have a high saturation field which limits its utility in
certain applications. The incremental change in resistance in these systems
decreases with increasing filed strength. However, the change in resistivity (and
therefore GMR) continues to increase for fields two orders of magnitude greater
than the magnetic saturation field and the GMR effect increase towards a limit

with bilayer number (Figure 1.3).

 

 

 

 

. . . . T
30 l- ./ T
l 0 ~
A Cu10AI(Co1OA/Cu10A)
be 20 - T=4.2 K '
Z .
E. . .
<
10 - -
I . "
o .r/ . I 1 I n
0 50 100 150

n (Number of Bilayers)

Figure 1.3. The relationship between number of bilayers and GMR observed
with the current in the multilayer plane (redrawn from [14])

This may be due to a degree of Co disorder at the higher layer number interfaces
that contributes significantly to the electron scattering but not much to the
magnetism at higher fields [14]. Copper/cobalt MLs have been prepared on a
number of different substrate types and by several methods. The first Co/Cu

magnetic ML examples were deposited on single crystal copper via electron-

beam evaporation [15,16], while others have used a similar evaporation
techniques to build the MLs on oxidized silicon [17] and metal-sliced substrates
[18]. Another approach has been molecular bean epitaxy (MBE) [e.g.19-22].
However, probably due to the faster growth rate and economics, the vast majority
of preparations have been made with various sputtering systems [e.g. 23-27].
More recently, pulsed electrodeposition (plating) techniques have been used

[28].
1.5 Fabrication considerations in Co/Cu GMR multilayers:

To be useful as a model or as an engineering material, the behavior and
stability of 3 ML system during fabrication and in subsequent application needs to
be known. Aside from representing materials with the appropriate ferromagnetic
and non-magnetic properties, the more fundamental metallurgical characteristics
of the Co/Cu system and its substrates are important. Obviously, layer thickness
is important as evidenced by the oscillatory exchange coupling. As can be seen
in Figure 1.2, the AF period is about 1.0 nm and an additional i 0.5 nm in the NM

layer thickness could destroy the GMR effect. As mentioned above, the number

of bilayers also is an important characteristic. For 3 Co 1nm/Cu 1m, bilayer
system, the GMR effect appeared to grow to a finite maximum with the number of
bilayers (Figure1.3). Thus, the ability to confirm layer geometry and quality is

important.

A second consideration in the fabrication of MLs is the effect of
crystallographic orientation and grain size on the GMR effect. Earlier work by

Cebollada, et al. [16] found that there was significant AF coupling in Co/Cu (100)

layers grown by MBE on Cu (100). Others found that the response was absent
in MBE Co/Cu MLs grown on Cu (111) [22,29], leading to the speculation that the
(100) grains in the sputtered examples were related to the GMR effect [29].
However, Parkin, et al. [6] found that there was significant AF coupling in a
system of Co/Cu MLs grown in a highly <111> texture on Pt buffer layers
deposited on (0001) sapphire. They suggested that the absence of AF coupling
seen in other <111> preparations, including those gown epitaxially on GaAs, may
be a result of structural imperfections. Grelg, et al. [19] found that the GMR in
<111> epitaxial films grown on GaAs was highly dependent on the seed I buffer
layer. Their low angle X-ray data also show strikingly more distinct layers when
the buffer layer was Au rather than Cu. In a sputter deposited system, Modak, et
al. [30] found that by changing the order of growth larger single crystals could be

grown through 3 ML stack (short range epitaxy, Figure 1.4).

 

Figure 1.4. Proposed grain structures of Co/Cu MLs. A) Grown directly on 3 Fe
seed layer with a thick Cu capping layer. B) Grown on a 35 nm Cu base layer.
From [30]

Films grown on a 35 nm Cu buffer layer had larger crystals and exhibited
about 42% higher GMR than similar MLs grown directly on the Fe seed layer
(with the 35 nm Cu layer deposited on the top of the ML stack. Their TEM data
also show considerably larger single crystal domains in the Co/Cu ML when

grown on the thick Cu buffer layer.

The interfacial quality of the ML is a fabrication consideration that can also be
important in the GMR response. One of the variables in the interface quality is
roughness. There have been conflicting reports concerning the effect of
interfacial roughness on the GMR response. For example Chlédek, et al. [31]
mention that the increased GMR seen in post deposition annealed systems is
due to roughness caused by partially broken MLs. In a study of the effect of post
deposition annealing, Ratzke, et al. [25] found that a 30 min anneal at 350°C
caused a layer break-up that was associated with an increased coercivity and an
initial increase in GMR. They consider the annealing effect to be one of lowering
crystallographic defects while also causing the break-up of the Co layer into
smaller domains. An annealing effect was also seen in Co/Cu multilayers, of
varying Cu thickness, grown on glass substrates. In this study post deposition
annealing over a range of Cu layer thicknesses, lead first to a slight increase in
MR then a monotonic decrease. This was ascribed to the resolution of crystal
defects , in the first part of the annealing, then to a decrease in MR with grain
size increase [32]. On the other hand, Miura, et al., [20] have demonstrated a
sharp drop in GMR with the increase of interfacial roughness in Co/Cu ML

systems. An imperfection related to layer break-up is in the form of "pinholes" in

10

the Cu layers. These ferromagnetic material "channels" can easily happen in the
thin (ca. 1 nm) layers near the first AF peak and the theoretical maximum AF
exchange can be masked. Thus, the GMR is affected by FM coupling through

these defects [33].

Plaskett and McGuire [14] found that GMR increased rapidly at first then with
a diminishing rate with the number of layers. They postulate that, as roughness
increases with higher layer number, the electron scattering caused by these
rough high-order layers dilutes the overall GMR effect. This is supported by in-
situ X-ray diffraction analysis of a Permalloy / Ag ML system during fabrication in
[31], where it was found that interfacial roughness increased with layer number.
Kubinski, et al., present micrographs showing a slight serial increase in interfacial
roughness in a 16 bilayer stack (in another system) deposited on a Ru buffer
over an oxidized Si substrate [35]. However, in Co/Cu and a COI CuX - Au1.x
systems [34], TEM images of specimen cross-sections show no discernable
increase in roughness with layer number. Lastly, in an electrodeposited system

[28], the final surface roughness decreased at higher layer number.

lnterdiffusion is a ubiquitous concern in thin films in general [36]. For sensor
applications, the inter-diffusion rates between elements forming separate layers,
having distinct functions, are crucial concerns. Elements that have similar crystal
structures and lattice parameters generally form solid solutions [37]. However,
copper and cobalt are immiscible at temperatures below 422°C [38] and the
associated positive heat of mixing [39] should promote highly stable interfaces

(Figure 5). Nevertheless, the formation of ordered interdiffused regions has been

11

reported [40] in Co/Cu systems. However, in other studies involving annealing,
coalescence leading to breakdown of the Co/Cu ML rather than diffusion has
been reported. This coalescence is implied by a decrease of saturation
magnetization field [12] and the presence of what appears to be Cu migrating
along Co grain boundaries as seen in TEM images [25]. Diffusion and/or

reactions with certain substrates has also been reported in GMR MLs [18,25].

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

'C MXCO
Co to 20 30 a0 50 60 "0 '0 °° 0’
1 l l l l l J
rm
MP
r¢6°
ism ‘ - ‘E‘
N g
I.
face) IO) \ O
667‘ (aCo)¢UO 2
12m 1'
// ioun- g
I o
I (OCO)O(CUI ‘ E
I i
am i
I
———————————— h——r-—-Ii—-~>——<I-—-o-—-d o l l LL41 L LL 11 l l l l l l l l 1
JED 0 25 50 75 I”
r: Co) . cw Co emanation (am)
a B
A 0 r0 :0 so a so co )0 an co m
AfZCu

Figure 1.5. The immiscibility of copper and cobalt. A) The binary phase diagram
for Cu and Co [38]. B) The calculated free energy of mixing for Co and Cu [39].

12

Reactions with Si substrates leading to the formation of silicides has been

reported and used to form a basis for epitaxial growth [18].
1.6 Epitaxial Co/Cu GMR MLs:

Although the GMR effect was first seen in MBE grown MLs, most of the work
in the Co/Cu system has been done with polycrystalline sputter deposited
samples. However, epitaxially grown Co/Cu MLs have also been prepared by
sputtering [e.g. 41]. The possibility of producing a more homogenous medium
through the fabrication of single crystal (i.e. epitaxially grown) layers in GMR
devices was early recognized as an intriguing approach for (at least)
experimental modeling [e.g. 6]. ln polycrystalline Co/Cu MLs, involving identical
GMR layer stacks grown with differing crystallinity, Modak, et al. [30] found that
the CIP GMR showed a distinct increase (about 42%) with a grain size increase
from about 15 to 35 nm. These data may appear contradictory to the results of
Tomlinson, et al. who reported an increase in GMR with decreasing crystal size
[24]. However, in the large crystal samples of Modak, et al., the film was <111>
textured and the large grains were usually epitaxial [30]. Early on, single crystal
epitaxial superlattices of Co/Cu had been grown using a single crystal Cu (100)
substrate [15], however this was accomplished by electron beam evaporation.
Sputtering systems are the practical method of choice in many thin-film
fabrication endeavors owing to their higher deposition rate, lower capital and
operation costs [36]. Co/Cu superlattices with large epitaxial areas have been
grown on M90 (110) and (100) substrates using Pt or Pd as a seed layers and

on sapphire (0001) using Pt as a seed layer [41]. Epitaxial growth in the Co/Cu

13

ML system has been shown to be highly dependent on the initial composition,
dimensions, and structural quality of the initial layer and thus on the deposition
conditions. The copper-cobalt system should be amenable to the formation of
such epitaxial layers. Although the equilibrium structure of copper is fcc and
cobalt is predominately hcp, cobalt can also will exist as an fcc structure and this
is generally the form that has predominated when Co thin films are grown on Cu

in MLs [e.g. 30,42].

The discovery of the GMR effect and most subsequent GMR measurements
have been done by measuring the AR along the planes of the ML. Sensitive
measurement of the GMR effect, especially for experimental purposes, is more
effective in the current perpendicular to plane (CPP) geometry [43,44]. Super
conducting Nb contacts are fabricated on top and bottom of the specimen since
the resistances are very small in this geometry [43]. Nb is the contact material of
choice for measurements in this temperature range due to its relatively high (9.5
K) superconducting behavior, which remains intact well above the He boiling
point even in the presence of a strong magnetic field. The GMR across the
sample is then measured using a super conducting quantum interference device

magnetometer (SQUID) with a sample measuring temperature of 4.2 K.

The epitaxial growth of Nb on sapphire is quite facile, with the (110) planes of
Nb growing epitaxially on the (1 120) of sapphire [45] with the Nb atomic lattice
patterning on the Al atoms of the sapphire. Extended Nb single crystals have
been made in this orientation by sputtering the Nb on heated single crystal

sapphire [44] substrates followed by a post deposition annealing treatment. The

14

thermal expansion coefficients of the two materials are similar (Table1.1) and a
smooth-surfaced Nb layer is retained after cooling. The misfit between the Nb
and the sapphire has been explored for a number of orientations and the lattice
strain is accommodated by a network of dislocations [47-50]. However, the
mismatch between the Nb and the Cu of the next deposited layer is more sever.
The mismatch between the Nb (110) surface and the close packed planes of Cu
(111) is accommodated with a Kurdjumov-Sachs orientation relation [46,51]. The
fcc form of Co and Cu have a lattice mismatch of less than 2% that appears to be

elastically accommodated in epitaxial thin films.
1.7 Characterization of magnetic MLs:

To correlate the structural concerns outlined above with device performance
and predict their theoretical significance, unambiguous characterizations are
desired. However, there is no one technique that provides a comprehensive
measurement of all of the important characteristics of ML structures. For GMR
MLs, low angle (29 < 10°), and high angle (29>10°) X-ray diffraction of the
fabricated ML are attractive methods. Both can be accomplished during the

thermal annealing steps of the fabrication process [e.g. 31].

Low angle scans of ML devices are a rapid and nondestructive means of
characterizing the layer frequency, overall texture, and crystal structure. High
angle diffraction can be used to obtain the crystal structure, orientation (texture),
and average grain size. The rapid through-put and non-destructive nature of this
technique has made it the most commonly used characterization technique [1,

37]

15

Table 1.1 Physical Characteristics of the multilayer materials

 

 

 

 

 

 

 

 

Material Al203 Nb Cu a-Co B-Co
Characteristic (sapphire)
structure corrundum bcc fcc hcp fcc
Strukturbericht 051 A2 A1 A3 A1
space group 167 229 225 194 225
point group R-30 lm3m Fm3m P63/m m Fm3m
c
lattice a=0.4763 a=0.3307 a=0.3615 a=0.2507 a=0.3544
parameters c=1 .303 c=0.4070
Atomic - 41 29 27 27
number (Z)
melting point 2030°C 2477°C 1085°C 1495°C 1495°C
coefficient of a=5.0 x 10'6 7.2-7.6 17.0-18.3 12.1-14.0 12.1-14.0
”pans” c= 9.03 x 10’6 x 10'6 x 10'6 x 105 x 10"

 

 

 

 

 

Nuclear magnetic resonance (NMR) is a second technique that has yielded
important structural information in multilayers [21,42]. By comparing the
resonance frequency (the frequency causing nuclear spin flipping) in a pulsed,
variable-frequency electromagnetic field, a spectrum of echo frequency maxima
is generated. These resonance frequencies are dependent on the environment of
the lattice atoms and, by comparisons to standards, the atomic environment
(crystal structure) of the atoms can be determined. The low frequency signal can

also give a measure of surface roughness in terms of Co nearest neighbors.

Reflection high energy electron diffraction (RHEED) is a diffraction technique
that involves the use of grazing incident electron waves to probe the surface

layers of the target. RHEED patterns are especially useful in determining the

16

 

quality of layer surfaces. RHEED is a non-destructive technique and has the add
advantage of being applicable during the film growth process. RHEED devices
are de n'gueur in commercial MBE systems and as such are the most common

film qualifying technique for MBE epitaxial studies [e.g. 19.]

Scanning electron microscopy (SEM) generally has not the resolution or
penetration to yield much information on completed devices. However, SEM with
polarization has been used to characterize both the crystalline and magnetic
structure of sputtered Co/Cu MLs [13]. SEM is also useful in the crystallographic
orientational analysis of MLs component by component [45] via electron back-
scatter diffraction (EBSP) techniques. Automated indexing systems allow

relatively large areas to be mapped for crystal orientation.

Transmission electron microscopy (TEM) of cross-sectioned materials has
also proven useful. Ordinary diffraction contrast can be used to differentiate
between thin metal layers in establishing layer dimensions normal to the electron
beam. This process is straightforward for materials with significantly different
elastic cross sections and for differences on the order of Nb or Ag and Cu or
NiFe (AZz17) [41, 54]. However, in the case of the layers in the Co/Cu system,
there is little differential scattering contrast due to closeness of atomic numbers
(AZ=2). One approach that has been used to delineate overall layer quality in
such systems is Fresnel defocus contrast [25, 28]. This technique makes use of
differences in the mean inner potential of the two materials (which is a function of
their densities and fonivard scattering factors) to generate focus sensitive

interference images. This interference is seen as a band of contrast change at

17

the interface between regions with differing mean inner potentials [51,52]. Such
techniques allow the quantitative analysis of layer interfaces by evaluation of the

focus dependent fringe amplitude.

High resolution transmission electron micrOSCOpy (HRTEM) pushes the
resolution of the TEM to the atomic level. HRTEM alloWs crystallographic details
of such features as crystal defects, layer interfaces, local non-equilibrium
structures and grain boundaries to be explored at the lattice level [32,41,54].
Extraction of periodic information from HRTEM micrographs has long been
enhanced by the use of optical diffraction techniques [55]. Sinclair, et al. [56]
have presented extensive analysis of metal lattice images and note that the
technique allows the extraction of crystal orientation and local lattice variations
from very small areas of the HRTEM micrograph, giving spatial resolution far
better than practical with electron diffraction techniques. Combining HRTEM of
metal foils with image analysis via computer generated fast Fourier
transformation (FFT) allows rapid quantitative crystallographic analysis of regions

a few square nanometers from HRTEM lattice images [e.g. 54].

In the latest, intermediate-voltage field-emission analytical electron
microscopes (AEM)s, spatial analytical resolution has decreased to the
nanometer or smaller scale in the past few years. This allows the near atomic
level detection of individual metal ions [57]. Surprisingly few GMR ML systems
have been studied using AEM [9.9.56]. AEM coupled with high resolution
imaging, which is increasingly important in the study of magnetic MLs, can

provide corroborative independent data for structural model testing [53].

18

1.8 Relevance of Previous Work to This Study:

Considerable work has been done on the preparation, GMR characterization,
and structural analysis of variously prepared Co/Cu ML systems. Conflicting
results appear with respect to the effects of a number of structural variables on
the final device performance. It has been noted that, for a number of fabrication
techniques and structures, the preparation of uniform material with known
magnetic transport properties can be somewhat site (i.e. investigator) specific

[35, 37].

Epitaxial Co/Cu MLs have been fabricated by a number of groups. The
current study concerns the epitaxial growth of metal MLs like those used in GMR
devices, using the Co/Cu system as a model, between the Nb contacts required
for CPP measurement. The formation of an initial epitaxial layer is crucial to the
subsequent ML development. Recent work [46] has confirmed that sputter
deposited Nb will grow epitaxially on a (1 120) sapphire surface. Samples with
extensive layers of NB contact metal outside of the ML device to be tested make
CIP GMR measurement difficult due to the dilution effect of current through these
layers. However, preliminary qualitative comparative CIP GMR measurements
on epitaxial Co/Cu MLs revealed values much lower than the similarly configured
polycrystalline sample. It was important to determine the degree of epitaxi in
these samples. Possible structural correlations between the GMR observations
and comparisons to the polycrystalline sample were also sought. In the present
study, an initial series of 3 nominally identical epitaxially-grown Co/Cu MLs were

analyzed. A second Co/Cu ML, of identical layering dimension, was fabricated

19

during the same production run, with the entire deposition conducted at room
temperature. This provided a polycrystalline sample for comparison. Previously,
CTEM has been used mainly to qualitatively analyze Co/Cu ML systems [e.g. 25-
28]. HRTEM has shown numerous short-range crystal defects exist in Co/Cu
systems [30,41]. Given the importance of structure, on many scales, to the
performance and of Co/Cu GMR devices, CTEM, HRTEM and HREDS were
used to analyze the structure and composition of the samples. The resulting
information should help to determine the fabrication variables to be considered in

attempting to increase the precision of epitaxial growth of GMR ML devices.

20

CHAPTER 2 EXPERIMENTAL PROCEDURES AND MATERIALS
2.1 Magnetron sputtering ML device fabrication

The GMR ML devices investigated in this study were fabricated by Dr. Reza
Loloee in the MSU Department of Physics and Astronomy. The detailed features
of the fabrication system has been described elsewhere (60). Briefly, an UHV
compatible DC-magnetron sputtering system, with multiple sputtering heads, was

used to apply alternating metal layers on (1 120) surfaces of Al203 single crystal

coupons (Union Carbide). Prior to film deposition, the sapphire surfaces were
washed with hot water/surfactant (Alconox) followed by serial rinses in acetone,
ethanol and de-ionized water. Cleaning was carried out in a class 1000 clean
room adjacent to the sputtering facility. The sapphire substrates received an
additional IR absorbing Nb coating (200 nm) on the back surface, prior to device
fabrication, to help maintain the temperature during deposition. Nominal
thicknesses of the layers were determined by timing of the sputtering process.
The timing was based on the independent analysis, by profilometery and atomic
force microscopy, of previously sputtered films. Up to four samples were
prepared in a single system pump-down (Base pressure = 2x10'8T).
2.2 Co/Cu GMR ML devices

The four GMR ML samples (F igure2.1) of the nominal form sapphire/szso nm
lCu20nm (Comm/Cu5nm)x6/Nb100nm investigated here were fabricated in a single
vacuum cycle in the apparatus described above. In addition, two single bilayer
samples of the form sapphire/Nbgso nm ICu20nm /COzonm fabricated in a prior run, as

fabrication trials, were prepared for TEM.

21

 

Figure 2.1 Schematic cross-section diagrams of the test specimens. A) The
Co/Cu GMR ML pattern. B) A preliminary Co/Cu bilayer sputtered on a Nb
substrate.

22

The GMR ML samples, serially fabricated, were denoted as 20-1, 20-2, 20-3 and
20-4. The two preliminary bilayer samples were 19—3 and 19-4. The growth
conditions for the MLs and bilayers examined are listed in Table 2.1.

In all samples, the first layer deposited was a nominally 250 nm thick Nb base
layer equivalent to that which is required for superconducting contacts in EBSVs.

This film was deposited on the pre-annealed (1 120) surface of the sapphire.

Ideally, in the coherent areas, the Nb (110) planes align with the Al atoms in the
sapphire. The generation of epitaxial growth is dependent on the establishment
of a suitable seed or buffer layer. Cu was chosen for this and for the NM layer in
the GMR ML. The growth of epitaxial Cu layers on epitaxial Nb base (contact)
layers has recently been investigated [46,60]. In those reports, the lower-energy
Nishiyama-Wasserman (N-W) relationship was determined to exist. The
interface between Cu and Nb has also been found to follow the Kurdjumow-
Sachs (K-S) relation (<110>{111}c;,,||<111>{110}n.> ) in thin Cu films grown on Nb
base layers and in multilayers [51,61]. This should form a close packed (111)
surface of Cu for further epitaxial growth. Figure 2.2A shows a schematic of the
K-S relation viewed in the plane of the TEM samples.

Ideally subsequent epitaxial layers of Co and Cu should form with uniform
planes of (111) fcc Co growing on (111) planes of Cu. The hcp to fcc structure
change in Co is generally found to occur martensitically at a temperature of 422°
C (Figure 1.5). The lattice mismatch is about 2% and in thin layers this should be
accommodated elastically. Figure 228 shows the relative relationship in the

TEM sample plane.

23

Table 2.1 Co/Cu bilayer and GMR multilayer fabrication temperatures

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

SAMPLE 1 Layer [ Deposition T I Anneal t@T LComments
193 Nb 750 10 min @ 950 Epitaxial
Cu 350 1 bilayer
Co 450 CUZOnm/COZOnm
19-4 Nb 750 10 min @ 950 Epitaxial
Cu 470 1 bilayer
CO 100 CUZOnm/COZOnm
20-1 Nb 750 10 min @ 950 Epitaxial
CU1 1 50 350 (COSnm/Cusnm)x
Bal 90-100 - 6 layers
20-2 Nb RT RT Polycrystalline
(006nm, C UanmIX
Cum RT RT 6 layers
C01-6nm RT RT
20-3 Nb 750 10 min @ 950 Epitaxial
cm 450 - (006nm/CUGnm)x
Bal 90-100 - 6 layer
20-4 Nb 750 10 min @ 950 Epitaxial
CU‘I 450 - (COGnrn/CUGnm)x
001 490 - 6 layer
Bal 90-100

 

 

 

 

 

24

 

 

 

 

Figure 2.2. Schematics of ML stacking. A) Fcc Cu over bcc Nb in the
Kurdjumow-Sachs orientation. B) The non-relaxed incoherent spacing between
Co and Cu in the fcc crystal structure.

25

2.3 CIP GMR measurements

Although the Nb contact layer prevents direct comparison with previously
reported data collected for similar MLs, current in plane (CIP) measurements
were made between one epitaxial sample and one polycrystalline sample for a
relative comparison. Samples 20-1 and 20-2 were tested in conventional 4-
probe CIP configuration, at room temperature with a maximum applied field of 2
kOe. The general form (Figure2.3) of the CIP GMR response for sample 20-2
was similar to that reported previously for polycrystalline ML samples of this motif
[13, 62]. Sample 20-1 showed a AR/R of 1.3% while 20-2 showed 3.1% AR/R.
While these values cannot be directly compared to conventional CIP devices,
due to the large Nb conduction, these values are in agreement with those seen
previously in Co/Cu MLs with a 6 nm motif [13]. The two remaining samples, 20-

3 and 20-4, were not tested.

 

  
    
 

 

1 ' '
A 0008 _ _ 1 .0 z
8 __ - o
c as-prepared-s» _. 3
g 0.06 - _ g
o- N
U) .. CD
93 ' a.
2 0 04 " coerdve'é — O m
0 . 9..
r: ,— m
a - s-
E 0.02 . to
E remanent _ g
5 0.00 ------ -1 0:0

 

 

 

Magnetic Field (0e)
Figure 2.3. The general CIP MR response in Cos nm ICus nm MLs [13].

26

2.4 TEM sample preparation

The cross sectional samples for TEM, AEM, and HRTEM were prepared
essentially by the methods of Bravman and Sinclair [63] as modified by Howell
[64] and Geng, et al. [54]. This was accomplished by scoring the as-fabricated
ML substrate (backside) and snapping the wafer into strips roughly 0.3 x 1.0 cm.
The resulting pieces were cleaned in acetone and ethanol and glued, with the ML
surfaces facing each other, with Gatan G-1TM Epoxy cement. The sample was
then clamped with gentle pressure in a spring-loaded TeflonTM lined clamp and
cured for about 1 hr at =130°C. Following polymerization, the resulting
“sandwich” was backed in two layers of silicon (or sapphire) wafer cemented to
the initial assemble with the same technique. The use of sapphire backing made
a more robust sample after final ion milling owing to the lower sputtering rate
than Si, which often perforated before the region of interest at the bottom of the
dimple. However, the thickness of the clear and colorless sapphire cannot be
judged as easily as Si when dimpling.

The resulting stack was temporally mounted on a 25 x 40 mm glass
microscope slide with CrystalbondTM mounting wax. This was then sliced
perpendicularly to the wafers surface with a high-speed diamond saw (Strueres
Acutome-5), operated at 3000 rpm with a 5 pm/ sec feed rate, to yield slabs
approximately 800 pm thick. These slabs were polished to a 50 nm finish with a
series of diamond lapping films (15,5,3,1 pm 3M lmperial‘iM diamond lapping film)
followed by 0.05 pm colloidal diamond. After washing in detergent and water

followed by water and ethanol rinses, the polished side was cemented (G1

27

 

epoxy) to a Mo support ring 3 mm in diameter with a 2mm hole. This slab-and-
ring assembly was wax mounted in the center of a 19 mm optically-flat sapphire
disk and lapped, with the same series as the first side, leaving a final specimen
thickness of 5 80pm. The samples were then mounted and centered on the
platen of a VCR group D500I dimpler and dimpled with 1 pm diamond slurry
using a 15.5 mm stainless steel wheel. Alternatively, a similar diameter, 1 pm
embedded-diamond lapping wheel was used with VCR lapping vehicle as the
lubricant. The termination point, a thickness of z 10 pm, was determined by
period removal of the sample and observation via light microscopy, with a stereo
microscope, every 10 to 20 um of dimpling. The samples were then polished to a
specular surface with colloidal diamond (0.050 nm). Following the final polish,
the samples were dismounted from the dimpler supports with acetone, washed in
ethanol, air dried and mounted on a Gatan DuoPostTM support and ion milled to
perforation in a Gatan M691 PlPSTM ion mill. The samples were milled at : 3°
beam incident-angle with 5 keV Ar ions, while turning at 3 rpm until a small
perforated area was seen at the bottom of the dimple. Figure 2.4 shows an

overview of the preparation sequence.

28

   
     

Score
and Break

Glue on
Mo-ring

Low-Angle Dimpling
Ion milling

Figure 2.4. Schematic of the main steps in specimen cross-section preparation.

29

2.5 TEM experimental methods:
CTEM Imaging

Routine brightfield and darkfield CTEM imaging, to judge the overall
morphology, crystallinity, quality of the sample and preparation, was
accomplished with a Hitachi H-800 TEM operated at 200 keV. Some CTEM
images were recorded on a JEOL 4000EX 400 keV microscope with a double-tilt
top entry stage. An aperture small enough to exclude ordered diffraction from
the component materials used in the samples was used for CTEM imaging
(brightfield and darkfield), (Figure 2.5). Selected area electron diffraction (SAED)
was used to roughly characterize the degree of polycrystallinity and to orient the
samples on low order zones. The SAED patterns were recorded with the
smallest aperture that would completely cover the whole ML stack, including the
Nb base and capping layers. Images and SAD patterns were recorded on Kodak
4489 EM film (200 keV) and SO-163 (400keV) processed as suggested by

Kodak.

 

Figure 2.5 Objective apertures used in TEM imaging in the JEOL 4000EX. A)
The CTEM aperture, a 7.2 nm'1 diameter. B) The Scherzer aperture, 20.8 nm“.
The diffraction pattern is Si, B=<110> showing 36 transmitted beams (including
the direct spot).

30

HRTEM

HRTEM analyses of thin areas of interest were accomplished using a JEOL
4000EX TEM with a LaBe cathode operated at 400 keV. The HRTEM was
performed at the Electron Microbeam Analysis Laboratory (EMAL) located at the
University of Michigan, Ann Arbor, MI. For lattice imaging, an objective aperture
(#2 aperture - 20.8 nm‘1) was used to remove frequencies significantly above the
Scherzer resolution limit (Figure 2.5). Lattice image micrographs were recorded
at a nominal 2.5 to 8.0 x 105 times magnification on Kodak SO-163 sheet film,
with the objective lens set to approximately Scherzer defocus. For Fresnel
imaging, the 7.2 nm'1 aperture was used to exclude all but the desired beam.
Specimens were tilted to the desired zone axis using the sapphire substrate
(dynamical) diffraction pattern from material immediately below the area under
investigation. This should allow the crystal tilt to be aligned to within about 04° of

the zone axis using the symmetrical brightness of the reflections of the {1 0T0}.
This is calculated using the angle to these first order reflections 0 = tan"(R/L),
where R is measured from the diffraction pattern direct spot to any {1 0T0}
reflection, and L: Rd/it. The value of d is known to be 0.233 nm and x is the
relativistically corrected 400 keV i. of 0.00164 nm.
HREDS

Compositional analysis of the MLs, collectively as well as regions of individual
various layers, was accomplished via high-resolution energy-dispersive X-ray
spectroscopy (HREDS). The spectra were acquired with a JEOL 2010F

TEM/STEM, also at the EMAL. The microscope was operated in a spot EDS

31

mode using a probe that was varied between z 70 nm and 0.5 nm at the
specimen surface, depending on the area for spectral acquisition. An EDAX
Phoenix windowless SiLi detector system was used to capture spectra after
which spectral files were translated to Microscopy Society of America (MSA)
format for subsequent analysis using Desk Top Spectrum Analyzer DTSA [65].
No direct method was available for specimen-level beam-current measurement
and known standards were not available so comparisons among specimens are
qualitative.
Foil Thickness Measurements

Specimen thickness normal to the incident beam is desirable for accurate
assessment and modeling of HRTEM images and for quantitative EDS. Foil
thickness can be determined in several ways. The easiest is to generate a
feature on either side of the foil (e.g. carbon contamination) and geometrically
calculate the thickness by separating the projection of the features with a know
angle of specimen tilt. This method generally over estimates the specimen
thickness and is of limited use in the case of very thin (HRTEM) specimens [66].
One of the more widely used methods of calculating foil thickness is the use of
CBED. Indeed, the technique was verified using 3 Co alloy [67,68]. Attempts
were made to use this method. However, a satisfactory CBED pattern was not
generated from the ML and the entire area of interest on the ML was apparently

less than 15.00%} for sapphire. The method of choice, if available, is electron

energy loss spectrometry (EELS). Not only can the thickness at a given position

be measured from the ratio of zero -Ioss to total intensities but, if a scanning

32

mode is available, a thickness map can be generated [65]. Unfortunately, the
EELS features on the analytical microscope used were not available during the
analysis period. A second analytical method of thickness determination, and
one that the ML sample geometry lends itself to is differential X-ray absorption
(DXA) [69]. The foil thickness of a single element can be determined from the
EDS X-ray spectrum if the spectrum contains two detectable, sufficiently
separated characteristic peaks. The Nb base layer provides an ideal internal
thickness standard as its EDS spectrum contains widely separated L and K
peaks. The Nb layer also tapers from a thickness of zero at the margin of the ion
milling perforation to electron opaque at some distance away along the ML.

A series of EDS spectra collected from this foil yielded a linear relationship
between L, and K1 intensities when plotted against distance from the ion milled
edge of the Nb layer. However, attempts to match the slope of the thickness to
simulated absorption ratios failed. This appears to be due to either differences in
modeled vs. actual detector geometry and/or the presence of an extensive
amorphous contamination layer on the specimen. The system could be
calibrated with a known thickness of Nb foil, providing a rapid method of
thickness determination, in near real time without altering the specimen tilt
conditions being used for high-resolution work.

2.6 TEM image analysis
Structural analysis of the HRTEM images was accomplished using calibrated
magnification prints, digitized with a flat-bed scanner coupled with a

microcomputer. The programs used were NIH image [70] and Adobe

33

._

PhotoShopTM (equipped with the Image Processing Tool Kit [71] suite of
analytical plug-ins). Digitized enlargements of HRTEM images were calibrated
using lattice images of known orientations of the Nb contact layers. Similarly, the
fast Fourier transforms (FFTs) of selected areas of the digitized images were
calibrated with transforms of the Nb lattice images. The spacing ofthe {110}
lattice spots seen in the [111] Nb projection transform were found to be

indistinguishable from the {1 0T0} lattice spots seen in [0001] sapphire

projection (Figure 2.6). Sapphire is regarded as undistorted by the Nb layer
[48,49]. Both the NIH image fast Hartley transform [72] and the propitiatory FFT
program code used in the IPTK suite yielded identical result when applied to the
same real space pixel maps. Thus the latter was used, for convenience, in most
cases. Transforms for orientational and structural analysis were made from 128
x128, 256x256 and 512 x512 pixel arrays selected from 600 pixel / inch images
from photographic prints printed at final magnifications from about 2.7 x10° to 107
times magnification. The choice pixel density in the analysis was a function of
the final image size and the expanse of a particular analysis. The angles
between resolvable lattice planes and their respective spacings for the
equilibrium structures of the elements used in these MLs are shown in Figure

2.7.

34

      

Figure 2.6. The Sapphire/Nb boundary. The transform of this area shows that

there is no detectable difference in the Nb {110} and sapphire {1 0T0} spacings.

Thus Nb lattice images were used to calibrate the HRTEM images.

35

60°

     

002
o

O
sitar"
u= . nm .
. !_ 3525
111

 
 

  
 
 

      

   

Cu=0 205 nm
Cu=0 208 nm
0- ‘3.
111 111

. O

 

.-

002 -
ZoneAxrs=[110] ZoneAxis=[111]

FCC BCC

  
    

0002
O
28.62°

  

Co=0. 192 nm

0002
Co=0176 nm

Zone Axrs = [100]
HCP

Figure 2.7. Resolvable lattice planes and their interplaner angles for equilibrium

structures of the elements used in the ML fabrication.

36

 

CHAPTER 3 EXPERIMENTAL RESULTS

Four ML samples (three epitaxial and one polycrystalline) and two epitaxial
bilayer samples were evaluated using TEM methods. The results of these
observations are presented, sample by sample, below. CTEM and HRTEM were
performed on representative thin areas of ion-milled cross -sections of each of
the samples. HREDS was performed on selected areas and structures in two of
the epitaxial samples . These were the polycrystalline ML sample and one of the
epitaxial bilayer samples. Within each sample, the order in which individual
layers were analyzed by HRTEM was from the upper-most "top" of the multilayer
progressing to the base layer. Thus, after the Nb capping layer, the first stratum
of the multilayer analyzed was the last copper layer deposited (Cu6). This was
followed by C06, the last cobalt layer deposited, etc.
3.1 Sample 20-1 (epitaxial)
CTEM of sample 20-1

CTEM, provided useful information about the overall uniformity, growth
texture, degree of epitaxy, and continuity of layers in all of the Co/Cu MLs in this
study. Figure 3.1 shows constant magnification brightfield and darkfield images
of the epitaxial sample 20-1. The images in Figure 3.1 A & B, taken with the

beam normal to the sapphire (ITOO) show a large and variable amount of

contrast in the Nb layer. This indicates that the Nb base contact layer is
crystalline but with a high defect density. The absence of grain boundaries
suggests that it is a single epitaxial crystal. The measured Nb contact layer

thickness was 201 nm (calculated from HRTEM lattice images) which is about

37

20% thinner than the nominal value of 250 nm. An SAED pattern from this
sample is presented as an insert in Figure 3.1. It shows that there is good long-
range order, with no observable radial blurring of the diffraction spots, which
would be the result of multiple grains oriented close to a given axis. There is little
interlayer contrast evident when imaged on the zone axis at Gausian focus
(Figure 3.1A). However, a weak layer modulation is evident in the darkfield
image obtained using the Nb (200) reflection. Defocusing the objective lens
reveals the layering much better (Figure 3.1 C). The ML stack shows a consistent
growth direction modulation, over the entire observed length, consistent with 6
bilayers. The Cu "buffer" layer thickness is also reasonably close to the nominal
value (20 nm) for the first layer. Across the field of view measured, this thickness
varied between 26 nm and 20 nm. The (Cue nm/Cos nm) x 6 ML averaged about
79 nm over the area viewed which is about 10% thicker than the nominal 72 nm '
thickness. The C0 (2:27) layers of the stack appear darker than the Cu (2:29)
layers (Figure 3.2), which is inconsistent with elastic scattering contrast
(variations in transmitted intensity) which is proportional to 22. In addition,
although the total ML thickness is close to the nominal sum, the individual Cu
layers in the stack appear thinner than expected.
HRTEM of sample 20-1

Regions of interest in the sample MLs were also examined using high
resolution transmission electron microscopy (HRTEM). Images were recorded, ‘
near Scherzer defocus, from thin areas having a complete ML device cross

section remaining.

38

 

Figure 3.1. CTEM images of sample 20-1. A) Brightfield minimum contrast

focus image, normal to the sapphire (10T0). B) (200) Darkfield on the same
axis. C) Over-focused brightfield image showing Fresnel contrast parallel to
the layer interfaces.

 

 

Figure 3.2. Multibeam (HRTEM) approximately Scherzer focus images and
average density plots along MLs. A) Sample 20-1. B) Sample 20-2. C.
Sample 20-4.

39

Figure 3.3 shows an overview HRTEM image of a representative area of the
20-1 ML. In this image, the general layering seen in the CTEM image (Figure
3.1) is still evident, despite the similarity in the scattering power of the Co and Cu
constituents. Selected areas in the three regions delineated were analyzed via
FFT analysis to assess lattice structure. The apparent local roughness and
diffusiveness of the interfaces decreases with layer number towards the Cu
buffer layer. A general increase in the crystallographic order of the layers and in
their overall width can also be seen towards the buffer layer. Additionally, both
the top and bottom Nb-Cu interfaces exhibit apparently diffuse zones. Some
recording artifacts are also present. A small, white ,"aster-Iike" featurescan be
seen in the second delineated area. This was the result of a static electrical
' discharge during film handling.

The upper part of the ML (Figure 3.3 inset 1) is detailed in Figure 3.4. The
darkregion at the top of the micrograph is the capping Nb layer. Below that can
be seen the top two Co layers (also relatively dark) and the adjacent Cu layers.
Insets A through F delineate the square (256 X 256 pixels) areas used for FFT
analysis. In Figure 3.5, the regions A-C are enlarged and presented (growth
direction vertical) along with their contrast-inverted (lattice spots appear white)
FFTs. In Figure 3.5A, the top Nb layer shows some lattice fringes in the growth
direction which are indicated by the two vertically oriented maxima in the
adjacent FFT spaced at 0.23 nm, consistent with the Nb {110} spacing . Other

then these maxima, the FFT is relatively diffuse, indicating a lack of resolvable

4o

 

@110
639
cm
C%

_ ,en
es
en
@9
@9

Iw —’ .7]
10) @1011]

 

Figure 3.3. HRTEM image of a selected region of sample 20-1. The growth
direction is about 30° off the depicted vertical. The three regions selected for
lattice image analysis are indicated.

41

 

Figure 3.4 An HRTEM image of the upper layers in sample 201. This figure is
oriented with the growth direction running vertically. The top dark region is the
capping Nb layer. The insets are areas that were tested by FFT analysis.

42

 

arrimm‘“

 

Figure 3.5. Selected HRTEM images A-F and their FFTs from the upper layers
of sample 20—1.

43

crystal periodicities at other angles or frequencies. Figure 358 shows the image
and transform of the diffuse region between the last deposited Cu layer (Cu6)
and the Nb capping layer. The F FT of this area shows two weak maxima,
oriented about 38° from the growth direction, spaced at 0.18 nm, consistent with
the Cu (200) interplanar spacing. This implies that the direction of the close
packed planes was about 17° off the nominal growth direction for the whole ML.
The third area, Figure 3.5C, is an area that should be contained in the Cu6 layer.
The FFT shows a diffuse contrast consistent with an amorphous phase rather
then discrete maxima indicating order. Inset Figure3.5 D is taken from the
topmost Co layer (C06). This transform shows two fairly strong reflections
spaced at 0.20 nm in the growth direction, consistent with Com (111) spacing.
However, there are also two weaker reflections, corresponding to spacings of
0.25 nm, at an angle of 57° to the growth direction. Inspection of the transformed
area for this sample shows lattice fringes corresponding to this angular
orientation in mainly in the lower left quadrant. This spacing is inconsistent with
those expected in the ideotype ML. It most closely matches the "a" value for on-
Co, 0.251 nm (twice the (1 120) spacing). In Figure 3.5 E, the transformed area
from the Cu5 layer shows a wavy band of contrast across it. The F FT of this
region yielded the 0.20 nm spaced reflections in the growth direction and a
second wider set (0.25 nm), as did the previous area. However, in this area the
angle for the second set of spots was 55° relative to the growth direction.
Additionally, the spots on the growth axis appeared elongated; nearly, but

indistinctly, two closely spaced reflections. In the final area analyzed in the top

portion of the ML, Figure 3.5 F, the selected area is situated centrally on layer
005. The elongated growth direction reflections, seen in the previous selection,
are resolved into spacings of 0.20 and about 0.25 nm.

In Figure 3.6, the middle field of the overall HRTEM image (Figure 3.6 inset
2), progressively more distinct layer interfaces are seen from the Co5||Cu4
interface down through the layers to the C02||Cu2 interface near the bottom of
the figure. An area enlargement and transform from the Co5||Cu4 interface is
shown in Figure 3.7A. In addition to the to the 0.20 nm and 0.25 nm spots
oriented in the growth direction, there appears to be a discontinuous
(polycrystalline) ring at about the 0.29 nm spacing. This spacing is not from any
known orientation of the component materials. . In the center of the area layers
C04 and C03 are of nearly uniform contrast. In Figure 3.7B, (004), the strongest
intensities are along the growth direction consistent with the lattice images visible
in the micrograph. The 0.20 nm and 0.25 nm spots, resolvable in the previously
inspected area are less distinct, however. Figures 3.7 C and D are two samples
from the Cu3 layer. In the first of these (C), in addition to the growth oriented
0.20 nm spots consistent with the Co and Cu {111} spacing, there are two other
sets of resolvable spots. These are both from a lattice spacing of 0.29 nm. They
are oriented 72° from each other and 18° from the growth axis. In the
micrograph these appear as a "texture" parallel to the growth axis. Across the
same Cu3 layer, Figure 3.7D shows lattices in the growth direction at the Co or
Cu {111} 0.20 nm spacing, 32° from this are planes at the 0.29 nm spacing.

There is another strong set of spots 120° from the growth axis and 92° from the

45

previously mentioned spots that correspond to a 0.24 lattice, this is just slightly
wider than the Nb {110} (0.23 nm) spacing. Figure 3.70, the lattice image that
produced this transform, appears to contain at least two areas with distinctly
different lattice orientations. Figure 3.7E is an area of layer Cu2. However, most
of the lattice image seen in this area appears to be a contribution from the
adjacent Co layers. The transform shows growth direction oriented spots and an
additional weaker set at about 47° from the growth axis. The final area from the
middle region of the ML, Figure 3.7F, shows an area from the 002 layer. The
only distinct spots in this image are at spacings of 0.20 and 0.23 nm along the
growth axis. These spacings suggests the presence of a Nb {110} lattice and the
Co/Cu {111} lattice in the same area. The HRTEM image (Figure 3.7F) appears
to have two different lattices in it and a wider spacing is seen in the upper right-
hand corner of the image compared to the lower left-hand side.

The 001 layer, the Cu buffer layer, and the top of the Nb epitaxial base layer
are shown in Figure 3.8 (inset 3 in Figure 3.3). Areas with significantly different
and distinct lattice structures are evident across this field of view. In Figure 3.9A,
an area of the nearly uniform C01 layer is shown. In the adjacent transform,
there is a set of strong, sharp spots at 0.20 nm spacing along the growth axis,
which are again suggestive of the Co/Cu {111} planes. The area of the Cu buffer
layer adjacent to the 001 (Figure 3.9B) layer shows apparent lattice order in
several directions and spacings. The F FT of this area shows 5 distinct sets of

spots. Along the growth axis, there are sets at the 0.20 and 0.25 nm spacings.

46

C03

'-.~-. ; 1“ d l.‘ ‘2'. ‘ ,.~ .
Easiness-rm ‘31-- ,-.
._" .7_V..3'~_..;‘ _..:'.

 

Figure 3.6 HRTEM image of the middle layers in sample 20-1. This figure is
oriented with the growth direction running vertically.

47

 

Figure 3.7. Selected HRTEM images A-F and their FFTs from the middle layers
of sample 20-1.

48

H." ,,,,

.._ ”9.”... 1.. _1-, .'
_ ;)'r.w<) 4’”: «I ._ '
‘ r - . .

 

Figure 3.8. HRTEM image of the bottom layers in sample 20-1. This figure is
oriented with the growth direction running vertically. The lowermost area, F, is
the bottom Nb layer.

49

a m m”

 

Figure 3.9. Selected HRTEM images A-F and their FFTs from the bottom
layers of sample 20-1.

50

At an angle of 25° to these growth direction spots, there is a second 0.20 nm
spaced pair. At 55° there is a second pair of 0.25 nm spacings. Finally at 90° to
the growth axis, there are two weaker spots that correspond to a spacing of 0.30
nm. In Figure 3.9C, which is also from about the middle of the Cu buffer layer,
the most distinct lattice image in the buffer layer is seen. There are 0.20 nm
spacings along the growth axis, consistent with the expected {111} lattice, and an
additional set of 0.25 nm spaced spots at 55° to the growth axis. At 72° from
these spots area third pair with a spacing of 0.280 nm. The FFT in Figure 3.9C
also shows higher order reflections of lower intensity that appear to be second
order spacings from the lower frequency lattices. In an adjacent area of the
buffer layer (Figure 3.9D), strong fringes 55° from the growth axis with a spacing
of 0.25 nm are again evident and are more intense than the weak the growth
direction spots. Again in this area, there are spots 72° from the highest intensity
55° set (122° from the growth axis). The inner most set of spots in this series
corresponds to a spacing greater than 0.6 nm. The final area of the buffer layer
selected was the interface with the Nb base layer (Figure 3.9E). This area was
remarkably uniform throughout the area observed. Its contrast was intermediate
between the Cu buffer layer and the Nb base layer. An FFT of a representative
area of this band shows no indication of any crystalline order. Figure 3.9F is an
area from the Nb base layer. Although only the top portion is shown in Figure
3.8, the entire square area in 3.9F was used for the FFT. This area appeared as

expected for the Nb in this orientation, with only the (110) planes resolvable.

51

HREDS of sample 20-1.

HREDS of Co/Cu ML samples were acquired in spot mode on a JEOL 2010F
with a windowless detector as described in Chapter 2. Two minute live-time
spot-mode spectra were acquired. This was accomplished using an incident spot
size that just covered the ML in cross sectional aspect. Three of these spectra,
from near but not overlapping areas of the ML, were averaged. This averaged
spectrum, along with a DTSA simulation using the microscope, spectrometer,
and specimen parameters outlined in Chapter 2, are presented in Figure 3.10. In
addition to the expected X-rays from Co and Cu, small peaks are visible in all of
the real spectra consistent with C, O, Al, Nb and Mo characteristic X—rays. The
carbon contribution is at least partially due to unavoidable contamination build-up
in the microscope. The oxygen and aluminum are likely due to the resputtering
of trace. amounts of the sapphire substrate. The Nb is likely sputtered from the
base layer of the ML and the Mo from the support ring. A hole count revealed no
detectable levels of these elements. Most strikingly, the relative peak heights for
the Co and Cu K X-rays are far from the yield ratios implied by the simulation.
Further analysis of 36 spectra (18 each from Co and Cu layers) selected from
arbitrary locations along the ML is presented in Figure 3.11. Figure 3.11A shows

a spectrum of the averaged Co and Cu peaks from 18 "light" (nominally Cu)

52

8000

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Co Ker A
7000‘
6000
5000
I»
~
EEAOOO
O
3000
2000‘
l000<
0 , , , T , , M
II 12 13 14 l5 l6 l? 18 I9 20
9000
Cu K
scoot B
7000. C0 K01
6000‘
3 5000‘
i
D 40001
3000-
mwd COKB CuKB
1000~
e f T r

 

Figure 3.10. EDS spectra from a cross-section the 20-1 ML structure. A) The

 

spectrum formed by averaging three whole-ML spectra. B) DTSA simulation of
a Co/Cu sample with the same volume fractions as a circular sample of the 20-1

ML motif

53

2000

.800] A

16004

 

12004

10004

' Count:

8004

600, C0 K

 

 

 

 

0 1 2 3 4 5 6 7 8 9 10 ll 12 l3 14 IS 16 I7 18 19 20

2000

 

16W+

14m ‘

12004 COK

Counts
§

CK
CoK u“

B Cu KB
4 5 6 7 8 9 10 11 12 13 14 13 16 17 18 19 20
W ‘

Figure 3.11. EDS spectra from a cross-section the 20-1 ML structure. A) The

spectrum formed by averaging 18 Cu layer spectra. B) The spectrum formed by
averaging 18 Co layer spectra.

 

 

 

 

A A;

 

 

 

54

layers. This spectrum shows a low total count number and a very low Cu
proportion. The total X-ray yield was nearly an order of magnitude less than that
seen for the ML as a whole (cf. Figure 3.10A). Figure 3.1 B is a spectrum from
the mean of the complementary 18 Co layers showing about twice the Co K X-
ray intensity and nearly the same Cu intensity as the mean for the Cu layers.
Summary for sample 20-1

At first observation in CTEM, sample 20-1 appeared to be a reasonably well
grown epitaxial sample when viewed at lower magnifications. In the electron

diffraction pattern normal to the (ITOO) (i.e. the [1 T01) both the ML and the Nb

base layer appeared as a single crystal. However, imaging in the HRTEM mode,
especially at the magnifications used for lattice imaging, revealed highly
anomalous growth in the ML. The crystallinity of the nominally Cu layers was
either largely missing or appeared to be continuations of lattices originating in the
much more strongly crystalline Co layers throughout most of the ML. Lattice
spacings inconsistent with those of the equilibrium structures of the sample
components were seen in all regions. The Cu buffer layer had what appeared to
be an amorphous interface with the Nb base layer, which is not expected with a
pure Cu||Nb interface. HREDS showed that the sample, as a whole, contained
no unexpected elements, was Cu deficient, and that the Cu layers appeared to
be "mass deficient" which corroborated the lower than expected electron
scattering seen in the nominally Cu layers when viewed in HRTEM (cf. Figure

3.2A and 3.2B).

55

3.2 Sample 20-2 (polycrystalline)

Sample 20-2 was grown during the same operation cycle and under the same
sputtering conditions as the epitaxial samples (Table 2.1), with the exception that
the entire sample was grown at room temperature and there was no post
deposition anneal.

CTEM of sample 20-2

In Figure 3.12 the ML is shown in cross section with the sapphire substrate
oriented in the [0001] direction. The SAED pattern (inset), which included only
the ML and Nb contacts, shows the ML is highly textured in the [111] growth
direction. Figure 3.12A is a brightfield image of the polycrystalline 202 sample
showing the columnar growth. In Figure 3.128, the same area is presented in a

darkfield image using the Nb (0T1) reflection. As expected, in either brightfield

or darkfield mode, there is little interlayer contrast at near Gausian focus. The
bottom Nb layer is distinguishable from the rest of the ML, due to interruptions in
the contrast of the columnar crystals at its interface with the Cu buffer layer and
greater electron scattering. In Figure 3-120 (the same area recorded 11 pm
below minimum contrast focus) the layering is quite evident. Consistent with the
nominal deposition, the first layer above the Nb contact is a thicker Cu layer.
However, there appears to be only 5 bilayers above this, as opposed to the 6
pairs planned. Fresnel defocus contrast also shows that there is a sample-wide
unevenness in the bilayer stack. The amplitude of this unevenness appears to

increase with layer number. It also appears to show layerwise coincidence, with

56

 

Figure 3.12. CTEM images of sample 20-2. A) Single beam brightfield image
of a cross section of the 20-2 ML. B) The darkfield image created using the
(0T0) Nb spot. C) A single-beam brightfield image taken 11pm below
minimum contrast.

57

roughness features arising from the first layers deposited being propagated
through the ML stack.
HRTEM of sample 20-2

The overall interlayer contrast in the polycrystalline sample 20-2 was much
less obvious than that seen in the HRTEM images of sample 20-1( Figure 3.28),
although a wide range of layer-dimension crystallographic features were
apparent. Figure 3-13 shows a low magnification HRTEM image of a selected
region of the ML cross section. The region outlined in the center was chosen for
lattice image analysis as it appeared to contain examples of the major
crystallographic features. The most striking of these features is the striated
appearance of the nominally Co layers, Co2-Co5, in the center of Figure 3.13.
The images were recorded along the sapphire [0001] rather than normal to the

(O 1 To) sapphire, as was the case for sample 20-1. Four 26 nm square regions

were selected from the HRTEM image field for analysis (Figure 3.14). The first
two of these areas (insets 1 and 2 ) were chosen to incorporate examples of
most of the obvious types of crystal defects observable in low magnification
micrographs. As before base-layer Nb lattice image {110} planes, from the same
negative as the ML areas, were used for calibration. In Figure 3.14, the nominal
layer positions, judged by local differences in interlayer contrast are also
indicated. Consistent with the results of the Fresnel contrast image (Figure
3.12C), only 5 bilayer pairs appear to have been deposited. However, the total

ML thickness in the growth direction of the area evaluated (Figure 3.14) is slightly

58

 

5300110170

 

Figure 3.13. Low magnification HRTEM image of the cross-sectioned 20-2
sample. The ML area (ML) and lattice analysis area (inset) are shown.

59

 

Figure 3.14. HRTEM of the ML 20-2. Four regions were selected for lattice
analysis of the metal layers. Based on crystal diffraction contrast there appears
to be 5 rather than 6 bilayers.

60

more than 100 nm, actually thicker than the 92 nm nominal thickness. Although
there is some unevenness, as mentioned in the CTEM description, this overall
dimension is consistent over extended ranges (of. Figure 3.12). Figure 3.15 is a
region near the top of the ML stack and Figure 3.16 is in a region incorporating
the intralayer transition to the striated horizontal contrast modulation. The areas
analyzed in these two regions are presented in Figure 3.17.

The upper half of the micrograph in Figure 3.15 is the Nb capping layer and
the presence of nearly uninterrupted lattice planes perpendicular to the growth
direction is evident. In Figure 3.17A, a nearly prefect [110] growth direction. is
seen in this area of the Nb layer. The FFT shows growth axis reflections
consistent with the 0.2338 nm spacing of the Nb {110} planes. The lower
amplitude of the off axis {110} reflections is due to the higher distortion in the off
axis planes. At the Nb||Cu5 interface, an irregular band of lower contrast can be
seen (Figure 3.15B). Figure 3173 shows the enlargement of this area. The
image dimensions were that of the previous area, to improve frequency
sampling. The FFT shows the expected (110)||(111) growth-axis spots which are
seen as a doublet, due to the 0.2338 nm Nb {110} vs. 0.2084 nm Cu{111}
spacing. However, the spot due to the Nb (0 1T) appears to be displaced
towards the growth axis and is seen at 57° rather than he expected 60°. At 67°,
a reflection consistent with a Cu {111} frequency is seen. The final two spots,
nearly co-radial, at 122° are at 0.18 and 0.23 nm are consistent with the Cu

{200}||Nb {110} relationship. The final area from the top of the ML analyzed

61

 

Figure 3.15. HRTEM of the upper region of sample 20-2. The image
corresponds to the inset (1) from figure 3.14. The upper half of the region is the
Nb capping layer.

62

was-

.. , . .1.
. - .. .
. .
g .
-..
..

“1".

A 5.0 nm

 

Figure 3.16. HRTEM of the middle region of sample 20-2. The image
corresponds to the inset (2) from figure 3.14.

63

8.le min I“

 

Figure 3.17. High magnification HRTEM areas and their corresponding FFT
transforms from Figures 3.15 (A-C) and 3.16 (D-F).

64

be a nearly prefect example of the (111) growth of Cu expected.

In the middle region of the ML (Figure 3.16) distinctive differences in crystal
structure are apparent within the individual layers. In the left side of this region,
the crystal is consistent with a single fcc structure across the layer boundaries
oriented, as expected, with the growth on the (111) planes of fcc Cu and Co. A
representative area (Figure 3.16D) in layer Cu3 is shown in detail in Figure
3.17D. The image and its FFT are consistent with expected (111). Figure 3.16E
shows an area of layer Co3 that exhibits extensive short-range contrast
modulation in the growth direction. The high magnification image and its FFT
(Figure 3.17E) show what appears to be extremely short range twinning on the I
(111) planes. The FFT shows a maximum growth-direction frequency consistent
with the (111) planes of either Co or Cu (the bulk equilibrium spacings of 0.205
nm for Comc and 0.208 nm for Cu cannot be differentiated with the sampling
space used). Along the growth direction are two other sets of lower frequencies
one corresponding to a frequency of twice the (111) spacing and to 5 times the
(111) spacing. The contrast modulation responsible for these spots can be
discerned in Figure 3.17E. The final area analyzed is shown in Figure 3.16F and
the high magnification image is presented in Figure 3.17F. This area of the Cu2
layer shows a single crystal pattern with the expected [111] growth direction.
The crystal in this area has its growth plane stacking switch from an "abcabc" to
a “cbacba” motif compared to that seen in Figures 3.17D from the left side of the

analysis area. Thus it appears as a mirror image with respect to the zone axis.

65

The first Co/Cu bilayer deposited on the Cu buffer layer showed little
differential contrast with the Cu buffer layer (Figure 3.14 inset (3)). This region
shown in, Figure 3.18, was chosen to compare the structures within the layers
cm and 001. Figure 3.19A is a sample area in the Cu1layer. It represents an
essentially perfect domain with the expected (111) growth and equilibrium lattice
spacings. The predominant defect in the Figure 3.18 region appeared to be
small twins on the (1 1T) planes (Figure 3.19B). In the C01 layer (Figure 3.19C)
the F FT shows additional spots at a frequency corresponding to twice the {111}
spacing. Careful inspection of the Figure 3.19C image that generated this
transform shows a subtle alternate plane modulation in contrast.

At the junction of the Cu buffer layer and the C01 layer, there are regions of
differential contrast across the boundary. However, the crystal orientation on
either side is nearly the same. Some slight distortion may be present along the
(1 11') at the interface as is evidenced by the fine splitting of the (1 1T) spot in
Figure 3.19D. Figures 19E and F are selections from the Cu buffer layer shown
in the lower half of Figure 18. The growth throughout the buffer layer is in the
[111] direction with little deviation from the orientation described above.

The quality of growth in the ML is sequentially dependent on the quality of the
previously grown layer. The last region of the 20-2 ML (Figure 3.20)
characterized is the lower part of the Cu buffer layer and the Nb interfacial region
In areas A and B from Figure 3.20, the regular spacing of the (111) planes seen

throughout the area of the sample studied are evident. There is a slight

66

 

Figure 3.18. HRTEM of the first Co and adjacent Cu layers from sample 20-2.
The area corresponds to the inset (3) from figure 3.14.

67

 

I rlfikllirm'

 

Figure 3.19. High magnification HRTEM areas and their corresponding FFT
transforms from Figures 3.18 (A-C) and 3.16 (D-F).

68

difference in rotation along the [1 1 1] in the buffer layer. The areas A and B of
Figure 3.20 both show clear lattice images in the [1 1 1] but only the area
highlighted in B (of. Figure 3.21 A and B) shows the (002) and (1 1T)

reflections. The interfacial region, seen in the middle of Figure 3.20, appears as
an undulating band of contrast change. In the area selected for lattice analysis
(Figure 3.21C) the growth direction spots correspond to spacings of 0.21 nm,
consistent with the Cu {111} spacing. The interfacial region, seen in the middle
of Figure 3.20, appears as an undulating band of contrast change. In the area
selected for lattice analysis (Figure 3.21C) the growth direction spots correspond
to spacings of 0.21 nm, consistent with the Cu {111} spacing. At 50° to the
growth direction there is a strong frequency at 0.23 nm, consistent with the Nb
(110} spacing and at 110° from the growth direction there is a second spot at this
spacing. This is consistent with a <111> viewing zone axis in the Nb layer. In
Figure 3.21 D, the Nb crystal can be seen to be imaged on a <111> zone as all
lattice angles are 60° to each other and the spacing is the 0.23 nm consistent
with the {110} in Nb. Towards the center of the field of view in Figure 3.20, there
is a band of increased contrast. An area (Figure 3.20 inset E) straddling this is
analyzed in Figure 3.21 E. The FFT of the area shows a set of frequencies, both
corresponding to the Nb {110} spacing, oriented at 46° and 60° to the growth
direction spot (also Nb {110} spacing). In the area to the right of the boundary,

the area is oriented with the 46° motif (Figure 3.21 F).

69

 

Figure 3.20. HRTEM of the Cu buffer layer and adjacent Nb base layer from
sample 20-2. The area corresponds to the inset (4) from figure 3.14.

70

7111/} 011117

 

Figure 3.21. High magnification HRTEM areas and their corresponding FFT
transforms from Figures 3.19 (A-C) and 3.16 (D-F).

71

 

Cu K0:

CoK

Cu L peaks

/

 

 

 

 

kaa

_ FA a MoKaNb \
0 I 2 3 4 S 6 7 8 9 10 11 12 13 14 15 16 17 16 19 20

1:09
Figure 3.22. EDS spectrum from a cross-section of the 20-2 ML structure.

 

 

 

 

HREDS of Sample 20-2

Although sample 20-2 consisted of only 5 bilayers, the nominal volumetric
proportions of Co and Cu were still very close to the 40 00:60 Cu ratio seen for
6 bilayers. Figure 3.22 shows a 2 min spot mode spectrum collected under
similar conditions to those used above for sample 20-1. The peak ratios for the
Co and Cu K peaks are essentially the same as those for the simulated spectrum
(Figure 3.108).
Summary for sampleZO-Z

Sample 20-2, being grown exclusively at room temperature, provides an
important comparison with sample 20-1 and the subsequent epitaxially grown

samples. The overall contrast (Figure 3.2) and the overall ML uniformity seen in

72

CTEM mode implies that the deposition process per se does not appear to be
responsible for the growth anomalies seen in sample 20-1. Similarly, since the
general highly textured polycrystalline growth observed is consistent with that
reported by others, it is doubtful that the structures seen in sample 20—1 resulted
from TEM sample preparation artifacts. Foils from both samples were prepared
by essentially identical methods.

The overall Co-Cu composition, based on qualitative evaluation of the Co and
Cu K peaks was nearly identical to the simulated ratio. Although the low
interlayer contrast prevented spot-mode layer by layer analysis, there were
indications of compositional layering due to crystal diffraction.

. Throughout the ML and buffer layer, the Cu grains oriented to yield lattice
images had an apparently low defect density. Co layers, on the other hand, had
a high density of stacking faults in the growth plane that HRTEM images suggest
increases towards the center of the layer, away from the interfaceses. This
appears to be due to a tendency for the fcc Co to revert to the hop, a
phenomenon commonly seen in relatively thick Co layers. The Nb layers forming
the capping layer and the base layers for the ML also were strongly textured in
the growth direction. Most grains were found to be within just a few degrees of

the overall growth direction.

73

3.3 Sample 20-3 (epitaxial)

The third sample prepared from the studied sputtering run (20-3) was grown
with a different temperature regimen than the first epitaxial sample (201). While
the Nb base layer was applied under the same conditions (as shown in table
2.1), the Cu buffer layer was applied at a higher temperature (450°C) and was
given no post deposition anneal in contrast to the Cu buffer layer in 20-1. The
rest of the ML was deposited at 90°-100°C.

CTEM

In contrast to the continuous film morphology seen with samples 20-1 and 20-
2, sample 20-3 had a discontinuous morphology above the Nb base layer (Figure
3.23). This bottom Nb layer appeared qualitatively similar to that seen in 20-1
(Figure 3.1) and was the close to the nominal 250 nm thickness. The brightfield
image (Figure 3.233) shows that the entire Cu seed layer and ML stack was
broken into island-like regions. The SAED pattern and dark field image confirm
that the ML, with the exception of the capping Nb layer, was nevertheless highly
<111> textured, if not entirely epitaxial. The combined Co/Cu strata of the ML
varied in total thickness by 85 nm. The thinnest regions were 73 nm and the
thicker areas as much as 158 nm, with the mean thickness being 124 nm. The
islands, viewed in cross section, had an average width (along the ML) of 374 nm.
The layering was especially evident in these island regions (Figure 3.230) when
viewed at a high objective lens defocus. The Cu buffer layer in these regions
had three distinct morphologies. In some areas there were anomalously thick

buffer layers. These generally subtended what appeared to be the clearest

74

 

Figure 3.23. CTEM images of sample 20-3. A) Single beam brightfield image
of a cross section of the 20-3 ML. B) The darkfield image obtained with the Nb
(0T0) reflection. C. A single-beam brightfield image focused 11pm below the
minimum contrast setting. Arrows 1-4 highlight a prominently layered region, an
anomalously thick buffer region, a region apparently sans buffer layer and an
apparent hole, respectively.

75

layering (Figure 3.23C). In other areas, often adjacent to the islands, the buffer
layer appears to be absent, with the remainder of the ML in direct contact with
the Nb base layer. In these areas, Fresnel defocus contrast reveals the
persistence of layering in the remaining ML. A third layering variation observed
was the complete lack of contrast and apparent voids subtending regions of
some of the islands. Fresnel contrast also indicates that most of the differences
in total ML thickness is due to local differences in the Cu buffer layer.
HRTEM of sample 20-3

Two regions of interest in this ML were also examined using high resolution
transmission electron microscopy. The first area is of a region showing an .
island-like structure and the second is the area surrounding an apparent hole
(Figure 3.24). Images were recorded near the JEOL 4000 EX Scherzer defocus
(-52 nm) from thin areas that were far enough from the thinned edge to have a
complete ML device cross section remaining. As before, the HRTEM
magnification calibrations were made using Nb lattice images, from areas within
the micrograph. Figure 3.25 is a HRTEM image of the area for analysis of the
ML, taken from a typical island area (Figure 3.24A (inset)). In this region the Cu
buffer layer is 70 nm thick, 3.5 times the nominal value. The rest of the ML ((Coa
nm ICU6 nm) x6) is 76 nm, which is in good agreement with the 72 nm nominal
dimension. Although the ML cross-section had been aligned to within a degree
of the sapphire [0001] in the microscope, structural images were only apparent in

the upper and lower Nb layers and some of the areas of the Cu buffer layer.

76

 

Figure 3.24. CTEM image of the cross-sectioned 20-3 sample. A) An area
within an island. B) The area around an apparent hole.

77

 

 

Figure 3.25. HRTEM image of an island area of 20-3 selected for structural
analysis. Areas A-F are representative areas of the upper and lower Nb layers
and the first and last Co/Cu bilayers.

78

 

 

‘ x. ‘
rinl’llilnl

Figure 3.26. High magnification HRTEM areas and their corresponding FFTs
from areas A-F in Figure 3.25.

79

However, the ML stack showed prominent layer delineation in the region over
the Cu buffer island. This appeared to be due, in large part, to diffraction from
the same intralayer faulting seen in the Co layers in some areas of specimen 20-
2. Figure 3.26 shows the images and FFTs from areas A-F in Figure 3.25.
Although structural images were obtained from the Nb regions above and below
the ML, the only discernable lattices evident in the Co/Cu strata of the ML were
attributable to the (111) planes of fcc Co and Cu. The effect is shown here for
both the first and last bilayer pairs (cf. Figure 3.26 B,C and DE). The buffer layer
was analyzed in four areas (Figure 3.256-J). These were selected to investigate
the most obvious areas with contrast differences in them. The areas selected
were twice the dimensions (6.6 nm X 6.6 nm) of those selected for the bilayer
region of the ML. An area showing anomalous contrast is seen in Figure 3276.
This figure shows a prominent reflection a spacing of 0.20 nm, which is
consistent with the fcc {111} spacing for Co (0.205 nm) and very near that for Cu
(0.208 nm). This spot has a diminishing tail through a rotation of about 20° from
the growth direction. This tail also has a diminishing radius, in the FFT, implying
a dilatation of the lattice spacings with angle from the growth direction. There is
a second spot that corresponds to a lattice spacing of 0.24 nm which is about 4%
less than the {110} spacing for Co and 6% less than the Cu {110}. In addition to
these spots, there is a second weaker set of reflections at 118° from the growth
axis, which may be a second set of {1 10} reflections. However, these lattices
should not be visible along the [110] zone axis. The second inset in the Cu
buffer layer (Figure 3.27H) is from an area very near the Co1 boundary. In the
image, there is a slight periodic modulation in contrast along the growth direction
and the FFT shows the same arrangement of reflections as did area G above.
However, the 0.24 spots are fairly distinct and are on the growth axis. This

implies that the modulation is a moire effect. This is

80

 

 

Figure 3.27. High magnification HRTEM areas and their corresponding FFTs
from areas G-J in Figure 3.25.

81

further supported by the image in Figure 3.27l. The HRTEM shows a clear
periodic contrast modulation and the FFT from this region shows the two
interfering spots and the pair of spots near the center of the FFT which
correspond to the period of the fringes. Figure 3.27.], an area spanning the Cu
buffer||Nb interface shows an interesting transform. The growth direction spots
are fairly sharp and correspond to lattice spacings of 0.22 nm. This is
intermediate between the 0.234 nm Nb{110} spacing and the 0.208 Cu {111}
spacing. At angles of 56° and 117° there are two additional spots corresponding
to the Nb{110} spacing.

In addition to the buffer layer morphology described above, some areas,
under some "islands", appeared to be voids when viewed in CTEM (Figure
3230, arrow 4). One such area was isolated for HRTEM (Figure 3.248). Under
HRTEM conditions, this area can be seen to consist of a hole surrounded by
areas of differing structure (Figure 3.28). Similar moiré patterns are seen, near
the hole, as were seen in the previous island Cu buffer layer HRTEM image. In
Figure 3.29A, a reference area from the C01 layer is presented. It shows the
common growth-direction lattice fringes and the FFT confirms a 0.20 nm spacing
consistent with the {111} growth of the COfcc. The area in Figure 3.28 inset (B) is
enlarged in Figure 3.298. It has a contrast intermediate between the Co1 layer
and the hole and both the image and the adjacent F FT show no indication of
periodic structure. The structure seen in Figure 3.29C is similar to that seen in
the middle of the Cu buffer layer in the island. There are a series of moiré fringes

arising from reflections having the same spacing and on-axis orientation as that

82

 

Figure 3.28. A HRTEM image of a Cu buffer layer hole and surrounding area
from sample 203.

83

.I / I"
Mil llll‘

 

Figure 3.29. High magnification HRTEM images and their corresponding FFl's.
from areas around the hole shown in Figure 3.28

84

seen in Figure 3.26. The area in Figure 3.28 inset (D) has 5 pairs of lattice spots
and a set of low frequency moire spots. Along the growth direction, the 0.21 Cu
{111} spacing is evident as is a coaxial inner pair, corresponding to a lattice
spacing of 0.24 nm. A similarly spaced pair is evident at 107° from the growth
direction. At 55° from the growth axis, there are two more spot pairs. These are
at 0.20 nm and 0.18 nm. These are similar to the Co {111} and Cu
{200}.spacings respectively. The order disappears in the image near the margin
of the hole as shown in Figure 3.29E and along the buffer layer Nb boundary
(Figure 3.29F)
Summary for sample20—3

Sample 20-3, the second epitaxial sample studied, differed markedly from the
first epitaxial sample (20-1). The most striking feature is the presence of a
grossly discontinuous Cu buffer layer. This layer appears as a series of islands
across the Nb base layer. The Nb base layer was, in fact, the only feature even
qualitatively similar to sample 20-1. The Cu buffer layer islands were variable.
Some of the islands had holes, or apparent holes, that seem to have formed after
the start of ML fabrication, since there are areas where a continuous ML can be
seen following the hole contour. In the Cu of both the continuous islands and
those with holes in the Cu buffer layer, numerous moiré fringes were seen.
These fringes appear to arise from the overlap of crystals not oriented with their
closest packed planes along the growth axis. Around island areas with holes, the
crystalline Cu buffer layer appeared to taper down towards the hole margin,

finally appearing amorphous at the lateral margin of the holes. At the upper

85

(Cu||Co interface) and lower (Cu||Nb interface) margins of the hole diffuse layers
of intermediate contrast, with no apparent lattice structure are seen. These
interfacial features are not seen in the island areas with continuous buffer layers.
The ML stack in sample 20-3 showed a interlayer contrast that appears to
arise from the same COfcc -Colmp strain relations that gave the interlayer contrast
in 20-2; that is a tendency towards reversion to the hop Co structure. However,

the ML region analyzed was not optimally oriented for full structural analysis.

86

3.4 Sample 20—4 (epitaxial)

The last sample from the studied sputtering run was fabricated with yet a
different temperature regimen. As shown in Table 2.1, the Nb base layer was
applied at an elevated temperature and given post deposition annealing identical
to samples 20-1 and 20-3 described above. The Cu buffer layer was also
applied at an elevated (450°C) temperature. However, in contrast to the second
epitaxial test (20-3), the C01 layer was deposited at 490°C, rather than the 90°-
100°C regimen of the previous sample. The remainder of the ML was deposited
at 90°-100°
CTEM of sample 20-4

Sample 20-4 had a discontinuous morphology above the Nb base layer, like
sample 20—3. The Nb base layer was similar to that seen in the other two
epitaxial samples (20—1 and 20-3), but measured 180 nm thick. The brightfield
image (Figure 3.30A) also shows that the entire Cu seed layer and ML stack
were again broken into islands that had a mean length of 257 nm when viewed
in cross-section. The SAED pattern (of. insets Figure 3.23 and 3.30) was
essentially the same as that recorded for the previous sample and the darkfield
image shows similar defect contrast (Figure 3.308). There were some lapses in
continuity in the layers of this sample, that resulted from faults that passed
through the entire ML to the Nb layer surface. The total ML thickness averaged
106 nm with a maximum of 122 nm and a minimum of 0 nm (at the trans-ML
faults). As with sample 20-3, there were distinct buffer layer morphologies. In

some of the island structures, a dilated buffer layer was present.

87

 

Figure 3.30. CTEM images of sample 20-4. A) Single beam brightfield image
of a cross section of the 20-4 ML. 8) The darkfield image obtained with the Nb
(0T0) reflection. C) A single-beam brightfield image focused 11pm below the
minimum contrast setting. Arrows 1-3 highlight: A region apparently sans
buffer layer, an apparent hole, and a prominently layered region, respectively.

88

In other areas (Figure 3.30A, arrow 2) holes were present. In yet other areas,
the ML had no apparent buffer layer. Imaging at high defocus (Figure3.30C)
revealed layers, that were the most uniform on the broader islands (Figure
3.30C, arrow 3), but present throughout the ML. The capping layer appears as
an apparently amorphous film, rather than the expected 100 nm Nb layer (cf.
Figures 3.23-3.30).
HRTEM of sample 20-4

An area from a typical island in this ML was examined using HRTEM. As
before, images were recorded near the JEOL 4000 EX Scherzer defocus (-52
nm) from thin areas that were far enough from the thinned edge to have a

complete ML cross-section remaining. The HRTEM magnification calibrations

_ , were made using Nb lattice images contained this area. The F FT analyses were

done using 3 to 6 nm scaled areas from constant magnification micrographs.
Figure 3.31 is a HRTEM image of the area for analysis of the ML, showing the
typical island area from which it was taken. In this region the Cu buffer layer is
43 nm thick, roughly twice the nominal value. The exact location of the Cu buffer
layer 001 interface is difficult to discern. The rest of the ML ((Coe nm /Cu5 r.m) x6)
is 81nm, which is in good agreement with the total ML thickness seen in sample
20—3 nm and in reasonable agreement with the expected thickness of 72 nm.

The ML cross-section was aligned to within a degree of the sapphire [0001] in the
microscope and structural images were apparent in all layers. Figure 3.32 shows

the HRTEM analysis area and the sampling areas within it insets (AM). Good

89

 

 

Figure 3.31. Low magnification HRTEM image of the cross-sectioned 20-4
sample. The area within a typical island chosen for structural imaging is
outlined.

90

MIDI) III II. (II

 

Figure 3.32. HRTEM image of an island area of 20-4 selected for structural
analysis. Areas AM are representative areas throughout Co/Cu bilayers.

91

N l

74/101111“

 

Figure 3.33. High magnification HRTEM areas and their corresponding FFTs
from areas A-F in Figure 3.32.

92

structural images were obtainable from representative areas across the ML. All
areas in the ML oriented to display ordered structure showed the sharpest
frequency along the (111) planes of fcc Co and Cu.

The topmost layer, which should have been the 100 nm Nb capping layer,
appeared as a thin (5 to 6 nm) apparently amorphous layer (Figure 3.33A). The
Cu6 layer (of the last deposited bilayer pair) grew as a single crystal at least
across the range of the HRTEM area viewed ( = 125 nm). The C06 layer, of this
last bilayer pair, showed a transform that indicated the presence of hcp Co
towards the center of the layer in some regions. The spacing in the growth
direction appears the unaffected, but this may be below the resolution level of the
sample size used. The Cu5 layer was clearly fcc but the area sampled was
oriented as if twinned on the [111] growth axis compared the Cu6 layer (Figure
3.33D). In the 005 layer, the arrangement of atomic layers is more complicated
Figure 3.33E). There is a strong growth-direction spot with the expected spacing
of 0.21 nm. At 55° to this is a very weak spot corresponding to 0.18 nm which is
expected for a {200} plane spacing. At 72° there is another spot at the 0.21 nm
spacing, which implies that a twin is imaged in the field of view. Finally, at 78°
there is a pair of spots that correspond to a spacing of 0.25 nm, which is
approximately the {110} spacing for Cu or C0, fcc. The angular relationship and
existence of this frequency in the zone viewed are less clear. The Cu4 layer,
excerpted in Figure 3.33E, shows the identical structure to that seen in figure
3.33D, except that the overall stacking has changed from the abcabc to the

cbacba (a simple twin relationship).

93

This pattern of nearly equilibrium Cu layers and more disordered Co layers is
also seen in bilayers 3 and 2 (Figures 33.3H-K). Layers Cu1 and C01 are hard
to distinguish. The areas shown in Figure 3.33 L and M appear nearly uniform
when viewed in the context of the whole HRTEM area (Figure 3.32). However,
even in the Co1 layer (which was deposited at 450), the HRTEM image and FFT
(Figure 3.34M) show there are lattice images and frequencies inconsistent with a
completely fcc structure. In addition to the prominent growth direction lattices
which have spacings (0.20 nm) exactly consistent with the Com {111}, there is an
additional pair similarly spaced at 70° and a pair at 125° spaced at 0.18 nm. This
implies a prefect {111} fcc growth motif for the Co1 layer. However, there is an 5
additional set of spots 102° from the growth direction (38° from the (1 1T) spot).
These correspond to a lattice spacing of 0.22 nm which is consistent with the
{1 0T0} spacing in the Co.1CD motif, although the angles from the growth direction
differ by 8 degrees. Thus it appears that in both the epitaxial and polycrystalline
Co||Cu layers 6 nm thick exhibit significant hcp character.

HREDS of sample 20-4

To determine local elemental composition in this sample, HREDS spectra
were collected from selected areas of the ML and over the whole layer stack, as
was previously shown for sample 202. The results of this acquisition are shown
in Figure 3.35. The Cu K, /Co K,,, ratio differed from the theoretical (of. Figure
3.118) by less than 6%. In addition to the overall spectrum, attempts to collect

spectra from individual layers and regions in a representative island showing a

94

o" ‘
H..." .
‘.fi‘ "1‘-

 

Figure 3.34. High magnification HRTEM areas and their corresponding FFTs
from areas G-M in Figure 3.32.

 

Cu L peaks

/

Co K0:

Counts

 

 

 

CuKa

 

 

Figure 3.35. An EDS spectrum collected from the whole ML cross-section of
sample 20—4. The collection time and beam parameters were essentially the

same as those used previously.

 

2000

1800‘

1600-

1400''

12m d
Nb 8: Mo L peaks

 

Counts
§

 

600 u
A! K Cu and Co L peaks

  

 

 

Ire-V

kaa

MOKu MOKB

Figure 3.36. An EDS spectrum from the top amorphous layer covering the ML.

96

 

 

combination of the hole and solid structure were made. Figure 3.36 shows the
spectra of the amorphous layer above the MLs in an island having a hole in the
Cu buffer layer. This layer was structurally identical to the area seen in Figure
3.32 (inset A). The spectrum showed mostly Nb in spite of the generally
structureless character observed.

Individual layers in the ML were hard to discern on the viewing screen when
the beam convergence angle and spot size were in the analytical mode. To
properly position the probe on specimen, the image (objective lens) needed to be
near focus and there were few visual clues to the position within the ML.
Nevertheless, representative spectra were recorded for a bilayer pair (5) near the
top of the ML (Figure 3.37). The upper (Cu) layer in the pair shows some Co
content in the sampled volume, but this may be due to the afore mentioned
difficulties in aligning the probe in the middle of a Cu layer. The lower total Cu
counts may be due to slightly different beam geometry or an effect of the
reduction in sample thickness in the beam direction towards the top of the ML.
The Co layer below the Cu layer shows a spectrum rich in Co and having a small
amount of Cu in it.

A final area tested with HREDS was near the margin of a Cu buffer-layer
hole. The structure probed is visible towards the margin of Figure 30.3A (arrow
2). The structure of this hole was essentially the same as that seen for sample
20—3 (Figure 3.28). The EDS spectrum (Figure3.38) shows a nearly pure Cu
buffer-layer with traces of Al, 0, Co, Nb, and Mo. These were most likely re-

deposited from elsewhere on the sample during the ion milling process.

97

 

3500
A Co K

s:
In

3000 .

2500 a

2000 t

Counts

1500 -

1000 *

 

 

 

 

 

key

Figure 3.37. EDS spectra from adjoining Cu and Co layers near the top of the
ML. A) A Cu layer and B) a C0 layer.

98

 

4300

3500 1 Cu L peaks

W /

CuK

 

1°00 ‘ Co L peaks

500 4 Nb 8. Mo L peaks Co KB Cu KB 13 OKB
Co KCl \ \ MoKa\ \

0 1 2 3 4 5 6 7 8 9 10 11 12 13 I4 13 16 17 18 19 20
keV

Figure 3.38. An EDS spectrum from the Cu buffer layer of sample 204.

 

 

 

 

Summary for sample 20-4

Sample 20-4, the final epitaxial ML sample studied, also differed markedly
from the first epitaxial sample, 20—1. As with sample 20-3, the presence of a
grossly discontinuous Cu buffer layer, appearing as a series of islands across the
Nb base layer, is its most striking feature. The Cu buffer layer islands were
variable within the sample and very similar to those seen in 20-3.

The ML stack in sample 20-4 showed interlayer contrast similar to but less
pronounced than that seen in that sample 20-3. However, in sample 204,
structural images were seen throughout the ML. The Cu appeared to alternate in
stacking order on either side of a C0 layer and all layers appeared to be growing

on the {111} planes with occasional stacking faults evident. All of the Co layers

99

displayed lattice images, especially near the middle of the layers, where the
growth appeared to switch to an hcp stacking. Thus, much of the layer contrast
appears to arise from the same Com-Col,Cp strain relations that gave the
interlayer contrast in 20-2, and 20-3. HREDS of the sample showed that the
overall composition is consistent with the modeled ML system. The topmost Nb
(capping) layer'was not well formed. It appeared as a thin (5-6 nm) amorphous
layer that was mostly Nb but also contained significant Cu.

3.5 Samples 19-3 and 19—4 (epitaxial bilayers)

In an attempt to further understand the events that lead to the anomalous
structure of the ML devices, two Cu/Co bilayers, prepared as initial test
examples, were also prepared for cross-sectional TEM analysis. These samples
consisted of a single layer of Cu, 20 nm thick and a single layer of Co, also 20
nm thick, deposited on a Nb epitaxial layer (Table 2.1). In both samples, the Cu
layer was deposited at an elevated temperature (350° and 470° C for 19-3 and
19—4 respectively). In sample 19-3 the Co layer was deposited at 450°C while in
19—4 the Co was deposited at 100°C.

CTEM of samples 19—3 and 19-4

In both of these samples, as with the epitaxial ML samples, the Nb base
layer grew as a continuous crystal (Figure3.39), with a high defect density. The
thickness of this layer was 239 nm in sample 19-3 and 165 nm in sample 19-4.
Cross sections of both samples showed an island-like morphology. The upper
layer of Co, which grew over the Cu islands and the intervening surface, was

within 5% of 20 nm thick in both cases and nearly continuous. The islands in 19-

100

3 covered 62% of the observed surface and were approximately uniform in height
with edge angles of = 45° from the Nb surface normal. The Cu layer height was
49 nm (Figure 3.39A). The islands in 19-4 covered 69% of the observed surface
and the nominal Cu layer height was 62 nm with island edge angles of === 60°
from the surface normal. In both samples islands were seen that varied in
internal contrast, with some appearing to have voids or amorphous areas in
either part or all of the island (similar to 20-3 and 20-4). At the base of the
islands a thin (z 5nm) uniform layer is visible in the CTEM micrographs ( 9.9.
Figure 3.39). A typical island, having a region of nearly continuous Cu with
apparent "holes" included in the island, was chosen for HRTEM analysis (Figure
3.39C). The island was 350 nm along the plane of the bilayer and stood 75 nm
above the Nb base layer.
HRTEM of sample 19—4

Three zones were chosen for analysis within the island structure chosen.
Figure 3.40 illustrates the area outlined as inset (1) in Figure 3.39C. At this
magnification, the "hole" area can be seen to be a continuous thin film. In
addition, the Co layer on the top of the island and that covering the Nb adjacent
to it can be seen to have markedly different dominant crystal defects. Six
representative areas were selected from this side of the island, These areas and

their FFTs are presented in Figure 3.41.

101

 

Figure 3.39. Brightfield CTEM cross-section images of samples 19—3 and 19-4.
A) Sample 19-3. 8) Sample 194. C. An island structure from sample 19-4
showing three regions analyzed via HRTEM.

102

The areas selected were 512 X 512 pixels (about 13 nm square) to capture short
range differences in order. Figure 3.41A corresponds to the top Co layer on the
island. The transform shows the very prominent lattice repeat at 0.20 nm , which
is consistent with the fcc {111} or the hcp {0002} spacing. There is a second pair
of spots, 10° from the growth direction at the same spacing, perhaps due to a
second crystal or low angle boundary being captured in the field analyzed. There
is a third set of spots at 128° from the growth direction. This is roughly three
degrees from the angle (125.26°) where the {200} lattice spots should lie if the
crystal were fcc. However, these spots also correspond to a 0.20 nm spacing
and thus do not appear to be from an fcc arrangement. In Figure 3.418, the
HRTEM image shows isolated areas where lattice images are visible. The FFT
shows that there is considerable layering in the growth direction with spacings
consistent with either Co or Cu {111}. At 150° to the axis there is a set of weak

spots that correspond to the Cu (or COfcc) (002) planes. The area at the ”gap”

near the outer base of the island shows a transition from the Co structure seen in
the Co film distant from the islands and the apparently less structured interior of
the island. In the Figure 3.41C the lattices of the Co film appear to fade to an
amorphous background towards the island center. There are three discernable
pairs of spots in the FFT. On the growth direction, there are the 0.20 nm (111)
spots and at 125° there is a doublet of weak spots corresponding to lattice
spacings 0.19 nm and 0.21 nm. The latter spacing is consistent with the Cu
(111) but the 0.19 nm is further enough apart than the {200} that it doesn't seem

to be a misidentification of these planes. Near the top of the inside of the island

103

 

Figure 3.40. HRTEM overview of the “inset 1' area of Figure 3.39C. Six
representative areas were selected for lattice image analysis.

. . I l
l‘l mm 74111111111 “

 

Figure 3.41. High magnification HRTEM areas and their corresponding FFTs
from areas A-F in Figure 3.40.

105

there is an abrupt change in contrast below the Co layer. This is shown in detail
in Figure 3.41D. The FFT of this area shows a strong 0.20 nm growth direction
spacing with a second coaxial set of spots corresponding to a 0.247 nm lattice.
At 125° to the growth direction, there is a set of weak spots at 0.18 nm, the CUfcc
{200} spacing. At the boundary of the light, apparently amorphous area and the
darker area below the Co layer topping the island, the boundary displays a band
of apparently randomly oriented nano-crystals. There are two rings of spots, one
with strong intensity in the growth direction at 0.20 nm and a second with a
spacing of 0.25 nm. These latter spots correspond to {110} fcc spacings for Cu
or C0. The final area tested on the left side of the island (Figure 3.40F)was the A
band of transition contrast between the bottom of the island and the Nb base
layer. No lattice fringes were evident in the HRTEM image and the transform
was accordingly diffuse.

A sample from the mid-section of the island was also analyzed via HRTEM.
The section (inset (2) from Figure 3.39C) is presented below. Three areas, on a
transect from top of the ML to the Nb surface, were selected for lattice image
analysis and are shown in Figure 3.42. The first area (inset A), which is the Co
layer at the top of the island, shows the 0.20 nm spacing in the growth direction.
A second set of spots at 57° to the growth direction, also at the 0.20 spacing is
also visible. This second set is within about 2° of being consistent with the Corcc

structure and 3° from the hcp {0111} position. The (002) spots are not seen.

However, their spacing is right at the resolution limit. Two crystals separated by a

thin structureless band appear further towards the Nb base layer (inset B) of

106

Figure 3.42. The bottom part of the inset displays lattice fringes in the growth
direction and the FFT shows a pair of spots perpendicular to the growth direction
at a spacing corresponding to a 0.25 lattice spacing, close to the {110} spacing in
Cu or Cofcc. The third area analyzed, just above the continuation of the
structureless band described in Fig 3.41 F, again showed what appears as
multiple crystal forms. In this case they may be either interlocking or
overlapping. The F FT of the high magnification inset C of Figure 3.42 shows the
expected 0.20 growth direction spacing. It also'shows a pair of spots at 62° from
the growth direction corresponding to a lattice spacing of 0.23 nm, the Nb {110}
spacing. At 66° there is a third, stronger set of spots that corresponds to the
0.20 growth direction spacing. Finally, at 97° there is a set of spots arising from a
lattice spacing of 0.25 nm, close to the fcc {110} spacing for this system. On the
right-hand side of the island, the junction between the island and the Co layer
deposited directly on the Nb appeared different from the overall morphology seen
on the left-hand side. Figure 3.43 shows a HRTEM overview of this region. On
this side, rather than being continuous with the Co layer on the top of the island,
the Co layer appears to continue under the edge of the island. There appears to
be a gap at the edge that was continuous of the left side (of. Figure 3.40). Six
areas were selected for HRTEM lattice analysis, as was done with the previously
examined edge. In the Figure 3.44A, an area in the Co layer atop of the island is
shown. The shows strong contrast perpendicular the growth direction with the

expected 0.20 nm lattice spacing.

107

 

Figure 3.42. HRTEM overview of the “inset 2" area of Figure 3.39C. The three
inset areas and their FFTs are also shown.

108

This motif is continued in Figure 3.448 where the strong growth direction layering
prevails. However, significant regions exist in these areas that show a second
lattice. These are especially obvious in the upper right-hand corner (towards the
outside if the island). The orientation (125° from the growth direction) and the

spacing as shown in the FFT are consistent with the Cu (002) lattice. In the

"gap" area (Figure 3.440), the enlarged micrograph appears to be nearly
amorphous. However, superimposed on this background are lattice images in
the growth direction. The FFT shows two spots. The higher frequency
corresponds to the {111} of the fcc Cu or C0. Interestingly the inner spots arise
from a lattice, oriented nearly in the same direction, with a spacing of 0.23 nm. 5
This is the {110} spacing in Nb. Towards the interior of the island (Figure 3.44D)
the HRTEM image again appears to be nearly amorphous. However, careful
observation of this image and the resulting FFT show a set of lattice planes
consistent with Cu in the (111) growth orientation. The HRTEM image from near
the Nb boundary (Figure 3.44E) shows a number of small lattice areas in at least
three orientations. Reference to the FFT from this area shows growth direction
lattices consistent with Nb and Cu growing on their (110) and (111) planes
respectively. The second major set of lattices shows spots oriented 125° to the
growth direction. The inner set corresponds to the 0.20 nm {111} spacing while
the outer (0.18 nm) matches the {200} Cu spacing. Careful inspection of the FFT
shows two sets of weaker spots. The first is seen at 40° to the growth axis and

arises from a spacing of 0.33 nm this is reasonably close to

109

a.
... as

1w.
u.

A.
as
t
.

1...
he?
.1.

by.
u

.....
.

C

as.

A

 

inset 3' area of Figure 3.39C. Six

representative areas were selected for lattice image analysis.

Figure 3.43. HRTEM overview of the

110

 

Figure 3.44. High magnification HRTEM areas and their corresponding FFTs
from areas A-F in Figure 3.43.

111

Nb bcc lattice constant, 0.331 nm. The second weak set occurs at 70° from the
growth axis and the 0.26 nm spacing they represent is a good fit to the 0.255 nm
Cu {110}. The final area analyzed was at the edge of the Nb layer, where it
appears to have grown over the quasi-amorphous layer that spans the island at
the Nb interface. The HRTEM image shows some moire fringes implying
overlapping lattices. In this case, they appear to be rotational in nature. The
typical (111) growth direction planes are seen as is a second set of similarly
spaces lattices at 102° from the growth direction. At 125° a third set of spots can
be seen in the FFT. These correspond to lattices with a 0.18 nm spacing, close
to the Cu {200} planes.

HREDS of sample 19-4

To determine if the local elemental composition and distribution matched that
implied by the morphologies, spot mode HREDS was employed using the same
operation protocols as before. Figure 3.45 shows an EDS spectrum of the top

(striated) Co layer on the top of an island in sample 20-4. It shows a nearly pure
Co composition. Similarly, the central parts of the island in both the "denser" and
"lighter" areas were nearly pure Cu. (Figure 3.46). Figure 3.47 is a
representative spectrum from the nearly transparent region seen near the gaps
at the lateral margin of the islands. Again it shows a nearly pure Cu
composition, but very little mass. The amorphous area seen at the interface of
the island Cu layer and the Nb base layer (e.g. 3.40F) is composed of a low

density Cu-Nb mix as the spectrum in Figure 3.48 shows.

112

 

 

 

 

 

 

 

 

6000
50001
' Coand Cu L eaks
p CoK0L
wool
41
530004
9
O
zomd
CoK
10004 /13
Mo L peaks Cu Ka Cu K MoK
AIK / / B \fl
_ / MoKa
0'1 . Y A. . . . v v r 1 1‘. v
01234567891011121314151617181920

Figure 3.45. HREDS spectrum from the Co layer on an island from sample 19-4.

5000

 

A B

4500‘

C K peak

/ C K peak

Cu and Co L peaks

4000‘

Cu and Co L peaks

200°. CuKa

 

 

 

 

 

 

 

 

O 1 2 3 4 3 6 7 8 9 O 1 2 3 4 5 6 7 O 9 10
M

Figure 3.46. An HREDS spectrum from the thicker (A) and thin (8) middle areas
of the Cu layer on a island from sample 19-4.

113

 

 

 

 

 

4500

4000 -

mo. ypeak

3000]
520004
3 Cu and Co L peaks

2000 «

1:00 a

1000 -

50° 1 l Nb & Mo L peaks Cu K01 CU KB

04 VA 1 Y T 1 VJ $ V I Y I f I W I U
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

keV

Figure 3.47. An HREDS spectrum from the "gap" area in an island from sample
19-4.

 

3300

3000‘ Cerak

/

‘3 Cu and Co L peaks
0

Nb & M0 L peaks
MoK0L

 

500 i OK NbKa

1 Al K C” K0 C" K];
| "DEE
0 'I v v v : A; . 4 . '7‘ . ‘ .‘T‘ 5"“ t ‘
0 3

I 2 3 4 8 6 9 10 11 12 13 14 15 16 17 18 19 20
100V
Figure 3.48. An HREDS spectrum from the Cu||Nb interface at the base of an

island from sample 19-4.

 

 

 

114

Summary for samples 19-3 and 19-4

Samples 19—3 and 19-4 were designed to be simple 20 nm bilayers of Cu with
Co on top. Uniform structure like this was nowhere evident in either sample.
The island morphology seen in these samples was similar to that seen in the
epitaxial samples. The holes seen in the epitaxial samples were not seen.
However, the copper content of the electron Iucent island margins was
considerably reduced. In contrast with the epitaxial MLs, there was a continuous
amorphous layer that subtended the entire island. This type of formation was
only seen at the upper and lower margins of holes in the ML samples. The Co
layer on top of the islands appeared to have the same striated contrast
mechanisms seen in other Co layers in contact with Cu. However, this was not
seen in the Co layer that grew on the Nb directly, implying that it assumed its hcp

structure.

115

CHAPTER 4. DISCUSSION

4.1 TEM of Multilayer Cross-sections

One of the first questions to be considered, in any TEM study, is the fidelity of
the data with respect to the actual specimen. Regardless of the material being
studied, the recognition of artifacts in TEM images is always a problem. TEM
image artifacts can be broadly grouped into those arising from specimen
preparation and those arising from TEM operation. In the first group, artifacts
can be generated in the initial sample-isolation process, by oxidation or
contamination. As the routine sample preparation continues, large free surfaces
are generated and elevated temperatures may be required to cure bonding
agents. These factors may induce chemical or diffusional changes that alter the
structure and thus the TEM image. Thinning the sample requires mechanical
abrasion that can cause deformation artifacts and ion milling that can cause
changes in the projected structure due to heating, surface amorphization, and

changes in the surface topography by differential milling,

Imaging in the TEM is a second major area for artifacts to accumulate.
Careful attention to beam geometry, focus and sample tilt are all crucial to the

acquisition of unambiguous images.

One method of containing the artifacts is to carry through controls with the
experimental treatments. The comparison of features on the controls with the
experimental treatment should allow extraction of the artifactual effects (barring a
treatment X artifact interaction). Thus, the inclusion of data from polycrystalline

sample 20-2, which showed structures consistent with previous work done using

116

TEM on GMR MLs, provides a degree of assurance in assigning causality to

results of the various fabrication treatments.

The Co/Cu multilayer system investigated in this study is an important and
relatively mature example of a GMR system. Therefore, it was a good starting
point in developing production protocols for creating epitaxial structures for CPP
GMR investigations. Very similar epitaxial multilayer systems, grown by MBE, on

(1 120) sapphire have been previously been reported [e.g 75]. Thus, a main

consideration in this investigation was to determine the factors effecting the
quality of the final ML product and transferring the approach to a specific
magnetron sputtering system. However, no parametric separation of the causes
of variability in many of the fabrication steps is strictly possible due to lack of
replications. On the other hand, the qualitatively uniform epitaxial growth of Nb
films, the one replicated process step, suggests that the run-to-run variability is

relatively low in this system.
4.2 Growth of the Nb Base layer

Nb contact layers are required for low temperature (4.2K) CPP GMR
experiments. This is due to the high super-conducting temperature of Nb (9.5K)
and its ability to remain super conducting well above the 4.2K experimental
environment temperature even in a strong applied magnetic field. Therefore, for
an epitaxial GMR device to be prepared for a system using Nb superconducting

contacts, the Nb contact (base) layer must also be epitaxial.

117

Epitaxial Nb films

Nb epitaxial films grown in this investigation varied from the nominal layer
thickness of 250 nm by more than 20°/o. However, there were no observed
qualitative differences in the epitaxial Nb layers among the samples, within the
field of view restrictions of the TEM technique. A length of roughly 25 pm (along
the multilayers) near the perforation of ion milled samples, is of "usable"
thickness for TEM. Within this range the Nb base layers appeared to be single
crystals for all epitaxial samples. This is in comparison to the often less than 50
nm Nb grain size seen in cross sectional view of the polycrystalline Nb capping
layer formed atop the epitaxial ML samples (e.g. Figure 3.23). The Nb||sapphire
interfaces (data not shown) in the epitaxial samples show extended regions that
are either coherent or semi-coherent (strain relaxed by dislocations). No
amorphous layer [48] was seen along any Nb||sapphire interface, supporting the

proposition that annealing prior to Nb deposition recreates the sapphire surface.

In the epitaxial samples, the upper surface of the Nb base layer was also
smooth. Previous work [44, 57, 76] has shown that the upper surface of epitaxial
Nb layers, prepared as those studied here, remains smooth to a RMS roughness
of 0.35 nm or better. Similar results were recorded in this investigation.

However in this investigation, the Nb layer was always subsequently coated with
Cu. In some areas the exact upper boundary of the Nb layer was indistinct (e.g.
Figures 3.8, 3.40). This could have been due to either intermixing (chemical
lnterdiffusion) or intercalation (due to local physical roughness) between the Nb
and the subsequently deposited layer. Separation of the two contributions by
HRTEM, if both are present, is not always straightforward [75]. If the
concentration gradient is constant along the interface and the roughness is

isotropic, the two contributions should be separable using a Fresnel focus series,

118

[50,51,77]. It is important to note that the majority of these interfacial areas are
ordered to within a few atomic radii. The two crystals on either side have
characteristically distinct contrasts and there is no appreciable region between

the two showing complete loss of resolvable order.
Polycrystalline Nb films

In an SAED pattern of a circular field covering just the metal layers of the
polycrystalline sample 20-2 (2450 nm) arcs are seen in the vertical orientation of
the Nb {110} frequency ring. These indicate an z20° spread around the growth
direction, implying that some low angle grain boundaries exist (Figure 3.12,
inset). The exact location of these grain boundaries (whether they are in the top
or bottom Nb layer) cannot be determined from the SAED pattern. However,
along the Nb||sapphire interface in this sample, coherent or semi-coherent
regions, at least 50 nm long, were also seen (data not shown) and neither the
base or capping layer crystals are randomly oriented. Thus, even with room
temperature deposition, Nb films grown on annealed sapphire surfaces are highly

textured in the [111] growth direction.

In the case of room temperature Nb layer deposition (sample 20-2), CTEM
(Figure 3.12) has shown that Nb upper surface is also nearly flat. The HRTEM
image (Figure 3.14), interpreted with the interfacial roughness caveat mentioned
above, also shows little surface roughness at the boundary with the Cu seed
layer. This is in contrast to the surface roughness that was seen by AFM with a
uncoated epitaxial Nb upper surface [44,57]. In these studies Nb deposited at
room temperature on preannealed sapphire surfaces showed a 3.2 nm RMS

roughness.

The epitaxial Nb lattice forms a facile but slightly strained continuation of the

Al lattice sites in the (1 120) of the sapphire. This leads to a system of

119

dislocations that accommodate this strain [46]. The samples prepared in this
study should be able to provide a comparison between the textured
polycrystalline strain accommodation and that seen due to mismatch dislocations
in the epitaxial system. The differences in the surfaces resulting from the

dislocations may affect the behavior of Cu on them.
4.3 Cu buffer layers

A 20 nm thick buffer layer of Cu was deposited on the Nb base layer in all of
the ML systems tested. A Cu buffer or seed layer is prescribed to accomplish
two tasks. First, the buffer layer should provide a diffusion barrier between the
Nb contact layer and the ML. The Co (or other ferromagnetic layer) used in the
GMR device might react with the Nb layer if it were deposited directly on the Nb
contacts. In the case of Co and Nb, the binary phase diagram indicates that
some solubility exists and a variety of intermetallic compounds may be formed.
Their phase stability appears to continue to low temperatures (Figure 4.1A). In
the case of Cu and Nb the situation is clearer (Figure 4.18) with no low
temperature solubility. Thus, Cu should be a good choice as a diffusion barrier

for the subsequently deposited Co/Cu GMR device.
Epitaxial Cu buffer layer

Some regions of the Cu||Nb interface from epitaxial samples had an
amorphous layer indicative of an lnterdiffusion zone. In the sample 20-1 buffer
layer, there was an approximately 5 nm wide structure-less zone along the
Nb||Cu interface, in addition to some anomalous ordered inclusions. This type of
amorphous zone was also seen under the Cu islands in samples 19-3 and 4 and
under the buffer layer holes in ML samples 20-3 and 2-4. A similar zone was
shown, but not evaluated, previously for the epitaxial Cu||Nb interface [74].

Occasionally, amorphous interfacial regions have been reported for multilayers of

120

materials that should not interdiffuse. For example, in a Cu/T a multilayer system,
Lee, et. al. [76] found that at 500°C (substantially below the Cu or Ta melting
points) there were 2 nm amorphous regions at interfaces between layers of these
metals. In co-evaporated films in the Co/Cu system, the two materials generally

tend to segregate on heating [77]. Paradoxically, while the direct deposition of

 

2600,‘

‘vv'Iv Vvv V'v Vvvvvvv'vv' >-
.

.

P

.

.

.

.

.

A ~
I

.

.

.

    

2200-?

_ N
1.-.. ....1......

'I'ernperaturc‘f?

 

 

 

 

 

 

 

Nb Atomic Percent Cobalt (To

 

 

Temperature‘ti

 

E

iosots‘tf

e
9"
‘39-.

 

 

30 40 50 60 70 80 90 1011
Atomic Percent Niobium Nb

Figure 4.1. Binary phase diagrams. A) Co-Nb. B) Cu-Nb. Below 800°C there
should be no solubility of Nb in Cu

 

121

the first Co FM layer on to a Nb contact layer could lead to some degree if
lnterdiffusion, this was not supported by the Co||Nb interfaces between the
islands in the bilayer systems in this study (Figure 3.39 and 3.40). Although the
Co was sputtered on the Nb at roughly 30% of Tm for Co the interface between
Co and Nb shows no observable signs of lnterdiffusion. The diffuse Nb||Cu
regions were seen only in some areas of the epitaxial GMR samples and never in
the polycrystalline samples. Thus, the creation of the diffuse region at the Nb||Cu
interface by physical mixing, due to the sputtering process, also seems unlikely

as the same deposition system was used in all cases.

A second reason for using a Cu seed layer is to form a crystallographic motif,
in the growth direction, to facilitate the epitaxial growth of the GMR device. It is
supposed to translate the {110} bcc closest packed planes of the epitaxial Nb
base (contact) layer to the close packed fcc planes of the active ML or spin valve
constituents. The Cu seed layers in this study were deposited under 4 different
temperatures in the epitaxial samples (Table 2.1). These deposition

temperatures resulted in at least 3 different buffer layer morphologies.

In the case of sample 20-1, the Cu buffer layer was deposited at 150°C and
subjected to a 10 minute anneal at 350°C. This resulted in the most uniform
gross morphology of the epitaxial systems. However, areas in the buffer layer
itself exhibited lattice spacings that corresponded to Nb {110} well away from the
Nb base layer interface, and other lattice spacings, inconsistent with any known
equlibrium structures of the constituent elemental species, were also observed.
These anomalous structures may or may not have affected the growth and
performance of a GMR device but, there is reason to doubt that the deposition

protolcol used is the best approach to buffer layer deposition. EDS showed that

122

the multilayer had no unaccounted for constituents. However, the possibility

of Cobe1-x phases cannot be ruled out.

The remaining epitaxially grown samples (20-3, 20—4, 19—3 and 19-4) all
showed gross disturbances in Cu seed layer confluence (Figures 3.23. 3.30 and
3,39). In these cases, the Cu seed layers had aggregated into what appeared (in
cross section) to be islands with their margins acutely tapered in the growth
direction. In these islands, the Cu buffer layer varied from 2 to 4 times the
nominal thickness. Given the apparent partial covering of the Nb layer by these
islands (e.g. Figure 3.23 and 3.39), the increased height of the Cu buffer layer
under these islands may well account for the mass of Cu deposited. The form
and distribution of these islands is entirely consistent with the plan view AFM
images of similarly prepared 20 nm Cu layers grown on epitaxial Nb base layers

(Figure 4.2) [47].

    

I
, 1. I iii# Hilli,
61:0 20’ $011.0 so 10 2.0 3.0 4.0 5.0
pm pm
Figure 4.2 Atomic force microscope images of 20 nm Cu layers grown on the
surface of 250 nm epitaxial Nb layers. A) Deposited at 350°C. 8) Deposited at
150°C and annealed at 350°C for 10 min. From [47].

In that investigation, a 20 nm Cu film grown on a Nb substrate showed what
appears to be island (Volmer-Weber) growth and that that growth form persists

through 100 nm film thickness at a 350°C. Volmer-Weber growth followed by

123

cluster migration could account for the widely spaced Cu islands seen in sample
194 [34]. In the samples with epitaxially grown Cu-buffer layers, there are no
grain boundaries evident (Figures 3.8, 3.25, 3.32, 3.42). This observation differs
from some epitaxial Cu buffer layers previously fabricated with this sputtering
system [57,74] where twins were found close to the Nb interface. Samples 19-3
and 19-4, were fabricated in the same process run as the sample 19-1 used in
the EBSP analysis of the Nb/Cu/Co system [47]. In that investigation, twins were
seen but could not be quantitated, due to the thinness of the film and dominance
of diffraction from the Nb underlayer. A similar 20 nm Cu||Nb system fabricated

in an MBE system yielded X-ray texture scans that also showed twinning [19].

Polycrystalline Cu buffer layer

In contrast to the Cu buffer layers seen in the epitaxial samples, the sample
prepared at room temperature (202) showed normal layer growth. This included
a strong <111> texture and sharp, ordered interfaces between the Nb base layer
and Cu buffer layer grain sizes consistent with those seen in previous
investigations [54] (Figures 3.20 and 3.21). Low angle (<10°) grain boundaries
were seen in the Cu and Nb layers (Figures 3.12, 3.13), but all grain boundaries
were not continuous through the Cu||Nb interface (e.g. Figure 3.20). Although
only a limited number of crystals were analyzed in the Cu seed layer, no 000
structured Cu was found either in the buffer layer or elsewhere in the ML.
However, considering the intermittent occurrence of columnar bcc non-
equilibrium grains previously reported for Cu [51,54], (and other normally fcc

metals) [54] this finding should be accepted with grave caution.

124

4.4 ColCu multilayers:
Growth of the epitaxial Co/Cu multilayer

The sample 20-1 epitaxial multilayer shows a pronounced overall layer
contrast modulation (Figure 3.2A). Even in this image, taken at near Gausian
focus, the layering is obvious. The most immediate cause of this abnormal and
inverted contrast modulation is the deficiency of Cu in the nominally Cu layers
(Figure 3. 11). While no specific mechanism has yet been derived to account for
this, it is interesting to compare this effect with that seen in the islands formed in
the bilayer samples, 193 and 19-4. In these samples, islands of nearly pure Cu
develop that have lower contrast margins and considerable amorphous
character. It seems unlikely that the Cu should have diffused into the Co layers,
although some Cu was found in all of the Co layers in sample 20-1. Firstly, Cu is
insoluble in Co well above any of the temperatures used in the fabrication of this
multilayer. Secondly, in the case of sample 20-4, the first Co layer was
deposited at a roughly 370°C higher temperature than that used in the Co
deposition in sample 20-1. Even in this temperature regimen, the (greatly
dilated) Cu buffer layer interface with the first Co layer shows no sign of
amorphous material in sample 20-4. Thirdly, the EDS spectrum for the whole
multilayer stack in 20-1 shows afar lower Cu/Co ratio than the simulation of the
multilayer, the sample 20-4 epitaxial sample or the 20-2 polycrystalline sample
(Figures 3.10, 3.11, 3.22, and 3.35). The lower contrast areas in the 20-1
sample, for example those seen near the CoSllCu4 interface, show a widely

spaced lattice structure of undetermined origin, in addition to spacings consistent

125

with the expected {111} of Cu. However, the distinct possibility exists that the
fringes arise from the adjacent Co layer. The tortuous interface between these
Co and Cu layers (Figure 3.6) supports this. The existence of a high degree of
local Co crystal incursions into the Cu layers could give the impression of a much
higher Co content than actually present. A possible structure that is consistent
with the data gathered so far would be some form of amorphous (glass-like)
copper compound (cupric carbonate, for example, can be amorphous [78]).
However, any proposed explanation must account for both the low overall Cu

concentration and the nearly nominal ML dimensions.

In samples 203 and 20-4 the multilayers grown on the top of Cu buffer
islands showed little overall inter-layer contrast (e.g. Figure 3.20). However,
especially in last deposited layers (nearest the Nb capping layer), layers can be
delineated. As can be seen in HRTEM (e.g. Figures 3.32—3.34) this
differentiation is largely due to abrupt changes in crystal orientation between
layers. In the copper layers, there appears to be a tendency for the orientation
along the [111] growth axis to switch stacking between sequential Cu layers.
Additionally, the Co layers tend to become increasingly hcp towards the center of
the layers. In thin Co layers on {111} Cu, the transition from foe to hcp has been
shown to be essentially complete within 5 atomic layers (1.03 nm) [79]. This was
also seen in epitaxial MBE Co/Cu (111) superlattices grown on Nb base layers
on sapphire (1 120) substrates for Co thicknesses of 1.6 to 2.0 nm separated by
2.6 and 2.0 nm Cu layers [73]. Much of the interlayer delineation that is seen

appears to be due to the combined effect of these intra-Iayer stacking changes.

126

Growth of the polycrystalline Co/Cu multilayer

The multilayer stack, (C06 nm/Cue nm) x5, grown on top of the polycrystalline Cu
buffer / Nb layer (sample20-2) showed no overall contrast difference between
individual Co and Cu layers (Figure 3.28). This is as expected since in the
HRTEM (multibeam) objective lens configuration, the differential contrast is due
to high angle (Z dependent) scatter of the beam electrons. Cu and Co, differing
in Z by two, should be nearly indistinguishable in a focused image. The layering
becomes more evident with objective lens defocus, which creates contrast
fringes related to the mean inner potential at the interface of the materials in the
sample [80]. More obviously, the upper layers in sample 20-2 can be
distinguished in some areas in high resolution imaging mode due to layer specific
crystal defects (Figure 3.14). In the room temperature sample, grains in the Cu
layers were consistently observed to be growing on the (111) planes with
extensive areas oriented to allow structural images (Figure3.14). In the Co
layers, extensive in plane twinning was seen in many areas of the field of view.
This appears to have lead to moire patterns in some areas of the Co layers, as
described for Co-Pt multilayers [81], while in adjacent areas, the film appears to
be strictly fcc across an entire Co layer (Figure 3.16). While extensive Corcc
regions have been characterized in sputtered films by EBSP [57], the results
seen here, in both the epitaxial and polycrystalline ML samples, demonstrate that
the stable range of Com can be much less. Interestingly, in the polycrystalline

sample (20-2), the grain column adjacent to the region with the pronounced

127

stacking fluctuations appears as a continuous fcc crystal through an equivalent
vertical range.

In the two bilayer samples, the 20 nm Co layer topping the islands showed
strong striations qualitatively similar to those seen in sample 20-2. However,
there also appear to be some crystals oriented with their close packed planes off
axis (Figures 3.40 and 3.41). Away from the islands, the Co layer showed an

apparent mix of hcp and fcc orientations (data not shown).

Summary:

The growth of epitaxial Co/Cu multilayers on Cu buffer layers via sputter
deposition appears to be tenable. The uniform persistence of fcc Co in epitaxial
layers as thick as 6 nm appears less likely, due to conversion to mixed or the hcp
form a few nanometers from a Cu boundary. As suggested by the phase
diagrams, Co/Cu interdiffusion appears to be insignificant within the temperature
regimens used in this investigation. However, others have reported or suggested
that interdiffusion may occur. This needs to be carefully separated from diffusion
through a film that can also occur [e.g. 25, 82]. Cu/Nb diffusion appears to be
more likely, based on the presence of amorphous interfacial layers seen in some
instances in epitaxial buffer layers. Amorphous layers, of similar appearance,
have been described in Cufl' a systems [76] despite similar indicated
insolubilities. However, the single overriding impediment to success in growing
epitaxial ColCu GMR multilayers, or any GMR device, on epitaxial Nb layers in
the deposition system used herein, seems to be the initial formation of reliable
flat single crystal Cu-buffer layers. Either a reassessment of the thermal
treatments used and/or, perhaps, the choice of Cu, alone, as the buffer layer

element are worth consideration.

128

CHAPTER 5. CONCLUSIONS AND RECOMENDATIONS

Conclusions:

1.

3.

The combination of high deposition temperature and post deposition
annealing of the Nb base layer produces uniform, epitaxial, single crystals
with a high defect density which appears qualitatively different than the room
temperature preparation.

The Cu buffer layers of all of the epitaxial samples showed gross deviations
from the nominal structure and distribution on the epitaxial Nb base layer.
The deviation from the ideotypic deposition took two forms, which may
depend on the use of a post deposition anneal for the Cu buffer layer. Lower
temperature deposition, with post deposition annealing has lead to laterally
continuous buffer layers that appear to have a diffuse interface with the Nb
base layer. The deposition of the Cu buffer at higher temperatures, or the
deposition of layers subsequent to the Cu layer at high temperatures, may
lead to the formation or persistence of buffer layer island growth. No buffer
layer material is generally seen between islands. In some islands, voids may
appear in the Cu buffer layer that are adjacent to less electron dense less
ordered Cu areas.

When grown on islands where a uniform epitaxial Cu buffer layer existed, the
ML growth appeared epitaxial. The elemental composition appeared

consistent with the nominal components and component ratio.

129

4. The first epitaxial sample (20-1) ML structure shows an unaccountable lower

than nominal mass of Cu, despite having layer dimensions consistent with
those of the nominal motif. This is corroborated by the lower than expected
contrast and lack or order in the nominally Cu layers.

The polycrystalline sample (20-2) was entirely crystallographically consistent
with other groups samples of this type. The EDS elemental composition was

also the same as the model for the ML.

Recommendations:

1.

The island growth seen in the 20 nm sputtered Cu buffer layers in the
samples in this study, and other samples produced in the sputtering system,
needs to be eliminated to form extended epitaxial layers. Currently the
individual effects of many system variables cannot be separated and the
cause of island growth is unknown. To separate out the deposition
temperature X Nb epitaxial layer interaction effect, various thickness of Cu
should be deposited on epitaxial Nb at room temperature. Ideally one sample
in this test should also have the same nominal layer dimensions used herein
with all of the layers prepared at room temperature (much like sample 20-2),
but grown on an epitaxial Nb substrate. This would help elucidate the main
effect of the epitaxial Nb surface on subsequent layer growth.

Prior to TEM sample preparation, these samples should characterized via
X-ray diffraction (Mo tube). Low and high angle X-ray diffraction data,
analyzed with the appropriate software [cf. 31] can characterize inter and intra

layer crystallinity. This would provide an expeditious method of determining

130

when further (detailed) TEM studies are warranted and provide a
corroborative method for the TEM analyses. Fractions of these samples
should be subjected to a series of inert environment annealings at least up
the temperature at which the sample 204 encountered. This would be
followed by further X-ray and AEM evaluation to determine crystalline
structure and chemical changes.

The possibility exists to expand these treatments to include a limited number
of additional layer thicknesses .

. For more complicated device designs, it would be useful to routinely deposit
the total layer series used in the differentially masked device, on an adjacent

area of the substrate coupon, for later TEM and XRD analysis if required.

131

[1]

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[9]

LIST OF REFERENCES

A. Barthélémy, A. Fert, F. Petroff (1999) Giant magnetoresistance in
magnetic multilayers In: K.H.J. Buschow, ed. Handbook of Magnetic
Materials V 12 p1-95 Elsevier Science B.V.

N.M. Baibich, J.M. Broto, A. Fert, F. Nguyen Van Dau, F. Petroff, P.
Eitenne, G. Creuzet, A. Friederich, and J. Chazelas. (1988) Giant
Magnetoresistance of (001)Fe/(001)Cr Magnetic superlattices, Phys. Rev.
Lett. 61, 2472-2475.

G. Binasch, P. Grunberg, F. Saurenbach, and W. Zinn, (1989) Enhanced
Magnetoresistance in Layered Magnetic Structures with Antiferromagnetic
Interlayer Exchange, Phys. Rev. B 39, 4828-4830.

S.S.P.Parkin, N. More, and KP. Roche, (1990), Oscillations in exchange
coupling and magnetoresistance in metallic superlattice structures: ColRu,
Co/Cr, and Fe/Cr, Phys. Rev. Lett. 64, 2304.

R. Allenspach and W. Weber (1998), Oscillatory magnetic properties, IBM
Journal of Research and Development, 42 (1) ONLINE

S.S. Parkin, R. F. Marks, R.F.C. Farrow, G.R. Harp, Q.H. Lam, and R.J
Savot (1992) Giant magnetoresistance and enhanced antiferromagnetic
coupling in highly oriented Co/Cu (111) superlattices, Phys. Rev. 8. 46(14)
9262-9265.

M. Suzuki, T. Ohwaki and Y Taga (1997) Durable giant magnetoresistive
sensors using ColCu superlattices, Thin Solid Films 304 (1-2) 333-338.

J.M. Daughton, A.V. Pohm, R.T. Fayfield and CH. Smith (1999)
Applications of spin dependent transport materials, J. Phys. D-Appl. Phys.
32(22): 169-177.

08. Jardine, N.D. Mathur, MG. Blamire and J.E. Evetts (1998) Increased
field sensitivity in Co/Cu multilayers with soft adjacent layers, IEEE
Transactions on Magnetics

132

[10]

[11]

[12]

[13]

[14]

[15]

[16]

[17]

[181

S.S.P.Parkin, (1991) Systematic variation of the strength and Oscillation
period of indirect magnetic exchange coupling through the 3d, 4d, and 5d
transition metals, Phys. Rev. Lett. 67, 3598-3601.

S.S.P.Parkin, R. Bhadra and KP. Roche (1991) Oscillatory Magnetic

Exchange Coupling through Thin Copper Layers, Phys. Rev. Lett. 66,
2152-2155

D. Bisero, G. Bordin, M. Minelli. F. Ronconi, F.Spizzo, A. Baraldi, S. Lizzit,
G. Paolucci, L. Pareti and G. Turilli, (1999) Effects of Atomic Diffusion
Processes in Co-Cu Multilayer Granular Films, NanoStructured Materials
11(6): 769-774.

J. Unguris. D. Tulchinsky, M.H. Kelly, J.A. Borchers, J.A. Dura, C.F.
Majkrzak, S.Y. Hsu, R. Loloee, W.P. Pratt, Jr. and J. Bass (2000)
Magnetic depth profiling Co/Cu multilayers to investigate
magnetoresistance. J. Appl. Phys, 87(9): 6639-6643.

T.S. Plaskett and TR. McGuire (1993) Magnetoresistance in
(Co10A/Cu1OA)n multilayer films as n increases, J. Appl. Physics, 73(1)
6378-6380.

A. Cebollada, J.L. Martinez, J.M Gallego, J.J. de Miguel, R Miranda, S.
Ferrer , F. Batallan, G. Fillion and JP. Rebouillat, (1989)
Antiferromagnetic ordering in Co-Cu single-crystal superlattices, Phys.
Rev (8) 39(13), 9726-9729.

A. Cebollada R. Miranda, C.M. Schneider, P. Schuster and J. Kirschner
(1991) Experimental evidence of an oscillatory magnetic coupling in
ColCu/Co epitaxial layers, J. Magn. Magn. Mat, 102:25-29.

H.A.M. de Gronckel, K. Kopinga, W.J.M. de Jonge, P. Panissod, J.P.
Schillé and F.J.A. den Broeder (1991) Nanostructure of ColCu multilayers,
Phys. Rev. 8., 44(16) 9100-9103.

C.M. Emmerson and T.-H. Shen (1996) A study of ColCu multilayer
growth on Si(111) with silicide buffer layers, J. Magn Magn. Mat. 156:15-
16.

133

[19]

[201

[211

[221

[231

[24]

[25]

[26]

[27]

D. Greig, M.J. Hall, C. Hammond, B.J. Hickey, H.P. Ho, M.A. Howson,
M.J. walker, N. Wiser and DC. Wright (1992) The giant
magnetoresistance of Col Cu superlattices grown by MBE, J. Magn
Magn. Mat. 110:L239-L246.

A.J.R. Ives, R.J. Hicken, J.A.C. Bland, C. Daboo M. Gester, and SJ. Gray
(1994) High-field polar MOKE magnetometry as a probe of interlayer
exchange coupling in MBE-grown ColCu/Co (111) and Fe/Cr/Fe (001)
wedged trilayers, J. Appl. Phys. 75(10) 6458-6460.

J.S. Lord, H. Kubo, P.C. Riedi, and M.J. Walker (1993) Nuclear magnetic
resonance of molecular beam epitaxilly grown Co-Cu superlattices that
exhibit large magnetoresistance, J. Appl. Phys., 73(10): 6381-6383.

W.F. Egelhoff, Jr. and MT. Kief (1992) Antiferromagnetic coupling in
Fe/Cu/Fe and ColCu/Co multilayers on Cu(111), Phys. Rev. 8., 45(14)
7795-7804.

D.H. Mosca, F. Petroff, A.Fret , P.A. Schroeder, W.P. Pratt, Jr., and R.
Loloee (1991) Oscillatory interlayer coupling and giant magnetoresistance
in Co/Cu multilayers, J. Magn Magn. Mat. 94: L1-L5.

M.E. Tomlinson, R.J. Pollard, D.G. Lord and P.J Grundy (1992) Giant
magnetoresistance in sputter deposited Co/Cu multilayer systems, J.
Magn. Magn. Mat, 111: 79-82.

K. Ratzke, M.J. Hall, 0.8. Jardine, W.C. Shih, RE. Somekh and AL.
Greer, (1999) Evolution of microstructure and magnetoresistance in ColCu
multilayers during annealing, J. Magn. Magn. Mat.204: 61-67.

S. Miura, M. Tsunoda, T. Nagatsuka, S. Sugano, and M. Takahashi (1999)
Drastic change of giant magnetoresistance of Co/Cu multilayer by
decreasing residual impurities in sputtering atmosphere, J. Appl. Phys.
85(8): 4463-4465.

R.J. Pollard, M.J. Wilson, and P.J. Grundy (1995) Microstructure and
GMR in (111) sputter-deposited ColCu multilayers, Mat. Res. Soc. Proc.
384 365-370.J

134

[28]

[29]

[30]

[31]

[32]

[33]

[34]

[35]

[36]

[37]

P. Nallet, E. Chassaing, M.G. Walls, and M.J. Hytch (1996) Interface
characterization in electrodeposited Cu-Co multilayers, J. Appl. Physics,
79(9) 6884-6889.

W.F. Egelhoff, Jr., and MT. Kief (1992) Fe/CuFe and ColCu/Co
multilayers on Cu(111): The absence of oscillatory antiferromagnetic
coupling, IEE Trans. Magn. 28(5): 2742-2744.

A.R. Modak, D.J. Smith, and S.S.P.Parkin, (1994) Dependence of Giant
Magnetoresistance on Grain Size in Co/Cu Multilayers, Phys. Rev. (8).
50(6), 4232-4235.

M. Chadek, C. Dorner, A. Buchal, V. Valvada, and H. Hoffmann (1996)
Quantitative in situ X-ray diffraction analysis of magnetic multilayers
during annealing, J. Appl. Phys 80(3) 1437-1445.

H. Zhang, R.W. Cochrane, Y Huai, M.Mao,X. Bian and WE. Muir (1994)
Effect of annealing on the giant magnetoresistance of sputtered ColCu
multilayers, J. Appl. Phys. 75(10): 6534-6536.

H. Kikuchi, J.-F. Bobo, and R.L. White (1997) The effect of pinholes on the
properties of Co/Cu multilayers, IEEE trans. Mag. 33(5) 3583-3585.

D.J. Smith, Z.G. Li, A.R. Modak, S.S.P. Parkin, R.F.C. Farrow and RF.
Marks, (1994) High resolution imaging of magnetic multilayers, Scripta
Met Mat. 30(6) 689-694.

D. Kubinski, M. Parsons and J. Hangas (2000) Magnetoresistive properties

of 1““ AFM Co/Cu multilayers sputtered in an Af‘l’Oz atmosphere, Mat. Res.
SOC. Proc., 6142 F8.8.1F-8.8.6.

M. Ohring (1992) The Materials Science of Thin Films, Academic Press,
San Diego

I.K. Schuller, S. Kim and C Leighton (1999) Magnetic supperlattices and
multilayers, J. Magn. Magn. Mat. 200: 571-582.

135

[38]

[39]

[40]

[41]

[42]

[43]

[44]

[45]

[45]

[47]

C.J. Smithells, E.A. Brandes and GB. Brook (1998) Smithells Metals
Reference Book 7th Ed. Buttervvorth-Heinemann.

C. Gente, M. Oehring, and R. Bormann (1993) Formation of
thermodynamically unstable solid solutions in the Cu-Co system by
mechanical alloying. Phys. Rev. B 48(18) 13244-13252.

P. Wu, E.Y.Jiang, H,L. Bai, H.Y. Wang (1997) Evolution of Microstructure
in Co-Cu Multilayers during Thermal Annealing, Phys. Stat. Sol. (a)
161:389-397

D.J Smith, AR. Modak, T.A. Rabedeau, S.S.P.Parkin, (1997) Growth
and Structural Characterization of Highly Oriented Sputter-Deposited
(111), (110), and (100) Co/Cu Superlattices, Appl. Phys. Lett. 71(11)
1480-1482

C. Mény, P.Panissod and R. Loloee (1992) Structural study of cobalt-
copper multilayers by NMR, Phys. Rev. (B). 45(21), 12269-12277.

J.M. Slaughter, W.P. Pratt, P.A. Schroeder. (1989) Fabrication of Layered
Metallic Systems for Perpendicular Resistance Measurements. Rev. Sci.
lnsts. 60(1) 127-131.

S-F Lee, Q. Yang, P.Ho|ody, R. Loloee, J.H. Hetherington, S. Mahmood,
B. lkegami, K. Vigen, L.L. Henry, P.A. Schroeder, W.P. Pratt, Jr. and J.
Bass, (1995) Current-Perpendicular and Current-Parallel Giant
Magnetoresistance in Co/Ag Multilayers Phys. Rev (B) 52(21), 15426-
15441.

S.M. Durbin, J.E. Cunningham and GP. Flynn (1982) Growth of single-
crystal metal superlattices in chosen orientations, J. Phys. F: Met. Phys.
12: L75-78.

R. Loloee, W.P. Pratt, Jr., and MA. Crimp (2001) Crystallographic
characterization of sputter-deposited epitaxial Nb-Cu-Co and Nb-Cu-

Permalloy multilayers using electron back-scatter diffraction patterns,
Philos. Mag. A., 81(2) 261-273.

V. Vitek, G, Gutekunst, J. Mayer and M. RUhle (1995) Atomic structure of
misfit dislocations in metal-ceramic interfaces, Philos. Mag. A 71(6): 1219-
1239.

136

[48] G. Gutekunst, J. Mayer, V. Vitek and M. RiJhle (1997) Atomic structure of
epitaxial Nb-Al203 interfaces ll. Misfit dislocations, Philos. Mag. A, 75(5):
1357-1382

[49] A. Levay, G. Mbbus, V. Vitek, M. Rilhle and G. Tichy (1999) Structure of
misfit dislocations in niobium-sapphire interfaces and the strength of
interfacial bonding: an atomistic study, Acta Mater. 47(15):4143-4152.

[50] J. Yuan, V. Gupta, and M.Kim, ( 1995) Structure and Chemistry of
Nb/Sapphire Interfaces with and without lnterlayers of Sb and Cr, Acta
Metall. Mater. 43(2) 769-779

[51] H. Kung,Y-C. Lu, A.J. Griffin, Jr., M. Nastasi, TE. Mitchell, and JD.
Embury, (1997) Observation of Body Centered Cubic Cu in Cu/Nb
Nanolayered Composites, Appl. Phys. Lett. 71(15) 2103-2105.

[52] J.N. Ness, WM. Stobbs and TE Page (1986) A TEM Fresnel diffraction-
based method for characterizing thin grain-boundary and interfacial films,
Philos. Mag. A., 54(5): 679-702

[53] PM Ross and WM. Stobbs (1991) Computer modeling for Fresnel
contrast analysis, Philos. Mag. A., 63(1): 37-70.

[54] H. Geng, J.W. Heckman, W.P. Pratt, J. Bass, F.J. Espinosa, S.D.
Conradson, D. Lederman, and MA. Crimp (1999) Occasional “long-range"
nopnequilibrium body-centered-cubic structures in NiFe/Cu spin valves, J.
Appl. Phys., 86(8), 4166-4175.

[55] R.W. Horne and R. Markham (1972) Optical diffraction techniques, In A.
M. Glauert, ed., Practical Methods In Electron Microsc0py, Vol1.,
American-Elsevier Publishing Co., NY

[56] R. Sinclair, R. Gronsky and G. Thomas (1976) Optical diffraction from
lattice images of alloys, Acat Metall. 24: 789-796.

[57] M. Watanabe and 08. Williams (1999) Atomic-level detection by X-ray
microanalysis in the analytical electron microscope, Ultramicroscopy
78:89-101.

137

[58]

[59]

[50]

[61]

[62]

[63]

[64]

[55]

[56]

[67]

PA. Stadelmann (1987) EMS-A software package for diffraction analysis
and HREM image simulation in materials science. Ultramicroscopy
21:131-146. [66]

H. Geng, J.W. Heckman, R. Loloee, W.P. Pratt, J. Bass, and MA. Crimp,
(2001) The Effect of Indium contamination and Nb Contact Layers on the
Structure of Ag/NiFe GMR SpinValves., (Submitted to Mat. Sci and Eng).

R. Loloee, (2000) Growth and Characterization of Epitaxial Films and
Magnetic Multilayers Grown by DC. Sputtering, Doctorial Dissertation,
Michigan State University, E. Lansing, MI.

H. Geng, R. Loloee, J.W. Heckman, J. Bass, W.P. Pratt, Jr. and MA.
Crimp (2000) Structural characterization of epitaxial GMR magnetic

multilayers and spin valves grown by sputter deposition, Mat. Res. Soc.
Symp. Proc. 614: F4.7.1F4.7.6.

 

R. Loloee, (2001) personal communication

J.C. Bravman and R. Sinclair (1984) The Preparation of Cross-Section
Specimens for Transmission Electron Microscopy J. Electron Micro. Tech,
1, 53-61.

D.A. Howell, (1995) Conventional and High-Resolution Transmission
Electron Microscopy Study of Spin-Glass/ Amorphous-Silicon Multilayers,
Masters Thesis, Michigan State University, E. Lansing, MI.

NIST/NIH Desk Top Spectrum Analyzer, available from the National
Institute of Standards and TechnoloSJY. Gaithersberg, MD 20899, USA.
(http://micro.nistoov/DTSA).

.._.
..A‘AAAD. . i

Tit---

D.B. Williams and CB. Carter (1996) Transmission Electron Microscopy:
A textbook for materials science. Plenum Press, NY. 729 pp.

S.M. Allen (1981) Foil thickness measurements from convergent-beam
diffraction patterns, Phil. Mag. A 43(2): 325-335.

138

[58]

[69]

[70]

[71]

[72]

[73]

[74]

[75]

[76]

[77]

SM. Allen and EL. Hall (1992) Foil thickness measurements from
convergent-beam diffraction patterns An experimental assessment of
error, Phil. Mag. A46(2): 243-253.

Z. Horita, K. Ichitani, T. Sano and M Nemoto (1989) Applicability of the
differential X-ray absorption method to the determination of foil thickness

and local composition in the analytical electron microscope, Phil. Mag. A,
59(5): 939-952.

W. Rasband NIH Image (developed at the US. National Institutes of
Health and available on the internet at http://rsb.info.nih.govlnih-imagel)

C. Russ and J. Russ (1996) The Image processing Tool Kit, Reindeer
Games, Inc. Asheville, NC 28801 _

A. A. Reeves (1990) Optimized fast hartley transformation for the
MC68000 with applications in image processing, Masters Thesis,
Dartmouth College, Hanover, New Hampshire.

P. Bodeker, A. Abromeit, K. Brohl, P. Sonntag, N. Metoki, and H. Zabel
( 1993) Growth and X-ray characterization of Co/Cu (111) superlattices,
Phys. Rev. 8. 47(4) 2553-2361.

H. Geng, R. Loloee, J.W. Heckman, J. Bass, W.P. Pratt, Jr. and MA. I
Crimp (2000) Structural characterization of epitaxial GMR magnetic l

multilayers and spin valves grown by sputter deposition, Mat. Res. Soc.
Symp. Proc, Vol 614. F4.71-F4.76.

Hytch, M.G. Walls, and J.-P. Chevalier (2000) Measurement of roughness I
and diffuseness of interfaces, Ultramicroscopy 83: 217-225

H.-J. Lee, K.-W. Kwon, C. Ryu and R. Sinclair (1999) Thermal stability of
a Cu/T a multilayer: An intriguing interfacial reaction, Acta. Mater. 47(15):
3965-3975.

K. Barmak, G.A. Lucadamo, C. Cabral, Jr., C. Lavoie, and J.M.E. Harper
(2000) Dissociation of dilute immiscible copper alloy thin films, J. Appl.
Phys. 87(5): 2204-2214.

139

 

[78]

[79]

[80]

[81]

[82]

P. G. Stecher (ed) (1968) The Merck index; an encyclopedia of chemicals
and drugs 8th ed. Merck, Rahway, NJ.

K. Heinz, S. Milller and L. Hammer (1999) Crystallography of ultrathin
iron, cobalt and nickel films grown epitaxially on copper, J. Phys:
Condens. Matter 11: 9437-9454.

R.E. Dunin-Borkowski (2000) The development of Fresnel contrast, and
the interpretation of mean inner potential profiles at interfaces,
Ultramicroscopy 193-216.

K.M. Krishnan (1999) Magnetism and microstructure: The role of
interfaces, Acta mater. 47(15): 4233-4244.

D.-Y. Shih, C.-A Change, J. Paraszczak, S. Nunes and J. Cataldo (1991)
Thin-film interdiffusions in Cu/Pd, Cu/Pt, Cu/Ni, Cu/NiB, Cu/Co, Cu/Cr,
Cu/T i and Cufl’iN bilayer films: Correlations of sheet resistance with
Rutherford backscattering spectrometries, J. Appl. Phys. 70(6) 3052-306

140

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

iiii]ii]i][]ii[i][i[]ii[i[i[[ii