Thermal Stability of Nanocrystalline Copper for … · Thermal Stability of Nanocrystalline Copper...
Transcript of Thermal Stability of Nanocrystalline Copper for … · Thermal Stability of Nanocrystalline Copper...
Thermal Stability of Nanocrystalline Copper for Potential Use in Printed
Wiring Board Applications
by
Patrick Kai Fai Woo
A thesis submitted in conformity with the requirements
for the degree of Doctor of Philosophy
Graduate Department of Materials Science and Engineering
University of Toronto
Copyright by Patrick K.F. Woo, 2011.
ii
Thermal Stability of Nanocrystalline Copper for Potential
Use in Printed Wiring Board Applications
By Patrick Kai Fai Woo
Doctor of Philosophy, 2011
Department of Materials Science and Engineering
University of Toronto
ABSTRACT
Copper is a widely used conductor in the manufacture of printed wiring boards
(PWB). The trends in miniaturization of electronic devices create increasing challenges
to all electronic industries. In particular PWB manufacturers face great challenges
because the increasing demands in greater performance and device miniaturization pose
enormous difficulties in manufacturing and product reliability. Nanocrystalline and ultra-
fine grain copper can potentially offer increased reliability and functionality of the PWB
due to the increases in strength and achievable wiring density by reduction in grain size.
The first part of this thesis is concerned with the synthesis and characterization of
nanocrystalline and ultra-fine grain-sized copper for potential applications in the PWB
industry. Nanocrystalline copper with different amounts of sulfur impurities (25-
230ppm) and grain sizes (31-49nm) were produced and their hardness, electrical
resistivity and etchability were determined.
To study the thermal stability of nanocrystalline copper, differential scanning
calorimetry and isothermal heat treatments combined with electron microscopy
iii
techniques for microstructural analysis were used. Differential scanning calorimetry was
chosen to continuously monitor the grain growth process in the temperature range from
40C to 400C. During isothermal annealing experiments samples were annealed at
23C, 100C and 300C to study various potential thermal issues for these materials in
PWB applications such as the long-term room temperature thermal stability as well as for
temperature excursions above the operation temperature and peak temperature exposure
during the PWB manufacturing process. From all annealing experiments the various
grain growth events and the overall stability of these materials were analyzed in terms of
driving and dragging forces. Experimental evidence is presented which shows that the
overall thermal stability, grain boundary character and texture evolution of copper is
greatly related to changes in driving and dragging forces, which in turn, are strongly
depended on parameters such as annealing temperature and time, total sulfur impurity
content and the distribution of the impurities within the material. It was shown that a
simple increase in the sulfur impurity level does not necessarily improve the thermal
stability of nanocrystalline copper.
iv
ACKNOWLEGEMENTS
First of all, I would to thank Professor Uwe Erb for his guidance, encouragement and
supervision given to me throughout my academic years starting from my undergraduate
studies to this doctoral research program. I was able to significantly extend my
understanding in the areas of synthesis and characterization of nanocrystalline materials.
I also would like to acknowledge Dr. Yijian Zhou (Integran Technologies Inc.), Dr.
Simon Kim (AECL Canada Inc.), Prof. Karl T. Aust (Professor Emeritus – University of
Toronto), Mr. Ben Yu (SNC Lavalin Canada), Mr. Charles Soong (SGS Lakefield
Canada) and Mr. Charles Kwan for valuable discussions and support concerning many
aspects of my thesis. I also thank Mr. John Calloway for providing laboratory technical
assistance and Mr. Fred Neub and Mr. Sal Boccia for teaching me in operating both TEM
and SEM at the beginning of my research program.
Financial support from the Natural Sciences and Engineering Research Council (NSERC)
of Canada and the University of Toronto Open Fellowship is gratefully acknowledged.
Finally, I would like to give my deepest gratitude to my father Hok Yim Woo for
countless years of financial support, teaching, worry and guidance. His ethics and
enthusiasm toward his job taught me how to be successful in many parts of my life.
v
“Yours, O LORD, is the greatness and the power and the glory and the majesty and
the splendor, for everything in heaven and earth is yours. Yours, O LORD, is the
kingdom; you are exalted as head over all.
Wealth and honor come from you; you are the ruler of all things. In your hands are
strength and power to exalt and give strength to all.
Now, our God, we give you thanks, and
praise your glorious name.”
1 Chronicles 29:11-13 (KJV)
vi
TABLE OF CONTENTS
ABSTRACT ii
ACKNOWLEDGEMENTS iv
TABLE OF CONTENTS vi
LIST OF FIGURES AND TABLES ix
LIST OF ACRONYMS AND ABBREVATIONS xvi
CHAPTER 1 – Introduction 1.1 – Research Motivation 1
1.2 – Research Objectives and Organization of this Report 5
1.3 – References for Chapter 1 7
CHAPTER 2 – Materials Synthesis and Characterization 2.1 – Electrodeposition of Copper 8
2.2 – Electrodeposition of Nanocrystalline Copper 9
2.3 – Electrodeposition of Ultra-fine Grained (UFG) Copper 15
2.4 – Synthesis of Electrodeposited Nanocrystalline and
Ultra-fine Grained Copper 18
2.5 – Synthesis Results 21
2.5.1 – Transmission Electron Microscopy 22
2.5.2 – Chemical Analysis 28
2.5.2.1 – Chemical Analysis by LECO/IRA 30
2.5.2.2 – Chemical Analysis by ICP-ITV 30
2.6 – Materials Comparisons 31
2.6.1 – Microstructure Comparisons (TEM) 33
2.6.2 – Texture/Crystalline Orientation
Comparisons (XRD) 35
2.7 – Mechanical and Electrical Measurements 38
2.7.1 – Hardness Measurements by Nanoindentation 38
2.7.2 – Electrical Resistivity Measurements
Using the 4-Point Probe 41
2.8 – Summary 45
2.9 – References for Chapter 2 47
vii
CHAPTER 3 – Previous Work on Thermal Stability of
Nanocrystalline Materials 3.1 – Introduction 49
3.2 – Thermal Instability of Nanocrystalline Materials 51
3.3 – Thermal Stability of Electrodeposited
Nanocrystalline Materials 53
3.3.1 – Investigation Techniques 53
3.3.2 – Review on Thermal Stability of
Nanocrystalline Electrodeposits 55
3.3.3 – Thermal Stability of Nanocrystalline Copper 65
3.4 – References for Chapter 3 71
CHAPTER 4 – Experimental Methods Used to Study Thermal Stability of
Nanocrystalline Copper 4.1 – Annealing Strategies 72
4.2 – References for Chapter 4 77
CHAPTER 5 – Iso-kinetic Analysis of Nanocrystalline Low-S
and High-S Copper 5.1 – Calormetric Analysis 78
5.2 – Microstructural Evolution for
Nanocrystalline High-S Copper 84
5.3 – Microstructural Evolution for
Nanocrystalline Low-S Copper 93
5.4 – Structural Evolution Differences Between Low-S and
High-S Copper 99
5.5 – Discussion 101
5.6 – Summary of DSC Annealing Experiences 104
5.7 – References for Chapter 5 106
CHAPTER 6 – Isothermal Annealing Analysis 6.1 – Microstructural Evolution at 100°C 108
6.2 – Microstructural Evolution at 300°C 112
6.3 – Microstructural Evolution Differences
Between Low-S and High-S Copper 117
6.4 – References for Chapter 6 125
CHAPTER 7 – Room Temperature Thermal Stability of Electrodeposited
Nanocrystalline Copper 7.1 – Literature Review 126
7.2 – Experimental Results 127
7.3 – Discussion 130
7.4 – Chapter Summary 133
viii
7.5 – References for Chapter 7 135
CHAPTER 8 – Grain Boundary Character Evolution in
Electrodeposited Copper
8.1 – Definition of Grain Boundary Character Distribution 136
8.2 – Types of Grain Boundaries 136
8.3 – Grain Boundary Character Distribution (GBCD)
of a Random Polycrystalline Aggregate 143
8.4 – Properties of Low Σ Boundaries 144
8.5 – Grain Boundary Engineering (GBE) 148
8.6 – Factors Affecting the Formation of Annealing Twins 152
8.6.1 – Stacking Fault Energy 153
8.6.2 – Solute Concentration 153
8.6.3 – Pre-anneal Deformation/Strain 155
8.6.4 – Annealing Condition 155
8.7 – CSL Frequency as a Function of Grain Size 156
8.8 – Experimental Details 160
8.9 – Results and Discussion 163
8.10 – Effect of Grain Size on GBCD 168
8.11 –Texture Evolution 170
8.12 – References for Chapter 8 176
CHAPTER 9 – Conclusions 180
CHAPTER 10 – Recommendations for Future Work 183
APPENDICES
Appendix 1 – Summary of Plating Experiments 185
Appendix 2 – Calculating the Intercrystalline Volume Fractions
as a Function of Grain Size 187
Appendix 3 – Definitions for Normal and
Abnormal Grain Growth 189
Appendix 4 – Calculations to Estimate the Sulfur Concentration [S]
Required to Form a Monolayer of Sulfur
at the Grain Boundary 191
ix
LIST OF FIGURES AND TABLES
Page Fig. 1.1.1
Schematic representation of PWB illustrating location of copper (a) and cracks due to the CTE mismatch effects (b).
2
Fig. 1.1.2 Schematic representation of PWB’s circuit trace after lithography and etching. Grain size reduction potentially enhances trace resolution and achievable wiring density of the PWB [Woo and Erb (2006)].
3
Fig. 1.1.3 Typical lead-free solder reflow profile 4
Fig. 2.1.1 Optical micrographs showing the effect of TIPA addition agent on the structure of copper deposits produced from sulfate type bath. Fig. 2.1.1a: no additives and Fig. 2.1.1b: with 3.5g/L TIPA. Both electrodeposits were plated at 30°C with an applied current density of 50mA/cm
2 [Lamb et al., (1970)].
9
Table 2.1.1 Influence of TIPA addition agent on strength, ductility and electrical resistivity of electrodeposited copper [Lamb et al., (1970)].
9
Fig. 2.2.1 Generalized pulsed current electrodeposition current waveform. 11
Fig. 2.2.2 XRD patterns showing the change in grain size (Dv) with varying citric acid contents [Natter and Hempelmann (1996)].
13
Fig. 2.2.3 Schematic diagram of a deposit surface. Active sites: #2-5; Passive site #1 [Natter and Hempelmann (1996)].
13
Fig. 2.3.1 HRTEM Micrograph of several "nanocrystallites" (A,B,C and D) separated by low angle (1-10°) grain boundaries in as-deposited copper [Lu et al., (2000)].
16
Fig. 2.3.2 a) TEM micrograph of the microstructure in an as-deposited UFG copper sample. b) TEM image showing dislocation pile-ups at twin boundaries after micro-tensile testing [Lu et al., (2004)].
17
Fig. 2.3.3 TEM micrographs showing typical grain structures for commercially available electronic grade copper foils. a: typical 18 µm thick ED Cu foil, b: 25 µm thick ED Cu oil plated with no additives, c: 18 µm thick heavily twinned ED Cu foil, and d: 18 µm thick twin-free ultra-fine grain ED Cu foil. Fig. 2.3.3a, 2.3.3b and 2.3.3d - [Merchant et al. (2004)] and Fig. 2.3.3c - [Weil and Hong (1996)].
18
Fig. 2.4.1 Schematic representation of the pulse current electrodeposition setup. 19
Table 2.4.1 Bath compositions for the electrodeposition of nanocrystalline low-S, nanocrystalline high-S and ultra-fine grained copper.
21
Table 2.4.2 Bath and selected operating parameters for the electrodeposition of nanocrystalline low-S, and ultra-fine grained low-S, nanocrystalline high-S and pure polycrystalline copper.
21
Fig. 2.5.1 TEM dark field micrographs, diffraction patterns and grain size distribution of high-S nanocrystalline deposit characterized at center (left) and corner (right) region of the
24
x
sample. Fig. 2.5.2
Corresponding TEM bright field (first row), dark field (second row) and diffraction patterns (third row) micrographs for nanocrystalline low-S (left), nanocrystalline high-S (center) and UFG low-S (right) copper samples.
25
Fig. 2.5.3 Steps involved in determining the average grain size (AGS), grain size distribution
(GSD) and cumulative volume fraction (CVF) of the microstructure. 26
Fig. 2.5.4 Grain size distributions (GSD) for nanocrystalline low-S (blue), high-S (yellow) and low-S UFG (red) copper.
27
Table 2.5.1 AGS, SD and ALGD for nc low-S, UFG low-S and nc high-S copper electrodeposits.
27
Fig. 2.5.5 Cumulative volume fraction plots for nanocrystalline low-S (blue), nanocrystalline high-S (yellow) and UFG low-S (red) copper.
28
Fig. 2.5.6 Schematic diagram of the ITV setup [Karanassios and Greib (2006)]. 30
Table 2.5.2 Results obtained from LECO/IRA analysis. 30
Table 2.5.3 Results obtained from ICP-ITV analysis. 31
Table 2.6.1 Summary of electrodeposited copper samples made in this study. 32 Fig. 2.6.1
TEM DF micrographs, average grain size and standard deviations for: nc high-S (top-left), UFG low-S (top-right), EG-ED (bottom-left) and EG-CRA (bottom-right) copper.
34
Fig. 2.6.2 Cumulative volume fraction vs. grain size plot for nc high-S, UFG low-S, EG-ED and EG-CRA copper.
35
Fig. 2.6.3 Top: XRD patterns for synthesized (nc low-S, nc high-S and UFG low-S) and commercial grade copper foils (EG-ED and EG-CRA). bottom: magnified view of the {111} peaks showing peak broadening as grain size decreases.
37
Table 2.7.1 Results of hardness and electrical resistivity measurements obtained in this study. 39
Fig. 2.7.1 a) Nano-indentation showing dynamic hardness for low-S, high-S and commercially available copper as function of grain size. b) Hall-Petch plot of the synthesized electrodeposited copper.
41
Fig. 2.7.3 Schematic representation of the four-point probe used in electrical resistivity measurements.
42
Fig. 2.7.4 Electrical resistivity results for copper foils made in this study and commerical available copper foils as a function of grain size.
43
Fig. 2.8.1 Cross sectional optical images of etched EG-ED (left) and nc low-S (right) copper [Kim (2007)].
46
Fig. 3.1.1 Effect of grain size (d) on the volume fraction for intercrystalline regions (grain boundaries and triple junctions), assuming a grain boundary thickness of 1 nm [Palumbo et al., (1990)].
50
xi
Fig. 3.3.1
DSC scans of 10nm Ni-0.158%S; heating rate 10°C/min. (heat exotherm downward) [Klement et al., (1995)].
57
Fig. 3.3.2 Modified Kissinger analysis for Ni-S, Ni-Fe, Ni-P and Co [Hibbard et al. (2002)].
59
Fig. 3.3.3 Peak temperature [Tp] as a function of: (a) grain size, (b) carbon concentration, and (c) sulfur concentration [Hibbard et al. (2006)].
60
Fig. 3.3.4 DSC scans of Ni~2.5wt.%P of sample: (left) smaller grain size with narrow grain size distribution [first group], and (right) larger grain size with wide grain size distribution [second group] [Zhou (2006)].
62
Fig. 3.3.5 Grain size as a function of dwell time at room temperature for IGC&C Cu made with different residual porosities. Sample 1 - 7%, Sample 2 - 4%, Sample 3 - 3% [Gertsman and Birringer (1994)].
67
Table 5.1.1 Grain size and sulfur levels for nc low-S and nc high-S copper used for DSC experiments.
78
Fig. 5.1.1 DSC heat release curves for nc low-S and nc high-S copper annealed to 400°C
using a heating rate of 40°C/min.
79
Table 5.1.2 Summary of peak temperatures and activation energies for nc low-S and nc high-S samples.
81
Fig. 5.1.2 Modified Kissinger analysis to determine the activation energies for grain growth (highest point of the heat release - Tp) in nanocrystalline low-S and high-S Cu.
82
Fig. 5.1.3 DSC heat release curves (heating rate: 40°C/min) with regions of interest for both nc low-S and nc high-S copper. The symbol T indicates where the microstructure was observed ex-situ in the TEM. Tp is the peak temperature that was used in measuring the activation energy for grain growth using the modified Kissinger analysis.
83
Fig. 5.2.1 TEM darkfield micrographs and diffraction patterns for nanocrystalline high-S
copper: as-deposited (top), annealed to 85°C (center) and annealed to 175°C
(bottom).
85
Fig. 5.2.2 Grain size distributions for nanocrystalline high-S copper of as-deposited material,
and material annealed to 85°C and 175°C, respectively, at 20°C/min.
86
Table 5.2.1 Grain size statistics for high-S annealed to different temperatures. 86
Fig. 5.2.3 TEM darkfield micrographs and diffraction patterns for nanocrystalline high-S
copper: annealed to 275°C (top) and 400°C (bottom).
87
Fig. 5.2.4 Grain size distributions for nanocrystalline high-S Cu annealed to 275°C and annealed to 400°C.
88
Table 5.2.2 Percent change in mean and right extreme grain size with respect to 89
xii
previous stage.
Fig. 5.2.5 Magnified view of selected area diffraction patterns for high-S samples annealed to
175°C (left) and 275°C (right). Samples annealed to 275°C showed formation of
second phase particles as indicated by extra spots the circles.
90
Fig. 5.2.6 Cu-S phase diagram showing different CuS and Cu2S second phase particles that could form at different temperatures [Chakrabarti et al. (1994)].
91
Fig. 5.2.7 Enlarged view of the Cu-S equilibrium diagram [Chakrabarti et al. (1994)]. 92
Fig. 5.2.8 EBSD map showing microstructure of nc high-S copper DSC-annealed to 400°C at
20°C/minute.
93
Fig. 5.3.1 DSC curve of nanocrystalline low-S copper DSC-annealed to 400°C at 20°C/minute.
94
Fig. 5.3.2 TEM darkfield micrographs and corresponding diffraction patterns for nc low-S copper in the as-deposited state (top) and DSC-annealed to 110°C (bottom).
95
Fig. 5.3.3 Grain size distributions for as-deposited low-S copper and material annealed to
110C.
96
Table 5.3.1 Grain size statistics for low-S copper: as-deposited and material annealed to different temperatures.
97
Table 5.3.2 Percent change in average grain size and largest grain diameter w.r.t the previous stage.
97
Fig. 5.3.4 Darkfield micrographs (left), and diffraction patterns (right) for nc low-S annealed to 250°C and 400°C.
98
Fig. 5.3.5 Grain size distributions of low-S materials annealed to 250°C and 400°C. 98
Fig. 5.3.6 EBSD map for low-S copper annealed to 400°C. 99
Fig. 5.4.1 Cumulative volume fraction curves for nc low-S copper annealed to different temperatures.
100
Fig. 5.4.2 Cumulative volume fraction curves for nc high-S annealed to different temperatures.
101
Fig. 6.1.1 TEM micrographs showing the as-deposited (left), and isothermally annealed (100ºC) microstructures of nc low-S (top) and nc high-S (bottom) copper for 5 (center) and 30 (right) minutes.
109
Fig. 6.1.2 Grain size distributions for nc low-S and nc high-S copper annealed at 100°C for 5
and 30 minutes.
110
Table 6.1.1 Results for average grain size, average largest grain size and standard deviations of nc low-S and nc high-S copper deposits in the as-deposited stage and after annealing at 100ºC for 5 and 30 minutes.
110
Fig. 6.1.3 Grain size distributions of nc low-S copper annealed at 100ºC. 112
xiii
Fig. 6.1.4 Grain size distributions of nc high-S copper annealed at 100ºC. 112
Fig. 6.2.1 EBSD maps showing the microstructures for nc low-S (top) and nc high-S (bottom)
copper electrodeposits annealed at 300°C for 5 (left) and 30 (right) minutes.
114
Table 6.2.1 Results for average grain size, average largest grain size and standard deviation of nc low-S and nc high-S copper deposits in the as-deposited state and annealed at 300ºC for 5 and 30 minutes.
114
Fig. 6.2.2 Grain size distributions of nc low-S and nc high-S copper annealed at 300°C for
different times.
115
Fig. 6.2.3 Grain size distributions of nc low-S copper annealed at 300ºC. 116
Fig. 6.2.4 Grain size distributions of nc high-S copper annealed at 300ºC. 116
Fig. 6.3.1 Maximum sulfur concentration at grain boundaries as a function of grain size. 119
Fig. 6.3.2 Grain size estimation for the formation of sulfur containing second phase particles as copper deposits reached 21.3at.% along the grain boundaries.
122
Fig. 6.3.3 Combined isothermal annealing data for nanocrystalline low (<10ppm) and high (~850ppm) sulfur nickel [Soong (2009), Hibbard (2002)].
123
Fig. 7.2.1 Darkfield TEM micrographs showing microstructures of nanocrystalline low-S and high-S copper: as-deposited (left), and stored at room temperature for 6 months (centre) and 2 years (right).
128
Fig. 7.2.2 Grain size distributions for nanocrystalline low-S (top) and high-S (bottom) copper of the as-deposited material, and material left at room temperature for up to 2 years.
129
Table 7.2.1 Grain size analysis summary: average grain sizes and standard deviations. 130
Fig. 7.3.1 Maximum sulfur concentration at grain boundaries as a function of grain size.
133
Fig. 8.2.1 Grain boundary energy measurements obtained from tin bicrystals [Aust and Chalmers (1950)].
137
Fig. 8.2.2 Measured grain boundary energy for [001] twist boundaries in Cu [Miura et al. (1990)].
138
Fig. 8.2.3 Schematic diagram of the sphere plate technique showing the grain boundary and the rotation of the sphere on a single crystal plate with the corresponding interfacial energy vs. misorientation curve [modified after Shewmon (1966)].
139
Fig. 8.2.4 XRD results of the sphere plate before (left) and after (right) annealing [McCafferty (1995)].
140
Fig. 8.2.5 Example of a CSL boundary obtained by a 21.8° rotation across a (111) plane
[Reed-Hill and Abbaschian eds. (1992)].
141
Fig. 8.4.1 SEM micrograph showing surface morphologies of attacked random and 145
xiv
unattacked special (Σ9) grain boundaries in nickel [Palumbo and Aust (1990)].
Fig. 8.4.2 Cracked grain boundaries (solid black lines) in alloy 600 fracture area after intergranular stress corrosion test [Langer (1995)].
146
Fig. 8.5.1 GBE and conventional as-cast lead acid battery grid after corrosion tests [Palumbo et al., (1999)].
150
Fig. 8.5.2 Schematic representation of a coherent twin boundary generated in a grain corner [Fullman and Fisher (1951)].
150
Fig. 8.5.3 Formation of a non-coherent twin that forms along the side a coherent twin boundary [Fullman and Fisher (1951)].
152
Fig. 8.7.1 Relationship between the frequency of special boundaries and grain size in various polycrystalline metals and alloys produced by thermomechanical processing [Watanabe (1993)].
157
Fig. 8.7.2 Frequency of CSLs as a function of grain size for materials susceptible to twinning [Lin (1997)].
158
Fig. 8.7.3 Variation of annealing twin density with grain size during isothermal annealing of Cu-Sn alloy at 700°C [Liu (1982)].
158
Fig. 8.7.4 CSL frequency as a function of average grain size for Ni-15wt.%Fe. Time in bracket indicate the annealing times [Palumbo and Aust. (1998)].
159
Fig. 8.7.5 %fSB as a function of average grain size for ultrafine-grained Ni [Kobayashi et al. (2010)].
160
Table 8.8.1 Grain size and chemical composition of samples used in this section. 160
Table 8.8.2 Annealing conditions of investigated samples. 162
Fig. 8.9.1 EBSD map of low-S nc Cu annealed at 300C for 5 minutes. This map is constructed using band contrast (BC), grain boundaries (GB) and CSL components. Low angle boundaries (LABs) and random high angle boundaries
(HABs) are shown in yellow and black. Special boundaries (SBs) including 3, 9,
and other boundaries (29) are also displayed and labelled as red, green and blue lines, respectively.
164
Fig. 8.9.2 EBSD map showing GB and CSL components used for GBCD analysis. 165
Fig. 8.9.3 GBCD’s of low-S copper investigated in this study. 166
Fig. 8.9.4 GBCD’s of polycrystalline copper used in this study. 166
Table 8.9.1 Grain size, line fractions of special boundaries and line fractions of high angle boundaries of the annealed structures used in this study.
167
Fig. 8.9.5
EBSD images illustrating samples with a low (left) and a relatively higher (right)
fraction of CSL grain boundaries retained in the annealed microstructure. 3 boundaries are labelled in red and random boundaries are indicated in black lines.
The yellow, green, and blue lines respectively represent 1, 9 and 29.
168
xv
Fig. 8.10.1
Percentage of special boundaries (fSB) as a function of grain size for low-S nc copper, PPC copper and sulfur containing nc Ni-Fe obtained by Palumbo and Aust [Palumbo and Aust (1998)].
169
Fig. 8.11.1 Selected {100} {110} and {111} pole figures obtained from EBSD measurements for
low-S copper annealed to grain sizes of: a)400nm, b) 500nm, c) 1.62m and d)
5.12m.
171
Fig. 8.11.2 Selected {100} {110} and {111} pole figures obtained from EBSD for PPC copper
with the grain sizes of: a) 3.58m – starting structure, b)7.95m, c) 9.78m and d)
18.9m.
172
Table A1 Effect of electrodeposition parameters and bath additives on deposit grain size and surface quality.
186
Fig. A2-1 Schematic representation of the model used to evaluate the intercrystalline (grain boundary and triple junction) volume fraction of material. The diagram shown here is a cross-sectional view of the intersection of three grains (three adjoining tetrakaidecahedra along a polyhedral edge) [Palumbo et al., (1990)].
187
Fig. A2-2 The effect of grain size (d) on calculated volume fraction for intercrystalline, grain boundaries and triple junction components. The calculations assumed a grain
boundary thickness () of 1nm [Palumbo et al., (1990)].
188
Fig. A3 Schematic diagrams of normal and abnormal grain growth (t=annealing time, D=grain size) [Haessner, (1978)].
189
xvi
LIST OF ACRONYMS AND ABBREVATIONS
3DAP – 3-dimensional atomic probe
ALGD – Average largest grain diameter
AGS – Average grain size
APFIM – Atomic probe field ion microscopy
BC – Band contrast
BF – Bright field
CRA – Cold-rolled annealed
CS – Chemical synthesis
CSL – Coincidence site lattice
CTE – Coefficient of thermal expansion
CVF – Cumulative volume fractions
DSC – Differential scanning calorimetry
EBSD – Electron backscattered diffraction
ED – Electrodeposited
DF – Dark field
DP – Diffraction patterns
FWHM – Full width at half maximum
GB – Grain boundary
GBCD – Grain boundary character distribution
GBE – Grain boundary engineering
GSD – Grain size distributions
High-S – High sulfur
IGC&C – Inert gas condensation and compacting
KDP – Kikuchi diffraction pattern
Low-S – Low sulfur
MA – Mechanical attrition
MAD – mean angle deviation
MS – Magnetron Sputtering
nc - Nanocrystalline
PED – Pulsed electrodeposition
PWB – Printed wiring boards
SD – Standard deviations
TEM – Transmission electron microscope
TMP – Thermomechanical processing
TIPA – Tri-isopropanol-amine
UFG – Ultra-fine grained
XRD – X-ray diffraction
1
Chapter 1 – Introduction
1.1 – Research Motivation
Copper is a widely used conductor in the manufacture of printed wiring boards
(PWB). The continuing trends in the miniaturization of electronic devices create
increasing challenges to all electronic industries. The PWB manufacturers are under
pressure because the performance capabilities and miniaturization of the integrated circuit
(IC) devices are improving at a much faster rate than the PWB technology required to
support them. Nanocrystalline and ultra-fine grain copper can potentially offer improved
reliability and functionality of the PWB due to the increases in mechanical strength and
achievable wiring density by grain size reduction.
In order to consider potential improvements in the reliability of PWBs, it is
important to address factors related to the coefficient of thermal expansion (CTE) and
how these contribute to PWB failure. CTE is the physical expansion in the PWB over a
specific range of temperature. Thermal expansion is a reversible process: the material
expands upon heating and contracts during cooling. Fig. 1.1.1a illustrates the stresses
placed on a PWB as a result of the difference in thermal expansion coefficients for the
copper foil and the FR-4 glass epoxy board (Cu=17 PPM/C and FR-4=58 PPM/C).
The copper inside the PWB, whether in the form of copper vias or copper foil, is stressed,
placing pressure onto the copper interconnects. These stresses can lead to foil cracking
or layer/via separations; ultimately causing failure of the electronic device. In fact,
copper foil/via cracking is one of the most common failures in electronic packages
2
[Viswanadham and Singh., (1998)]. Fig. 1.1.1b is a schematic diagram of a PWB cross
section showing foil and via cracks due to the CTE mismatch.
The first part of this research deals specifically with potential improvements of
the copper foil performance by grain size reduction. Typical dimensions for an electronic
grade (EG) copper foil used today are as follows: 18 by 24 inch wide, with thicknesses
ranging from 10 to 35m. There are two methods of making EG copper foil: cold rolling
+ annealing (CRA) and electrodeposition (ED).
a b
Fig. 1.1.1 – Schematic representation of PWB illustrating location of copper (a) and cracks due to the CTE mismatch effects (b) [GE Electromaterials (2001)].
A decrease in the grain size increases the grain boundary volume fraction of the
material. This provides constraints for dislocation movement and ultimately increases the
strength of the copper foil. As the strength of the copper foil increases, the circuit is
expected to become more resistant to deformation and cracking, hence potentially
increasing the reliability of the PWB.
Copper via Copper foil
3
Based on the isotropic dissolution characteristics of copper during the lithographic
etching process, it is further expected that the achievable wiring density can also be
increased through grain size reduction. Fig. 1.1.2 shows schematic diagrams of circuit
traces produced by lithography, using grain sizes of 1 m (a) and 100nm (b),
respectively. Circuit traces in the copper foil with smaller grains would have a better
spatial resolution than what is achievable with coarser grained material. Secondly, as the
grain size decreases, the trace width can be reduced and more circuits can be lithographed
into the same area, resulting in increased PWB wiring density. Moreover, the increased
spatial resolution is expected to make traces straighter and more uniform, resulting in
decreased localized stress field.
a b
Fig. 1.1.2 – Schematic representation of PWB’s circuit traces after lithography and etching. Grain size reduction could potentially enhances trace resolution
and achievable wiring density of the PWB [Woo and Erb (2006)].
Average grain size of 1m Average grain size of 100 nm
Circuit trace
Etched copper
4
However, thermal stability is one of the main issues that may limit the use of
nanocrystalline materials at elevated temperatures. With such a small grain size,
nanocrystalline materials have very high densities of grain boundaries. Consequently, they
are metastable and have a high driving force for grain growth. A good understanding of the
thermal stability of nanocrystalline copper is not only of importance from a basic science
point of view. It is of also necessary for the potential application of nanocrystalline copper
in printed wiring boards. Fig. 1.1.3 shows a typical reflow profile of lead free solder during
printed circuit board manufacturing. From this figure, it can be seen that the printed wiring
board is subjected to annealing and the reflow process (i.e: temperature above 217C in
order to solder the electrical components onto the printed wiring board) takes about 60-150
seconds with a peak temperature that can reach about 260C. The room temperature
thermal stability is also of particular interest here because some studies have shown grain
growth of nanocrystalline copper even at room temperature. Moreover, during device
usage the printed wiring board could reach temperature up to ~50°C.
Fig. 1.1.3 – Typical lead-free solder reflow profile
5
1.2 – Research Objectives and Organization of this Report
There are two main objectives for this thesis research. The first part deals with
the synthesis and microstructure characterization of ultra-fine and nanocrystalline copper
foils. The copper foils are then compared with commercially available electronic grade
electrodeposited (ED) and cold-rolled annealed (CRA) copper foils in terms of
mechanical and electrical properties.
The second part deals with the main objective of this thesis which is focused on
developing a better understanding of the thermal stability of nanocrystalline copper. For
this part of the work nanocrystalline copper electrodeposits were synthesized with
different amounts of sulfur impurities and annealed at various temperatures.
This thesis is written and organized in the following manner. Chapter 2 deals
with the material synthesis, microstructural, mechanical and electrical characterizations
of copper as a function of grain size. Chapter 3 of this report presents a literature review
regarding the thermal stability of nanocrystalline materials, with particular emphasis on
materials that are made by electrodeposition. Annealing strategy and experimental
techniques for the thermal stability studies are presented in Chapter 4. Chapter 5
summarizes the results of all the calorimetric annealing experiments, followed by
isothermal annealing experiments results in Chapter 6. Chapter 7 presents a brief
discussion in regards to room temperature grain growth of nanocrystalline copper.
Chapter 8 provides an introduction to grain boundary character, grain boundary
engineering, and the change in grain boundary character and texture during isothermal
6
annealing. Conclusions from research and recommendations for future work will be
presented in Chapters 9 and 10, respectively.
7
1.4 – References for Chapter 1
GE Electromaterials – technical bulletin on HTE copper foil, Nov (2001), 1
Viswanadham P., and Singh P., (ed.), Failure Modes and Mechanisms in Electronic
Packages, 1st edition, Chapman and Hall Publishing, New York (1998)
Woo P., and Erb U., IPC APEX Printed Circuit Expo Proc., (2006) S37 1
8
Chapter 2 – Materials Synthesis and Characterization
2.1 – Electrodeposition of Copper
Electrodeposition is a powerful synthesis technique which can produce materials
with a wide range of property combinations in terms of strength, hardness, ductility,
fatigue resistance and electrical resistivity [Safranek (1974)]. More specifically, copper
electroplating has long been used in many engineering and decorative applications.
Combinations of different electrolytes, plating bath additives and operating parameters
yield a wide range of mechanical and physical properties for the resulting electrodeposits.
In the early 1970s, Lamb et al. published a comprehensive review on the physical
and mechanical properties of electrodeposited copper produced from a wide variety of
plating experiments [Lamb et al. (1970)], revealing several correlations between structure
and mechanical strength of the deposits. Optical microscopy showed that the increase in
mechanical strength and hardness of the deposit was associated with grain refinement.
For example, Fig. 2.1.1 shows the effect of the addition agent TIPA (Tri-isopropanol-
amine) on the microstructure of the copper deposits. Without any addition agent (Fig.
2.1.1a), the deposit is columnar with coarse grain structure. On the other hand, copper
deposits from a bath that contained 3.5 g/L TIPA (Fig. 2.1.1b) showed significant grain
refinement. Moreover, the samples plated with 3.5g/L of TIPA experienced a two-
fold increase in tensile strength and about 10% increase in electrical resistivity. A
summary of the mechanical properties and electrical resistivity for both materials is
given in table 2.1.1.
9
Fig. 2.1.1a – No additives Fig. 2.1.1b – with 3.5 g/L TIPA
Fig. 2.1.1 – Optical micrographs showing the effect of TIPA addition agent on the
structure of copper deposits produced from a sulfate bath. Fig. 2.1.1a: no additives and Fig. 2.1.1b: with 3.5g/L TIPA. Both electrodeposits were plated at
30C with an applied current density of 50mA/cm2 [Lamb et al., (1970)].
Table 2.1.1 – Influence of TIPA addition agent on strength, ductility and electrical resistivity of electrodeposited copper [Lamb et al., (1970)].
This example clearly shows that a small amount of bath additive has a significant
effect on the deposit’s microstructure. It should be noted that the actual grain size were
not given in this paper. Research that specifically concentrated on making
nanocrystalline and ultra-fine grained copper is summarized in the following sections.
2.2 – Electrodeposited Nanocrystalline Copper
In recent years there has been considerable interest in producing nanocrystalline
copper for various applications. There are several studies on nanocrystalline copper
Copper sulfate
solutions Property
Tensile strength at -78C
and (325C) [GPs]
Elongation at -78C
and (325C) [%]
Electrical resistivity
[ohm-cm] @23C
No addition agents 27, (8) 54.5, (8.5) 1.73
With 3.5 g/L TIPA 41, (7) 14, (4) 1.89
10
produced by various techniques such as inert gas condensation and compacting (IGC&C)
[Sastry et al. (1998) and Nieman et al. (1989)], chemical synthesis (CS) [Suryanarayanan
et al. (1996)], magnetron sputtering (MS) [Lu et al. (2004)] and mechanical attrition
(MA) [Wang et al. (2003) and Ma (2003)]. This report is limited to studies dealing with
nanocrystalline copper made by electrodeposition (ED).
In 1995, Cziraki et al., synthesized nanocrystalline copper using electrodeposition
from a copper sulfate bath containing proprietary additives [Cziraki et al. (1995)]. The
samples (50 mm in diameter, up to 5 m thick) were deposited onto Ti substrates using
direct current (DC) plating (Current density: 50mA/cm2). TEM observations found the
average grain size of the deposits to be around 100nm. It was also observed that the
electrical resistivity of the deposits was almost identical to conventional polycrystalline
copper. However, reasons/mechanisms for the observed grain size reduction were not
given in the paper.
Natter and Hempelmann, in 1996, produced nanocrystalline copper using pulsed
current electrodeposition [Natter and Hempelmann (1996)]. In this technique, a square
wave current pulse is used that is switched on and off periodically. Fig. 2.2.1 shows a
typical waveform used in this technique to produce nanocrystalline metals [El-Sherik and
Erb (1994) and Erb et al. (1994) and (1995)]. The electrolytic bath used in the work by
Natter and Hempelmann contained copper sulfate and ammonium sulfate. The results
showed that plating under certain conditions of Ipeak, Ton and Toff produced
nanocrystalline copper with an average grain size of around 50nm. Further grain size
11
reduction was demonstrated by adding organic additives such as citric acid into the
plating bath.
Fig. 2.2.1 – Generalized pulsed current electrodeposition waveform.
There are two important findings in the paper by Natter and Hempelmann. First,
they re-confirmed that pulsed electrodeposition (PED) can refine the grain structure of
copper deposits. Second, they presented yet another example on grain size reduction by
introducing an organic additive into the plating bath. The technique of using PED with or
without bath additives in making nanocrystalline metals was previously demonstrated by
El-Sherik and Erb in 1995 [El-Sherik and Erb (1995)].
The fundamental requirements of nanocrystal electrodeposition are to promote
grain nucleation and impede growth of existing grains. Natter and Hempelmann
referenced the work by Glasstone to explain grain refinement realized by using pulse
current electrodeposition [Glasstone (1935)]. Glasstone related the activation energy for
grain nucleation (Ak) and cathodic overpotential (g) by equation 2.1;
12
2])/'[(
1
MeMe
kaag
A (Eqn 2.1)
where a’Me+ is the Cu2+
activity near the cathode (Nernst diffusion layer) and aMe+ is the
Cu2+
activity of the bulk electrolyte. This equation shows that a high cathodic
overpotential (g) will result in low activation energy for grain nucleation (Ak),
ultimately promoting nucleation rate and assisting in nanocrystal electrodeposition.
Pulse current electrodeposition can be utilized to allow higher nucleation rates
during electrolysis [Erb et al. (1994) and (1995)]. This is achieved by permitting
electrodeposition to be carried at very high current density (high overpotential) for a short
on-time (Ton). The peak current density (Ip) and on-time (Ton) determine the reduction
rate of the metallic species during a pulse. However, high nucleation rate will strongly
decrease the ion concentration near the cathode. Hence, an off-time (Toff) is used here to
allow ions to diffuse back into the ion-depleted layer near the surface of the cathode.
Natter and Hempelmann also explained the reduction in grain size for Cu from 50
to 11nm with increasing citric acid contents (0 to 100g/L) [Natter and Hempelmann
(1996)]. The XRD patterns in Fig. 2.2.2 show the grain size - citric acid concentration
relationship. According to this study, citric acid acts as a copper complexing agent that
impedes lateral diffusion by occupying active sites on the deposit surface. A schematic
diagram of a deposit surface is shown in Fig. 2.2.3. Active sites can be described as
edges and peaks of the deposit (Fig. 2.2.3 sites 2-5). Copper complexes tend to occupy
these active sites and metal ions are forced to nucleate on passive sites (Fig. 2.2.3, site 1).
13
Consequently, diffusion of copper ions is blocked, leading to a disturbance in grain
growth and smaller overall crystallite size.
Fig. 2.2.2 - XRD patterns showing the change in grain size (Dv) with varying citric
acid contents [Natter and Hempelmann (1996)].
Fig. 2.2.3 – Schematic diagram of a deposit surface. Active sites: #2-5; Passive site #1 [Natter and Hempelmann (1996)].
Natter and Hempelmann also observed an increase in grain size with increasing
bath temperature. Increasing bath temperature increases the ion mobility and decreases
the cathodic overpotential (g). Thus, the energy for nucleation (eqn 2.1) is increased
resulting in lower nucleation rate and the formation of coarser grains.
14
In this publication, Natter and Hempelmann demonstrated the synthesis of
electrodeposited nanocrystalline copper. However, the deposits were only produced as
thin films (actual deposit thickness was not given here) and a special process was
required in order to remove the deposits from the Ti substrate. In addition, surface
quality, sample thickness, mechanical and electrical properties of nanocrystalline copper
were not given in this paper.
Compounds containing sulfur have been common additives used in the process of
electrodeposition for a long time [Dini (1993)]. For example, compounds such as
saccharin or trisulfonated naphthalene are known to increase the yield strength in nickel
electrodeposits through grain size reduction. Many other sulfur containing compounds
such as thiourea and cyclic thioureides also showed to have a leveling effect which makes
the deposit surface smooth which is important for decorative coatings. Moreover,
nanocrystalline nickel electrodeposits often contain sulfur impurities in the range of 400
to 2000ppm due to the use of saccharin as grain refiner [El-Sherik and Erb (1995)].
Thiourea is a common sulfur-containing bath additive used for enhancing the strength of
the copper electrodeposits through grain refinement [Lamb et al. (1970)]. The use of
thiourea as a bath additive in making nanocrystalline/ultra-fine copper deposits has also
been demonstrated by Huang et al., Tao and Li, and Hakamada et al. [Huang et al. (2004),
Tao and Li (2005) and Hakamada et al. (2006)]. The deposit thicknesses ranged from 15-
35m, with average grain sizes varying from 30 to 160nm. Both Huang et al. and
Hakamada et al. showed that the hardness of the as-electroplated copper deposits
increased with thiourea content. For example, Hakamada et al. showed that hardness
15
increases linearly with thiourea concentration up to 0.02g/L. Beyond this point, the
hardness values remained constant as thiourea concentration increased. Tao and Li,
compared the process difference between D.C. plating and pulse plating, with the same
thiourea concentration and average current density [Tao and Li (2006)]. Their results
showed that a much finer microstructure is achieved through utilizing pulse plating.
There are several other studies that described the synthesis of nanocrystalline
copper. However many of these electrodeposits were extremely thin. For example, Paik
et al., electroplated 1.5 m thick copper with an average grain size of around 100nm,
using DC plating with a current density of 70 mA/cm2 and an acid sulfate bath that
contained proprietary grain refiners [Paik et al. (2003)]. Seah et al. also electroplated thin
copper films (<0.5 m thick) with an average grain size of 100nm, using a current density
of 50-60 mA/cm2 and 30 seconds of deposition time [Seah et al. (1998)].
2.3 – Electrodeposited Ultra-fine Grain Copper
There are numerous studies dealing with synthesis and characterization of ultra-
fine grained copper. K. Lu et al., reported that electrodeposited “nanocrystallite” copper
subjected to room temperature rolling exhibited extremely high elongation to fracture
(exceeding 5000%) [Lu et al. (2000)]. However, high-resolution transmission electron
microscopy (e.g. Fig. 2.3.1) showed that the as-deposited copper was not truly
nanocrystalline. The microstructure consisted of ultra-fine crystalline domains ranging in
size from a few nanometers to about 80nm and separated by low angle boundaries.
16
Microhardness results indicated that the “nanocrystallite” copper had the same Vickers
hardness as conventional coarse grain copper.
Fig. 2.3.1 – HRTEM micrograph of several “nanocrystallites” (A, B, C and D)
separated by low angle (1-10) grain boundaries in as-deposited copper [Lu et al., (2000)].
Recently, Lu et al., produced 200 m thick ultra-fine grained (UFG) copper
electrodeposits with an average grain size of 400nm [Lu et al. (2004)]. Most of the grains
contained high densities of nano-scale twins. These deposits were also synthesized using
pulse current electrodeposition from an acid copper sulfate bath. A TEM electron
micrograph for an as-deposited copper sample with many nano-scale twins is shown in
Fig. 2.3.2a. Microtensile testing of these samples displayed excellent tensile strength (1
GPa) and ductility (10% elongation to failure) for this material. A TEM micrograph
taken near the fracture surface (Fig. 2.3.2b) revealed that the twin boundaries act as
barriers for dislocations to pile up, impeding deformation and ultimately strengthening
the material. Moreover, Lu et al. also measured the room temperature electrical
resistivity of the copper electrodeposit. The result showed that the material retained a
similar electrical resistivity (1.69ohm-cm) as its polycrystalline counterpart.
17
Fig. 2.3.2a Fig. 2.3.2b
Fig. 2.3.2a: TEM micrograph of the microstructure in an as-deposited UFG copper sample. Fig. 2.3.2b: TEM image showing dislocation pile-ups at twin
boundaries after micro-tensile testing [Lu et al., (2004)].
It is also worth mentioning that some commercially available electronic grade
copper foils also have a very fine grain structure, with grain sizes that are well below
1m. Merchant et al. measured the mechanical properties of different electronic grade
copper foils. TEM analysis on these copper foils revealed that their grain sizes ranged
from 400nm to 3m [Merchant et al. (2004)]. Weil et al. determined fatigue properties
and structural characteristics of electronic grade thin copper foils. TEM analysis showed
that some of the copper foils have average grain sizes of around one micrometer [Weil
and Hong (1996)]. As can be seen in Fig. 2.3.3, commercially available EG copper foils
exist in many different grain structures [Merchant et al. (2004) and Weil and Hong
(1996)].
18
Fig. 2.3.3a Fig. 2.3.3b
Fig. 2.3.3c Fig. 2.3.3d
Fig. 2.3.3 - TEM micrographs showing typical grain structures for commercially
available electronic grade copper foils. a: typical 18 m thick ED Cu foil, b: 25
m thick ED Cu foil plated with no additives, c: 18 m thick heavily twinned ED
Cu foil, and d: 18 m thick twin-free ultra-fine grain ED Cu foil. Fig. 2.3.3a, 2.3.3b and 2.3.3d - [Merchant et al. (2004)] and
Fig. 2.3.3c - [Weil and Hong (1996)].
2.4 – Synthesis of Electrodeposited Nanocrystalline and Ultra-fine Grained
Copper
In the current study, nanocrystalline and ultra-fine grain copper foils were
synthesized by the pulsed current electrodeposition (PED) technique. A schematic
representation of the electrodeposition setup is shown in Fig. 2.4.1. Pulse
electrodeposition is carried out by applying a periodic current, with on-times and off-
times in the milliseconds range. The electrolyte is mechanically stirred and a bath
temperature of 252C is maintained throughout the process. The average current
19
density (Iavg=Ipeak*Ton/(Ton+Toff)) ranged from 150-550mA/cm2 depending on the target
grain size. Phosphorized copper (Cu-4wt.%P) was used as the anode material. The
cathode is made from a piece of titanium (3x3cm2) which provides a stable electrically
conductive oxide layer on the surface, for easy mechanical stripping of the copper deposit
after deposition [Lowenheim (1978)].
Fig. 2.4.1 – Schematic representation of the pulse current electrodeposition setup.
Acid copper sulfate type plating baths were used because they are economical to
prepare, operate and widely used in the electronic industries. Copper sulfate and sulfuric
acid comprise the basic ingredients for these electrolytes. Copper sulfate acts as the
source of copper ions, and sulfuric acid increases the conductivity of the electrolytic bath.
Three sulfate-type copper baths were used in this study. The first additive-
containing bath generated deposits with relatively low sulfur impurity concentrations
ED Cu onto
Ti substrate
(cathode)
Chemical bath
(electrolyte)
Raw Cu rounds
(anode)
Pulse-Plating
Power supply
20
while the second bath produced deposits with much higher sulfur content due to a sulfur-
containing additive. The third bath contained no additives and was used to produce
conventional sulfur-free fine grained copper for comparison. For the remainder of this
report, the two types of nanocrystalline materials with sulfur will be referred to as low-S
and high-S copper deposits and PPC will be used to denote pure polycrystalline copper
made from an additive-free acid-sulfate bath.
Sulfur-free polyethylene-glycol (PEG) additive was used as the grain refiner in the
low-S copper electrolyte. For high-S copper samples, thiourea was used as the bath
additive. Thiourea is a copper sulfate bath additive that has been used for many years to
enhance the strength of copper electrodeposits through grain refinement [Lamb et al.
(1970)]. The combination of high nucleation rate, low crystal growth rate and different
bath additives ultimately support the synthesis of nanocrystalline copper with different
sulfur impurities and grain sizes. Tables 2.4.1 and 2.4.2 summarize the bath
compositions and operating parameters used for producing all the deposited materials
used in this study.
21
Table 2.4.1 – Bath compositions for the electrodeposition of nanocrystalline low-S, nanocrystalline high-S and pure polycrystalline copper (PPC).
Bath composition
Bath Type Low-S High-S PPC
Component Content in g/L (unless otherwise stated)
Copper sulfate pentahydrate 250 250 250
Sulfuric acid (ml/L) 125 125 125
Polyethylene-glycol (PEG) 10 10 0
Thiourea (TU) 0 0.05 0
Table 2.4.2 – Bath and selected operating parameters for the electrodeposition of nanocrystalline low-S, ultra-fine grained low-S, nanocrystalline high-S and pure
polycrystalline copper.
Bath parameters
pH 0.8-1.0
Temperature [C] 202
Agitation 3 inch magnetic stirrer at ~ 700rpm
Operating parameters
Type low-S UFG
low-S high-S PPC
Ton/Toff [ms] 2.5/16.5 2.5/16.5 5/95 D.C 2/65
IPeak [mA/cm2] ~3000 ~1000 ~4000 300 ~5800
Iavg. [mA/cm2] ~400 ~150 ~200 300 ~180
Sample thickness ~25m
2.5 – Synthesis Results
The synthesis of the desired copper deposits was not a trivial process. It required
more than 300 plating runs and a total of 13 plating baths in order to achieve acceptable
deposits with required microstructures and different sulfur levels. There are many factors
that affect the grain size of copper electrodeposits including current density, length of the
22
periodic pulse (Ton and Toff), bath additives, bath conductivity, bath temperature and
agitation. A summary of synthesis results and the effect of different plating variables on
grain size and the appearance of the deposits is presented in Appendix 1.
In this section, details of the microstructural characterizations will be given for
three representative materials, i.e: low-S nanocrystalline, high-S nanocrystalline and
ultrafine grained low-S copper.
2.5.1 – Transmission Electron Microscopy
The microstructures of as-plated copper samples were characterized using a
Hitachi H-800 transmission electron microscope (TEM), operated at 200kV. Thin foils
for TEM analysis were prepared using electrolytic jet polishing in an electrolyte
containing 30% orthophosphoric acid (85% aq. soln.) and 70% distilled water at room
temperature, polishing at a voltage of ~15VDC.
For all experiments, the center 2x2cm areas of 3x3cm deposit coupons were used.
To assess the uniformity of the microstructure, samples from different regions (center and
corner) were characterized using the TEM. Fig. 2.5.1 shows TEM dark field (DF)
electron micrographs, diffraction patterns (DP) and grain size distributions from the same
high-S nanocrystalline copper deposit, taken from the center and the corner regions,
respectively. It can be seen that there are no major differences in the microstructures for
the two areas. The average grain sizes for the center and corner were measured to be 31
and 33nm, respectively. The standard deviations of the two deposits are 13.3nm and
23
13.6nm. Grain size distribution histograms also confirmed that the grain size difference
between the regions is negligible.
Fig. 2.5.2 shows examples of TEM bright field (BF) and dark field (DF) electron
micrographs, as well as the corresponding electron diffraction patterns (DP) for
nanocrystalline low-S, high-S and low-S UFG copper samples, respectively.
To better quantify the differences in grain structure between these samples, the
grain size distributions (GSD), cumulative volume fractions (CVF), average grain sizes
(AGS), standard deviations (SD), and average largest grain diameters (ALGD) were
determined. These five quantitative structure descriptors were determined by first
manually outlining more than 300 grains from dark field electron micrographs. Twin
boundaries were ignored in grain size measurements, i.e. a grain that contained many
twin boundaries was considered as only one grain. Using analytical software, ImageJ
(version 1.31u, source code written by Wayne Rasband, National Institutes of Health,
USA), and assuming that each grain is circular, the area of each grain was then
determined. Methods for calculating average grain size (AGS), grain size distribution
(GSD) and cumulative volume fraction (CVF) are described graphically in Fig. 2.5.3.
The average largest grain diameter (ALGD) was calculated based on averaging the five
largest grains found in the grain size distributions.
24
centre
corner
0
5
10
15
20
25
30
35
40
0-1010-20
20-3030-40
40-5050-60
60-7070-80
80-9090-100
Pe
rc
en
t F
re
qu
en
cy (%
)
Grain Size Range (nm)
Fig. 2.5.1 – TEM dark field micrographs, diffraction patterns and grain size
distributions of high-S nanocrystalline deposit characterized at center (left) and
corner (right) region of the sample.
25
Fig. 2.5.2 – Corresponding TEM bright field (first row), dark field (second row) and diffraction patterns (third row) micrographs for nanocrystalline low-S (left),
nanocrystalline high-S (center) and UFG low-S (right) copper samples.
Fig. 2.5.4 and 2.5.5 show the grain size distribution and cumulative volume
fraction distribution relationships for the nanocrystalline (nc) low-S, nc high-S and
ultrafine grained (UFG) low-S electrodeposits. The results of the microstructural analysis
are presented in table 2.5.1. It can be seen from Fig. 2.5.4 and table 2.5.1 that high-S
copper has a finest microstructure, along with the most narrow grain size distribution.
High-S, low-S, and low-S UFG copper were determined to have average grain sizes of 31,
49 and 198nm, respectively. All of the as-deposited samples seem to show a lognormal
26
grain size distributions. Electrodeposited nanocrystalline and ultra-fine grained metals
with log-normal distributions were also observed and described elsewhere [e.g. Wang et
al. (2003), Natter and Hempelmann (1996) and Cheung et al. (1995)].
Fig. 2.5.3 – Steps involved in determining the average grain size (AGS), grain size distribution (GSD) and cumulative volume fraction (CVF) of the microstructure.
Another observation that can be made from the grain size distributions is that
low-S UFG copper has a much wider GSD (larger standard deviation) compared with the
nanocrystalline copper samples. It is important to point out that as the GSD widens, the
use of AGS in explaining the mechanical and electrical properties of the material
becomes less reliable. For example, Zhang et al. observed a difference in mechanical
strength for ultra-fine and nanocrystalline zinc that had the same average grain sizes but
different grain size distributions [Zhang et al. (2004)]. A better way to describe the
microstructure of the bulk material is to use the cumulative volume fraction (CVF) plot.
Scanned TEM micrograph
Manually outline the perimeter of each grain
(300+ counts)
Compute average grain size and its
area using analysis software ImageJ
Determine grain size distribution (GSD) and
cumulative volume fraction (CVF) plots GSD CVF
27
Fig. 2.5.5 shows the CVF plots for nanocrystalline low-S (blue), nanocrystalline
high-S (yellow) and UFG low-S (red) copper. This plot basically gives the volume fraction
of the bulk sample with certain grain sizes. For example, the solid black line shown in Fig.
2.5.5 indicates that for the nc low-S sample, 60% of the material is comprised of grains that
are less than 100nm. On the other hand, for low-S UFG copper, shown as a solid red line,
only 10% of the grains are less than 200nm. In other words, for low-S UFG copper, grains
that are larger than 200nm fill up the remaining 90% of the bulk volume.
50-75
100-125
150-175
200-225
250-275
300-325
350-375
400-425
450-475
500-525
550-575
600-625
650-675
700-725
750-775
800-825
850-875
900-925
950-975
1000-1025
Low-S
High-S
Low-S UFG0
5
10
15
20
25
30
35
40
45
% F
req
ue
ncy
Grain Size Range (nm)
Fig. 2.5.4 - Grain size distributions (GSD) for nanocrystalline low-S (blue), high-S
(yellow) and low-S UFG (red) copper.
Table 2.5.1 - AGS, SD and ALGD for nc low-S, UFG low-S and nc high-S copper electrodeposits
nc
Low-S
UFG
Low-S
nc
High-S
Average grain size (AGS) 49nm 198nm 31nm
Standard deviation (SD) 34nm 189nm 15nm
Average largest grain diameter (ALGD) 1012nm 1087nm 135nm
28
0
10
20
30
40
50
60
70
80
90
100
0 200 400 600 800 1000 1200 1400
Cu
mu
la
tive V
olum
e Fra
ctio
n
Grain Size (nm)
UFG Low-S
nc High-S
nc Low-S
Fig. 2.5.5: Cumulative volume fraction plots for nanocrystalline low-S (blue), nanocrystalline high-S(yellow) and UFG low-S (red) copper.
2.5.2 – Chemical Analysis
For both nc low-S and high-S copper electrodeposits, as well as the pure
polycrystalline copper (PPC), LECO analysis/infrared spectroscopy analysis (IRA) and
in-torch vehicle inductively coupled plasma (ITV-ICP) were carried out for determining
impurity contents.
A LECO CS-442 IRA was used to measure the bulk carbon and sulfur
concentrations of the electrodeposits. This technique is capable of quantifying carbon
and sulfur concentrations higher than 10ppm [LECO instrument manual 2007]. LECO
supplied pure analytical carbon and sulfur standards were used for machine calibration.
29
An average of three runs was used in order to obtain the bulk sulfur and carbon
concentrations and about half a gram of sample is required for each analysis.
An IRIS Intrepid II XDL ICP system with an in-torch vehicle setup was also used
as another technique to quantify the impurities of these electrodeposits. A conventional
ICP system generally requires bulk/solid samples to be dissolved in acidic media such as
sulfuric or nitric acid. Because of the acid dilution, conventional ICP usually requires
several grams of material, and measuring organic impurities is usually problematic since
most of the acid always contain large amount (>10ppm range) of C,S,O and N impurities.
With the in-torch vehicle (ITV) setup, developed by Professor Karanassios at the
University of Waterloo, solid samples can now be directly used as feedstock to the
evaporation chamber [Karanassios and Greib (2006)]. A much smaller amount of sample
is needed for each measurement (~10mg). Most importantly, quantifying minute organic
elements can now be achieved because acid dilution is no longer required.
The in-torch vehicle system (ITV), shown in Fig. 2.5.6, uses a cup shaped
filament, made of rhenium, that is capable to heat and evaporate the sample. The ITV is
connected right before the plasma chamber and is sealed and purged with inert argon-3%
helium gas. After the sample of interest is put inside the rhenium cup, a current is
applied to rapidly heat the cup and vaporize the sample. The sample vapour then flows
into the plasma chamber. The generated plasma then travels to the spectrometer (charge
injection device (CID) detector), where the wavelength of each plasma particle is
analyzed and recorded for chemical analysis.
30
0 0
ICP Torch
Induction Coil
ITV Chamber
Ar (3% H2)
Supporting Rod
Seal
Teflon
Support
Power
Supply
Sample
Re Cup
Ar
PlasmaSpectrometer
with CID
Detector
Power
Transfer CablesSupporting
Rod
Ar
Fig. 2.5.6 – Schematic diagram of the ITV setup [Karanassios and Greib (2006)].
2.5.2.1 - Chemical Analysis Results by LECO/IRA
LECO analysis was performed to specifically measure minute carbon and sulfur
impurities. An average of three runs were used for each sample. Results from LECO
analysis are shown in the table 2.5.2.
Table 2.5.2 – Results obtained from LECO/IRA analysis.
Carbon concentration
[ppm]
Sulfur concentration
[ppm]
Low-S 38043 307
High-S 40027 25014
PPC 2137 <10ppm
(below detection limit)
2.5.2.2 - Chemical Analysis Results by ICP-ITV
Again three runs were used to obtain average sulfur and carbon concentrations for
each sample. The results are shown in table 2.5.3.
31
Table 2.5.3 – Results obtained from ICP-ITV analysis.
Sample Carbon concentration
[ppm]
Sulfur concentration
[ppm]
Low-S 436 255
High-S 808 23343
PPC 223 52
From Tables 2.5.2 and 2.5.3 it can be seen that the sulfur contents of the three
samples as measured by LECO and ICP are very similar. However, there are
considerable differences in the carbon concentrations measured by the two techniques.
The ICP analysis likely provides a more accurate measurement for carbon compared to
the LECO analysis. This is because in ICP the sample is first pre-annealed at 250C to
evaporate all surface contamination prior to the chemical measurements. On the other
hand in LECO analysis, samples are not pre-annealed to eliminate any surface
contamination before the analysis. As a result, the values as measured by ICP will be
used for the remainder of this report.
2.6 – Materials Comparisons
In addition to the samples described in section 2.5, several other deposits were
used for various investigations in the following areas of this thesis:
Area 1 - microstructural characterization
Area 2 – thermal stability studies
Area 3 – grain boundary character evolution.
32
Details of the microstructural characterization were presented only for the samples
with 31, 49 and 198nm average grain size. For the analysis of hardness and electrical
resistivity all samples were considered. Chemical analyses were presented for samples
that were used for the thermal stability and grain boundary character evolution studies
(Areas 2 and 3).
Table 2.6.1 – Summary of electrodeposited copper samples made in this study.
Grain size
(nm)
Standard
deviation (nm) Notes
Area of study
Area
1
Area
2
Area
3
3580 1700 Pure polycrystalline
copper X
1048 1465 UFG Cu produced
from low-S bath X
198 189 UFG Cu produced
from low-S bath X
86 67 nCu produced from
low-S bath X
73 60 nCu produced from
low-S bath X
49 41 nCu produced from
low-S bath X X X
31 21 nCu produced from
high-S bath X X X
Area 1 – Microstructural characterization – TEM, electrical resistivity and nano-indentation (chapter 2) Area 2 – Thermal stability studies – DSC, TEM (chapters 5, 6 and 7) Area 3 – Grain boundary character evolution studies - EBSD (chapter 8)
The remainder of this chapter will be devoted to the mechanical and electrical
property characterization of all synthesized nanocrystalline (low-S and high-S) and ultra-
33
fine grain copper (UFG Cu). Pure electrodeposited polycrystalline copper will also be
used here to serve as the standard for comparisons. Moreover, two different commercial
electronic grade electrodeposited (EG-ED) and cold-rolled annealed (EG-CRA) copper
foils were also used for comparison. These materials were supplied by PWB
manufacturer Coretec Inc. and the copper foils were originally manufactured by Gould
Electronics.
2.6.1 – Microstructure Comparisons (TEM)
Fig. 2.6.1 shows TEM dark-field electron micrographs, together with data for
average grain size (AGS) and standard deviation (S.D) for nc High-S (a), low-S UFG (b),
EG-ED (c) and EG-CRA (d) copper, respectively. High-S nc copper has the smallest grain
size; about 10 and 100 times smaller compared to EG-ED and EG-CRA copper foils.
The cumulative volume fraction plots of all four types of copper foils are shown
in Fig. 2.6.2. This plot shows that 80% of the bulk volume of the nc high-S sample
consists of grains that are less than ~50nm. On the other hand, 80% of the bulk EG-ED
copper sample consists of grains that are less than one micrometer. Only a small
fraction of the grains in EG-CRA copper are less than one micrometer.
34
nc High-S AGS=31nm UFG Low-S AGS=198nm S.D=21nm S.D=189nm
EG-ED AGS=325nm EG-CRA AGS=3.2m
S.D=261nm S.D=1.7m
Fig. 2.6.1 - TEM DF micrographs, average grain size and standard deviations for: nc high-S (top-left), UFG low-S (top-right), EG-ED (bottom-left) and EG-CRA
(bottom-right) copper.
35
0
10
20
30
40
50
60
70
80
90
100
0 1000 2000 3000 4000 5000
Cu
mu
la
tive V
olum
e Fra
ctio
n (%
)
Grain size (nm)
nc High-S
UFG low-S
EG-ED 325nm
EG-CRA 3238nm
Fig. 2.6.2 – Cumulative volume fraction vs. grain size plot for nc high-S, low-S UFG, EG-ED and EG-CRA copper.
2.6.2 – Texture/Crystallographic Orientation Comparisons (XRD)
For crystal orientation analysis, x-ray diffraction (XRD) patterns of the as-
deposited or as-received samples were collected using a Rigaku Miniflex diffractometer
with Co-K radiation (=0.17902nm). XRD scans were made by scanning from 40 to
140 at 0.2/s to cover the first five diffraction peaks from (111) to (222) planes.
X-ray diffraction patterns for the nanocrystalline low-S, high-S, low-S UFG, and
commercially available copper foils are shown in Fig. 2.6.3. In addition, the theoretical
36
intensities of a diffraction pattern that would be obtained for a random texture copper
standard sample as per structure factor and intensity calculations (e.g. Cullity B.D.) are
also included [Cullity (1978)]. A comparison of the XRD patterns shows changes in
peak widths and peak intensities, which will be addressed in the following. First, when
the average grain size of the deposit is less than 100nm (e.g. low-S and high-S nc copper),
the peaks show line broadening. For example, the full width at half maximum (FWHM)
for the (111) peak for polycrystalline EG-CRA, low-S 49nm and high-S 30nm are 0.22,
0.36 and 0.41, respectively, as shown more clearly at the bottom of Fig. 2.6.3. Second,
the XRD patterns for the electronic grade (EG) and as-deposited samples show some
changes in the peak intensities in comparison with the copper standard. Both the
nanocrystalline and ultra-fine grain samples produced in this study have a weaker (200)
peak intensity. The deposits tend to become random in texture as grain size decreases.
For EG-ED, EG-CRA and UFG low-S copper, the (220) peak intensity is enhanced over
the copper standard. This is likely due to variations in processing parameters for these
materials.
37
Fig. 2.6.3 – Top: XRD patterns for synthesized (nc low-S, nc high-S and UFG low-S) and commercial grade copper foils (EG-ED and EG-CRA). bottom:
magnified view of the {111} peaks showing the peak broadening as grain size decreases.
nc high-S 31nm
nc low-S 49nm
EG-CRA Cu 3.2µm
38
2.7 – Mechanical and Electrical Property Measurements
2.7.1 – Hardness Measurements by Nanoindentation
The hardness for all deposited materials, as well as for EG-ED and EG-CRA
copper were determined using the nano-indentation technique. Nano-indentation was
performed using a Shimadzu DHU-W201S dynamic ultra-microhardness tester equipped
with a 115 triangular pyramid (Berkovich) indentor.
Dynamic microhardness is different from conventional microhardness in that it
determines the hardness by indentation depth, rather than by measuring the diagonal
lengths of an indentation. The depth sensing capability is controlled via a piezo-electric
motor with a depth sensing accuracy of ±10nm. Dynamic hardness determines the
strength of the material based on both plastic and elastic deformation of a sample. In
contrast, conventional hardness is measured from the indentation dimension after the test
force is removed; the hardness of the material is derived solely from plastic deformation.
Nanoindenation is used over conventional microhardess here because a relatively small
applied load (<100mN) can be used in order to achieve an indentation depth that is less
than 10% (~2.5m) of the sample thickness. A small-applied load is required in this
study in order to avoid any substrate effects that could cause errors in the hardness
measurements.
Prior to hardness measurements, samples were electropolished at room
temperature using 70% phosphoric acid at 30VDC. The sample was then secured onto a
glass slide with acrylic based glue. Cyclic loading was used as the loading mode for
39
hardness measurements. Each measurement consisted of four loading and unloading
cycles. Each sample was measured three times using a maximum load of 70mN. The
load was fixed at 10% of the maximum load during the unloading period. A loading rate
of 6.67mN/s and a holding time of 60 seconds were used during each load-unload period.
The cyclic loading mode can yield reliable measurement for hardness.
Dynamic hardness as a function of grain size is listed in table 2.7.1 and shown in
Fig. 2.7.1a.
Table 2.7.1 – Results of hardness and electrical resistivity measurements obtained in this study.
Grain size
(nm)
Sample
description
Average hardness
value (GPa)
Standard
deviation
(nm)
Average
electrical
resistivity
(µohm-cm)
3238 EG-CRA Cu 0.47 3.3 1.78
1048 UFG Low-S Cu 0.84 11.8 1.84
325 EG-ED Cu 1.33 8.7 1.89
198 UFG Low-S Cu 1.30 5.1 1.96
86 nc Low-S Cu 1.52 6.9 2.1
73 nc Low-S Cu 1.58 2.0 2.26
49 nc Low-S Cu 1.69 7.2 2.34
31 nc High-S Cu 2.30 13.2 2.51
Nanocrystalline copper exhibits a significantly higher hardness than the
polycrystalline EG-CRA copper. A four-fold increase in hardness is observed as the
grain size is reduced from 3.2m to 31nm. EG-CRA copper with a grain size of 3.2µm
has a hardness value of 47 GPa, while the EG-ED copper, with an average grain size of
40
325nm, exhibits a hardness increase to about 1.33GPa. The hardness for high-S nc
copper with an average grain size of 31nm reached approximately 2.3GPa. In other
words, nanocrystalline copper, with a grain size of 31nm, showed an increase of
approximately 75% in the hardness compared to today’s industrial leading edge (EG-ED)
copper foil.
The Hall-Petch plot (H=H0+kd-1/2
) shown in Fig. 2.7.1b illustrates the hardness for
all electrodeposited copper samples as a function of the inverse square-root of grain size
(d-1/2
). The k and H0 values of the Hall-Petch plot shown in Fig. 2.7.1b are 8.82
GPanm1/2
and 0.56GPa, respectively. Chokshi et al., also observed the Hall-Petch
behavior of copper as grain size decreased from 25 to 8 m with measured k and H0
values of 18.6GPanm1/2
and 0.51GPa, respectively [Chokshi et al., (1989)]. The
calculated base hardness value (H0) in Fig. 2.7.1b corresponds well with the number
obtained by Chokshi et al. However, there is a considerable difference for the k values
which correspond to the slope of the Hall-Petch curve. The reason for the difference is
likely that Chokshi et al. used polycrystalline OFHC copper to achieve an average grain
size down to about 6µm, whereas the samples made for this current study covered a much
smaller grain size range. Fig. 2.7.1b also shows the hardness values of polycrystalline
copper obtained by Chokshi et al.
41
High-S
0
50
100
150
200
250
10 100 1000 10000
Dynam
ic H
ardness (G
Pa)
Grain size (nm)
This study
EG-ED
EG-CRA
2.5
2.0
1.5
1.0
0.5
H = 0.56+8.82d-1/2
0
0.5
1
1.5
2
2.5
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
Dyn
am
ic H
ard
ness (G
Pa
)
d-.5
(nm-.5
)
This study
Chokshi et al.
H-P curve fit for Cu samples made in this study
2000 600 200 60 30100
Fig. 2.7.1a Fig. 2.7.1b
Fig. 2.7.1a: Nano-indentation showing dynamic hardness for low-S, high-S and commercially available copper as function of grain size. Fig. 2.7.1b: Hall-Petch
plot of the synthesized electrodeposited copper.
2.7.2 – Electrical Resistivity Measurements Using the Four-Point Probe
Electrical resistivity testing is incorporated in this study to measure the resistance
to passage of electrical current in the materials with different grain sizes. The
conventional four-point probe technique was used to measure the direct current (DC)
electrical resistivity at room temperature. A schematic representation of the four-point
probe is shown in Fig. 2.7.3. Electrical resistivity () is the product of the resistance (R)
and cross-sectional area (A) per unit length (L) as shown by the following equation:
.IL
VA
L
RA (eqn 2.2)
Grain Size (nm)
42
The value for resistance is the measured voltage (V) divided by the applied
current (I). The four-point probe consists of a pair of outer current leads that provide
constant current (e.g. I=150mA) and a pair of inner potentials leads that are connected to
a voltmeter. For all electrical resistivity measurements, the samples were 30mm in length,
4mm in width with a cross sectional thickness of around 25µm. To measure the cross-
sectional area of each sample, the thicknesses and widths were measured using the
scanning electron microscope (SEM).
Fig. 2.7.3 – Schematic representation of the four-point probe used in electrical resistivity
measurements.
Table 2.7.1 and Fig. 2.7.4 show the electrical resistivity of the copper samples as
a function of their grain size. EG-CRA copper with an average grain size of 3.2µm
showed the lowest electrical resistivity value (1.77 cm). In comparing to EG-CRA
copper, 31nm High-S, 49nm low-S, 198nm low-S UFG and 325nm EG-ED copper
showed increases of respectively 41%, 31%, 10% and 6% in electrical resistivity.
The increase in electrical resistivity is likely due to the decrease in grain size and
impurity effects. Atoms located at grain boundaries and impurity atoms both are
effective scattering centers for electrons. They decrease the mean free path for the
electron and, ultimately, increase the electrical resistivity of copper.
43
1.65
1.75
1.85
1.95
2.05
2.15
2.25
2.35
2.45
2.55
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Ele
ctric
ia
l re
sistivity (m
oh
m-c
m)
Grain size (nm)
High-S
Low-S
PPC
EG-ED
EG-CRA
McCrea's Interpolation
Fig. 2.7.4 – Electrical resistivity results for copper foils made in this study and commercially available foils as a function of grain size.
McCrea quantified the effect of the increase in electrical resistivity as grain size
decreases [McCrea, (2001)]. The electrical resistivity of the material, according to
Matthiessen’s rule, is the sum of the resistivity due to lattice vibrations (phonons), and
the resistivity of crystals imperfections such as vacancies, dislocations, impurities, and
grain boundaries [Matthiessen and Vogt (1864)]. McCrea combined the effects of
dislocations, impurities and temperature into one general resistivity component, 0. The
total resistivity of the material as a function of grain size can then be given as:
gbtotal 0 (eqn 2.3)
where the resistivity due to grain boundaries, ρgb, can be expressed as:
dSGBRgb
37.2 (eqn 2.4)
44
where SGBR is the specific grain boundary resistivity and 2.37/d is the grain boundary
surface area per unit volume based on an idealized grain shape of a 14-sided
tetrakaidecahedron [DeHoff and Rhines (1968)]. Assuming that the dislocation density
and impurity levels are constant for different grain sizes, 0 is the electrical resistivity of
pure annealed electrodeposited copper and the SGBR value for copper taken from the
literature [Andrews et al., (1969)], the equation for the total electrical resistivity in copper
at room temperature can be expressed as:
cm )37.21012.3
667.1(6
d
total (eqn 2.5)
Equation 2.5, i.e. the McCrea electrical resistivity interpolation, is plotted in Fig.
2.7.4. It can be seen that in the polycrystalline regime (i.e: grain size >1000nm), the
resistivity interpolation is lower than the actual measured values. This is probably due to
the fact that ρ0 used in this equation was taken from copper with negligible impurities.
On the other hand, in the sub-100nm regime, the interpolation tends to give a higher
resistivity value. The exact reason for this is not known. One contributing factor could
be the difference in the grain boundary character distributions. It is known that special
grain boundaries have lower resistivity than random high angle grain boundaries
[Nakamichi (1990)]. If the special grain boundary fraction increases as grain size
decreases (e.g. through higher twin densities), the results will be a lower resistivity value
compared to the McCrea’s resistivity interpolation which did not take into consideration
any changes in grain boundary resistivity due to grain boundary character distribution.
45
2.8 – Chapter Summary
This chapter provided a brief review of previous research on the synthesis and
properties of electrodeposited nanocrystalline copper. This was followed by
experimental work dealing with the synthesis and characterization of various as-deposited
materials with different grain sizes and sulfur/carbon contents. It was shown that, as the
average grain size decreased to about 30nm, the hardness and electrical resistivity of
copper increased by 380% and 49%, respectively, compared to fully annealed pure
polycrystalline copper with an average grain size of 3.2µm.
This part of the research also provided materials for the study of the corrosion
behaviour and etchability of nanocrystalline copper. Yu studied the corrosion behaviour
of the electrodeposited copper foils with grain sizes ranging from 45nm to 1µm [Yu,
(2007)]. He also compared the corrosion behaviour of nanocrystalline copper and
commercially available cold-rolled annealed (EG-CRA) and electrodeposited (EG-ED)
copper. The results showed that grain size reduction had no significant effect on the
potentiodynamic polarization curves in both 0.1M NaOH [Yu et al., (2007)] and 0.1M
NaCl solutions [Yu et al. (2010)].
Kim studied the etchability of low-S nanocrystalline copper for use in printed
circuit boards [Kim (2007)]. His results showed that for nanocrystalline copper, the trace
resolution significantly improves over conventional polycrystalline cold-rolled annealed
and electrodeposited copper as predicted in section 1.1. Fig. 2.8.1 presents cross-
sectional optical images showing the etch profile improvements over polycrystalline EG-
46
ED copper foil as grain size decreases. As the copper line-width continues to shrink in
PWB design, the straight edge etching behaviour of nanocrystalline copper will be
beneficial in reducing bridging/shorting defects during PWB manufacture.
Fig. 2.8.1 – Cross sectional optical images of etched EG-ED (left) and nc low-S (right) copper [Kim (2007)].
The increase in strength and improvement in etch profile proximity found in
nanocrystalline copper should also provide considerable improvements in the reliability
of PWB’s. With respect to CTE, nanocrystalline copper offers additional strength for
bearing additional load and the improvement in etch proximity reduces points for stress
concentration which ultimately causes failure of electronic device (refer to section 1.1).
The remainder of this thesis will focus on the thermal stability of electrodeposited
nanocrystalline copper. The next chapter will first present a literature review on the
thermal stability of electrodeposited nanocrystalline materials in general. Previous work
on the thermal stability of nanocrystalline copper will be given at the end of the chapter.
47
2.9 – References for Chapter 2
Andrews P.V., West M.B. and Robenson C.R., Phil. Mag., 19 (1969) 887
Cheung C., Djuanda F., Erb U. and Palumbo G., Nanostr. Mater., 5 (1995) 513
Chokshi A.H., Rosen A., Karch J. and Gleiter H., Scripta Mater., 23 (1989) 1679
Cullity B.D., Elements of X-ray diffraction, 2nd
Ed., Addison-Wesley Publishing
Company, Massachusetts (1978) 140
Cziraki A., Gerocs I., Toth-kadar E. and Bakonyi I., Nanostr. Mater., 6 (1995) 547
DeHoff R.T. and Rhines F.N. (ed.), Quantitative Microscopy, McGraw-Hill, New York
(1968)
Dini J.W. (ed.) Electrodeposition – The Materials Science of Coatings and Substrates,
William Andrew Puslishing/Noyes, NY, (1993)
El-Sherik A.M. and Erb U., J. Mater. Sci., 30 (1995) 5743
Erb U., El-Sherik., Cheung C. and Aus M.J., US Patent #5433797, (1995)
Erb U., Nanostr. Mater., 6 (1995) 533
Erb U. and El-Sherik A.M., US Patent #5352266, (1994)
Glasstone., S. Trans. Faraday Society, 31 (1935) 1232
Hakamada M., Nakamoto Y., Matsumoto H., Iwasaki H., Chen Y., Kusuda H. and
Mabuchi M., Mater. Sci. and Eng., A457 (2007) 120
Huang C.A., Kao Y.L., Tu G.C. and Chang J.H., Mater. Sci. and Eng., A382 (2004) 104
Karanassios V. and Greib L., Acta Spectrochemica, 61B (2006) 164
Kim S., 4th
year thesis, University of Toronto (2007)
Lamb V. A., Johnson C. E and Valentine D. R., J. of Electrochem. Soc., 117 (1970)
291c-318c, 341c-352c, 381c-404c
LECO induction furnace instrument manual – carbon and sulfur determination (2007)
Lowenheim F.A., Electroplating, McGraw-Hill, Inc. USA. (1978)
48
Lu K., Qian L.H., Lu Q.H. and Kong W.J., Scripta Mater., 50 (2004) 1407
Lu K., Qian L.H., Chen X., Shen Y. and Lu L., Science 304 (2004) 422
Lu K., Sui M. L. and Lu. L., Science, 287 (2000) 1463
Ma E., Scripta Mater., 49 (2003) 663
Matthiessen A. and Vogt C., Ann. Phys., 122 (1864) 19
McCrea J., Ph.D. thesis, University of Toronto (2001)
Merchant H.D., Liu W.C., Giannuzzi L.A. and Morris J.G., Mater. Chara., 53 (2004) 360
Nakamichi I., J. Sci. Hiroshima Univ., Ser A, 54 (1990) 49
Natter H. and Hempelmann R., J. of Phys. Chem., 100 (1996) 19525
Nieman G.W., Weertman J.R. and Siegel R.W., Scripta Met., 23 (1989) 2013.
Paik J.M., Park Y.J., Yoon M.S., Lee J.H. and Joo Y.C., Scripta Mater., 48 (2003) 683
Safranek W.H., The Properties of Electrodeposited Metals and Alloys, American Elsevier
pub. Co. (1974)
Sastry S.M.L., Suryanarayanan. R., Frey C.A., Waller B.E. and Buhro W.E., Mater. Sci.
& Eng., A264 (1999) 210
Seah C.H., Mridha S. and Chan L.H., J. of Mater. Proc. Tech., 89 (1998) 432
Suryanarayanan R., Claire A., Shankar M.L., Waller B.E., Bates S.E. and Buhro W.E., J.
of Mater. Res., 11 (1996) 439
Tao S., and Li D.Y., Nanotechnology, 17 (2006) 65
Wang Y.M., Wang K., Pan D., Lu K., Hemker K.J. and Ma E., Scripta Mater., 48 (2003)
1581
Weil R. and Hong S., Thin Solid Films, 283 (1996) 175
Yu B., Woo P., and Erb U., Scripta Mater., (2010) Submitted
Yu B., Woo P., and Erb U., Scripta Mater., 56 (2007) 353
Yu B., Master thesis, University of Toronto (2007)
Zhang X., Wang H. and Koch C. C., Rev. Adv. Mater. Sci., 6 (2004) 53
49
Chapter 3 – Previous Work on Thermal Stability of Nanocrystalline Materials
3.1 - Introduction
Nanostructure science has become one of the most popular areas in research and
development over the past two decades. Gleiter, in the early 1980’s first produced
nanocrystalline materials by inert gas condensation in reducing their grain size to less
than 100nm [Gleiter (1981) and Gleiter (1984)]. Since that time, enormous research
efforts have been concerned with the study of properties of nanocrystalline materials
produced by various synthesis methods.
Three dimensional nanomaterials are generally defined as polycrystals with grain
sizes below 100nm. Numerous studies have shown that, as grain size decreases into the
nanocrystalline regime, materials will start to exhibit desirable properties such as
increased yield strength, hardness and wear resistance [Gleiter (1989), Erb et al. (1997)
and Koch (2007)]. To explain some of the desirable properties, Palumbo et al. developed
a model to estimate the change in the intercrystalline (grain boundaries and triple
junctions) volume fraction of the material with respect to changes in grain size [Palumbo
et al. (1990)]. Their results showed that, as grain size decreases from 1000nm to 2 nm,
the intercrystalline volume fraction increases from 0.3% to 87.5% (Fig. 3.1.1 and
appendix 2). In other words, as the grain size decreases, the structure of the material
changes from a crystal-dominant to an intercrystalline defect-dominant solid. Hence,
many properties of the material will change considerably.
50
Fig. 3.1.1 - Effect of grain size (d) on the volume fraction for intercrystalline regions (grain boundaries and triple junctions), assuming a grain boundary thickness of 1
nm [Palumbo et al., (1990)].
There are many techniques for making nanocrystalline materials, such as inert gas
condensation (IGC), chemical synthesis (CS), magnetron sputtering (MS), mechanical
attrition (MA) and electrodeposition (ED). IGC based techniques can produce a wide
range of metals and alloy powders but the final consolidated products often suffer from
problems associated with residual porosity. For example, residual porosity has been
shown to be the main cause to the reductions in the Young’s modulus [El-Sherik et al.
(1993) and Nieman et al. (1991)]. CS also yields products mainly in powder form.
Consequently the synthesis of bulk material requires secondary processing such as
powder consolidation and compacting. Materials synthesized by MA often have the risk
of incorporating high amounts of impurities due to the abrasion from the attrition
apparatus [Koch (1996)]. On the other hand, ED is able to produce fully dense,
nanocrystalline metals and alloys using large scale production facilities with minimal
extra capital costs [El-Sherik et al. (1993b)].
51
3.2 – Thermal Instability of Nanocrystalline Materials
Despite their unique property enhancements, applications of nanocrystalline
materials are often limited due to their susceptibility to grain growth at elevated
temperatures. Nanocrystalline materials have a high driving force for grain growth. A
diffusion study done by Birringer et al. showed that the increase in the intercrystalline
volume fraction in nanocrystalline materials enhances the self and solute diffusion rates
compared to their polycrystalline counterparts [Birringer et al. (1987)]. This
enhancement in diffusivity promotes grain boundary migration and allows the material to
reduce its excess interfacial energy associated with the high volume fraction of grain
boundaries.
In order to study the growth phenomena of nanocrystalline materials, it is
important to first consider the fundamentals of grain growth. The driving force for grain
growth, F, can be expressed mathematically in the following form [e.g. Haessener
(1978)]:
d
GBF
(eqn 3.1)
where is the grain shape factor, GB is the grain boundary energy and d is the grain size.
It can be seen that the driving force for grain growth increases with decreasing grain size.
A simple version of a grain growth kinetics equation is as follows:
d-do=Ktn
(eqn 3.2)
where d is the grain size after time t, do is the initial grain size, and n is the grain growth
exponent. Since grain growth is a diffusion process, it follows an Arrhenius type
behaviour which the constant K expressed as:
52
K=A exp(-Q/kT) (eqn 3.3)
where A is the pre-exponential factor, Q is the activation energy for grain growth, k is the
Boltzman constant and T is the temperature.
Equation 3.3 shows that materials with higher activation energy will exhibit better
thermal stability. Several studies have reported that the activation energies for
electrodeposited nanocrystalline materials is close to the activation energy value for grain
boundary self-diffusion [Mehta et al. (1995), Wang et al. (1997), Hibbard (2002), and
Ebrahimi and Li (2006)].
There are many investigations that have demonstrated the use of dragging forces
to counteract the driving force for grain growth, and hence to stabilize the nanocrystalline
microstructure [Malow and Koch (1996), Mehta et al. (1995) and Hibbard et al., (2002)].
The common dragging mechanisms in porosity-free nanomaterials are solute drag and
Zener particle drag. The dragging forces for solute (Fs) and Zener (Fz) are given by the
following equations:
Fs=C0GB/rs (eqn3.4)
Fz=2GBf/R (eqn3.5)
where GB is the boundary energy, C0 and rs are the average concentration and atomic
radius of the solute, f and R are the volume fraction and radius of the second phase
particles, respectively [Haessner, (1978)]. From these equations, it can be seen that
solute drag can often achieve a stronger dragging force compared to Zener particle drag
53
because the atomic radius of the solute is usually much smaller than the radius of the
particle (rs<<R).
There are numerous studies that examined the thermal stability of nanocrystalline
materials. However, considerable differences have been reported in terms of stability
temperature, differential scanning calorimetry (DSC) heat exotherm profiles, and grain
growth kinetics. These discrepancies can likely be attributed to different synthesis
techniques and impurity levels. The same material made by different synthesis
techniques can show significant differences in its thermal stability due to differences in
defect density, bulk density and grain size distribution. Thermal instability could become
a potential obstacle for applications of nanocrystalline copper in PWB design due to the
potential of microstructural instability at elevated temperature.
This literature review will focus on the thermal stability of electrodeposited
nanocrystalline materials. Since the remainder of the thesis deals with the thermal
stability of nanocrystalline copper, a short summary on the thermal stability of
nanocrystalline copper obtained by other synthesis techniques will also be provided at the
end of this chapter.
3.3 – Thermal Stability of Electrodeposited Nanocrystalline Materials
3.3.1 – Investigation Techniques
There are several techniques that have been used to study the thermal stability of
nanocrystalline materials. These techniques are basically divided in two groups: in situ
54
and ex situ monitoring. In situ techniques are characterization techniques that
continuously monitor the microstructural evolution upon annealing. Over the years,
several in situ techniques have been used including transmission electron microscopy
(TEM) [Boylan et al. (1991), El-Sherik et al. (1992) and Klement et al. (2010)], X-ray
diffraction (XRD) [Natter et al. (1998)], electrical resistivity measurements [McCrea et al.
(2001)] and differential scanning calorimetry (DSC) [Klement et al. (1995), Choi et al.
(2005), Hibbard et al. (2001, 2002 and 2006), Zhou (2006) and Klement and Da Silva
(2007)]. DSC is among the most popular in situ monitoring techniques for grain growth
studies because the annealing conditions (i.e: heating rate, start and end temperature) can
be accurately controlled. DSC analysis also provides many valuable quantitative data
such as the peak grain growth temperature (Tp), total heat enthalpy releases (H), and as
well as the activation energy for grain growth (Q).
In ex situ techniques, the material is first subjected to different pre-determined
temperature/time annealing conditions and subsequently prepared for microstructural
characterization. Hence, ex situ annealing is not a real-time monitoring technique and
certain stages of microstructural evolution could be missed. In this approach, the
annealed structures can be examined using a variety of characterization methods such as
TEM [Boylan et al. (1991), and El-Sherik et al. (1992)], atomic probe field ion
microscopy (APFIM) [Hentschel et al. (2000)], and 3-dimensional atomic probe (3DAP)
[Farber et al. (2000) and Choi et al. (2005)].
55
In this thesis, DSC will be the main technique used to give a real-time overview
of the thermal stability of nanocrystalline copper. Ex situ TEM was also performed for
microstructural characterization after different DSC annealing treatments.
3.3.2 – Results of Previous Studies
The thermal stability of nanocrystalline electrodeposits has been investigated in
numerous studies [e.g: Boylan et al. (1991), El-Sherik et al. (1992), Cziraki et al. (1994),
Klement et al. (1995), Mehta et al. (1995), Bryden et al. (1997), Czerwinski et al. (1997),
Turi (1997), Palumbo et al. (1992), Wang et al. (1997), Natter et al. (1998), Farber et al.
(2000), Hentschel et al. (2000), Thuvander et al. (2000), Abraham et al. (2002), McCrea
(2001), Hibbard (2002), Hibbard et al. (2002), Choi et al. (2005), Hibbard et al. (2006),
Zhou (2006), Klement et al. (2007), Klement and Da Silva (2007) and Klement et al.
(2010)]. Nanocrystalline nickel and nickel alloys (Ni-S, Ni-P, Ni-Fe and Ni-Co) are
among the most widely studied materials in terms of grain growth evolution upon
annealing. Boylan et al. and El-Sherik et al. were among the very first in studying the
thermal stability of electrodeposited nanocrystalline nickel [Boylan et al. (1991), and El-
Sherik et al. (1992)]. Boylan observed that Ni-1.2wt.%P with average grain size of about
5-10nm is stable at 200C with no grain growth during in situ TEM annealing. Upon
annealing to 350C, the grain size initially increased by a factor of 2-3 times, and then the
microstructure rapidly stabilized. El-Sherik et al., also conducted in situ TEM annealing
on Ni-0.12wt.%S and Ni-1.2wt.%P, with respective initial grain sizes of 10 and 6 nm.
Both materials were isothermally annealed at 300C and the microstructure displayed a
uniform transition from initial rapid grain growth (the initial 60 minutes) to a stabilized
56
grain structure (after 60 minutes) with final grain sizes of less than 25nm. Isothermal
annealing at 350C was also conducted on the Ni-1.2wt.%P samples and showed rapid
growth from the initial 7nm to a final grain size of 75nm in 50 minutes. Boylan et al. and
El-Sherik et al. suggested that the stabilized microstructures were mainly due to pinning
of grain boundaries by precipitation/Zener drag and triple junction drag.
In 1995, Klement investigated the thermal stability of nanocrystalline Ni further
by using differential scanning calorimetry (DSC) [Klement et al. (1995)].
Electrodeposited nickel containing approximately ~0.158wt.%S, with 10nm initial grain
size, showed a two stage exothermic heat release (Fig. 3.3.1). The first heat release
consisted of a broad reaction in the temperature range of 80-260C, followed by a more
dominant exothermic peak at around 290C. TEM was also conducted using samples
isothermally annealed for 30 minutes at different temperatures ranging from 120 to
420C. Klement et al. was able to identify that the first broad exothermic process (80-
289C) corresponded to a transformation where abnormal grain growth begins, followed
by the main heat release which corresponded to normal grain growth (~290-370C) and
grain growth toward equilibrium (370-500C). The definitions for normal and abnormal
grain growth are given in appendix 3. In contrast to the observation by El-Sherik et al.,
Klement et al. did not observe any microstructure stabilization as the material was
annealed to 300C. This suggests that the grain size stabilization observed by El-Sherik
et al. by in situ TEM was likely due to grain boundary pinning in thin TEM specimens.
57
Klement et al. also suggested that the initial abnormal grain structure was possibly
due to sub-grain coalescence and grain rotation. Wang et al., also reported that the
mechanism leading to abnormal grain growth is probably subgrain coalescence [Wang et
al. (1997)]. Observations using TEM showed that many of the grains in nano Ni in the
as-deposited state are only slightly misoriented from each other. If these subgrains rotate
slightly upon annealing, they will merge and form grains with significantly larger
dimensions.
Fig. 3.3.1 – DSC scans of 10nm Ni-0.158%S; heating rate 10C/min (heat exotherm downward) [Klement et al., (1995)].
Klement suggested that the normal/abnormal grain growth morphology is highly
depended on the annealing temperature [Klement et al. (1995)]. Her work also showed
that in situ TEM may not be a proper technique in studying the true mechanisms for grain
growth in reflect to its bulk form due to grain boundary pinning from the thin foil effect.
In 2002, Hibbard et al., conducted a comprehensive ex situ grain growth study
that covered a broad range of annealing conditions, followed by TEM for each annealing
condition. It was suggested that many of the previously observed microstructures (e.g.
stabilized nanostructure: [Boylan (1991) and El-Sherik (1992)], abnormal to normal
58
[Klement et al. (1995)] grain structures) are all particular snapshots of a multi-staged
growth process upon annealing.
Following isothermal annealing of nano Ni (starting grain size: 20nm, S content:
850ppm, C content 350ppm) at 420C, his results showed that large grains were formed
to the sub-micron range in ~1 second of annealing. After 30s of annealing, the
abnormally growing grains had completely consumed the nanocrystalline matrix and the
average grain size rapidly changed from the initial 20nm to ~350nm. After the initial
rapid grain growth, the growth rate slowed down significantly and the growth
morphology became much more uniform. An average grain size of ~450nm was reached
after 60 minutes of annealing. As the isothermal annealing reached 11 hours, the
microstructure showed an abnormal grain structure to re-appear, with the large grains
ranging in sizes of 5-50 m and having planar growth fronts. These planar growth fronts
were likely due to the accumulation of the sulfur in the grain boundaries [Hibbard (2001)].
As annealing reached 120hrs, these abnormally growing grains disappeared and a
uniform polycrystalline microstructure was observed. The stabilized nanostructure
observed by Boylan and El-Sherik was never observed here through all annealing
conditions. Hibbard suggested that the stabilized nanostructures observed by Boylan
[Boylan et al. (1991)] and El-Sherik [El-Sherik et al. (1992)] were probably due to grain
boundary pinning by the free surfaces during in-situ TEM annealing.
Hibbard et al., also performed calorimetric studies on nanocrystalline Ni(-S), and
Co electrodeposits [Hibbard (2001)]. Conventional DSC was conducted to measure peak
59
grain growth temperature as a function of heating rates. Fig. 3.3.2 show a modified
Kissinger analysis that gives the activation energy for grain growth. From calorimetric
measurements, higher activation energies represent higher thermal stability of the
materials. This graph also shows that the thermal stability of nanocrystalline Ni
electrodeposits can be substantially increased with the addition of alloying elements such
as P [Mehta et al. (1995)] and Fe [Turi (1997)]. Hibbard et al., also showed that the
thermal stability increases as the cobalt concentration increases in the Ni-Co alloy system
[Hibbard et al. (2006)].
Fig. 3.3.2 – Modified Kissinger analysis for Ni-S, Ni-Fe, Ni-P and Co
[Hibbard et al. (2002)].
Moreover, the experimental work for Ni and Co alloys presented in Fig. 3.3.2 also
showed that the activation energy for grain growth is closely related to the activation
energy value for grain boundary self-diffusion. This is a good indication that the growth
mechanism for nanocrystalline electrodeposits is probably controlled by grain boundary
diffusion [Hibbard et al. (2002)].
60
The work by Mehta et al., Turi, and Hibbard et al. showed that the thermal
stability and microstructural evolution in nanocrystalline materials is highly influenced
by the alloying/impurity level. In a continuation of this work, Hibbard et al., in 2006,
showed that the thermal stability of the material is highly dependent on the sulfur
impurity level [Hibbard et al. (2006)]. In this experiment, Hibbard et al. used a series of
nanocrystalline Co samples with similar grain size distributions but different sulfur and
carbon impurity levels. These Co samples were annealed through DSC with various
heating rates to obtain the activation energies for grain growth. By plotting peak
temperature (Tp), versus either grain size (d), carbon (C) and sulfur (S) concentration, as
shown in Fig. 3.3.3, Hibbard et al. were able to show that sulfur has the dominant effect
on the thermal stability of the material.
Fig. 3.3.3 – Peak temperature [Tp] as a function of: (a) grain size, (b) carbon
concentration, and (c) sulfur concentration [Hibbard et al. (2006)].
Hibbard et al. also proposed a segregation mechanism in which abnormally
growing grains sweep through the nanocrystalline matrix and collect sulfur as they
migrate [Hibbard et al. (2002)]. It was suggested that the segregation is dynamic in a
way that sulfur impurities and grain boundaries are both moving into each other. As a
61
migrating boundary collects these impurities, the excess energy associated with that
migrating boundary decreases; hence the thermal stability of the material increases.
Impurity segregation, in this case, phosphorus in nickel or cobalt were also
observed by using tomographic atomic probe (TAP) [Hentschel et al. (2000)] and
3 dimensional tomographic atomic probe (3DAP) [Farber et al. (2000), and Choi et al.
(2005)]. All three studies showed very similar results. For example, Choi et al., used
3DAP and showed that the P-distribution in the as-deposited nanocrystalline Co-P was
non-uniform throughout the sample. Isothermal annealing at 400C for 1h showed that
the material was thermally stable. At 460C for 1h, abnormal grain growth began and the
3DAP analysis showed that the P concentration along the grain boundaries increased with
grain size. Further analysis also showed that phosphorus was segregated and CoP/Co2P
was formed in the grain boundaries after isothermal annealing at 480C for 1hr. The
authors further observed that the formation of these second phase particles was then
followed by a change from abnormal to normal grain growth.
These studies are in agreement with Hibbard’s dynamic segregation theory in
which segregation upon annealing slows down the rate of grain boundary migration.
Moreover, second phase particles begin to form when the impurity level at the grain
boundaries increases to a certain threshold value.
In 2006, Zhou performed a series of calorimetric annealing experiments using
nanocrystalline Ni~2.5wt.%P with various heating rates (5-80C/min) [Zhou (2006)].
62
DSC analysis showed that samples with small average grain size (<7nm) and a narrow
grain size distribution exhibited exothermic peaks which consisted of a first broad flat
plateau (100-400C), which was followed by a major heat release (400-550C), as shown
in Fig. 3.3.4a. In contrast, samples with a larger average grain size (12-29nm) and a
wider grain size distribution revealed a peak superimposed on the flat plateau, followed
by a major heat release (Fig. 3.3.4b). In conjunction with DSC, TEM analysis was
conducted on some samples DSC annealed to certain temperatures, predetermined based
on the changes in the DSC thermal signature. For the first group of samples with narrow
grain size distributions, TEM showed that grain growth was uniform as samples were
annealed up to 550C. On the other hand, for samples with a wider starting grain size
distribution, the annealed structures showed abnormal grain growth after annealing up to
400C. Further analysis showed that the larger grains underwent tremendous growth
from the initial size, while the majority of the small grains showed only minimal growth.
Zhou’s experiments indicated that larger grain size and/or wider grain size distribution
can lead to abnormal grain growth upon annealing.
Fig. 3.3.4 – DSC scans of Ni~2.5wt.%P of materials with: (left) smaller grain size
and narrow grain size distribution [first group], and (right) larger grain size with wide grain size distribution [second group] [Zhou (2006)].
a b
63
Zhou et al. also showed that the abnormally growing grains form rapidly during
the first heat release, and their growth slowed down substantially during the major heat
release. However, all of the samples showed that the matrix grew substantially after
annealing through the major (second) heat release. Zhou et al. explained this on the basis
of the solute dragging force that retards the growth of the matrix grains during the first
heat release; in stage II this dragging force diminished significantly resulting in growth of
the matrix. From TEM diffraction patterns it was shown that the loss in dragging force
coincided with the formation of Ni3P particles during the major heat release. With the
formation of second phase particles, the dragging force changed from the initially strong
solute drag to a much weaker Zener drag. In addition to the weakening of the dragging
force, the driving force for grain growth also increases as temperature increases, these
two combined effects ultimately causing rapid grain growth.
From this literature review, it can be concluded that a low concentration of
alloying elements (i.e: sulfur or phosphorus) can greatly enhance the thermal stability of a
nanocrystalline material. Hibbard et al. have shown that the thermal stability of
nanocrystalline cobalt is not strongly dependent on either average grain size and carbon
impurity [Hibbard et al. (2006)]. In contrast, the addition of sulfur greatly enhanced the
thermal stability of the material. Hibbard’s study also indicates that the planar growth
front in late stage abnormal growth is probably caused by the sulfur enrichment and
subsequent formation of a wetting phase along the grain boundaries [Hibbard (2002)].
64
Zhou studied thermal stability of nanocrystalline material also from another
perspective. He focused on how the initial grain size distribution affects the
microstructural evolution upon annealing. His results showed that during annealing up to
the first heat exothermic range (100-400C, Fig. 3.3.4), abnormal grain growth only sets
in if the initial microstructure has a larger initial grain size and a wider grain size
distribution [Zhou (2006)]. Rapid grain growth occurred in all samples during the second
exothermic event (400-550C) where the dragging force decreases as a result of Ni3P
precipitate formation.
The experimental results presented by Hibbard and Zhou showed that sulfur
impurities and initial grain size distribution can have a huge effect to the thermal stability
and microstructural evolution of nanocrystalline nickel and cobalt. It is also important to
note that all the nickel samples that have been studied contained high amounts of sulfur
ranging from 200-1300ppm.
In order to better understand how sulfur affects the thermal stability of
nanocrystalline electrodeposits, one needs to produce materials with the ability to control
the level of sulfur incorporated into the deposits. In the current study, copper
electrodeposits were produced with grain sizes down to 30nm, and sulfur impurity
concentrations ranging from 30-250ppm.
The next section provides a brief literature review on the thermal stability of
nanocrystalline copper produced by different methods.
65
3.3.3 - Thermal Stability of Nanocrystalline Copper
There are only a few previous studies that have specifically addressed the thermal
stability of nanocrystalline copper. More specifically, most of these studies were done
using nanocrystalline copper that was made by inert gas condensation (IGC) rather than
electrodeposition.
There are a few previous studies that reported the thermal stability of
electrodeposited nanocrystalline copper through DSC measurements. Lu et al. showed
DSC curves consisting of a very minor exothermic peak at around 155°C, with a total
heat release of 0.05J/g [Lu et al. (2000)]. The authors mentioned that the weak thermal
response of grain growth was likely due to the fact that the material was mainly
comprised of low angle boundaries. The authors also mentioned that the heat release was
likely influenced by a stress relief in the microstructure during DSC annealing. Kao et al.,
conducted calorimetric studies of electrodeposited ultra-fine grain copper (AGS<200nm)
produced from plating baths with varying thiourea concentration (0-8mg/L) [Kao et al.
(2004)]. The results showed that the DSC peak temperature and heat release increase
with thiourea concentration.
Gunther et al., produced IGC and compacted copper with an average grain size of
40nm, with ~2% residual porosity [Gunther et al. (1992)]. TEM examination showed
that abnormal grain growth began even at ambient temperature (20C). Gertsman and
Birringer. further looked at the room temperature grain growth behaviour of IGC and
compacted copper and found that the microstructure can be stabilized by incorporating
66
porosity into the copper matrix [Gertsman and Birringer. (1994)]. Fig. 3.3.5 illustrates
that for nanocrystalline copper with 7% residual porosity (sample 1), the grain size has
changed from 22 to 25nm 17 days after synthesis. In contrast, nanocrystalline copper
with only 3% residual porosity (sample 3) showed a grain size change from 32 to 55nm.
Thermal stability studies were also conducted using nanocrystalline copper
synthesized by filter cathodic arc [Cao and Zhang (2006)] and gas deposition [Okuda et
al. (2001)]. A one micron thick copper film was produced using the cathodic arc
technique. Cao and Zhang observed rapid initial grain growth followed by stabilized
growth structures. This stabilized microstructure was likely due to surface pinning of the
grain boundaries similar to what has been observed by El-Sherik et al. and Boylan et al.
[Boylan et al. (1991), El-Sherik et al. (1992)]. For the gas deposited nanocrystalline
copper samples [Okuda et al. (2001)], the results showed that specimens with smaller
initial grain size and strongest preferred orientation had the highest thermal stability
(stable microstructure up to 300C). The thermal stability was much lower for samples
with <100> or no preferred orientation, compared to samples with <111> preferred
orientation.
It is important to note that gas deposited material may contain impurities in
addition to porosity. For instance, gas deposited nanocrystalline gold can contain He
impurities from the inert helium gas used during the synthesis process in addition to a
large percentage of residual porosity [Okuda et al. (2001)]. Both porosity and helium
impurities would likely enhance the thermal stability of the material.
67
Fig. 3.3.5 – Grain size as a function of dwell time at room temperature for IGC&C Cu made with different residual porosities. Sample 1: 7%; Sample 2: 4%; Sample 3:
3% [Gertsman and Birringer (1994)].
Pantleon and Somers, electrodeposited thin nanocrystalline copper films from
commercially available copper electrolytes [Pantleon and Somers (2006)]. The deposited
film thickness ranged from 400nm to 5m. All synthesized samples were self-annealed
at room temperature and the results showed that the grain growth is strongly depended on
the thickness of the copper deposit. In this paper, grain size measurements were done
using the Scherrer formula in conjunction with XRD. Grain size measurements based on
peak broadening revealed that the grain size for the {111} oriented grains was much
larger than grains in the {100} orientation (50 vs. 15nm). A faster growth kinetics was
found for the smaller {100} oriented grains, and shortly after they started growing, they
become even larger than the grains in the {111} orientation. This self-annealing
experiment by Pantleon and Somers indicated that the grain growth kinetics also depends
68
on the film thickness in addition to grain size and the orientation of the grains. The
thinner the copper layer, the slower the growth kinetics due to surface pinning effects.
The result also showed that self-annealing at room temperature is completely suppressed
when the material thickness of the sample was reduced to ~400nm.
69
3.4 – References for Chapter 3
Abraham M., Thuvander M., Lane H., Gerezo A. and Smith G.D.W., Mater. Res. Soc.
Symp. Proc., 581 (2002) 517
Birringer R., Gleiter H. and Horvath J., Sol. State Comm., 62-5 (1987) 319
Bryden K.J. and Ying J.Y., Nanostr. Mater., 9 (1997) 485
Boylan K., Ostrander D., Erb U., Palumbo G. and Aust K.T., Scripta Met. et Mater., 25
(1991) 2711
Cao P. and Zhang D., J. of Mod. Phy., 20 (2006) 3830
Choi P., da Silva M., Klement U., Al-Kassab T. and Kirchheim R., Acta. Mater., 53
(2005) 4473
Czerwinski F., Li H., Megret M. Szpunar J.A., Clark D.G. and Erb U., Scripta Mater., 37
(1997) 1967
Cziraki A., Tonkovics Z., Gerocs I., Fogarassy B., Groma I., Toth-Kadar E., Tarnoczi T.
and Bakonyi I., Mater. Sci. Eng. A, A179/180 (1994) 531
Ebrahimi F. and Li H., Scripta Mater., 55 (2006) 263
El-Sherik A.M., Erb U., Palumbo G. and Aust K.T., Nanostr. Mater., 2 (1993a) 383
El-Sherik A.M., Erb U., Krstic V., Szpunar B., Aus M.J., Palumbo G. and Aust K.T.,
Mater. Res. Soc. Symp. Proc., 286 (1993b) 173
El-Sherik A.M., Boylan K., Erb U., Palumbo G. and Aust K.T., Mater. Res. Soc. Symp.
Proc., 238 (1992) 727
Erb U., Palumbo G., Szpunar B. and Aust K.T., Nanostr. Mater., 9 (1997) 261
Farber B., Cadel E., Menard A., Schmitz G. and Kirchheim., Acta Mater., 48 (2000) 789
Gertsman V.Y. and Birringer R., Scripta Met. et Mater., 30 (1994) 577
Gleiter H., in Deformation of Polycrystals: Mechanisms and Microstructures, Proc. 2nd
Riso. Int. Symposium on Metall. and Mater. Sci. National Laboratory, Roskilde,
Denmark (1981) 15
Gleiter H., Phys. Stat. Sol. B., 172 (1992) 41
70
Gleiter H., Prog. In Mater. Sci., 33 (1989) 223
Gleiter H., Mater. Res. and Adv. Techniques., 75 (1984) 263
Günther B., Kumpmann A. and Kunze H.D., Scripta Mater., 27 (1992) 833
Haessner F. (ed.), Recrystallization of Metallic Materials, Dr. Riederer Verlag GmbH,
Stuttgart, Germany (1978)
Hentschel T., Isheim D., Kirchheim R., Muller F. and Kreye H., Acta Mater., 48 (2000) 933
Hibbard G., Aust K.T. and Erb U., Acta Mater., 54 (2006) 2501
Hibbard G., Aust K.T. and Erb U., Mater. Sci. and Eng. A, 433 (2006) 195
Hibbard G., in Processing and Fabrication of Adv. Mater. XV, MS&T06, Cincinnati, OH,
(2006) 287
Hibbard G., Erb U., Aust K.T., Klement U. and Palumbo G., Mater. Sci. For., 386-388 (2002)
387
Hibbard G., Ph.D. Thesis, University of Toronto (2002)
Hibbard G., Aust K.T., Palumbo G. and Erb U., Scripta Mater., 44 (2001) 513
Kao Y.L., Tu G.C., Huang C.A. and Chang J.H, Mater. Sci & Eng., A382 (2004) 104
Kim B., Jane’s Defence Weekly., Nov (2005) 373
Klement U., Hollang L., Dey S.R., Battabyal M., Mishin O.V. and Skrotzki W., Solid
State Phenomena, 160 (2010) 235
Klement U., Da Silva M., Wille C., Choi P., and Al-Kassab T., Mater. Sci. & Eng.,
A445-446 (2007) 31
Klement U., and Da Silva M., Sino-Swedish Struc. Mater. Sym. 2007 Proc., (2007) 173
Klement U., Erb U., El-Sherik A.M. and Aust K.T., Mater. Sci. & Eng., A203 (1995) 177
Klement U., Erb U. and Aust K.T., Nanostr. Mater., 6 (1995) 581
Koch C. C (ed.), Nanostructured Materials – Processing, Properties and Potential
Applications, 2nd
edition, William Andrew Publishing, Norwich, New York (2007)
Koch C.C. and Whittenberger J.D., Intermetallics , 5 (1996) 339
Lu K., Lu L., Wang L.B. and Ding B.Z., Mater. Sci & Eng., A286 (2000) 125
71
Malow T.R. and Koch C.C., Mater. Sci. For., 225-227 (1996) 595
McCrea J. L. Aust K.T. Palumbo G. Erb U., Mater. Res. Soc. Symp. Proc., 581 (2000)
461
Mehta S.C., Smith D.A. and Erb U., Mater. Sci. and Eng. A, 204 (1995) 227
Natter H., Schmelzer M. and Hempelmann R., J. of Mater. Res., 13 (1998) 1186
Nieman G.W., Weertman J.R. and Siegel R.W., J. of Mater. Res., 6 (1991) 1012
Okuda S., Kobiyama M., Inami T. and Takamura S., Scripta Mater. 44 (2001) 2009
Palumbo G., Gonzalez F., Brennenstuhl A.M., Erb U., Shmayda W. and Lichtenberger
P.C., Nanostr. Mater., 1 (1992) 77
Palumbo G., Thorpe S.J. and Aust K.T., Scripta Met. et Mater., 24 (1990) 1347
Pantleon K. and Somers M.A.J., J.Appl. Phys., 100 (2006) 114319-1
Thuvander M., Warren P.J., Abraham M., Lane H., Gerezo A. and Smith G.D.W., Mater.
Sci. For., 701 (2000) 343
Turi T., Ph.D. Thesis, Queen’s University (1997)
Wang N., Wang Z., Aust K.T. and Erb U., Acta Mater., 45 (1997) 1655
Zhou Y., Ph.D. Thesis, University of Toronto (2006)
72
Chapter 4 – Experimental Methods Used to Study Thermal Stability of Nanocrystalline Copper
This chapter outlines the experimental methods that were used in studying the
microstructural evolution upon annealing of nanocrystalline copper made by
electrodeposition in this study. The annealing experiment strategy employed in this work
will be presented.
4.1 - Annealing Experiment Strategy
To better understand the microstructural evolution of electrodeposited
nanocrystalline copper, bulk freestanding foils were synthesized with different amounts
of sulfur impurities (~25 and ~230ppm). Upon annealing, the microstructural evolution
can be monitored by measuring the following characteristics:
1) Heat release upon annealing
2) Grain boundary migration
3) Changes in average grain size (AGS) and average largest grain size (ALGS)
4) Changes in grain size distribution (GSD)
5) Changes in grain boundary character distribution (GBCD)
An ideal characterization technique would allow the monitoring of all these
events in one experiment. But such a technique does not currently exist. As a result, a
73
variety of characterization techniques and combinations of different annealing methods
were employed in this study.
An important technique is in situ transmission electron microscopy (TEM)
because it allows for continuous monitoring of several characteristics (2 to 4) as a
function of time and temperature. However, problems can arise when samples are
thinned to electron transparency. As the samples become thinner (e.g: <100nm), grain
boundaries can intersect the free surface and an additional dragging force is introduced
when thermal grooves anchor the grain boundaries. This surface effect impedes grain
boundary movement, resulting in an overestimated thermal stability of the material
[Greenough and King (1957) and Mullins (1958)]. A good example of surface pinning
effects during in-situ annealing of nanocrystalline Ni was recently described by Hibbard
et al. [Hibbard et al. (2008)].
On the other hand, differential scanning calorimetry (DSC) is a great alternative
to continuously monitor the microstructure evolution of nanomaterials during annealing.
DSC is capable of heating samples from sub-ambient to 650C to continuously monitor
microstructural changes through measuring the heat release from the sample [e.g. Chen
and Spaepen (1990), Klement et al. (1995) and Hohne (2003)]. Heat release is associated
with the large amount of stored enthalpy in the interfaces and the reduction of the
interfacial volume fraction during grain growth. DSC is a widely accepted method in
studying grain growth of nanocrystalline materials [e.g. Tschope et al. (1992), Klement et
al. (1995), Hibbard (2002), and Zhou (2006)]. Unlike in-situ TEM studies, DSC only
74
provides heat flow as an indicator for any microstructural change during annealing. In
order to observe the microstructure associated with heat flow events, samples are usually
first annealed to certain temperatures in the DSC at predetermined heating rates.
Subsequently, samples are prepared for ex-situ TEM microstructural analysis. In order to
capture the important growth events during annealing, DSC annealing followed by ex-
situ TEM was used extensively in this study.
A TA instrument DSC Q1000 was used for DSC annealing. A high purity indium
standard was used for temperature calibration, while high purity sapphire was then used
for heat enthalpy and baseline calibration. Each sample was first cut into pieces of
4x4mm, and a weight of ~20mg was used for each DSC run. Aluminum sample holders
were used and all DSC runs were conducted in nitrogen-purged environment. Several
trial measurements were conducted to determine both the heating rate (b) and the
temperature range so that all thermal events were captured. For the total heat release
measurements, samples were quenched and subsequently heated again for a second cycle
to obtain a baseline for enthalpy measurement. Using the modified Kissinger analysis
[Kissinger (1956)], proposed by Chen and Spaepen [Chen and Spaepen (1990)], DSC at
multiple heating rates of 5-40C/min were used from -40C to 400C to obtain the
activation energies for grain growth. The activation energy (Q) is obtained by measuring
the shift in peak temperature (Tp) caused by changes in the heating rate (b). Equation 4.1
is the modified Kissinger analysis that describes the relation between heating rate (b),
peak temperature (Tp) and activation energy (Q):
75
CkT
Q
T
b
pp
ln (eqn 4.1)
where k and C are the Boltzmann and translation constant, respectively. By plotting
pp kTT
b 1 vs.ln , the negative slope then corresponds to the activation energy for grain
growth (Q). Chapter 5 will present all the results from the DSC annealing.
Chapters 6 and 7 will present the thermal stability of nanocrystalline copper under
iso-thermal annealing, and respectively at elevated and room temperature. To anneal iso-
thermally at elevated temperature, a 55%KNO3-45%NaNO2 salt bath was used to anneal
copper samples at temperatures ranging from 100C to 300C. Salt bath annealing has
the advantage of high heating rates with minimum oxidation compared to annealing in
conventional furnaces. All annealed samples were sealed in aluminum bags during
annealing and subsequently water quenched to room temperature. Isothermal salt-bath
annealing followed by TEM or electron backscattered diffraction (EBSD) was also
carried out to investigate the microstructural transformations as a function of time at
specific annealing temperatures.
Lastly, for annealed samples with sufficiently large grain sizes, EBSD was also
conducted in order to understand the change in grain boundary character and texture of
the material upon annealing. An overview on the importance and significance of grain
boundary character and texture will be presented in Chapter 8 which will also present the
EBSD results obtained from isothermal annealing of nanocrystalline (nc) low-S, and pure
polycrystalline (PPC) copper. For EBSD sample preparation, samples were initially
76
punched out into 3mm discs and mounted onto an aluminum specimen holder. Samples
were then prepared by mechanical polishing beginning with 1200 silica grid, and
subsequently further polished using 6 and 1 m diamond pastes. Fine 50nm colloidal
silica was used as final polish in order to remove the surface deformation layer. A low
deformation surface layer is critical in obtaining strong backscattered diffraction patterns
for orientation imaging microscopy.
77
4.2 – References for Chapter 4
Chen L.C. and Spaepen F., J. Appl. Phys., 69 (1990) 679
Greenough A.P. and King R., J. Inst. Metal., 79 (1951) 415
Hohne G., Differential Scanning Calorimetry, 2nd edition. Springer Publishing, London (2003)
Hibbard G., Radmilovic V., Aust K.T. and Erb U., Mater. Sci. Eng. A., 1-2 (2008) 232
Hibbard G., Ph.D. Thesis, University of Toronto (2002)
Kissinger H.E., J. Res. Nat. Bur. Stand., 57 (1956) 217
Klement U., Erb U., El-Sherik A.M. and Aust K.T., Mater. Sci & Eng., A203 (1995) 177
Mullins W.W., Acta Metal., 6 (1958) 414
Tschope A., Birringer R. and Gleiter H., J. Appl. Phy., A204 (1992) 5391
Zhou Y., Ph.D. Thesis, University of Toronto (2006)
78
Chapter 5 – Iso-kinetic Analysis of Nanocrystalline Low-S and High-S Copper
5.1 – Calorimetric Analysis
Both nanocrystalline low-S and high-S copper samples were made in order to
investigate the thermal stability and structural evolution differences upon annealing. All
samples were annealed from -40C to 400C at constant heating rates: 5, 10, 20, 30 and
40C/min. The nc low-S and nc high-S copper are the same samples as presented in table
2.7.1, which had average grain sizes of 49 and 31nm, respectively. Table 5.1.1 again
summarizes the average grain sizes and sulfur impurities used for this calorimetry study.
Fig. 5.1.1 shows typical examples of DSC heat release curves for low-S and high-
S copper annealed using a heating rate of 40C/min.
Table 5.1.1 – Grain size and sulfur levels for the nc low-S and nc high-S copper used for DSC experiments.
Grain size (nm) Standard deviation
(nm) Sulfur impurities (ppm)
nc high-S 31 21 23343
nc low-S 49 41 255
79
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
50 100 150 200 250 300 350 400
He
at F
low
(W
/g)
Temperature (°C)
low-S
high-S
T
T
T
Fig. 5.1.1 – DSC heat release curves for nc low-S and nc high-S copper
annealed to 400C using a heating rate of 40C/min.
For the high-S nanocrystalline material, the DSC curve consists of two major heat
releases, labelled as „T‟ in Fig. 5.1.1. The first exothermic peak starts at ~85C and peaks
at around 175C. It is then followed by another major heat release that peaks at around
~275C. Exothermic heat release then drops sharply as annealing progresses to 325C.
Many sulfur-containing nanocrystalline electrodeposits have shown similar heat
release curves as observed here for high-S copper. DSC curves showing two stages of
heat release were presented in the literature for sulfur-containing nanocrystalline Ni
[Klement et al. (1995), Wang et al. (1997), and Hibbard et al. (2001)], Ni-P [Isheim et al.
(2000), and Zhou (2006)], Ni-Co [Hibbard et al. (2006)], Co [Hibbard et al. (2001)], and
Co-P [Choi et al., (2005)]. However, it should be noted that the general shape of the DSC
80
curve for high-S copper is still considerably different from the curves presented in these
previous studies, which generally showed a much narrower second heat release.
On the other hand, the low-S copper, with a much lower sulfur impurity content
than the previously studied Ni, Ni-P, Ni-Co and Co-P materials, showed a significantly
different DSC curve. Here, well-defined exothermic events were not observed. Instead a
low heat exothermic reaction occurred over a broad temperature range (eg: 135-400C in
Fig. 3.1.1). Similar DSC curves have been shown elsewhere for inert gas condensed and
compacted nanocrystalline copper [Günther et al. (1992)].
Looking at high-S electrodeposits, the shape of the heat release profile is very
similar to DSC curves presented for copper electrodeposits from baths with thiourea
additives produced by Kao et al. [Kao et al. (2004)]. These results showed that the DSC
peak temperature and heat release increased with thiourea concentration.
Using the results obtained by DSC annealing of the copper samples at different
heating rates, the activation energy for grain growth (Q) was determined using the
modified Kissinger analysis [Chen and Spaepen (1990)] (eqn4.1), which uses the shift in
peak temperature (Tp) corresponding to different heating rates (b). Table 5.1.2
summarizes the peak temperatures and total heat releases for low-S and high-S copper
electrodeposits annealed at different DSC heating rates. For the high-S material the
major peak at the higher temperature was used for this analysis, see Fig. 5.1.1.
81
The high-S electrodeposits showed a higher total heat release compared to the
low-S copper. This can be attributed to the smaller initial grain size of the high-S
material and the possible formation of second phase particles due to the higher impurity
levels.
Table 5.1.2 – Summary of peak temperatures and activation energies for nc low-S and nc high-S samples.
Peak temperature Tp (C) Q (eV)
Heating rate, b 5C/min 10C/min 20C/min 30C/min 40C/min
Peak Temp: Low-S N/A 222 243 254 260 0.67
Heat release (J/g) N/A 3.12 3.06 3.27 2.38
Peak Temp: High-S 232 245 254 259 272 1.29
Heat release (J/g) 9.54 9.02 7.60 6.49 5.94
Fig. 5.1.2 shows the results of the modified Kissinger analysis. The activation
energies for the main heat release event for low-S and high-S copper deposits were
determined to be 0.67 and 1.29eV, respectively. For polycrystalline copper, the
activation energies for lattice, grain boundary and surface diffusion are 1.98-2.04eV,
1.06-1.1eV and 0.69eV, respectively [Kuper et al. (1954), Gust et al. (1985), Cousty et al.
(1981)]. The activation energy for self-diffusion in inert gas condensed nanocrystalline
copper was previously measured by different groups and ranged from 0.64-0.85eV
[Gleiter (1985), Günther et al. (1992) and Horvath et al. (1987)]. For high-S copper the
activation energy for the major heat release is close to the activation energy for grain
boundary diffusion in copper. On the other hand, for low-S copper the value is about
30% lower compared with the values for grain boundary diffusion. The activation energy
obtained for high-S indicates that sulfur has a significant influence in increasing the
82
activation energy for grain growth, and hence this may increase the thermal stability of
the electrodeposits. Increasing activation energy for grain growth with increasing sulfur
content has also been observed by Hibbard et al. for electrodeposited nanocrystalline
cobalt [Hibbard et al. (2006)]. Their results have shown that an increase in the sulfur
concentration from 240 to 980ppm resulted in an increase in the activation energy for
grain growth in Co from 1.44 to 2.3eV.
Low-S = 0.67eV
High-S = 1.29eV
-10
-9
-8
-7
-6
-5
-4
-3
-2
-1
0
20 21 22 23 24 25
In(b
/Tp)
1/kBTp
Low-S
High-S
Fig. 5.1.2 – Modified Kissinger analysis to determine the activation energies for grain growth (highest point of the heat release - Tp) in nanocrystalline low-S and high-S Cu.
The measured activation energy for high-S nanocrystalline copper relative to the
grain boundary diffusion is comparable with the activation energies of other
nanocrystalline electrodeposits reported in the literature. For example, in the case of
nanocrystalline nickel electrodeposits, the activation energies for the major DSC heat
release were reported to be in the range of 1.2-1.46eV [Wang (1997), Turi (1997),
Hibbard et al. (2000) and McCrea (2001)]. In electrodeposited nanocrystalline cobalt
83
deposits, the activation energy was reported in the range between 1.44-2.41eV [McCrea
(2001), and Hibbard et al. (2001)]. It should be noted that for both nanocrystalline Ni and
Co, the activation energies are shifted to higher values compared to nanocrystalline
copper because of their higher melting points.
In order to understand the changes in microstructural evolution due to sulfur
impurities, nanocrystalline copper with different sulfur concentrations (low-S and high-S)
were annealed at 20C/min in the DSC to specific temperatures and subsequently cooled
down to room temperature and prepared for TEM analysis. The characteristic
stages/regions of interest are labelled in Fig. 5.1.3. They correspond to either the start,
the end or the maximum of each exothermic reaction.
50 100 150 200 250 300 350 400
Re
lati
ve
He
at R
ele
as
e (W
/g)
Temperature (°C)
low-S
high-S
T (Tp)
T
T
T
T (Tp)
T
T
Fig. 5.1.3 – DSC heat release curves (heating rate: 40°C/min) with regions of interest for both nc
low-S and nc high-S copper. The symbol T indicates where the microstructure was observed ex-situ in the TEM. Tp is the peak temperature that was used in measuring the activation energy for
grain growth using the modified Kissinger analysis.
Region I
Region I Region II
84
5.2 – Microstructural Evolution for Nanocrystalline High-S Copper
Using a heating rate of 20C/min, regions of interest in high-S copper were
selected which correspond to the beginning of the first heat release (~85C), the first
major heat release (~175C), the second major exothermic peak (~275C) and the end of
the annealing treatment when heat release becomes negligible (400C).
Fig. 5.2.1 shows the TEM dark field images (left), and corresponding diffraction
patterns (right) of high-S in the as-deposited state (top), annealed to 85C (center) and
175C (bottom), respectively. Fig. 5.2.2 shows the grain size distributions of the
corresponding annealed microstructures in comparison with the as-deposited structure.
At 85C, negligible heat release was observed and the micrographs and grain size
distributions show that the microstructure is more or less the same as observed in the as-
deposited state.
As the annealing temperature exceeds 85C, the heat release increases steadily
and reaches the first plateau at around 175C (Fig. 5.1.3). TEM darkfield micrographs
(Fig. 5.2.1 bottom) and grain size distribution (Fig. 5.2.2) show that some grain growth
occurred as the temperature reached 175C. Looking at the grain size statistics (Table
5.2.1), the average grain size has grown from ~31nm to ~48nm. Most importantly, the
average largest grain diameter has increased by a factor of two, reaching ~257nm, as
compared to ~130nm as measured at 85C. DSC curves showing a first broad heat
release were previously related to the beginning of abnormal grain growth as observed in
85
several sulfur-containing Ni electrodeposits [Klement et al. (1995), Hibbard et al. (2001)],
and Co-P [Choi et al. (2005)]. However, for the high-S copper studied here, the size
difference between the matrix grains and the largest grains are much smaller compared to
the sulfur-containing Ni electrodeposits observed in the above studies.
Fig. 5.2.1 – TEM darkfield micrographs and diffraction patterns for nanocrystalline high-S copper:
as-deposited (top), annealed to 85C (center) and annealed to 175C (bottom).
86
0-25
50-7
5
100-
125
150-
175
200-
225
250-
275
300-
325
350-
375
400-
425
450-
475
0
10
20
30
40
50
Grain Size
High-S As-dep.
High-S Stage 1 (85°C)
High-S Stage 2 (175°C)
Coun
ts F
raction (
%)
Fig. 5.2.2 – Grain size distributions for nanocrystalline high-S copper of as-deposited material,
and material annealed to 85C and 175C, respectively, at 20/min.
Table 5.2.1 – Grain size statistics for high-S annealed to different temperatures.
Annealing
stages (C)
Average
grain size
(nm)
Standard
dev. (nm)
Average
largest grain
size (nm)
As-deposited 31 15 135
1 (85C) 33 28 130
2 (175C) 48 29 257
3 (275C) 272 120 1673
4 (400C)
718
(measured by
EBSD)
451 2515
Fig. 5.2.3 and 5.2.4 show TEM dark field micrographs, corresponding diffraction
patterns and grain size distributions for the high-S deposits annealed to 275C (top) and
400C (bottom), respectively. Upon annealing to the second major heat release (275C),
significant grain growth is observed. Therefore, this major heat release can be attributed
to rapid grain growth and, potentially, the formation of second phase particles. Rapid
87
grain growth results in the annihilation of many grain boundaries and, therefore, in a
considerable amount of heat release.
Fig. 5.2.3 – TEM darkfield micrographs and diffraction patterns for nanocrystalline high-S copper:
annealed to 275C (top) and 400C (bottom).
2µm
2µm
88
0-25
125-
150
250-
275
375-
400
500-
525
625-
650
750-
775
875-
900
1000
-102
5
1125
-115
0
1250
-127
5
1375
-140
0
1500
-152
5
1625
-165
0
1750
-177
5
1875
-190
0
2000
-202
5
2125
-215
0
2250
-227
5
0
5
10
15
Grain SizeHigh-S Stage 3 (275°C)
High-S Stage 4 (400°C)
Coun
ts F
raction (
%)
Fig. 5.2.4 – Grain size distributions for nanocrystalline high-S Cu annealed to 275C and
annealed to 400C.
Table 5.2.1 summarizes the grain size statistics for all high-S samples annealed to
different temperatures. Besides grain growth, grain size distributions show that the
growth mode consisted of combinations of normal and abnormal growth. The definitions
for normal and abnormal grain growth can be found in appendix 3 of this thesis. Based
on the given definition, the growth mode (normal vs. abnormal) can be quantitatively
evaluated based on the difference between the change in both average grain size (Dm)
and average largest grain size (Dmax). When the average largest grain size grows at a
rate that is significantly faster than the grain size of the matrix (Dmax>>Dm), abnormal
grain growth occurs. Table 5.2.2 tabulates the change in the average grain size (Dm)
and the change in the average largest grain diameter (Dmax) with respect to the previous
annealing stage. These results show that in stages 1 and 3, the change in average grain
size (Dm) and the change in average largest grain diameter (Dmax) increase at a very
89
similar rate, showing a normal growth mode. In contrast, annealing to 175C (stage 2)
shows that the average largest grain size (Dmax) grows at a rate that is twice as fast as for
the average grain size (Dm), indicated by the shaded row in table 5.2.2. The much
higher growth rate for the larger grains indicates that some moderate abnormal growth
has occurred during this stage of microstructural evolution.
Table 5.2.2 – Percent change in mean and right extreme grain size with respect to previous stage.
Annealing stages (C) Dm - Change in A.G.S.
w.r.t previous stage (%)
Dmax - Change in largest grain
dia. w.r.t. previous stage (%)
1 (85C) 7% negligible -4% negligible
2 (175C) 46% moderate 99% fast
3 (275C) 469% very fast 550% very fast
4 (400C) 164% fast 50% moderate
Rapid grain growth occurred when the samples were annealed from 175 to 275C.
Careful analysis of the diffraction patterns taken after annealing to 175 and 275C, as
presented in Fig. 5.2.5, shows that, in contrast to the 175C annealing, second phase
particles have formed as the annealing temperature reached 275C. However, these
particles are too small to be imaged in either bright or dark field imaging. It should be
noted that the samples became very difficult to electro-jet polish as they were annealed
beyond 275C. This is likely caused by the second phase particles which resulted in non-
uniform dissolution between the particles and the copper matrix during electrolytic jet
polishing.
90
Fig. 5.2.5 – Magnified view of selected area diffraction patterns for high-S samples annealed to
175C (left) and 275C (right). Samples annealed to 275C showed formation of second phase particles as indicated by the extra spots in circles.
To identify the sulfur containing second phase particles, the extra diffraction spots
(spots that did not correspond to the copper matrix) were analyzed to calculate their
interplanar spacing (d) values. The measured interplanar spacing values were then
compared to JCPDS (Joint Committee of Powder Diffraction Standards) interplanar
spacing values for different Cu/S compounds. Current indexing results have shown that
the majority of the extra-spots correspond to Digenite (Cu2S-Dg) and Covellite (CuS-Cv)
second phase particles.
It should be noted that indexing the extra diffraction spots from second phase
particles is not trivial because sulfur can form many metastable phases in this temperature
range (i.e: 275C and 400C). For example, annealing could form Cu2S (high
Chalcocite-ch) or CuS (Covellite – Cv), both with hexagonal crystal structure.
Moreover, the atoms can also rearrange and form another type of Cu2S second phase
(Digenite – Dg) with a face-centred cubic structure. Fig. 5.2.6 is the Cu-S phase diagram
91
showing the possible second phases that can be formed at this annealing temperature
range. It should be noted that other second phase particles such as low Chalcocite (ch,
another form of Cu2S), Djurleite (Dj – Cu~2S) and Anilite (An – Cu1.75S) were never
observed in the diffraction patterns. This could be due to the fact that these second
phases are only thermally stable at low temperature (<100C), as shown in Fig. 5.2.7,
which presents an enlarged view of the Cu-S phase diagram from 0-160C with atomic
percent sulfur ranging from 32 to 37 at.%.
Fig. 5.2.6 – Cu-S phase diagram showing different CuS and Cu2S second phase particles that could form at different temperatures [Chakrabarti et al. (1994)].
92
Fig. 5.2.7 – Enlarged view of the Cu-S equilibrium diagram [Chakrabarti et al. (1994)].
Finally, when the high-S electrodeposits were annealed to 400C, the
microstructure continued to grow and the grain size distribution became even wider as
shown in the TEM micrograph and grain size distribution presented in Fig. 5.2.3 and
5.2.4, respectively. Because of the difficulty in preparing TEM samples of annealed
material containing second phase particles, EBSD was used in order to obtain a good
statistics in grain size measurements for this annealing treatment. Fig. 5.2.8 is an
example of an EBSD map showing the microstructure of a sample that was annealed to
400C. Grain growth from 275C to 400C occurred in such a way that the change in
average grain size was much greater than the change in largest grain size (Dm>Dmax).
This indicates that the microstructure evolution is dominated by a normal type grain
growth. Tables 5.2.1 and 5.2.2 show that the average grain size increased by 164% from
272nm to 718nm, while the size of the largest grains only changed by 50%, from 1.7m
to 2.5m.
Tem
pera
ture
ºC
Atomic Percent Sulfur
Atomic Ratio, Cu/S
93
Fig. 5.2.8 – EBSD map showing microstructure of nc high-S copper DSC-annealed
to 400C at 20C/minute. Low angle boundaries (LABs) and random high angle boundaries
(HABs) are shown in yellow and black. Special boundaries (SBs) including 3, 9 and other
boundaries (≤29) are also displayed and labelled as red, green and blue lines, respectively.
From Table 5.2.2, one can see that abnormal grain growth sets in as early as in
stage 2 (175C) of annealing. The largest grains grow at a rate twice as fast as the grains
closer to the average grain size. Upon annealing to stage 3 (275C), the temperature was
high enough that it activated both the matrix and the abnormally grown grains. It is
interesting to see that despite a smaller grain curvature with larger grain diameter, larger
grains continued to grow at a similar rate compared to the matrix. As the largest grains
continue to grow, segregation continues as the migrating growth front keeps collecting
impurities and hence decrease the mobility of the migrating boundary. As annealing
progresses to stage 4, the matrix finally grows at a rate that is faster than the larger grains.
5.3– Microstructural Evolution for Nanocrystalline Low-S Copper
The DSC curves for low-S copper electrodeposits showed a significantly different
heat release profile compared to high-S copper, and many other sulfur-containing
1µm
94
electrodeposits [e.g. Klement et al. (1995), Isheim et al. (2000), Hibbard (2002), and Choi
et al. (2005)]. Here a very low broad single heat release is observed which begins at
around 110C and peaks at around 250C (Fig. 5.1.3 and Fig. 5.3.1).
Three specific temperatures were used here to observe how the microstructural
transformation is related to the heat release profile. These three selected temperatures are
110C, 250C and 400C, which again correspond to the beginning of the exothermic
reaction, the highest exothermic heat release and the end of the annealing where
exothermic reaction becomes negligible.
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0.016
0.018
0.02
50 100 150 200 250 300 350 400
Temperature (°C)
He
at
Flo
w (
W/g
)
Fig. 5.3.1 – DSC curve of nanocrystalline low-S copper DSC-annealed
to 400C at 20C/minute.
Fig. 5.3.2 shows TEM darkfield micrographs and corresponding diffraction
patterns for low-S nanocrystalline copper in the as-deposited state (top) and annealed to
110C (bottom). Fig. 5.3.3 shows the grain size distributions for both cases.
T
T
T
95
Table 5.3.1 shows the grain size statistics for low-S copper electrodeposits
annealed to different temperatures, while table 5.3.2 shows the relative change in both the
average grain size and the average largest grain size at different annealing temperatures.
The DSC heat profile showed only a small heat release as annealing progressed to
110C. However, grain size measurements in table 5.3.1 showed that there is already a
small change in the microstructure. The standard deviation increased from ~34 to ~55nm
and the grain size distribution also showed some broadening and a shift to the right.
Fig. 5.3.2 – TEM darkfield micrographs and corresponding diffraction patterns for nc low-S copper
in the as-deposited state (top) and DSC-annealed to 110C (bottom).
1 m
1 m
96
Fig. 5.3.4 and 5.3.5 show TEM darkfield micrographs, corresponding diffraction
patterns and grain size distributions for low-S copper annealed to 250C and 400C,
respectively.
Upon annealing to 250C, the average grain size increased by 227%, from 53nm
to 171nm, whereas the average largest grains grew at a significantly slower rate, only
46%, from ~752nm to ~1.1m (Tables 5.3.1 and 5.3.2).
It should be noted that no extra diffraction spots were observed in the diffraction
patterns of annealed low-S copper, regardless of the annealing temperature. However, this
does not mean that there are absolutely no second phase particles in these materials. If
present their concentration and size are likely too small to be detectable in electron diffraction.
0
75-1
00
175-
200
275-
300
375-
400
475-
500
575-
600
675-
700
775-
800
875-
900
975-
1000
1075
-110
0
1175
-120
0
1275
-130
0
0
5
10
15
20
25
30
35
Co
un
t F
ractio
n (
%)
Grain Size (nm)
LowS-Asdep
LowS-110°C
Fig. 5.3.3 – Grain size distributions for as-deposited low-S copper
and material annealed to 110C.
97
Table 5.3.1 – Grain size statistics for low-S copper: as-deposited and material annealed to different temperatures
Annealing
stages (C)
Average
grain size
(nm)
Standard
dev. (nm)
Average
largest grain
size (nm)
As-deposited 49 34 631
1 (110C) 53 54.5 752
2 (250C) 171 130 1100
3 (400C)
246
measured by
EBSD
220 1370
Table 5.3.2 – Percent change in average grain size and largest grain diameter w.r.t the previous stage
Annealing stages (C) Dm - Change in A.G.S.
w.r.t previous stage (%)
Dmax - Change in largest grain
dia. w.r.t. previous stage (%)
1 (110C) 8% very small 12% small
2 (250C) 227% very fast 46% moderate
3 (400C) 44% moderate 25% small
With further increase in DSC temperature, the exothermic heat release slowly
diminished as the temperature approaches 400C. Again, EBSD was used in order to
obtain micrographs at low enough magnification for better grain size measurements. Fig.
5.3.6 is an EBSD map showing the microstructure of low-S copper annealed to 400C.
At 400C, the average grain size increased by 44%, from 171nm to 246nm (Table 5.3.1
and 5.3.2). In contrast, the average largest grain size only increased by 25%, from 1.1m
to 1.37m (Table 5.3.1 and 5.3.2).
98
Fig. 5.3.4 – Darkfield micrographs (left), and diffraction patterns (right) for nc low-S copper
annealed to 250C and 400C.
0
75-1
00
175-
200
275-
300
375-
400
475-
500
575-
600
675-
700
775-
800
875-
900
975-
1000
1075
-110
0
1175
-120
0
1275
-130
0
1375
-140
0
1475
-150
0
1575
-160
0
0
2
4
6
8
10
12
Coun
t F
raction (
%)
Grain Size Range (nm)LowS-250°C
LowS-400°C
Fig. 5.3.5 – Grain size distributions of low-S materials annealed to 250C and 400C.
300nm
300nm
99
Fig. 5.3.6 – EBSD map for low-S copper annealed to 400C.
5.4 – Structural Evolution Differences Between Low-S and High-S Copper
By comparing the percent changes in the average grain sizes and the average
largest grain diameters (table 5.2.1 and table 5.3.1), it can be seen that abnormal grain
growth was more pronounced in the case for high-S copper electrodeposits. During
abnormal grain growth, the larger grains grow faster than the matrix grains. This means
the bigger grains experience a lower dragging force compared to the smaller grains.
In the case for low-S copper, the grain growth mode is more continuous in which
the grains with size near its mean value are growing during grain growth. This can be
explained by the reduced dragging force for the case of low-S Cu electrodeposits. In
low-S copper, the sulfur concentration is much lower than in high S copper (table 5.1.1).
The lower the impurity concentration the higher the mobility of the boundaries which
results in grain instability even at low temperature. Moreover, since the driving force is
100
inversely proportional to the grain size, grains that are small will have the tendency to
grow faster than the larger grains.
Fig. 5.4.1 and 5.4.2 show the cumulative volume fractions for low-S and high-S
copper annealed to different temperature stages. Direct comparison shows that high-S
copper initially has a higher thermal stability than low-S copper. Up to 175C, the
cumulative volume fraction curves for high-S copper show a slight shift to the right,
whereas for the case of low-S copper the cumulative volume fraction curve was already
shifted significantly by annealing up to 110C.
0
10
20
30
40
50
60
70
80
90
100
0 500 1000 1500 2000 2500 3000
Cu
mu
lati
ve
Vo
lum
e F
rac
tio
n (
%)
Grain Size (nm)
Low-S as-dep. Low-S 110C Low-S 250C Low-S 400C
Fig. 5.4.1 – Cumulative volume fraction curves
for nc low-S copper annealed to different temperatures.
101
0
10
20
30
40
50
60
70
80
90
100
0 500 1000 1500 2000 2500 3000
Cu
mu
lati
ve
Vo
lum
e F
rac
tio
n (
%)
Grain Size (nm)
High-S As-dep. High-S 85C High-S 175C High-S 275C High-S 400C
Fig. 5.4.2 – Cumulative volume fraction curves for nc high-S copper annealed to different temperatures.
As annealing progresses, microstructural transformation continues and low-S
copper eventually becomes more thermally stable. The average grain size for low-S and
high-S copper after annealing to 400C were measured to be 246 and 718nm, respectively.
The reasons for this could be impurity segregation and sulfur-induced formation of
second phase particles causing the dragging force to change from the initially stronger
solute drag to a weaker type Zener drag as will be discussed in the following section.
5.5 – Discussion
When looking at the change in the average grain size (Dm), and the average
largest grain size (Dmax) at different temperatures during DSC annealing in low-S
copper (table 5.3.2), it can be seen that the matrix grew at a rate that is faster than for the
102
average largest grain size (Dm>Dmax). On the other hand, for the case of high-S copper,
as shown in table 5.2.2, the average largest grain size grew at a faster rate compared to
the matrix grains during abnormal grain growth (stage 2). The differences in growth
mode observed can be explained if the dragging and driving forces for grain growth are
examined in more detail.
In low-S copper, the sulfur concentration is only about 10% compared to high-S
copper (Table 5.1.1). Also, looking at the diffraction patterns taken from the as-deposited
state (Fig. 5.2.1), a ring pattern is observed which corresponds to the electron diffraction
from different copper planes only. This indicates that the majority of the sulfur atoms is
likely distributed in a solid solution in the as-deposited copper matrix. Therefore low-S
copper has a much lower solute dragging force, Fs, compared to high-S copper as per
equation 5.1 [Haessener (1978)]:
Fs=[C0]GB/rs (equation 5.1)
where [C0] is the sulfur concentration in atomic percent, is the interfacial energy and r is
the atomic radius of sulfur.
On the other hand, if the segregated sulfur atoms form copper sulfide precipitates
(e.g. Cu2S) as shown in high-S copper DSC annealed to 275C (Fig. 5.2.5), a change
from solute drag to Zener drag would occur with a dragging force given by equation 5.2
[Smith (1948)]:
Fz=2f GB/R (equation 5.2)
103
where Fz is the Zener dragging force against grain growth, f is the particle volume
fraction, is the grain boundary energy, and R is the diameter of the precipitate. A
transition from strong solute drag to the much weaker Zener drag was also found in
nanocrystalline Ni-1.2wt.%P electrodeposits during grain growth at elevated
temperatures [Mehta et al. (1995)]
The driving force for grain growth, FD, is inversely proportional to grain size:
FD=/d (equation 5.3)
where is a shape factor, is the interfacial energy and d is the grain size. Assuming
constant dragging forces, equation 5.3 shows that smaller grains have a higher driving
force for grain growth. This is what was observed in table 5.3.2 for low-S copper where
the matrix grains (Dm) always grew at a rate faster than the average largest grains (Dmax).
On the basis of solute drag alone, one would expect a better thermal stability with
the addition of higher sulfur impurity levels. However, this was not observed as the final
grain size after DSC annealing to 400°C showed that low-S copper had a significantly
smaller grain size as compared to high-S copper deposits (low-S=246nm vs. high-
S=718nm). The reason for the change in thermal stability is likely associated with a
change in dragging forces. Sulfur impurities segregate to the grain boundaries and the
subsequent formation of second phase particles can lead to a change from the initially
strong solute drag to a weaker type Zener (particle) drag.
104
When high-S copper was annealed to 175C, table 5.2.2 showed that despite a
smaller driving force for grain growth that is associated with the larger grains, they still
grew at a significantly faster rate compared to the matrix. This is likely due to the local
variation in dragging force where boundaries associated with the larger grains are more
mobile, causing the larger grains to grow faster than the matrix (Dmax>>Dm). As
annealing progressed to 275C, the matrix experienced a higher driving force for grain
growth due to its much smaller grain size. In contrast, the already mobilized larger grains,
experienced a relative decrease in dragging force caused by segregation that ultimately
lead to the formation of second phase particles. These combinations ultimately lead the
entire microstructure to grow at a much greater rate and with a small growth rate
difference between the matrix and the average largest grains at 275C (Dm≈Dmax). As
annealing progressed to 400C, the matrix grew at a faster rate compared to the larger
grains (Dm>>Dmax). This is due to the fact that at high annealing temperature, the
driving force for grain growth exceeds the residual dragging force that hinders grain
growth.
5.6 – Summary of DSC Annealing Experiments
In summary, the general microstructural transformation for low-S copper
exhibited a continuous (normal) growth morphology. Microstructural analysis showed
that the grain growth evolution changes with higher sulfur concentrations. Abnormal
grain growth, or significant grain growth of the larger grains is observed in the samples
with high sulfur impurities within a certain temperature range. This abnormal grain
growth is observed when the larger grains (grain size>>AGS) grow at a much faster rate
105
than the matrix (stage 2 in high-S copper). The difference in transformation morphology
can be explained using a driving force and dragging force analysis.
In order to develop a better understanding between dragging mechanism induced
changes in the growth mode (continuous vs. abnormal) and thermal stability in general
with the addition of sulfur impurities, additional (salt bath) annealing experiments were
conducted by isothermally annealing both low-S and high-S copper at temperatures of
100C and 300C, respectively. Results of the isothermal annealing work will be
presented in Chapter 6.
106
5.7 - References for Chapter 5
Chakrabarti D.J., Laughlin D.E., and Sabramanian P.R., “Phase Diagrams of Binary
Copper Alloys”, 1st ed., ASM international, Ohio, USA, 1994
Chen L.C. and Spaepen F., J. Appl. Phys., 69 (1990) 679
Choi P., Da Silva M., Klement U., Al-Kassab T. and Kirchheim R., Acta. Mater., 53
(2005) 4473
Cousty J., Peix R. and Peraillon B., Surf. Sci., 107 (1981) 586
Gleiter H., Mater. Res. and Adv. Techniques., 75 (1984) 263
Günther B., Kumpmann A. and Kunze H.D., Scripta Mater., 27 (1992) 833
Gust W., Mayer S., Bogei A. and Predel B., J. de Physique, C4 (1985) 537
Haessner F. ed., “Recrystallization of Metallic Materials”, 2nd
ed., Hasselkus, Stuttgart,
Germany, 1978
Horvath J., Birringer R. and Gleiter H., Solid State Comm., 62 (1987) 319
Hibbard G., in Processing and Fabrication of Adv. Mater. XV, MS&T06, Cincinnati, OH,
(2006) 287
Hibbard G., Aust K.T. and Erb U., Mater. Sci. and Eng. A, 433 (2006) 195
Hibbard G., Ph.D. Thesis, University of Toronto (2002)
Hibbard G., Aust K.T., Palumbo G. and Erb U., Scripta Mater., 44 (2001) 513
Hibbard G., Erb U., Aust K.T. and Palumbo G., Mater. Res. Soc. Symp. Proc., 580 (2000)
183
Isheim D., Hentschel T., Kirchheim R., Muller F. and Kreye H., Acta Mater., 48 (2000)
933
Kao Y.L., Tu G.C., Huang C.A. and Chang J.H, Mater. Sci & Eng., A382 (2004) 104
Klement U., Erb U., El-Sherik A.M. and Aust K.T., Mater. Sci & Eng., A203 (1995) 177
Kuper A., Letaw H., Slifkin L. and E Sonder., Phys. Rev., 96 (1954) 1224
Lu K., Lu L., Wang L.B. and Ding B.Z., Mater. Sci & Eng., A286 (2000) 125
107
McCrea J., Ph.D. Thesis, University of Toronto (2001)
Mehta S.C., Smith D.A. and Erb U., Mater. Sci. and Eng., A204 (1995) 227
Smith C.S., Trans. Metal. Soc. A.I.M.E., 175 (1948) 15-24
Turi T., Ph.D. Thesis, Queen‟s University (1997)
Wang N., Wang Z., Aust K.T. and Erb U., Acta Mater., 45 (1997) 1655
Zhou Y., Ph.D. Thesis, University of Toronto (2006)
108
Chapter 6 – Isothermal Annealing Analysis
In this section, results obtained from isothermal annealing experiments will be
presented to develop a better understanding of the relationship between sulfur impurity
content and microstructural stability at different temperatures of low-S and high-S copper
nano-deposits.
Two annealing temperatures and times were used in these experiments. DSC
annealing and ex-situ TEM studies from the previous chapter have shown that grain size
instability was first observed at around 100C, while rapid grain growth occurred
between 250C and 300C. In addition the 300C temperature is of practical importance
for the potential application of nanocrystalline copper in PWB applications because many
typical lead-free solder reflow process as require PWB’s to withstand a peak reflow
temperature that’s above 260C for a few minutes. Consequently, isothermal annealing
experiments were conducted at 100C and 300C. The as-deposited low-S and high-S
nanomaterials described in table 5.1.1 were annealed for 5 and 30 minutes at each
temperature. Microstructural characterization was performed after each annealing
treatment. TEM was used for samples that were annealed at 100C, while
microstructures of the samples annealed at 300C were characterized using EBSD in
order to achieve better grain count statistics.
6.1 - Microstructural Evolution at 100C
Fig. 6.1.1 shows TEM micrographs of low-S (top) and high-S (bottom) copper, in
the as-deposited (left) states and their corresponding annealed microstructures after 5
109
(center) and 30 (right) minutes at 100C. Fig. 6.1.2 shows the corresponding grain size
distributions for the different annealing conditions. Table 6.1.1 shows the average grain
size, average largest grain size and standard deviations of low-S and high-S copper
isothermally annealed at 100C for 5 and 30 minutes, respectively.
Fig. 6.1.1 – TEM micrographs showing the as-deposited (left), and
isothermally annealed (100ºC) microstructures of nc low-S (top) and nc high-S (bottom) copper for 5 (center) and 30 (right) minutes.
In the case of low-S copper, the microstructure showed some grain growth with
the average grain size increasing from 49nm to 62nm after 5 minutes, and to 83nm after
30minutes of annealing. The grain size distributions also widen and shift to the right, as
compared to the as-deposited state. The grain statistics revealed that the standard
deviation increases as annealing time increases.
110
0-25
125-150
250-275
375-400
500-525
625-650
750-775
875-900
1000-1025
1125-1150
1250-1275
1375-1400
Low-S - As-dep.
Low-S - 5min
Low-S - 30min
High-S - As-dep.
High-S - 5min
High-S - 30min
0
10
20
30
40
50
60
70
Grain Size Range (nm)
Count fraction
(%)
Fig. 6.1.2 – Grain size distributions for nc low-S and nc high-S copper annealed at 100C for 5 and 30 minutes.
Table 6.1.1 – Results for average grain size, average largest grain size and standard deviations of nc low-S and nc high-S copper deposits in the as-deposited stage and
after annealing at 100ºC for 5 and 30 minutes.
Sample Annealing
condition
(C-min)
Average
grain size
(nm)
Standard
deviation
(nm)
Average
largest grain
size (nm)
Standard
deviation
(nm)
nc low-S Cu As-dep. 49 34 631 435
100-5 62 39 876 477
100-30 83 62 1066 205
nc high-S Cu As-dep. 31 15 135 8
100-5 34 13 133 18
100-30 34 15 124 22
111
The micrographs of the annealed structures for high-S copper show that the
average grain size and its standard deviation are more or less unchanged through 30
minutes of annealing at 100ºC (Fig. 6.1.2 and Table 6.1.1). This confirms that high-S
copper is more thermally stable than low-S copper at this low annealing temperature.
Moreover, for both low-S and high-S copper, isothermal annealing showed negligible
increase in the average largest grain size (Table 6.1.1). This confirms that at 100ºC, only
the smaller grains (i.e: Dm) will exhibit grain growth.
By comparing the microstructural evolution between the low-S and high-S copper
annealed at 100ºC, the experimental results show that the low-S copper sample grows
toward what appears to be the onset of a bi-modal distribution as annealing progresses.
This can be seen by plotting the percent count fraction as a function of grain size on the
log scale (log D), as shown in Fig. 6.1.3 and 6.1.4, respectively, for low-S and high-S
copper. In low-S copper, it can be seen that a minor second peak, highlighted by the
broken circle, begins to emerge after 5 minutes annealing which grows slightly as the
sample is annealed to 30 minutes. For the high-S copper (Fig. 6.1.4), however, the grain
size distributions still remain unimodal throughout the annealing treatments.
112
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
0
10
20
30
40
50
Log D
Co
un
t F
ractio
n %
Low-S As-dep.
100ºC-5min
100ºC-30min
Isothermal Annealing of Low-S at 100C
Fig. 6.1.3 – Grain size distributions of nc low-S copper annealed at 100ºC.
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
0
10
20
30
40
50
Log D
Co
un
t F
ractio
n %
High-S As-dep.
100ºC-5min
100ºC-30min
Isothermal Annealing of High-S at 100C
Fig. 6.1.4 – Grain size distributions of nc high-S copper annealed at 100ºC.
6.2 – Microstructural Evolution at 300C
Fig. 6.2.1 shows the EBSD maps for low-S (top) and high-S (bottom) copper
annealed for 5 (left) and 30 (right) minutes at 300C, respectively. The black lines in the
EBSD maps indicate random high-angle boundaries (>15 misorientation between
adjacent grains). The red, green, blue and yellow lines indicate 3, 9 and 27 and low
113
angle boundaries (<15), respectively. Brandon’s criterion (=15Σ-1/2
) was used here
for Σ value determination. More details on EBSD map analysis will be presented in
chapter 8.
Table 6.2.1 shows the corresponding average grain size, average largest grain
diameters and standard deviations for low-S and high-S deposits isothermally annealed at
300C. Fig. 6.2.2 shows the grain size distributions of low-S and high-S copper deposits
corresponding to the annealing conditions as given in table 6.1.1.
With respect to the annealed structures for low-S copper after 5 minutes, the
average grain size changed from an initial 49nm to 524nm, while the average largest
grain size increased from 1012 to 1415nm. As annealing continues to 30 minutes, the
material experiences negligible change in both average grain size and average largest
grain diameter.
In the case of high-S copper, the microstructures obtained from EBSD show grain
growth at a much more pronounced rate compared with low-S copper. The high-S
copper annealed at 300C for 5 minutes shows rapid grain growth from an initial 31nm to
1.25m. Grain growth continues as annealing time increases to 30 minutes when the
average grain size reached ~1.56m.
114
Fig. 6.2.1 – EBSD maps showing the microstructures for nc low-S (top) and nc high-S (bottom)
copper electrodeposits annealed at 300C for 5 (left) and 30 (right) minutes.
Table 6.2.1 – Results for average grain size, average largest grain size and standard deviation of nc low-S and nc high-S copper deposits in the
as-deposited state and annealed at 300ºC for 5 and 30 minutes.
Sample Annealing
condition
(C-min)
Average
grain size
(nm)
Standard
deviation
(nm)
Average
largest
grain size
(nm)
Standard
deviation
(nm)
nc low-S As-dep. 49 34 1012 435
300-5 524 281 1415 149
300-30 491 286 1314 584
nc high-S As-dep. 31 15 135 8
300-5 1250 545 2740 202
300-30 1570 707 3090 152
1 m 1 m
2 m 2 m
115
Fig. 6.2.2 – Grain size distributions of nc low-S and nc high-S copper annealed at 300C
for different times.
After annealing for 5 minutes, the microstructure of the low-S sample grows
toward a weak bimodal distribution. Fig. 6.2.3 represents a grain size distribution plot
with the grain size range plotted on the log scale. As circled on this figure, a small
secondary peak appeared as a result of annealing up to 30 minutes. In contrast, annealed
high-S copper (Fig. 6.2.4) shows a unimodal grain size distribution with a much more
uniform grain structure (Fig. 6.2.1).
116
0
10
20
30
40
50
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Log D
Low-S As-dep.
300ºC-5min
300ºC-30min
Isothermal Annealing of Low-S at 300C
Count
Fra
ction %
Fig. 6.2.3 – Grain size distributions of nc low-S copper annealed at 300ºC.
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
0
10
20
30
40
50
Log D
High-S As-dep.
300ºC-5min
300ºC-30min
Isothermal Annealing of High-S at 300C
Co
un
t F
ractio
n %
Fig. 6.2.4 – Grain size distributions of nc high-S copper annealed at 300ºC.
117
6.3 - Microstructural Evolution Differences Between Low-S and High-S
Copper
With isothermal annealing, high-S copper deposits were also found to be more
thermally stable than low-S copper at low temperature (100C). However, the superiority
in terms of thermal stability switched to low-S copper as annealing takes place at a higher
temperature (300C). As annealing temperature increases, the change in better thermal
stability from high-S to low-S can be explained by the dragging force mechanism.
In order to examine the solute dragging force in more detail, a better knowledge
of sulfur solute concentration at grain boundaries is needed. With known bulk sulfur
concentrations (section 4.6) and the intercrystalline volume fraction equations proposed
by Palumbo et al. [Palumbo et al. (1990), see appendix 2], we can calculate the maximum
possible grain boundary solute concentration for a given grain size by assuming that all
sulfur impurities reside in the grain boundaries as proposed in the solute dilution modal
by Palumbo and Erb [Palumbo and Erb (1999)].
The intercrystalline volume fraction equation is based on a 14-sided
tetrakaidecahedron as the grain shape (appendix 2). The outer skin of the
tetrakaidecahedron is the intercrystalline region with a thickness of /2, where is the
outer skin (grain boundary) thickness. The grain size (d) is the diameter of the
tetrakaidecahedron fitted into an inscribed sphere. Using this tetrakaidecahedron model,
the total intercrystalline (i.e: grain boundary) volume fraction as a function of grain size
(d) can now be expressed as:
118
Vtic = 1-[(d-)/d]
3. (eqn. 6.1)
Using a grain boundary thickness of 1nm, the above equation can be used to
estimate the intercrystalline volume fraction of materials (Vtic) as a function of grain size
(d). Hence, the relationship between maximum solute concentration [Sgb,max] along the
interface as a function of grain size can be derived by using the bulk sulfur concentration
[Sbulk] divided by the intercrystalline volume fraction for a given grain size [Vtic(d)]. This
relationship can be expressed as:
[Sgb,max] = [Sbulk]/Vtic(d). (eqn. 6.2)
The results are shown in Fig. 6.3.1, with two diagonal lines showing the
maximum sulfur impurity levels at the grain boundary corresponding to low-S (30ppm)
and high-S (250ppm) deposits, respectively.
Assuming the upper limit is where all the sulfur impurities reside in the grain
boundaries, Fig. 6.3.1 shows that, in the as-deposited state, high-S (grain size: 31nm) will
have a close to 6 times higher sulfur concentration at the grain boundaries, compared
with the low-S deposits (grain size: 49nm): 0.28at.% vs. 0.05at.%.
119
0.01
0.1
1
10
100
10 100 1000
Ma
xim
um
Gra
in B
ou
nd
ary
So
lute
Co
nc
en
tra
tio
n
[Sg
b,m
ax]
(at.
%)
Average Grain Size (nm)
25ppm 232ppm
0.05
0.28
Low-S
High-S
31 49
8334
starting structure
annealed at 100C, 30min
Fig. 6.3.1 – Maximum sulfur concentration at grain boundaries as a function of grain size.
As grain size increases upon annealing, the maximum grain boundary solute
concentration increases as the intercrystalline volume fraction decreases. For example,
Fig. 6.3.1 also shows that for samples annealed at 100C for 30 minutes, the maximum
grain boundary solute concentration changes from the initial 0.05at.% to 0.08at.% and
0.28at.% to 0.3at.%, respectively for low-S and high-S copper. Since low-S have a much
lower sulfur concentration, isothermal annealing at 100C resulted in moderate grain
growth, whereas the high-S copper showed very little change in microstructure.
In the DSC annealing experiments, it was shown that second phase particles are
present when high-S copper was annealed to 275C (Fig. 5.2.5). As precipitates are
formed, the dragging force changed from the stronger type solute drag to a weaker type
Ma
xim
um
Gra
in B
ou
nd
ary
So
lute
Co
nce
ntr
ati
on
(a
t.%
)
120
second phase (Zener) drag, ultimately causing more rapid grain growth for high-S copper.
The formation of second phase particles which subsequently induced more rapid grain
growth was also described elsewhere for nanocrystalline Ni-P [Mehta et al. (1995),
Farber et al. (2000), and Zhou (2006)]. In the case of low-S copper, as annealing takes
place at high temperature (300C), due to its lower initial impurity level, impurity
concentrations are insufficient for the formation of second phase particles. As a result,
the solute dragging force is still in effect even at higher temperature.
The maximum sulfur concentration plot (Fig 6.3.1) can also be used to explain the
change in thermal stability induced by the change in dragging forces. Fig. 6.3.2 is a plot
similar to Fig. 6.3.1, with the addition of a horizontal line that estimates the minimum
concentration required for sulfur to form second phase particles (CuS and Cu2S) along
grain boundaries. This concentration, 21.3at.%, is the upper limit calculated based on the
minimum concentration required to form a sulfur monolayer along the closest packed
(111) copper plane. A detailed mathematical derivation can be found in Appendix 4,
attached at the end of this report. This type of order of magnitude calculation has been
used previously by Zhou to estimate the minimum phosphorus concentration required to
form precipitates in annealed electrodeposited nanocrystalline nickel phosphorus alloys
[Zhou (2006)]. His result showed that the calculated value agreed quite well with
experimental values [Farber et al. (2000), Abraham et al. (2000), and Warren et al.
(2000)]. Looking at Fig. 6.3.2, in the case for high-S copper, precipitates are likely to
form when the sulfur concentration reaches 21.3at.%, which corresponds to a threshold
grain size (dt high-S) of ~2.4m. In contrast, low-S copper needs to reach to a grain size
121
(dt low-S) of ~20m in order for the sulfur concentration to reach the 21.3at.% threshold
value. However, annealing low-S copper at 300C for 30 minutes only resulted in grain
growth to ~0.49m. This corresponds to a maximum boundary concentration of
~0.5at.%, which is far below the threshold value required for the formation of second
phase precipitates. In contrast, annealing high-S copper at 300C resulted in grain
growth to 1.57m after 30minutes of annealing. With this grain size, Fig. 6.3.2 shows
that the maximum grain boundary solute concentration reached is about 15at.%. This
value is close to the calculated value (21.3at.%) and indicates second phase particles are
much more likely to form and consequently induce a change from the initially strong
solute drag to the much weaker type second phase particle drag in high-S copper.
It should be noted that this type of calculations can only demonstrate general
trends expected for nc low-S and high-S copper. The concentrations will change when
other planes are considered. Also the calculations assume that the entire grain boundary
area is covered with sulfur impurities. In reality this is likely not the case and the sulfur
concentrations may vary considerably from location to location. Nevertheless, this
analysis shows that high-S copper is more likely to change the dragging forces from
strong solute drag to weaker Zener drag.
122
0.01
0.1
1
10
100
10 100 1000 10000 100000
Average Grain Size (nm)
25ppm
232ppm
0.05
0.28
21.3
Fig. 6.3.2 – Grain size estimation for the formation of sulfur containing second phase particles as copper deposits reached 21.3at.% along the grain boundaries.
Klement and Da Silva studied the thermal stability of Co-1.1at.% P and Co-
3.2at.%P and showed that sample with higher solute content is relatively less stable with
lower grain growth temperature [Klement and Da Silva (2007)]. Their results showed
sample with 3.2at.% of P, phosphorus segregation has occurred and concentration of up
to 25at.% was found in the as-deposited state. For this reason, the change from solute to
Zener drag will occur earlier upon annealing and consequently sample with higher P
content resulted in a lower overall thermal stability.
Soong also observed that a decrease in the sulfur impurity level can lead to an
increase in thermal stability of nanocrystalline nickel electrodeposits [Soong (2009)].
0.5
d low-S300-30=491nm
d high-S300-30=1.57m
15
dt low-S=20m dt high-S=2.4m
Ma
xim
um
Gra
in B
ou
nd
ary
So
lute
Co
nce
ntr
ati
on
(a
t.%
)
123
Soong synthesized low-S (<10ppm) electrodeposited nanocrystalline nickel. Iso-thermal
annealing was conducted on these low-S electrodeposits and the results were compared
with conventional sulfur containing (~850ppm) nanocrystalline nickel studied by Hibbard
[Hibbard (2002)]. Fig. 6.3.3 shows the grain size as a function of annealing time at
similar annealing temperatures. With the low-S nickel deposits, isothermal annealing
showed moderate grain growth as samples were annealed isothermally at 350C for 1
hour (3600s). In contrast, high sulfur-containing nanocrystalline nickel deposits showed
grain stability for the first 30 seconds followed by a much higher growth rate after 30
seconds of annealing at 320ºC. After about 30 seconds of annealing, the low-S nickel
deposits become much more thermally stable as annealing continues.
0
50
100
150
200
250
300
350
400
450
500
1 1000 1000000
Gra
in s
ize
(n
m)
Time (sec)
Hibbard 320 C nNi (~850ppm S)
Soong 350 C nNi (<10ppm S)
30 3600 39600
°
°
432000
Fig. 6.3.3 – Combined isothermal annealing data for nanocrystalline low (<10ppm) and high (~850ppm) sulfur nickel [Soong (2009), Hibbard (2002)].
124
6.4 – Chapter Summary
The results obtained from the isothermal annealing experiments have shown that
the thermal instability of nanocrystalline copper is associated with a change from strong
solute to the weaker Zener type drag. Therefore, for printed wiring board applications, it
is important to consider the amount of sulfur incorporated into the copper during
synthesis because a high concentration (i.e: 250ppm – high-S) would cause its thermal
stability to decrease at the elevated temperature during the solder reflow step (Fig. 1.1.3).
However, it should be noted that a moderate amount of sulfur in the system is necessary
in order to impede growth during electrodeposition, and to obtain nanocrystalline copper
with the desired mechanical and etching properties.
125
6.5 - References for Chapter 6
Abraham M., Thuvander M., Lane HM., Cerezo A. and Simth DW., Mater. Res. Soc.
Symp. Proc., 581 (2000) 517
Farber B., Cadel E., Menand A., Schmitz G. and Kirchheim R., Acta. Mater., 48 (2000)
789
Hibbard G.D., Ph.D. Thesis, University of Toronto, (2002)
Klement U., and M. Da Silva., Sino-Swedish Struc. Mater. Sym. 2007 Proc., (2007) 173
Mehta S.C., Smith D.A., and Erb U., Mater. Sci. and Eng., A204, (1995) 227-232
Palumbo G. and Erb U., MRS Bulletin, 24-11 (1999) 27
Palumbo G., Thorpe S.J. and Aust K.T., Scripta Mater., 24 (1990) 1347
Soong C., Master Thesis, University of Toronto, (2009)
Warren P.J., Thuvander M., Abraham L.H., Cerezo A. and Smith D.W., Mater. Sci. For.,
343 (2000) 701
Zhou Y., Ph.D. Thesis, University of Toronto, (2006)
126
Chapter 7 – Room Temperature Thermal Stability of Electrodeposited Nanocrystalline Copper
7.1 - Literature Review
There are many studies which have shown that nanocrystalline material can
exhibit grain growth at ambient temperature. Many of these studies used materials
produced by inert gas condensation (IGC) such as aluminum [Birringer (1989)], copper
[Günther et al. (1992), Schwitzgebel and Heim (1999), Gertsman and Birringer (1994)],
lead [Birringer (1989)], magnesium [Birringer (1989)], palladium [Birringer (1989),
Günther et al. (1992)], silver [Günther et al. (1992)] and tin [Birringer (1989)]. The
results of these studies have shown that microstructural stability is highly influenced by
melting temperature, impurity level and residual porosity of the materials. For example,
significant grain growth was observed at room temperature for materials with melting
temperatures less than 600C [Birringer (1989)]. In the case of copper, despite its much
higher melting temperature, room temperature grain growth was also observed for
nanocrystalline copper produced by IGC [Günther et al., (1992), Schwitzgebel and Heim
(1999), and Gertsman and Birringer (1994)]. A modified Kissinger analysis work
performed by Gunther et al. showed that the activation energy for grain growth in IGC
copper was enhanced by introducing impurities such as oxygen before powder
compaction [Günther et al., (1992)]. The work performed by Gertsman et al. on IGC
copper showed that the room temperature thermal stability can be enhanced by increasing
the residual porosity of the starting nanostructure, as was shown in Fig. 3.3.5 [Gertsman
and Birringer (1994)].
127
Room temperature grain growth was also found in nanomaterials produced by
other synthesis techniques such as electrodeposition. For copper electrodeposits,
Brongersma et al. synthesized ultra-fine grained (actual grain size was not given here)
copper foil with different thicknesses and impurity contents [Brongersma et al. (1999)].
They observed room temperature microstructure instability and growth kinetics which
decreased with increasing foil thickness and impurity contents. Similar observations
were also reported elsewhere which indicated that fine-grained electrodeposited copper is
susceptible to grain growth at room temperature [Pantelon and Somers (2006a) and
Vas’ko et al. (2004)]. There are also a few cases where nanocrystalline copper deposits
were found to be thermally stable. These cases were attributed to very thin specimen
thickness (<2m) [Pantelon and Somers (2006a)] and/or high residual porosity (26%)
[Okuda et al. (2001)]. By reducing the specimen thickness, the thermal stability of the
material is enhanced by providing an additional dragging force that is generated by grain
boundary surface pinning [Greenough et al. (1951), Mullins (1958) and Pantelon and
Somers (2006b)]. The self annealing of electrodeposited nanocrystalline copper was also
observed by Militzer et al. and their results were able to show that the self-annealing rates
increase with decreasing initial grain size. Most importantly, they also suggested that the
growth kinetic can be significantly influenced by the interaction/amount of plating
additives incorporated into the deposits [Militzer et al. (2004)].
7.2 – Experimental Results
This present work presents the room temperature thermal stability results for fully
dense, electrodeposited nanocrystalline copper foils with different sulfur impurities.
128
Here, the low-S and high-S copper specimens were stored at room temperature
(23C) in a dry desiccator. The microstructures of the deposits were characterized by
transmission electron microscopy (TEM) after different storage times (up to two years).
To avoid grain boundary pinning due to surface effects [Hibbard et al., (2008)], samples
were only thinned to electron transparency just before each TEM characterization.
Fig. 7.2.1 shows representative TEM darkfield micrographs for low-S (top) and
high-S (bottom) in the as-deposited state, and stored at room temperature for 6 months
and 2 years, respectively, and Fig. 7.1.2 shows their grain size distributions. Table 7.2.1
provides a summary of the average grain size and standard deviation of both materials.
Fig. 7.2.1 – Dark field TEM micrographs showing microstructures of nanocrystalline low-S (top)
and high-S copper (bottom): as-deposited (left), and stored at room temperature for 6 months (centre) and 2 years (right).
129
0-10
20-3
0
40-5
0
60-7
0
80-9
0
100-
110
120-
130
140-
150
160-
170
180-
190
200-
210
220-
230
240-
250
Low-S as-dep (49nm)
Low-S 1day (49nm)
Low-S 6mths (51nm)
Low-S 1yr (55nm)
Low-S 2yrs (61nm)
0
5
10
15
20
25
30
Grain Size Range
Perc
en
t F
req
uen
cy
0-10
20-3
0
40-5
0
60-7
0
80-9
0
100-
110
120-
130
140-
150
160-
170
180-
190
High-S as-dep (31nm)
High-S 1day (30nm)
High-S 6mths (31nm)
High-S 2yrs (30nm)
0
5
10
15
20
25
30
35
40
Grain Size Range
Pe
rce
nt
Fre
qu
en
cy
Fig. 7.2.2 – Grain size distributions for nanocrystalline low-S (top) and high-S (bottom) copper of the as-deposited material, and material left at room temperature for up to 2 years.
130
Table 7.2.1 – Grain size analysis summary: average grain sizes and standard deviations.
Time Average grain size
(nm)
Standard deviation
(nm)
Low-S As-deposited 49 34
1 day 49 35
6 months 51 32
1 year 55 33
2 years 61 55
High-S As-deposited 31 15
1 day 30 17
6 months 31 16
2 years 30 14
7.3 – Discussion
After room temperature storage, the grain size distributions showed that the
amount of sulfur impurities incorporated into the deposits has a significant effect on the
thermal stability of nanocrystalline copper at room temperature. The high-S copper
deposits showed no change in the microstructure after 2 years at ambient temperature.
On the other hand, results for the low-S copper showed that some grain growth had
occurred after about 1 year. Selected area diffraction patterns taken for both materials
showed no indications of second phase particle formation after storing for up to 2 years.
The difference in thermal stability of the nanocrystalline copper with different
impurity concentrations can again be explained if one examines the driving and dragging
forces for grain growth. Assuming constant dragging forces, the driving force for grain
growth equation (eqn. 3.1) shows that deposits with a finer microstructure will exhibit a
131
higher driving force for grain growth. However, results obtained here have shown that
high-S deposits, with finer as-plated microstructure, have a better room temperature
thermal stability compared to the low-S copper electrodeposits. This indicates that the
microstructural stability found in high-S deposit is likely the result of the increase in
dragging force because of a much higher initial impurity content.
Pantleon and Somers further investigated the effect of thickness to the room
temperature grain growth of nanocrystalline copper with grain size of ~30nm [Pantleon
and Somers (2007)]. Their results showed that kinetics of room temperature grain growth
increase with the sample thickness. They also showed that the microstructure can be
stabilized when the deposit thickness is about 400nm.
Looking at the absence of extra diffraction spots in the selected area diffraction
pattern obtained from the low-S and high-S as-deposited materials (Fig. 2.5.2), it can be
assumed that all the sulfur impurities are distributed as a solid solution in the copper
matrix. Due to the sulfur level difference between low-S and high-S copper, it can
therefore be assumed that low-S has a much lower solute dragging force Fs, compared to
high-S copper.
Here, the dragging force equation (eqn. 3.4) assumes that all sulfur impurities are
distributed evenly across the deposits. However, it should be noted that there are several
references which showed that for as-electrodeposited materials, there is already some
132
accumulation of impurities along the grain boundaries [Hentschel et al. (2000) and Choi
et al. (2005)].
Following the solute dilution model discussed in section 6.3, a graph showing the
maximum grain boundary solute concentration as a function of average grain size is
shown in Fig. 7.3.1 for the case of the grain size changes observed at room temperature.
For the nanocrystalline low-S and high-S copper the maximum grain boundary solute
concentrations correspond to 0.04 and 0.28 at.% for the as-deposited state, again
assuming that all the sulfur impurities are segregated to grain boundaries. After two
years at room temperature the maximum sulfur concentration in the high-S copper is
unchanged, while it increased from 0.04 to 0.05at.% in low-S copper. Again it should be
noted that the solute dilation model can only give an order of magnitude estimate since
the actual segregation state in the material is unknown. Nevertheless it can account for
higher thermal stability of high-S copper compared to low-S copper on the basis of a
higher solute dragging force, even though the driving force for high-S copper is
significantly higher.
133
0.01
0.1
1
10
100
10 100 1000
Average Grain Size [d] (nm)
25ppm
232ppm
0.04
0.28
Low-S
High-S
31 49 61
[Sbulk]
Fig. 7.3.1 – Maximum sulfur concentration at grain boundaries as a function of grain size.
7.4 - Chapter Summary
This chapter has presented the effect of sulfur impurities on the room temperature
thermal stability of nanocrystalline copper produced by electrodeposition. The high-S
copper samples showed no change in the microstructure at ambient temperature for 2
years. On the other hand, the material with a much lower sulfur impurity content already
showed a small change in grain size after around 1 year. The experimental evidence
presented shows that the room temperature thermal stability of the nanocrystalline
material is greatly related to the initial solute dragging force.
Ma
xim
um
Gra
in B
ou
nd
ary
So
lute
Co
nc
en
trati
on
[S
gb
,max]
(at.
%)
134
For printed wiring board applications, the results obtained in this chapter showed
that copper with higher solute impurity concentrations provide better stability against
grain growth at room temperature. From all the annealing experiments obtained in this
study, it can be seen that a compromise between sulfur impurities and initial grain size is
required. Such a compromise would be increasing the initial grain size (e.g. to 50nm) to
reduce the driving force for grain growth and to adjust the sulfur level (e.g to 100ppm) to
provide an adequate solute dragging force to prevent grain growth during the solder
reflow processing step.
135
7.5- References for Chapter 7
Birringer R., Mater. Sci. Eng. A117 (1989) 33
Brongersma S.H., Richard E., Vervoort I., Bender H., Vandervorst W. and Lagrange S.,
J. of App. Phys., 86-7 (1999) 3642
Choi P., da Silva M. Klement U., Al-Kassab T., and Kirchheim R., Acta Mater., 53
(2005) 4473-4481
Gertsman V.Y. and Birringer R., Scripta Mater. 30 (1994) 577
Greenough A.P. and King R., J. Inst. Met. 79 (1951) 415
Günther B., Kumpmann A. and Kunze H.D., Scripta Mater., 27 (1992) 833
Haessner F. ed., “Recrystallization of Metallic Materials”, 2nd
ed., Hasselkus, Stuttgart,
Germany, (1978) 14
Hentschel T., Isheim D., Kirchheim R., Muller F., and Kreye H., Acta Mater., 48 (2000)
933-941
Hibbard G., Radmilovic V., Aust K.T. and Erb U., Mater. Sci. Eng., A1-2 (2008) 232
Militzer M., Freundlich P. and Bizzotto D., Mater. Sci. For., 467 (2004) 1339
Mullins W.W., Acta Metall., 6 (1958) 414
Okuda S., Kobiyama M., Inami T. and Takamura S., Scripta Mater. 44 (2001) 2009
Pantleon K. and Somers M.A.J., Mater. Sci For., 558-559 (2007) 1261
Pantleon K. and Somers M.A.J., Scripta Mater., 55 (2006a) 283
Pantleon K. and Somers M.A.J., J. App. Phys., 100-114319 (2006b) 1
Schwitzgebel G. and Heim U., Nanostr. Mater., 12 (1999) 19
Vas’ko V.A., Tabakovic I., Riemer S.C. and Kief M.T., Microelectronic Eng.,
75 (2004) 71
136
Chapter 8 – Grain Boundary Character Evolution in Electrodeposited Copper
The first part of this chapter (sections 8.1-8.8) deals with the structure and
properties of grain boundaries. To describe and quantify different types of grain
boundaries within a material, the term grain boundary character distribution will be
introduced and the general concept of grain boundary engineering will be discussed. The
second part of this chapter (sections 8.9-8.12) is concerned with the change in grain
boundary character distribution of low-S nanocrystalline and pure polycrystalline copper
produced in this research as samples were heat treated during iso-thermal annealing.
8.1 – Definition of Grain Boundary Character Distribution
The grain boundary character distribution (GBCD) is a material characteristic that
describes the interface content by sorting different types of grain boundaries within the
material according to structure types. Different types of grain boundaries can be
classified in many ways such as using mis-orientation (angular differences between two
grains) or coincident site lattice (CSL) (common atom position) relationships between
two adjacent crystals.
8.2- Types of Grain Boundaries
Grain boundaries are complex defects in a polycrystalline aggregate. In the early
1900s, it was believed that all grain boundaries have the same uniform structure and
properties. For example, in 1912 Rosenhain and Ewen published the amorphous cement
137
theory which stated that grain boundaries are like a quenched-in liquid with no long-
range order [Rosenhain and Ewen (1912)]. On the other hand, the unit cell model by
Hargreaves and Hill considered that grain boundaries formed between two crystals with
different crystallographic orientations should exhibit different magnitudes of disorder
within the structures [Hargreaves and Hill (1929)]. In 1950 Aust and Chalmers published
some critical data which showed that grain boundaries are not uniform in nature but of
various different structures [Aust and Chalmers (1950)]. Here, Aust and Chalmers used
synthetic tin bi-crystals to measure the grain boundary energy as a function of increasing
boundary misorientation. Fig. 8.2.1 is the grain boundary energy curve obtained in this
study which showed that the grain boundary energy changes with crystal misorientation,
in particular in the low angle (0-7°) region.
Fig. 8.2.1 – Grain boundary energy measurements obtained from tin bicrystals
[Aust and Chalmers (1950)].
In numerous other studies, several drops or cusps in the energy versus
misorientation curves have been observed [e.g: Chadwick and Smith (1976)]. For
example, Fig. 8.2.2 is a plot of the relative boundary energy versus misorientation angle
for <001> twist boundaries in copper [Miura et al. (1990)].
138
Fig. 8.2.2 – Measured grain boundary energy for [001] twist boundaries in Cu
[Miura et al., (1990)].
The two graphs shown in Fig. 8.2.1 and 8.2.2 illustrate that:
1. Not all boundaries should be treated the same in terms of properties, and
2. Special high angle boundaries at some critical misorientations exhibit much lower
boundary energies than general or random high angle boundaries.
Extensive research has focused on explaining the significant drop in grain boundary
energy at these critical misorientations [e.g Chadwick and Smith (1976), Wolf and Yip
(1992)]. Hermann et al., used the so-called sphere plate technique to detect these
misorientations in several metals [Hermann et al. (1976)]. In this technique a set of
randomly oriented single crystal spheres (~100µm diameter) are annealed on a single
crystal plate. A schematic diagram of this experimental technique for one sphere is
shown in Fig. 8.2.3. If the misorientation between the sphere and the plate is initially
139
given by 1 with grain boundary energy of E2 then the system can lower its grain
boundary energy to E1 by rotating the sphere to a new orientation 2. The driving force
for sphere rotation is then E=E2-E1. Experimentally, this sphere rotation can be
achieved by annealing the sphere plate arrangement at very high temperature to induce
the required mass transport for sphere re-orientation.
In order to detect all minima in the energy versus misorientation curve
simultaneously, the experiment can be modified such that a large number of single crystal
spheres (e.g: 106 /cm
2) is annealed on the same plate, as shown schematically in the left
of Fig. 8.2.4. Upon annealing all spheres rotate into misorientations that correspond to
the minima in the energy versus misorientation curve. Fig. 8.2.4-right shows the
accumulation of spheres in orientation space after annealing. The misorientations of low
energy can be detected by x-ray diffraction (e.g. pole figures) as indicated in Fig. 8.2.4.
Fig. 8.2.3 – Schematic diagram of the sphere plate technique showing the grain boundary and the rotation of the sphere on a single crystal plate with the corresponding interfacial energy vs.
misorientation curve [modified after Shewmon (1966)].
Grain boundary
140
Fig. 8.2.4 – XRD results of the sphere plate before (left) and after (right) annealing
[McCafferty (1995)].
The misorientations of the spheres after annealing, correspond to the grain
boundaries that have the lowest interfacial energies. These misorientations were found to
correspond to coincidence site lattice (CSL) boundaries, according to a model that was
first described by Kronberg and Wilson in 1949 [Kronberg and Wilson (1949)].
In this study, copper plates were heavily cold rolled and subsequently annealed at
400ºC to produce a microstructure with a strong cube texture. The textured copper sheets
were subsequently annealed to an even higher temperature to cause secondary
recrystallization which resulted in a new set of larger grains with different orientations.
Kronberg and Wilson were able to show that many of these grains were related by a grain
rotation of 21.8 about the [111] axis, as shown in Fig. 8.2.5.
141
Fig. 8.2.5 – Example of a CSL boundary obtained by a 21.8 rotation across a (111) plane [Reed-Hill and Abbaschian eds. (1992)].
In Fig. 8.2.5, net A represents the atomic arrangement of the (111) plane for the
bottom crystal (indicated by open circles). Net B represents the atomic arrangement of
the (111) plane in the top crystal rotated by 21.8° with respect to the bottom crystal (Net
A). The points where atomic positions of the two nets coincide in the grain boundary
between the two nets are shown as larger open circles. Kronberg and Wilson pointed out
that this type of boundary with a large fraction of coincidence sites would support rapid
grain growth.
Kronberg and Wilson further showed that the number of coincidence sites in Fig.
8.2.5 is equal to 1/7 of the number of atoms in either net A or B. They also used the
142
Greek symbol to describe the density of coincidence sites. As a result, 7 was used to
describe the type of boundary shown in Fig. 8.2.5.
Another significant advance in the understanding of CSL boundaries was
published by Ranganathan in 1966 [Ranganathan (1966)]. Ranganathan introduced
mathematical equations to describe the coincidence site lattices for interfaces between
two crystals. The following three equations can be used to describe any particular CSL
boundary:
)))(/(tan2( 2/11 Nxy (eqn 8.1)
Nyx 22 (eqn. 8.2)
222 lkhN . (eqn. 8.3)
where is the rotation angle about the rotational axis [hkl]. The Greek symbol Σ denotes
the reciprocal of the density of the coincidence sites and (x,y) are the coordinates of a
coincidence site in the plane (hkl).
Ranganathan pointed out that all grain boundaries can be classified as CSL
boundaries. However, many studies have shown that CSL boundaries with high Σ values
have no special structures and properties. Only boundaries with Σ≤29 are usually
considered special boundaries. For example, Palumbo and Aust have shown that grain
boundaries with Σ≤29 exhibit enhanced or beneficial properties such as corrosion and
creep resistance [Palumbo and Aust (1992)]. On the other hand, boundaries with Σ>29
can be considered as random because the density of coincidence sites is low (i.e:
143
Σ29=1/29=3.45%) resulting in less physical significance of coincident sites as compared
to low-Σ boundaries.
For all CSL boundaries, a small deviation from the exact angle of misorientation
between the two lattices is permitted. Brandon [Brandon (1966)], Dechamps et al.
[Dechamps et al. (1987)], Ishida et al. [Ishida and McLean (1972)] and Palumbo and
Aust [Palumbo and Aust (1990)], have all proposed slightly different criteria to define the
maximum allowed deviation for different CSL boundaries. All these criteria are based on
the dislocation model earlier described by Read and Shockley [Read and Shockley
(1950)]. Brandon’s criterion is the most widely used criterion to describe the maximum
deviation, , from CSL:
2/115 . (eqn 8.4)
In recent years, Palumbo and Aust introduced a more restrictive criterion,
6/515 , which is more consistent with experimental observations for the
correlation of low Σ boundaries and their susceptibility to integranular corrosion in high
purity nickel and several other grain boundary properties [Palumbo and Aust (1990)].
8.3 – Grain Boundary Character Distribution (GBCD) of a Random
Polycrystalline Aggregate
Low Σ (Σ≤29) CSL boundaries are only created when the two adjacent crystals
are oriented in some specific misorientation. Statistically, the probability of forming low
CSL boundaries within a material is relatively low. For instance, Warrington and Boon
144
have calculated that, in a material with random crystal orientation distribution, there is
only a ~11% chance for a boundary to occur as a low CSL boundary (1-25) [Warrington
and Boon (1975)]. This calculated value is generally accepted as the standard grain
boundary character distribution (GBCD) generated from a random polycrystalline
aggregate. Warrington and Boon also commented that in many metals, the frequency of
CSL occurrence can be higher than the 11% calculated value. Experimental evidence
also showed that the frequency of the CSL boundary occurrence is highly influenced by
several factors such as the materials synthesis technique, texture, stacking fault energy,
impurities and heat treatment [i.e: Liu (1982), Watanabe (1993), Lin (1997), Palumbo et
al. (1998), Shimada et al. (2002), and Randle (2004).]
8.4 – Properties of Low Boundaries
Because of the more ordered atomic arrangements between the two adjacent
crystals, the low-sigma boundaries have been shown to behave very differently compared
with random (Σ>29) high-angle grain boundaries in particular in terms of corrosion,
electrical and mechanical properties. The early bi-crystal studies have shown that low
boundaries have significantly lower boundary energy than random high-angle boundaries
[e.g: Chadwick and Smith (1976)]. In an effort to relate the grain boundary energy to
another property of a material, Arora and Metzger conducted a corrosion study using bi-
crystals with different [100] tilt boundaries in high purity aluminum [Arora and Metzger
(1966)]. Their results showed that the depth of the corrosion grooves formed along the
grain boundaries decreased significantly as the bi-crystals were tilted to ~23 and ~37,
which correspond to 13a and 5 grain boundaries, respectively.
145
Palumbo et al., studied the grain boundary corrosion of nickel in different aqueous
environments [Palumbo and Aust (1990)]. Their results showed that the severity of the
integranular corrosion is highly related to the grain boundary type. Fig. 8.4.1 is a SEM
micrograph of nickel polarized in 2N H2SO4 at 1200mV showing that the two random
boundaries (R) are heavily corroded while a Σ9 boundary remained unattacked.
Fig. 8.4.1 – SEM micrograph showing surface morphologies of attacked random and unattacked special (Σ9) grain boundaries in nickel [Palumbo and Aust (1990)].
Langer showed that CSL boundaries are very effective in preventing/disrupting
integranular crack propagation [Langer (1995)]. Fig. 8.4.2 is a schematic diagram of a
cracked region in alloy 600 after an intergranular stress corrosion test. Langer showed
that low sigma boundaries (1 and 3) are much less prone to integranular fracture than
random (R) boundaries. Here, the crack propagation is interrupted by these low sigma
boundaries. It should be noted that Σ1 boundaries are low angle boundaries with
misorientations angles less than 15°.
146
Fig. 8.4.2 – Cracked grain boundaries (solid black lines) in alloy 600 fracture area after intergranular stress corrosion test [Langer (1995)].
Results presented by Watanabe showed that creep induced integranular fracture
preferentially takes place along random boundaries [Watanabe (1993)]. He also showed
that boundaries that were close to low Σ CSL orientations are more resistant to creep
intergranular fracture than random boundaries.
Low Σ CSL boundaries also displayed a lower intrinsic electrical resistivity
compared to other random high angle boundaries. Nakamichi used bi-crystals to study
the electrical resistivity of different types of grain boundaries in pure aluminum and an
aluminum-silver alloy [Nakamichi (1990)]. His results showed that the low Σ CSL
related boundaries have a much lower grain boundary resistivity value compared with
other high Σ CSL high-angle boundaries with the same rotational angle. The grain
boundary resistivity of a coherent twin boundary (Σ3) was found to be an order of a
magnitude lower than the average grain boundary resistivity of general high angle
boundaries.
147
The mechanical properties of a metal are strongly influenced by the movement
and interaction of dislocations with grain boundaries. The degree of misorientation
between two grains can have a significant influence on the mechanical properties of the
material. For example, increasing the misorientation between two grains will increase the
ability of the boundary to impede dislocation movement since a dislocation passing into
the adjacent grain requires a change in its direction of motion. In addition, a fine-grained
material is harder and stronger than one that is coarse grained. Fine-grained materials
have a much higher frequency of grain boundaries, causing the material to be effective in
hindering dislocation movements.
The coherent twin boundary (Σ3) is a high angle boundary that provides a good
ability to impede dislocation movement and yet contributes little to electron scattering
(i.e: increase in electrical resistivity). While many strengthening mechanisms (e.g:
solution hardening, precipitating hardening, strain hardening) often cause a pronounced
increase in electrical resistivity, Lu et al. showed that pure copper with high densities of
twin (3) boundaries resulted in a tensile strength increase of a factor of 10, yet with
minimal increase in electrical resistivity, compared to conventional coarse grained copper
[Lu (2004)].
Many low boundaries have also shown a greater resistance to impurity
segregation and precipitation of second phase particles. For example, several studies
have shown that coherent twin boundaries (3) display a high resistance to boron
segregation in stainless steel [Karlsson and Norden (1988), Kurban et al. (2006)] and Ni-
148
Cr superalloys [Guo et al. (1999)]. Other low sigma boundaries (1 and 9) also showed
some resistance to impurity segregation and precipitation. Zhou et al., showed that low
sigma boundaries are less prone to carbide precipitation in stainless steel [Zhou et al.
(2001)]. Their results showed that boundaries provided the greatest resistance to
carbide formation compared to random high-angle boundaries.
8.5 - Grain Boundary Engineering (GBE)
The concept of grain boundary design and control was first introduced by
Watanabe in 1984 [Watanabe (1984)]. This idea involves the development of special
synthesis and processing techniques to control a material’s grain boundary structures and
characters in order to achieve improved overall properties. In other words, modifying the
grain boundary character distribution of conventional material towards higher fractions of
CSL boundaries eliminates less-ordered random boundaries with undesirable properties.
Thermomechanical processing (TMP) is one of the most common processing techniques
used in making GBE materials. As the name implies, this processing technique involves
a combination of several annealing (thermo) and deformation (mechanical) steps in order
to increase the proportion of low sigma boundaries (3 in particular) within the material.
There are mainly two types of thermomechanical processes, which involve a
deformation (pre-strain) stage followed either by a recrystallization or strain annealing
heat treatment. Palumbo has shown that a recrystallization heat treatment is preferred
over the strain annealing type heat treatment because of its relatively short annealing time
which helps the final microstructure to retain a relatively much finer grain size and near-
149
random texture [Palumbo (1994)]. This thermomechanical process is often repeated
several times because each subsequent deforming and annealing steps will progressively
increase the low sigma boundary fractions. However, Randle et al. showed conflicting
results that iterative thermomechanical treatment could be detrimental in promoting high
proportions of 3’s into the material [Randle et al. (2007)]. In general, conditions for the
thermomechanical treatments are usually determined empirically by several sets of
experiments, and are highly dependent upon the type of material.
Although there’s no unique answer as to whether one step or iterative annealing
would provide better optimization in grain boundary character, both methods have shown
that deformation followed by some sort of annealing treatment would yield an increase in
special boundary fractions. More specifically, it has been shown that low-sigma
boundaries (mainly 3s in materials with low stacking fault energy) are promoted
through the thermomechanical processing technique. Through increasing special
boundary fractions, GBE materials showed many property improvements such as a
significant enhancement in stress corrosion cracking. Fig. 8.5.1 shows an example of
conventional cast (right) and GBE (left) Pb-Sn battery grid material after identical
corrosion tests [Palumbo et al. (1999)]. While the conventional as-cast material has been
completely disintegrated by integranular degradation, the GBE processed material is still
fully intact, providing a much longer in-service lifetime for the GBE material.
150
Fig. 8.5.1 – GBE and conventional as-cast lead acid battery grid after corrosion tests
[Palumbo et al., (1999)].
The formation of the 3 boundaries is best explained by considering the growth
accident model proposed by Fullman and Fisher, who studied in detail the formation of
annealing twins along a triple junction [Fullman and Fisher (1951)]. In their grain
growth experiments, they observed that annealing twins are formed during grain growth
resulting from the overall decrease in the interfacial free energy of grain boundaries
through twinning. A schematic representation showing a twin boundary formation near a
triple junction is shown in Fig. 8.5.2.
Fig. 8.5.2 – Schematic representation of a coherent twin boundary generated in a grain corner
[Fullman and Fisher (1951)].
151
The decrease in total interfacial free energy by the formation of the twin boundary
in grain A can be mathematically written as:
CA'BA'CA'BA'AA'AA' aE+aE<aE+aE+aE ACABCA'BA' , (eqn. 8.5)
where E’s and a’s represent the interfacial energies and areas of the boundaries
indicated by the subscripts. The formation of the twin boundary AA’ would also lower
the interfacial energies of A’B and A’C boundaries. The decrease in the interfacial
energy of the grain boundary network would also improve corrosion (integranular) and
mechanical (creep) properties of the material.
Fullman and Fisher further explained the formation of non-coherent twin
boundaries. Fig. 8.5.3 is a schematic representation showing that initially a coherent twin
could form in the corner of a growing grain (Fig. 8.5.3a). The twin boundary grows to a
point that it is tangent to grain boundary A (Fig. 8.5.3b). A new junction point would
then form (Fig. 8.5.3c) and a non-coherent twin boundary is created as grain growth
progresses (Fig. 8.5.3d). Fullman and Fisher also discussed that a non-coherent boundary
could transform into a coherent twin boundary when a second twin boundary is formed
parallel to the first one and extended towards to the non-coherent twin boundary.
152
Fig. 8.5.3 – Formation of a non-coherent twin that forms along the side a coherent twin boundary [Fullman and Fisher (1951)].
8.6 – Factors Affecting the Formation of Annealing Twins
The grain growth model proposed by Fullman and Fisher showed that twin
boundaries could form simply by minimizing the energy of the system during annealing
(eqn. 8.5). In other words, the formation of twin boundaries is the result of decreased
interfacial energy of the microstructure that thermodynamically favours the formation of
these special Σ3 boundaries. Grain boundary engineering essentially makes use of this
driving force to manipulate the GBCD of a given material. It should be noted that there
are several factors that could affect the formation of twins in a given material. There are
many studies that have looked at these factors. The important parameters are listed
below and will be discussed in detail in the following sections:
1. stacking fault energy
2. solute concentration
3. pre-anneal deformation/strain
4. annealing condition (time and temperature)
153
8.6.1 - Stacking Fault Energy
The formation of annealing twins in an annealed structure is highly depended on
the stacking fault energy of the metal. Metals with low stacking fault energy usually
have much higher twin densities compared with higher stacking fault energy materials
[Fullman and Fisher (1951)]. The reduction in total grain boundary free energy is the
essential driving force for the formation of annealing twins. The coherent twin boundary
has the lowest interfacial energy, γΣ3, and the twin density increases with decreasing ratio
of twin boundary energy relative to the general random boundary energy, γgb:
Twin density as 3/gb
8.6.2 - Solute Concentration
The amount of impurity/solute in the material can also greatly affect the formation
of twin boundaries. Experimental results from Gallagher and Murr showed that in many
metals the increase in solute concentration decreases the stacking fault energy of the
material [Gallagher (1970) and Murr (1975)].
It is important to note, however, that the increase in solute concentration does not
always result in an increase in the frequency of twinning in the annealed structure. This
is because solute additions also change the grain boundary energy of the material. For
the case for equilibrium segregation, the Gibbs equation is obeyed where the excess
solute at the boundary, α, can be expressed as:
154
dx
dγx=α
gb
RT (eqn. 8.6)
From the above expression, it can be seen that the increase in solute concentration
(x) will cause a decrease in grain boundary energy (gb), ultimately causing an decrease in
twin density (i.e: 3/gb ). Bolling and Winegard showed that the twin density
decreased as silver was added to zone-refined lead alloy under equilibrium segregation
conditions [Bolling and Winegard (1958)]. Similar results were also observed by Lin
were silicon with higher impurities showed a lower fraction of special boundaries when
compared with higher purity grade silicon [Lin (1997)].
However, there are also cases where the increase in solute concentration causes an
increase in the overall twin density of the structure. This was explained in terms of non-
equilibrium segregation for which the increase in solute concentration causes an increase
in grain boundary energy (gb), leading to an increase in twin density (i.e: 3/gb ).
Here, non-equilibrium segregation causes the solute concentration in the boundary to be
much higher compared to the equilibrium segregation case. The additional solute atoms
migrate to the boundaries during grain growth and increase the grain boundary energy.
Simpson and Aust showed an increase in twin density when a small addition of tantalum
(~0.02 at.%) was added in zone-refined lead [Simpson and Aust (1970)].
8.6.3 - Pre-anneal Deformation/Strain
Several studies have shown that a moderate degree of pre-strain will yield higher
growth twin densities after annealing [e.g. Irving et al. (1964) and Silcock et al. (1966)].
155
This was attributed to the increase in dislocation density which provides sites for
nucleating annealing twins during recrystallization [Dash and Brown (1963) and Murr
and Meyers (1978)].
8.6.4 - Annealing Condition
In grain boundary engineering, the annealing temperature and time have been
shown to have a great influence on the grain boundary character of the annealed
microstructure. Baro and Gleiter measured the density of annealing twins as a function
of temperature for Cu-3wt.%Al [Baro and Gleiter (1972)]. Their results showed that the
density of annealing twins increases with temperature, up to about 600°C. Viswanathan
and Bauer observed a similar trend that the twin density of copper initially increased with
temperature to about 700°C and then decreased at higher temperatures [Viswanathan and
Bauer (1973)]. Simpson and Aust observed a similar behaviour in which the twin density
of pure lead near the melting point was much lower than at room temperature [Simpson
and Aust (1970)]. It was suggested that this is due to the fact that regardless of whether a
boundary is CSL or randomly oriented, they all converge towards a common grain
boundary energy value as the annealing temperature approaches the melting point of a
material. At very high temperatures, this would lead to a very low twin density because
the formation a twin boundary no longer reduces the overall boundary energy of the
system according to the theory proposed by Fullman and Fisher [Fullman and Fisher
(1951)].
156
Charnock and Nutting studied the effect of time on the twin density of pre-
deformed brass under isothermal annealing at 600°C, 750°C, 850°C and 930°C for
different times [Charnock and Nutting (1967)]. Their results showed that the twin
density initially increases linearly with time. However, as annealing progressed, the twin
density reached a maximum value and then decreased slowly.
Liu also showed that the twin density of Cu- 1at.%Sn and Cu- 1at.%Zn initially
increased with time, then reached a maximum value, followed by a decrease after longer
annealing times [Liu (1982)]. He also showed that the time period in which the increase
in twin density was the strongest was closely related to the period of relatively high grain
boundary migration rates, that is during the initial stage of rapid grain growth.
8.7 – CSL Frequency as a Function of Grain Size
Results obtained by Grabski and Watanabe on Al and Fe based alloys showed that the
frequency of special boundaries (%fSB) tend to increase with decreasing grain size for materials
produced by thermomechanical processing [Grabski (1985) and Watanabe (1993)]. Fig. 8.7.1
shows the relationship between the %fSB and grain size for various polycrystalline metals and
alloys produced by thermomechanical processing [Watanabe (1993)].
157
Fig. 8.7.1 – Relationship between the frequency of special boundaries and grain size in various polycrystalline metals and alloys produced by thermomechanical processing [Watanabe (1993)].
Lin studied the %fSB as a function of grain size on materials that are susceptible to
twinning (Fig. 8.7.2). His results also showed that the %fSB increases initially with
decreasing grain size from ASTM4 (~90µm) to ASTM9 (~16µm). However, %fSB
decreased with further grain size reduction [Lin (1997)].
The work conducted by Liu on annealing as-cast Cu-Sn alloys also showed a
similar trend where the twin density increases with grain size at the beginning of
annealing and then decreases as annealing continued (Fig. 8.7.3) [Liu (1982)].
158
Fig. 8.7.2 – Frequency of CSLs as a function of grain size for materials susceptible to twinning [Lin (1997)].
Fig. 8.7.3 – Variation of annealing twin density with grain size during isothermal annealing of Cu-Sn alloy at 700°C [Liu (1982)].
It should be noted that the starting materials used in all of the above studies were
polycrystalline materials with grain size >>1µm. In contrast, Palumbo and Aust,
conducted a series of %fSB analyses on isothermally annealed nanocrystalline Ni-15%Fe
159
(average starting grain size ~30nm) at 600°C from 5 to 900s. Their results also showed
(Fig. 8.7.4) that %fSB initially increases with increasing grain size but then decreased
after the grain size reached about 500nm.
Fig. 8.7.4 – CSL frequency as a function of average grain size for Ni-15wt.%Fe. Time in bracket indicate the annealing times [Palumbo and Aust. (1998)].
In a recent study, Kobayashi et al. studied the CSL frequency as a fraction of
grain size on ultrafine-grained nickel [Kobayashi et al. (2010)]. Their results showed that
%fSB decreases with increasing grain size upon annealing. However, their data also
showed that %fSB and grain size dependence presents only significant when the annealed
grain size is less than ~400nm. Fig. 8.7.5 is the results obtained by Kobayashi et al.
showing the relationship between the %fSB as a function of average grain size for
ultrafine-grained nickel.
160
Fig. 8.7.5 – %fSB as a function of average grain size for ultrafine-grained Ni [Kobayashi et al. (2010)].
8.8 – Experimental Details
The focus of this section is to understand the change in grain boundary character
distribution as a fuction of grain size of electrodeposited copper during isothermal
annealing. Here, low-S nanocrystalline and pure polycrystalline copper were selected for
this study. Table 8.8.1 provides a summary (i.e: grain size and ICP chemical analysis) of
the samples used in this EBSD study.
Table 8.8.1 – Grain size and chemical composition of samples used in this section
Sample Low-S nc Cu PPC Cu
Grain Size 49nm 3.58 µm
Sulfur content (ppm) 307 <10
Carbon content (ppm) 38043 214
Deposit thickness ~25µm
Low-S nanocrystalline Cu was used because of its small initial grain size, and low
impurity levels (relative to typical nanocrystalline electrodeposits which often contain
161
250-900ppm of S). With a much lower impurity level, the grain boundary character of
the annealed structure is expected to be improved because the lower impurities would
correspond to less segregation that would cause a decrease in the general grain boundary
energy for the random boundaries. Moreover, the use of nanocrystalline/fine-grained
materials as the starting microstructure is to provide high grain growth rate conditions for
manipulating boundaries and to retain a relatively fine grain structure upon annealing.
This could potentially provide a new processing technique to GBE by eliminating the
need of repetitive pre-straining and annealing.
For comparison, pure electrodeposited polycrystalline copper (PPC) samples were
also used for this part of the study. The pure polycrystalline sample was made to yield
the smallest grain size without using any bath additive. This was done by manipulating
various plating parameters such as average current density, duty cycle and pulse current
Ton and Toff. ICP chemical analysis showed that the deposits contained less than 10 and
~20 ppm of sulfur and carbon, respectively (Table 8.8.1).
These two types of materials should provide a better understanding on how grain
size, and annealing conditions would affect the grain boundary character evolution of
copper.
The two sets of samples were isothermally annealed in a salt bath at temperatures
and times ranging respectively from 200-500°C, and from 1 to 60 minutes. Table 8.8.2
gives an overview of the annealing conditions investigated in this section.
162
Table 8.8.2 - Annealing conditions of investigated samples.
Sample type
Low-S nCu PPC Cu
Annealing
condition
Abbreviated name
200°C-60min Low-S 200-60 PPC200-60
300°C-5min Low-S 300-5 PPC 300-5
300°C-30min Low-S 300-30 PPC 300-30
300°C-60min Low-S 300-60 PPC 300-60
400°C-5min Low-S 400-5 PPC 400-5
400°C-30min Low-S 400-30 PPC 400-30
500°C-1min Low-S 500-1 PPC 500-1
500°C-5min Low-S 500-5 PPC 500-5
The GBCDs of these materials for the different annealing conditions were
determined using the electron backscattered diffraction (EBSD) technique. For EBSD
sample preparation, annealed samples were first punched out into 3mm discs and then
polished mechanically using 6 and 1µm diamond. Ultra-fine (50nm) colloidal silica was
used as a final polishing step to remove most of the residual deformation during coarse
polishing to ensure strong Kikuchi patterns for orientation indexing. The typical
polishing times were about 10, 2 and 15 minutes, respectively for 6µm, 1µm and final
colloidal silica polish. The EBSD system was attached to a Hitachi S4500 field emission
SEM. Kikuchi pattern for EBSD analysis were collected at 70° tilt using 20kV
accelerating voltage and a probe current of ~2nA. The hkl EBSD system and the channel
5 EBSD software were used to collect and index all Kikuchi patterns. The Brandon
criterion (=15-1/2
) was used to characterize all grain boundaries [Brandon (1966)].
Each EBSD map typically collected 500 grains in order to ensure statistical significance
163
while minimizing beam instability and stage drift. The scanning step size ranged from
20nm to 500nm and was selected based on about 10% of the sample's average grain size.
For example, a 100nm step size was used for a grain size of ~1µm. The indexed
percentage ranged from 75%-97%. The captured Kikuchi patterns were rejected when
the mean angle deviation (MAD) value deviated by more than 1.5°. The MAD value is
the angle measurement built into the EBSD software to determine orientation index
misfit. It is used to check whether the Kikuchi pattern is indexed correctly in order to
provide reliable data to access the grain boundary character of the material. Experimental
evidence from hkl Technologies showed that using an acceptance MAD value of 1.5°
would provide both accurate pattern indexing and high indexing rate (low zero solutions)
for cubic structure material [EBSD reference guide from hkl Technologies].
8.9 – Results and Discussion
Fig. 8.9.1 shows a typical EBSD map of low-S nanocrystalline copper that has
been annealed at 300C for 5 minutes. This EBSD map is composed layer-by-layer using
band contrast (BC), grain boundaries (GB) and coincidence site lattice (CSL)
components. The BC component is used to generate a visual gray scale image by
assigning a gray level scale (0-255) based on Kikuchi diffraction pattern (KDP) quality.
That is, the higher the value, the better the KDP quality. As a result, disordered regions
(e.g. grain boundaries) would have a low BC value (i.e: black color), whereas perfect
crystalline regions (i.e: grains) would have a relatively high BC value (i.e: white color).
The GB component is used to differentiate between low and high angle grain boundaries
by measuring the misorientation between the neighbouring crystals. Low angle
164
boundaries (LAB’s) are defined as misorientations between two adjacent crystals of less
than 15. High angle boundaries (HAB’s) are indicated when the misorientations
between two crystals are higher than 15. The LABs and general HABs are labelled as
yellow and black lines, respectively, throughout this report. The low CSL boundaries
(Σ≤29) component defines and labels all the special boundaries (SBs) using Brandon’s
criteria (=15-1/2
) for the acceptable deviation from the exact misorientation. For
simplicity, all low CSL boundaries are organized into three categories: 3, 9, and other
low boundaries (29). These three categories are displayed as red, green and blue
lines, respectively.
Fig. 8.9.1 – EBSD map of low-S nc Cu annealed at 300C for 5 minutes. This map is constructed using band contrast (BC), grain boundaries (GB) and CSL components. Low angle boundaries
(LABs) and random high angle boundaries (HABs) are shown in yellow and black. Special
boundaries (SBs) including 3, 9, and other boundaries (29) are also displayed and labelled as red, green and blue lines, respectively.
165
Fig. 8.9.2 is a simplified EBSD map (with the BC component removed) that is
used to illustrate different grain boundary types. Here, the different colours correspond
to the different types of grain boundaries as mentioned above. The grain boundary
character distribution of the material can now be measured by extracting the line fractions
for each type of grain boundary and displayed in bar charts.
Fig. 8.9.2 – EBSD map showing GB and CSL components used for GBCD analysis.
Fig. 8.9.3 shows bar charts of the GBCD’s of low-S nanocrystalline copper,
annealed at different conditions as indicated in Table 8.9.2. Fig. 8.9.4 shows the GBCD’s
of PPC copper used in this study. Both figures also display the GBCD of a material with
random crystal orientations as given by Warrington and Boon [Warrington and Boon
(1975)]. Table 8.9.1 summarizes the GBCD’s of all different annealing conditions and
samples used in this study.
1µm
166
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Lin
e fra
cti
on
s o
f d
iffe
ren
t b
ou
nd
ari
es
(%
)
Annealing conditions and measured grain size (µm in bracket)
HABs
Other Sigma
Sigma 9
Sigma 3
LABs
(≤29)
(>29)
Fig. 8.9.3 – GBCD’s of low-S copper investigated in this study.
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Lin
e fra
cti
on
s o
f d
iffe
ren
t b
ou
nd
ari
es
(%
)
Annealing conditions and measured grain size (µm in bracket)
HABs
Other Sigma
Sigma 9
Sigma 3
LABs
(>29)
(≤29)
Fig. 8.9.4 – GBCD’s of polycrystalline copper used in this study.
167
Table 8.9.1 – Grain size, line fractions of special boundaries and line fractions of high angle boundaries of the annealed structures used in this study.
Sample Grain
size
(m)
fSBs 1-
29 (%)
fΣ3
(%)
HABs
(%)
Low-S 200-60 0.4 38.7 20.2 61.3
Low-S 300-5 0.52 39.6 23.6 60.4
Low-S 300-30 0.5 37.8 20.7 62.2
Low-S 300-60 0.63 36.0 21.3 64.0
Low-S 400-5 0.91 45.9 34.9 54.1
Low-S 400-30 1.62 47.6 31.5 52.4
Low-S 500-1 3.12 45.4 29.3 54.6
Low-S 500-5 5.12 49.1 28.6 50.9
PPC As-dep. 3.58 67.2 47.6 32.8
PPC 200-60 5.95 72.5 47.8 27.5
PPC 300-5 4.25 72.2 49.7 27.8
PPC 300-30 6.18 60.1 45.5 39.9
PPC 300-60 7.83 62.3 46.7 37.7
PPC 400-5 8.19 71.6 59.1 28.4
PPC 400-30 9.78 62.6 47.6 37.4
PPC 500-1 11.2 68.6 58.1 31.4
PPC 500-5 18.9 66.7 54.0 33.3
Looking at Fig. 8.9.3 and table 8.9.1, it can be seen that annealing low-S copper at
various conditions results in variety of different grain boundary character distributions.
For example, the total special boundaries (1-29) line fractions (%fSB) ranged from 36%
(300°C-60minutes) to 49.1% (500°C-5minutes). It should be noted that the grain
boundary character distribution of as-deposited low-S copper could not be determined by
the EBSD technique due to the small grain size (<50nm). Fig. 8.9.5 shows two EBSD
images of low-S copper for a sample with a low fraction of special boundaries (200C,
60mins) and another sample with relatively much higher fraction of special boundaries
(400C, 30mins).
168
Fig. 8.9.5 – EBSD images illustrating samples with a low (left) and a relatively higher (right)
fraction of CSL grain boundaries retained in the annealed microstructure. 3 boundaries are labelled in red and random boundaries are indicated in black lines. The yellow, green, and blue
lines respectively represent 1, 9 and 29.
In comparison, the GBCD results for annealed pure polycrystalline copper (PPC)
(Table 8.9.1), generally showed annealed structures with much higher final %fSB values.
Here, in the case for the PPC samples, different annealing conditions resulted in %fSB
values ranging from 60.1% (300°C-30minutes) to 74.6% (200°C-30minutes).
8.10 - Effect of Grain Size on GBCD
Fig. 8.10.1 shows the %fSB as a function of grain size for all the nc low-S and
PPC copper samples used in this study. For comparison, results obtained by Palumbo
and Aust on the change in %fSB value as a function of grain size upon annealing on
nanocrystalline Ni-Fe are also shown in this figure [Palumbo and Aust. 1998].
1m 1m
169
0
10
20
30
40
50
60
70
80
90
100
0.1 1 10 100
% f
SB
Average grain size (µm)
Low-S nc Cu PPC Cu Palumbo and Aust nc Ni-Fe
Fig. 8.10.1- Percentage of special boundaries (fSB) as a function of grain size for low-S nc copper, PPC copper and sulfur containing nc Ni-Fe obtained by Palumbo and Aust
[Palumbo and Aust (1998)].
In the case for low-S nc copper, it can be seen that the increase in grain size
would generally yield a higher final %fSB value. In fact, samples with higher %fSB value
were all annealed at higher temperatures (400°C and 500°C). One contributing factor
could be the rapid impurity segregation leading to the formation of very small second
phase particles at these temperatures even through they were not detected in electron
diffraction (see section 5.3). When second phase particles are formed, the energetic
differences between random and special boundaries re-emerge and promote the formation
of more special boundaries which ultimately leads to an increase in fSB values. In
contrast to the change in fSB value observed in nc Ni-Fe, Palumbo et al. observed a slight
decrease in %fSB value as grain size increases during annealing following the initial
minor increase [Palumbo et al. (1998)]. However, it should be noted that both low-S
170
copper and the nanocrystalline Ni-Fe in the study by Palumbo et al. contained sulfur
impurities and resulted in a similar range of %fSB value (i.e: 30-50%) upon annealing.
For PPC copper, results from Fig. 8.10.1 showed no strong influence of grain size
affect on the %fSB value. This is probably attributed to the low impurities in the starting
material and the less significant changes in grain boundary energetics due to segregation
effects.
From the literature review, it can be seen that there is no definite relationships in
relating frequency of special boundaries (fSB) to grain size. The results from the current study
indicate that the evolution of fSB is likely related to the impurity level of the starting structure,
and support the general idea that grain boundary character distributions in thin foils (as used
in the current study) could be different from distributions found in bulk materials [Watanabe
(1989)]. In the case of thin foils, surface energy anisotropy may have an influence on grain
growth and grown boundary character evolution [Wantanabe et al. 1989)].
8.11 – Texture Evolution
Fig. 8.11.1 shows typical {100} {110} and {111} pole figures for low-S copper
annealed to grain sizes of 400nm (200C-60min.), 500nm (300C-30min.), 1.62m
(400C-30min.) and 5.12m (500C-5min.), respectively. These pole figures all show
that the annealed materials maintain the same texture throughout annealing. In other
words, all the low-S electrodeposits annealed as per table 8.9.1 resulted in a random
texture, regardless of their final grain size and annealing conditions. It should be noted
171
that the texture for the as-plated low-S nc copper is not available as the grain size was too
small for EBSD analysis. However, from the -2 diffraction scan shown in Fig. 2.6.3, it
can be assumed that the as-plated material also had a random crystalline texture.
Fig. 8.11.1 – Selected {100} {110} and {111} pole figures obtained from EBSD measurements for
low-S copper annealed to grain sizes of: a)400nm, b) 500nm, c) 1.62m and d) 5.12m.
a)
b)
c)
d)
172
Fig. 8.11.2 shows the {100} {110} and {111} pole figures for PPC copper at the as-
deposited state, and annealed to grain sizes of 7.95µm (200C-60min.), 9.78m (400C-
30min.) and 18.9m (500C-5min.), respectively. It can be seen that a weak preferred {110}
orientation exist for the polycrystalline copper in the as-deposited state. Here, as annealing
progresses, the {110} texture component becomes stronger with increasing grain size as seen
from the sharpening of the intensity in Fig. 8.12.2a-d.
Fig. 8.11.2 – Selected {100} {110} and {111} pole figures obtained from EBSD for PPC copper
with the grain sizes of: a) 3.58m – starting structure, b)7.95m, c) 9.78m and d) 18.9m.
a)
b)
c)
d)
45
90
35
90 90
0
60
173
Earlier computer simulation work performed by Gabacz and Grabski showed that
materials which exhibit a high %fSB would have a strong crystallographic texture [Gabacz
and Grabski (1989)]. Their results were consistent with the experimental work
performed on Fe-6.5% Si and pure Al which both showed a significant fraction of CSL
boundaries along with a strong material texture [Watanabe et al. (1989) and Shibayanagi
et al. (1992)].
Szpunar et al. studied the thermal stability of a nanocrystalline Ni-45%Fe alloy
using X-ray pole figure analysis. Their results indicated that when grain growth occured
the intensity of the {111} component increased at the expense of the {200} component
[Szpunar et al. (1997)]. Somers et al., used EBSD to study the influence of annealing on
the grain growth and texture change in electrodeposited nanocrystalline nickel [Somers et
al. (2004)]. Their results also showed that abnormally growing grains develop a partial
{111} texture from the as-plated {200} texture obtained from XRD. As annealing
temperature increased, rapid grain growth occurred and resulted in a relatively strong
{111} texture component. There were other instances where grain growth also caused a
texture change in electrodeposited nanocrystalline materials during annealing. For
instance, Pantleon and Somers showed that room temperature grain growth of copper
occurred and the initial texture changed from a {111} to a {100} fibre texture [Pantleon
and Somers (2006)].
In a recent study, Klement et al. also used the EBSD technique to determine
changes in the crystallographic texture of nanocrystalline nickel and Ni-20at.%Fe upon
174
annealing [Klement and Da Silva (2007) and Klement et al. (2008)]. Their results
showed that some of the grains began to grow abnormally from an initial grain
orientation of ~{411} to {111}. As annealing continued, grains grew further and the
texture of the electrodeposited material changed to a dominant {111} texture component.
Seo et al. and Kim et al. studied the texture evolution upon annealing for several
electroplated Ni-Fe alloys with varying Fe content [Seo et al. (2007), Kim et al. (2007)].
In these structures the as-deposited materials had a strong {100} fibre component and a
minor {111} component. However upon annealing a strong {111} texture developed at
the expense of the {100} component.
There is currently no theory to explain the texture evolution of nanocrystalline
materials upon annealing. Most of the nanocrystalline electrodeposits studied thus far
often begin with a strong initial texture which changes upon annealing. However, it is
interesting to point out that the low-S nanocrystalline copper studied here behaved very
differently in that it begins and finishes with a random texture before and after annealing.
Perhaps this result could be attributed to the random texture found in the as-deposited
microstructure. It should be noted that the texture of as-electroplated nanomaterials can
be strongly influenced by the choice of plating parameters and the impurities
incorporated into the deposits. For example, results from El-Sherik and Erb showed that
the preferred orientation of the starting structure changed from a strong {200} texture for
material produced from a sulfur-additive free bath to a {111}, {200} double fibre texture
175
for deposits made from a bath with 10g/L of sulfur-containing additives [El-Sherik and
Erb (1995)].
176
8.13 - References for Chapter 8
Arora O.P. and Metzger M., Trans AIME, 236 (1966) 205
Aust K.T. and Chalmers B., Proc. Royal Soc., A201 (1950) 210
Bolling G.F. and Winegard W.C., J. Inst. Metals, 86 (1958) 492
Baro G. and Gleiter H., Z. Metallkde, 63 (1972) 663
Brandon D.G., Acta Metall., 14 (1966) 1479
Chadwick G.A. and Smith D.A., (eds), Grain Boundary Structure and Properties,
Academic Press, London (1976)
Charnock W. and Nutting J., Met. Sci. J., 1 (1967) 78
Dash S. and Brown N., Acta Metall., 11 (1963) 1067
Dechamps M., Baribier F. And Marrouche A., Acta Metall., 35 (1987) 101
EBSD Reference Manual, hkl technology, 73
El-Sherik A.M. and Erb U., J. Mater. Sci., 30 (1995) 5743
Fullman R.L. and Fisher J.C., J. of. App. Phy., 22 (1951) 1350.
Gallagher P.C.J., Met. Trans., 1 (1970) 2429
Gleiter H. and Chalmers B., High Angle Grain Boundaries, Pergamon Press, Oxford
(1972)
Garbacz A. and Grabski M.W., Scripta Metall., 23 (1989) 1369
Grabski M.W., J. De. Phys., 46 (1985) C4-567
Guo H., Chaturvedi M., Richards N. and McMahon G., Scripta Mater., 49 (1999) 3
Hargreaves F. and Hill J.R., J. Inst. Met., 41 (1929) 2257
Hermann G., Gleiter H. and Baro G., Acta Metall., 24 (1976) 353
Irving C., Miodownik A.P. and Towner J.M., J. Inst. Metals, 93 (1964) 360
Ishida Y. and McLean D., Phil. Mag., 27 (1972) 1125
177
Karlsson I. and Norden H., Acta Metall., 36 (1988) 13
Kim J.K., Choi S.H., and Park Y.B., Adv. Mater. Res., 15-17 (2007) 923
Klement U., Da Silva M. and Skrotzki W., J. Microscopy, 230 (2008) 455
Klement U. and Da Silva M., J. of alloys and Compounds, 434-435 (2007) 717
Kobayashi S., Tsurekawa S., Watanabe T., and Palumbo G., Scripta Mater., 62 (2010)
294
Kronberg M.L. and Wilson F.H., Trans AIME, 185 (1949) 501
Kurban M., Erb U. and Aust K.T., Scripta Mater., 54 (2006) 1053
Langer R.A., M.A.Sc. thesis, Queen’s University (1995)
Lin P., Ph.D Thesis, University of Toronto (1997)
Liu D., M.A.Sc. Thesis, University of Toronto (1982)
Lu K., Science, 304 (2004) 422
McCafferty K., M.A.Sc. Thesis, Queens University (1995)
Miura H., Kato M. and Mori T., Colloque de phys., 51-C1, (1990) 263
Murr L.E. and Meyers M.A., Acta Met., 26 (1978) 951
Murr L.E., in Interfacial Phenomena in Metals and Alloys, Addison-Wesley, Reading,
Mass. (1975)
Nakamichi I., J. Sci. Hiroshima Univ., Ser A, 54 (1990) 49
Palumbo G., Lehockey E.M., Limoges D., Tomantschger K. and Vincze A., J. of Power
Sources, 78-1 (1999) 79
Palumbo G. and Aust K.T., in Grain Growth in Polycrystalline Materials III, Weiland H,
Adams B.L., and Rollett A.D. (eds), The Minerals, Metals and Materials Society (1998)
311
Palumbo G, Lehockey E.M. and Lin P., JOM 2 (1998) 40
Palumbo G. and Aust K.T., Can. Met. Quart. 34 (1995) 3
178
Palumbo G., Thermomechanical Processing of Metallic Materials, US Patent
#19930167188, 1994.
Palumbo G. and Aust K.T., in Materials Interfaces, Wolf D. and Yip S.(eds), Chapman
and Hall, London (1992)
Palumbo G. and Aust K.T., Acta Metall. Mater., 38 (1990) 2343
Pantleon K. and Somers M.A.J., J. Appl. Phys., 100 (2006) 114319
Randle V., Colemand M. and Owen Gregory., Mater. Sci. For., 550 (2007) 35
Randle V., Acta Mater., 52 (2004) 4067
Ranganathan S., Acta Cryst., 21 (1966) 197
Read W.T. and Shockley W., Phys. Rev., 78 (1950) 275
Reed-Hill R.E. and Abbaschian R. (eds.), Physical Metallurgy Principles, 3rd
edition,
PWS-Kent Pub., Boston (1992) 195
Rosenhain W. and Ewen D., J. Inst. Met., 8 (1912) 149
Sautter H., Gleiter H. and Baro G., Acta Metall., 25 (1977) 467
Seo J.H., Kim J.K., and Park Y.B., Process and Properties of Structural Nanomaterials.
Edited by Shaw L.L., Suryanarayana C and Mishra R.S. TMS (2003) 213
Shewmon P., in Recryst. Grain Growth and Textures, American Society for Metals,
Metals Park, Ohio, (1966) 165
Shibayanagi T., Takatani H. and Hori S., Mater. Sci. For., 94-96 (1992) 495
Shimada M., Kokawa H., Wang Z.J., Sato Y.S., and Karibe I., Acta Mater., 50 (2002)
2331
Silcock J.M., Rookes R.W. and Barford J. J. Iron Steel Inst., 204 (1966) 623
Simpson C.J. and Aust K.T., Met. Trans., 1 (1970) 1782
Somers M.A.J., Rasmussen A.A., Gholinia A. and Trimby P.W., Mater. Sci. For., 467
(2004) 1345
Szpunar J.A., Megret H. F., Czerwinski F., Clark D.G., and Erb U., Mater. Res. Soc.
Symp. Proc., 451 (1997) 501
179
Viswanathan R. and Bauer C.L., Met. Tran., 4 (1973) 2646
Warrington D.H. and Boon M., Acta Metall., 23 (1975) 599
Watanabe T., in Grain Boundary Engineering, U. Erb and G. Palumbo (eds.), CIM,
Montreal (1993)
Watanabe T., Mater. Sci. For., 94-96 (1992) 209
Watanabe T., Fujii H., Oikawa H. and Arai K.I., Acta Metall., 37 (1989) 941
Watanabe T., Kawamata Y. and Karashima S., Trans JIM, suppl., 27 (1986) 601
Watanabe T., Res. Mechanica, 11 (1984) 47
Wolf D., and Yip S., Materials Interfaces, Chapman and Hall, London (1992)
Zhou Y., Aust K.T., Erb U. and Palumbo G., Scripta Mater., 45 (2001) 49
180
Chapter 9 – Conclusions
Through this thesis, the following scientific/technical contributions have been made.
1. Free standing, copper foils were produced using the pulse electrodeposition
technique. These materials covered the grain size range from 3.58µm down to
31nm and contained sulphur impurities from <10ppm to 230ppm.
2. A property comparison for conventional polycrystalline copper and
nanocrystalline copper was presented. Nanocrystalline copper offers up to a four-
fold increase in the hardness, with only about 40% increase in the electrical
resistivity value. The increase in hardness of copper could help in stiffening the
PWB against warpage, and the additional strength provided could also reduce foil
cracking due to CTE mismatch between copper and the fibre-glass epoxy board.
3. For these materials, several other studies were able to show that the decrease in
grain size significantly enhanced the etchability of copper potentially allowing
PWB’s to achieve smaller copper line widths for high wiring density applications.
4. By varying plating parameters and bath additives, the amount of sulfur impurities
and the grain size in the copper electrodeposits can be controlled. This allowed
for a study of the effect of sulfur impurities on the thermal stability of the
electrodeposited nanocrystalline copper.
181
5. In the as-deposited state, the sulfur impurities were found to be mainly in solid
solution within the copper matrix which provided a strong solute dragging force
to counter the curvature-induced driving force for grain growth.
6. At room temperature, nanocrystalline copper with higher amounts of sulfur
impurities displayed a better thermal stability. Similar results are also found
when deposits were annealed isothermally at 100C.
7. However, using isothermal annealing at 300C, samples with lower sulfur
impurities were found to exhibit better thermal stability. Therefore, the increase
in sulfur impurities does not necessary help in improving the thermal stability of
the electrodeposits.
8. This study showed that the thermal instability of electrodeposited nanocrystalline
copper is associated with a change from strong solute drag to the weaker Zener
drag. The change in dragging force is believed to be related to the maximum
grain boundary solute concentration as explained using the solute dilution theory.
9. For some annealing conditions a mild form of abnormal grain growth was
observed. Abnormal grain growth is likely associated with differences in local
dragging force, causing a growth rate difference between the matrix and the larger
grains.
182
10. Isothermal annealing of low-S nanocrystalline copper (starting grain size: ~50nm)
up to a grain size of ~5µm showed that the frequency of special grain boundaries
slightly increased with increasing grain size within the range from ~36% to ~49%.
Annealing pure polycrystalline copper showed insignificant changes in special
grain boundary frequencies upon increasing their grain size from ~3.58µm to
~19µm .
11. Pole figure analysis showed that there is no change in texture for nanocrystaline
low-S copper upon grain growth, i.e. the texture remains random throughout
annealing. In contrast, pure polycrystalline copper, showed a texture change from
a weak {110} to a strong {110} texture.
183
Chapter 10 – Recommendations for Future Work
1. Production scale-up for nanocrystalline copper foil. Work is currently underway
in collaboration with Integran Technologies, Toronto. To date, much larger foils
(e.g. 8x8”) can be made in their pilot plant facility and hopefully a continuous
drum plating system can be implemented in the near future.
2. Implementation of using nanocrystalline/ultrafine grained copper for plated
through hole (PTH) applications. A proper grain size optimization would likely
increase the mechanical properties of copper which ultimately improves the
reliability of PWBs against PTH cracks.
3. Interconnect stress test (IST) of PWB using nanocrystalline copper. Interconnect
stress testing is one of the standard test methodology for the assessment of PWB
reliability. Here, a prototype PWB made using nanocrystalline copper should be
subjected to thermal cycling, typically from 25 to 150C. IST is one of the
industrial standard tests for PWB failure such as foil delamination and PTH/via
cracks.
4. Grain boundary character distribution and texture analysis of as-deposited
nanocrystalline copper. More recent EBSD systems are now capable to study the
GBCD of materials with spatial resolution of ~30nm. This opens up the
opportunity to characterize the GBCD of the low-S and high-S copper in the as-
184
deposited state. EBSD also provides information in regards to the texture of the
material. This would greatly help to develop a better understanding of the initial
grain boundary structure of the material in the as-deposited states and the
subsequent GBCD evolution upon annealing.
5. With the ability in producing nanocrystalline copper with known impurity levels,
starting microstructures and the thermal stability characteristics at elevated
temperatures, the effect of impurities on GBCD of nanocrystalline copper upon
annealing should be further studied by producing copper foils with various sulfur
contents. Together with thermal profiles (i.e. from DSC peak temperature Tp),
this would help in understanding the change in GBCD upon annealing due to
solute and Zener drag.
185
Appendices
1. Summary of Plating Experiments
An increase in current density generally promotes the nucleation of new crystals
and hence reduces the grain size of the electrodeposits. However, experimental work in
this study showed that an increase in current density alone is insufficient to reduce the
grain size of copper to below 100nm. In fact, applying very high current densities (e.g.
Iavg=500mA/cm2) will result in powdery/brittle deposits because of the insufficient
copper ion supply near the cathode. As the copper ion concentration near the cathodic
surface (i.e: ions in the Nernst diffusion layer) is depleted, other reduction reactions take
place such as hydrogen evolution. An increase in bath agitation, together with pulsed
electrodeposition was found to improve the quality of the deposit by helping copper ions
to diffuse back into the depleted zone.
The length of the periodic pulse also affected the grain size and overall quality of
the electrodeposits. Experimental work showed that a decrease in duty cycle (Ton/(Ton
+Toff) x 100) reduced both the grain size and the internal stress of the deposits. A
decrease in the plating bath temperature also had an effect on the grain size reduction.
Organic additives in the plating bath had a significant effect on grain size reduction. In
fact, the experiments showed that grain refinement into the nanocrystalline regime can
only be achieved by using a small amount of PEG.
186
In order to produce nanocrystalline copper foils of acceptable quality, compromises
were required in terms of many electroplating parameters. Table A1 summarizes how
various factors affected the grain size and surface quality of electrodeposits.
Table A1 – Effect of electrodeposition parameters and bath additives on deposit grain size and surface quality.
Plating variable Grain size Surface quality
Increase in current density Decreased Decreased
Increase in duty cycle Increased Decreased
Increase Toff Increased Increased
Increase in bath temperature Increased Increased
Increase in bath agitation Decreased Increased
Increase in bath pH Decreased Decreased
Increase in PEG concentration Decreased Negligible
Increase in TU concentration Decreased Increased then decreased
187
2. Calculating the Intercrystalline Volume Fractions as a Function of Grain
Size
Palumbo et al. evaluated the relationship between the intercrystalline volume
fraction and grain size (d) [Palumbo et al., (1990)]. The intercrystalline volume fraction
is calculated based on a 14-sided tetrakaidecahedron as the grain shape. The outer skin of
the tetrakaidecahedron is the intercrystalline region with thickness equal to /2, where
is the outer skin (grain boundary) thickness. The grain size (d) is the diameter of the
tetrakaidecahedron fitted into an inscribed sphere. Fig. A2-1 is a schematic
representation of the intercrystalline region using a tetrakaidecahedron model with a
grain boundary thickness of .
Fig. A2-1 – Schematic representation of the model used to evaluate the intercrystalline (grain boundary and triple junction) volume fraction of material. The diagram shown here is a cross-
sectional view of the intersection of three grains (three adjoining tetrakaidecahedra along a polyhedral edge) [Palumbo et al., (1990)].
188
Using the tetrakaidecahedron model, the total intercrystalline volume fraction
(Vtic) as a function of grain size (d) is expressed as:
Vtic = 1-[(d-)/d]
3.
The intercrystalline volume fraction is the sum of the grain boundary volume
fraction (Vtgb
) and triple junction volume fraction (Vttl), which can also be
mathematically expressed as:
Vtgb
=[3(d-)2]/d
3, and Vt
tl= Vt
ic-Vt
gb.
Fig. A2-2 is a graphical representation of these three volume fraction components,
assuming a boundary thickness () of 1nm. It can be seen that the intercrystalline
volume fraction increases very rapidly as grain size decreases to below 100nm.
Fig. A2-2 – The effect of grain size (d) on calculated volume fraction for intercrystalline, grain boundaries and triple junction components. The calculations assumed a grain boundary
thickness () of 1nm [Palumbo et al., (1990)].
189
3. Definitions for Normal and Abnormal Grain Growth
Definitions for normal and abnormal grain growth can be found in many material
science textbooks. The following definition is taken from Dr. Askeland and P.P Phulé, in
“The Science and Engineering of Materials”, 4th
ed., Thomson-Brooks/Cole, Pacific
Grove CA, 2003. “In normal grain growth the average grain size increases steadily and
the width of the grain size distribution is not affected severely. In abnormal grain
growth, the grain size distribution tends to become bi-modal (i.e., we get a few grains that
are very large and then we have few grains that remain relatively small).”
Furthermore, Fig. A3 represents schematic diagrams of the two cases of grain
growth (F. Haessner, in “Recrystallization of Metallic Materials”, 2nd
ed., Riederer-
Verlag, Stuttgart Germany, 1978).
Fig. A3 – Schematic diagrams of normal and abnormal grain growth (t=annealing time, D=grain size) [Haessner, (1978)].
In normal grain growth, grain growth occurs as a mechanism to lower the grain
boundary area of a material and therefore its total grain boundary energy. In normal
grain growth, the grain size distribution remains unimodal and the average grain size
190
increases steadily as annealing progresses (Dm=Dmax). On the other hand, abnormal
growth occurs in such a way that a bimodal grain size distribution is created as some
grains become very large at the expense of the smaller grains (Dm<<Dmax).
191
4. Calculations to Estimate the Sulfur Concentration [S] Required to Form a
Monolayer of Sulfur at the Grain Boundary
In order to form sulfur containing second phase particles at a grain boundary,
there must be an adequate amount of sulfur in the boundary’s vicinity. During grain
growth, dynamic segregation occurs as moving boundaries accumulate impurities.
Furthermore, diffusion within the interface would cause rearrangement of atoms that
ultimately form the second phase particles.
This calculation assumes that a sulfur monolayer has the same crystal structure as
copper, and the interface region (grain boundary) has the same atomic density per volume
Dv,IN as the crystalline region. The thickness of the interface is assumed to be =1nm.
The estimated concentration, in atomic %, is calculated based on the ratio
between the (111) planar density (111) and the average atomic density of the interface
Dv,IN with an interface thickness () of 1nm (i.e: the number of atomic sites per cubic
meter of interface material). Haasz et al. observed a density reduction of 2.3% in the
intercrystalline region compared to the crystalline region [Haasz et al. (1995)]. As a
result, a correction factor of 2.3% will be used for calculating the average atomic density
of the interface.
192
21.3at.%or 213.0D
]S[
atoms/m1028.8
.977 m10g/mol55.63
/mcm10atoms/mol10023.6g/cm94.8
factor correction 97.7% hicknessBoundary tCopper of weight Atomic
number sAvogadro'Copper ofDensity D
m
atoms107619.1
))nm362.0(2(4
3
atoms 2
(111) ofarea Planar
(111) in atoms of #
INv,
111
219
9336233
INv,
2
19
2
111
The calculated value tells us that when the [S] concentration along the grain
boundary reaches 21.3at.%, a monolayer of sulfur is created and second phase particles
would likely form by some atom rearrangements. This is a very conservative estimation
because the 111 plane requires the most sulfur atoms in order to fill up a monolayer at the
interface. A sulfur monolayer can be formed at a much lower concentrations for all other
planes because of their lower planar density. This calculation also assumes a uniform
distribution of sulfur along the interface. The sulfur monolayer is likely to form locally
in regions, and then reacts to form precipitates well before a uniform sulfur monolayer is
created. Since sulfur can locally coalesce along the grain boundary, second phase
particles are likely to form when [S] is less than 21.3 at.%.
References:
Haasz T.R., Aust K.T., Palumbo G., El-Sherik A.M., and Erb U., Scripta Mater., 32
(1995) 423