This item is only available as the following downloads:
THIN FILM METALLIZATION
FOR MICRO-BIMETALLIC ACTUATORS
JONATHAN FRANK GORRELL
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
I would like to express my appreciation to advisor Dr. Paul H. Holloway for his
support, guidance and encouragement. I would also like to give special thanks for pro-
viding a new perspective to some of the problems I encountered in this research. Thanks
are also due to my doctoral committee members Dr. DeHoff, Dr. Singh, Dr. Adair, and
Dr. Kumar for their assistance and interest in my work.
This work would not have been possible without the support of EG&G ICSensors
and the DARPA MEMS research program. I am also grateful to Dr. Hal Jerman for his
efforts in obtaining this grant and his support of my work.
I would also like to thank my friends and coworkers for their support and encour-
agement and who made my return to school most enjoyable. I would like to thank Mark
Davidson for providing me with his insight into the workings of a university and scien-
Lastly, I would like to thank my parents James and Billie Gorrell for instilling a
love of learning and the belief in myself that I can achieve the goals I set for myself from
a very early age. Without these values I would have never tried much less have com-
pleted this program.
TABLE OF CONTENTS
ACKNOWLEDMENTS.................... .............. .......... ......................... i
A B ST R A C T.................................................................. ............... ................... vi
1 INTRODUCTION....................... ................... ..............
M otivation and Objective.......................... ................ ..................... 1
Scope of Present W ork.................................................. ..................... 6
2 LITERATURE REVIEW............................. .. ....................... 8
Introduction....................... ............. .... ...................... 8
The Bimetallic Strip..................................................... 8
The Simple Bimetallic Cantilever................................. ............ 9
The Bimetallic Disk..................... .. .................... 10
The Effects of Plastic Deformation........................ ................. 12
Atomic Bonding.................................................................... 13
Plastic Deformation In Metals............................. ...................... 17
C reep in M etals.................................................. ................................. 19
Deformation Mechanism Maps (DMMs)............................................ 22
Annealing ................................................................ ...................... 22
Strengthening of Bulk Metals........................................ 24
Ordered Intermetallic Compounds.................... ...................... 27
Thin Film s............................................. ....................... .................... 28
Environmental Stability of Materials....................... ..................... 35
3 EXPERIMENTAL PROCEDURE................................................... 37
Introduction....................... .. ................. ..................... 37
Thin Film Deposition......................... ........ ........................... 38
Copper Gold Heat Treatment........................ .... ................. 39
Stress and Stress Relaxation Measurements...................... ........... ... 39
Powder X-Ray Diffraction (XRD)........................... .................... 40
Atomic Force Microscopy (AFM)................................................... 42
Scanning Electron Microscopy (SEM)...................... ..... ........... 42
Electron-Probe Microanalysis (EPMA)..................... ...... ........... .. 43
Transmision Electron Microscopy (TEM)................................... ............ 43
E tch ing ................................. ................................. ............................. 44
Auger Electron Spectroscopy (AES).......................... ......... ........... .. 44
Curve Fitting........................................................ 45
Numerical M odeling.................................................... 46
4 R E SU LT S............................................................................ ............. 47
Introduction....................... .................................... 47
T201 Aluminum..................................................... 47
5052 Aluminum............................................... 59
2090 Aluminum............................ ............................ 68
C opper..................... ................................. .................................. 70
Titanium, Manganese and Nickel....................... ........ .......... .... 72
Copper-Gold Intermetallics........................................... 74
Aluminum-Titanium Intermetallics................................................ 77
Modeling the Effects of Oxide Thickness on a Bimetallic
Actuator's Curvature................................................ 78
5 D ISCU SSIO N .......................................... ................... ..................... 81
Introduction................ ........................... .... .................... 81
Thin Film Strengthening............................................. ...................... 81
Solid Solution Strengthening..................... ...................... 82
Precipitation and Multiphase Hardening............... ........ ........... .. 84
Ordered Phases and the Order Disorder Transition................................ 85
Stress Relaxation Mechanism in Thin Films........................................... 86
Mode 2 Stress Relaxation......................... .................... 88
Mode 3 Stress Relaxation.................... ..... .................. 89
High Temperature Application.................... ...................... 90
O xidation ................................................................ ...................... 91
Strength at Elevated Temperatures....................... ................. 91
Figure of M erit................................................................................... 92
6 CONCLUSIONS ..................................................................... 94
7 FUTURE WORK............ ........ .................................. 96
A DERIVATION OF RELATIONSHIP BETWEEN YOUNG'S
MODULUS AND COEFFICIENT OF THERMAL EXPANSION.... 98
B CURVE FITTING OF STRESS RELAXATION DATA........................ 101
C CALCULATION OF CHANGE IN CURVATURE OF A
BIMETALLIC STRIP BASED ON OXIDE THICKNESS................ 106
D MECHANICAL PROPERTIES OF EXAMINED METALS............... 108
BIBLIOGRAPHY ...................................................................................... 109
BIOGRAPHICAL SKETCH......... .................................................. 113
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
THIN FILM METALLIZATION
FOR MICRO-BIMETALLIC ACTUATORS
Jonathan Frank Gorrell
Chairperson: Dr. Paul H. Holloway
Major Department: Materials Science and Engineering
In this study, eleven different thin film metallization systems were evaluated for
use in micro-bimetallic actuators for microelectromechanical structures. These films
were evaporated or sputtered onto silicon wafers. The film stress and stress relaxation
were determined by measuring changes in the wafer curvature. The phases and micro-
structure of these films were evaluated with, scanning electron microscopy, transmission
electron microscopy, Auger electron spectroscopy, electron probe micro-analysis, X-ray
diffraction and line shape analysis, and atomic force microscopy.
Bimetallic actuator may be operated to generate either force or displacement. The
displacement mode is dominated by the coefficient of thermal expansion while the force
mode is a function of both Young's modulus and coefficient of thermal expansion of the
active layer material. In both modes the maximum displacement or force is determined
by the material's yield strength. A figure of merit was developed to aid in material
The 5052 aluminum alloy films showed that solid solution strengthening can
double the yield strength of a thin film. The T201 aluminum alloy films showed that
precipitates can increase yield strength by 2.5 times. The 2090 alloy film oxidized during
the first heating. Based on isothermal stress relaxation data and changes in the micro-
structure of the 5052 and T201 alloy thin films, two mechanisms involving logarithmic
creep have been postulated to cause stress relaxation. One mechanism is movement of
dislocations in slip systems that terminate at the surface while the other is dislocations
moving in slip systems that terminate at grain boundaries.
Copper gold intermetallics films oxidized and plastically deformed before the
order-disorder transformation occurred, but showed that ordered intermetallics have a
lower stress relaxation rate than the solid solution phase. The Al Ti films showed no
stress relaxation at 4500C, plastically deformed only above 500C, and had limited oxida-
tion up to 800C.
Nickel, copper, titanium, and manganese films all oxidized on their first heating to
3500C. The copper film also oxidized at 500C over 48 hours. Calculations also showed
that the passivation oxide on aluminum alloys can significantly reduce preformance a
bimetallic actuator. Thus oxidation resistance is a significant requirement for materials
for thermal actuation.
Motivation and Objective
The miniaturization of electronics has led to what is now called the information
age. More and faster electronics packaged in ever smaller volumes allow massive
information processing and exchange. Information processing systems are, however,
limited by the devices that allow them to perceive and affect the physical world. Sensors
and actuators have been connected to computers and control systems from the early days
of computing, but these sensors and actuators have been, until recently, large electrome-
chanical devices. Researchers have begun to miniaturize electromechanical devices using
the technologies developed for microelectronics [Hog96]. These new micro sensors and
actuators have created a new class of devices called MEMS, microelectromechanical
systems [Mas95, Fra97, Ang83, Bea96, Mad97]. Integration of MEMS sensors and
actuators is enabling the development of mechanical, chemical and biological "smart
systems" that are able to interact with the physical world.
MEMS have already become a significant industry. Yearly sales are in the billions
of dollars and the market continues to grow rapidly [Mas95]. This market is currently
dominated by pressure sensors and accelerometers. The miniaturization of pressure
sensors has reduced sensor cost to the point that in medical applications the sensors can
be discarded after a single use. Accelerometers are predominately used to determine
when to deploy automobile air bags. Micro actuators have yet made as many inroads into
industry. Currently the biggest application of MEMS actuators is for ink jet print heads
Figure 1.1: SEM image of an EG&G IC Sensors bimetallic
thermally actuated micro-valve.
[ROW95]. The market of MEMS actuators will continue to develop as actuators are
needed to produce a truly "smart system" that can not only monitor processes but also
control them. However, additional research and development are needed to improve the
performance of micro actuators, and this is the general focus of this work.
Micro actuators have been designed based on a number of physical properties
including electrostatic, electromagnetic, piezoelectric, magnetostriction and thermal
activation bimetallicc, shape memory alloys, and thermopneumatic) [Mad97]. Table 1.1
shows a comparison of the different actuation techniques and their differences in perfor-
mance. While all of these techniques have advantages for particular applications, this
study focuses on the thermally activated bimetallic actuator. Bimetallic actuators are of
interest because they can produce a relatively large force or displacement, the force is
relatively constant throughout the travel of the actuator and bimetallic actuators are easily
manufactured using semiconductor processing technology. Micro bimetallic actuators
have already been employed in commercially available products [Mad97, Jer94]. Figure
1.1 shows a bimetallic actuated micro valve produced by EG&G ICSensors produced by
Table 1.1: Comparison of actuation principles [Bel97].
Par r B c Electro- Piezo- Electro- Thermo-
static electric magnetic hydraulic
Fore pr High(O.I 1) Iow(0'. 10') High High High
I rotor, N
Deflection, mm High OOl I) Low (10'. 10') LUw' High High
Work per cycle per L () gh H
l m--2, J High (10') Low (10 ) High High
I mm'. J
Frequency range, Low 01-100 (Mech Resonal (Mcch Resonat L.w
Khz (limiting factorr)(Heat Tanfer frequ y) higheqey) Hea Tranfer
\bltage, V Law (3 12) High (100 300) High Low Low
Current, mA High (0.1 10) Lw (10 10') Low High High
Power Consunption. H,
Power Consumption High(10' 0.1) Low (10 W. 1-) Law High High
Efficiency, % 0.01 0.5 000l o.01
Size: Length, mm Semll (1. 1)
Width, mm Sal ( 0.1) Sl Lage Large Lrge
rAddeiional s lih lVtage (Mageti field Media separaton
Cost Low Low High High High
aluminum metallization on silicon. Even though silicon is not a true metal, this structure
is still referred to as a bimetallic actuator.
The objective of this research is to improve the performance of micro-bimetallic
actuators and broaden their range of operating parameters (stress, displacement, tempera-
ture range and stress relaxation) by identifying better or improve the materials of con-
struction. In order to understand how to improve the performance of the bimetallic
actuator, it is important to understand how the device works. A bimetallic actuator is
made by bonding together two different materials that have a large difference in thermal
expansion, such as aluminum and silicon, to create a bimetallic strip or spring that is
heated or cooled. Upon heating, the material with the larger coefficient of thermal expan-
Aluminum Active Layer
Figure 1.2: Diagram of a bimetallic strip, showing how the differencein
thermal expansion bewteen the active layer, aluminum, and
passive layer, silicon, causes the bimetallic strip to bend.
sion, called the active layer, expands more than the material with the smaller coefficient,
called the passive layer (see Figure 1.2). As these two materials are bonded together, a
shear stress develops between the layers due to the different rates of expansion, producing
a force that causes the structure to bend. Upon cooling the reverse processes occurs and
the bimetallic strip returns to its original position, or deflects in the opposite direction
with continued cooling. The geometry may be a disk, cantilevered beam, or more com-
While the current generation of bimetallic actuators works well for a number of
applications, their use could be significantly expanded by improving the position and
force stability of these actuators and increasing the operable ambient temperature limit
from 75C to 500C. Force and position stability have been found to be a problem in
... :.: ..L-
* I i;r: ..; .
:' .:''5;H '.
Aluminum Valve Closed
\ i,3: Dkiphragm
Valve Body \
Heating Stress Relaxation
Aluminum Valve Open
\ _\ 4/, -* _
Figure 1.3: Cross Sectional Diagram of bimetallically actuated micro-valve.
some bimetallic actuators such as the one shown in figure 1.1. It is difficult to maintain a
constant flow from these valves over extended periods of time. The opening of the valve
decreases with time. Force and position instability in bimetallic actuators are caused by a
relaxation in the shear stresses between the active and passive layers of the bimetallic
strip which in turn reduces the force bending the structure. Figure 1.3 shows a cross
sectional diagram of a bimetallic actuated micro-valve and the effects of stress relaxation
on the valve operation. The inability to maintain a constant force or position for extend
periods of time complicates the usefulness of the device. These limitations result prima-
rily from creep in the active metal layer of the bimetallic actuators. Thus, improvements
can be achieved through the identification and or development of better metal thin films.
While improvement of materials for bimetallic actuators may at first seem to be a
narrow subject, it requires examining much broader materials questions. How can a thin
metal film be strengthened? What are the atomic mechanisms responsible for the stress
relaxation in thin films? What materials are sufficiently chemically stable at elevated
temperatures to be useful? These questions are of general interest to the entire MEMS
field because micro-machines use thin metal films as structural components, and struc-
tural failure and environmental stability are concerns in any mechanical device. Also, the
continued reduction of interconnect widths in microelectronic integrated circuits lead to
greater stresses in thin metal lines that may fail due to stress induced voiding (SIV)
[Bow74]. Therefore, this research will examine questions of interest to many disciplines
The temperature change of an actuator is normally accomplished by resistive
heating. The passive layer of most micro-bimetallic actuators is silicon, and a resistive
heater is easily made by doping the silicon. A thin silicon oxide layer grown on the
silicon is all that is needed to keep the resistive heater from shorting to the active layer.
Cooling is accomplished by conduction of heat to the valve body and the surrounding
structure. The thermal masses of micro-bimetallic actuators are so small that they can be
cycled rapidly. Other heating and cooling methods could be used, such as a Peltier
element [Cad60]. In the present work, the mechanisms to change the temperature of the
bimetallic actuator is not of direct interest. Instead the present research is concerned with
the mechanical and thermal properties of the materials in the bimetal strip, and how they
affect actuation. Optimization of the performance of bimetallic strips will be the focus,
which directly translates into optimizing the performance of a bimetallic actuator.
Scope of Present Work
The three specific objectives of the present work are as follows. (i) The first
objective is to determine the effectiveness of different strengthening techniques for thin
film metallization. There are a number of techniques used to increase the yield and
ultimate tensile strength of bulk metal alloys, but their effectiveness in thin films has yet
to be determined. Increased yield strengths of the metal thin films would increase the
maximum force and displacement produced by an actuator, and could increase force and
displacement stability. (ii) The second objective is to evaluate use of materials at el-
evated temperatures. The primary requirements for elevated temperature applications are
high strength over their operation temperature range and resistance to creep and oxida-
tion. (iii) The third objective is to identify the mechanism responsible for isothermal
stress relaxation in thin films. It is this relaxation that causes the force and position
instability seen over extended time periods.
To meet these objectives, eleven different materials were deposited and evaluated.
The evaluation focused on determining the material's strength, environmental stability
and rate of stress relaxation. For those materials that showed promise and to understand
the underlying mechanisms responsible for the stress relaxation, the morphology and
microstructure of several samples were examined. The materials that were evaluated fall
into three groups: (i) aluminum alloys, (ii) elemental metals, and (iii) intermetallics. The
aluminum alloys were used to identify the mechanisms responsible for isothermal stress
relaxation, and to determine the effectiveness of the different strengthening techniques.
The elemental metals and intermetallics were evaluated at elevated temperatures. Some
of the intermetallics also exhibited ordering and non-linear thermal properties. The
evaluation of several materials systems allowed a wide variation in properties and there-
fore provided significant insight into the materials requirements for MEMS devices.
The organization of this dissertation is as follows. Chapter 2 is a review of the literature
pertinent to this study. The experimental procedures used to deposit, test, and analyze the
samples are presented in chapter 3. Chapter 4 contains the results of these tests and
analyses. The results are discussed in chapter 5. The conclusions are summarized in
chapter 6 and suggestions for future work are presented in chapter 7.
Thin film metallization for bimetallic actuators are the focus of this research,
therefore a number of topics need to be reviewed. The first topic is the theory of the
bimetallic strip. In examining literature, the concern is to determine how the different
materials properties affect the performance of a bimetallic strip. With an understanding
of the effects of materials properties, improved materials selection can be used to opti-
mize performance. After their identification, the origin and interdependence of these
properties can be discussed using the theories of atomic bonding and thermal expansion.
In addition, there are a number of processes that may affect the metallization of a bimetal-
lic strip, such as creep, plastic deformation, and oxidation. Lastly, these bimetallic
actuators are manufactured with thin film technology and the effects of these processes
are discussed. There is no perfect material for use in all bimetallic actuators, since
different applications require different properties. With an understanding of the interde-
pendence of the different materials properties, the tradeoffs that are needed for any given
application can be better selected.
The Bimetallic Strip
The formulation of the bimetallic strip problem was first reported, in part, by G.
Gerald Stoney [Sto09]. Stoney determined the relationship between curvature and stress
in a bimetallic strip. S. Timoshenko [Tim25] then derived the equations for displacement
and force of a bimetallic strip versus change in temperature. Lastly, Townsend [Tow87]
generalized the equations for a multi-layer structure.
The formulation and solution of the bimetallic strip problem are dependent on the
geometry and boundary conditions used, therefore no single solution is generally appli-
cable. Roark's Formulasfor Stress and Strain [You89] provide solutions for many shapes
and boundary conditions. This complexity of solutions and lack of solutions for unusual
shapes has led to the widespread use of finite element analysis to design bimetallic
actuators [Bar94] [Tsa92].
The Simple Bimetallic Cantilever
To determine the effects of materials properties on performance, the simple
bimetallic cantilever, simply supported on only one end, is examined. The simplifying
assumptions made in this case are that the length is much greater than the width or thick-
ness, the materials are perfectly elastic, and the material returns to its original dimension
once the load is removed [Mey84]. Figure 2.1 shows a diagram of this case where E, and
E2 are the Young's moduli, L is the length of the strip, h is the thickness of the strip, a and
b are the thickness of the first and second layer, w is the width of the strip, a, and ca are
the coefficients of thermal expansion, and the curvature of the bimetallic cantilever, k,
for a given change in temperature, AT, is[Tim25]
6(a2 -a )(AT)(l+ )2
h[3(l+b))2 +( 2 bE2 (1)
For the case of both layers of the strip being the same thickness a = b and equal moduli
the equation for curvature reduces to [Tim25]
G F Iaver 1 a
LayrEer 2 h
LoVer 2 b
--- L >
Figure 2.1: Diagram of bimetallic strip
3 (a,- 2)
k = AT. (2)
The error induced by assuming the moduli are equal is only three percent for a factor of
two change in either material's modulus. Therefore, the primary material's property
affecting the displacement in this case is the difference in the coefficients of thermal
The equation for the restraining force, R, that is needed to keep a bimetallic strip
from moving for a given temperature change DT is the maximum force a bimetallic strip
can produce. This restraining force is given by equation 3:
(E + E2) h (3)
R= h 2. ( -a,)AT.
Thus the force produced by a bimetallic strip is a function of both the thermal expansion
and the Young's moduli of the materials.
The Bimetallic Disk
To evaluate the effects of materials properties on performance for a more complex
case, the bimetallic disk is examined. Due to the complexity of the equation describing
\ CA 172
c Contour of
Constant Force Contour of / SMg
0 10 20 30
Expansion Coefficient (XE-6/OC)
Figure 2.2: Equal potential lines for force and displacement for a bimetallic disk [Jer95].
the bimetallic disk the evaluation was performed by calculating and plotting the
equalpotential lines for maximum displacement (curvature) and force of a bimetallic disk
as a function of Young's moduli and the coefficient of thermal expansion, as shown in
Figure 2.2 [Jer95]. This model is based on a bimetallic strip where the passive layer is
silicon and both the active and passive layers are 5ltm thick. In agreement with equations
(2) and (3) this figure shows that displacement is predominately limited by the thermal
expansion of the active layer material. There is some dependence on the Young's modu-
lus but this occurs only for a modulus lower than that found in most engineering materi-
als. The equal potential line for force shows that thermal expansion is not as significant a
factor as in displacement. Therefore, while the relationships shown in Figure 2.2 for a
bimetallic disk are far more complex than the relationships determined by equations 2
and 3 for a single fixed cantilever, the trends and limiting materials properties are the
same. Displacement in a bimetallic structure is primarily limited by the thermal expan-
sion of the active layer material. In contrast, the force generated by a bimetallic strip is
dependent on both the thermal expansion and Young's modulus of the active layer mate-
rial. These trends and limitations are expected to hold for other configurations as well.
The Effects of Plastic Deformation
The assumption that a material performance is perfectly elastic is only valid for a
range of applied stresses. Beyond this range of stress the material will not return to its
original shape when the stress is removed and the material is said to have undergone
plastic deformation [Mey84](see Figure 2.3). Plastic deformation in either layer of a
bimetallic device invalidates the relationships in equations (1), (2) and (3), and would
reduce the force or displacement produced from the strip. The effects of plastic deforma-
Ultimate Tensile Strength, UTS
Offset Yield Strength,
Yield Strength, YS B
u\/ Breaking Point
Young's Modulus, E, Stress/Strain
Figure 2.3. Typical stress strain curve for a ductile metal.
tion can be modeled as a non-linear reduction of the thermal expansion or Young's modu
lus of a material. While not considered in the above equations that predict the force or
displacement of a bimetallic strip, the stress at which plastic deformation begins, called
the yield strength [Ask89, Mey84] (see Figure 2.3) is also a critical materials parameter.
A material's yield strength determines the maximum force it can exert on another layer
and in turn may determine the maximum force or displacement a bimetallic strip can
Therefore, the materials properties of interest depend upon whether the applica-
tion requires force or displacement. In the displacement mode, the material's thermal
expansion and yield strength are the primary limiting properties. In the force mode,
thermal expansion and Young's modulus both contribute to the force generated by a
bimetallic strip, and yield strength is a limiting materials property.
The atomic bonds that hold a material together also determine its thermal expan-
sion and Young's modulus. There are a number of aspects of atomic bonding that affect
these physical properties, and these aspects will be examined in this section. This review
is not intended to cover all aspects of thermal expansion, but only those that affect the
performance of a material in a bimetallic actuator.
The first aspect of atomic bonding that determines a material's rate of thermal
expansion and its Young's moduli is the strength of the atomic bonds. Figure 2.4 shows
the asymmetric potential energy well that an atom is drawn into when it bonds with
another atom [Ask89]. The depth of this potential well determines the strength of the
bond. The deeper the potential well, the stronger the bond. A deep potential well nor-
mally has a larger dE/dx which is proportional to the Young's modulus.
The rate of thermal expansion of a material is due to the asymmetry of the poten-
tial energy well. The addition of energy to a material causes an increase in vibration of
Stime-averaged position of atom
5 Increase in interatomic distance,
Figure 2.4: Variation in potential energy with distance between atoms.
the atoms. This increase in atomic vibration causes the atoms to move up in the potential
well. As the atoms move higher their time-averaged position increases (see Figure 2.4)
and the material expands [Tou75]. The deeper the potential well, the smaller the change
in the time-averaged position of the atoms for a given increase in temperature, and a
lower thermal expansion is observed. The rate of thermal expansion is frequently ex-
pressed as the coefficient of thermal expansion, CTE. A material's CTE is the derivative
of its thermal expansion curve, AL/L versus temperature. To a first approximation, the
thermal expansion is inversely proportional to the atomic bond strength. Thus, the
Young's modulus and thermal expansion of a material are inversely proportional. Appen-
dix A shows the thermodynamic proof of this relationship.
The second aspect of atomic bonding that affects the rate of thermal expansion
and the Young's Modulus is the directionality of the bonding. The directionality of the
bonding results from the type of bonds that form between atoms, the crystallography of
the material, the number of nearest neighbors, the coordination number, and the bond
length. Graphite is a good example to illustrate the effects of directionality of bonding.
Graphite has a layered structure where the bonds within a layer are strong covalent bonds,
but the bonds between atoms in different layers are much weaker van der Waals bonds
[Eva79]. This difference in bonding causes the c-plane coefficient of thermal expansion
(CTE) to be 2.8xl0-60C- while between the planes the CTE is 4.4x 106OC'.
In many instances, the directionality of bonding does not change the bond type but
still has a major affect on physical properties. Zinc illustrates this point well. The atoms
in zinc are bonded into a hexagonal close packed crystal structure, with metallic bonds
between all atoms. Along the c-axis the CTE for zinc is 5x10-60C, but along the a-axis
it is 65xl0-60C-1 [Tou75]. Thus there is a 13-fold difference in CTE between different
planes due to the crystallography that is determined by the bonding.
There are other materials with small anisotropy but unexpected physical proper-
ties due to unusual crystallographic structure. Manganese is one such metal [Dea52].
While o-manganese has a cubic structure, it does not have a simple face centered cubic
(FCC) or body centered cubic (BCC) crystal structure [Ask89]. The unit cell is based on
the BCC structure but contains 58 atoms in 29 pairs [Dea52]. The CTE for manganese is
22x 106 C-', almost double what would be expected for a material with a melting tem-
perature of 12440C. In comparison, beryllium melts at 12780C and has a CTE of only
12x10-6 'C- [Wea76].
The directionality and crystallographic structure of a material are not always
static. Pure iron undergoes two allotropic transformations between solidification at
15380C and room temperature. Iron first solidifies into a BCC structure at 15380C,
converts to an FCC structure at 13940C, and then converts back to a BCC structure at
9120C [Kra90]. At each allotropic transformation, there is a significant change in thermal
expansion and other physical properties. Discontinuities in a material's CTE could be
useful in a bimetallic actuator. Unfortunately the transitions in iron are far above the
-2 0 20 0 40 6W0
Figure 2.5: Linear thermal expansion for intermetallic compound CuAu and
Cu1Au. The discontinuity in thermal expansion is due to the order-disorder
transion of these compounds [Tou75].
temperature currently being considered for practical devices. However, transformations
are used in shape memory alloy actuators [Mad97].
Another change in bonding that can significantly affect properties is the order-
disorder transformation exhibited by some intermetallic compounds [Tou75]. Intermetal-
lic compounds are made up of two or more elements, A and B. In ordered alloys, the A-B
bond may be stronger or weaker than the A-A or B-B bonds and thus there is a driving
force to form one type of bond over the other [Cah83]. Ordered intermetallic compound
may form directly upon solidification, as in Al Ti and NiAl. However, there are other
intermetallic compounds that first solidify as a solid solution and then order at a lower
temperature. CdMg3, CuAu CuAu, and Cu3Au are all examples of materials that
undergo a solid phase disorder/order transition [Hov64, Fed58, Cul78, War69]. The
material may undergo a substantial contraction upon ordering (see Figure 2.5). This
contraction is due to the ability of the atoms to pack tighter in the ordered phases due to
the stronger atomic bonding. This transition may also increase the material's yield
strength, reduce the creep rate, and increase electrical conductivity.
Plastic Deformation In Metals
As mentioned above, plastic deformation in either layer of a bimetallic actuator
will limit the force and displacement of the actuator. Therefore, the yield strength deter-
mines the maximum force and displacement of an actuator. Atomic motion becomes
rapid as the temperature of a metal exceeds half its melting point, therefore the homolo-
gous temperature, TH, is defined to be the ratio of the temperature to the melting point of
the material in degrees Kelvin. At TH < 0.5 dislocation motion is the dominant mecha-
nism of plastic deformation. Inhibition of the motion of dislocations would strengthen
the material. This is in contrast to the thermal expansion and Young's modulus of a
material that can not be changed by inhibiting dislocation motion. In addition, many of
the methods used to inhibit plastic deformation do not significantly alter the CTE or
Young's modulus of the material. Therefore, improved yield strength can be used to
produce a better actuator. In this section, the atomic processes responsible for plastic
deformation will be examined, along with methods to inhibit these operations.
The two main processes that take place in a material during plastic deformation
are: dislocation movement and multiplication. Movement of dislocations produces a
change in shape of the material. However, the dislocation density in an annealed sample
is not great enough to produce the observed plastic deformation. Thus the dislocation
density is increased during plastic deformation by dislocation multiplication [Ree92,
With respect to dislocation motion, they can move through a crystal by either
dislocation glide or climb. Dislocation glide occurs when a dislocation moves on a slip
plane and in a slip direction of the crystal [Ree92]. The combination of a slip direction
lying along a slip plane creates a slip system. There are a limited number of slip systems
and they are determined by the crystallography of the material. Dislocation may not be
able to glide in response to an applied force because a slip system is blocked. On the
other hand, the stress resolved onto that slip plane in the slip direction may be below the
critical value to cause motion of dislocation. This is known as Schmid's law [Ask89].
The second way for a dislocation to move is by climb, in which it moves between parallel
slip planes. Climb allows dislocations to move off blocked slip planes. It occurs by
vacancy movement [Ree92], but both the vacancy concentration and movement are very
dependent on the temperature of the material [Deh93]. Dislocation climb only becomes a
dominant process at high homologous temperature T,. For most metals dislocation climb
becomes significant at T,>0.5 and for most intermetallics at T,>0.7. Above these tem-
peratures dislocations may rapidly climb.
The process of dislocation multiplication can occur in many ways. One of the
more significant is the Frank-Read source [Ree92] by which dislocation glide can in-
crease the dislocation density. In a Frank-Read source, a dislocation line on a slip plane
is pinned or blocked at two points. In response to an applied resolved shear stress the
dislocation bows out between these two points and forms an incomplete dislocation loop.
A critical stress exists, dependent on the dislocation line length, above which complete
dislocation loops form. This mechanism is complicated by the fact that dislocations can
pin one another, thus forming more Frank-Read sources. This is one of the reasons that
metals harden when they are cold worked, which is known as "work hardening" [Bro82].
There generally exists an inverse relationship between temperature and yield
stress independent of the mechanisms responsible for the deformation. For dislocation
glide, random thermal excitation helps dislocation overcome the Peierls-Nabarro barrier
[Mey84] and thus lowers the yield stress. The effects of temperature on dislocation
density and mobility have already been discussed and these effects also lead to reduced
yield stress with increased temperature. While an inverse relationship exists between
yield strength and temperature it is a non-linear function specific to each material.
The stress at which plastic deformation begins is also dependent on whether the
force is applied in tension or compression. There is no general relationship between the
compressive and tensile yield strength and most yield strength data for metals are for
tensile yield strength. As the thin films in a bimetallic actuator can be in tension or
compression the variation in the compressive versus tensile yield stresses could be sig-
Lastly, the crystallographic texture (which is the preference for one crystallo-
graphic orientation) of a material also affects its yield strength [Mey84, Bro82]. Baldwin
[Bal46] showed up to 30% change in tensile yield strength based on the direction of the
applied stress relative to the texture of rolled copper sheet.
Creep in Metals
In the preceding section the mechanisms responsible for the rapid plastic deforma-
tion of materials in response to an applied stress were examined. However, materials
continue to plastically deform in response to applied stresses over extended periods of
time (from minutes to years). Time dependent plastic deformation is known as creep. In
this section the different types of creep and the mechanisms responsible for the time
dependent plastic deformation will be discussed.
Figure 2.6 [Cah83] shows the four major types of creep that occur in materials as
a function of the homologous temperature versus the resolved stress, o, normalized by the
shear modulus, i.
At stress levels below which dislocation can move (ao /g < 10-8) anelastic creep
occurs. In anelastic creep, interstitial atoms move to interstitial sites that have been
elongated by the applied stress [Cah83]. This movement of interstitial atoms creates a
0.5 1.0 Theoretical
10 \ Strength
10 HighTemperature St
10 \ (Andrade Creep)
Low Temperature Creep
Figure 2.6: The creep diagram defining the conditions of temperature and stress which
produce the four principle types of creep. The temperature is plotted as the homologous
temperature and the stress is normalized by the shear modulus. a = applied resolved
shear stress, p = shear modulus, and oas. = critical resolved shear stress of a well
annealed crystal [Cah83].
small deformation of the material that is not permanent as the interstitial atoms will
randomly redistribute once the applied force is removed. Since the materials examined in
this study do not contain a significant percentage of interstitial atoms, this creep mecha-
nism is of little concern.
Herring-Nabarro-Coble creep [Her50, Cob63] produces a deformation in the
material by a net mass diffusion. In the Herring-Nabarro model, a net flux of vacancies
move away from the axis of the applied stress produces a net flux of atoms to the axis of
applied tensile stress, thus elongating the sample in the direction of the applied force.
This mechanism of creep requires a high concentration of vacancies with high mobility,
and only occurs at T. > 0.9 [Cah83]. Coble expanded this idea of bulk diffusion to
include the diffusion of atoms along grain boundaries. Grain boundary diffusion has a
lower activation energy than bulk diffusion because a grain boundary is an array of
dislocations. This lower activation energy for diffusion slightly reduces the temperature
at which the creep rate is significant. Coble creep still requires a T > 0.85 [Cah83] to
become active and is therefore not of concern in this study.
The third type of creep is low temperature or logarithmic creep. In this mecha-
nism, deformation occurs through dislocation multiplication, glide and climb. However,
the sources for dislocation multiplication become exhausted. Therefore the creep rate
starts at some initial higher valve and logarithmically approaches zero as the number of
mobile dislocations approaches zero [Cah83]. The rate of logarithmic creep is not a
function of the applied stress, since the rate limiting steps are dislocation climb or multi-
plication which are driven by random thermal excitations. This type of creep is typically
observed at low temperatures (T, < 0.5).
The fourth creep mechanism shown in figure 2.6 is high temperature or Andrade
creep [Cah83]. In this mechanism dislocation glide, climb and multiply in response to
the applied force with the help of thermal excitation to overcome the higher activation
energies. There are three stages to Andrade creep. In stage one, a stress applied to the
sample at a given temperature causes an initial high strain rate that immediately begins to
decline. The reason for the declining strain rate is that the sample begins to work harden.
After a finite strain, equilibrium is established between the rate of work hardening and the
rate of dynamic annealing. Annealing is a heat treatment that eliminates the effects of
cold work, in dynamic annealing this processes is occurring at the deformation tempera-
ture as described below. Second stage creep is a steady state process that produces a
constant strain rate. In the third stage of Andrade creep the sample begins to neck, grain
boundary sliding occurs [Mey84] and voids form inside the sample. This reduces the area
over which the force is carried thus increasing the stress which increases the strain rate,
ultimately resulting in failure. This type of creep is typically observed for T > 0.6.
Deformation Mechanism Maps (DMMs)
In the proceeding sections, several mechanisms have been discussed which lead to
plastic deformation of metals. H.J. Frost [Fro82] has created deformation mechanism
maps (DMM) relating stress, temperature and deformation mechanism. All DMMs are
based on numerical models describing the different deformation mechanisms. This
model is then used to generate a two dimensional contour plots (temperature and stress
axis) that shows which deformation mechanisms are active. These plots are useful in
selecting materials for a given application. If the stress and temperature levels for a given
application are known, then the deformation mechanisms expected for a given application
can be projected. The primary limitation to the use of DMMs is that detailed data re-
quired by the model are not available for all materials and conditions.
DMMs have been used to explain deformation and stress relaxation in thin film
metallization [Kol86, Fro92, Tho93, She96]. These models have been able to accurately
predict plastic deformation in thin films at high strain rates, but none of these DMMs
were able to accurately predict the rate of isothermal stress relaxation. However, only
bulk material properties are currently available to use in DMMs.
In the preceding sections the mechanisms responsible for plastic deformation at
high and low strain rates were discussed. However, since stress relaxation is of concern
in this study, other mechanisms that transform metals must be considered, such as anneal-
Annealing is the process by which the stored energy in cold worked metal is
released. When a metal is cold worked, that is plastically deformed below the tempera-
ture at which dislocation climb becomes significant, part of the deformation energy is
stored in the metal as defects and lattice distortion [Ree92]. Hundreds of Joules per mole
can be stored in a metal that has been heavily cold worked. The annealing process re-
Figure 2.7: Distribution of dislocations in a bent crystal (a) before polygonization, (b)
after polygonization and subgrain coalescence.
leases this stored energy by reducing the defects, lattice distortion and residual stress. In
the case of a thin metal film deposited on a substrate, the shear stresses between the film
and the substrate may also distort the lattice and induce defects. The process of annealing
should result therefore in stress relaxation.
There are four stages in annealing: recovery, recrystallization, grain growth, and
secondary grain growth [Ree92]. For this study the first three stages are of most interest
since they dominate the release of stored energy from cold work.
In recovery there is little change in the mechanical properties of metal, but the
electrical resistance of the metal decreases. This reduced electrical resistance indicates
that random dislocation tangles have begun to polygonize [Cah49] (see Figure 2.7). The
polygonization of the dislocation tangle into subgrain boundaries reduces the free energy
of the system. Ordering also reduces the number of electron scattering centers, resulting
in a lower electrical resistivity. No dislocations are destroyed, therefore the yield stress
does not change.
Polygonization and subgrain coalescence also lead to the second stage of anneal-
ing, recrystallization. In recrystallization new unstressed crystals nucleate and grow to
replace the strained, disordered cold worked grains. One theory on the formation/nucle-
ation of these new unstrained crystals is that subgrains coalesce to form high angle grain
boundaries leaving an unstrained crystal in their wake [Cah83]. The difference in ener-
gies provides the driving force for the high angle grain boundary to move outward,
consuming the strained crystals until they are all consumed. This is the growth stage of
the annealing process and it is at this point that the energy of cold work has been removed
and therefore stress has been relaxed.
Grain growth also affects the size and distribution of precipitates within a sample
[Bro82]. Precipitates strain the matrix of the material in which they exist. There is
therefore a driving force for precipitates to coalesce and reduce the strain in the matrix.
In age hardened alloys this is called "over aging" as it reduces the yield strength of the
material. This process would also produce a relaxation of stress.
Strengthening of Bulk Metals
In the preceding sections', mechanisms responsible for stress relaxation in thin
metal films have been reviewed. We will now review the methods used to strengthen
bulk metals to determine which could be used to strengthen thin metal films in bimetallic
actuators. There is limited information on strengthening of thin films.
The key to strengthening any non-brittle metal is to inhibit dislocation glide. This
can be done in several ways. First, the matrix of the metal can be affected to increase the
force needed to move a dislocation in a slip system. This can be done a number of ways,
including straining the matrix with cold work or solid solutions. Second, hard particles or
phases can be placed in the matrix of a material that will physically block the movement
of dislocations. There are five commonly identified strengthening mechanisms that use
one or both of these general approaches. Also, many commercial alloys use two or more
alloying elements and several of the five strengthening mechanisms to ensure that several
means of inhibiting dislocation motion are used. The five strengthening methods used in
bulk alloys and the means by which they inhibit dislocation motion are described below.
The first strengthening method is solid solution strengthening, in which soluble
alloying elements are added to the host metal [Cah83]. Since most alloying elements
have an atomic size different from the host element, the matrix of the metal is strained,
and inhibits dislocation glide. The more the host matrix is strained, the harder it is for
dislocations to move. However, alloying elements generally have limited solubility in
host metals which limits the strengthening that can be achieved. For most of the com-
monly used engineering metals, the solubility and strengthening effects of many alloying
elements have been determined experimentally. The Metals Handbooks Volumes 1 and 2
[Bak 97] contain most of this information. For more unusual materials, the Hume-
Rothery [Ree92] rules for substitutional solid solution strengthening are helpful.
Solid solution strengthening has a number of properties that make it unique
compared with the other strengthening mechanisms. First, the solute atoms are in equi-
librium with the host matrix atoms. Thus there are no driving forces to change the atom
distribution in the metal and solid solution strengthened alloys are expected to be more
stable at elevated temperatures than alloys using some of the other strengthening tech-
niques. This is of particular importance for this study since materials will be used at
elevated temperatures. Secondly, a solid solution strengthened material is expected to be
homogenous. Thus there is limited possibility that electrolytic half cells would cause
The second strengthening mechanism is to reduce grain size. Dislocations gener-
ally pileup at the grain boundaries as they cannot cross them. The strength of a material
has been found to be proportional to the inverse of the square root of the average grain
size [Mey84, Ree92]. This relationship is called the Hall-Petch relationship. A potential
limitation to using this strengthening technique is that grains coarsen at elevated tempera-
tures (T, > 0.6), and this generally reduces the strength of the material.
The third strengthening techniques is work hardening. When a metal is cold
worked (i.e. worked below a temperature at which annealing occurs) the dislocation
density increases greatly. With an increased density, dislocation will be pinned by other
dislocations, which limits dislocation slip. Work hardening can produce seven fold
increase in yield strength for pure aluminum. Fully annealed aluminum has a yield
strength of 15-20 MPa that increases to 50-60 MPa for 40% cold work and 100-120 MPa
for 90% cold work [Bak97]. A concern with work hardening is that the material will
soften as dislocations climb at elevated temperatures during recrystallization.
The fourth way to strengthen some materials are by a martensitic transformation
[Kra90]. This type of transformation occurs only in a few alloys and ceramics. It is
widely known and used in ferrous alloys, but does occur in some copper, aluminum and
titanium alloys. The martensitic transformation is a diffusionless non-equilibrium pro-
cess that occurs by a coordinated shear displacement of atoms. This transformation
stresses the matrix of the material which inhibits dislocation motion. The use of marten-
sitic transformations to strengthen alloys is limited because there are few non-ferrous
alloys which exhibit them [Pet70]. In addition, a martensitic transformation leads to a
non-equilibrium state so there is always a driving force to transform to a lower strength
The fifth way to strength a metal is by precipitation or multi-phase hardening.
Alloying elements are added above their solubility limit in a host and precipitates form
out of the solid solution to block dislocation motion [Bro82]. In certain cases, a fine
ceramic powder is added to a molten metal in place of an alloying element to provide the
particles in a modification called dispersion hardening. This leads to improved thermal
stability of the alloy. For the greatest precipitation hardening, a large number of small
precipitates should be uniformly distributed throughout the host matrix. Thermal stability
is a concern at elevated temperatures since precipitates may coarsen (over age) resulting
in a weakened material.
The presence of a second phase can also strengthen a host matrix by straining it.
A second phase with a coherent interface to the host matrix will increase the strain in the
matrix. A coherent interface does not an array of dislocation between the precipitate and
the matrix [Bro90]. Thus, in addition to blocking dislocation motion, a coherent precipi-
tate strains the matrix, increasing the force needed to move a dislocation. Thermal
stability is a concern, as precipitates may grow by thermally activated diffusion and the
coherent interface may be lost. The loss of the coherent interface substantially reduces
the strength of the alloy.
Ordered Intermetallic Compounds
Ordered intermetallic compounds [Cah85] are of interest in this study because
they are noted for high strength at both room and elevated temperatures with good resis-
tance to creep and oxidation [Wes95]. There are several reports on the use of intermetal-
lics at high temperatures in jet engines [Lim97]. The basic structure of intermetallics was
discussed above in the section on thermal expansion. The mechanical properties are
Intermetallics are a class of materials with structures and properties between pure
metals and ceramics [Pop87]. The general composition on an intermetallic is A B where
A and B are metallic elements and x and y are normally small integers. This more com-
plex composition creates a more complex crystallographic structure which in turn make
dislocation movement more difficult. There are several reasons for this. First, the num-
ber of atomic distances a dislocation has to move to reach an equivalent crystallographic
site is much greater. In a simple AB intermetallic a dislocation has to move two atomic
distances as compared with one atomic distance for a non-ordered metal. In many
instances a dislocation will break into pairs of partial dislocations which only move one
atomic distance forming an anti-phase boundary [Bro82]. The creation of an anti-phase
boundary requires energy and increases the critical resolved shear stress for dislocation
glide. In a normal metal a dislocation must break A-A bonds and then reform them to
move. In an intermetallic a dislocation must break the stronger A-B bond to move. Once
the A-B bond is broken the weaker A-A and B-B bonds form (the anti-phase boundary),
which then have to be broken to form the new A-B bond. This process requires more
energy, and thus intermetallics have high yield strengths.
The more complex structure also inhibits creep. The self diffusion rate in inter-
metallics is unusually low [Pop87] due to the large distance between equivalent site, and
a high activation energy for this process. As the creep rate for a material is proportional
to the rate of self diffusion [Cah83], intermetallics have a low creep rate.
Some intermetallics are also resistant to oxidation. Oxidation resistant intermetal-
lics form an adherent passivating oxide such as A12O, on compounds of aluminum,
nickel and titanium.
A concern in using intermetallics in bimetallic actuators is their sensitivity to
composition fluctuations. In A1Ni, as little as 0.5% change in the composition of the
intermetallic can cause dramatic changes in the strength of the metal [Lim96]. Therefore,
the compositional stability of any intermetallic compound should be tested before selec-
Thin films have a number of unique properties versus their bulk counterparts.
Sinse the materials examined in this study are use as thin films in bimetallic actuators, an
understanding of their properties is needed. There are three primary reasons for the
unique properties of thin films. First, thin films are deposited onto a substrate, resulting
in substrate-thin film interactions. Second, the surface area-to-volume ratio for a thin
film is orders of magnitude higher than for a bulk material, so surface energies and
kinetics play a much greater role. Last, the techniques used to produce thin films are very
much different from the techniques used in the processing of bulk metals, which affects
There are at least three common categories of thin film deposition processes:
chemical vapor deposition (CVD), physical vapor deposition (PVD), and electroplating
[Ohr92]. In this study we have only used physical vapor deposition methods consisting
of sputtering or thermal evaporation.
In all physical vapor deposition (PVD) processes, a vapor of the source material is
created and condensed on solid surfaces, including the substrate of interest. All PVD is
done in a high vacuum chamber so that the material vapor can travel from the source to
the substrate without reacting with atmospheric gases. In thermal evaporation, the source
material is simply heated until it vaporizes, a very simple and clean process useful to
deposit single element films. The vapor pressure of different material varies so greatly
that the composition of an alloy can not be maintained with thermal evaporation [Ohr92].
In the sputter deposition processes, an inert gas, normally argon, is ionized and ions are
accelerated into the source material or target. When the inert gas ion strikes the target
atoms, its momentum is transferred and by development of a momentum cascade, some
of the target atoms are ejected from the surface. The sputter process is based on momen-
tum transfer, therefore the vapor pressure of the target material does not control the rate
at which atoms are ejected from the target surface. Sputtering can therefore be used to
deposit metal alloys containing a number of elements [Ohr92]. However, differences in
composition may be seen between the deposited film and source material [Zhe97].
After being sputtered or sublimed, the vaporized atoms condense and agglomerate
to form a film [Sor95]. The morphology of the developing film is primarily determined
by the mobility of the condensing atoms [Mac95]. For thermal evaporation, the condens-
Figure 2.8: Diagram of Movchen and Demishin's model of thin film morphology
ing atoms have little kinetic energy and the surface mobility is primarily controlled by the
substrate temperature. Based on this, Movchen and Demishin [Kra90, Thor74] devel-
oped a model that predicts film morphology based on the homologous temperature of the
substrate (Figure 2.8). In the Movchen and Demishin model, the film morphology is
divided into three zones. In zone 1 (TH < 0.3) the adatom mobility is so low that tapered
crystals form with voids between the crystals. In zone 2 (0.3 < TH < 0.5) the adatom
mobility is great enough that most of the voids fill in and columnar grains form. In zone
3 (TH > 0.5) the mobility is high enough that nearly equiaxed crystals form.
In the sputtering process, the kinetic energy of the condensing atoms can be
controlled by the gas pressure and by biasing the substrate [Win92]. The effects of
substrate temperature and gas pressure have been studied by J.A. Thomton [Tho74], who
developed a model to predict the morphology of sputtered deposited films (see Figure
2.9). Thornton did not divide the sputtered morphology into zones as Movchen and
Demishin did. There is some correlation between these two models as Krauss illustrated
in figure 2.9 [Kra90]. It should also be noted that post deposition annealing of a thin film
may cause the microstructure to develop similar to that which would have developed had
the film been grown at that same temperature [Mac95, Kno91].
Figure 2.9: Thornton's model of sputtered thin film morphology with a comparison
to Movchen and Demishin's model of thin film morphology [Kra90].
During the deposition process and post deposition annealing, thin films can
develop a crystallographic texture [Mac95, Kno95]. There are two competing processes
responsible for the development of the thin film texture. In the growth of a columnar
microstructure, even from a melt, the crystallographic planes with the fastest growth
cause the elongated axis of the column [Bro94]. In thin films there is also a large driving
force to reduce surface energy. So the crystallographic planes that have the lowest sur-
face energy have a thermodynamic advantage over faster growing planes with higher
surface energies. The final texture that develops in a thin film is the balance of these two
There are two types of stresses that develop in thin films: intrinsic and extrinsic.
Intrinsic stresses are inherent to the deposition process and conditions, while extrinsic
stresses are due to the difference in the thermal expansion of the film and substrate.
Extrinsic stresses cause actuation of the bimetallic element.
Intrinsic stresses develop during the deposition process due to limited adatom
mobility [Mac95]. This limited mobility causes the formation of voids and vacancies in
the film. Voids tend to collapse and form grain boundaries. Vacancies migrate to grain
boundaries and are destroyed. This causes the film to attempt to contract, but instead it
may develop a state of biaxial tensile stress because it is attached to a rigid substrate. In
sputter deposition it is possible to increase the kinetic energy of the sputtered atoms, by
reducing the sputter gas pressure or biasing the substrate. Under bias, ions may penetrate
the first few layers of the deposited film. This "atomic peening" can reduce the intrinsic
tensile stress and even produce a compressive stress [Mac95]. However, atomic peening
injects a large number of trapped vacancies and implants gas [Mac95]. If the film is
heated to a temperature where vacancies or gasses are mobile, they will migrate to the
grain boundaries, be destroyed or evolved, and change the stress back to tension [Tow87].
In MEMS it is often desirable to have a residual stress near zero at room tempera-
ture. This allows for greater flexibility of design. While it is possible to achieve low
stress by adjusting the sputter parameters, it is sometimes possible to reduce tensile stress
by thermal cycling. If the thin film and substrate are thermally cycled to -196 C (liquid
nitrogen temperature) the metal film will normally contract more than the substrate. If
the film is ductile, it will plastically deform and upon returning to room temperature, the
film will experience a lower tensile stress or even a compressive stress [Bal94]. Unfortu-
nately, elevated temperature cycling of the film will counteract this compressive shift in
stress and re-introduce a tensile stress [Bal94].
A primary concern of this study is stress relaxation in the thin film, defined above
to be a time dependent change in the state of intrinsic plus extrinsic stress. Stress relax-
ation changes the force or displacement of a bimetallic actuator, and may be caused by
plastic deformation, creep, or diffusional processes including precipitation. The process
of plastic deformation and dislocation movement in thin films is restrictive as compared
with bulk metals. There are several reasons; first, a thin film is under a biaxial state of
stress. Thus dislocations can only glide in slip systems that have a component in the z
direction, normal to the film surface, since only these planes have a non-zero resolved
shear stress [Nix89]. If the thin film is textured there will even fewer slip systems with a
...U ----.-- --------------- --- ---- ---- --.-- --- --- -.-. ..... .....-- -
i} ^ i
0 50 100 150 200 250 300 350
Figure 2.10: Stress temperature profile for pure sputtered aluminum and
microelectronic alloy Al-Si-Cu [Jer95].
resolved shear stress greater than the critical value. Second, a dislocation moving along a
slip system must nucleate misfit dislocations at the film/substrate interface, which re-
quires additional energy. If a passivating oxide forms on the surface of the film, as in the
case of aluminum, misfit dislocations must be created at both interfaces [Nix89], further
reducing dislocation glide.
The primary method used to test the mechanical properties of thin films is mea-
surement of stress as a function of temperature. As the thin film expands at a different
rate than the substrate, changing the temperature changes the stress in the film. This is an
easy and non-destructive way to test a thin film, but it is limited because the stress and
temperature cannot be varied independently. Instruments that measure stress in this way
include the Tencor Flexus which was used in this study (see description in Chapter 3).
The result of this type of testing is a plot of stress as a function of temperature, shown in
Figure 2.10, for sputtered pure Al and Al-1.5%Si-2%Cu. To evaluate stress relaxation
the sample can be heated to a temperature and the stress plotted as a function of time.
For bimetallic actuators, a film that shows no hysteresis in stress versus tempera-
ture over the range used is desired. Hysteresis in this plot indicates that the film is plasti-
cally deforming. Figure 2.10 shows that the metals currently being used to manufacture
bimetallic actuators [Jer96], both exhibit significant hysteresis over the temperature range
from 20C to 300-3500C. The pure aluminum shows little strength over 120C, while the
Al-Si-Cu alloy is only slightly stronger and plastically deforms above 170C under a
compressive stress of 50 MPa. The room temperature residual tensile stress for pure
aluminum is 150 MPa, while the Al-Si-Cu is 200 MPa. Isothermal stress relaxation tests
were also performed on pure aluminum and Al-Si-Cu alloy thin films [Jer96], with both
films showing a 15% reduction in stress over 1000 minutes. At 50'C the pure aluminum
film was initially under a tensile stress of 75 MPa, while the Al-Si-Cu alloy film was
initially under a higher tensile stress of 150 MPa.
There have been only a few studies to examine isothermal, long term stress
relaxation in thin films. The first was performed in 1985 by Hershkovitz, Blech and
Komem. They identified three modes of stress relaxation in thin aluminum films. The
first mode was identified as dislocation glide. The mechanism responsible for the other
two modes were not determined. In 1991, Drapper and Hill studied stress relaxation in
Al-Si-Cu thin films. They concluded that logarithmic creep was the dominant mecha-
nism responsible for stress relaxation. However, Drapper and Hill assumed a single
mechanism was responsible for the stress relaxation and fit there data with a single
exponential function, ignoring the conclusions of Hershkovitz, Blech and Komem. The
single figure of the data and curve fit in their article showed great inaccuracies between
the data and the fit. In 1995 Witvrouw, Proost, Beweerdt, Roussel, and Imec also re-
ported isothermal stress relaxation data from highly tensile stressed Al-Si-Cu thin films.
At low temperature (70C) they attribute stress relaxation to dislocation glide. At higher
temperatures (120C to 140C) they propose that the dislocations are cutting the Al2Cu
precipitates. From these limited studies, no complete model has been developed to
explain isothermal stress relaxation in thin films.
A new mechanism proposed by J. Tersoff for the stress relaxation in epitaxial
films is surface roughing [Ter94]. In this new mechanism the surface of the epitaxial
layer becomes rough, allowing easy nucleation of dislocations. This mechanism has not
yet been tested in polycrystalline thin films.
While strengthening of thin films to reduce stress relaxation and plastic deforma-
tion has not been extensively studied, work to reduce failures caused by thermal cycling
has been reported. Much of this work was done to improve the reliability of Josephson
superconducting devices (SQUID Superconducting Quantum Interference Device)
[Kir80], which were of interest for building ultrahigh-speed computers. These devices
operate at temperatures below 10K (-263C), and need to be able to withstand tempera-
ture cycles form 200C to -263C. Basavaiah and Greiner found the addition of gold to the
lead indium alloy reduced failures, but the mechanisms responsible for the reduction in
failures was not determined [Bas77].
Environmental Stability of Materials
Bimetallic actuators may operate in atmospheric gases at elevated temperatures,
and the materials of construction must be stable. Environmental instabilities of these
films would compromise their physical integrity and cause the device to fail. There are
two environment reactions that need to be considered for this application: oxidation and
Most metals oxidize, and some self-passivate while others require some protective
coatings. This study has focused on metals which self passivate, such as aluminum,
copper, titanium, nickel. For a passivating oxide to be effective it should have the follow-
ing four properties: high thermodynamic and kinetic stability, slow growth rate, adherence
to the metal, and easily form or re-form [Wes95]. The growth rate is of greater impor-
tance in thin films as an oxide could consume the entire film. Most passivating oxides
generally follow a parabolic growth rate equation given by
x = kt (4)
where x is the thickness of the oxide, k is the parabolic-growth-rate constant, and t is
time. Aluminum typically exhibits an inverse logarithmic dependence at low tempera-
tures and parabolic at high temperatures. Thus for thin film applications, materials with a
low parabolic-growth rate constant are needed, such as aluminum and silicon, for el-
evated temperature applications. In addition, the formation of an oxide and or a metal-
oxygen solid solution will introduce some stress into the metal substrate. This change in
stress could affect the force or displacement of a bimetallic strip. This change may be
time dependent as can be seem from equation 4.
Corrosion is also a concern because these devices may be operated in moist air
where the possibility of a galvanic cell exists. One area where a cell could exist is be-
tween the active and passive layers of the bimetallic strip. In this application, there is
normally an oxide layer between these two layers of the bimetallic strip that electrically
isolates them and so a galvanic cell cannot be formed. Disruption of this oxide would
cause this to be a concern. Also, formation of micro-galvanic cells in a two phase alloy,
should be tested [Ask89].
Eleven different materials were deposited as thin films and analyzed in this study
(see Table 3.1). The deposition methods and procedures will be covered in this sections.
The films were first tested to determine their mechanical properties and the best materials
were analyzed further to determine their microstructural and chemical compositions. A
number of analysis techniques were used as discussed below.
Table 3.1: Materials examined in this study.
Material Deposited Thickness Composition or Purity, (wt. %)
T201 Alminum 1.3 A -Al 4.6Cu 0.57Ag 0.36Mn -
T201 Aliminum 1.3 0.2Mg-0.27T.
5052 Aluminum 1.3 Al 2.5Mg 0.25Cr
2090 Aluminum 1.3 Al 2.57Cu 2.1Li -0.12Zr
Nickel 0.8 99.98
Titanium 0.55 99.99
Manganese 0.4 99.9
Copper 0.9 99.9
AI,Ti 0.4 Al-99.99. Ti-99.99
CuAu 1.3 Cu-99.99 Au-99.99
CuAu 1.3 Cu-99.99 Au-99.99
CuAu, 1.3 Cu-99.99 Au-99.99
Thin Film Deposition
The thin films examined in this study were deposited by either sputtering or
electron beam evaporation. All the films were deposited on 100mm diameter, 
oriented, single crystal silicon wafers. The thickness of each film is shown in table 3.1.
The wafers used for the aluminum alloy and copper gold intermetallics had a 100 nm
thermal oxide grown on them before film deposition. All wafers were cleaned using the
following procedure: 5 minute in an ultrasonic bath for each solvent with a de-ionized
water rinse between solvent for, trichroloethane, acetone, and methanol, then five minutes
in a solution of 75% sulfuric acid and 25% hydrogen peroxide-30% followed by a de-
ionized water rinse and blown dry with dry nitrogen.
The aluminum alloy and copper gold intermetallic films were sputter deposited by
Sputtered Thin Films, Inc., Santa Clara, CA, using an 8-inch DC magnetron sputter gun
running at 5kW of power. The chamber was first pumped down to 2.8x10-7 Torr, then
back filled with argon to a pressure of 8 mTorr. The substrates were heated to 1000C.
The aluminum alloy sputter targets were from bulk sheet stock. Copper gold intermetal-
lics were produced by depositing four layers of pure copper and gold in the proper pro-
portion to produce the required composition. Table 3.2 shows the thickness of the indi-
vidual copper and gold films that were deposited for each of the average compositions.
Table 3.2: Layer thickness deposited for the copper/gold intermetallic films.
C d Copper Layer Gold Layer
Thickness, gm Thickness,4m
Cu3Au 0.1132 0.4868
CuAu 0.2465 0.3533
CuAu, 0.4059 0.1941
The copper, titanium, nickel, manganese and aluminum-titanium intermetallic
films were deposited by electron beam evaporation. The evaporation system contained
three Telemark Model 211 single pocket electron-guns in the vacuum chamber. These
guns are powered and controlled by a Sloan PAK 12 series 12 kW power supply operated
at 10 kV. The deposition rate and total deposited thickness were measured with quartz
crystal monitors, either an Inficon XTC or Sycon model 100. The deposition rate was
manually controlled and varied between 10 to 15 angstroms per second. The vacuum
chamber was pumped down to 5xl0- Torr before each deposition. Due to outgassing, the
chamber pressure would rise to 5x 10 6Torr during deposition. The substrates were heated
to 1500C +/- 100C with resistive heaters, manually controlled with a variable transformer.
Copper Gold Heat Treatment
The copper gold compounds required a special heat treatment to homogenize the
layers and to develop the long range ordered intermetallic structure [Fed58]. These heat
treatments were performed in a vacuum of 5x 10-6 Torr or better to inhibit oxidation of the
copper. The heat treatments were 12 hours at 500C to homogenize the film, and 12
hours at 2750C followed by 12 hours at 225C to develop the long range order. All
temperatures changes were ramped at I1C per minute using an Omega CN3000 controller.
Stress and Stress Relaxation Measurements
Stress and stress relaxation in the samples were measured with a Tencor Flexus
model 2320. The Flexus 2320 uses an optical lever to measure the curvature of the
sample [Tur96]. Stoney's equation [Sto09], Equation 4, was used to calculate the average
biaxial stress, a, in the film:
where R is the curvature of the substrate, t is the film thickness, h is the substrate thick-
ness, E is the Young's modulus for the substrate, and v is Poisson's ratio for the substrate.
The silicon substrate's thickness and curvature were measured before the film was
deposited. The curvature of the substrate was then measured after the film was deposited,
and the change in curvature is used in Stoney's equation to calculate the film stress. In
addition to measuring stress at room temperature, the Flexus could heat a sample to
500C or cool it down to -60C under computer control and measure curvature. The
Flexus could also temperature cycle a sample, taking measurements during the cycle to
generate a stress versus temperature plot (see Figure 2.10).
The isothermal stress relaxation in aluminum alloy thin films was also measured
with the Flexus. To obtain a similar starting condition for different tests, each sample
was heated to 3500C at PlC/minute, held at 350C for 30 minutes, and cooled back down
to room temperature at I C/minute. The sample was then taken to the test temperature, at
lC/minute, and held at the test temperature for forty eight hours. Every 15 minutes the
stress in the sample was measured, allowing stress to be plotted versus time (see Figure
3.1). Once the mechanical properties of a film had been measured with the Flexus, the
samples was divided for chemical and microstructural evaluation.
Powder X-Ray Diffraction (XRD)
Powder x-ray diffraction was used to determine the long range ordering in the
copper gold films [War69] and to evaluate crystallite size in the aluminum alloy films
[Cul78]. The system used was a Philips APD 3720 x-ray diffraction systems controlled
with an IBM-PC type computer running Microsoft windows 3.11 with Philips PW1877
Automated Powder Diffraction software, version 3.6g. X-rays from a copper anode
operated at 40 kV and 2 milliamps with a nickel filter were used.
Figure 3.1: Isothermal stress relaxation in 5052 aluminum alloy sample at 125C.
X-ray diffraction is effective for detecting the change in structure that occurs in an
order-disorder transformation. For the copper gold system, the disordered crystallo-
graphic structure is FCC, with atoms sitting randomly on lattice sites. In an ordered
phase, one type of atom, either copper or gold, sits in the corner sites and the other atom
will sit in the face sites. This creates two inter-linked simple cubic or simple tetragonal
crystallographic structure. This change in the crystal structure causes the original diffrac-
tion peaks to change, and the different structure factor causes a number of new peaks to
appear [Cu178]. With a true randomly oriented sample, it is possible to calculate the
degree of ordering in the sample by comparing different peak heights [War69]. However,
thin films have a crystallographic texture that complicates this procedure. As a result it
was difficult to quantify the degree of order in the films.
For aluminum alloys the dislocation structure inside the aluminum crystals, the
grain size and the strain were all evaluated using x-ray diffraction line shape analysis. In
80 -----------1 ----- 1 ----- 1 ------ 1 ------
0 S00 000 1500 2000 2500 3000
Figure 3.1: Isothermal stress relaxation in 5052 aluminum alloy sample at 1251C.
X-ray diffraction is effective for detecting the change in structure that occurs in an
order-disorder transformation. For the copper gold system, the disordered crystallo-
graphic structure is FCC, with atoms sitting randomly on lattice sites. In an ordered
phase, one type of atom, either copper or gold, sits in the comer sites and the other atom
will sit in the face sites. This creates two inter-linked simple cubic or simple tetragonal
crystallographic structure. This change in the crystal structure causes the original diffrac-
tion peaks to change, and the different structure factor causes a number of new peaks to
appear [Cul78]. With a true randomly oriented sample, it is possible to calculate the
degree of ordering in the sample by comparing different peak heights [War69]. However,
thin films have a crystallographic texture that complicates this procedure. As a result it
was difficult to quantify the degree of order in the films.
For aluminum alloys the dislocation structure inside the aluminum crystals, the
grain size and the strain were all evaluated using x-ray diffraction line shape analysis. In
this technique, the (111) and (220) diffraction peaks were selected for evaluation. The
(111) peak was scanned from 28 = 380 to 39.20, and the (220) peak was scanned from 20
= 64.240 to 66, using 0.02 increments and 10 second dwell time. For line shape analysis,
an unstressed quartz standard was scanned in the same diffractometer under the same
conditions. The quartz standard was scanned from 20 = 38.70 to 39.80, and from 20 =
63.50 to 64.7.
Line shape analysis extracts the factors that cause the x-ray diffraction peaks to
shift and/or broaden. There are three main causes of peak broadening in x-ray diffraction.
The first is the diffractometer itself. Instrumental broadening is eliminated by compari-
son to the quartz standard. The other two causes of peak broadening are strain and
crystallite size. Each of these produce a different shape. Strain effects produce a
Gaussian distribution while the crystallite size produce a Cauchy distribution [War69].
The Phillips diffraction software can deconvolute these effects and determine the strain
and crystallite size.
Atomic Force Microscopy (AFM)
A Digital Instruments Nanoscope III atomic force microscope (AFM) was used to
determine the surface topology of the aluminum alloy samples [Hud92]. A silicon tip in
the tapping mode was used to probe the surface of the samples. Digital Instruments
software was used process the scanned images and calculate the RMS surface roughness.
Scanning Electron Microscopy (SEM)
A JEOL 6400 scanning electron microscope was also used to image the surface of
samples. In addition to the secondary and back scattered electron detectors, this system
has an energy dispersive spectrometer (EDS) that allowed qualitative elemental analysis
of surface features [God92]. Samples were first viewed in the secondary electron mode
to evaluate the surface topology. Very bright areas were viewed in the back scattered
electron mode. The true secondary electron yield from an area is more strongly depen-
dent on the topology, whereas the back scattered electron intensity is dominated by the
atomic mass of the excited atoms. Areas that appeared to have different atomic composi-
tions were examined with EDS to determine the concentration of elements.
Electron-Probe Microanalysis (EPMA)
EPMA was used to quantitatively determine the compositions of the aluminum
alloys and the copper gold intermetallic films [Gol92]. A JEOL 733 Superprobe with
four wavelength despersive spectrometers was used. For the aluminum alloy films the
bulk aluminum alloy was used as the analysis standard to determine difference in compo-
sition between the bulk and thin film composition. For the copper gold intermetallic
films, pure copper and gold standards were used with the ZAF (atomic number effects, X-
ray absorption affects, and X-ray fluoresces effects) analysis technique to determine the
composition of these films [Gol92].
Transmision Electron Microscopy (TEM)
A JEOL 200CX analytical transmission electron microscope was used to examine
precipitates and the crystal size of the aluminum alloys. The samples were prepared by
first cutting a 3mm disk from a silicon substrate with a deposited film. These samples
were then thinned from the silicon side to between 150tm and 250.m. Using a dimpling
grinder, the back sides of the samples were hollowed out so that the center of the samples
were approximately 50glm thick. The thinned samples were encapsulated in paraffin and
a small opening (1mm) was made in the paraffin over the dimple. The samples were
etched in a solution of 50% HF and 50% HNO,. The etching process was monitored by
viewing the samples through a low power binocular microscope. Once the acid had
etched through the silicon and reached the aluminum film, etching was stopped. These
samples were ion milled until a small opening appeared in the aluminum, then viewed in
In the TEM, the samples were first viewed in the bright field mode to evaluate the
general structure of the sample [Lor94]. The dark field mode was used to better identify
different phases. Electron diffraction images and patterns were also taken to identify the
phases present by calculating the interplaner spacing. The inter plane spacing, dhkl was
where L is the camera length (for the JEOL 200CX L = 82 cm), X is the electron
wavelength which is based on the accelerating voltage (X = 0.0251nm based on an accel-
erating voltage of 200keV) and R is the radius of the diffraction pattern.
Grain size is an important parameter when evaluating the strength of a metal.
Etching is the primary means by which the grain size may be made visible in bulk metals,
but this technique was found to be ineffective in this study. Etching thin films to define
grains is problematic. Most etchants attach different crystallographic planes at different
rates. In a material with a randomly oriented crystals, this type of etchant works well.
However, thin films are strongly textured and tend to etch more uniformly, complicating
the determination of crystal size. Some etchants that attack grain boundaries preferen-
tially [Smi67, Gif70] and two were tested for aluminum alloys: 1% hydrofluoric acid and
a 10% sodium hydroxide solution. Satisfactory definition of the grain boundaries was not
Auger Electron Spectroscopy (AES)
The Perkin-Elmer PHI 660 Scanning Auger Microprobe was used to evaluate the
oxide thickness on the aluminum alloy films. Auger analysis is very surface sensitive,
containing information primarily from the top three or four atomic layers [Hol80]. The
Auger system contained an argon sputter gun that enables depth profiles to be collected.
To do so, an Auger spectrum was taken, the sample sputtered for 15 seconds, and another
Auger spectrum taken. This process was repeated to the required depth. The aluminum
alloy samples were sputtered until the oxygen signal had decreased to 10% of its original
value. This technique gives a thickness in terms of sputter time. The conditions used
result in an estimated sputter rate of 250 angstroms per minute These data are primarily
used to compare differences in oxide thickness from sample to sample.
To curve fit the stress relaxation data, curve fitting techniques were evaluated:
polynomial, single exponential, and double exponential. The double exponential tech-
nique fit the data using the equation:
a(t)=A+Be"' + Ce"2 (5)
where a is stress, t is time, and A, B, C, m, and m2 are constants that are determined from
the data. This technique fit most of the data with the least error.
The process for determining the constant was the linearization of the data. This
was sometimes complex as the two exponential functions needed to be separated to be
linearized. Separation is accomplished by first fitting the stress relaxation data for t > 380
minute. To linearize these data, the value of A in equation 5 was first manually selected
and subtracted from the remaining data. The natural log for the remaining data was then
taken and least squares linear fit to determine the slope, intercept and correlation factor.
The A constant was iterated to obtain the best possible correlation factor. With this step
completed, the first exponential was subtracted from the original data. The natural log of
the difference was then taken and the resulting data again fitted with a linear least square
fit to determine the slope and intercept, m and C.. All calculations were done in
MathCad Version 4.0.
The effects of oxide thickness were found to be significant, therefore a numerical
model was developed to quantify this effect. This model was based on the equation
developed by Townsend [Tow87] and implemented in MathCad version 4.0. In this
analysis a three layer structure was modeled. The first layer was a 5tm thick layer of
silicon, and the second layer was 5ltm of aluminum. The third layer of Al203 was varied
from 0 to Inm in thickness. This model was used to calculate the curvature of the struc-
ture, therefore the length was not required. The output was plotted as a percentage
change in curvature for the structure with varying oxide thickness.
The results of the testing and analysis will be presented in this chapter which is
organized by material to provide an understanding of their performance.
Figure 4.1 shows the stress versus temperature plot for the T201 alloy film and
may be compared to the literature data shown in figure 2.10. These data were collected
after an initial temperature cycle to 3500C to stabilize the film's microstructure after
deposition. While the intrinsic stress in these films were a function of deposition param-
eters, the first cycle to 350C annealed and consolidated the films and the extrinsic
stresses were so high that they dominated the stresses reported as in Figure 4.1. As can
be seen from these data, the room temperature residual tensile stress is almost 400 MPa
and the film begins to plastically deform at approximately 2000C under a compressive
stress of 50 MPa. This sample was also cooled to -1960C in an attempt to reduce its room
temperature residual stress, but this procedure had no effect on the room temperature
The isothermal stress relaxation seen in these samples at different temperatures
are shown in figures 4.2 through 4.6. These data were curve fit as describe above and the
equation is shown in each plot. Below 1251C the stress relaxation data was fit well by a
double exponential decaying function that suggesting that are two stress relaxation pro-
cesses were occurring. At 1251C and 150C the noise on the data are to large to allow
fitting. Rather than heating at a rate of lbC/minute samples were also heated rapidly
300 ------------... ...- ----- -------,----- ------------------------
( 200 ------ --- ---- ---- --------- ------------
(i 100 v -- S .- ^ -
0 50 100 150 200 250 300 350
Figure 4.1: Stress versus temperature plot of T201 aluminum thin film.
390 I -
o(t) = 330.2 + 48.9e -5.45o1 + 9.2e -.096"1
Figure 4.2: Isothermal stress relaxation of T201 aluminum thin film at 50C.
Figure 4.3: Isothermal stress relaxation of T201 aluminum thin film at 75C.
Figure 4.4: Isothermal stress relaxation of T201 aluminum thin film at 100oC.
Figure 4.5: Isothermal stress relaxation of T201 aluminum thin film at 125C.
46 1--------1----------------1------- 1 -----
0 0 %0 .0
I 00^ 0 *' ^<
36 I I I I
0 500 I000 1500 2000 2500
Figure 4.6: Isothermal stress relaxation of T201 aluminum thin film at 1500C.
(10C/min) to the testing temperature to evaluate the effect of ramping rate on stress
relaxation. The stress relaxation was little changed, showing that ramping rate had little
affect on the stress relaxation seen in the samples. A compilation of the curve fitting
parameters is shown in table 4.1.
In addition to the mechanical evaluation of the T201 thin films, microstructure
and chemical composition was evaluated. Table 4.2 shows the results of the EPMA
analysis of the thin films and the composition of the bulk alloy, as determined by an
external lab. There are systematic differences in composition between the sputtered thin
film and the bulk alloy presumably due to the sputter deposition process [Zhe97].
Figure 4.7 is a SEM micrograph of a T201 thin film sample that was heat treated
for 60 hours at 100C. Back scatter images of the same regions showed that the bright
areas had a different atomic composition than the areas around them. Figure 4.8 is the
EDS spectrum for the dark areas showing only aluminum. Figure 4.9 shows the EDS
spectrum for the bright areas and both copper and aluminum are detected while silicon is
not. Therefore, the bright areas on the SEM image are due to both topographic features
and differences in chemical composition of the sample, presumably due to precipitates.
Table 4.1: Residual stress after infinite relaxation time, o, change in stress at short (AG)
and long (Aoa) times, and the time constants (t) for the two exponential equations
for T201 aluminum.
Temp. Ao T" AzC
(C) (MPa) (MPa) (min) (MPa) (min)
50 330 9.2 196 49 1820
75 240 13.8 220 26 1940
100 160 608 220 29 1400
125 33 0 0 81 960
150 40 -3 200 0 0
Table 4.2: Chemical composition of sputter deposited T201 aluminum alloy thin
films determined by EPMA, and compared with the bulk alloy composition.
Alloying Bulk Alloy Sputtered Thin Film
Element Weight Percent Weight Percent
Al 93.91 93.19+/-2
Ag 0.57 0.48 +/- 0.04
Cu 4.6 4.59 +/- 0.2
Ti 0.27 0.53 +/- 0.04
Mg 0.29 0.34+/- 0.1
Mn 0.36 0.87+/- 0.1
Figure 4.7: SEM micrograph of T201 aluminum sample after 60 hours at 100"C.
1 2 3 4 5 6 7 8 9 10
Figure 4.8: EDS spectra of dark area of SEM image shown in figure 4.7,
showing a lack of copper and silicon.
1 2 3 4 5 6 7 8 9 10
Figure 4.9: EDS spectra for bright areas in SEM image, figure 4.7, consistent with
a precipitate rich in copper.
Atomic force microscopy (AFM) was used to evaluate the surface morphology of
the samples, with an image shown in Figure 4.10. Mounds or hillocks about 0.6p1m high
were present on the sample surface. The AFM was also used to evaluate the surface
roughness of these films after different heat treatments. Due to the large mounds that
form on the surface, the RMS roughness varied greatly from position to position on the
same sample. An effort was made to scan areas unaffected by the mounds, but this
required very subjective selection and exclusions of areas of the sample. Thus, the RMS
roughness data were not reliable and are not reported.
Figure 4.10: AFM amplitude image of T201 aluminum thin film after stress relaxation at
125C for 2 hours. The scanned area is 15 lm by 15 ltm.
The primary strengthening techniques used in the T201 alloy are solid solution
strengthening (due to alloying elements of magnesium and manganese [Dav90, Bro82])
and precipitation hardening (due to alloying element copper). Silver is added to this alloy
(0.57 % weight percent) to stabilize the copper precipitates [Dav90]. The effectiveness of
precipitation hardening is controlled by the size and distribution of the precipitates in the
host crystals [Cah83]. TEM was used to evaluate the formation of precipitates in the
aluminum crystals. Figure 4.11 a shows a plan view TEM image of a T201 sample. The
small (10 nm to 30 nm) dark spots in this image are the CuA12 precipitates in the alumi-
num matrix. Figure 4.1 lb shows the indexed electron diffraction pattern from this sample
with the aluminum diffraction rings and the CuAl, diffraction spots identified.
The bright field TEM images also give some indication as to the size of the
aluminum crystals. Caution must be used in evaluating grain size from this image be-
cause the etching process used in the sample preparation would likely attack small crys-
Figure 4.11: TEM micrographs of T201 thin film sample. (a) bright field image showing
aluminum crystals and CuAI2 precipitates; (b) indexed electron diffraction pattern show-
ing the diffraction rings for the Al (111) and (200) planes and for the AICu2 (422) plane.
Percipiijiel f C(uAI.
tals faster than large crystals. Also, a TEM image only examines a small area of the
sample, and a large area must be viewed to determine the true grain size distribution. The
apparent average crystal sizes was between 1 and 2 R.m.
While direct measurement of the aluminum grain size by etching was unsuccess-
ful, formation of the copper rich precipitates gives some indication of the aluminum grain
size. Precipitates tend to nucleate and grow at triple points where three crystals come
together [Ree92]. While the direct calculation of grain size from the precipitation density
is questionable, the comparison of different samples and inferring changes in crystal size
based on changes in precipitation density is reasonable. Figure 4.12 shows a SEM micro-
graph of a T201 sample that was heated to 350C and then cooled to 25C over 48 hours
to test the stability of the microstructure. The number of copper rich particles in figure
4.12 are of the same order as that of figure 4.7, i.e. 150 +/- 10 particles in the 20 by 10
gim area of the micrograph. Therefore, the precipitate size and density of these samples
is apparently stable over several temperature cycles to 350C.
Figure 4.12: SEM image of T201 aluminum sample after temperature
cycling from 3500C to 25oC over 48 hours.
0 I 0
2S A] Al
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5,0
SPUTTER TIME, min.
Figure 4.13: Auger depth profile of T201 thin film. The film was sputtered
with 3 keV argon in a 3x3 mm raster.
The effect of oxidation on a bimetallic actuator's performance is a concern. AES
analysis was used to measure the oxide thickness on a T201 thin film sample that had
been cycled to 3500C 5 times. Figure 4.13 shows the Auger depth profile. The rastered
ion gun used in this analysis sputtered A120, at an approximate rate of 25 nm per minute.
Thus the oxide thickness on this sample is approximately 25 to 30 nm.
Powder x-ray diffraction analysis was also performed on the T201 thin film
samples. A broad 20 scan was taken of a sample cycled to 3500C once to determine the
phases present in the sample. Only elemental aluminum and silicon from the substrate
were detected in these scans. XRD scans were then run on samples that were heat treated
at 50C or 1251C for up to 32 hours. The samples were scanned for their (111) and (220)
peaks. Figures 4.14 and 4.15 show the diffraction peaks after different stress relaxation
times at 1250C. Note that there is a substantial shift in the peak position between the
initial scan and those that have undergone stress relaxation for the shortest time of two
40K 2 Hours
S Hou s
38.0 38.2 38.4 38.6 388 39.0 39.2
Figure 4.14: Aluminum ( 11) x-ray diffraction peaks for T201 thin film samples
after various stress relaxation times at 1250C.
2 HOURS 16 s
65.6 65.8 66.0 66.2 66.4 66.6
Figure 4.15: Aluminum (220) x-ray diffraction peaks for T201 thin film samples
after various stress relaxation times at 125C.
hours. Line shape analysis was performed on the X-ray data to determine crystallite size
as a function of stress relaxation, but the crystallite sizes varied randomly and no correla-
tion was found.
Figure 4.16 shows the stress versus temperature plot for the 5052 aluminum alloy
thin film. These data were collected after an initial temperature cycle to 3500C to stabi-
lize the film's microstructure. As can be seen from these data, the room temperature
residual tensile stress in the film is 300 MPa versus 400 MPa for T201, 200 MPa for Al-
Si-Cu, and 150 MPa for pure AL. The film begins to plastically deform under a compres-
sive stress of 25 MPa at 2001C. A 5052 sample was also cooled to -1960C in liquid
nitrogen to reduce the room temperature residual stress. Upon warming to room tempera-
200 --- ------ -----------
Fe 4.16: St------------- ------ress versus temperate pt fr 52
0 50 100 150 200 250 300 350
Figure 4.16: Stress versus temperature plot for 5052 aluminum.
ture, the tensile stress was reduced from 300 MPa to 125 MPa, a 58% reduction. Starting
at the reduced room temperature stress of 125 MPa the film begins to plastically deform
at a lower temperature of 125oC under a compressive stress of 25 MPa. Upon cooling
from 350C to room temperature, the tensile stress returned to 400 MPa.
The isothermal stress relaxation seen in these films at different temperatures is
shown in figures 4.17 through 4.21, along with the fitted curve and the equation. At 50C
the stress relaxation was best fit with a single exponential curve. For temperatures
between 75C and 125C the stress relaxation data was fit well by a double exponential
curve. At 150C the noise in the stress relaxation data is so large that the fitted curve is
unreliable. Samples were also heated at 10C/minute to the testing temperature to evalu-
ate the effect of ramping rate on stress relaxation. Again, this had little affect on stress
relaxation. The curve fitting parameters are compiled in table 4.3.
Figure 4.17: Isothermal stress relaxation in 5052 aluminum thin film at 50oC.
Figure 4.18: Isothermal stress relaxation in 5052 aluminum thin film at 75C
S1000 1500 2000 2500
Figure 4.19: Isothermal stress relaxation in 5052 aluminum thin film at 100C
85 wI I
0 500 1000 1500 2000 2500
Figure 4.20: Isothermal stress relaxation in 5052 aluminum thin film at 1250C.
u U00 1000 1500 2000 2500
Figure 4.21: Isothermal stress relaxation in 5052 aluminum thin film at 1500C.
Table 4.3: Residual stress after infinite relaxation time, i. change in stress at short (Ao)
and long (A() times, and the time constants (7) for the two exponential equations,
for 5052 aluminum.
Temp. o. Ao ACa, "
("C) (MPa) (MPa) (min) (MPa) (min)
50 240 0 0 30 2800
75 150 12 200 33 1500
100 110 19 170 29 1380
125 80 10 160 17 220
150 0 0 0 58 50000
Table 4.4: Chemical composition of sputter deposited 5052 aluminum alloy determined by
EPMA, compared with the bulk alloy composition.
Alloy Element Bulk Alloy Sputtered Film
Weight Percent Weight Percent
Al 97.25 98.9 +/- 0.5
Mg 2.5 1.73 +/- 0.03
Cr 0.25 0.16 +/- 0.06
The sample that was cooled to -196"C was tested for stress relaxation at 75C. At
this temperature the tensile stress in the film was 50 MPa and there was no stress relax-
ation observed over forty-eight hours of testing.
The EPMA chemical analysis of the 5052 samples is shown in Table 4.4. Again,
there are differences in composition between the sputtered thin film and the bulk alloy,
presumable due to sputter deposition effects [Zhe97].
Figure 4.22: SEM micrograph of 5052 aluminum alloy thin film sample that was
temperature cycled to 3500C at 10C/minuteand held at 350C for 30 minutes
and then cooled to room temperature over 48 hours.
Figure 4.23: SEM micrograph of 5052 aluminum alloy thin film
that was held at 100C for 60 hours.
Figure 4.24: AFM height image of 5052 aluminum alloy thin film without heat treat-
ment. Scan area is 5 im by 5 gpm, Z-range 32nm and RMS roughness is 3.77 nm.
Figure 4.25: AFM height image of 5052 aluminum alloy thin film heat treated at 125C
for 32 hours. Scan area is 5 pm by 5 tm, Z-range 60nm and RMS roughness is 7.09 nm.
Table 4.5: RMS roughness and Z-range for 5052 aluminum alloy thin films that have
undergone different heat treatments.
Time at RMS
Tempe atur R Z-Range RMS Roughness Z-Range
Temperature Roughness (nm) (nm)
,TTs (m (nm) (nm) (nm)
0 3.77 32.4 3.77 32.4
2 5.9 47.5 5.9 60.2
4 4.6 35.7 5.6 55.2
8 5.8 51.1 5.6 55.5
16 6.3 49.4 4.6 58.9
32 5.6 51.7 4.7 43.6
Figure 4.22 and 4.23 show SEM micrographs of 5052 alloy thin films that have
undergone different heat treatments. EDS was used to examine the difference between
the light and dark areas seen in figure 4.23. No difference in composition between these
two areas was found.
The changes in AFM surface roughness with stress relaxation in samples that
were heat treated at 50C and 1250C for up to 32 hours is shown in Table 4.5 and figures
4.24 and 4.25. The RMS roughness and Z-range (the height difference between the
highest and lowest point on the sample) are reported in Table 4.5. Figure 4.24 shows the
AFM height image of a 5052 alloy sample with no heat treatment and figure 4.25 shows
the AFM height image of a sample that had been held at 125C for 32 hours. The size of
features in these images are nearly the same, but the RMS roughness and the Z-range
have doubled for the sample held a 125C (Table 4.5). The data in Table 4.5 shows that
the change in surface roughness occurred in the first two hours of the heat treatment, and
that heat treatment temperature has little effect on the rate or the magnitude of the rough-
2 -Mg Al A] A]
0 I I i II
0.0 0,5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4,5 5.0
SPUTTER TIME, min.
Figure 4.26: Auger depth profile of 5052 alloy thin film. Film sputtered with
3 keV argon in a 3 mm x 3 mm raster.
ness change. AES depth profile data in Figure 4.26 shows that the surface oxide was
between 20nm and 30nm thick.
Powder x-ray diffraction analysis of 5052 thin films heat treated at 500C or 125C
for up to 32 hours showed that the (111) peak shifted low 0.1" for heat treated (2 to 32
hours) versus unheated samples. No additional peak shifts were detected for samples
heat treated longer than two hour. The peak shift for the (220) showed the same trend
with a shift of 0.1. Samples were also scanned to detect changes in the A13Mg2 peak
heights or widths that would indicate a change in these precipitates. No measurable
change was seen in these peaks indicating that the precipitates had not coarsened. Line
shape analysis was also performed on the X-ray data to determine crystallite size as a
function of stress relaxation, but no correlation was found with stress relaxation time or
150 ----- --i- ------ ---- -------- ----
100 ---- --- ---- ---- ----
0 50 100 150 200 250 300 350
Figure 4.27: Stress versus temperature plot of 2090 aluminum alloy thin film.
Figure 4.27 shows the stress versus temperature plot for the 2090 aluminum alloy
thin film for the first and second temperature cycle to 350C. The as deposited tensile
stress in this film is 250 MPa before the first cycle. During the first temperature cycle,
after deposition the film stress decreases linearly with increasing temperature up to
140C. Between 140C and 2000C the stress in the film increases with increased tempera-
ture, presumable due vacancies migrating to and being destroyed at the grain boundaries
[Tow87]. This behavior was typical for T201 and 5052 films also during the first thermal
cycle. Once the excess vacancies were eliminated, the slope again becomes linear up to
275C. At 2750C the slope increases and becomes unstable up to 350C, where the film is
under a compressive stress of 70 MPa. Upon cooling, the slope of the stress-temperature
curve is much less than during heating and the film only reaches a tensile stress of 80
0I I i I
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
SPUTTER TIME, min.
Figure 4.28: Auger depth profile of 2090 aluminum alloy thin film. Film sputtered with
3 keV argon in a 3x3 mm raster.
MPa at room temperature. On the second temperature cycle the film attains a slightly
higher compressive stress of 80 MPa at 3500C and returns to a tensile stress of 80 MPa
upon returning to room temperature. When the sample was removed from the Flexus it
was noticed that the specular reflection of the film was greatly reduced due to surface
Based on the reduced specular reflection and the lower slope of the stress tem-
perature curve it was assumed that the film had oxidized badly. This was confirmed by
the Auger depth profile data from that sample, (see Figure 4.28). As can be seen, the
oxygen concentration falls off very gradually. This indicates that the film had oxidized to
a depth greater than the 5052 or T201 films (see figures 4.26 and 4.13).
The stress versus temperature plot for the copper thin film is shown in figure 4.29.
While this plot is quite complex the data are consistent with the film being oxidized
during heat treatment. As the stress-temperature plot indicates, the film was stable during
the first thermal cycle to only 1000C. Another film was tested for stress relaxation at
50oC for 22 (see figure 4.30). The copper film oxidized in this test, as indicated by a
greatly reduced the specular reflection, and the film stress increased after 14 hours due to
To inhibit the oxidation of the copper thin film, a 75 nm thick protective alumi-
num layer was deposited on the copper. This sample was then heated to 500C for 6
hours in a tube furnace to test the stability of this protective coating. The film retained its
specular reflection during this test. From an Auger depth profile on this sample the oxide
thickness is approximately 200nm (see figure 4.31).
0---"V ^--------------- -----
0 200 ..........
-3 o o .. .. .. .....-- -- '- --- .- -"-: ---. .. ......-- .- -.. --. -... -..... .....
-400 --------.--- -- ----------- -.------ -----,-
0 100 200 300 400 500
Figure 4.29: Stress versus temperature plot for a copper thin film.
0- 400 .-------------- --- ------------------- 4-- ---------------------
23 0 -^ .^ .------ .-------J---I- --------.--|---------------.-------- -- ----------- -------
-1 0 --------------------- ----------- -------------------- -------- -------------
0 5 10 15 20 2
Tim e, H ours
Figure 4.30: Isothermal stress test of a copper thin film conducted at 500C.
0.0 0.5 10 1.5 2.0 2.5
SPUTTER IME, min.
30 3.5 4.0 4.5 5.0
Figure 4.31: Auger depth profile of a copper film coated with a protective layer of alumi-
num 750 nm thick. The film was sputtered with 3keV argon in a 3x3 mm raster.
Titanium. Manganese and Nickel
The stress versus temperature plots for titanium, manganese, and nickel are shown
in figures 4.32 through 4.34. All of these films oxidized severely during their first tem-
perature cycle up to 500C, 250C, and 500oC for titanium, manganese and nickel respec-
tively. This was evident from a change in specular reflection, color change of the samples
and by the stress-strain data. For titanium (figure 4.32) the rapid reduction in tensile
stress at 350C to a compressive stress of over 600 MPa at 500C and the film maintain-
ing a high tensile stress (375 MPa) upon cooling to room temperature indicates some
changes in the film from oxidation. Oxidation of the manganese film (figure 4.33) results
in an 800 MPa shift in the stress, becoming highly tensile at just over 2000C. The nickel
film (figure 4.34) shows a change in slope of the stress strain curve at 2750C and the
residual stress at room temperature dropped from 1000 MPa before to 600 MPa after the
600 ------ ---- ---- ------ ----I------ ----------------- ---- -----^--- ------------
2 0 0 ------ --------- -------- ------------ ---- ---- -,-- .----- --------------
2 200............. ....... ... -..............
-4 0 0 C o oli. ----- I --... .. .
-600 ---------- ------------------------ --------------
0 100 200 300 400 500
Figure 4.32: Stress versus temperature plot for a titanium thin film.
800 ----------------------- W..... .. .
S600---------- --- ------- ---------
-100 ------------ ,-'-
0 50 100 150 200 250
Figure 4.33: Stress versus temperature plot for a manganese thin film.
100 --------------------- 0 -------------
600 -------- ----------------.^ ^-----------------S--------------- ---------------------
200 -------------------------- -----------------.------
-200 ------------------------------------------------------------------ -----
0 100 200 300 400 500
Figure 4.34: Stress versus temperature plot for a nickel thin film.
thermal cycling due to oxidation. An aluminum protective coating was deposited onto a
manganese sample, but did not inhibit oxidation of the sample.
After the copper-gold films had undergone the homogenization and ordering heat
treatment, their chemical composition was analyzed by EPMA (see Table 4.6). The films
were within 2 to 3 percent of the desired composition.
In powder x-ray diffraction all of the films showed a shifts in the diffraction peak
positions form disordered to the ordered state, and additional peaks were observed from
the long range order. As reported above, for textured thin films it was not possible to
calculate the degree of ordering.
The films were thermally cycled in the Flexus up to 450C in an argon atmo-
sphere. Figures 4.35 through 4.37 show the results of these tests. The CuAu films did
not show a change in stress near the order-disorder temperature, 4000C figure 4.35).
The CuAu film showed a marked change in stress at the transition temperature (figure
4.36) as did to a lesser degree the CuAu, film (figure 4.37). However, the order-disorder
transition occurs well above the temperature (300C) at which the films begin to plasti-
Table 4.4: Chemical composition of copper-gold thin films.
Atomic % Atomic %
Cu3Au 77 +/- 0.2 23 +/- 0.2
CuAu 53 +/- 0.4 47 +/- 0.5
CuAu, 22 +/- 0.4 78 +/- 0.4
500 --- ------------------
40 0 ----------------- --^ ---- ------- .--- .- ---- ^-- ^---- --- .-- -- .-- ---------------- I --------------.---.---
300 --------------------- ------- ------
10 0 ^- ^-^--- ^-------- --: --------------- .------- :^ ^----- ------------------ ---------------- -
0 100 200 300 400 500
Figure 4.35: Stress versus temperature plot for a Cu3Au thin film.
S 3 0 0 ------------------ -- -- ---- --------- .------------------ -------. ----------------------^ -----^-----------
0 100 200 300 400 500
Figure 4.36: Stress versus temperature plot for a CuAuthin film.
500 --- ---------------------------------------------------------------------------
500 --- -------- ---- -- -
0 O ------------ -- -- ^ -- ^ ^^ -- -- -
3 0 0 ---------. -------- -.----------- .. ---- ---. .--.---.-------- --------------- -- -- ---
^= 200 ---- ----- ------------ .-------- --------------------|-- -------------------4--- ------ .---- ---- .-- --
0 100 200 300 400 500
Figure 4.37: Stress versus temperature plot for a CuAu3 thin film.
320 5-------- -----------------------------..................................
0 3 15 .. -
310 --... ..
0,5 1.0 1.5
Figure 4.38: Isothermal stress relaxation of ordered and disordered CuAu,
thin films at 150"C.
A series of isothermal stress relaxation test were performed on these samples.
These tests were conducted in argon since these films oxidize if heated to 150C in air.
The tests were run at 150C to maximize mobility in the samples. Figure 4.38 shows the
isothermal stress relaxation data for an ordered and a disordered CuAu, thin film. The
ordered film has a 2 percent reduction in stress over two days while the disordered film
has a 7 percent reduction in stress over the same period. The isothermal stress relaxation
in the CuAu and Cu3Au films showed no difference between the ordered and disordered
The as-deposited tensile stress in the layered AlTi film was 200 MPa. On the
first temperature cycle to 500C the slope of the stress temperature curve changed at
350"C, indicating a change in the structure of the film. At 500C the film was in a state of
no stress. When the film cooled to room temperature it was under a tensile stress of 800
MPa as shown in figure 4.39. The film was cycled up to 500C two additional times and
showed no hysteresis in the stress-temperature plot. An isothermal stress relaxation test
was performed on this sample at 350C and the film showed no stress relaxation over a 48
The sample was then tested at Tencor up to 8000C. The results of this test are
shown in figure 4.39. The AlITi film begins to plastically deform above 500"C under a
compressive stress of 200 MPa. At 8000C the film still maintains a compressive stress of
300 MPa. Upon returning to room temperature the film was under a tensile stress of
nearly 1000 MPa. An isothermal stress relaxation test was also performed at 450C and
again the film showed no stress relaxation over a 48 hour period.
The native passivating oxide held up well for all tests at all temperatures. The
specular reflection of the sample was high after all the tests. There was no visible change
in the oxide after cycling to 500C. Figure 4.40 shows the Auger depth profile for an
800 ... .
600 : "
2 0 -" ---...-- -. -- -.
4200 .. ....... ............. ^' .. .. .
*4000-----------i -- ***
0 100 200 300 400 500 600 700 800
Figure 4.39: Stress versus temperature for Al3Ti thin film.
Al Ti film that was heated to a maximum of 500oC. The oxygen concentration drops
sharply after 5 minutes of sputtering. Figure 4.41 shows the Auger depth profile for a
sample heated to 800"C. The oxygen concentration drops to near zero after 25 minutes of
sputtering, indicating that the surface is passivated, but the oxide layer is about three
times thicker for samples heated to 8000C are than samples heated only to 5000C.
Modeling the Effects of Oxide Thickness on a Bimetallic Actuator's Curvature
As has been seen with many of the materials examined in this study oxidation has
a significant impact of the performance of the material. For the aluminum alloys, it was
not possible to test for the effects of the oxide layer since one always forms when alumi-
num is exposed to air. To evaluate the effect of the oxide on the performance of a bime-
tallic strip, the change in curvature of a bimetallic strip as a function of the Al203 layer
thickness was calculated. This calculation is based on a bimetallic strip that consists of 5
SPUTTER TIME, min.
Figure 4.40: Auger depth profile of AITi sample that was heated to 5000C.
0.0 5 10 15 20 25 30
SPUTTER TIME, min.
Figure 4.41: Auger depth profile of AI3Ti sample that was heated to 8000C.
gm of aluminum on 5 gm of silicon with an oxide whose thickness varied from 0 to 1
% of Max
0 ---- I I
0 02 04 0.6 0.8
Figure 4.42: Curvature of a bimetallic strip as a function of oxide thickness as a percent-
age of maximum curvature.
pm. Figure 4.42 shows the results of this calculation. From this calculation it can be
seen that the 30 nm oxide layer found on the aluminum alloy samples reduces the curva-
ture of the bimetallic strip by 0.5 percent, but thicker oxides would have a more signifi-
cant impact. Therefore, controlling the oxide thickness is critical to maximizing the
performance of a bimetallic strip.
In this chapter the results and conclusions that have been drawn from this research
will be presented and discussed with respect to the three objectives of this research. (i)
effectiveness of strengthening techniques in thin films. (ii) identification of the mecha-
nism responsible for isothermal stress relaxation in thin films. (iii) identification of
materials for high temperature applications. In addition, to assist the development of
bimetallic actuators, a figure of merit has been developed and a number critical param-
eters have been identified to assist in the selection of materials for this application.
Thin Film Strengthening
Of the five strengthening techniques used in bulk metals, this study examines the
effectiveness of two of these, solid solution strengthening and multiphase hardening
(precipitation/age hardening) in thin films. While not normally considered a strengthen-
ing technique, the effectiveness of ordered intermetallics to resist stress relaxation was
also examined. Solid solution and precipitation/age hardening were selected because of
their effectiveness in strengthening bulk metals and because these mechanisms are rela-
tively insensitive to small variation in composition, which improves manufacturability.
The order/disorder transition that occurs in some intermetallics was also of interest due
to the large change in volume that frequently occurs at this transition. The order / disor-
der transition provided a means of evaluating the effectiveness of the ordered phase to
resist stress relaxation.
With respect to measuring the strengthening effects of these techniques, there are
a number of parameters that are of interest. With bulk materials, strengthening is usually
evaluated by measuring the yield strength or ultimate tensile strength of the material at a
temperature. In thin film bimetallic structures, this is not possible since stress and tem-
perature are dependent on one another. Thus the temperature and stress at which plastic
deformation begins are interrelated, and is the point on the stress versus temperature
curve where the slope significantly departs from linear. This study is also concerned with
the stress relaxation that occurs in thin films, which determines the stability of a bimetal-
lic actuator. So the strengthening techniques must be evaluated with respect to their
effectiveness at increasing both the temperature and therefore the stress at which plastic
deformation occurs and at reducing isothermal stress relaxation in the film.
While it would be reasonable to assume that the bulk strengthening techniques
would be effective in thin films, the AlSiCu alloy used in microelectronics would suggest
that the strengthen mechanisms are less effective in thin films. In the AlSiCu alloy,
silicon could provide some solid solution strengthening, but the solubility of silicon in
aluminum at room temperature is very low. Copper could provide precipitation strength-
ening, but due to processing is generally in the overaged condition. As shown in Figure
2.9, the strength of the AISiCu alloy is only 30% greater than pure aluminum. In contrast,
age hardened bulk aluminum copper alloys can have yield strengthens of five times those
of pure aluminum. While this suggest that bulk strengthening mechanism may not be as
effective in thin films, it must also be noted that the AISiCu alloy was not developed for
strength but to reduce aluminum spiking into silicon wafer (Si) and to reduce
electromigration (Cu) [Bow74]. On the other hand, the alloys evaluated in this study
were developed for strength.
Solid Solution Strengthening
The aluminum alloy 5052 was selected to test solid solution strengthening. This
alloy is solid solution strengthened by the addition of magnesium. While the majority of
the magnesium stays in solution, some AIMg2 and AlIMg precipitates form. However,
these precipitates do not provide the primary strengthening. Comparing the stress tem-
perature curve for 5052 aluminum (Figure 4.16) with the curves for pure aluminum and
the AISiCu alloy (Figure 2.10) clearly shows that solid solution strengthening is effective
in thin films. The room temperature residual stress for the 5052 alloy is 300MPa as
compared to 150MPa for pure aluminum and 200MPa for the AISiCu alloy. Also plastic
deformation begins in the 5052 alloy sample at 225C under a compressive stress of
40MPa while the pure aluminum and AISiCu alloy begin to deform under the same stress
but at much lower temperatures of 1250C and 175C, respectively. Therefore solid solu-
tion strengthening has significantly increased the performance of the film.
Testing also showed that the 5052 alloy film is more resistant to stress relaxation
than a pure aluminum or AlSiCu films. At 500C, pure aluminum and the AlCuSi alloy
films experience a 17% reduction in stress over 48 hours while the 5052 alloy film only
experience a 6.7% reduction in stress over the same time period. This is even more
significant because the 5052 alloy film was under a tensile stress of 150 MPa while the
pure aluminum was under only a 80 MPa compressive stress.
In addition to the 5052 aluminum alloy the copper / copper gold systems were
examined. The copper gold system was examined primarily because it undergoes an
order disorder transition. However, in the disordered phase it can be considered a solid
solution of the two elements and it can be compared to pure copper. From (Figure 4.29)
the room temperature residual tensile stress for pure copper is 50 MPa and the films
began to plastically deform above 2500C under a compressive stress of 300 MPa. The
stress temperature plot for CuAu (Figure 4.35) shows a residual room temperature tensile
stress in the film of 575 MPa and the film begins to plastically deform at 300C under a
compressive stress of 50 MPa. As with aluminum, solid solutions of copper and gold
nearly doubled the strength of the film. Stress relaxation tests were not done on the pure
copper films due to oxidation.
Given the results of the 5052 aluminum and copper gold, it is clear that solid
solutions are effective in strengthening thin films. It has also been shown that this
strengthening mechanism is effective at reducing stress relaxation in aluminum alloys.
Precipitation and Multiphase Hardening
The T201 aluminum alloy was selected to determine the effectiveness of precipi-
tation hardening to reduce stress relaxation. The T201 aluminum alloy is not strength-
ened only by precipitation hardening. Magnesium and manganese are added for solid
solution strengthening and titanium is added to reduce grain size. Copper is the primary
alloying element and it leads to strengthening by precipitates. Silver is added to stabilize
the copper precipitates. Table 4.2 shows the complete composition of the T201 alloy.
Since this alloy is precipitation and solid solution strengthened, comparisons must be
made to pure aluminum, AISiCu alloy, and the 5052 alloy to evaluate the effectiveness of
the precipitation strengthening.
Figure 4.1 shows the stress-temperature plot for the T201 alloy. The room tem-
perature residual stress is nearly 400 MPa, which is 100 MPa higher than the 5052 alloy.
(Figure 4.16), 2.5 times greater than pure aluminum and 2 times greater than the AlSiCu
alloy (Figure 2.10). The T201 alloy began to plastically deform at 2250C under a com-
pressive stress of 75 MPa, again much higher than pure aluminum or the other aluminum
alloys. At 50oC the T201 alloy thin film undergoes a 12% reduction in stress over 48
hours (Figure 4.2) as compared to 7% for the 5052 samples (Figure 4.17) and a reported
value of 15% for the pure aluminum and AlSiCu alloy. However, at 500C, the T201 film
is under a tensile stress of 390 MPa as compared to 270 MPa for the 5052 alloy films, 75
MPa for the pure aluminum film and 150 MPa for the AlSiCu alloy film. At 125C, the
T201 sample undergoes a 2% stress relaxation over 48 hours under a 114 MPa tensile
stress (Figure 4.5) while the 5052 alloy film undergoes an 18% stress relaxation under a
104 MPa tensile stress (Figure 4.20). Therefore, precipitation/age hardening is effective
at reducing stress relaxation in thin films.
The plane view TEM photomicrograph in Figure 4.11 verifies that precipitates
were present in the aluminum thin film. In this image, precipitates are clearly seen within
aluminum crystals and the diffraction pattern for CuAI2 is seen. The size of these precipi-
tates is on the order of 20nm to 30nm which is appropriate for age hardening [Bro82].
Based on the testing and microstructural examination of the sample it can be
concluded that precipitation/age hardening is effective in thin films. It should also be
noted that the addition of silver to this alloy stabilizes the CuAlI precipitates. Without
this stabilization, the copper would precipitate at the grain boundaries, have a coarse
dispersion rather than a fine dispersion, and be much less effective at strengthening the
film. A coarse dispersion of "overaged" precipitates is the reason for lack of strength in
the AISiCu alloy [Bro82].
Ordered Phases and the Order Disorder Transition
While not normally considered a strengthen technique, ordered phases generally
have a higher strength than their disordered counterparts. As mentioned above, order
disorder transitions were also of interest due to a large change in volume that occurs
during this transition. This volume change could generate large stress or deflection
values in the design of thermally actuated MEMS. Figures 4.37, 4.38 and 4.39 show the
stress temperature curves for the three copper gold compounds tested. While there is a
substantial change in stress seen near the order disorder temperature, this occurs beyond
the point at which the yield strength / temperature conditions lead to plastic deformation
for these films and the order/disorder transition is not of use in this system.
Ordered phases were presumed to have a lower stress relaxation rate than disor-
dered phases since the self diffusion rate, which determines the creep rate and stress
relaxation rate for many materials, were expected to be lower in an ordered phases versus
a disordered phase [Cah83]. Figure 4.38 clearly shows that this presumption was true, for
Cu3Au, the ordered phase showed a 1.6% reduction in stress over two days compared
with a 6.5% reduction in stress for the disordered phase. The disordered film's stress
relaxation rate is almost four time faster than the ordered phase, while the stress is only
8% higher on the disordered film.. The AlITi ordered intermetallic also showed no stress
relaxation at 4500C in two days.
Stress Relaxation Mechanism in Thin Films
While increasing the strength of the aluminum thin films reduced the stress
relaxation rates and magnitudes, it did not eliminate stress relaxation. Thus the mecha-
nisms responsible for stress relaxation in thin films needs to be discussed. In this study
three modes of stress relaxation have been observed. The first mode is plastic deforma-
tion of the thin film [Her85]. If the stress applied by the substrate is greater than the yield
strength of the material, the film will plastically deform and relax the stress. This is seen
as a change in slope of the stress-temperature curves upon heating (Figures 4.1, 4.16,
4.35, 4.36, 4.37, 4.39). The second mode of stress relaxation occurs over several tens of
minutes after a sample has reached temperature. This is seen in the stress relaxation
curves of the T201 and 5052 aluminum alloy films (Figures 4.2-4.4 and 4.18-4.20). The
third mode of stress relaxation occurs over a number of days and can again be seen in the
stress relaxation plots of the T201 and 5052 samples.
The challenge in determining the stress relaxation mechanisms in thin films on a
substrate is that film stress and temperature can not be independently varied. This elimi-
nates the use of an Arrhenius equation to calculate the activation energies of the processes
under investigation. It is possible to manipulate samples at different temperatures to the
same stress, however the microstructure and dislocation array of a material are not state
functions. Therefore, their structures will be quite different at different temperatures even
if the stress is the same. In addition, removing the film from the substrate would, in many
cases, substantially change the film in a number of ways. For self passivating materials,
such as examined in this study, the back side of the film would oxidize upon exposure to
air. It has been shown that dislocations interact differently with oxidized surfaces
[Nix89]. It would also be difficult to create and apply a biaxial stress to a free standing
thin film and to measure strain in the film. Therefore, direct determination of an activa-
tion energy and thus the mechanisms responsible for the different modes of stress relax-
ation is not possible. Other data are needed to determine the mechanisms responsible for
stress relaxation in thin films.
It should be noted that the change in total strain and temperature to reduce the
stress in these films is very small. A change in stress of 10 MPa is roughly equivalent to
a temperature change of only five degrees. For aluminum on silicon, this is a strain of
only 0.00012. Thus very small changes in the dimensions of a film can result in large
changes in the stress in the film.
The inability to calculate activation energies for stress relaxation modes 2 and 3
makes it very difficult to identify their mechanisms. All plausible relaxation mechanism
must be considered including: recovery recrystallization, logarithmic creep, Andrade
creep, bulk diffusion, gain boundary migration, grain boundary sliding, surface to grain
boundary diffusion, polygonization/subgrain coalescence, and precipitate formation or
coarsening. Several of these mechanism can be eliminated as possible causes of both
modes 2 and 3 stress relaxation.
For the aluminum alloys at 150C, the T, is less than 0.5, which is slightly lower
than the temperature at which Andrade creep would be expected. In addition, none of the
stages of Andrade creep would produce an exponential decay as is seen in the stress
relaxation data. Samples heated at higher rates showed no transition from stage 1 creep
to stage 2 creep. In the early stages of relaxation the stress relaxation rate decrease, but
does not go to zero, and the stage I strain rate is not a simple exponential [Cah83] decay
as seen in the data. Stage 2, steady state creep has a constant strain rate that would
produce a constant stress relaxation rate that is not seen in the data. The strain rate in
stage 2 creep has some dependence of the applied stress, but this relationship is not a
simple exponential [Cah83]. In the third stage of Andrade creep the stain rate increases
and so is not consistent that data that has been collected. Therefore, Andrade creep does
not appear to be responsible for the stress relaxation.
Diffusional creep processes are generally significant at temperatures of, TH > 0.6,
which is well above the temperatures at which the current stress relaxation is seen (TH <
0.5). Diffusional process increase exponentially with temperature [Cah83] while the
present stress relaxation shows an inverse relationship to temperature. Diffusional pro-
cesses should result in a zero stress at infinite time instead of the finite residual stress
projected from the furve fits (Tables 4.1 and 4.3).
X-ray line shape analysis did not detected polygonization or subgrain coalescence,
so these mechanism can be eliminated. In addition, subgrain coalescence [Cah83] is a
preliminary step to recovery and recrystallization, and the microstructure of the samples
did not change during stress relaxation (Figures 4.7 and 4.12). Thus recovery, recrystalli-
zation and grain boundary migration do not appear to be responsible for the stress relax-
ation. The formation or coalescence of precipitates does not appear to be a factor. X-ray
diffraction showed no significant change in precipitate concentration over time. In
addition, the same two modes of stress relaxation have been seen in pure aluminum
[Her85]. Having eliminated these potential mechanism of stress relaxation it is necessary
to separately consider the mechanism applicable to each mode.
Mode 2 Stress Relaxation
Mode 2 stress relaxation occurs relatively quickly. It is virtually complete in two
hours. During this time there is a measurable increase in the surface roughness of the
film (Table 4.3). There is also a 0.10 shift in the (111) x-ray diffraction peak. The in-
crease in surface roughness equates to a 0.15% change in the thickness of the 1.2p.m film
which is of the same order as the change in the d spacing calculated from the x-ray peak
shift of 0.1 This indicates that the processes affecting the surface is occurring throughout
the thickness of the film. Surface to grain boundary diffusion is not the likely cause of
this stress relaxation since diffusional processes are temperature sensitive as discussed
above [Ask89], while the roughness and peak shifts are temperature insensitive. Grain
boundary sliding is also not consistent with the results obtained since it normally has a
very high activation energy [Mey84]. A high activation energy is not consistent with a
process that occurs rapidly at low temperature. The film also has a columnar grain
structure [Kra90, Tho74] so the resolved shear stress on most of the grain boundaries will
be very small. If some crystals have a tilted grain boundary, these grains should be
visible with the AFM. Therefore, grain boundary sliding is eliminated as a potential
explanation of stress relaxation.
The only mechanism not yet eliminated to explain stress relaxation is logarithmic
creep which results from dislocation glide on slip systems. It is postulated that mode 2
stress relaxation is caused by the movement of dislocation on slip planes that terminate as
the surface of the film. Dislocation moving in these slip systems should not encounter
dislocation pile ups as would be encountered in slip system that terminate at grain bound-
aries. Thus these dislocations should move and be eliminated quicklt. The degree of
relaxation should only be limited by the number of sources of mobile dislocations. The
increase in roughness could be caused by dislocation bands. Also because of the stress
level and temperature of these films, the DMM's predict that logarithmic creep would be
the dominant process (Figure 2.5) [Cah83].
Mode 3 Stress Relaxation
The mechanisms not yet eliminated for mode 3 stress relaxation are grain bound-
ary sliding, surface to grain boundary diffusion, and logarithmic creep. Surface to grain
boundary diffusion is a diffusional process that should be exponentially dependent upon
temperature. It could increase surface roughness as atoms migrate from the surface to the
grain boundary. However, the data do not show a large temperature dependence and the
surface roughness does not increase over stage 3. The temperature at which the stress
relaxation is occurring is also low for a diffusional process to be active (TH < 0.5). There-
fore, surface to grain boundary diffusion does not appear to be the mechanism responsible
for this mode of stress relaxation. Grain boundary sliding also not likely to be the mecha-
nism responsible for this stress relaxation, since grain boundary sliding should also
increase surface roughness. In addition, the activation energy for grain boundary sliding
is very high and it normally only occurs just before mechanical failure. Furthermore, in a
columnar structure the resolved stress on most grain boundary should be very low. This
only leaves logarithmic creep, which was concluded to be the mechanism for mode 2
stress relaxation. However, how can the same mechanism have a different rate constants?
Different resolved stresses on different slip directions could explain this in part, but, the
slip systems are all equivalent within each of the (100),(110) and (ll1) oriented crystals.
For this to be the cause of change in relaxation rate, the different oriented crystals would
have to relax at different rates, and this should be noticeable in the AFM images. Also,
no increase in the roughness of the film was seen in mode 3. It is known that disloca-
tions pile up at grain boundaries [Mey84]. Dislocation climb could move some of these
dislocations into the grain boundary. This would result in a reduction of stress and no
increase in surface roughness. Therefore this mechanism is consistent with the observa-
tion made in this study and is postulated to be the origin of the mode 3 stress relaxation.
High Temperature Application
Micro actuators are needed that can operate in severe environments, such as inside
jet engines. In these severe environments oxidation was assumed to be a major concern.
High temperature application also require materials that maintain their strength at the
operating temperatures. These topics will be examined in this section.
During this study it was found that oxidation is a pervasive problem affecting all
temperatures ranges. As is shown above and in Appendix C, a 30nm thick oxide forming
of the surface of a Sm Al 5pm Si bimetallic strip reduces the curvature of the strip by
0.5%. The 2090 aluminum lithium alloy was found to be unsuitable for bimetallic actua-
tors because the lithium increased the oxidation rate (Figure 4.29). The other elements
tested also behaved poorly. At elevated temperatures, copper, titanium, manganese, and
nickel all oxidized (Figures 4.31, 4.32, 4.34, 4.35). In bulk metals application the thick-
ness of the passivating oxides is of little concern as long it remains intact. In thin films
the passivating oxide can consume a large fraction of the thickness. This is very damag-
ing to a bimetallic actuator because most oxides have a very low CTE which will reduce
the displacement of the device. Thus, a material's oxidation rate is a critical parameter
and those metals with the slowest oxidation rate are best suited for this application.
Aluminum and silicon both have very low oxidation rates due to the formation of passi-
vating oxides and both are well suited for elevated temperature applications [Wes95]. A
thin aluminum layer was also found to be effective at protecting less resistant materials,
such as pure copper.
Strength at Elevated Temperatures
As discussed above, intermetallics have a high resistance to creep due to their low
self diffusion rates. This is of interest for elevated temperature application where a lower
melting point intermetallic with a higher CTE could be used in place of a standard higher
melting point alloy with a low CTE. The Al1Ti intermetallic performed well, showing no
stress relaxation at 450C and no oxidation. Unfortunately nickel and copper oxidized, so
comparisons of stress relaxation could not be made.
Figure of Merit
To assist engineers who design bimetallic actuators in selecting the active layer
material, a figure of merit has been developed. The figure of merit is designed to provide
an initial evaluation of the type of material that should be used, i.e. aluminum or nickel.
The figure of merit, FOM, is expressed in equation 5.1 as:
FOM = (MHT -M i Y'S. [(FD CTE 10 'C) + ( E 101 Pa'. (1 FD))] (5.1)
-- W -- --- --P 1---W--
Temperature compensated Displacement Force
yield strength mode mode
where MHT is the maximum homologous temperature (the homologous temperature
above which the yield strength of the material rapidly declines, equal to 0.4 for most
metals and 0.65 for most intermetallics), MOT is the maximum operating temperature of
the actuator, YS is the yield strength of the material at room temperature, CTE is the
coefficient of thermal expansion for the active layer material, E is Young's modulus for
the active layer material and FD, determines whether the priority of the device is force or
displacement (0 to maximize force, and zero to maximize displacement). Table 5.1
shows a comparison of the different figures of merit for different materials at different
operating temperatures and for force or displacement. Note that the values for tin are
negative, indicating it is not suitable at the temperatures. In general the large CTE of
aluminum alloys give the FOMs of ~ 108 for displacement while the higher Young's
modulus and moderate CTE's of nickel and stainless steal give them FOMs of -101"-10"
for force. Note the FOM at 500oC of 1015 for cobalt. This figure of merit is complex,
but could be expanded to also include the environmental stability of the materials.
Table5.1: Figure of merit ratings for different materials at different operating tempera-
tures and optimized for displacement, D, or force, F.
Max. Temp. 100 (-C) 250 ('C) 500 ("C)
ce D F D F D F
Sn -5.8*106 -3.4*107 -2.9*106 -1.7*107 -1.5*106 -9.1*106
5052 Al 1.4*10' 8.8*109 -2.8*10' 1.7*109 -9.3*106 -5.7*10'
T210 Al 1.1*108 1.2*10 -2.4*107 -2.7*109 -8.3*106 -9.0*10'
Mg 8.1*10' 8.0*10' -3.2*107 -3.2*108 -9.6*106 -9.5*10'
Anneal Al 2.8*107 1.3*109 -1.3*107 -6.0*108 -3.7*106 -1.7*10'
75% c.w. A 5.0*107 2.4*109 -2.3*107 -1.1*109 -6.6*106 -3.1*10'
Cu 7.5*106 1.4*1010 2.1*10' 3.8*101' -1.1*106 -2.0*10'1
Ni 1.0*106 6.1*101 1.7*106 9.7*107" 1.1*10' 6.6*1012
Co 4.2*106 5.4*1013 6.5*106 8.3*10" 7.8*107 1.0*101"
410 S.S. 2.4*10' 2.3*10" 3.6*105 3.4*10" 2.6*106 2.5*10"
c.w. cold worked
S.S. stainless steal
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E41GQ6C9L_Q80D2K INGEST_TIME 2013-01-23T13:52:47Z PACKAGE AA00012979_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC