Thin film metalization for micro-bimetallic actuators


Material Information

Thin film metalization for micro-bimetallic actuators
Physical Description:
vii, 114 leaves : ill. ; 29 cm.
Gorrell, Jonathan Frank, 1958-
Publication Date:


Subjects / Keywords:
Materials Science and Engineering thesis, Ph.D   ( lcsh )
Dissertations, Academic -- Materials Science and Engineering -- UF   ( lcsh )
bibliography   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph.D.)--University of Florida, 1998.
Includes bibliographical references (leaves 109-113).
Statement of Responsibility:
by Jonathan Frank Gorrell.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 029233568
oclc - 39556976
System ID:

This item is only available as the following downloads:

Full Text








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-

tific instrumentation.

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.


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







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



Jonathan Frank Gorrell

May, 1998

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-
Parameter Bimetallic
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
reqauremtnts Co.nlbhaly)

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


Passive Layer





Silicon Expansion
Aluminum Expansion

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-
plex geometry.
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
Metallizatlon Inlet
\ i,3: Dkiphragm

Valve Body \


Cooling or
Heating Stress Relaxation
of Metal
Aluminum Valve Open
Metallization Inlet

\ _\ 4/, -* _

Valve Body


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
k= b
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]

Layer1a Loy-w
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)
2 h

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.
4 h-L

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

; Ni
L 2-
\ CA 172

o 1

CA T201
1 5052
c Contour of
Constant Force Contour of / SMg
Constant Deflection
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

Strain, AL/L
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.

Atomic Bonding

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


0 1.6

0 ,


-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


Cu Au

I /

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
Creep (approximate)
10 \ (Andrade Creep)

10 Herring
(/., Nabarro-Coble
Low Temperature Creep
10 Creep
(logarithmic Creep)
10 (,,,/P)
Anelastic Creep
(Recoverable 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-

(a) (b)
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

discussed below.

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

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

their microstructure.

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-

0.3 0.5

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
Temperature, C

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.
0.2Mg 0.27Ti
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, [100]

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:

E-h2 (4)

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.

Time, Minutes

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
Time, Minutes

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

the TEM.

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

calculated using:

SL=Rdhk (5)

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.

Curve Fitting

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.

Numerical Modeling

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.

T201 Aluminum

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

residual stress.

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 .- ^ -

O 0.


0 50 100 150 200 250 300 350

Temperature, C
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
380 -


Time, minutes

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.

Time, minutes
Figure 4.5: Isothermal stress relaxation of T201 aluminum thin film at 125C.

46 1--------1----------------1------- 1 -----

4O O
40 0
0 0 %0 .0

0 0\^^^X
I 00^ 0 *' ^<

36 I I I I
0 500 I000 1500 2000 2500

Time, minutes
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.


c s

1 2 3 4 5 6 7 8 9 10
Energy, keV
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
Energy, keV
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-


Aluminum Grains
Aluminum Grains

CuAl2 (422)

A (200)
Al (200)

Al (1ll)

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.




I 6

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


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

32 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.

0 Hours



2 HOURS 16 s

1000 Hou

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.

5052 Aluminum

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 --- ------ -----------

Temperature C

Fe 4.16: St------------- ------ress versus temperate pt fr 52

0 50 100 150 200 250 300 350
Temperature, C

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.

Time, minutes

Figure 4.17: Isothermal stress relaxation in 5052 aluminum thin film at 50oC.

Time, minutes
Figure 4.18: Isothermal stress relaxation in 5052 aluminum thin film at 75C

S1000 1500 2000 2500
Time, minutes
Figure 4.19: Isothermal stress relaxation in 5052 aluminum thin film at 100C

2 95


85 wI I

0 500 1000 1500 2000 2500

Time, minutes
Figure 4.20: Isothermal stress relaxation in 5052 aluminum thin film at 1250C.

u U00 1000 1500 2000 2500

Time, minutes
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.

50C 125C

Time at RMS
Tempe atur R Z-Range RMS Roughness Z-Range
Temperature Roughness (nm) (nm)
,TTs (m (nm) (nm) (nm)
(Hours) (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


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


200 ----

150 ----- --i- ------ ---- -------- ----


100 ---- --- ---- ---- ----

0 50 100 150 200 250 300 350
Temperature, C
Figure 4.27: Stress versus temperature plot of 2090 aluminum alloy thin film.

2090 Aluminum

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


9 \0


4 0


1 C
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


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).


100 --------------------------|-----------

0---"V ^--------------- -----

0 200 ..........

-3 o o .. .. .. .....-- -- '- --- .- -"-: ---. .. ......-- .- -.. --. -... -..... .....

-400 --------.--- -- ----------- -.------ -----,-
-500 ------___---------------------___

0 100 200 300 400 500
Temperature, "C
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

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.

pJ 0

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


800 -------------------------------------

600 ------ ---- ---- ------ ----I------ ----------------- ---- -----^--- ------------

2 0 0 ------ --------- -------- ------------ ---- ---- -,-- .----- --------------

2 200............. ....... ... -..............
) 00


-4 0 0 C o oli. ----- I --... .. .
-600 ---------- ------------------------ --------------

-800 _
0 100 200 300 400 500

Temperature, C

Figure 4.32: Stress versus temperature plot for a titanium thin film.


900 Coo

800 ----------------------- W..... .. .

S600---------- --- ------- ---------
S500 --......--.....-.-...-----..---.................-------

200----- ------

-100 ------------ ,-'-
0 50 100 150 200 250
Temperature C
Figure 4.33: Stress versus temperature plot for a manganese thin film.


1000 ------------------------------------
100 --------------------- 0 -------------

600 --
600 -------- ----------------.^ ^-----------------S--------------- ---------------------

200 -------------------------- -----------------.------


-200 ------------------------------------------------------------------ -----

0 100 200 300 400 500
Temperature, C

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.

Copper-Gold Intermetallics

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-

cally deform.

Table 4.4: Chemical composition of copper-gold thin films.

Atomic % Atomic %
Cu Al

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
Temperature, C
Figure 4.35: Stress versus temperature plot for a Cu3Au thin film.

500-------------- -------


S 3 0 0 ------------------ -- -- ---- --------- .------------------ -------. ----------------------^ -----^-----------


0 100 200 300 400 500
Temperature, C
Figure 4.36: Stress versus temperature plot for a CuAuthin film.

500 --- ---------------------------------------------------------------------------
500 --- -------- ---- -- -

0 O ------------ -- -- ^ -- ^ ^^ -- -- -

3 0 0 ---------. -------- -.----------- .. ---- ---. .--.---.-------- --------------- -- -- ---

^= 200 ---- ----- ------------ .-------- --------------------|-- -------------------4--- ------ .---- ---- .-- --

-100 ----------------------------------------------------------------

0 100 200 300 400 500
Temperature, C
Figure 4.37: Stress versus temperature plot for a CuAu3 thin film.

320 5-------- -----------------------------..................................


0 3 15 .. -
Disordered CuAu
310 --... ..
^3 305

0,5 1.0 1.5
Time, days

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


Aluminum-Titanium Intermetallics

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

hour period.

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
Temperature C
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


Figure 4.40: Auger depth profile of AITi sample that was heated to 5000C.

0.0 5 10 15 20 25 30


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

A1203 Thickness

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

Full Text
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