On the role of indium underlays for the prevention of thermal grooving in thin gold films


Material Information

On the role of indium underlays for the prevention of thermal grooving in thin gold films
Physical Description:
vi, 139 leaves : ill. ; 28 cm.
Lee, Soo Young, 1954-
Publication Date:


Subjects / Keywords:
Indium   ( lcsh )
Gold films   ( lcsh )
Grain boundaries   ( lcsh )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph. D.)--University of Florida, 1985.
Includes bibliographical references (leaves 132-138).
Statement of Responsibility:
by Soo Young Lee.
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 - 000870298
notis - AEG7375
oclc - 14471135
System ID:

Full Text







Tns cssertation.

The autor grate ac
and advice oy vrs. R ~. eHcff, ?. H. -oA 'cway and J. J. Hren.

Special :ranks c : Lee Grures in e<:ron'x rc K. K. <. r in 3GC

,"r ~neir assistance in ootaining XTEM pnctcmicrograpns.

Hyunsock and Nakyung Lee are thanked for their patience and


Final ly, -re a.tcr gratefully acknow'ecces the Iinancial s-p,-

by ARCD and assistance rendered by the Major Analytical Insrertaticn

Center ,AIC : at :ne University of Florida.


ACKNOWLEDGEMENT ........................................... 11ii

ABSTRACT ................................................ ...... v






INTRODUCTION.................................. 1

THEORETICAL BACKGROUNDS....................... 5

Brief Review of Driving Forces for
the Mass Transport............................ 5

Capillary Induced Mass Transport .............. 8
Introduction.................................. 8
Theories of Thermal Grooving.................. 9
Theories of Morphological Instability......... 14

EXPERIMENTAL PROCEDURES....................... 17

3.1. Film Preparation.............................. 17

3.2. Isothermal Annealing.......................... 20

3.3. Characterization of Film Microstructures...... 21
3.3.1. Scanning Electron Microscopy.................. 21
3.3.2. Auger Electron Spectroscopy................... 22
3.3.3. Transmission Electron Microscopy.............. 23
3.3.4. Cross Sectional Transmission Electron
Microscopy.................................... 24
3.3.5. X-Ray .Diffraction............................. 29


CHAPTER 4. RESULTS AND DISCUSSION........................ 30

4.1. Microstructures of Pure Au Films.............. 30
4.1.1. As Deposited Pure Au Films.................... 30
4.1.2. Annealed Pure Au Films........................ 38

4.2. Microstructures of In/Au Composite Films......44
4.2.1. As Deposited In/Au Composite Films............44
4.2.2. In/Au Composite Films Annealed in Air......... 59
4.2.3. In/Au Composite Films Annealed in Hydrogen ....83

4.3. Effects of In and In203 on the
Microstructural Evolutions................... 102
4.3.1. Effects of In and In 03 on the
Microstructural Evolution of In/Au
Composite Fi 1 lms .............................. 102
4.3.2. Effects of In 03 on the Microstructural
Evolution of In/Au Composite Films..........104


5.1. Derivation of the Critical Conditions ........113
5.1.1. Grain Boundary............................... 113
5.1.2. Grain Boundary Vertex........................ 116

5.2. Parameters to Control the Groove Evolution
in Au Thin Films............................. 123

CHAPTER 6. CONCLUSIONS.................................. 130

REFERENCES ................................................... 132

BIOGRAPHICAL SKETCH .......................................... 139

Abstract of Dissertation Presented to the Graduate School of
the University of Florida in Partial Fulfillment of the
Requirement for the Degree of Doctor of Philosophy





Chairman: Rolf E. Hummel

Major Department: Materials Science and Engineering

Preliminary studies have shown that grain boundary grooving in

thin Au films is prevented by inserting an indium underlay between gold

film and substrate. The objectives of this work was to investigate the

mechanisms for the prevention of grain boundary grooving in In/Au

composite films by comparing the microstructural evolutions of pure Au

films with In/Au composite films during isothermal annealing.

Microstructures were characterized in terms of grain size, grain

size distribution, texture and surface morphology utilizing TEM, XTEM,

SEM and X-Ray diffraction. The chemical reactions and the distribution

of the phases were monitored by SAD in TEM, and by AES sputter


It was found that the principal mechanisms to prevent grain

boundary grooving in In/Au composite films are as follows:

1) Indium underlays modify the microstructure of Au films by

randomization of the orientation of the grains, refinement of

the grain size, uniformity of the grain size distribution and

roughening of the surface of the gold films.

2) Indium is redistributed on gold films and forms In203 on the

free surface and in the gold film during air annealing.

3) The In203 on the surface "caps" the surface of gold film and

limits the mass transport process during annealing.

4) The In203 in the gold film, presumably residing near grain

boundaries, impedes the grain growth by pinning the grain

boundary migration.

Additionally, the critical conditions, where grain boundary grooves

reach the substrate, were derived from geometrical consideration. The

critical ratio of the thickness to grain diameter for Au films was

calculated to be 0.07 using experimental values for the surface energy

and the grain boundary energy from literature.


The continuing trends of increasing complexity of integrated

circuits have placed severe demands on the improved reliability of the

microelectronic devices. A significant fraction of reliability

problems in the microelectronic devices have been attributed to the

failures in thin film metallizations that interconnect individual

components. Aluminum is the most widely used principal metallization

material for providing necessary contacts and interconnections.

Despite the considerable improvements to increase the lifetime by

alloying with copper and silicon, aluminum metallizations are still

riddled by one dominant failure mechanism, such as electromigration,

which limits the reliability of the integrated circuit interconnections


The future trends towards further miniaturization with an

attendant decrease in the line width and the thickness of the

metallizations favor a shift from aluminum metallization to gold

metallizations because of their greater resistance to corrosion and

lower susceptibility to electromigration. Gold metallizations have

been known to have a longer life time under DC stressing, which might

be attributed to the difference in activation energy for electro-

migration being about 50 to 200% higher than that of aluminum [6-9].

However, it has been shown that several other failure mechanisms,

such as grain boundary grooving and thermotransport, also contribute

significantly to the failure of gold metallizations. In particular,

the grain boundary grooving which is driven by capillarity (surface

tension) has been known to be quite detrimental to the stability of

gold thin film metallizations [10-13].

A grain boundary groove, or thermal groove, is formed when a grain

boundary meets the free surface and its depth deepens with time. In a

sufficiently thin metallization film, this groove may reach the

substrate, leading to voids in the film or to an open circuit. Even

when the grooves do not extend all the way to the substrate, the

decrease in cross sectional area will cause increased Joule heating and

current crowding which results in a concommitant increase in

temperature and electromigration rate, so that the groove area will be

the preferred sites for the failure [14-16].

Hummel et al. have shown that indium underlays between gold and

the substrate inhibit the development of grain boundary grooving during

heat treatment in gold thin film metallizations [17,18]. Although

stabilizing effects of the indium on the grain network of gold films

have been are observed, the conclusive explanation why indium underlays

impede grain boundary grooving has not been provided.

The objective of this dissertation is to investigate the mechanism

for the inhibition of grain boundary grooving of gold thin films by the

addition of indium underlays.

In order to achieve this objective, microstructure evolutions,

chemical reactions and re-distribution of the phases during deposition

and the isothermal heat treatment are monitored. Microstructures are

characterized in terms of grain size, grain size distribution, texture,

and surface roughness, utilizing TEM (Transmission Electron

Microscope), SEM (Scanning Electron Microscope), X-ray diffraction and

XTEM (Cross-Sectional Transmission Electron Microscope). The chemical

reactions such as oxidation and the intermetallic compound formation

and the distribution of the phases are accomplished by SAD (Selective

Area Diffraction) in TEM and AES (Auger Electron Spectroscopy) sputter


Three types of specimens have been prepared: 1) pure Au films, 2)

In/Au composite films annealed in air, 3) In/Au composite films

annealed in hydrogen gas. The effect of indium underlays upon the

microstructure of gold films during deposition is studied by comparing

the microstructure of "as deposited" pure gold films with the

microstructure of "as deposited" In/Au composite films. The effect of

In203 on a microstructural evolution during annealing has been

investigated by comparing the microstructures of annealed pure Au films

with the microstructures of annealed In/Au composite films.

Additionally, some theoretical considerations that include

instability conditions for thin films caused by grain boundary grooving

as a function of the grain size and the groove angle and the parameters

which might change the evolution of grain boundary grooves in gold thin

films are discussed.

It will be shown that In203 on the surface and in the grain

boundary are responsible for stabilizing the surface structure and for


inhibiting the grain growth, respectively, which in turn reduce the

grain boundary grooving of gold thin films during heat treatment.


2.1 Brief Review of Driving Forces for the Mass Transport

There are several different kinds of driving forces to cause mass

transport in thin films. In general, the action of a driving force on

atomic motion can be written [19]

Ji = Di VCi + F ( Di / f k T ) Ci

where Ji = the atomic flux of ith constituent

Ci = the concentration of ith constituent

Di = the diffusivity of ith constituent

f = the correlation coefficient appropriate to the mechanism

responsible for the motion, and

kT has its usual meaning.

The first term represents the action of diffusion. The second term

shows the influence of an external driving force.

Driving forces derived from the electric current (electro-

transport) and by the thermal gradient (thermotransport) are briefly

reviewed below. Mass transport driven by capillarity may also be

important in thin films; capillarity effects will be discussed

separately in section 2.2.

2.1.1. Electrotransport (Electromigration)

Electrotransport, or alternatively electromigration, is the

phenomenon of mass transport arising from the driving force of electric

field or electric current. In electrotransport, the driving force is

considered to be the sum of two effects: 1) electrostatic interaction

between the electric field and the ionic core of the atoms, and 2) a

friction force between these ions and the flowing charge carriers,

which is often called the electron wind force. Accordingly, the

driving force can be expressed by [20]

F = Z* |e E = ( Zel + Z wd ) e| E

where Z = the effective charge number

e = the charge of an electron, and

E = the electric field.

From the theoretical considerations, Z* is given by

Z = + z [1 y (pd/Nd) (N /p) |m* / m*]

where z = the electron atom ratio of the material

(pd/Nd) (N /p) = the ratio of the specific
resistivity of the moving defects

to that of the lattice

m = the effective mass, and

y = a numerical constant = 0.5.

Usually the electron wind force is dominant and thus the mass

transport is presumed to occur in direction of the electron flow.

The flux of atoms due to electrotransport is given by

Ji = Ni Di / k T Z* e E

where Ni is the number of ith atoms per unit volume.

2.1.2. Thermotransport (Thermomigration)

Thermotransport, or thermomigration, is the transport of mass due

to a temperature gradient. The driving force for the thermotransport

can be expressed by [21]

Fi= Qi/T V T

where Qi = the heat of transport of ith constituent, and

VT = the temperature gradient = 3T/3x + 3T/Dy + 3T/az.

Q is the energy which flows per unit mass

transported; it has three contributions [22]:

Q = Qint + Qel + Qph

where Qint = the intrinsic contribution due to the motion

of atoms under the static temperature gradient,

Qel = the contribution due to the electron-moving
atom interaction under thermo electric field,


Qph = the contribution due to the heat carrier
interaction (phonon scattering).

The atomic flux along the temperature gradient can be written as

S= Ni Di Qi / k T2 V T.

The net transport of matter will be toward lower temperature if Qi

is positive and toward higher temperature if Qi negative.

2.2. Capillary Induced Mass Transport

2.2.1. Introduction

When polycrystalline metals and alloys are heated at an elevated

temperature the shape of the external surface will be altered. The

changes in surface morphology are caused by mass transport driven by

the capillarity; they have been analyzed rigorously in the theories of

thermal grooving and morphological instability. Thermal grooving

theory, or grain boundary grooving theory, deals with the formation and

the evolution of the groove which is formed at the intersection of a

grain boundary with a free surface. Morphological instability theory

deals with the conditions where the original geometry becomes unstable

due to perturbations driven by capillarity. Both thermal grooving and

morphological instability are caused by the surface curvature, which is

related to the chemical potential. However, the curvature, thus the

driving force is discontinuous at the grain boundary for the case of

the thermal grooving while it is continuous for the case of the

morphological instability.

In section 2.2.2., thermal grooving theories involving several

different mass transport mechanisms are presented. Morphological

instability theories for different geometries are given in section


2.2.2. Theories of Thermal Grooving

A thermal groove is formed at the intersection of a grain boundary

and the free surface as the surface energy and the grain boundary

energy establish an equilibrium satisfying the relation

2 ys sine = Y where e is the equilibrium groove angle, ys is the

surface energy and yg is the grain boundary energy (Figure 1).

The ultimate motivation for the formation of the thermal groove is

the minimization of the total free energy of the system by reducing the

interfacial free energy [23,24]. Although the grain boundary and the

surface may quickly achieve the proper equlibrium angle where they

intersect, the groove will generally continue to grow. Thermo-

dynamically, the growth will continue until the system reaches its

lowest free energy by eliminating the grain boundary. Mechanistically,

the growth occurs in response to the chemical potential gradient which

is caused by the curvature of the groove profile. The Gibbs-Thompson

formula relating the curvature and the chemical potential shows that a

curved surface has chemical potential that is different from a flat

surface. The gradient in chemical potential drives atoms from a point

of higher chemical potential to one of lower chemical potential, with

the result that the groove deepens with time [25].

Assuming isotropic surface properties, Mullins developed a theory

of thermal grooving in which mass transport occurred under the solitary

action of: 1) evaporation-condensation mechanism, 2) surface diffusion

mechanism, and 3) volume diffusion mechanism [26]. The differential

equations describing the evolution of the groove profiles are derived

using small slope approximation.




Figure 1. Profile of Thermal Grooving

The shapes of the groove profiles and the dimensional charac-

teristics produced by each transport mechanism are shown in Table 1.

Table 1. Shapes of Groove Profiles for Various
Mechanisms of Mass Transport



Evaporation and Condensation

Surface Diffusion

No Maxima


Volume Diffusion

The shape of the profile depends upon m, which is the tangent of

the equilibrium angle, but is independent of time. Furthermore, the

linear dimensions of the fixed shape are proportional to tI/2, tl/4 and

t1/3 for evaporation-condensation mechanism, surface diffusion

mechanism and volume diffusion mechanism, respectively.

The three mechanisms of mass transport may operate simultaneously

[27,28]. The relative importance of each mechanism depends upon the

experimental conditions, which may affect the vapor pressure, the

diffusivity and the diffusion distance. It is found that surface

diffusion is the dominant process in the initial stage whereas the

volume diffusion and evaporation-condensation mechanisms are important

at later stages.

The predictions of Mullins theory fit reasonably well with

computer calculations and the experimental data [29-31]. Huang and Lin

have extended the Mullins theory to cases where ridges or notches are

present on the initial surface [32]. It is shown that the grooving

takes place only if the initial slope of the notch is less than /2ys ,

or if the initial surface has a ridge at a grain boundary.

The thermal grooving theory in a system where a loss of matter due

to free evaporation or corrosion occurs simultaneously with surface

diffusion has been treated numerically [33,34]. It is shown that the

groove evolution is not steady state; thus the groove shape is time

dependent and also that the groove depth and the width do not follow a

t1/4 law. It is suggested that free evaporation and corrosion effects

should not be ignored for temperatures greater than 0.65 Tm or for

lower temperatures when the reaction between the adsorbed gases and the

specimen occurs.

Without the assumption of the small slope approximation and the

steady state evolution, a numerical analysis of the groove evolution

for a plane and a wire has been performed [35]. It is predicted that

the groove evolution can lead to the failure of the wire only by

surface diffusion, which indicates the significant effect of geometry

on the evolution of thermal grooving.

Most thermal grooving theories for thin films deposited on a

substrate have been studied in the system where the films are stressed

at high current densities. The thermal groove profile therefore is

influenced, not only by temperature but also by electromigration.

Using the approach of Mullins, Ohring has derived an equation for the

groove profile caused by the applied electric field [36]. It is shown

that the unidirectional electromigration flux establishes the asymmetry

in the profile which is characterized by a fixed shape where linear

dimensions change with time as tI/4. In the case where the temperature

gradient is dominant the groove depth varies with time as tl/3. When

the divergence of the mass flux induced by the electromigration is not

equal to zero (V J / 0 ), a local mass depletion or accumulation

occurs, which accelerates thermal grooving. The electrotransported

mass is depleted at or carried from a grain boundary groove, upsetting

the equilibrium angle and thereby establishing a driving force which

promotes further groove deepening.

Effects of annealing and electromigration on grain boundary

grooving has been investigated in a bicrystal film which is

isothermally annealed and subsequently stressed at a high current

density [37]. The combination of fluxes from capillarity-induced

surface diffusion and electromigration-induced grain boundary diffusion

is shown to cause open circuit failures by developing severe mass

depletion and accumulation at the grain boundary groove.

2.2.3. Theories of Morphological Instability

The morphological changes of the surface driven by capillarity may

cause the instability of the original geometry. The break-up of a long

cylinder into spheroidal particles, the blunting of a field emmision

cathode and the rupture of a wire by thermal grooving are examples of

this phenomenon [38-41].

Instability conditions are strongly dependent upon the geometry,

dominant mass transport mechanism, the surface energy, the amplitude

and the direction of the perturbation. While small perturbations in a

nearly planar surface will decay with time to make the flat surface

stable, some perturbations in a nearly cylindrical or spherical surface

increase their amplitudes with time to make the original surface


First-order perturbation analysis has been treated by Rayleigh for

inviscid fluids assuming the cylinder of infinite length [42].

Analogous treatment has been performed by Nichols and Mullins for

different geometries [43]. For an infinite cylinder with longitudinal

perturbation by surface diffusion the following instability conditions

are obtained.

For any longitudinal perturbation of wavelength less than o0,

which is 2TrRo, the cylinder is stable, i.e. such perturbations decay

with time. For A>\ the cylinder becomes unstable, i.e. such

perturbations increase in amplitude with time. For X= XM, the wave

length at which the perturbation develops the maximum value, the

cylinder breaks up into a line of particles with spacing xAM

The mechanism of break-up of plate-like particles has been also

predicted; a platelet will first develop a doughnut-shaped rim, which

is essentially a curved cylinder by a bulging process along its edges.

This rim will in turn break up into a ring of spheres separated by AM,

by the process discussed above. The experimental support for this

mechanism has accomplished by Yen and Coble [44]. The spherodization

of semi-infinite rods has been treated with a numerical method in which

the initially uniform circular section is spheroidized into a series of

egg-shaped particles. Nichols has analysed a finite cylinder with

hemisphere ends and has calculated the critical length to diameter

ratio (L/0D) of 7.2. Below this value only one spheroidal particle

results; above it the particle breaks up into two or more parts [45].

The energetic and the kinetics of the instability of a thin film

where an initially uniform film, which has been deposited on a

substrate, breaks up into an array of beads or islands, has been

studied by Srolovitz and Safaran [46,48]. They have calculated the

instability condition for a thin film with respect to large amplitude

perturbations, namely holes and islands using non linear perturbation

analysis. To evaluate the shape of the film, a quasi-static

approximation is employed, i.e. determine the shape of the film by

minimizing the energy of the system with respect to film shape. It is


found that holes which exceed a critical size, which is proportional to

the ratio of the thickness to equlibrium contact angle, grow and

eventually disconnect the film. For a potential source of these large

perturbations the grain boundary groove, especially at the vertex,

where three grain boundaries meet, is proposed.


3.1. Film Preparation

The pure Au and the In/Au composite films were prepared by using a

high vacuum deposition technique.

The substrates on which the films were deposited were optically

flat, commercial grade fused quartz plates (2.5 x 3.5 x 1 mm3), which

were polished on one or both sides. Films were deposited on the polished

side. Impurities of the quartz substrate were as follows:

A1203 : 60 100 ppm, Fe203 : 3-5 ppm, K20 : 3 ppm

Ti02 : 5 ppm, Na20 : 4 ppm, Ca : 0.5 ppm, B : 0.3 ppm

Prior to deposition the substrates were examined with the aid of

an optical microscope to discard the ones with gross defects such as

scratches, seeds and bubbles. The substrates were then ultrasonically

cleaned in the following sequences:

1) Alconox detergent (7.3% phosphorous by weight, pH 9.0 9.5)

and water in order to remove water-soluble and some organic

contaminants including oils, greases, soils and carbon products.

2) Micro and de-ionized water in order to remove water-soluble

contaminants which were introduced by the detergent.

3) Reagent grade acetone, a semi-polar solvent, acting as a vapor


4) Reagent grade alcohol, a less polar solvent containing fewer

inherent impurities used to continue the cleaning action initiated by

the acetone wash.

The substrates were then subjected to a stream of dry air and re-

examined optically for flaws. The substrates were mounted in spring-

loaded holders that enabled the placement of a molybdenum mask over the

substrate. These holders were then placed in the vapor deposition

chamber for film deposition.

Prior to opening the vacuum chamber to load the substrates and the

source materials, several heating tapes were wrapped around the chamber

in order to minimize impurity adsorption along the walls and the

constituent parts of the chamber. After the appropriate substrates and

source materials were placed in the chamber, the system was immediately

closed and pumped down. A mechanical pump was used to bring the

pressure in the chamber to 1.5 x 10-1 Torr within 5 minutes. This cut-

off pressure and the use of a zeolite-filled absorption trap minimized

the backstreaming of water and the hydrocarbon molecules from the

mechanical pump oil back into the chamber. Laboratory grade nitrogen

was then bled into the chamber to reduce the partial pressure of oxygen

and water vapor. This process was repeated several times before

turning on the adsorption pump. A liquid N2 cooled absorption pump

then reduced the chamber pressure 1.5 x 10-1 Torr to 2 x 10-3 Torr

(2.6 x 10-1 Pa ) by particularly reducing the partial pressures of

nitrogen, carbon dioxide, water vapor and residual hydrocarbons. After

outgassing the titanium filament of a sublimation pump, the adsorption

pump was closed off. Two sputter-ion pumps, in conjunction with the

sublimation pump, brought the system pressure down to the deposition

pressure within 3 to 5 hours. Base pressures prior to deposition were

in the range of 4-6 x 10-8 Torr (5-8 x10-6 Pa).

The Au was evaporated from a Mo boat that was resistively heated

by 180 amps generated by a 3 KVA power supply. The purity of Au wire

was 99.999%. A sufficient amount was loaded to prevent the poor

wetting of Au. Deposition was carried out with the ion pump on.

Pressure during the deposition was maintained below 1 X 10-6 Torr (1.3

x 10-4 Pa). The substrates were held at room temperature but the

radiation heating during the evaporation raised the substrate

temperature up to about 45 deg C, which was measured by the Alumel-

Chromel thermocouple attached to the substrate. Deposition rates of Au

varied from 2 A/s to 9 A/s with an average rate of approximately 3.3


For In/Au composite films, indium of 100-150 A thickness was

deposited first on the quartz substrate by applying 120 amps to a Mo

boat. Indium wire of 99.99% was used as the source material. Base

pressure prior to deposition of In was below 6 x 10-8 Torr (8 x 10-6

Pa) and the pressures during deposition were 1-4 X 10-7 Torr (1.3-5 x

10-5 Pa). Deposition rates ranged from 2 A/s to 8 A/s with an average

rate of 4.8 A/s. After deposition, about 2 to 15 minutes were required

for the system to reach the base pressure, 6 X 10-8 Torr (8 x 10-6 Pa)

for the subsequent deposition of Au. Gold film of 700-800 A thickness

then was deposited on the indium-covered quartz. Pressures during

deposition were 1-4 X 10-6 Torr (1-5 x 10-5 Pa) and the deposition

rates ranged from 2 to 8 A/s. The thickness of the films was measured

in situ with a piezo-electric film thickness monitor calibrated for


Upon completion of the deposition, the ion pump was turned off and

the films were stored in the vacuum chamber for more than 6 hours for

cooldown. The films were inspected with an optical microscope for

probable defects prod-uced during the deposition process such as

pinholes, scratches and noticeable contaminants. The specimen with

defects were discarded.

3.2. Isothermal Annealing

Three annealing temperatures, namely 300, 400 and 500 C, were

chosen to approximate those temperature produced during the typical

electromigration tests. A 500 C anneal was chosen to accelerate any

thermally activated processes. The annealing time, unless specified,

otherwise was 1 hour.

The specimens were heated in a tube furnace where the temperature

and the atmosphere were controlled. The furnace was pre-set to the

annealing temperature and allowed to stabilize for 2 hours before the

specimen was inserted.

Annealing in an oxidizing atmosphere was accomplished by heating

the specimens in air. No purging step was taken. Upon completion of

heating, the specimens were removed from the hot zone and were allowed

to cool in air at the cold zone for 30 minutes.

Annealing in a reducing atmosphere, that is in H2, required

several precautions due to potential oxidation of films during the heat

treatment. Preliminary experimental results showed that the use of an

inert gas, for example, Ar (99.999 % pure), for purging and cooling

caused some oxidation. Therefore, only ultra high pure H2 (99.999%)

was used for purging, heating and cooling. For heating and cooling,

the specimens were inserted into the hot zone and pulled out into the

cold zone with a stainless steel rod. To avoid any oxidation, the

specimen relocations were carried out while H2 was flowing. The H2

pressure during the heating was maintained to 4 psi (2.7 x 104 Pa).

The flow rate of H2 was controlled by monitoring the gas bubbles from

an oil bath.

3.3. Characterization of Film Microstructures

3.3.1. Scanning Electron Microscopy

Surface topography of pure Au and In-alloyed Au films were

examined in a JEOL scanning electron microscope JSM 35 CF. The

specimens were mounted on an aluminum specimen holder (5mm in diameter)

with silver paint and were dried for 24 hours before examining them in

the microscope. No conducting coatings were applied.

The grain size and the grain size distribution were measured using

a line intercept method, which allows calculation of the size of the

grain from counting the number of intersection of a line of known

length with the grains.

3.3.2. Auger Electron Spectroscopy

An analysis of the atomic species contained in the films as well

as the sputter profiles for In/Au composite films were accomplished by

using a Scanning Auger Microscope. Up to 8 specimens were mounted on a

carousel holder in a single loading and sequentially rotated in front

of the analyzer. The critical conditions for the analysis were as


E-Gun beam voltage: 3 KV

Peak to peak modulation amplitude: 3 eV

Sweep rate: 3 eV/s

Detection sensitivity: 25 V

Electron multiplier: 1000 V

Base pressure: <1 X 10-9 Torr (1.3 x 10-7 Pa)

Pressure, Ar ion sputtering: 4 X 10-5 Torr (5.3 x 10-3 Pa)

Ion Gun beam voltage: 2 KV

Ion Gun emission current: 5 mA.

Identification of peaks in the Auger spectrum was accomplished

through the combined use of a chart of Principal Auger Electron

Energies and the standard spectra in the handbook [49]. Transition

peaks of NVV (69 eV), MNN (404 eV) and KLL (510 eV) were used to
identify Au, In and 0 elements, respectively. Peak to peak heights for

these peaks were measured and plotted with respect to the sputter time.

The quantitative analysis, atomic concentration (%), was obtained

from the calculation using the formula:

Cx = Ix/Sx /(Ia/Sa)

where Cx = atomic concentration, %

Sx = relative sensitivity of element X

Ix = peak to peak height of element X

Sa = relative sensitivity of element A

Ia = peak to peak height of element A.

The relative sensitivity of element X was calculated from the peak

to peak heights of element X and silver in the handbook. The relative

sensitivities of Au, In and oxygen calculated were 0.417, 0.93 and

0.484, respectively. Atomic concentration for each elements were

plotted with respect to the sputter time and these profiles were

normalized to the thickness of Au films.

3.3.3 Transmission Elecron Microscopy

The films were examined in a JOEL 200CX STEM operated at 200 KV.

A given film for conventional top view transmission electron microscopy

was floated off the quartz substrate in a 40 % concentrated

hydrofluoric acid and rinsed in DI water. The floated film was then

put on a standard 3mm copper grid with 100 mesh and dried in the

dessicator before examination in the microscope.

A grain structure of the film was obtained by image mode and the

diffraction patterns were obtained by SAD (Selective Area Diffraction)

mode. The size and the distribution of the grains were measured by

line intercept method. Phase identification was accomplished by

measuring the distance of the diffraction spot or ring from the center

(transmitted beam) and using the formula [50]:

Rx d = X xL

where R = the distance of the diffraction pattern

d = the inter-planar spacing of a phase

x = the wave length of the electron

L = the camera constant of the microscope.

3.3.4. Cross Sectional Transmission Electron Microscope

The cross sectional TEM refers to examining the specimen in cross

section by TEM. That is, the surface normal of the specimen is made

perpendicular to the electron beam. The preparation of cross section

specimens of TEM has been described by several investigators mainly for

silicon based materials [51-52]. The same type of preparation

technique was adopted for annealed In/Au composite films, although

several modifications had to be made for following reasons:

1) A very thin area is required for TEM examination for an Au

based alloy because Au has a very high atomic number, which means that

it is very difficult for electrons to transmit the specimen.

2) Since sputter rates of Au and quartz are supposed to be quite

different, a uniform thinning during ion milling is difficult.

3) As Au has very poor adhesion to most surfaces, it is critical

to apply an adhesive compound which has a good adhesion and the

mechanical strength.

The specimens were prepared in the following sequence:

1. Sectioning

The specimens were sectioned into rectangular slabs that measured

2.5 mm in width and 9 mm in length. A diamond low speed watering saw

was used for this purpose. The slabs then were degreased and

thoroughly cleaned by ultrasonic scrubbing in acetone.

2. Gluing

Two slabs were glued face to face into a composite by applying a

thermosetting adhesive epoxy compound in between them. The epoxy

compound was prepared by mixing 16g of EPOK 812, 14g of NMA (Nadic

Methyl Anhydride) and 0.58g of DMP-30 (Tris dimethylaminomethyl

phenol). The compound was very fluid so that a smooth and thin layer

could be easily applied.

The composite was then inserted into a vise and pressure was

applied by moving the crosshead toward the end plate which was made out

of teflon. The vise which was designed to hold a composite together

under even pressure during curing was then put into a convectional oven

and cured at 70 deg C for 8 hours. Subsequently, the vise was allowed

to cool for 45 minutes. Then the composite was removed from the vise

by releasing the pressure. The glue on the surface of the composite

was removed by slightly grinding with the sand paper.

3. Molding

For protection and handling purposes, the composite was embedded

into the epoxy compound by molding. This was accomplished by filling

up a rubber mold with epoxy compound, followed by inserting the

composite into the mold. Care was taken not to introduce any bubbles

during inserting the composite into the mold. The mold was then put in

an oven and cured at 70 deg C for 8 hours. After curing has been

completed, the mold was allowed to cool for 1 hour. The molded epoxy

bar which contains the composite inside was removed then from the mold.

4. Slicing and Mounting

The molded epoxy bar was sliced into several discs of 0.3 mm

thickness with a diamond watering saw. In most cases, more than 6

discs were cut from a single epoxy bar.

Up to three discs were then simultaneously mounted onto a disc

holder (1 cm in diameter and 2 cm in height) on a hot plate using a

crystal-bond wax.

5. Grinding

The disc holder was then inserted into a DISC Grinder by which the

height of the specimen grinding was controlled with an inclement of 10

m. The discs were ground by using a low speed rotating wheel using 600

grit self adhesive sand paper. Less than 3000 rpm was used as rotating

speed. Plenty of water was supplied during grinding in order that the

wax did not overheat and melt. The thickness of the disc was

frequently checked with a micrometer while the discs were still

attached to the disc holder. The grinding was terminated at a

thickness of about 150 Pm. Both the disc holder and the DISC Grinder

were cleaned thoroughly with a detergent and water. They were gently

rubbed by a cotten swab during cleaning to remove any surface


6. Polishing and Mounting

Three successive polishing steps were carried out using 1 um

diamond paste, 0.3 um alumina and 0.03 pm alumina on separate polishing

cloths. Rotation speeds were maintained below 2000 rpm. It was

critical to clean the disc holder and the DISC Grinder thoroughly after

each polishing step in order not to introduce scratches during


Upon completion of the final polishing and the cleaning, the disc

holder was put on a hot plate and each disc was turned over to thin the

other side. More crystal-bond wax was applied if necessary.

7. Grinding

The other side of the discs were ground the same way as in step 5

until the thickness was less than 100 pm. Subsequent thinning the

discs from 100 pm to 20 pm was accomplished by grinding them manually

with plenty ofwater on sand paper. Special care was taken not to

apply excessive pressure during grinding. The cleaning was followed by

using the same detergent and rubbing the specimen with a cotten swab.

8. Polishing and Demounting

Final polishing was accomplished with 1 pm diamond paste on

polishing cloth. A 1000 rpm rotating speed and short polishing time

were used to avoid breakage of the disc. Cleaning followed the same

procedure as in step 7.

The disc holder was then put on a filter paper and was soaked in

acetone for 20 minutes to remove any wax. As the discs were demounted

from the holder, they remained on the filter paper, which was carefully

removed and dried. The discs were very fragile and were handled with

utmost care.

10. Gluing

In order to improve handling in the subsequent processes and the

microscope examination, the discs were glued on a standard 3 mm copper

grid having a single hole of 1 mm diameter. This was accomplished by

applying EPOXY(Double/Bubble), Harman Co., around the edge of the hole

and putting the disc on the grid. The position of the disc was

adjusted to make sure that the interface was located at the center of

the grid. The copper grid was then cured in an oven at 100 deg C for 2


11. Ion Milling

The final thinning was accomplished by ion milling the specimen in

a ion milling instrument with a terminator, where argon ions were

accelerated with 5 KV potential, a specimen current of 3 mA and a

specimen tilt of 13 degrees. The vacuum level was about 5 x 10-6 Torr

(6.6 x 10-4 Pa). The terminator measured the specimen current every 10

seconds and shut off the ion milling automatically when the current

reached the pre-set current. A total of 8 to 10 hours were taken

before the terminator stopped the milling. Once perforation was

achieved, the tilt angle was reduced to about 11 degrees for another 15

to 30 minutes ion milling without using the terminator.

Most of the cross sectional TEM examination were accomplished in

Hitachi H-800 TEM (200 KV). Some of the specimens were investigated by

EDX (Energy Dispersive X-ray Spectroscopy) analysis for the quantitative

information of the elements.


3.3.5 X-Ray Diffraction

X-ray diffraction patterns for pure Au and In/Au composite films

were obtained from an automated powder diffractometer operated at

45 KV. The position and the integrated intensity of the diffraction

peak was automatically measured and recorded on the chart by data

processing computer.


The following experiments were undertaken to study the effect of
indium underlays upon microstructural evolutions of gold thin films:
1) annealing of pure Au films at 300 C, 400 C and 500 C
for 1 hr in air
2) annealing of In/Au composite films at 300 C, 400 C and 500 C
for 1 hr in air
3) annealing of In/Au composite films at 300 C, 400 C and 500 C
for 1 hr in hydrogen gas
4.1. Microstructure Studies of Pure Au Thin Films
4.1.1. As Deposited Pure Au Films
a) Grain structure
Figure 2a depicts the grain structure of an as deposited Au film
examined by TEM. This micrograph reveals a mixture of fine and coarse
grains. It is noted that, apart from the predominant grains whose mean
grain diameter is about 1000 A, much bigger grains, of a few thousand A
in diameter can be easily distinguished.


The presence of such large grains may suggest the preferential

growth of large nuclei along certain crystallographic direction during

deposition. The nucleation and the growth of gold evaporated on the

non reacting surface, known as "non-wetting growth," has been known to

form large nuclei because of the high mobility of adsorbed atoms,

called adatoms, along the surface [53]. Similar grain structures have

been reported by other investigators in vapor deposited gold thin films

although the mean grain size was found to be different probably due to

different deposition conditions [54,55].

b) Surface morphology

Very little detail of the surface structure were resolved by SEM

as shown in Figure 2b, which suggested that the as deposited pure Au

film had a very smooth surface. The surface roughness in the vapor

deposited thin films is determined by the statistical process of

nucleation and growth and the surface mobility of adatoms during

deposition [56]. As the surface mobility increases, the tendency to

have a smooth surface increases since the condensation can occur

preferentially at the concavities and thus smoothens the surface.

The surface mobility of adatoms increases with increasing kinetic

energy of adatoms and the substrate smoothness. The effect of kinetic

energy on the surface mobility of evaporated gold adatoms has been

studied by Chopra [57]. He showed that critical nuclei can move a

considerable distance with the momentum imparted by gold adatoms.

The surface of the substrate used in the present experiment was

polished. A very flat surface with little irregularities was observed

by the examination of the cross section of the substrate in TEM. The

surface smoothness of the as deposited pure Au films, therefore, is

believed to result from the smooth surface of the substrate and the

high kinetic energy of gold atoms during deposition.

c) Texture

X-ray diffraction patterns revealed very strong (111}, {222} peaks

and a weak {311} peak (Figure 3). SAD patterns revealed that the

intensity of {220} ring was greater than that of the {111} ring (Figure


The preferred orientation, or fiber texture, in the film can be

determined by comparing the X-ray and SAD diffraction patterns with the

diffraction pattern of the randomly oriented gold powders which is

listed in JCPDS (Jointed Committee of Powder Diffraction Standard).

The randomly oriented gold powders have the most intensive {111} and

moderately intensive (200}, {220}, {311} peaks as shown in Table 2.

Therefore, the very strong (111} and (222} peaks in the X-ray

diffraction patterns suggest a (111) preferred orientation in the film.

Furthermore, since X-ray diffraction does occur from the crystals which

lie parallel to the substrate surface, it is suggested that the (111)

orientation is parallel to the substrate surface [58].

SAD patterns complement the presence of a (111) texture in the

film. The fact that the intensity of the {220} ring is greater than

that of {111} ring suggests that the incident electron beam direction

is likely to be [111] direction since electron diffraction in TEM does

occur'from the crystals which lie parallel to the incident beam




Figure 2. Photomicrographs of As Deposited Pure Au Films :
a) Transmission Electron Micrograph
b) Scanning Electron Micrograph




Figure 3. X-Ray Diffraction Patterns of As Deposited Pure Au Films

Figure 4. SAD Patterns of As Deposited Pure Au Films

Figure 4. SAD Patterns of As Deposited Pure Au Films

Table 2. X-Ray Diffraction Patterns of Au Powders Listed in JCPDS

d, A











lhkl/lllI %

direction [59,60]. In summary, as deposited pure Au films have a (111)

texture parallel to the substrate surface.

The preferred orientation in thin films may develop at various

stages, e.g., nucleation, growth, epitaxial growth and heat treatment.

When the substrate has a dominating influence on the orientation, in

which the substrate energetically favors the adsorption of one

geometrical arrangement atoms in nuclei over another, the epitaxial

growth will be most probable [61,62]. The structure of the substrate

used in the experiment, fused quartz, is amorphous. Thus, the

preferred orientation during epitaxial growth is not likely to occur.

The temperature of the substrate during deposition was maintained below

45 degrees C, hence texture development due to heat treatment is also

disregarded. Therefore, the (111) texture in the as deposited pure Au

film was likely to develop during the nucleation and growth stages.

According to the capillarity theory of nucleation, the texture

will occur for the orientation which gives lower interfacial free

energy, and hence a lower free energy of formation for the critical

nucleus, and a much higher nucleation rate than any other orientation

[63]. The (111) orientation of gold has been reported to have the

lowest interfacial energy [64]. Thus, the preferred nucleation of

(111) might occur during the nucleation stage.

Several empirical rules concerning texture formation during grain

growth stage have shown the principle of geometric selection during

growth stage. That is, in the growth of randomly oriented nuclei, only

those grains will survive in which the direction of maximum rate of

growth approximately coincides with the normal to the crystallization

front [65]. According to this principle, the closest-packed planes

will lie parallel to the substrate if the surface mobility of adatoms

during deposition is high. Since (111) is the closest-packed plane for

gold, the higher rate of (111) orientation than other orientation

could be achieved.

In summary, whether it is attributed to a higher nucleation rate

or a higher growth rate, or due to a combination of the two, the (111)

texture which is parallel to the substrate surface is likely to occur

during deposition. The presence of a (111) texture is in agreement

with other experiments in which gold films evaporated on the amorphous

substrate had a (111) texture after deposition [66].

4.1.2. Annealed Pure Au Films

a) Grain structure

The grain structures of pure Au films annealed at 300 C, 400 C and

500 C examined by TEM are shown in Figure 5. One observes a coarse-

grained structure throughout. A significant grain growth occurred

during annealing as shown in Figure 6, in which the mean grain

diameters are plotted with respect to the annealing temperatures. The

wide scattering in the grain size indicates the wide grain size

distribution. This becomes wider as the annealing temperature

increases. This might be caused by the higher growth of textured

grains during annealing at the expense of the small grains.

0.2/ m

Figure 5. Transmission Electron Micrographs of Pure Au Films
Annealed at : a) 300 C, b) 400 C, c) 500 C





300 400 500
T EMP ., C

Mean Grain Diameters measured from Transmission Electron
Micrographs of Pure Au Films Annealed
at Various Temperatures

Figure 6.

b) Surface morphology

Very significant morphological changes in the surface resulting

from the grain boundary grooving and the grain growth are shown in

Figure 7. Thermal grooves along the grain boundaries or at the grain

boundary vertices, where three grain boundaries meet, were observed as

dark lines or dark spots in the SEM micrograph (Figure 8a) and as

bright lines or spots in the TEM micrograph (Figure 8b). The

development of grain boundary grooving during annealing is evident and

the degree of the grooving seems to increase as the annealing

temperature increases.

The most severe grain boundary groove was observed in a gold film

annealed at 500 C, where a hole formed at the grain boundary vertex as

shown in Figure 8b. The presence of a hole at the vertex was confirmed

by the fact that there was no change in the contrast of the bright spot

during tilting the sample up to 45 degrees in TEM examination. Since

the grain boundary grooving is caused by the capillary-induced mass

transport, which is a thermally activated process, the effect of

annealing temperature on the development of the groove seems quite

significant. Development of similar grain boundary grooving in thin

gold films annealed at elevated temperatures or stressed at high

current densities has been reported [67-68]. However, large holes due

to the loss of adhesion or gas bubbles entrapped during deposition were

not observed in the present experiment.





Figure 7. Scanning Electron Micrographs of Pure Au Films Annealed
at : a) 300 C, b) 400 C, c) 500 C

n I


0.2 m


Figure 8. Thermal Grooves in the Grain Boundaries and in the Grain
Boundary Vertices observed by : a) SEM, b) TEM

c) Texture

X-ray diffraction patterns revealed very strong {111} and {222}

peaks, which suggested the (111) texture parallel to the substrate

surface in the annealed pure Au films (Figure 9). The change in the

amount of the texture during annealing can be monitored by comparing

the intensities of {111} peaks. The change in the intensity of 111

peak for annealed pure Au films as a function of the annealing

temperature is shown in Figure 10. The increase of the relative

intensity as the annealing temperature increases suggests a favored

development of grains with (111) orientation during annealing.

The (111) textures in the annealed pure Au films are also evident

in the SAD patterns, in which the {220} ring is more intensive than the

{111) ring, indicating that the [111] direction is the incident beam

direction (Figure 11). The presence of a (111) texture parallel to the

substrate surface is in agreement with other experimental data, in

which the (111) texture was reported in the annealed gold films


4.2. Microstructure of In/Au Composite Films

4.2.1. As Deposited In/Au Composite Films

a) Grain structure

Figures 12a and 12b depict the grain structure of as deposited

In/Au composite films examined by TEM and SEM, respectively. They

reveal a uniform and fine-grained structure with a mean grain diameter

of about 700 A.








Figure 9. X-Ray Diffraction Patterns of Pure Au Films Annealed
at : a) 300 C, b) 400 C, c) 500 C









x 10-


300 400 500


Figure 10. Intensities of (111) Diffraction of Pure Au Films
Annealed at Various Temperatures




Figure 11. SAD Patterns of Pure
a) 300 C, b) 400 C,

Au Films Aneealed at :
c) 500 C

For comparison purposes, the grain structure of a pure indium film

(800 A in thickness deposited on a quartz substrate) was examined by

SEM (Figure 13). A rather uniform and fine-grained structure with a

mean grain diameter of 850 A was observed. It is easily noticed that

the grain structure of the In/Au composite film is qualitatively

similar to that of a pure indium film. This suggests that the

microstructure of In/Au composite films may be influenced by the

microstructure of the indium underlay film.

More careful examination shows some structural difference between

In/Au composite films and pure In film, however. These differences are

1) the mean grain diameter of pure indium film (850 A) is larger than

that of In/Au composite film (700A) and 2) the pure indium film has two

maxima, at about 700 A and 1000 A, in the grain size distribution

whereas the In/Au composite film has a single maximum at about 700 A.

These differences may be attributed to grain growth of the pure indium

film at room temperature and/or to the thickness effect on the grain

size. Since indium has a low melting temperature, 155.4 C, the

homologous temperature, T/Tm, at room temperature is about 0.7. This

is high enough to cause normal grain growth [70]. As some grains grow,

the grain size distribution with two maxima may result.

The effect of the film thickness on the grain size of deposited

thin films has been reported before for many systems [71-73]. The grain

size was observed to increase with increasing thickness. Since the

thickness of pure indium films examined was about 800 A whereas the

thickness of indium underlay in the In/Au composite film was 100 A, a

smaller grain size in the In/Au film than in the pure indium film is

expected. However, this is not generally a linear effect; i.e. the

grain size of a 100 A film is not necessarily 100 A, but more likely

about 700 A. Considering these two effects, it is suggested that the

microstructure of indium underlay affects the microstructure of the

subsequent gold film during deposition.

b) Texture

X-ray diffraction patterns revealed all possible diffractions of

gold, namely {111}, {200}, {220}, {311} and {222} (Figure 14). The

relative intensities of each peak with respect to the intensity of the

{111} peak are compared with those of randomly oriented gold powders in

Table 3. The trend is similar, which indicates that as deposited In/Au

composite films have a relatively random orientation of gold grains

with the tendency of the (111) texture. SAD patterns complement this

information by observing the {111} ring to be the most intensive

(Figure 15).

The relatively random orientation of grains in the as deposited

In/Au composite film suggests the absence of a preferred nucleation and

growth along a certain crystallographic direction. A decreased surface

mobility of adatoms has been reported to cause the random orientation

in vapor deposited thin films [74]. In summary, as deposited In/Au

composite films have the random orientation of gold grains which may

result from the decreased mobility of gold adatoms during deposition.

c) Surface morphology

The surface morphology of the as deposited In/Au composite film

examined by XTEM is shown in Figure 16. A surface roughness similar to



Figure 12. Photomicrographs of As Deposited In/Au Composite Films:
a) Transmission Electron Micrograph
b) Scanning Electron Micrograph

Figure 13. Scanning Electron Micrograph of Pure Indium
of 800 A Thickness




Figure 14. X-Ray Diffractions of As Deposited In/Au Composite Films

Table 3. Values of Ihkl/111 for As Deposited In/Au Composite Films







Au Powder






I111, %






Figure 15. SAD Patterns of As Deposited In/Au Composite Films

Figure 16. Cross Sectional Transmission Electron Micrograph of
As Deposited In/Au Composite Films

/ Au "A




n\ I -n


TIME, min

Figure 17. AES Sputter Profile of As Deposited
In/Au Composite Films





Some co-deposition of indium and gold could occur either by

radiation during gold evaporation or by heat conduction through the

electrodes. This is possible because of the low melting point of

indium. It requires only about 40% electric power compared to gold

evaporation. Furthermore, the molybdenum boats were located very close

together with one of their ends connected to the same electrode.

However, if indium on the gold surface resulted from co-deposition of

indium and gold, a uniform concentration of indium within the gold film

would be expected. This has not been observed. Therefore, co-

deposition of indium during gold deposition seems not to have occurred

to a 1arge extent.

Interdiffusion of indium and gold during deposition may be aided

by the kinetic energy of the gold atoms which impinge upon the indium

surface during gold deposition. The kinetic energy imparted in this

way could promote the rapid formation of a mixed layer or intermetallic

compounds during the deposition depending upon the deposition

parameters such as deposition rate and substrate temperature. For

example, the formation and the growth of a AuIn2 phase in Au/In thin

film couples during evaporation has been reported [76,77]. A growth

mechanism was proposed which entails the rapid diffusion of indium

along the grain boundaries of AuIn2 grains to the surface. An

interdiffusion during deposition, thus, might have occurred in this

experiment. However, no intermetallic compounds or any other phases

were observed by SAD in TEM and X-ray diffraction analysis.

Interdiffusion after deposition may occur if the indium in the

underlay diffuses out to the free surface at room temperature. Rapid

diffusion of indium at room temperature has been observed by several

investigators [77,78]. Furthermore, some enrichment of indium on the

surface after deposition in Au/In composite films and Au-In alloy films

monitored by RBS and AES has been reported [79,80]. Particularly, the

accumulation of indium on the surface during AES analysis has suggested

the tendency of indium surface segregation at room temperature.

Out-diffusion of indium after deposition is most likely to occur

because of high diffusivity at room temperature. The presence of

oxygen on the surface may suggest the formation of indium oxide and

thus the oxidation might be the driving force for the out-diffusion of

indium. However, surface segregation or a concentration gradient in

the Au film also could be a driving force.

4.2.2. In/Au Composite Films Annealed in Air

a) Grain Structure

The grain structures of In/Au composite films annealed at 300 C,

400 C and 500 C in air as examined by TEM are shown in Figure 18. The

change in the mean grain diameter as a function of the annealing

temperature is shown in Figure 19, where the error bars indicate the

standard deviations. A true identification of grain boundaries in the

images of heavily faulted polycrystalline films was difficult, and the

results may be subject to some error. Nevertheless, the formation of

larger grains at higher temperatures showing grain growth during

annealing was evident.





Figure 18. Transmission Electron Micrographs of In/Au Composite
Films Annealed at : a) 300 C, b) 400 C, c) 500 C




300 400 500

Figure 19. Mean Grain Diameters measured from Transmission Electron
Micrographs of In/Au Composite Films Annealed at Various
Temperatures in Air

b) Surface morphology

Very little surface structural changes and hence no significant

grain boundary grooves in the annealed In/Au composite films were

revealed by SEM examination (Figure 20). This will be substantiated

below when XTEM results will be discussed. The mean grain diameter

after annealing measured with SEM is about 700 A, which is the same as

in the as deposited In/Au composite film. This could be interpreted

that no grain growth occurred during annealing, which is contradictory

to the result of our TEM examination in which the grain growth was


The apparent difference between the surface structure observed by

SEM and the internal structure observed by TEM can be explained by

considering some model structures:

1) Model A involves small grains of 700 A in diameter and large

grains of few thousands A in diameter at the bottom of these small

grains as shown in Figure 21a.

2) Model B involves large grains of about 1500 A in diameter

having rough surfaces whose sinusoidal "perturbation" wavelength is

about 700 A as shown in Figure 21b.

Image formation in SEM is achieved by the topographic contrast

resulting from secondary electrons which are emitted from the specimen

surface [81]. The topographic contrasts, therefore, in model A would

be caused by the fine grain structure on the surface whereas those in

model B would be caused by the surface roughness. Since the dimension

of the small grains in model A and the wavelength of the sinusoidal


Figure 20. Scanning Electron Micrographs of In/Au Composite Films
Annealed in Air at : a) 300 C, b) 400 C, c) 500 C



Figure 21. Model Structures to explain the Apparent Difference
between the Surface Structure and the Internal
Structure : a) Model A b) Model B

"perturbation," i.e. surface roughness, are identical, the same surface

structure may be obtained by SEM for both models. Similarly, large

grains as well as small grains may be observed by TEM examination

because image formation in TEM is achieved by the diffraction contrast

of the transmitted electrons [50].

Thermodynamically, the structure of model A is less likely to

occur because it possesses a large amount of grain boundaries although

grain boundaries are expected to be eliminated during annealing in

order to reduce the total free energy by reducing interfacial energy.

Also, knowing that the as deposited In/Au composite films have the

columnar structure, in which the grains are extending through the

entire film (see Figure 16) the granular structure in model A, in which

small grains are embedded on the surface of large grains, is not likely

to develop during annealing [82].

On the other hand, the structure of model B is more likely to

occur because the as deposited In/Au composite film has a rough

surface. The evidence for the existence of model B structure, however,

was achieved by XTEM examination of In/Au composite films annealed at

400 C and 500 C. Figure 22 illustrates the cross section of the In/Au

composite film annealed at 400 C in which some large grains of about

2000 A in diameter and some surface roughness whose sinusoidal

"perturbation" wavelength and amplitude are 750 A and 100 A,

respectively, are clearly seen. Several small grains near the surface

may be attributed to columnar grains whose grain boundaries are located

under some angle with respect to substrate surface normal so that only

a section of the large grain is seen in the projected image of the

cross section of the film. A similar grain size and surface roughness

in the sample annealed at 500 C are shown in Figure 23.

In summary, the origin of the apparent difference between surface

structure and internal grain structure is believed to be due to the

surface roughness of In/Au films annealed in air.

c) Texture

X-ray diffraction patterns for annealed In/Au composite films are

shown in Figure 24. All the possible diffractions of gold, namely

{111}, {200}, {220}, {311} and {222} peaks, are observed. The relative

intensity of each peak with respect to the intensity of the {111} peak

is tabulated in Table 4. By comparing these relative intensities with

those of gold powders as listed in JCPDS, a relatively random

orientation of the grains was found. Little change in the relative

intensities with annealing temperature increase was found. This

suggests that no significant preferential grain growth along certain

crystallographic directions during annealing occurred.

d) Distribution of In and phase formation during annealing

When In/Au composite films are annealed in air, interdiffusion of

the components is likely to occur. Indium will be redistributed to

form a solid solution, intermetallic compounds, or oxides depending

upon the annealing conditions. Identification of the phases formed and

their locations in the film is essential to the understanding of the

microstrutural evolution during annealing.

The equilibrium binary phase diagram for the Au-In system (Figure

25) shows at least 6 intermetallic compounds [83-85], e.g.


Figure 22. Cross Sectional Transmission Electron Micrograph of
In/Au Composite Films Annealed at 400 C in Air


I quartz

0,0 5pm

Figure 23. Cross Sectional Transmission Electron Micrograph of
In/Au Composite Films Annealed at 500 C in Air


(a) 3112



222 311 200



222 311 20


Figure 24. X-Ray Diffraction Patterns of In/Au Composite Films
Annealed in Air at : a) 300 C, b) 400 C, c) 500 C

Table 4. Values of Ihkl/1111 for In/Au Composite Films Annealed at
Various Temperatures in Air, As Measured by X-Ray Diffraction

hkl/l111' %
In/Au Composite
Powder 300 C 400 C 500 C

111 100 100 100 100

200 52 30 14 13

220 32 80 100 72

311 36 40 16 18

222 12 6 11 8

Au-In Gold-Indium

10 20 30 40

Atomic Percentage Indium
50 60 70

80 90

1200 -

1100 -
1900 F


800 -
1400 F

700 -

600 -

500 -

400 -
00F -

300 -

200 -
300 F
100 -

Figure 25. Equilibrium Binary Phase Diagram for Au-In System

30 40 50 60
Weight Percentage Indium

Au7In, Au4In, Au31n, AugIn4, AuIn and AuIn2. The solid solubilities of

indium in gold at 500 C and at room temperature are reported to be

about 10 at. % and 7 at. %, respectively. No solid solubility of gold

in indium is reported.

In addition to the intermetallic compounds, three different indium

oxides, which are stable in the temperature ranges in which the

experiments were conducted, are expected to occur [86]. These oxides

are InO, In20 and In203.

The amount of indium in the present In/Au composite films is less

than 5 at. %, assuming that all the indium would diffuse into the gold

without any chemical reaction. This may suggest that indium is

distributed in gold as a solid solution. However, since the structure

of In/Au composite films consists of a 100% gold film immediately on

top of a 100% indium underlay, any phases mentioned above may be formed

depending upon the annealing conditions such as annealing time,

temperature, and atmospheres. The identification, the distribution and

the characterization of the phase or phases formed during annealing

have been accomplished with the combinational use of TEM, AES and XTEM

with EDX techniques.

All In/Au composite films annealed in air revealed the same

additional diffraction rings in SAD patterns as shown in Figure 26.

Based upon the structural informations of the intermetal lic compounds

and oxides available in the literature and in the JCPDS data, the phase

which produces the observed rings is identified as In203. The

orientation of In203 crystals corresponding to each diffraction ring is



Figure 26. SAD Patterns of In/Au Composite Films Annealed in Air
at : a) 300 C, b) 400 C, c) 500 C

also shown in Figure 27. The sharp ring patterns suggest that In203

has a fine-grained, polycrystalline structure.

The distributions of indium in In/Au composite films after

annealing as monitored by AES sputter profiling are shown in Figure 28.

High concentrations of In and 0 on the surface were observed in all

cases (300 C, 400 C and 500 C), which indicates the formation of an

indium oxide layer on the surface. This suggests that most of the

indium did diffuse from the underlay through gold film to the free

surface. The thickness of the oxide on the free surface was found to

grow as the annealing temperature increased as shown in Figure 29. The

thickness was measured by the sputter time required to reduce the

indium signal to 50%. In order that indium oxide can grow in excess of

a monolayer, the diffusion of either the oxygen or indium through the

oxide layer, or both is required. This will depend on the ease of

diffusion of the species through the oxide in a given annealing

condition [87]. We note that the diffusion rates are temperature

dependent following an Arehenius form. Thus, the thickness increase

with increasing annealing temperature is believed to be associated with

a higher diffusivity of indium or oxygen through the oxide layer at

higher temperatures.

The physical dimensions of the In203 grains on the surface were

studied by XTEM examination of the In/Au films annealed at 400 C as

shown in Figure 30. We observe a fine-grained oxide layer, whose mean

grain diameter and thickness are about 200 A and 100 A, respectively. To

further investigate the indium oxide layer on the surface, an EDX

analysis was carried out using the STEM mode at two different locations




Figure 27. In203 Phase Identified in SAD Patterns of
In/Au Composite Films Annealed in Air


(a) 0 A / \

S \ / In

15 30 45 60

SN Au- A

80 s /,

/ o\ /

0 \ J "\/

(b) / I
20 I\
2/ \\ / \

15 30 45 60
S P U T T E R TIME min

F- -P" Au

80- / \ c
/ \ /

S' /
(c) I-"-- .
o40 /\

015 30 45 60
S P U T T ER TIME ,rmin

Figure 28. AES Sputter Profiles of In/Au Composite Films Annealed
in Air at : a) 300 C, b) 400 C, c) 500 C






300 400 500

T E M P. C

Figure 29. Thicknesses of Surface Indium Oxide in In/Au Composite
Films Annealed in Air measured from AES Sputter Profiles

I i
Figure 30. Cross Sectional Transmission Electron Micrograph of
In/Au Composite Films Annealed at 400 C in Air showing
Indium Oxide at the Free Surface

in the film, i.e., the surface region and the middle region of the gold

grain. The beam size of the probe was less than 20 A in diameter. The

analysis showed that the intensity of indium signal (at 3.29 KeV) which

is obtained from the surface region (Figure 31a) was about 50% higher

than that from the middle region of a gold grain (Figure 31b). This

again suggests the presence of indium oxide on the surface.

We attempt now to estimate the thickness of the oxide layer. The

maximum thickness of In203 which is theoretically possible when all

indium from the underlay of 100 A thickness diffuses to the surface and

forms In203 is calculated as follows.

Let tin = thickness of indium, tox = thickness of In203, din

density of indium, dox = density of In203, Min = molecular weight of

indium, Mox = molecular weight of In203, A = unit area, Nin = number of

moles of indium and Nox = the number of moles of In203. The density of

indium can be.written

din = Min x in/A x tin


Nin = din x A x tin/Min.

Similarly, the density of the indium oxide can be used to write

Nox = dox x A x tox/Mox.

From the reaction

2 In + 3/2 02 = In203,

in 2 Nox-

Inserting the values available in the literature [86,88], the thickness

of indium oxide is obtained by

tox = 1/2 (din/dox) (Mox/Min) x tin

= 120 A.

The maximum thickness of In203 on the surface from a 100 A indium

underlay is thus about 120 A. The experimentally observed thickness,

which is about 100 A (Figure 30), is in reasonable agreement with the

result of this calculation. We need to note that our AES sputter

profile data also showed small amounts of indium (and oxygen) in the

gold film. If we consider some indium oxide left behind in the film,

the agreement between calculation and experiment becomes even more


The location of indium oxide in the film was difficult to see in

electro-optical micrographs because of the small amounts involved.

However, in one case, a grain boundary as shown in Figure 32 revealed

nearly the same contrast as the surface oxide, which might suggest that

indium oxide in the film is located along some grain boundaries.

Further confirmation of indium enrichment along the grain boundaries

with EDX analysis was not possible because the area of interest was

usually too small for a reasonable analysis.

In summary, by utilizing the combination of TEM, XTEM with EDX and

AES techniques, the formation of polycrystalline In203 with a thickness










Figure 31. EDX Analysis of In/Au Composite Films Annealed in Air
from Two Different Locations in the Film :
a) surface region, b) middle region




0.02 m

Figure 32.

Cross Sectional Transmission Electron Micrograph of
In/Au Composite Films Annealed at 400 C in Air showing
Indium Oxide along the Grain Boundary

and mean grain diameter of about 100 A and 200 A, respectively, has

been observed for In/Au composite films annealed in air.

4.2.3. In/Au Composite Films Annealed in Hydrogen

a) Grain structure

The grain structures of In/Au composite films annealed at 300 C,

400 C and 500 C in H2 examined by TEM are shown in Figure 33. A

drastic increase in grain diameter with annealing is observed. The

change in the mean grain diameter as a function of the annealing

temperature is shown in Figure 34. The error bars indicate as usual

the standard deviation. The most significant grain growth was observed

in the sample annealed at 500 C, where very large grains with diameters

greater than 1 pm were easily distinguished. This substantial grain

growth accompanied by a wide grain size distribution may result from

the preferential grain growth along certain crystallographic

orientations, presumably (111) parallel to the substrate surface. As

these grains grow preferentially, other grains shrink in size and a

wider grain size distribution may result.

b) Surface morphology

SEM micrographs for In/Au composite films annealed at 300 C and

400C in H2 revealed very little structural changes of the surfaces

(Figure 35). The films annealed at 500 C will be discussed separately

below. In these films, the surface morphology of the as deposited

In/Au composite films seemed to be preserved; that is no significant

grain growth and grain boundary grooves were observed. We know from

,0.2 m

Figure 33. Transmission Electron Micrographs of In/Au Composite
Films Annealed in H2 at : a) 300 C, b) 400 C, c) 500 C

04 5-

0T M P. C

-Various Temperatures in H2
W f

3 T i


Various Temperatures in H2




Figure 35. Scanning Electron Micrographs of In/Au Composite Films
Annealed in H2 at : a) 300 C, b) 400 C, c) 500 C

the results reported above that the significant grain growth which

occurred during annealing can be observed mainly by TEM. The

difference between the surface structure and the internal grain

structure is again noticed. This apparent discrepancy is believed to

be result of the surface roughness of these samples. Indeed, for an

In/Au composite film annealed at 400 C, the bright field and the dark

field cross sectional TEM micrographs shown in Figure 36a and Figure

36b reveal large grains of about 3000 A in diameter and a rough surface

whose sinusoidal "perturbation" wavelength is about 750 A. Therefore,

the fine surface structure observed by SEM is believed to be caused by

surface roughness, as before, rather than by real grain structure.

A substantially different surface morphology was observed when

In/Au films were annealed at 500 C in H2. The SEM micrograph (Figure

37) shows a very coarse-grained structure and some severe grain

boundary grooves at the grain boundary vertices. The cross sectional

TEM micrographs (Figure 38) reveal very large grains whose diameter is

greater than 7500 A and an extremely smooth surface. This suggests

that the surface roughness has been relaxed during annealing.

The flattening of the rough surface and the evolution of grain

boundary grooves are both caused by the capillary-induced mass transport

during annealing [25-27,89-91]. The mass transport, either by volume

diffusion, surface diffusion or evaporation-condensation mechanisms, is

a thermally activated process and therefore temperature dependent.

Depending on the annealing conditions, the mass transport and the

subsequent flattening or thermal grooving may not occur due to the

kinetic limit. Apparently, annealing at 300 C and 400 C for 1 hr in H2


0.0 5 Am

Figure 36.

Cross Sectional Transmission Electron Micrographs of
In/Au Composite Films Annealed at 400 C in H2 :
a) bright field image, b) dark field image

0. 2 m

Figure 37. Scanning Electron Micrograph of In/Au Composite
Films Annealed at 500 C in H2


0.0 5 ym

Figure 38. Cross Sectional
In/Au Composite
a) bright field

Transmission Electron Micrographs of
Films Annealed at 500 C in H2
image, b) dark field image

I 01 1 ,69 P.Ia II8C I

did not provide the conditions to activate the mass transport so that

the relaxation of the roughness has not occurred during annealing. On

the other hand, annealing at 500 C for 1 hr in H2 did provide the

conditions to allow the flattening and the grain boundary grooving to


c) Texture

X-ray diffraction patterns for In/Au films annealed at 300 C, 400

C and 500 C are shown in Figure 39. The relative intensity of each

diffraction peak with respect to the intensity of the {111} peak is

listed in Table 5. In/Au composite films annealed at 300 C and 400 C

revealed a relatively random orientation; that is they displayed all

diffraction peaks (Figure 39a and Figure 39b). However, the In/Au

composite film which was annealed at 500 C revealed only {111} and {222}

diffraction peaks, which suggested a (111) texture parallel to the

substrate surface (Figure 39c).

SAD patterns complement this information. In/Au composite films

annealed at 300 C and 400 C in H2 revealed that the {111} diffraction

ring was the most intensive one (Figure 40a and 40b). This suggests

the random orientation of the grains in the film. However, the In/Au

composite film annealed at 500 C revealed that the {220} ring was more

intensive than the {111} ring and also showed some directionality as

well (Figure 40c), which suggests a (111) texture paral lel to the

substrate surface.



n222 20 nill

Figure 39. X-Ray Diffraction Patterns of
Annealed in H2 at : a) 300 C,

In/Au Composite Films
b) 400 C, c) 500 C

Table 5. Values of Ihk/111 for
Various Temperatures in


In/Au Composite Films Annealed at
H2, As Measured by X-Ray

In/Au Composite

300 C






400 C





500 C





hkl/111' %




Figure 40. SAD Patterns of In/Au Composite Films Annealed in H2
at : a) 300 C, b) 400 C, c) 500 C