DEPOSITION, CORROSION AND COLORATION OF
TUNGSTEN TRIOXIDE
ELECTROCHROMIC THIN FILMS

BY

SEY-SHING SUN

A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF

THE UNIVERSITY OF FLORIDA

IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE

DEGREE OF DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA
1983

To my mother
and
Mei-Hwei, my beloved wife

ACKNOWLEDGEMENTS

The author would like to acknowledge and thank his advisory
committee and, in particular, to express his gratitude to Professor
P. H. Holloway, who convinced the author of the importance of his
research and maintained an open and creative research environment
throughout the course of this work. The author is also grateful to
Professor J. R. Ambrose, J. Ali and R. U. Lee for their assistance
and useful discussion during corrosion studies. In addition, thanks
are extended to Drs. G. E. McGuire and R. T. Tuenge of Tektronix, Inc.,
who provided the author with internship and encouragement for the
development of practical electrochromic display devices.

Thanks are also due to A. Haranahalli, K. Shanker and Y.-X. Wang
for their technical assistance and encouragement.

Finally, acknowledgements are due to the Army Research Office,
Durham, NC, for research support under grant //DAAG-29-8O-C-00O7.

iii

TABLE OF CONTENTS

PAGE
ACKNOWLEDGEMENTS iii

ABSTRACT vi

CHAPTER

I . INTRODUCTION 1

II. CORROSION STUDIES AND MODIFICATION OF TUNGSTEN

TRIOXIDE THIN FILMS BY OXYGEN BACKFILLING 6

Introduction 6

Experimental H

Film Preparation H

Chemical Characterization 12

Physical Characterization 15

Corrosion Measurements 16

Electrochromic Properties Measurements 16

Results 20

Corrosion Studies 20

Modification of W0~ Films by Oxygen Backfilling 26

Discussion 44

Corrosion of WO Thin Films 44

Modification of Stability of WO Films by

Oxygen Backfilling 61

Modification of EC Properties of W0_ Films by

Oxygen Backf ill ing 62

Summary .

.80

III. DEVICE FABRICATION 84

Introduction °4

Experimental °5

Preparation °5

Characterization 86

PAGE
CHAPTER

Liquid Electrolyte Device 88

Results 88

Discussion 97

Solid Electrolyte Device 105

Results 108

Discussion HI

IV. SUMMARY • 113

Conclusions H3

Future Developments 119

APPENDICES

A. PRELIMINARY STUDIES OF CATHODOCHROMISM IN TUNGSTEN

TRIOXIDE FILMS AND THE EFFECT OF AIR BACKFILLING .. 121

B. CORROSION OF TUNGSTEN TRIOXIDE FILMS DURING

STORAGE IN MOIST AIR 128

BIBLIOGRAPHY 132

BIOGRAPHICAL SKETCH 140

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

DEPOSITION, CORROSION AND COLORATION OF
TUNGSTEN TRIOXIDE
ELECTROCHROMIC THIN FILMS

By

Sey-Shing Sun

December, 1983

Chairman: Paul H. Holloway

Major Department: Materials Science and Engineering

Electrochromism (reversible color change of a material , induced
by an electric field) has been investigated in tungsten trioxide
thin films. It is known that electrochromic device lifetime is
generally limited by corrosion of WO films in an acid electrolyte.
Two mechanisms of corrosion were observed, viz., general dissolution
and interfacial delamination. It was found that the dissolution of W03
films in acid could be attributed to high concentration of thermody-
namically unstable species (W02 and WO) in the as-deposited films.
These species were identified by X-ray photoelectron spectroscopy
data and are consistent with Rutherford backscattering spectroscopy
data which showed an oxygen deficiency (0/W = 2.76). The Pourbaix
diagram for tungsten indicated that WO was the thermodynamically
stable specie for the present storage condition (pH = 0.5,

E = 0.4 V„TTTJ . A corrosion mechanism was proposed consisting of
SHE

dissolution of W0? and WO and precipitation of crystalline W03 and
its hydrates. Interfacial delamination occurred when WO- and its
hydrates precipitated back onto the original films.

The oxygen content in the WO., films was increased by oxygen back-
filling during evaporation. Dissolution and interfacial delamination
of the oxygen enriched films were reduced to negligible rates due to
reduced concentration of WO„ and W„0 . However, the electrochromic
properties were degraded by oxygen enrichment. For example, increased
resistivity and decreased optical efficiency in the oxygen enriched
films resulted in slower coloration speed. The resistivity increase
and decreased optical efficiency were explained by postulating an
increased density of inactive electron trapping sites. The porosity
of the films could be increased by deposition at high background
pressure, resulting in increased surface area and absorbed water.
The bleaching speeds and self-erasure rates were increased since the
rates of removal of protons were increased by the increases in
porosity and absorbed water.

In another approach to increase the electrochromic device lifetime,
the electrolyte was modified. Devices using a solution of LiClO^ in
propylene carbonate exhibited excellent lifetime. Switching speeds
were increased by increased porosity, deposition of MgF2 overlayers,
and more conductive indium-tin-oxide layers. In addition, solid state
electrochromic devices using a hydrated MgF3 film were fabricated.

CHAPTER I
INTRODUCTION

Electrochromism is broadly defined as the ability of a material
or system to change color reversibly in response to an applied
potential. A great variety of electrochromic materials or systems
are know and can be conveniently classified into the following
categories according to the physical mechanisms involved:

- reversible electrodeposition systems

- organic ion-insertion materials

- inorganic ion- insertion materials.

Viologens, anthraquinones and electroplating of very thin films of

o 4

metals such as silver comprise the first class. Diphthalocyaniens

and tetrathiafulvalenes are included in the second class. Class

three is comprised of transition metal oxides, among which WO^ thin

films have received by far the most attention. Other materials

40-44
such as anodically deposited iridium oxide films of IrO,, , molyb-

denium oxide (MoO ) , " vanadium pentoxide (V^) , and nickel

oxide (NiO) have also been shown to color electrochromically , and

would be placed in this class. Besides coloring electrochromically,

some of the transition metal oxides may also be colored by heating

Although the vapor deposited tungsten trioxide films are substoichio-
metric as reported below, the term 'W03 thin films' will be used for
simplicity unless otherwise noted.

(thermochromism) , by exposure to ultra-violet light or X-ray
(photochromism)7 or by electron beam irradiation (cathodochromism) .
The W0„ electrochromic (EC) display device typically consists of a
transparent electrode such as a Sn02 coated glass substrate, a thin
layer of transparent WO,, an electrolyte and a counter electrode
(Pt or Au). When a small negative voltage ( -U.0 V) is applied to
the transparent electrode, the WO film is colored blue in trans-
mission.8,9 The induced optical change is proportional to the charge
injected during the coloring pulse. If the voltage is removed, the
W0„ film remains colored. Reversing the polarity of the applied
voltage will cause bleaching of the films. The electrochromic

coloration of W03 in acidic aqueous solution involves the injec-

9
tion of electrons and protons into W03 films. The blue coloration

in transmission is associated with the broad optical absorption band

which gives an absorption minimum at wavelength of 0.4 ym and an

absorption maximum at 0.95 Mm.8'9 The absorption band is generally

explained by electron transfer between neighboring tungsten atoms,

10-12
i.e., optical absorption by excitation of polaron hopping.

The protons injected are for the purpose of charge compensation.

T .+ 13-21 + 25-28
Therefore, any monovalent cation can be used, e.g., Li , wa ,

» + 29
or Ag .

The electrochromic device exhibits many attractive features as
displays, such as

- good color contrast especially under high level of ambient
light;

- wide-angle viewability which is an advantage over most
liquid crystal displays (LCD) ;

- low voltage operation (1.5 V down to approximately 0.5 V);

2

- low power, V9 mW/cm , somewhere between LCD and light-
emitting diode (LED) displays;

- storage of the display without power dissipation;

- potentially low cost.

The response time has been reported to be approximately 50 ms with

30
a modified device configuration in acid electrolyte, and is

comparable with LCD. However, there is a principal obstacle to

the commercialization of EC displays, viz., short device lifetime.

The general lifetime reported was about ^10 cycles at 0.5 Hz or

less than two weeks for static condition. To understand the cause

of degradation and to find a way to improve the lifetime was the

major goal in the present research.

The degradation of EC devices was reported to result from

32-35
corrosion of WO., thin films in the acid electrolyte. In

Chapter II, the corrosion mechanism and the results of using oxygen

backfilling during WO evaporation to prevent corrosion will be

discussed. The durability of the deposited films in sulfuric acid

solution will be shown to be limited by two mechanisms — a general

uniform film dissolution and an interfacial delamination. Due to

the deficiency of oxygen (0/W = 2.76), some of the configurations of

tungsten atoms in as-deposited films may be thermodynamically unstable.

A conversion process, viz., dissolution of these unstable

species and precipitation of the stable crystalline rW03 and its

hydrates, leads to the general film dissolution and interfacial

delamination of WO films. The delamination of films from the sub-
strates occurs when the internal stresses induced by precipitation
exceed the interfacial adhesion strength.

In order to improve the lifetime, two basic approaches were
adopted in the present study. In one approach, oxygen backfilling
during WO- evaporation was used to modify film structure and
reduce the attack by the acid electrolyte. By this method, the
0/W ratio was increased and the amount of degradation was reduced
due to the increased concentration of the thermodynamically stable
specie, WO . The current injection and optical efficiency during
coloration were lower, however, while the bleaching speed and
self-erasure rate were faster for oxygen-enriched films. The
mechanisms by which these modifications of EC properties were
achieved are discussed.

The second approach to improve the device lifetime involved
the use of non-aqueous electrolytes. Non-aqueous electrolytes
have been shown to provide an inert environment where the corrosion
of WO was not observed. In Chapter III, trial devices using
LiCIO, dissolved in propylene carbonate (Li/PC) and solid MgF
thin film electrolytes are discussed. Previous investigators
have reported that devices using Li/PC as an electrolyte exhibited
slow response speed. In this study, an improvement in speed was
achieved by nitrogen backfilling during WO deposition and by the
use of a MgF over layer. Layers of MgF„ were also used as elec-
trolyte to fabricate solid state display devices. These experi-
ments indicated that the increased water content in MgF films

resulting from air backfilling increased the coloration/bleaching
(C/B) speed, while the operational area of the device was limited
by the air backfilling during deposition of WO films.

Preliminary data on cathodochromism (electron beam induced
coloration) in WO thin films are reported in Appendix A. The
coloration density in WO films was found to be increased by air
backfilling, .especially for higher energy electron beams.
Results indicate that cathodochromism in amorphous WO thin
films may still be caused by polaron hopping absorption as in
elect rochromism.

CHAPTER II
CORROSION STUDIES AND MODIFICATION OF TUNGSTEN TRIOXIDE
THIN FILMS BY OXYGEN BACKFILLING

Introduction
Electrochromism (EC) has been investigated in tungstic oxide
films for possible application in display devices. The devices
typically used diluted sulfuric acid solutions as the electrolyte
With a modified device configuration (Figure 2.1), viz., a porous gold

overlayer on the WO films, it is possible to achieve a switching

30
time of approximately 50 ms. The time is comparable to that

of liquid crystal displays (LCD). However, the lifetime of EC

devices in such an electrolyte is normally short. The device

lifetime was found to be limited by the degradation of WO^ films

in the acid electrolyte in the form of general dissolution and

delamination. Short device lifetime is a principal obstacle

to the commercialization of EC displays.

Degradation of WO thin films has been studied and was

32—35
correlated with corrosion, However, the mechanism(s) by

which corrosion occurs and the thermodynamics of W03 corrosion

35
have not been considered in detail. Faughnan and Crandall

studied evaporated WO films in H^O^/water solution (pH = 0) and

found the dissolution rate was approximately 300 A per day. Cyclic

coloring and bleaching of the films led to faster dissolution rates.

In

Figure 2.1. Schematic representation of the
electrochromic cell with thin
gold overlayer on WO .

32
Randin replaced the water with glycerin as the solvent in the

electrolyte and found the dissolution rate was lowered by an order

of magnitude (approximately 20 A per day) . An increase in the

dissolution rate was also observed under cycling condition.

The film was eventually destroyed as a result of delamination

from the substrate. To explain this behavior, Randin listed a

number of possible mechanisms of degradation, e.g.:

a) proton insertion process causing hydrogen embrittlement
which undermined the internal strength of W0_,

b) anodic disintegration of the films,

c) oxidative degradation occurring readily on semiconductor
electrodes, such as WO + 6h — > W (sol'n) + 3/2 0^
where h designates hole carriers.

However, no specific data and discussion supporting the above

postulates were given. He also found WO thin films were more

stable in some non-aqueous solvents such as propylene carbonate,

acetone and Y-butyrolactone. However, the response speed of the

device using an electrolyte with such a solvent was too slow to

be practical.

33
Reichman and Bard reported that both the kinetic behavior

and the stability of the films in acid electrolyte depend on the

amount of water and the porosity of the films. From infrared

absorption spectra, they found the water content was increased

when the films were cyclically colored and bleached. On the basis of

these observations, they speculated that during operation, uptake of

water into the films caused expansion of the lattice, making it

more porous and allowing even greater water penetration. The

result was higher water content in the films after cyclic colo-
ration and bleaching. They also measured the dissolution rate of
WO film in acid by monitoring the decrease in current injected
during cyclic coloration and correlated it with the increase in
water content of W03 films. They concluded that the dissolution
was due to the formation of W03 hydrates (W03>H90 or W03.2H20)
since these hydrates are more soluble than WO,,. Hence, the higher
the water content of the film, the more WO dissolved. No model
of dissolution was proposed, nor was saturation of the electrolyte
discussed. In addition, no data were shown demonstrating the
formation of hydrates. Reichman and Bard also observed slow
dissolution in glycerin/H SO, (10:1) as did Randin; they attri-
buted this to the lack of water pickup during cycling.

Arnoldussen studied the dissolution of the evaporated W03
films in deionized (DI) water and concluded the films dissolve

to form metatungstate or pseudometatungstate ions. On the basis of

37
previous study of vapors generated by subliming WO.,, he postu-
lated that as-deposited WO- films are composed chiefly of trimeric
clusters weakly bound to one another through water-bridge or
hydrogen bonds. The dissolution of the film was therefore pos-
tulated to occur by these trimeric clusters' being dissolved as
3-, 6-, or 12-mer ionized complexes directly from the dissolving
surface. He also suggested cyclic conditions were just a natural
extension of the dissolution process, i.e., voltage enhanced
dissolution. Degradation in non-aqueous solvents was suggested

10

to result from formation of negative complexes with electrolyte
anions or from tungstate formation due to H„0 incorporated into
the films during manufacture. However, all of Arnoldussen ' s
data dealt with electrolyte of pH > 4, and the corrosion reaction

proposed was based on metatungstate ions; it has been shown that

2+
WCL is the dominant specie for dissolution of tungstic oxides

in acid solution of pH < 1. The O/W ratio in Arnoldussen' s

microstructure model is always > 3, since the oxygen is bound

either in W,0Q or HO. It will be shown in this study that the

0/W ratio was always < 3. In addition, the idea of H?0 stabilized

W.0q rings is unreasonable, since the disintegration of the films

would be expected if HO was removed. However, Zeller amd Beye-

ler reported HO was removed at 170°C without detectable change

in the structure of WO films. In fact, no structural change

32
was observed until crystallization at 350°C, which is well

above the HO removal temperature.

As is evident from reported studies, a variety of mechanisms

have been proposed for degradation of WO,, films. All of these

investigators have speculated that water is a major cause of

corrosion. The mechanisms proposed are normally not supported

by strong and direct evidence, and are sometimes contradictory

to existing data (e.g., Arnoldussen' s HO model discussed above).

However, it is generally recognized that WO thin films in acidic

electrolyte degrade in the form of general dissolution and dela-

mination. In cycling conditions the dissolution was increased.

11

In the following, new mechanisms which can fully explain these
phenomena without the shortcomings of the existing postulated
mechanisms will be discussed.

There are several ways to counter corrosion problems, e.g.,
1. modify the electrolyte, 2. put protective coating on the W03
film, 3. modify the W0_ thin film microstructure. In the second
half of the chapter, stability data will be reported for WO^ films
evaporated with oxygen backfilling (OB). The 0/W ratio will be shown
to be increased by backfilling. This led to reduced degradation.
However, the coloration speed and optical efficiency were decreased
while the bleaching and self-erasure rates were faster for the oxygen
enriched films. The mechanisms by which these EC properties were
modified are also discussed.

Experimental

Film Preparation

Thin films of WO were prepared by evaporation of 99.9% pure
WO powder (Cerac, Inc.) from a resistivity-heated, alumina-coated
tungsten spiral boat (Sylvania, Inc.). The substrate was normally
tin oxide coated glass (NESA glass, Pittsburgh Plate and Glass Co.)
with a resistance of 100 Q/p , but soda-silicate glass slides and pure
graphite substrates (Ernest Fullan, Inc.) were used occasionally.
Prior to deposition, the substrates were cleaned with Alconox/water
solution, rinsed in deionized water for 10 minutes, etched in

12

Chromerge (a registered brand name of Monostat Labs representing a

mixture of chromic acid and sulfuric acid, 1:10) and degreased in

boiling methanol vapors for 5 minutes. The evaporation system was

evacuated to a base pressure of 2 x 10 Torr and WO was evaporated

either at a residual pressure (P ) of 1 x 10 Torr. or with a

res .

-5 -3

partial pressure of oxygen (P ) between 5 x 10 and 1 x 10 Torr.

2

Oxygen (99% pure) partial pressure was manually controlled by a needle

valve. The deposition rate was typically 10 A/sec. The thickness
of the films ranged from 2,000 A to 4,000 A. The source to sub-
strate distance was 30 cm. No substrate heating was used during
evaporation.

Chemical Characterization

Rutherford backscattering spectroscopy (RBS) was used to deter-
mine the oxygen to tungsten ratio (0/W) of the evaporated films.
Briefly, a beam (300 nA, 300-600 uC) of monochromatic (2 MeV) alpha
( He ) particles was focused (rectangular area 2 mm x 4 mm) onto a
WO- films on a graphite substrates. Alpha particles were scattered

through 155° by W and 0 atoms in the films losing an energy which

53
depends on the masses of these atoms. Analysis of the number of

backscattered alpha particles versus their energy allowed the 0/W

ratio to be determined.

A typical RBS spectrum is shown in Figure 2.2, which is plotted

as the backscattered yield (counts per second) versus the channel

13

100

Figure 2.2

300

500

700

900

Channel Number
Rutherford backscattering (RBS) data for a WO

film

on graphite substrate using alpha particle of 2.0
MeV.

14

number. The channel number is related to the backscattered particle
energy by a conversion constant (eV/channel number) . The concentra-
tion of the elements in the films appears as square-top peaks in the
spectrum. The ratio of the different elements in the films can be
obtained by calculating the ratio of the area under the respective
peaks and correcting for variations in their scattering cross sections,
Since the width of the peak corresponds to the energy lost by alpha
particles which traveled through the films, the thickness can be
obtained by converting the peak width to depth using calculated
stopping powers.40 The shape of the peaks also provide information
about the structure of the films (uniformity, interdif fusion, etc.).

Compositional information about light constituents such as
alkali and water were obtained from secondary ion mass spectroscopy
(SIMS). The experimental system incorporated a UTI quadrupole
mass filter (UTI 100C) with a 3M prefilter (Model 610), a 3M mini-
beam ion gun, and a sample manipulator (Huntington PM-600XYZTRC) .

-9
In this study, the vacuum system was pumped down to 1 x 10 Torr

and was then backfilled to 5 x 10"5 Torr with argon. A primary ions

energy of 4 KeV and beam current of 100 nA was used during the analysis.

X-ray photoelectron spectroscopy (XPS) was used to obtain
the information on the concentration of different oxidation states
of tungsten ions in WC>3 thin films. The analysis was performed with

a Krato XPS system (XSAM 800) using Mg K-a X-rays (1253.6 eV) . The

-9
samples were maintained at room temperature and a pressure of 5 x 10

Torr.

15

Physical Characterization

The relative density of evaporated W0„ films was changed by
oxygen backfilling. A simple method using a quartz oscillator
thickness monitor (Piezo Technology Inc. Model 990) was used to
measure the areal mass change. The areal mass measured by the
monitor was fixed for deposition at all pressure. The physical
thickness of the deposited films was then measured by a profilo-
meter (Sloan Dektek II) . An arbiturary index was obtained when the
physical thickness was devided by the thickness calculated from the
areal mass by assuming the theoretical density of W0„. This index
was used to compare the relative densities of the films. Thin films
of W0-, with and without oxygen backfilling, were studied for surface
damage (before and after corrosion) with a JOEL scanning electron
microscope (Model JSM-35C) .

X-ray diffraction spectra were obtained from as deposited and
corroded films. Nickel-filtered Cu-Ka radiation at 40 KV and 20 mA
from a Phillips Electronics Instrument Model 690 spectrometer were
used. A range of 10° < 26 < 90° was scanned at a speed of 2*/min.,
with a counting time constant of 1.0 second. A Bicron model SR-
BP941 scintillation detector was used.

Films of WO deposited on glass substrates could be colored to

blue by touching an indium wire to the film surface through an acid

electrolyte (3.6 N H„S0,). The coloration occurs by the injection
/ 2 4

T 3+ -L. 1 ~

of electrons provided by the corrosion reaction: In — * in -t- je ,
and protons being injected into the film from the electrolyte to

16

maintain charge neutrality. The coloration progressed uniformly

outwards from the point of contact of the In wire with a circular

56
color front. This phenomenon prompted Crandall and Faughnan to propose

a simple method for determining the electron diffusion coefficient.

In this method, the radius of the color front was plotted as a function

of the square root of the time. The electron diffusion coefficient

(D ) was obtained from the slope which is equal to V4D .
e e

Resistivity was measured immediately after deposition without
breaking the vacuum. The electrodes were parallel etched stripes of
SnO„, 0.1 cm wide with a gap of 0.5 cm. The current upon application
of 10 V was measured by an electrometer.

Corrosion Measurement

In the corrosion study, the WO films deposited on graphite
substrates were immersed horizontally in the unstirred acid electro-
lyte (3.6 N H„S0, in water) contained in glass dishes at room
temperature. The electrolyte was not deaerated during the experiment.
The samples were removed at selected time intervals and examined by
RBS to measure the WO thickness. The average dissolution rate was
obtained by dividing the change in thickness by time.

Electrochromic Properties Measurements

Optical absorption spectra were obtained from films deposited
on glass slides both before and after coloration. The equipment

17

consisted of a quartz halogen lamp (Oriel Model 6140) , a mono-
chromator (Oriel Model 7240) and a silicon (for 400-1000 nm) or
PbS (for 1000-2000 nm) detector. The range scanned was 300 to
1500 nm with measurements being taken in transmission.

The electrochromic (EC) properties (coloration/bleaching
speed, optical efficiency and open circuit memory) were charac-
terized in the electrochromic cell shown in Figure 2.3. EC
properties are defined as follows

- Coloration time (t ) is defined as the time needed to color
a film to a contrast of 50% absorption;

- Bleaching time (t ) is that time required to erase the film
from 50% to 25% absorption;

- Optical efficiency is defined as optical density change per
unit charge injected;

- Open circuit memory is defined as the time needed to bleach
the film to half of the original contrast after the power was
disconnected.

In the following discussion, the C/B speed will be used which is

defined as the rate of coloration or bleaching to reach defined

contrast, and is inversely proportional to C/B times. Electrical

contact to the Sn0? layer was made with indium solder outside the

9

electrolyte. The active area was 1.5 cm". The cell was made of

Teflon® spacers and stainless steel holders for the glass plates.

An optical path through the cell allowed electro-optical measurements.

The electrochemical data were taken using the circuit shown in
Figure 2.4. A potentiostat (Princeton Applied Research Model 173)

18

Luggin
Capillary

-> Reference
Electrode

Steel
Plate

NESA
Glass

c

v\

Spacer
-i

-XElectrolyt

lyte

— I
WO,

Working Pt Counter
Electrode Electrode

u

.Steel
^— d Bolt

Glass

R

Figure 2.3. Schematic representation of the
electrochromic cell.

19

• • •-

WE RE CE

^vwv

potentiostat

r *•

^

Light
Source

EC Cell

x-t

recorder

Figure 2.4. Circuit diagram showing the connection between
EC cell, potentiostat and measuring devices.
WE: working electrode, RE: reference electrode.
CE: counter electrode.

20

was used as the power supply. Platinum wire and a calomel electrode
were used as the counter electrode and reference electrode, respect-
ively. Monochromatic light (700 ran) was used throughout the experi-

Results

Corrosion Studies

The durability of vapor deposited WO films in acid electrolyte
was limited by two mechanisms - a general uniform film dissolution
and an interfacial delamination. As prepared, W03 films exhibited
featureless smooth surfaces with some isolated pinholes detected with
SEM at magnification of x500 (Figure 2.5). After storage in the acid
electrolyte, WO films prepared without OB begin to degrade by general
film dissolution, formation of crystallite-like particles, wrinkles
spreading in a rosette-like pattern (Figure 2.6), and inter-
facial delamination of the films from the glass substrates (Figure 2.7
(A)). The dissolution rate was very high at VL6 A/hr (V350 A/day)
as shown in Figure 2.8. However, the dissolution rate in the static
condition (i.e., without electrochemical cycling) may have been
enhanced by the graphite substrate, as will be discussed in the
following section. The static dissolution rate is expected to be
lower for films on glass substrates.

The peaks in the RBS spectra of W03 films also exhibited a
significant change after corrosion (Figure 2.9). The original W

21

(B)

Figure 2.5. Scanning electron micrographs of as-deposited WC>3
films on glass substrates;
(A)

p = 1 x 10 ^ Torr, x 500,
(B) Pres-= 1 x 10 Torr, x 3,000
°2

22

y~\ %*

4 ^

(A)

")"/"' ' ^/±

-.;"

,

(B)

Figure 2.6. Scanning electron micrographs of corroded surface^
(3.6 N H SO,, 4 days) of WO films (P ^ = 1 x 10
Torr) showing the forms of interf aciales 'delamination.
(A) wrinkles, (B) rosette pattern.

23

Figure 2.7. Scanning electron micrographs of W03 films after storage

in 3.6 N H„SO, ;
(A) P

-5

(B) P

= 1 x 10 Torr, 2 days, x 500,
= 1 x 10 Torr, 8 days, x 3,000.

24

P (Torr)

Figure 2.8. Dissolution rate of WO films in 3.6 N H2S04
solution versus the oxygen partial pressure
during deposition.

25

100

300

500

700

900

Figure

Channel Number

Effect of corrosion on the RBS spectra for W03 films
deposited without oxygen backfilling;

(a) as-deposited,

(b) after 5 hr. in 3.6 N HtSOa>

(c) after 20 hr . in 3.6 N ft?S0^.

peak was rectangular in shape indicating a uniformly thick film.
After corrosion, the peak was triangular in shape which indicates a
variable film thickness consistent with micropore formation by corro-

Modification of WO- Films by Oxygen Backfilling

Chemical and physical properties

The compositional changes caused by oxygen backfilling are

shown in Figure 2.10 and 2.11. From RBS data, the average 0/W ratio

increased from 2.7 with no oxygen backfilling (Pres _ = 1 X 10 Torr)

to 2.9 at P = 1 x 10~3 Torr. Since the films were hydrated as shown

°2
later, the oxygen detected may have included that from the incorporated

water. The actual number of oxygen atoms that were directly bonded

between tungsten atoms might have been even lower. SIMS data indicated

that the content of water in the films increased by about a factor

of two between the two extreme cases (without oxygen backfilling and

P = 1 x 10~3 Torr) . The films prepared at the two extreme conditions

2
were examined by XPS. The spectra of the W4f core levels for these

films are shown in Figure 2.12. For comparison, the core level spectra

for WO on W representing a W6+ state and the positions of the reduced

W states (W5+, W4+) are also shown. The multiplet split W4f peaks

*For simplicity in the following discussion, *W0 ' (without OB)

and 'WO ' (P = 1 x 10 Torr) are used to dehdte W03 films prepared

at these' ?wo °2 extreme oxygen partial pressure conditions.

27

3.00

2.90

2.80

2.7

10

10

-4

10

0„(Torr)

Figure 2.10. Oxygen to tungsten atomic ratio of WO^ films from
RBS data versus the oxygen partial pressure during
deposition.

28

3. Or

1.0

0.0

P (Torr)
°2

Figure 2.11. Relative signal intensity for H^O from secondary ion
mass spectroscopy (SIMS) versus the oxygen partial
pressure during deposition of W03 thin films.

29

40 35 30

Binding Energy (eV)

Figure 2.12. XPS W(4f) core level spectra for WO.^ films
prepared at two extreme conditions, i.e.,
'WO ' and 'WO '. For comparison, ^he
spectra for WO on W representing a W sta£e_
and the position of the reduced W states (W ,
are also shown.

W )

30

of 'W0„ ' spectra are located at the same energies and exhibit

6+
similar intensity as those peaks from W in the reference spectra.

The spectra of 'WO ' therefore indicate that the tungsten exsited

mostly as W in these films. On the other hand, the spectra of

'WO ' exhibit a filled valley between and a shoulder on the low

energy side of the multiplet split W4f peaks from W . This indicates

that the concentration of subvalent W and W were high in 'w02>7'

films. Therefore, the oxygen backfilling increased the oxygen content

in the films as indicated by RBS data, and led to a concentration

increase in W states which was identified by XPS data.

The relative density of WO- films was found to decrease

as a result of OB as shown in Figure 2.13. The unmodified films

exhibited a relative density of 0.8 when compared to the normalized

-3
bulk density. For films prepared at PQ = 1 x 10 Torr, the

relative density decreased to 0.5 and the porosity was expected to
be substantially higher.

The surface topography of WO films was modified by oxygen
backfilling as shown by SEM photomicrographs taken at high magnifi-
cation (x 3,000, see Figure 2.5(B)) for films deposited at P = 5 x 10

2
Torr. The rough surface is consistent with the higher porosity in

these films.

31

1.0

0.5-

o.a

P (Torr)

Figure 2.13. The relative density of WO films versus the partial
pressure of oxygen during deposition

Stability

Interfacial delamination was not observed for films prepared at

-4
higher partial pressures of oxygen, i.e., >5 x 10 Torr. However,

a slightly corroded surface which was rougher than the original

surface was observed at higher magnification (x 3,000, see Figure

2.7(B)). The dissolution rates for these films were negligible as

shown in Figure 2.8.

The X-ray diffraction pattern for the as-deposited films with or

without OB showed a very broad diffused peak typical of an amorphous

structure (Figure 2.14(A)). After 18 days storage in acid electrolytes,

the X-ray diffraction patterns exhibited crystalline peaks on top of

the diffuse scattering from the amorphous film. These data are

interpreted to indicate a partial transformation into crystalline

phases of W03,5? WO^O58 and W03.2H2059 (Figure 2.14(B) and 2.14(C)).

The only observable difference between the spectra from the two

extreme OB conditions is that there is a distinct peak 3.01 A

which is associated with WO .2H 0 for films without 0B (Figure

2.14(B)). This peak is absent from the spectra from films deposited

at P = 1 x 10~3 Torr (Figure 2.14(C)). On the other hand, there is

°2
a peak at 5.37 A in Figure 2.14(C) (associated with W03-H20) which is

absent in Figure 2.14(B). This seems to indicate that the W03-2H20

phase was the dominant phase in the films without backfilling while

W0„.H„0 dominated the films with 0B.

33

*W0-

3W03 • H,0
• W03 • 2H20

(C) -a-wo-j

: WO . H,0
• W03- 2H20
it *

in 1 1

28 Ideqreei

Figure 2.14. X-ray diffraction spectra: (A) As deposited, showing
amorphous pattern, and (B) , (C) from WO films after
18 days in H„SO/ solution where the amorphous back-
ground intensity was substracted to emphasize the
crystalline peaks. The spectra are from samples which
were deposited with (B) residual pressure of 1 x 10
Torr, (CO oxygen partial pressure of 1 x 10 Torr.
The symbols (o, +,<r) show the origins of specific

diffraction peaks. (•&: WO,

W03.H20,

WO .2H20)

34

Electrical and electrochromic properties

The coloration times, t , for films deposited at various pressure
c

are shown in Figure 2.15. The t increased from 4.2 sec. for films

a c

prepared without OB to about 5.8 sec. for films prepared at PQ =
1 x 10 Torr. The increase in the coloration times may be due to a
decrease in charge injection rate and/or a decrease in optical effi-
ciency. Figure 2.16 shows injected current versus time (j vs. t)
during coloration for films prepared at the two extreme OB conditions.

The films exhibited similar behavior i.e., the current was almost

-1/2
independent of t at short time (t < 0.2 sec), and followed a t

dependence at long time (t > 2.0 sec). However, the magnitude of
the current was lower for films with OB. The charge injected versus
time may be calculated from Figure 2.16 and is shown in Figure 2.17.
Again it demonstrates the lower charge injection at a specified
time for the high oxygen backfilled pressure. In Figure 2.18 the
electron diffusion coefficients (D ) which were measured by the
color front growth rate technique are plotted versus the partial
pressure of oxygen. The D was decreased by about a factor of three
when the films were prepared at PQ = 1 x 10 Torr, as compared to
the films without OB. The thermal activation energy (Eth) for elec-
tron conduction in the uncolored films was also found to have been

changed by OB as shown in Figure 2.19. In this Figure, the tempera-
ture dependence of current is plotted versus T for films prepared
at the two extreme OB conditions. From the slope of the plot Eth

35

6.0-

5.0

4.0

P (Torr)

Figure 2.15.

Time needed to color WO films to obtain 50% absorp-
tion versus oxygen partial pressure during deposi-

tion. (E

-0 6V , Film thickness;
SCE'

3,000 A)

30^

10

1 _

0.5 1.0 5.0

Log Time (sec. )

10.0

Figure 2.16. Coloration current density versus time for WO^ films
prepared at two extreme conditions; solid circles:

'W0„ ' and open circles:

'WO

2.9

37

Time (sec . )

Figure 2.17. Injected charge versus time for
films prepared at
""1 ■— -■• '^2.7

films prepared at two
solid circles: 'WO,
'WO,

3
extreme conditions:

and open circles:

2.9

3iT

3.0-

1.0 -

0.0

P (Torr)

Figure. 2. 18. Electron diffusion coefficient in WO films versus
partial pressure of oxygen during deposition.

T9^

3.5

1000/T (°K )

Figure 2.19. Temperature dependence of current in W0? films pre-
pared at two extreme conditions: solid circles:

'WO,

' and open circles: 'WO

2.9

40

can be calculated, i.e.,

I - IQ exp (-Eth/kT) (2.D

where I is the current measured, k is Boltzman's constant and IQ is

the pre-exponential constant. For films without OB, Z^ was 0.52 eV

-3
but was increased to 0.67 eV for films prepared at P = 1 x 10

2
Torr. In Figure 2.20, the optical density change (AOD) is plotted

as a function of the injected charge (Q) for the films deposited

at various oxygen pressure. The optical efficiency is defined as the

change in optical density per unit charge injected, and is the slope

of the plot AOD versus Q. The optical efficiency was decreased by

OB as shown in Figure 2.21. The absorption spectra for films prepared

at the two extreme OB conditions for the same thickness (2 urn) and

colored by the same injected charge 06 mC/cm ) are shown m Figure

2.22. For ease of calculation, the absorption intensity and the

wavelength were converted into relative absorption coefficient and

photon energy, respectively. The decrease in optical efficiency by

OB is again apparent in these spectra since the relative absorption

coefficient (normalized to unity for 'W02 ?') at the absorption peak

of 'W0? ' is only 65% of that of 'WO, ?'. The energies of the

absorption maximum, E , and the widths of the absorption band were
r op '

not changed significantly by OB, being 1.47 eV vs. 1.45 eV and 1.6
vs. 1.5 for 'W0? ' and 'W0? g', respectively.

The variation of bleaching time and color retention time
(open circuit memory) with oxygen backfilling are shown in Figure

41

1.0

0.8_

0.4

0.0

0.2-

Figure

20.

5 10 15

2
Charge (mC/cm")

Optical density change as a function of injected

charge for WO- films deposited with: (A) residual

pressure of 1 x 10 Torr,_^B), (C) , (D)_gxygen

partial pressure of 5 x 10 Torr, 5 xlO Torr
and 1 x 10 Torr, respectively.

42

P (Torr)

Figure 2.21. Optical efficiency of WO films versus oxygen
partial pressure during deposition.

43

1.25

1.0

0.8

3 0.4

0.2

0.0

600

Figure 2.22.

'00 800 900

Wavelength (nm)

1,000

Absorption spectra for W03 films prepared ^ at two extren

conditions. ' ' represents 'WO _ ' and

represents 'WO, ' The films were of same taic^ness
(2.0 Urn) and colored with same injected charge ( -o
mC/cm ) .

44

2.23 and 2.24, respectively. The bleaching times were found to

-3
decrease from 4.7 sec. without OB to 1.6 sec. at P = 1 x 10

2
Torr. After coloration, the power supply was disconnected and the

films would self-erase at different rates depending on the back-
filling pressure. The time for self-erasure to half of the original

0D decreased from 1,200 sec. for films prepared without 0

0

-3
to 460 sec. for films prepared at Pn = 1 x 10 Torr, as shown

2

in Figure 2.24.

Discussion

The corrosion of WO- thin films in H-SO, electrolyte proceeds

3 - ■+

in the form of general dissolution and interfacial delamination .
Previously, it was speculated that the water in electrolyte and/or
W0 films allowed or caused the corrosion. However, the mechanism
proposed could not fully explain corrosion in the acid-organic
solvent systems, nor was it compatible with data showing that W0
thin films are oxygen-deficient and stable up to 350 °C (crystalliza-
tion temperature). A new corrosion model will be proposed below.

Corrosion of WO Thin Films

General dissolution

The thermodynamic equilibrium between crystalline WO- and water
predicts that WO dissolves by hydrolysis to form tungstate ion;

45

6.0

y 4.0

0.0

2.0 -

P (Torr)

igure

2.23. Time needed to bleach the WO films to 50" of
the original optical density versus oxygen
partial pressure during deposition.

(E:

-0.6 V Thickness

3,000 A)

t+5

1,000

800

600

400 -

200

P (Torr)

Figure 2.24. Time for self-erasure to half of the original OD

of WO films versus oxygen partial pressure during
deposition.

47

WO, + H,0 = WO,2" + 2H+
3 2 4

log (W042_) = -14.05 + 2pH.

-4.9

(2.2)
(2.3)

The solubility at 25 °C is 10~ ' moles/liter in neutral pure
water and decreases exponentially with decreasing pH. In acid
solution of pH ^ 0.5 such as used in the current studies, the
solubility would be decreased to VL0_1 moles/liter if this reaction
is still valid at such a low pH. The equilibrium potential of W03
in this solution is 0.4 V with respect to the standard hydrogen
evolution potential (SHE).60 According to the Pourbaix diagram
for W (Figure 2.25),61 the experimental condition for storage
condition (pH = 0. 5 , E = 0. 4 V^) is located in the W03
stability region. That is, WO- is the thermodynamically stable
specie under these storage condition. In contrast, the vapor
deposited WO thin films dissolved comparatively fast even in the
small volume of low pH solution used in the present tests. To
explain this discrepancy, a thorough understanding of structure and
composition of the film and of the thermodynamics of corrosion is
necessary.

Previous studies and present results have shown that thermally

evaporated WO., films are

1. amorphous - X-ray diffraction pattern of these films
exhibited broad diffuse peaks typical of amorphous
structure;

2. oxygen deficient - XPS data also indicated that the^
tungsten atoms exhibited various oxidation states of
W , W and W due to the substoichiometry in these
films;

^E

Figure 2.25. Pourbaix Diagram for tungsten

Atlas of Electro-

chemical Equilibria in Aqueous Solutions, by M. Pourbaix :
Per^amon Press (1966). Reproduced with permission.

3. porous - the relative density is approximately 50% to
80% of that of crystalline WO ;

4. hydrated - films were reported to contain as much as 0.5
HO per WO^

The water might have been incorporated during W03 evaporation since

the partial pressure of water is the primary contributor to the mea-

35
sured total residual pressure. Water could also originate from

absorption of moisture from air after deposition since the films

were porous.

39
From the X-ray scattering experiments, Zeller and Beyeler

obtained the radial distribution function of these films and

concluded that the amorphous WO films consist of a disordered

network of corner sharing W-0 octahedra. The substoichiometry and

the existence of the W , W and W oxidation states in these

films suggests that tungsten atoms may exist with a configuration not

only as WO but also as W02 and W^. Therefore, it is logical to

consider the evaporated W03 films as a random solution of W02, W^

and WO . These molecules are connected with each other through

corner sharing O-bonds39 and forming a network-like structure (Figure

2.26). The water incorporated during evaporation or after deposition

is not expected to be important in structure formation, since

no observable structural change was observed at 170 °C when most

water was evolved/9 Water may be bonded in the open space between

tungstic oxide molecules by hydrogen bond or van der Waals forces.

The corrosion of these films in acid electrolyte can be

explained by examining the Pourbaix diagram for tungsten. Both

Figure 2.26. Proposed microstructure of vapor deposited WO^ films-
a random network composed of WCL . W^O and WO^ units
linking each other through corner-sharing oxygen bonds.
Note water molecules may occupy the empty space between
tungstic oxide molecules.

WO and W 0r are not thermodynamically stable in the storage
condition (pH = 0.5, E = 0.4 VSRE) . They should be converted to
higher oxidation states, but the conversion processes and products

"J Q

are both unknown. However, Nazarenko et al. have shown that for

2+
dissolution of tungstic oxide in acid (pH < 1) , the cationic, W02

2-

form dominates. The anion forms, e.g., WO (OH) 3 , WO^ only dominate

at higher pH (> 2). Di Paola et al. ' studied anodic W03 films
on tungsten and proposed that the corrosion of these WO^ films took
place by hydration of WO •

WO + 2H+ = WO "+ + H20 (2.4)

WO 2+ + (x + 1)H 0 = W03.xH20 + 2H+ (2.5)

where x = 1 or 2.
However, the free energy of hydration for forming both the mono- and
di- hydrates was shown to be positive. The hydration reaction was
observed if electrochemically driven. ' Although the mechanisms
for corrosion of WO- and WO, are unknown, it seems reasonable to
postulate similar reactions. For example, the dissolution processes
may be

W0 — * W09"+ + 2e" (2.6)

WO + 2H+ — • 2WO?2+ + H?0 + 2e" (2.7)

pH < 1.

2+
The solubility of W00 under these conditions is unknown, but the

ions as postulated precipitate as WO and its hydrates when the

solubility limits are reached. The total reaction can then be

WO + (x+l)H?0 = W03.xH20 + H2
WO + (2x+l)H 0 = 2W03-xH20 + H?

written as

(2.8)
(2.9)

where x = 0, 1, ar 2
and would result in the crystalline phases detected by X-ray diffra-
ction after 18 days storage in acid.

Even though XPS data (Figure 2.20) indicated the existence of

5+ 4+
high concentration of subvalence states (W , W ) in the evaporated

;ent? Gerard et al.

64

WO,, films, how much W0? and W 0 does this repre<

calculated the concentration of each valence states (W , W and W )

from XPS spectra of reactively sputtered WO- with varied 0/W ratio

6+ , ,
as shown in Figure 2.27. The concentration of W in a W09 g

films was 80% and decreased to only 40% for a 'W0? ' film. It

is reasonable to expect that the same relation could be found in

evaporated WO., films. For a film prepared without OB, the 0/W

ratio was about 2.7. Based on the data from Gerard et al. the

concentration of W is expected to be ^ 60% with the remaining 40%

tungsten atoms existing as W or W states. As discussed, the

W0 films can be considered to be a random network consisting of

WO , WO. and WO., molecules linking each other through corner sharing

0-bonds. Then every one out of three tungsten links in the network

exists with a configuration of W0„ or W 0^ . Since the oxygen

detected may have included that from the incorporated water, the

actual number of oxygen atoms that were directly bonded between

M

,5+

.4+

.0+

.WO.

WO,

wo

2.5

WO,

3

igure _.

Chemical Shift (eV)
6+ 5+

4+
W and

>+■

27. Concentration of W , W , W and w versus

corresponding chemical shifts for the sputtered
WO films with different O/W ratio. From Gerard
et3al., J. Appl. Phys. 48. 4253 (1977). Reprodu-
ced with permission.

tungsten atoms might have been lower as discussed in the last section.
Therefore, the concentration of W09 and WO could be even higher than
proposed here. In the acidic environment, these WO and W?0_ links
are selectively attacked and the whole network solid may be broken
down into a distribution of random length of W0~ chains. These
broken WO chains presumably can no longer stay in the bulk but drift

into solution. These WO chains may be the polymer ions in the

34
solution observed by Arnoldussen. As a result, the dissolution

rate is quite high. Once the solubility limits in solution are

reached, precipitation occurs on surfaces as crystalline WO^ or

hydrates, since the free energy of hydration may be negative starting

with WO and W^ .

Interfacial delamination

In addition to general dissolution, flaking by delamination at
the interface were also observed and may be induced by:

1. the internal stresses caused by precipitation of these cry-
stalline phases on the original sample;

2. gas evolution, e.g., 2H + 2e~ -* H? , accompanying the
dissolution and re-precipitation processes;

3. surface contamination and defects which weaken the adhesion
strength of the film to the substrate.

The first two mechanisms can be viewed as side effects of the genral

dissolution mechanism proposed. It is not uncommon to find some

pinholes O 10 urn) in the evaporated WO films. In the corrosion

processes, the dissolved ion concentration in these pinholes will

reach saturation limit much earlier than over the flat surface and
lead to the rapid precipitation of crystallites of hydrated tungstic
oxide. These crystallites serve as nucleation sites for further
precipitation of crystalline phases, and eventually grow to a size
larger than the volume of the pinhole. The induced stresses from the
crystallites may be transmitted radially outward in the films and
along the film-substrate interfaces since the pinholes go completely
through the films. The rosette-shaped wrinkles in Figure 2.6(B) are
typical result of this effect. The delaminated films can be easily
broken into fragments by physical disturbance of the electrolyte
(Figure 2.7(A)). Small bubbles were observed on the samples during
corrosion and are assumed to be hydrogen gas evolved during the
corrosion process. The gas evolution may also enhance the inter-
facial delamination.

Interfacial delamination was rare for films deposited on graphite
substrates (semi-polished). The rough surfaces on these substrates
may provide a mechanical interlocking through pore and valley. Gas
bubbles were most often observed on films deposited on graphite sub-
strates, since the rims of graphite substrates provided sites for
cathodic reactions. However, gas evolution is not thought to be a
major cause for film delamination. This is particularly true for
films deposited on NESA glass substrates because little gas evolution
was observed.

Corrosion during cyclic coloration and bleaching
The reported increase in the dissolution rate for cyclic
coloration and bleaching condition can be explained by nixed potential
theory60'66 as shown in Figure 2.28. Since the WC>3 specie in the
films should remain stable in this polarization range according to
the Pourbaix diagram, only W0„ and W20 species are involved in the
dissolution process. Figure 2.28 schematically illustrates the current/
voltage relations for the W02~H2S04 system, but the W^-l^SO^ system

is expected to be similar. The two reactions occurring are assumed

2+ -
to be W0„ dissolution (W02 -*> W02 + 2e ) and hydrogen evolution

(2H+ + 2e~ —* H„). The static corrosion potential and current for

this particular system is represented by the intersection between

the polarization curves for WO a dissolution and hydrogen evolution.

This represents the storage conditions discussed above. During the

bleaching pulse, the system is polarized anodically and the anodic

current density (I_T7 for W09 dissolution) is much higher than the

cathodic current density (X,. for hydrogen evolution) . Thus more

on

rapid dissolution of WO (and W^) is expected during bleaching.
However, the WO dissolution current, I , is much lower than
hydrogen evolution current, I , during the coloration pulse. As
a result, W09 (W 0.) dissolution is less but gas may be evolved
during coloration.

BH CW CH BW
Log Current Density

Figure 2.28. Combined activation polarization curves for W02 disso-
lution and hydrogen evolution show the effect of anodic
and cathodic polarization on the dissolution of WCy
E is the equilibrium potential (storage condition).
riC°rl and I are the overpotential , dissolution
current7 densily of W0? and reduction current of H?
during bleaching pulse, respectively. While nc, Icw

and I are those for coloration pulse.
CH

35
In addition, Faughnan and Crandall observed an increase in

dissolution rate when device were cycled at higher contrast. Higher

contrast would require either operating at increased C/B voltage at the

35 35

same frequency, or an increased C/B time at the same voltage. As

discussed above, an increase in the dissolution rate can be expected

32
when the bleaching voltage is increased. On the other hand, Rand in

reported shorter lifetimes when the devices were cycled at low

frequencies (< 0.5 Hz). This low frequency effect is consistent

with the contrast effect reported by Faughnan and Crandall if the

higher contrast was obtained by prolonged C/B time at the same

voltage. The causes of increased dissolution rates at low frequency

can be attributed to a number of possibilites . For example, at higher

frequency, the short bleaching pulse may allow a sharp concentration

gradient of WO 2+ to occur near the W03 film surface (due to limited

2+
dif fusion). This limits subsequent dissolution ot W02 and hence

the kinetics of dissolution would be impeded.

Non-aqueous electrolyte

Attempts to use non-aqueous electrolytes have been reported by
Randin32 and Reichman et al.33 The results are summarized in Table
2.1. The WO., thin films dissolved not only in sulfuric acid, but
also in formic acid and acetic acid, but remained stable in organic
solvents such as acetone, y-butyrolactone , and propylene carbonate.
The WO films dissolved when these organic solvents were combined
with sulfuric acid, e.g., methanol plus H0S0 or acetonitrile plus

Table 2.1

Stability of vapor deposited W0_ films in
various non-aqueous solvents or electrolytes.

Electrolyte (solvent)

Dissolution

Reference

Formic acid

Yes

Acetic acid

Yes

Acetone

No

n j. 32
Randin

Y-butyro lac tone

No

Propylene carbonate

No

H'SC^/glycerin (1:10)

Yes

32
Randin, ..Reichman

and Bard

*

HnS0; /methanol
2 4

Yes

33
Reichman and Bard

*
H„S0, /acetonitrile
2 4

Yes

concentration unknown

H„SO, . The dissolution rates were not specified for these mixtures.
2 4

32
Randin reported the dissolution rate in H-SO, /glycerin (1:10) was

an order of magnitude lower. These results indicate that in addition

to the substoichiometry of the films, the pH of the electrolyte is

also the culprit in corrosion. The water in the electrolyte can not

32-35
be the main factor as most investigators claimed, since a) the

corrosion does not occur until the organic solvents were combined with

acid, b) the dissolution rate in H„S0, /glycerin was decreased by only

an order of magnitude while the water content in the electrolyte

was reduced by many (^ 20) orders of magnitude lower. The water

incorporated during evaporation is also unlikely to be the cause

, 34
of corrosion in non-aqueous electrolyte as Arnoldussen claimed,

otherwise the corrosion should have been observed in acetone and

propylene carbonate. The lowered dissolution rate of WO film in

2+
H„S0, /glycerin could be caused by lowered mobility of WCL ions

33

because of the high viscosity and/or by decreased proton concen-
tration in the electrolyte because of the decreased dissociation
constant. However, no conclusion can be made as to which cause is
dominant .

bi

Modification of Stability of WO Films by Oxygen Backfilling

Since the substoichiometry in the WO. films is responsible for
the corrosion, an improved stability is expected at higher stoichiometry ,
By backfilling during WO, evaporation, the oxygen content of WO, films

was increased. The XPS data also indicated the concentration of W

5+ 4+
states in these films were increased relative to W and W . Since

there are less dissolution-prone species available, the dissolu-
tion rate of oxygen enriched films in acid electrolyte should decrease.
As shown in Figure 2.28, the dissolution did decrease and became
negligible when the 0/W ratio reached 2.9. Interfacial delamination
was not observed as a result of the reduced corrosion rates (Figure
2.7b). However, the dissolution and precipitation of small amount
of W0? and W 0_ still rearranged the surface portion of the films to
give crystalline peaks of WO- and its hydrates in X-ray diffraction
pattern from OB samples (Figure 2.14c). The bulk of the films remained
intact when OB was used. As indicated by Figure 2.7b; only at higher
magnification can the corrosion effect (the slightly roughened
surface) be observed. The X-ray diffraction data also indicated that
WOVH90 was the dominant phase in oxygen enriched films while W0^.2H90
was in the films without OB. This difference may be another reason
for lower dissolution in oxygen enriched films since WO..H 0 dissolves
less readily than W0,.2H_0. The cause for inducing different
dominant phase by backfilling is unknown.

Modification of EC properties of WO Films bv Oxygen Backfilling

The performance of electrochromic display devices is generally
determined by four electrochromic properties: 1. coloration speed
2. optical efficiency, 3. bleaching speed, and 4. coloration
retention time (self-erasure rate) . The effect of oxygen backfilling
on the EC display performance will be discussed in this order.

Coloration speed

.30 35
The rate controlling mechanism for coloration was proposed

to be current limited by the WO resistance at short time. The

duration of WO. resistance control increases as the thickness of WO was

increased. In addition, the current was affected by the Schottky and

Helmholtz double layer (HDL) barriers. At long time ( >1 sec.) for

thickness of <0.4 um, the-current control was dominated by the HDL.

As shown in Figure 2.16, the coloration current for films prepared

with or without oxygen backfilling showed a time independent behavior

-1/2
at the onset of coloration, and approached a t dependence at

I/O

t >3 sec. This t dependence is characteristic of HDL control.
In general, the charge injection process is not affected by OB.
However, the magnitude of injected current was lowered for oxygen -

enriched films. This can be explained by considering the Butler-

60
Volmer equation:

30

r (l-';)e~ -sen
JP - J0 [ 6XP M 6XP kT

(2.10)

where J is the proton current density across the double layer, JQ
is the exchange current density, 6 is the barrier symmetry factor,
k is Boltzman's constant, T is temperature, e is electronic charge
and n is the overpotential (the potential drop across the HDL when
J f 0) . Since 6 is a property of the ionic solution, it is not
expected to be affected by changes in the film properties. The
overpotential, n , can be expressed as

Au(x)
a NF Pi

(2.11)

where V is the applied potential, j! is the reduction in chemical

potential due to the proton concentration (x) in the W03, N is Avogadro's

number, F is Faraday's constant and R is the film resistivity. As

shown in Figure 2.18, the electron diffusion coefficient in WC>3 was

found to be decreased by OB. The conductivity (resistivity) measured

at room temperature was also found to be decreased (increased) by a

factor of three (Figure 2.19). Since Rf is higher at the onset of

the coloration (x = 0) , n would be smaller for oxygen enriched films

at the same applied bias V . The current injected would be lower as

a

indicated by equation 2.13.

Resistance control was dominant at t < 1-2 sec. However, the
effect on the magnitude of current injected persisted into the HDL
control region, i.e., the decrease in current by OB was almost
invariant with time even in the HDL controlling region (Figure 2.16).
However, the optical efficiency, i.e., the change in optical density

per unit charge injected, was also decreased by OB as shown in Figure
2.21. This effect together with the lower current injection led to a
slower coloration speed as shown in Figure 2.15. The mechanism of
degraded optical efficiency is discussed in the following section.

Optical efficiency

An understanding of the electronic structure of WCL thin films is
necessary before discussing the change in optical efficiency by oxygen
backfilling. In the following, the mechanisms proposed for optical
absorption and DC conduction are reviewed and compared with the
present results.

Electronic structure for WO thin films. Many models have been
proposed to explain the optical transition in electrochromic W03 films.
Optical absorption in these films is known to arise from a transition
of an electron from one trapping site to an adjacent site. A trapping
site is a position in the lattice where localization of an electron
will result in a lowering of the total free energy of the system. The
trapping sites are probably W sites as identified by electron spin
resonance studies. Faughnan and Crandall proposed an inter-valency
charge transition model which can be expressed as shown below:

W5+(A) + W6+(B) + hv > W6+(A) + W5+(B) (2.12)

where A and B represent two neighboring sites of tungsten atom,
and hv is the photon energy. This model explains qualitatively the
features of the optical absorption including the magnitude of oscillator

strength. However, the XPS data have shown the presence of large
concentration of W + and W O" 40%) even in the transparent W03 thin
films and hence appear to disagree with this model.

An alternative model based on small-polaron theory has been
proposed by Shirmer et al. " In a systematic study of the optical
properties of amorphous and crystalline WO films, they found a
shift in the maximum absorption peak from ^900 nm to \,1,200 nm upon
crystallization. This shift, plus the asymmetry of the high-energy
side of the absorption maximum (Figure 2.22) in amorphous films has
led Shirmer et al. to postulate a self-trapping term (i.e., a polaron
formation) which would lower the energy of the site with the trapped
electron relative to other W + sites. In a disordered system, the
W6+ sites do not all have equivalent energy because of the existence
of a range of W-0 bond lengths and angles. The injected electrons
will be trapped primarily at those sites of lower energy. Hence,
optical transitions will occur with a range of energies. This explains
the asymmetry in the optical-absorption peak. The shift Co lower
energy for crystalline films occurs because the disorder term
disappear for crystalline films.

The optically excited polaron hopping between two neighboring
sites can be described by a diagram showing the electron potential
energy as a function of a single, one-dimensional lattice configu-
ration coordinate (Figure 2.29). It has long been realized that

op

'th

MA MB
Conf igurational Coordinate

Figure 2.29. Conf igurational Coordinate model for the polaron
hopping absorption in the WO. films.

optically induced transitions between two sites can be considered as

72

Frank-Condon transitions, ~ indicated by the vertical arrow in

Figure 2.29. These processes take energy E which can be measured
from the peak energy of the absorption band. The change of the
lattice distortion accompanied by the electron transfer from site A
to site B is represented by Aq = q - q . The energy E is the
thermal energy of the small polaron hopping from site A to site B,
which can be obtained from the measurement of conductivity change
over a temperature variation as discussed below.

The DC conductivity of WO. amorphous films have been reported
by several authors. Faughnan et al. reported a logl - T
dependence at low coloration (x < 0.3 for H WO ) between 10.2 and
300 °K, and concluded that the electron transport in this condition

was best described by a variable-range-hopping model (electron

- 74,75 . . ..
hopping between trapping sites). Benci et al . in a similar

study reported a logl - T ' dependence for transparent film

between 4.2 and 300 °K, although a logl - T depencence could

described the data at temperature near 300 °K. They concluded the

conduction in transparent WO films at T<300 °K is in agreement

with the hopping between two localized states (a donor state and a

trap state). Gritsenko et al. studied the DC conductivity of WO^

films at a higher temperature range, 353 - 773 °K, and reported a

logl - T~ dependence and concluded that the overlapping of the

potential of deep centers was responsible for the conductivity.

The mechanism of electron transport between these centers was
thermally assisted tunneling and the conductivity increased exponen-
tially with temperature. Although the detailed nature of these
"donor states" ' and "deep centers" was not explained, they are
expected to be associated with the active polaron states where a
localized electron is allowed to contribute to conduction. The
exact nature of active polarons will be discussed in a later section.

It seems, however, that the DC conduction mechanisms of electrons
in WO films vary with the temperature range investigated. The
mechanism is associated with polaron hopping at low temperature
(T < 300 °K) , and with polaron tunneling at high temperature (353 <
T < 773 °K) . Hence, the thermal activation energy E h obtained at
high temperature will not have the same physical significance as
that from low temperature measurement. One needs to be cautious
in interpreting data to identify the controlling mechanism.

Effect of oxygen backfilling on EC characteristics and
electronic structure of WO- films. As discussed, the optical

efficiency (or the optical absorption per unit injected charge),
the DC conductivity and the electron diffusion coefficient of WO
films were decreased by oxygen enrichment, while the thermal acti-
vation energy (between 300 - -+50 °K) was increased. The decreased
optical absorption for more oxygen enrichment was also reported by

Arnoldussen. He observed that oxygen ion implantation of
evaporated WO films resulted in films which could no longer be
colored. This is the extreme case. He attributed this phenomenon
to the creation of electron traps by structural change as a result
of oxygen implantation. However, no detailed discussion on the
nature of electron traps or the mechanisms causing decreased optical
absorption was given.

In the discussion of optical properties of the trapped or
localized center as in WO thin film, the Smakula's equation is
applicable

N'f = 8.7 x 10

(K L)W.

; 2 . ,,2 max 1/2
(n + 2)

(2.13)

where N' is the number of absorption centers per cm , n is the index

of refraction of the film, K is the absorption coefficient at

max

the band peak in cm" , L is the thickness of the film and W^9 is the
full width at half maximum (FWHM) in eV. The index of refraction, n,

does not vary significantly between films deposited from W00 or WO^

,78
powders (which were assumed to result in a different O/1* ratio; or

9
between amorphous films and single crystal W0_. The absorption

spectra in Figure 2.22 are from film of the same thickness and

injected charge. As reported earlier, the width of the absorption

bands >' ,?) are nearly the same (1.6 eV for films prepared without

-3
OB, 'WO ', 1.5 eV for films prepared at P = 1 x 10 rorr,
2 . / ^9

'W0o '). Thus, equation 2.12 can be written as:

K = Constant \" f (2.14)

The analysis of degredation in optical efficiency therefore, can be
initiated by either of the following two hypotheses:

1. The charge in K was caused by the change in f, while N'
remained constanfXat constant charge injection (i.e., every
electron injected resulted in the formation of a polaron) .
The decrease in the absorption coefficient was therfore
caused by a decrease in the oscillator strength of the
absorption center.

2. The change in K was caused by the change in N ' , while f
remained constantX(i. e., the oscillator strength of absorp-
tion center remained unchanged by the backfilling) . The
decrease in absorption coefficient was caused by a decrease
in the concentration of absorption centers, possibly by
creation of electron traps by OB.

i 78
The first hypothesis was adopted by Yoshimura et ai. to

explain the increase in optical efficiency of tungstic oxide films

deposited from crystalline W0? powder. They observed the absorption

coefficient K was increased for films deposited from WO powder
max

as compared with those from WO powder (for the same film thickness,
injected charge and nearly equal FWHM's of the absorption peak). 3y
assuming every injected electron resulted in an absorption center
(polaron) and the use of equation 2.13, they concluded that the
oscillator strength, f, of films deposited from W0„ powder was
higher than that of films deposited from WO powder. They also
concluded that the increased oscillator strength was caused by an
increase in the degree of extension of electron wave function for
films from WO powder. Using a conf igurational coordinate model,
they suggested the change of lattice distortion Aq was decreased as
a result of an increase in electron wave function extension and

predicted a decrease in both E and E for films deposited

from WO- powder versus those from WO- powder. The predictions

were consistent with experimental observations. Therefore, they

concluded that the decrease in optical efficiency for films

deposited from WO powders was caused by a decrease in oscillator

strength of the absorption centers. The decrease in f was caused by

a change in the lattice configuration. Although they did not measure

the 0/W ratio in these films, they speculated that the change in

lattice condition was introduced by the deficiency of oxygen in films

deposited from WO- powder.

At first thought it would seem that the result in the present

study, i.e., the decrease in optical efficiency in 'W02 9' films can

be explained by a decrease in oscillator strength, ' f, using the

model of Yoshimura et al. The decrease in f was in turn caused by

a decrease in the degree of electron wave function extension caused

by oxygen enrichment. However, an increase in EQp was not observed.

The increase in E might not have been measurable due to a number
op

of factors that affect such measurements. For example, EQp was

reported to increase in energy as the optical density was increased

(as much as 0.15 eV) . However, the optical density at constant

charge injection decreased for 'W02 g' versus 'W02>?'. Since data

were taken at constant charge injection, the conf igurational coordinate

model predicts a higher E for 'WO. ' films, but the lower optical
r op ^- • j

72

density may have resulted in the wavelength of absorption peaks being
shifted lower and fortuitously coinciding with that of 'WO^-,'.

On the other hand, the derivation of the model by Yoshimura et al.
was based on a number of assumptions. One assumption was that the
concentration of injected electrons was equal to that of absorption
center. As will be shown later, there are a number of tungsten
sites where electrons may become highly localized and are not able
to cuase optical absorption or contribute to electronic conduction.
Therefore, the concentration of the electrons which can participate
in optical absorption is expected to be lower than that measured by
current integration.

A second assumption was that the DC conduction mechanism was
assumed to be associated with polaron hopping even for the tempera-
ture range 300 - 500 K, where the measurements were performed.
As discussed, the DC conduction mechanisms of electrons in WO^ films
varied with the temperature range investigated. The electronic
conduction in the studies of Yoshimura et al. was expected to be
controlled by thermally assisted tunneling as opposed to the hopping
they assumed. In addition, the polaron hopping mechanism predicts that

E = 1/4 E while their measured relationship was E ^ 2 E
th op °P

(i e E versus E were 0.75 eV and 0.55 eV versus 1.4 eV and 1.2
v ' ' ' th op

eV for films deposited from WO and W02 powders, respectively).

Therefore the E Yoshimura et al. measured does not have the same
th

TX

physical significance as the E in Che model they proposed (i.e., the

polaron hopping model) . The prediction by their model of an increase

in E which was not observed in this study further indicating that
op

the model may be invalid and that a different approach is necessary.

The decrease in the absorption coefficient Kmax of 'W02.9'
films (for the same charge injection), therefore, could have
been cuased by a decrease in the concentration of optical abosrption
cneters instead of a decrease in oscillator strength. However, the
decrease in the concentration of absorption centers must be associated
with an increase in the concentration of inactive electron traps as

a result of OB.

The existence of inactive trapping sites have been indicated by a
number of authors. Gritsenko et al. reported that the calculated
electrically active "deep center" concentration was two orders of magni-
tude lower than that of injected charge in WCy They speculated that
not all injected charge formed "deep center". However, no explanation

for such behavior was given. In addition, most plots of optical density

35
versus injected charge (e.g., Figure 2.20 and reported data ) indicated

2
that there is a threshold charge (y 2 mC/cm ) needed before coloration

can be observed. The threshold charge strongly suggests the presence

of inactive trapping sites. Thus, the decreased absorption center

density could be explained by an active-inactive polaron site model.

The electrons that are trapped at inactive sites will become highely

localized, and the possibility for such electrons to re-escape and

contribute to absorption or electron conduction is very small
compared to active polaron hopping. The energies of such inactive
sites are expected to be lower than those of active sites. Electrons
will preferentially occupy the lower energy states.

The origin of such inactive centers can be correlated with the

4+ ,5+
increase in water concentration or the decrease in W and VT by

oxygen backfilling. Deneuville et al. have proposed that cation
sites are improtant to coloration and that there are two types
of sites in WO films which protons can occupy. One type is
optically inactive, i.e., the proton does not contribute to colora-
tion. Protons incorporated during evaporation occupy these sites and
the as-deposited films are transparent. The other site is optically
active; coloration may occur when the protons in the optically inactive
sites are redistributed into active sites by changes in the chemical
potential of H within W0_. This in turn may be caused by changes in

the chemical potential of H in the hydrated dielectric electrolyte

79
as found for WO /MgF„/Au devices, or from the absorption of photons

80

as in ultra-violet irradiation. The essentially invariant change

in the H/W ratio in the films before and after coloration (reported
by Deneuville et al. to be a 3% increase) was used to support this
postulate of two sites. However, this model can not be used
verbatim to explain electrochromism since electrically-induced
coloration is accompanied with a large quantity of charge passed

through the cell and an opposite amount of charge passed in bleaching.

2
This charge is tens of mC/cm for deep coloration, whereas in the

75

redistribution model, the only charge transported could be that to

2 65
charge the double-layer capacitance, i.e., ^20 mC/cm". " Should the

charge be going into some other electrochemical side reaction, it

is extremely unlikely that such a reaction would be completely

reversible while having such a large capacity.

The idea of two types of cation sites is still acceptible, and

supported by the observed aging effect in W0_/Li device reported

14
by Knowles. The aging consisted of a gradual decrease in optical

contrast obtained during a fixed time potentiostat ic coloration

pulse. It was reported that some Li was present but did not make

a contribution to coloration in the bleached state. This residual

Li content increased with both the number of C/B cycles and the

depth of coloration, and was dependent on the degree of hydration

of the W0_ films. The dependence of residual Li content on the

degree of hydration of the films was postulated to result from the

evaporated WO,, film behaving as an ion-exchange material. The H

ion of an 0H~ group within the film was postulated to exchange in

a colored film for an Li ion from either the electrolyte or from

solid solution within the WO . In the latter case, a film colored

only by Li insertion would then be bleached by extraction of the

same total amount of charge comprised of both Li and H leaving a

residual concentration of Li . This would limit the rate of further

14
injection and result in the observed aging effect. Knowles

reported that a solution to this problem was ultra-violet illumination

76

of WO films held in as-bleached state. As a result of UV
illumination, the aged cell could be restored to close to its
initial performance.

Therefore, it seems that the idea of active-inactive cation
site can be related to the active-inactive polaron site model
proposed here. The inactive cation incorporated during deposition
might have introduced an extra lattice distortion on" the local
arrangement of W-0 entities. The extra lattice distortion might

cause 'an increase in the self-trapping energy for the trapped electron

71 8 T
by Coulumbic interaction. ' The electron trapped at such an

inactive site would require much higher energy to escape and the

possibility for such electron to contribute to conduction and

optical absorption would be low. Since the water concentration was

increased by OB, there were more inactive protons incorporated and

hence an increase in inacitve electron trapping sites.

On the other hand, the electrons that were localized around W

sites (W , W°+ states) in an_ as-deposited transparent film might cause

a decrease in the self-trapping energy for the electrons trapped

at neighboring W+ sites (e.g., by Coulumbic interactions). Therefore,

the electrons localized at these neighboring sites might have a higher

. ,.6+ .
tendency to escape and behave as active polaron. For the W sites

that were not adjacent to a W or W , the electrons that were

localized might self-trap in a deeper well and act as the inactive

4+ 5+
polaron. Therefore, when the concentration of W and M was

decreased by OB, there were less active polarons formed upon

TT

electron injection. As a result, the conductivity of the oxygen
enriched films was lowered. The increased thermal activation energy
in oxygen-enriched films might be caused by an increase in lattice
distortion as shown in conf igurational cordinate model. The lattice
distortion was increased by the decrease in active polaron
density, since the average distance between polarons was increased
and hence an increase in the lattice distortion would be required
for polaron motion between the sites.

In conclusion, the decrease in optical efficiency in oxygen
enriched films is best described as the result of an increase
in the active trap density by OB. However, the oscillator
strength of absorption center might have been decreased by oxygen
enrichment; present data are not sufficient to preclude a decrease
in oscillator strength but do show that the decreased optical
efficiency is not a result of changing only the oscillator
strength. While reasons for optical inactivity of selected
sites are uncertain, it may be associated with the lower energy
of the site and/or increased lattice distortion in the conf igurational

cordinate model. Increased distortion in this model could

4+ 5+
result from the presence of W , W and protons.

^w

Bleaching speed

The bleaching has been reported to be limited by proton

35
diffusion in a space-charged region. The current as a runction

of time is described bv

ICO - (P3kVp)1/4va1/2(4t)-3/4

(2.15)

where p is the volume charge density of protons (equal to proton

charge per unit area multiplied by thickness of the film), k is the

-12
relative dielectric constant of WO (e = 8.85 x 10 Farads/m) ,

u is the proton mobility and V is the applied voltage. The time
P ' a

for complete erasure is

4 2

t, = pL / 4kenu V
r 0 p a

(2.16)

where L is the thickness of the film. If the thickness of the films.

the applied voltage and the injected charge density are all kept

constant, t is inversely proportional to the proton mobility,

There are two major factors which affect the diffusion of protons

in the W0_ thin films:

1. amorphous structure - crystallization of the films is
known to decrease the proton diffusion coefficient by
an order of magnitude; '

in faster proton diffusion

'W

The microstructure of oxygen-enriched WO films remained amorphous,
but the water content was higher in these films as shown in Figure
2.11. Thus, proton mobility is expected to be higher. In addition,

the porosity in the oxygen enriched films was significantly increased
(Figure 2.13). The increased surface area and decreased diffusion
length caused by porosity will also facilitate the bleaching

process. As shown in Figure 2.23, the bleaching time was reduced

-3
by a factor of three for the films prepared at P = 1 x 10 Torr

2
as compared to the films without backfilling.

Self-erasure

The effect of increased proton mobility and reduced diffusion
length is also reflected in the self-erasure rate of OB films.

Q O

Hitchman has proposed that self-erasure is caused by the oxidation
of the so-called 'hydrogen tungsten bronze' HWO (which can also be
viewed as a solid solution of protons in WO ) back to WO^ according
to the reaction:

'.HWO + 1/2CL = 2W03 + H20

(2.17)

This is, however, a general description when the process is considered
from a thermodynamically point of view, e.g., it could be written as

2H+ + 2e + 1/20- — * 2H„0

(2.18)

where the H and the e are both removed from W0_ films. In

reality, the injected protons remained ionized and the colored WO,

35
films contained a large concentration of proton. " A chemical

potential is developed by the proton in the films resulting in a

back emf of 0.7 V for an OD of 1.0, and this is the driving force
for the self-erasure process. Therefore, similar to bleaching, self-
erasure is controlled by concentration-dependent proton diffusion in
WO films. Since the proton mobility was increased and the
diffusion path was reduced in the oxygen enriched films, the
coloration retention time was decreased by approximately a
factor of three (Figure 2.24). The magnitude of the reduction
was the same for bleaching and self-erasure. This again indi-
cates that proton transport was the controlling mechanism in both
processes and was modified by the oxygen backfilling.

Summary

1. Vapor deposited WO. films have been confirmed to be amorphous,
oxygen deficient (0/W = 2.76), and porous (>80% of bulk density).

2. X-ray photoelectron spectroscopy data have confirmed that the
tungsten atoms in the as-deposited films existed not only

as W , but also as W and W . The total concentration of
subvalence states can be ^40% or higher.

3. The tungsten atoms in the as-deposited films are postulated to be
present as WO , WO and WO according to their valence states.
The films structure can be visualized as a random network
composed of WO and lower valent oxide molecules linking with
each other through corner sharing oxygen bond.

4. The Pourbaix diagram for W indicates that WO is the thermo-
dynamically stable specie for WO films stored in the acid

01

electrolyte (pH = 0.5, E = 0.4 V ), while W02 and W?0_ are not.
The conversion of WO and W?0 into WO were observed as dissolu-
tion of WO adn WO in acid to form crystalline WO. and its

hydrates (i.e., WO.. HO and W0..2H 0). The reactions are

2+
uncertain, but appear to involve the formation of WO ions in

solution and saturation of these solution.

5. The dissolution of as seposited films was rapid ( ^16 A/hr.) due
to the presence of these unstable species. In addition to

the general dissolution, interfacial delamination between
the film and substrate was observed. The precipitation of
crystalline WO and hydrates on the original film, particularly
in the pinholes is postulated to cause delamination by intro-
ducing internal stresses which exceed interfacial adhesion
strength between film and substrate. The hydrogen evolution
accompanying the corrosion processes might also contribute to
the delamination.

6. Corrosion was increased when films were cycled between colored
and bleached states because the dissolution rates of WO. and W^O^
were greatly increased by anodic polarization during bleaching.
The corrosion was still observed when the Ho0 in the acid electro-
lyte was replaced by glycerin indicating pH rather than H90 content
of the electrolyte defined whether corrosion would occur.

7. The oxygen content in WO films was increased by oxygen back-

4+ 5+
filling. The concentration of subvalence states (W , W )

^7T

decreased with respect to W states. The stability of the films
in acid was improved becuase of decreased concentrations of
dissolution-prone species in the films. The dissolution and

delamination was negligible for films prepared at P = 1 x 10

-3

Torr, (0/W = 2.9).

8. The coloration speed was slower for oxygen enriched films

because of lower current injection and reduced optical efficiency.
The current injection was found to be controlled by the same
mechanisms as for films without backfilling (i.e., WO- resistivity
and Helmholtz double layer). However, the magnitude of the
current passed was lower because the effective potential across the
HDL was decreased by the increased film resitivity from back-
filling. The increased resistivity and decreased optical
efficiency in the oxygen enriched films was explained by postu-
lating an increased density of inactive electron trapping sites.
This increased density of inactive trapping sites limited
the formation of active polaron for optical absorption and DC
conductivity. However, the possibility that a decreased oscilla-
tor strength resulted in a degraded optical efficiency could not
be ruled out .

9. The porosity of the films was increased by deposition at high
background pressure. Due to increased effective surface area,
the films absorbed higher amounts of water after deposition as
shown by secondary ion mass sepctrometry . The bleaching speed

^5T

and self-erasure rates were increased since proton mobility was
increased and the diffusion path was shortened by this porosity
effect.

CHAPTER III
DEVICE FABRICATION

Introduction

In Chapter II, the corrosion of EC display system based

on WO -H„SO,/H„0 was investigated. A solution to corrosion,
3 2 4 2

viz., oxygen backfilling, was studied and shown to increase the device
lifetime. However, the improvement was accompanied with degredation
of EC properties.

In this chapter, modification of the electrolyte to increase
the device lifetime is considered. Two types of ECD systems were
investigated and fabricated. The first type involved the use of a
liquid electrolyte consisting of LiC104 dissolved in propylene
carbonate (Li/PC). Since the devices using this electrolyte were
reported to exhibit slow switching speeds, " two approaches were
adopted to increase the speed, viz., nitrogen-backfilling and MgF2 over-
layer. The switching speed was found to be increased by depositing the
WO films at high partial pressure of nitrogen and by depositing a thick.
dense MgF. overlayer. The second device configuration involved the
use of a hydrated solid dielectric layer, i.e., a MgF2 film deposited
with air-backfilling. The results indicated that air backfilling
during MgF film deposition was effective in achieving fast
switching, while air backfilling during deposition of W03

84

^x

films limited the operational area of the device and may not be
desirable. The mechanisms will be proposed to explain these results,

Experimental

Preparation

WO films were prepared by thermal evaporation of pure WO^
polycrystalline powder (Cerac, Inc., 99.9%) from a resistively
heated source onto indium- tin-oxide (ITO) coated glass. The source
used was an alumina coated tungsten basket (Sylvania CS-1010) . The
surface resistance of the ITO coated glasses ranged from 15 Q/ to
100 Q.I . However, during the experiment, the resistance of the
ITO within the group of the samples to be analyzed was kept constant.
The deposition rate was varied between 3 A/s and 30 A/s using a
thickness monitor (Sloan DDC1000) and manually adjusting the power
to the evaporator. Before evaporation, the chamber was evacuated
to a base pressure of 6 x 10_6 Torr. The evaporation was accomplished
either with a partial pressure of air or nitrogen controlled by a
gas flow controller (Vacuum General) or directly without gas back-
filling. Thin films of MgF2 were prepared by thermal evaporation
MgF powder from Ta boat onto W03 coated ITO glass. Deposition rate
was kept between 5-10 A/s. The pressure during the deposition was
varied between 4 x 10~6 and 5 x 10_4 Torr of air or nitrogen. Thin
gold films were prepared by DC sputter deposition from a small coating

86

system for SEM sample preparation, or by vapor deposition from an
electron beam deposition system. The substrates were cleaned by
rinsing with soapy water (Alconox with water) for 15 minutes
followed by hand scrubbing with cotton wool. The substrates were
then rinsed in deionized water for approximately 10 minutes,
isopropyl alcohol (I?A) for one minute, and vapor degreased for
5 minutes in a IPA/Freon tank. The liquid electrolyte was prepared
by dissolving LiCIO, in propylene carbonate (PC) to make a 1 M
solution. Neither LiCIO, nor PC was processed to eliminate
moisture before or after mixing.

Characterization

The relative density of the evaporated WO films was changed by-
nitrogen backfilling and was measured by the same method reported in
Chapter II, i.e. mass change versus surface step height. The
surface topography of the films was examined by scanning electron
microscopy (Cambridge XSEM-9) . For Li/PC EC devices, the glass
substrates deposited with WO. films or W03/MgF? films were assembled
into an EC cell as shown in Figure 2.3. The coloration/bleaching
(C/B) speeds were measured with the apparatus shown in Figure 3.1.
A gold wire loop was used as the counter electrode. A constant
potential, 2.75 V between Sn09 and Au electrodes, was used to

drive the cell (Heathkit IP-28) . An Ar-Ne laser (633 nm) was used

2
in the measurements. The operating area was 3.88 cm". The solid

*<%

-ens z 3.T.Z
Voltage

v7777}

//77? >

u

1 \

Spacer

■ ■- * z

'

K s

■ \

S-e Recorder

r-

He-Ne Lise:
6 33 ~a

«'0-,/:T0/Glass [OZ)-

<ecorce r

Ci " Councar ^.dc.r:ce
OE - Display Ilec:rcae

Figure 3.1. Electrochromic cell circuit diagram for
the measurements of C/3 speeds.

state devices were fabricated by successive deposition of WO^,
MgF and Au films as described above (Figure 3.2). The thickness
of WO , MgF and Au was 4.0 KA, 1.5 KA and 0.2 KA, respectively.
The C/B speed was measured in the reflection mode.

Liquid Electrolyte Device

Results

As shown on Table 3.1, the relative density of the W03 thin

films increased as the deposition rate (DR) increased (for a fixed

partial pressure of nitrogen), or increased with decreasing partial

pressure of nitrogen (P ) for the same DR. As shown in Figure 3.3,

2 _4

the films prepared at high background pressure (P = 7 x 10 Torr)

2
exhibited a more rough surface topography than those prepared at

low pressure. The films prepared without backfilling were feature-
less, even at a magnification of x 20,000.

The effect of deposition rate, on the coloration time (t^) for
2,000 A thick WO films to a contrast of 30% absorption are shown in

Figure 3.4. The t remained nearly constant up to a DR of 10 A/s

° c

then increased drastically for a DR of 30 A/s. The data in the same

figure also show that t decreased as the partial pressure increased

-5 "4

from the residual pressure (1 x 10 Torr) to P =7 x 10 Torr.

2
This effect of backfilling pressure on t is again illustrated in

Figure 3.5 for films deposited at the same DR (5 A/'s) at three

^T

ITO

WO /MgF,

ITO

Au(0.2 KA)

MSF (1.5 KA)

2

WO (4.0 KA)
3

Figure 3.2. Schematic representation of the ITO/
WO /MgF /Au EC device.

Table 3.1

Effect of nitrogen partial pressure and deposition rate
on the relative density of W0_ film

D.R.

P„ (Torr)
N2 v

1 x 10

-5

2 x 10

-4

7 x 10

-4

3 A/s 5 A/s 10 A/s 30A/s

0.8

1.25

0.91

1.33 1.33

0.67 0.67 0.67 0.73

There are several possibilities for the relative densities of
the WCL films to be greater than one; (1) Calibration errors,

(2) The presence of lower oxides, e.g,
density of W0„ is 12.1 g/cm ).

WO..

and W„0- (bulk
2 3

91

00 094

Figure 3.3. Scanning electron micrographs of WO films showing

the effect of nitrogen backfilling during deposition
on the surface topography of the films;
(A) P = 1 x 10, Torr, x 20,000,

(B) P

'= 7

10

-U

Torr, x 20,000,

92

Deposition Rate (A/S)

Figure 3.4. Effect of deposition rate variation on coloration time for
films prepared at two extreme conditions. solid circles:
residual pressure of 1 x 10 Torr and_4open circles:
partial pressure of nitrogen at 7 x 10 Torr. The films
were colored to a contrast of 30% absorption (E = -2.75 V
on WO,). Substrates used were high resistance ITO coated
glasses (75 Q/Q ) .

^x

7 x 10

P.. (Torr)
N2

Figure 3.5. Effect of variation of nitrogen partial pressure during
deposition on coloration time for WO films colored to
different contrast: solid circles: 40% and open
circles: 30% absorption (E - 2.75 V on W0-) . Substra-
tes used were low resistance 1T0 glasses ( 15 .:/ a ) .

94

different partial pressures of nitrogen (P = 2 x 10 , 7 x 10 ,

1 2
2 x 10~3 Torr) . The t generally decreased as the partial pressure

of nitrogen was increased.

The effect of resistance in the ITO layer upon t is evident

when comparing data in Figure 3.4 and 3.5. The average tQ for films

prepared at 7 x 10~4 Torr and 5 A/s on an ITO glass substrate with

a resistance of 75 Q/n (Figure 3.4) was 9.3 sec. The average tc

for the films with the same deposition condition but on a more

conductive ITO glass substrate (15 fi/a ) was only 6.0 sec. (Figure

3.5).

The effects of a MgF, over layer is shown in Table 3.2. The W0^

thin films were prepared under the same condition (7 x 10 Torr

of nitrogen, 2,000 A thick). Magnesium floride overlayers of

500 A or 1,500 A thickness were deposited at P = 6 x 10 Torr.

2
Other films with thickness of 1,000 A and 2,000 A were deposited at

-4
a pressure of P, = 5 x 10 Torr. The device were colored to

N2
a contrast of 30% absorption. Data in Table 3.2 show that the tc was

generally lower for MgF„ films prepared at lower pressure and for

larger thickness. The maximum decrease was from 12 sec. for WO^ films

without MgF overlayer to 8.5 sec. for WC>3 with a 1,500 A MgF, deposited

at 6 x 10~3 Torr.

The bleaching time, t, , also varied in the same manner with
b

increased DRorP, as shown in Figure 3.6. However, the change from

2
10 A/s to 30 A/s was similar compared to those for t .

95

Table 3.2

Effect of MgF overlayer on the coloration
time of WO -Li/PC devices

\\Th i c kne s s

N2 ^\

0.5 KA

1.0 KA

1.5 KA

2.0 KA

6 x 10~5

10.75 s

8.50 s

-4
5 x 10

_i

10.00 s

9.50 s

J- '

^w

^ 20 _

Deposition Rate (A/s)

Figure 3.6. Effect of deposition rate variation on bleaching time for
films prepared at two extreme conditions as described
in Figure 3.4. The films were bleached from 30% to AA
absorption (E = +2.75 V on WO^ . Substrates used were
high resistance 1T0 coated glasses (75 .7 D ) .

No corrosion of WO films in Li/PC electrolyte was observed.
The device was operable even after storage for one year.

3
15,16

Discussion

The charge transport mechanisms for Li insertion in WO

thin films has been discussed by several authors. Mohapatra

-9 2 -1 , . .+
reported a diffusion coefficient of 6 x 10 cm s for Li

in evaporated amorphous WO films, However, he concluded that

the rate limiting processes during coloration was charge

transport across the interfacial barrier between WO^ and the

electrolyte, similar to HDL control in acqueous electrolyte.

Greenlfi has .reported diffusion studies for sputtered polycrystalline

WO. thin films with a typical grain size of 250 A. Diffusion was

described as being relatively rapid along grain boundaries followed

by slow diffusion within the grains. Evidence for this model is

provided by the observed reduction in response time with reduction

in average grain size in WO films of equal thickness. The kinetic

data were interpreted as corresponding to chemical diffusion coe-

— ft 7 —1

fficients within the grain of 6 x 10 cm" s and those in the grain
boundaries of 2 x 10_12 cm2 s_1. Ho et al. have also recently
measured Li+ diffusion using ac impedance techniques in evaporated
and crystallized WO films. The films were porous (66% of bulk
density). The structure was either polycrystalline with a grain size
ranging from a few tens to a few hundred angstrons, or consisted of

98

some large grains embedded in amorphous material. Under the small
perturbation condition employed, the Li chemical diffusion coeffi-
cient was found to increase with Li content, from 2.4 x 10
cm2 s"1 at x = 0.097 to 2.1 x 10~ cm s~ at x = 0.26 (x refers
to the ratio of H to W atom) . No evidence for sequential grain
boundary/grain diffusion was found since the data could be adequately
fitted by a single diffusion coefficient. Ho et al. also
proposed that both diffudion and interfacial kinetics are important
for the injection of Li into W0„ films. He concluded that at long
times, the interface is at equilibrium and the diffusion of Li in
a concentration gradient limits the rate of coloration. At time
shorter than about 0.5 sec, however, charge transfer across the
WO. /electrolyte limits the rate of injection. Thus it appears as
if the controlling mechanisms for coloration processes in the
W0 /Li+ EC cells are different for amorphous versus crystalline W03
films. Since the WO films in this study were amorphous, the
controlling mechanism for coloration at long times was expected to
be interfacial barrier, i.e., Helmholtz double layer, control.

The bleaching process was generally postulated to be limited

by the diffusion of Li ions in the space charge regions by all

15-18
the above authors.

In the present study, the switching times decreased and the
porosity increased as a result of lower deposition rates or higher

99

partial pressures of nitrogen. Therefore, it seems reasonable to

correlate the decrease in switching times with the increase in

porosity. However, the decrease in times was not generally

consistent with the porosity change. For example, the relative

densities were 0.8 both for films prepared at 3 A/s, 1 x 10 Torr

and at 30 A/s, 7 x 10~ Torr, but the average t was 11 sec. for

the former and 26 sec. for the latter films. Hence, the porosity

is not sufficient to explain the change in switching speed and

detailed consideration of the DR and P on the nucleation, growth

2
and structure of thin films is necessary to explain these

phenomena.

It was expected that during film growth, the higher deposi-
tion rate could lead to an increased surface mobility of adatoms.

Higher adatom surface mobility would lead to a reduced island

- i 84
density which would increase the agglomeration of the rilms.

The resultant films would have a larger grain size and smaller number
of frozen-in structural defect, thus the film density would be
higher.

For films deposited at lower pressure, the surface contami-
nation by residual gas would be less and hence result in an
increased nucleation density. The surface mobility of the
adatoms at lower pressures was expected to be higher due to
increased mean free path since there was less impingement of gas

100

molecules. In addition, there was less gas incorporation
possible at such a low background pressures. Therefore, the
resultant films were expected to exhibit higher density.

Therefore, both higher deposition rate and lower background
pressure increase the film density but the mechanisms are not the
same. Since low density may corresponded to a much rougher surface
topography, the total effective areas of the films exposed to the
electrolyte was increased. The amount of charge injection per
unit time was expected to be increased by this increased surface
area and result in the decreased coloration time (t ) as shown
in Figure 3.4 and 3.5. In addition, the diffusion path was
shortened by porosity in the films deposited at high background
pressure. Therefore, the bleaching times were decreased.

However, different combinations of these two parameters
might have resulted in totally different film structures, but
with the same relative density. These decrease in relative density
might be caused by either an increase in the porosity or an increase
in the entrapment of elements with lower mass (i.e., nitrogen gas
molecules) . The gas molecules entrapped are expected to be physically
or chemically bonded to tungstic oxide molecules. The formation of
effective pores was not possible if these entrapped gas molecules
were finely dispersed and isolated. However, the gas molecules
might aggolmerate together and caused a pressure build-up

within the pores. Such pores became 'effective' when the pressure

was high enough to penetrate the tungstic oxide matrix, release

the gas and form interconnects between the pores. The surfaces of

these pores would be exposed to the electrolyte during operation

to increase the Li injection and extraction, and hence are termed

'effective pores'. For films deposited at high DR and P , the

2
film density was expected to be high because of high agglomeration

rate of island at high DR. However, a certain amount of gas

couls have been trapped during film growth since the background

pressure was high. The growth of the effective pores by gas

entrapment was expected to be limited by the high agglomeration rate

of islands. Therefore, the effective pore volume exposed to the

electrolyte was limited. The porosity effect on the C/B speed was

therefore limited. While the initial nuclei density was high

for film deposited at low pressure and DR, the surface mobility

of adatom was low due to low DR. The adsorption of residual gas

was comparable with that of adatoms. During film growth, the

pores could be well developed and interconnected. The total

active surface area could be much higher than that from high DR

and P . Therefore, the same pore volume but different type of
N2

structure led to variation in t 's.

c

The limitation of switching time by the resistance of the

transparent conductive layer (ITO) on the glass substrate has

25
been reported by several authors. Hamada et al . proposed an

102

emprical formula to illustrate this effect
t - Q

IV

Ernl

(3.1)

appl CD

where t is the response time, Q is the charge injected to achieve
the required contrast, R is the series resistance which is the sum
of the internal WO resistance and any external resistance (dominated
by the ITO layer) , and E is the difference between the equilibrium
potential of WO- and that of counter electrode in a given electrolyte.
The ITO resistance effect can be illustrated by examining the voltage
drop due to resistance between the electrical contact and the

active electrode area (i.e., IR drop). The value of the current

2
per unit area for coloration of WO was about 1.0 mA/cm". The

resistance of the electrical path from contact to electrode was

approximately 750 2 (for 75 Sty a from a contact 2 mm wide and

20 mm long). Therefore, the electrical current for coloration and

bleaching caused a relative large IR voltage drop (^ 0.75 V). As

the resistance of the ITO layer is reduced, the IR voltage drop

will be decreased and a larger voltage will occur across the HDL.

This will cause a more rapid charge injection. Therefore, higher

conductive ITO glass provides higher charge injection and lead to

faster coloration speed as illustrated in Figure 3.4 and 3.3.

8 5
Yoshimura et al. reported that WO^ is insulating for film

thickness below 5,000 A. For film thickness above 5,000 A, WO.,

103

exhibits a semiconductor nature and the conductivity increases

almost linearly with increasing film thickness. The reason for

85
this behavior was explained as the depletion layer formation

on the surface of WO_ films. The depletion layer was formed
because of the discontinuity at the surface, which allowed surface
states to be created. These states might trap the majority carriers
causing recombination and low conductivity. For W0_, which is an n-
type semiconductor, surface states trap electrons as shown schemati-
cally in Figure 3.7(A). However, it was reported that when a MgF2

layer was deposited onto WO,, the conductivity increased

85
dramatically. The MgF~ is an insulator and the resitivity

is at least several orders of magnitude higher than that of WO^.

It was suggested that deposition of MgF removed the surface

states as shown in Figure 3.7(B). This explains why the coloration

speed was increased when a surface layer of MgF2 was present

(Table 3.2). Since there is a smaller voltage drop across the WO^

films, the resistance limit at short times is reduced and the HDL

controlled charge injection occurred much earlier. The data in

Table 3.2 also indicate that this depletion layer removal effect

is less pronounced if the deposited MgF films are thin and

porous. The porous MgF films resulting from deposition at high

background pressure or thin MgF films may not be -able to provide

a continuous coverage to remove the surface states due to the

discontinuity at the WO surface, and hence the effect is less

significant. In conclusion, the stability of WO^ films in Li/PC

uw

(A)

C. B.

Surface
States

V. B.

WO.

(3)

C B.

V B.

WO-

MgF2

Figure 3.7. Energy band model of electrochromic device
(A) before and (B) after the deposition of
M^F^ overlaver.

105

electrolyte is excellent. The C/B speed could be increased by
a porous WO- film, by a MgF? overlayer and by more conductive
ITO substrate.

Solid Electrolyte Device

Results

A qualitative study was conducted to determine the effect
of air-backfilling on the WO and MgF films in the solid state
devices. The results are shown in Table 3.3. The device made

with WO without backfilling and MgF with or without backfilling

2
were easily colored, even over larger areas ( V36 mm ). The C/B

speed of these devices are shown on Table 3.4. With higher air-
backfilling pressure during MgF deposition, the t^'s and t 's both

decreased.

-4
The devices made of WO with air-backfilling (Pair = 5 x 10

Torr) and MgF with or without backfilling can only be colored in

a limited area ( ^2 mm"). The memory is even poorer in the device

made with MgF with air-backfilling than that without backfilling;

self-erasure to a completely bleached state occurred within several

seconds. The cyclic lifetime of such device was limited. The device

eventually failed to color when the Au film flaked off from the

MgF layer as a result of prolonged cycling ( '^100 cycles).

106

Table 3.3

Effect of air-backfilling during WO and MgF deposition
on the electrochromic colorability of W0-/MgF„/Au solid devices

^\ W03

MgF2^^^^^

No
air-backfilling

with
air-backfilling

No
air-backfilling

Colorable

Colorable in
limited area

With
air-backfilling

Colorable

Colorable in ex-
tremely limited
area, poor memory

107

Table 3.4

Effect of air-backfilling for MgF2 on Che C/B
times of WCL/MgF_/Au solid state devices.

Time (sec.)

P = 1 x 10 Torr
res.

4.00

6.75

P = 5 x 10
air

1.75

2.50

108

Discussion

Device based on WO~/MgF„/Au have been reported by various
authors. Yoshimura et al. reported that water in the ambient
atmosphere or water near the MgF„ film surface is dissociated at
the counter electrode, generating protons which move through the
MgF„ films into the WO films and thus contribute to coloration,
i.e.,

Electrolysis in MgF„

WO.

Au

2xH?0 --* 2xH+ + 2xOH" (3.2)

xH+ + W0„ + e~ -^ H WO., (3.3)
3 x 3

2xOH" -- * xH„0 + 1/20, + 2xe~ (3.4)

During bleaching, the protons are ejected from the WO, films and
H» gas is released from the semi-transparent Au electrode, i.e.,

WO

MgF,

Au

2H WO, -* 2xH + 2W0, + 2e
x 3 3

2xH + 2x0H -- • 2xH?0

2xH + 2xe -- »• :<H„

(3.5)
(3.6)
(3.7)

Measurements of H and 0 evolution and weight change of the
device during operation have supported these postulated mechanisms.

It is quite apparent that increasing the water content in
both MgF and WO films could increase the switching speed of
the device. However, it is difficult to increase water content
simply by water backfilling. Therefore, air backfilling seemed to

Q c

be a good substitute. Yoshimura et al. have shown that the effect

109

of air backfilling can be ascribed to the water in air. The

decrease in t and t, by increasing partial pressure of air during

c b

MgF„ deposition can be attributed to:

1. The porous structure resulted from deposition with

a high background pressure could increase the adsorption
of water during subsquent exposure to the atmosphere. The
total water content in the films, therefore, would be the
sum of the water incorporated during deposition and the
water absorbed after deposition. It will be much higher
than those without backfilling;

2. The porous structure of the MgF? films also facilitate
the transport of the dissociated ions.

However, air-backfilling during WO film deposition was
unnecessary, especially from the viewpoint of device fabrication.
It is believed that the porous structure and higher water content
of the WO,, films prevented the creation of uniform and dense
MgF films; dielectric breakdown occurred readily in the resulting
thin film sandwich configuration. In addition, even a 1 Ptn parti-
culate contamination was enough to make the device inoperable.
Therefore, the operational area of the device was limited. The

diffusion coefficient for proton in the WO films prepared without

-8 2-19
backfilling was already sufficiently large ( 'VLO cm s )

that the coloration was expected to be limited only by interracial

barrier. Hence, the modification of the interface between

WO and MgF will be a more proper approach to try to achieve faster

8 S
foshimura et al. reported the water absorption at the

response ,

WO /MgF interface does increase the coloration speed.

110

The poor open-circuit coloration memory for the devices
fabricated from WO and MgF„ both with air-backfilling again could
have resulted from the high porosity and water content in these
films. As mentioned in chapter II, self-erase resulted from protons
diffusing to the WO /electrolyte interface and reacting with oxygen.
The high porosity and water content in MgF„ films were expected to
more rapidly provide the oxygen needed for self-erasure.
Alternatively, the proton could also combined with OH to form
water as described above. The porous structure can expedite the
the removal of the reaction products from the interface region.
Hence, the self-erasure was faster in these devices. In addition,
gas evolution during operation would lead to a premature failure of
the device either by loss of contact between the MgF? films and the
counter-electrode or by exhaustion of the water in the MgF9 films
as reported by Yoshimura et al .

An ideal solid state EC display, therefore, should employ
a configuration that is free from such side reactions. A rever-
sible counter-electrode and a dielectric layer which has strong
holding power for water would be the main requirements to satisfy

these conditions. Devices based on this principle, such as

87
IT0/W0„/Ta„0r/Ir0 /ITO, have been reported and exhibited fast
3 2 5 x

response (50 ms for 0D of 0.5) and long lifetime ( >10 cycles).

Ill

Summary

1. In liquid WO -Li/PC devices, the porosity of WO films was
increased by deposition at high partial pressure of nitrogen
and low deposition rate. The coloration/bleaching speeds were
slightly increased by the increased porosity possibly due to
increased surface area for Li injection during coloration, and
shortened diffusion path for Li extraction during bleaching.

2. The increase in the speeds was not generally consistent with

the porosity change. It is possible that the pores introduced by

high deposition rate and high P were not effective as those

2
from low deposition rate and low P . Therefore, a variation

2
in the switching speeds was observed even at the same relative

density.

3. The coloration speed was found to be slightly increased when
a dense, thick MgF„ overlayer was deposited on the WO^ film.
Increased coloration speed was postulated to be caused by an
decrease in the duration of WO film resistivity control at
the onset of coloration.

4. The resistanceof the transparent indium tin oxide layer was
found to limit the effective potential across the WO /electrolyte
interface. Therefore, a highly conductive ITO layer is
desirable for a fast switching EC device.

5. In a solid WO /MgF„/Au device, the water content in the MgF2 film
was found to be increased by air backfilling during evaporation.

112

The coloration/bleaching speeds were increased because of
increased proton mobility caused by increased water content in
the MgF9 film. The current injection was limited by the
interfacial barriers between WO /MgF rather than by the proton
diffusion in MgF films.

6. Air backfilling during WO films deposition appeared to be un-
desirable for solid electrolyte devices, since the porosity and
rough surface prohibited a uniform sandwich layered structure
to be formed and led to local dielectric breakdown. The
operational area of such devices was limited.

7. The lifetime of WO /MgF./Au was limited ( "-100 cycles). The
device failed as a result of the loss of contact between the
Au electrode and the MgF„ film. Exhaustion of water in MgF?
film from gas evolution during operation may also have occurred

CHAPTER IV
SUMMARY

Conclusions

Electrochromic tungsten trioxide films have been studied in
relation to their application to non-emmisive display devices. Elec-
trochromic display devices exhibit a number of attractive features such
as no limitation on viewing angle, low power consumption, open-circuit
memory, and good contrast at high ambient light. Therefore, they have
varied applications not possible with other conventional display
materials. However, there are some drawbacks. For example, the
switching speeds are slow, lifetime is limited ( ^10 cycles) and they
color only to blue in transmission. Among these three, the lifetime
problem is the principal obstacle to the commercialization of electro-
chromic displays. The lifetime of the device was found to be limited
by the corrosion of WO- films in the H SO electrolyte. The goal of
the present research was to investigate the corrosion mechanisms and
seek methods to improve the stability of the device.

The tungsten trioxide films for this study were prepared by
vapor deposition. The as-deposited films were amorphous, oxygen
deficient (0/W = 2.7), and porous (80% of the bulk density).
Becuase of the oxygen deficiency, the tungsten atoms existed not only

as W but also as W4 and W as evident from the data of XPS . Cal-

4+
culations indicated the concentration of these subvalence states (W ,

113

114

W°+) could be very high (approximately 40%) . The tungsten atoms with
6+, 5+ and 4+ states may take the configuration of WO_, W^O^ and W02,
respectively. The microstructure of vapor deposited W0_ films, there-
fore, may be considered as a random network, of these molecules linking
with each other through corner sharing oxygen bonds. The equilibrium
potential of WO thin film stored in a 3.6 N H2S04 solution was 0.4 VSH£.
The Pourbaix diagram for tungsten indicated that WC>3 is the only stable
specie for these conditions (pH = 0.5, E = 0.4 V^) . Specifically, W02
and W„0S are not considered to be stable. The conversion of W02 and W^

occur by a dissolution and precipitation mechanism. It was postulated

2+ 2+ .

that W0„ and W.CL dissolved in the acid as WC>2 ions. These WC>2 ions

then reacted with H?0 and formed crystalline WO^ and its hydrates

(WO .HO and WO . 2H 0) , which precipitated within the cell. Since there

are very large concentration of dissolution-prone species in the vapor

deposited films, rapid dissolution in the acid (^ 16 A/hr) was observed.

The precipitation of crystalline WO- and its hydrates on the

original films, particularly in the pinholes is postulated to cuase

delamination by introducing internal stresses which exceed interfacial

adhesion strength between the film and its substrate. Gas evolution

was observed and attributed to hydrogen gas formation during the

dissolution-precipitation processes. This may have contributed to the

force causing delamination. Interfacial delamination was not observed

on the WO films deposited on the graphite substrates which exhibited a

rough surface, probably becuase the adhesion strength was increased by

mechanical interlocking of the film and substrate through pores and

vallevs .

113

The corrosion process was enhanced when the WO- films was bleached

after coloration becuase of the dissolution rate of WO- and W?0- was

increased in the bleaching condition. The W03 films were reported to be

extremely stable in most organic solvents, but became soluble when the

solvent was acidified. The corrosion of WO thin films in H2S04/gly-

cerin (1:10) electrolyte or other acidified electrolyte (H?S04/methanol

or H„S0, /acetonitrile) supported the postulate that the pH rather than
2 4

water content defined whether corrosion would occur, since the order of
magnitude of the reduction of the corrosion rate was not consistent
with the reduction in the water content.

Knowing the causes of corrosion (the oxygen deficiency and the
pH of the electrolyte) many solutions to the corrosion problem can
be developed. Generally, there are two types of solutions, one
involved the modification of the WO films and the other the modi-
fication of the electrolyte. Both types of solutions were attempted
during the sutdy. Since the corrosion occurred because of oxygen
deficiency in the W0_ films, oxygen backfilling during WO^ deposition
was used to increase the oxygen content. Rutherford Backscattering
Spectroscopy data indicated that the 0/W was increased from 2.7
for films prepared without backfilling to 2.9 for films deposited at

_ n

a P = 1 x 10 Torr. The increase in the 0/W ratio was accompanied
by an increase in the W states relative to the W and W states as
shown by XPS data. Since the concentration of unstable species (W ,
W5+) were reduced, the dissolution rate was decreased. Dissolution

116

-3
was reduced to a negligible rate for films deposited at P - 1 x 10

Torr. Interfacial delamination was not observed for films prepared

at P > 1 x 10 Torr because the corrosion process was reduced.

2 ,.,4+ rr5+

However, the small concentration of subvalence states (.W , W )

still allowed some small scale corrosion process to occur and the
films exhibited partial crystallization in X-ray diffraction data
after long term storage. In addition, a slight roughened surface
was observed at high magnification (x 3,000). Generally, the oxygen
enriched films exhibited superior stability as compared to the
unmodified films.

The electrochromic properties of these oxygen enriched films
were altered too, however, the coloration speed was decreased
because the magnitude of current injected and the optical efficiency
were decreased. The current injection during coloration was found to
be controlled by the same mechanisms as the films without backfilling
(i.e., WO- resistivity and Helmholtz double layer). However, the
magnitude of the current passed was lowered by backfilling due to an
increased film resistivity. The increased resistivity and decreased
optical efficiency in the oxygen enriched films resulted in slower
coloration speeds and were explained by postulating an increased
density of inactive electron trapping site. This increased density
of inactive trapping sites limited the formation of active polaron
for optical absorption and DC conductivity. However, the possibility
that a decreased oscillator strength resulted in degraded optical
efficiency could not be ruled out. The porosity of the films was

117

increased by deposition at high background pressure. Due to increased
effective surface area, the film absorbed higher amounts of water after
deposition as shown by secondary ion mass spectroscopy . The bleaching
speed and self-erasure rates were increased since the rate of proton
removal were increased by increase in porosity and absorbed waters.
In summary, the compositional changes from OB (0/W ratio)
resulted in a degraded coloration speed and optical efficiency while
the structural change (porosity) led to an increase in bleaching
speed and self-erasure rate. Of these four property changes, only
increased bleaching speed is desirable from the viewpoint of device
fabrication.

In a second approach to improve device lifetime, LiCIO, /propy-
lene carbonate (Li/PC) electrolyte and solid hydrated MgF dielectric
films were used in the place of the acid electrolyte. Since the
switching speeds of these devices were slow, nitrogen and air back-
filling during deposition was used to modify the structure and
composition of WO and MgF? films. In liquid WOy Li/PC devices,
the porosity of WO films was increased by deposition at high partial
pressures of nitrogen and low deposition rate. The coloration and
bleaching speeds were slightly increased by the increased porosity,
possibly due to the increased surface area tor Li' injection during
coloration, and shortened diffusion path for Li' extraction during
bleaching. However, the increase in the speeds was not directly
proportional to the relative density change. It was possible that the

118

pores were not equally effective in facilitating ion injection and
extraction in different deposition conditions, e.g., high deposition
rate and high partial pressure of nitrogen as compared to low deposi-
tion rate and low partial pressure of nitrogen. The resistivity of WO^
thin films was high and was reported to be caused by the formation of
a depletion layer by the presence of trapping states for electrons at
the WO surface. Deposition of a MgF„ overlayer was expected to
decrease the resistivity by decreasing the thickness of this depletion
layer. The coloration speed of devices with a MgF2 overlayer was
slightly increased because the duration of the resistivity control
was shortened. The resistance of the transparent indium tin oxide
layer was found to limit the switching speed. Therefore, a highly
conductive ITO layer will be desirable for a fast switching EC
device. In general, the stability of WO films in Li/PC electrolyte
is excellent. The switching speed could be increased by a porous
WO film, a MgF overlayer and more conductive ITO.

In the device using solid MgF„ films as an electrolyte (i.e.,
W0-/MgF?/Au) , air-backfilling during deposition of MgF2 film was
reported to increase the water content in these films. The
switching speeds were increased becuase of the increased proton
mobility in the MgF,, films. Air-backfilling during W03 evapora-
tion appeared to be undesirable, since the porosity and rough
surface prohibited a uniform sandwich layered structure to be formed
and led to local dielectric breakdown. The operational area of the
device was limited. A different degradation mechanism was postulated

in this type of device, i.e., gas evolution during the operation
resulted in loss of contact between Au and MgF films. Exhaustion
of water in the MgF layer may also have occurred. The cycling
lifetime was limited to ^100 cycles.

Future Development

1. Becuase of the nature of the vapor deposited W0_ films (amorphous,
porous and hygroscopic), the films are extremely unstable. In
addition to corrosion in the acid, the films are also susceptible
to the attack by the moisture in the air (see Appendix B) .

Future studies on crystalline WO. films with high ionic conduc-

88
tivity such as crystalline hexagonal W0_ thin films may

provide a more reliable device.

2. Recently, the technology of devices using Li/PC has experienced

19—25
vigorous development. The reproted long lifetime ( > 2 x

10 cycles and unlimited shelf lifetime), and fast switching

speed (C/B times <300 ms) look promising. However, the

switching speed could be further increased by a reduced 0/W

ratio, by a MgF„ overlayer and by increased porosity. A

switching time in the millisecond range (comparable with LCD)

is expected.

3. The ultimate goal is the development of a reliable solid state
device. Devices using solid state protonic conductors (HUP
Cr90. films, NAFION,""4 etc.) have reported long lifetimes and
fast switching speeds. The stability of such devices, however,

120

is questionable since a fast switching speed would require a
high water content in the electrolyte layer, and hence a

corrosive environment may be created. Future studies with Li-

89
electrolytes may provide more stable devices.

APPENDIX A

PRELIMINARY STUDIES OF CATHODOCHROMISM IN TUNGSTEN

TRIOXIDE FILMS AND THE EFFECT OF AIR BACKFILLING

The coloration in vapor deposited WO films can be induced not

only by the electrochemical method described earlier but also by

50-59
electron beam irradiation (cathodochromism) "" or by high energy

, , 70
photon (e.g., ultra-violet or X-ray) irradiation (pnotocnromism) .

The area that is bombarded by the electron beam is colored deep
blue and has a high contrast with the remaining unbombarded area.
The coloration is normally limited by the size of the electron beam.
The application of this phenomenon is limited by one's imagination.
However, the mechanism for cathodochromism is unclear at this time
due to the scarcity of studies. Lin and Lichtman studied catho-
dochromism in WO powder and oxidized W foil and suggested that the
coloration was caused by the preferentially removal of oxygen from
the bombarded area by electron-beam- induced out-diffusion or by
electron-beam- induced-desorpt ion . They attributed the blue colora-
tion to deficiency of oxygen atoms because polycrystalline WO^ powder
is yellow in color, while WO q powder is deep blue. Morita
et al."3 reported a similar study on vapor deposited WO^ r ilms
and also observed the ejection of oxygen atoms from the bombarded
area, and the blue coloration. Again they attributed the
coloration to deficiency of the oxygen in the films. However, it
is known that the vapor deposited WO films are amorphous and

121

TZT

already oxygen deficient, i.e. 0/W =2.7. But the films are still
transparent. Evidently, a different mechanism is needed to explain
the cathodochromism in amorphous WO thin films. In the following,
the results of a preliminary study of cathodochromism in WO thin
films are presented and discussed.

Results and Discussion
The WO,, films were prepared by thermal evaporation of pure WO
polycrystalline powder (Cerac Inc., 99.9%) from a resistively
heated source onto indium tin oxide coated glass (ITO glass) at
room temperature. Before evaporation, the chamber was evacuated
to a base pressure of 6 x 10~ Torr. The films were deposited at
a residual pressure (during evaporation) of 1 x 10 Torr, or at
a partial pressure of backfilled air of 5 x 10 Torr. The deposi-
tion rate was kept constant at 5 A/s, while the thickness was

either 4,000 or 8,000 A. The deposited films were placed in a

-6
demountable electron beam writing system evacuated to 2 x 10

Torr by a diffusion pump with liquid nitrogen trapping backed

by a mechanical pump. Electron dose was varied by changing the

irradiation time. The electron beam was rastered over an area

2
of "*■>! cm". The sensitivity of the film colored by electron

irradiation was characterized in terms of absorption of light at

at wavelength of 6,238 A (He-Ne laser).

Results are shown in Figure A.l and A. 2 for 2.0 and 3.8 KV beam.

respectively. Generally, there is no significant difference

TIT

Electron Dose (C/cm )

Figure A.l. Coloration intensity versus electron dose for WO films
prepared at residual pressure of 1.x 10 Torr and
partial pressure of air at 5 :■: 10 Torr. Solid circles
refer to films of 8 KA while open circles to those of
4 KA. Electron accelerating voltage, 'V ' was 2 KV .

124

7 0

60

V = 3.8 KV
ace

-4
p . = 5 x 10 Torr
air

50

4 0

30

20

P = 1 x 10 Torr
res .

L0

10

10

Electron Dose (C/cm )

Figure A. 2. The coloration intensity versus electron dose for WO

films prepared at two extreme conditions and thickness
as described at Figure A.l. V = 3.3 KV.

125

between films of the two thickness. However, increasing the accele-
rating voltage and air backfilling did increase the optical contrast
in WO films. For the same dose ( ^1.8 x 10 C) , the contrast
increased from 15% absorption for films prepared without back-
filling to ^0% for films prepared at 5 x 10 Torr. For films
prepared with air backfilling, the optical contrast was increased
even further to ^-65% when the accelerating voltage was approximately
doubled, while that in the films prepared without backfilling
remained nearly constant.

Baba et al . ~ have reported that the absorption spectra from
electron beam induced coloration is similar to that of electrochromic
coloration. The energy of the absorption peak was also identical,
and the electron beam colored area can be electrochemically erased
to a completely bleached state, indicating that the mechanism
responsible for coloration may be same.

As shown in Chapter III, the air backfilling is known to
increase the water content (or H content) in the WO films. The
result that the optical contrast was increased in the W0_ films
with air backfilling (higher H content) can be explained by the
active-inactive-cation-sites disscussed in Chapter II. Since
there were more protons in the inactive site by backfilling
the protons that were redistributed into active site upon electron
beam irradiation and contribute to coloration would increase. When
the electron accelerating voltage was increased, there was more

energy released during electron decelerating resulting in higher
probability for protons to transfer into active sites and hence a
denser coloration. However, the accelerating voltage effect is
contingent upon the air backfilling effect as evident from the fact
that the contrast remained the same when the accelerating voltage

was almost doubled for the films prepared without backfilling.

79
Deneuville, et al. reported that WO., films prepared without

backfilling contain less H when compared with the films prepared
at 5 x 10~ Torr (H/W =0.6 and 1.2. respectively). Therefore, the
coloration is solely controlled by the concentration of the protons
in the films. These data strongly support the active-inactive-
cation-sites model.

Increasing the thickness from 4,000 A to 8,000 A did not
increase the contrast. It is possible the thickness range in this
study was much higher than the electron penetration depth ( V300 A)
at the operating voltage used (2.0 and 3.8 KeV) . The optical contrast
may be enhanced when a thinner film was used so the electrons
scattered from the substrate can contribute to further coloration.

Judging from the features of the absorption spectra and the
dependence of optical contrast on the proton concentration, the
optical absorption mechanism may still be associated with the
the polaron hopping mechanism as proposed for electrochromism. The
electrons injected during the irradiation were expected to localize
around tungsten atoms and eventually form the polaronic band. The

127

preferential removal of oxygen may still occur but is not expected
to be the cause of coloration since the electron-beam- induced colo-
ration can be removed by ensuing electrochemical bleaching. The
fact that the electron beam colored area can be electrochemicaily
erased is also strong evidence supporting the polaron hopping
absorption mechanism in cathodochromism.

In conclusion, the mechanism that is responsible for cathodo-
chromism in amorphous WO thin films appears to be related more
to polaron hopping absorption as opposed to the oxygen extraction
model. The observed air backfilling and accelerating voltage
effects support the active-inactive-cation-sites model proposed by
Deneuville et al. , and indicate that the position of cations in
WO films play an important role in both cathodochromism and
electrochromism.

APPENDIX B
CORROSION OF TUNGSTEN TRIOXIDE FILMS DURING
STORAGE IN MOIST AIR

The WO chin films are hygroscopic due to the high porosity in
these films. The storage of these films before device fabrication

Q O

constitutes a major problem. Hitchman reported that evaporated WO,
films were crystallized after standing in air for 12 months. In
our study, the conversion processes were expedited by moisture
saturation of the atmosphere.

Films of WO prepared with and without oxygen backfilling
were stored in a sealed box containing a cup of water for moisture
saturation. After three months, the box was opened, the surface
topography was examined using a scanning electron microscope.
Crystallite-like particles were distributed on the films. There
was no significant difference between the films prepared with
and without oxygen backfilling (Figure B.l). However, the
surface damaged by moist air corrosion were different from those
corroded by acid. The corrosion mechanisms in these conditions
are expected to be different. In the air storage, the water
condensed on the WO, films surface, especially in the pinholes,
was the corrosive agent. As discussed earlier, the W0„ is readily

123

129

m

f- * . * .**~^ M • ■ ~ J?, i i' - J- ' **«• ~4 ■•■1*1 '^/V-*'. ••

Figure B.l. Scanning electron micrographs of corroded surrace of WO
films after storage in moisture saturated air for three
months; _r

(A) P = 1 x 10 Torr, x 500,
(3) PneS*= 1 x 10 Torr, x 500.

130

dissolved in neutral water, and W0_ and W00_ exhibit even higher
solubility:

WO, + H„0 = 2H + WO.

3 2 4

WO- + 2Ho0 = WO, +■ 4H ++ 2e~

2 2 4

W„0„ + 3H„0 = 2WO , 4- 6H+ + 2e"
2d 2 4

(3.1)
(B.2)
(3.3)

The oxygen dissolved in water could also generate hydroxyl ions by
HO + 1/20- + 2e~ — » 20H" (3.4)

and render the condensed water more basic. This would lead to
higher dissolution rates of tungstic oxides. When the saturation
limit was reached, the crystalline WO- and its hydrates started to
precipitate as shown in Figure B.l. The population of the crystallites
on the surface was initially small and limited to the pinholes.
Gradually they spread out and covered the surface. The mechanical

strength of these films was therefore weakened. The proton

,35 , 65 , . , ,
diffusion coefficient was also reported to be Lowered due to

crystallization, the switching speed was therefore slowered. Even-
tually the films were useless for device fabrication.

In summary, because of the nature of the porous, hygroscopic and
amorphous vapor deposited WO films, they are extremely unstable.
In addition to corrosion in the acid electrolyte, the films are
also susceptible to the attack by air, or more specifically, by the
moisture in air. Therefore, a strictly controlled atmosphere is

131

needed for the fabrication of electrochromic display devices where
the moisture incorporated in the device (W0_ films, electrolyte) will
be reduced to a minimum.

BIBLIOGRAPHY

1. Schoot, C. J., J. Ponjee, H. T. Van Dam, R. A. Van Doom

and P. T. Bolwijn, "New Electrochromic Memory Display," Appl.
Phys. Lett. , 23, 64 (1973) .

2. Van Uitert, L. G. , G. J. Zydzik, S. Singh and I. Cambibel,
"Anthraquinoide Red Display Cells," Appl. Phys. Lett., 3_6
109 (1980).

3. Zaromb, S., "Theory and Design Principles of the Reversible
Electroplating Light Modulator," J. Electrochem. Soc . , 109 ,
903, 912 (1962).

4. Nicholsen, M. M. , and F. A. Pizzarello, "Charge Transport in
Oxidation Product of Lutetium Diphthalocyanine, " J. Electro-
chem. Soc, 126, 1490 (1979).

5. Kaufman, F. 3., A. H. Schroeder, E. M. Engler and V. V. Patel,
"Polymer-Modified Electrodes: A New Class of Electrochromic
Materials," Appl. Phys. Lett., 36, 422 (1980).

6. Deneuville, A., and P. Gerard, "Influence of Substoichiometry ,

Hydrogen Content and Crystallinity on the Optical and Electrical

Properties of H WO Thin Films," J. Electron. Mat., 7, 559 (1978).
x y

7. Golton, R. J., A. M. Guzman and J. W. Rabalais, "Photochromism
and Electrochromism in Amorphous Transition Metal Oxide Films,"
Accounts of Chemical Research, 170 (1978) .

8. Deb. S. K. , "Optical and Photoelectric Properties and Color Centers
in Thin Films of Tungsten Oxide," Phil. Mag., 2_7, 801 (1973).

9. Faughnan, 3. W. , R. S. Crandall and P. M. Heyman , "Electro-
chromism in WO Amorphous Films," RCA Review, 36, 177 (1975).

10. Schirmer, 0. F. , V. Wittwer, G. Baur and G. 3randt, "Dependence
of WO Electrochromic Absorption on Crystallinity." J.
Electrochem. Soc, 124, 749 (1977).

11. Wittwer, V., 0. F. Schirmer and P. Scholtter, "Disorder

Dependence and Optical Detection of the Anderson Transition

in Amorphous H W0_ 3ronzes," Solid State Commun. ,
x 3

977 (1978)

132

133

12. Schirmer, 0. F.,and E. Salje, "The W + Polaron in Crystalline
Low Temperature WO.: ESR and Optical Absorption," Solid State
Commun. , 33, 333 (1980) .

13. Hersh, H. N. , W. E. Kramer and J. H. McGee, "Mechanism of
Electrochromism in WO,," Appl. Phys. Lett., 2_7, 645 (1975).

14. Knowles, T. J., "Optical Regeneration of Aged WO- Electro-
chromic Cells," Appl. Phys. Lett., 31, 817 (1977;.

15. Mohapatra, S. K. , "Electrochromism in Li WO.," J. Electrochem.
Soc. , 125, 284 (1978). X

16. Mohapatra, S. K., and S. Wagner, "EMF Measurements on Electro-
chromic Amorphous Li WO- Films," J. Electrochem. Soc., 125 ,
1603 (1978). X

17. Ho, C. , I. D. Raistrick and R. A. Huggins, "Application of A-C
Techniques to the Study of Lithium Diffusion in Tungsten
Trioxide Thin Films," J. Electrochem. Soc, 127 , 343 (1980).

18. Green, M. , W. C. Smith and J. A. Weiner, "A Thin Film Electro-
chromic Display Based on the Tungsten Bronzes," Thin Solid
Films, 38, 89 (1976) .

19. Miyoshi, T., and K. Iwasa, "Electrochromic Displays for Watches,"
SID 80 Digest, 126 (1980) .

20. Kaneko, N., J. Tabala and T. Miyoshi, "Electrochromic Display
Watch Design," SID 81 Digest, 74 (1981).

21. Morita, H., and H. Washida, "A Modified Model for the Mechanism
of Symmetric Electrochromic Cells," Jpn. J. Appl. Phys., 19 ,
L228' (1980) .

22. Morita, H. , "Electrochromic Memory Degradation in WO ,-LiC10, /PC
Cells," Jpn. J. Appl. Phys., 21, 655 (1982).

23. Yamanaka, K.,"New Counter-Electrodes Made of Iron Compound-
Graphite Mixtures for WO /Liquid Electrolyte Electrochromic
Devices," Jpn. J. Appl. Phys., 21, 926 (1982).

24. Matsuhiro, K. , and Y. Masuda, "Transmissive Electrochromic Display-
Using A Porous Crvstalline WO, Counter Electrode," Proceedings
of SID, 21, 101 (1980) .

25. Hamada, H. , H. Kawagaki, K. Yano, H. Take, M. Matsuura, Y. Inami
and T. Wada, "A WCL-ECD with MnCL-Counter Electrode," Sharp
Tech. Report, 22, 53 (1983), in Japanese.

26. Green, M. , and K. S. Kang, "Solid State Electrochromic Cells:
the M- Alumina System," Thin Solid Films, 40, L19 (1977).

27. Dautremont-Smith, W. C, M. Green and K. S. Kang, "Optical and
Electrical Properties of Thin Films of WO Electrochromically
Coloured," Electrochimica Acta, _22> 7^1 (1977)

28. Barna, G. G. , "Material and Device Properties of a Solid State
Electrochromica Display," J. Electron. Mat., 8_, 153 (1979).

29. Green, M. , and D. Richman, "A Solid State Electrochromic Cell-
The RbAg,I System; Theory and Fabrication," Thin Solid Films,
24, S45 (1974).

30. Haranahalli, A. R. , and P.H. Holloway, "The Influence of Metal
Overlayers on Electrochromic Behavior of W0_ Films," J.
Electron. Mats., 10, 141 (1980).

31. McGee, J. H. , W. E. Kramer and H. N. Hersh, "A Tungsten Bronze
Electrochemical Character Display," SID 75 Digest, 50 (1975).

32. Randin, J. P., "Chemical and Electrochemical Stability of W03
Electrochromic Films in Liquid Electrolytes," J. Electron.
Mat., ]_, 47 (1978).

33. Reichman, B. , and A. J. Bard, "The Electrochromic Process at W03
Electrodes Prepared by Vacuum Evaporation and Anodic Oxidation
of W," J. Electrochem. Soc, 126., 583 (1979).

34. Arnoldussen, T. C, "A Model for Electrochromic Tungstic Oxide
Microstructure and Degredation," J. Electrochem. Soc, 128,
117 (1981).

35. Faughnan, B. W. , and R. S. Crandall, "Electrochromic Displays
Based on WO.," in Display Device (Topics in Appl. Phys., Vol. 40)
ed. by J. Pankov (Springer Verlag, New York, 1980), 181-211.

36. Di Paola, A., F. Di Quarto and C. Sunseri, "Electrochromism in
Anodically Formed Tungsten Oxide Films," J. Electrochem. Soc,
125, 1344' (1978).

135

37. Berkowitz , J., and M. G. Inghram, "Polymeric Gaseous Species
in the Sublimation of Tungsten Trioxide," J. Chem. Phys.,
1]_, 35 (1957).

38. Nazarenko,V. A., E. N. Poluektova and G. G. Shitareva, "Spec-
trophotometric Determination of the Formation Constants of
Monomeric Hydroxo-Complexes of Tungsten (VI)," Russian J.
Inorg. Chem'., 22_, 551 (1977).

39. Zeller, H. R. ,and H. U. Beyeler, "Electrochromism and Local
Order in Amorphous WO " Appl. Phys., 13, 231 (1977).

40. Beni, G.,and J. L. Shay, 'Electrochromism of Heat-Treated
Anodic Iridium Oxide Films in Acidic, Neutral and Alkaline
Solutions,' Appl. Phys. Lett., 33. 567 (1978).

41. Shay, J. L. , G. Beni and L. M. Schiavone, "Electrochromism
of Anodic Iridium Oxide Films on Transparent Substrates,"
Appl. Phys. Lett., 33, 952 (1978).

42. Gottesfeld, S . , and J. D. E. Mclntyre, "Electrochromism in
Anodic Iridium Oxide Films," J. Electrochem. Soc . , 126 ,
742 (1979).

43. Rice, C. E. , "A Comparison of the Behaviors of Tungsten
Trioxide and Anodic Iridium Oxide Film Electrochromics in

a Non-Aqu dus Acidic Medium," Appl. Phys. Lett., 35 , 563 (1979)

44. Dautremont-Smith, W. C, G. Beni, L. M. Schiavone and J. L.
Shay, "Solid State Electrochromic Cell with Iridium Oxide
Film Electrodes," Appl. Phys. Lett., 15, 565 (1979).

45. Deb, S. K. , "Physical Properties of a Transition Metal Oxide;
Optical and Photoelectric Properties of Single Crystal and
Thin Film Molybdenum Trioxide," Proc . Roy. Soc. A., 304 ,

211 (1968).

46. Deb, S. K., and J. A. Chopoorian, "Optical Properties and Color
Center Formation in Thin Films of Molybdenum Trioxide," J.
Appl. Phys., 27_. 4818 (1966).

47. Arnoldussen, T. C, "Electrochromism and Photochromism in MoO
Films," J. Electrochem. Soc, 123 , 527 (1976).

48. Gourilyvk, A. E.,and F. A. Chudnowder. "Electrochromism in vy^-
Films," Sov. Tech. Phys. Lett., 3, 69 (1977).

136

49.

50.

Mclntyre, J. D. E. , W. F. Peck, Jr. and G. P. Schwartz, "Electro-
chromism in Hydrous Nickel Oxide Films," Presented at 21st Elec-
tronic Materials Conf. (1979)..

Lin, T. T.,and D. Lichtman, "The Effects of Electron Bombardment
on the Surface Composition of
Appl. Phys., 50, 1298 (1979).

on the Surface Composition of WO , Oxidized W and Ta Foils," J

51.

Morita, H. , T. Miura and H. Washida, "Coloration of Tungsten
Oxide Films Induced by Ion or Electron Irradiation," Jpn. J.
Appl. Phys., 20, L233 (1981).

52. Baba, M. , and T. Ikeda, "A New Inorganic Electron Resist Using
Amorphous WO Films," Jpn. J. Appl. Phys., 20, L149 (1981).

53. Chu, W. K., J. W. Mayer and M.-A. Nicolet, Backscattering
Spectrometry (Academic Press, New York, 1978).

54. Werner, H. W. , "Introduction to Secondary Ion Mass Spectrometry
(SIMS)," in Electron and Ion Spectroscopy of Solids (Plenum,
New York, 1978), 324-441.

55. Hercules, S. H. , and D. M. Hercules, "Surface Characterization
by Electron Spectroscopy for Chemical Analysis (ESCA)," in
Characterization of Solid Surface (Plenum, New York, 1974), 307-
335.

56. Grandall, R. S., and B. W. Faughnan, "Measurement of the Diffusion
Coefficient of Electrons in WO- Films," Appl. Phys. Lett., 26_,

120 (1975).

57. ASTM Powder Diffraction File, Card 24-747, for W03-

58. ASTM Powder Diffraction File, Card 18-1418, for W03.H20.

59. ASTM Powder Diffraction File, Card 18-1419, for W03.2H20.

60. Bockris, J. 0. M. , and A. K. N. Reddy, Modern Electrochemistry
(Plenum/Roselton, New York, 1973).

61. Pourbaix, M., Atlas of Electrochemical Equilibria in Aqueous
Solutions (Pergamon Press, Oxford, 1966).

137

62. Di Paola, A. , E. Di Quarto and C. Sunseri, "Anodic Oxide Films

on Tungsten I: The Influence of Anodizing Parameters on

Charging Curves and Film Composition,'' Corrosion Science, 20,
1067 (1980).

63. Johnson, J. W.,and G. L. Wu , "The Anodic Dissolution of Tungsten,"
J. Electrochem. Soc, 118, 1909 (1971).

64. Gerard, P., A. Deneuville, G. Hollinger and T. M. Due, "Color
in Tungsten Oxide Thin Films," J. Appl. Phys . , 48_, 4252 (1977).

65. Dautremont-Smith, W. C, "Transition Metal Oxide Electrochromic
Materials and Displays: A Review; Part I," Displays, 4,

3 (1982).

66. Scully, J. C, The Fundamentals of Corrosion, 2nd Edition, Inter-
national Series of Materials Science and Technology, Vol. 17,
(Pergamon Press, Oxford, 1975).

67. Handbook of Chemistry and Physics, 61st Edition, The Chemical
Rubber Co., Cleveland (1980)'.

68. Dickens, P. G. , S. J. Hobble and R. H. Jarman, "Hydrogen
Insertion Compounds of Transition Metal Oxides," J. Electron.
Mats. , 10, 999 (1981) .

69. Gazzinelli, R. ,and 0. F. Schirmer, "Light-Induced V ESR in W03,':
J. Phys. C: Solid State Phys., 10, L145 (1977).

70. Hollinger, G-, T. M. Due and A. Deneuville, "Charge Transfer in
Amorphous Colored WO Films Observed by X-Ray Photoelectron
Spectroscopy," Phys. Rev. Lett., 37_> L56Z* (1976).

71. Mott, N. F.,and E. A. Davis, Electronic Processes in Non-
Crystalline Materials, (Oxford University Press, Oxford, 1971).

72. Austin, I. G.,and N. F. Mott, "Polarons in Crystalline and
Non-Crystalline Materials," Adv. Phys., 18_, 41 (1969).

73. Crandall, R. S . , and B. W. Faughnan, "Electronic Transport in

Amorphous H WO '' Phys. Rev. Lett., 39, 232 (1977).
r x 3 ' —

74. Benci, S. M. Manfredi and G. C. Salviati, 'Mechanisms of Electircal

>3

33, 107 (1980).

138

75. Dallacasa,V. , M. Manfredi and G. SehiachiT '"Tfie.-Colora.tion
Mechanisms of Amorphous W0_ Thin Films," Thin Solid Film,
191, 1 (1982).

76. Gritsenko, V. A., Y. 0. Roisin, L. E. Semenchuk and N. L.
Schwarz, "Semiconductor-Metal Transition Due to Deep Centers
Interaction in Tungsten Oxide Films," Solid State Commun. ,
38, 351 (1981).

77. Dexter, D. L., "Theory of the optical Properties of Imperfections
in Nonmetals," Solid State Physics, Vol.6, ed. by F. Seitz, and D,
T, Turnball (Academic Press, New York, 1958), 353.

78. Yoshimura, T. , M. Watanabe, Y. Koike, K. Kiyota and M. Tanaka,
"Enhancement in Oscillator Strength of Color Centers in
Electrochromic Thin Films Deposited From W0„ Powder," J. Appl.
Phys., 53, 7314 (1982).

79. Deneuville, A., P. Gerard and R. Billat, "Principles and

Operation of an All Solid State Electrochromic Display Based

on a-WO ," Thin Solid Films, 70, 203 (1980).
3 —

80. Gerard, P., A. Deneuville and R. Courths, "Characterization of
a-WO Thin Films Before and After Coloration," Thin Solid
Films, 71, 221 (1980).

81. Hill, R. H. , "Hopping Conduction in Amorphous Solids," Phil.
Mag, 24, 1307 (1971).

82. Hitchman, M. L. , "Analysis of Equilibrium Potential of Hydrogen
Tungsten Bronzes," J. Electro-Ana. Chem. , 8_5, 135 (1977)

83. Crandall, R. S., P. J. Wojtowicz and B. W. Faughnan, "Theory
and Measurement of Change in Chemical Potential of Hydrogen

in Amorphous H WO as a Function of the Stoichiometric Parameter
x," Solid State Commun., 18, 1409 (1976)

84. Chopra, K. L. , Thin Film Phenomena (Krieger Pub. Co., New York,
1979), 137-245.

85. Yoshimura, T. , M. Watanabe, Y. Koike, K. Kiyota and M. Tanaka,

"Effect of Surface States of WO- on the Operation Characteristics
of Thin Film Electrochromic Devices," Thin Solid Films, 101,
141 (1983) .

139

86. Yoshimura, T. , M. Watanabe, K. Kiyota and M. Tanaka, "Electrolysis
in Electrochromic Devices Consisting of WO- and MgF„ Thin Films,"
Jpn. J. Appl. Phys., 21, 128 (1981).

87. "Transmissive All Solid State Electrochromic Display Devices-
New Product Announcement," Nikkei Electronics, p. 87, August (1981)

88. Cheng, K. H. , A. J. Johnson, M. S. Wittingham, "Incorporation

of Alkali Metal Cations in Tungsten Oxide Compounds," Ext. Abstr.-
Electrochem. Soc, 80-601, 1499 (1980).

89. Yoshimura, T. M. Watanabe, Y. Koike, K. Kiyota and M. Tanaka,
"Electrochromism in a Thin-Film Device Using Li?W0, as an
Li-Electrolyte," Jpn. J. Appl. Phys., 22, 152 U983) .

90. Molnar, B. J., "Elechromic Behavior of Rhenium Oxide, Tungsten
Oxide and Metal Tungstates and Thermochromic Properties of
Rhenium Oxide," unpublished manuscript (1980).

BIOGRAPHICAL SKETCH

Sey-Shing Sun was born in Taipei, Republic of China, on July 9,
1955. He is the son of Yu-Ping and Ching-Der Sun. In June 1977,
he received his B.S. in materials science and engineering from
National Tsing Hua University, Hsinchu, Republic of China. After
graduation, he served in the Chinese Army ROTC as a second Lieutenant
for two years. In January 1980, he enrolled in the Graduate School
of the University of Florida and is now completing the requirements
for the degree of Doctor of Philosophy in materials science and
engineering.

His primary interest is in the area of thin film phenomena.
As an undergraduate, he was involved in a senior research project
studying the microstructure and properties of electro-deposited
nickel films on copper substrates for the Ti-Cu-Ni-Au metallization
system. After joining the Univeristy of Florida, he conducted
research in the electrochromism of vapor deposited W0_ thin films;
hence he has a strong background in the physical and chemical
properties of metal and oxide films, thin film deposition and vacuum
technology. In addition, he is familiar with surface analytical
techniques, including RBS, SEM, SIMS, AES and ESCA, particularly
with their application to thin film analysis.

140

He is a member of the American Vacuum Society, the American
Electrochemical Society and the International Society for Hybrid
Microelectronics .

Sey-Shing Sun was married to Mei-Hwei Wu in Gainesville, Florida,
on June 14, 1980.

I certify that I have read this study and that in my opinion
it conforms to acceptable standards of scholarly presentation and
is fully adequate, in scope and quality, as a dissertation for the
degree of Doctor of Philosophy.

Paul H. Holloway, Chairmin
Professor of Materials Science and
Engineering

I certify that I have read this study and that in my opinion
it conforms to acceptable standards of scholarly presentation and
is fully adequate, in scope and quality, as a dissertation for the
degree of Doctor of Philosophy.

Rolf E. Hummel

Professor of Materials Science and

Engineering

I certify that I have read this study and that in my opinion
it conforms to acceptable standards of scholarly presentation and
is fully adequate, in scope and quality, as a dissertation for the
degree of Doctor of Philosophy.

John R. Ambrose

Associate Professor of Materials

Science and Engineering

I certify that I have read this study and that in my opinion
it conforms to acceptable standards of scholarly presentation and
is fully adequate, in scope and quality, as a dissertation for the
degree of Doctor of Philosophy.

Christopher D. Batich

Associate Professor of Materials

Science and Engineering

I certify that I have read this study and that in my opinion
it conforms to acceptable standards of scholarly presentation and
is fully adequate, in scope and quality, as a dissertation for the
degree of Doctor of Philosophy.

Sheng S. Li

Professor of Electrical Engineering

This dissertation was submitted to the Graduate Faculty of the
College of Engineering and to the Graduate School, and was
accepted as partial fulfillment of the requirements for the
degree of Doctor of Philosophy -\

/ / 'ft i_ / 1 i'

December, 1983 L- U^ L y ' -

Dean, College of Engineering

Dean for Graduate Studies and
Research

UNIVERSITY OF FLORIDA

3 1262 08553 2108