Você está na página 1de 238

Organic Vapor Phase Deposition and

Vapor Jet Printing for


Electronic and Optoelectronic Device Applications


Max Shtein


A DISSERTATION
PRESENTED TO THE FACULTY
OF PRINCETON UNIVERSITY
IN CANDIDACY FOR THE DEGREE
OF DOCTOR OF PHILOSOPHY

RECOMMENDED FOR ACCEPTANCE
BY THE DEPARTMENT OF
CHEMICAL ENGINEERING



NOVEMBER 2004

UMI Number: 3143423













Copyright 2004 by
Shtein, Max

All rights reserved.









________________________________________________________

UMI Microform 3143423
Copyright 2004 ProQuest Information and Learning Company.
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
____________________________________________________________


ProQuest Information and Learning Company
300 North Zeeb Road
PO Box 1346
Ann Arbor, MI 48106-1346




Copyright 2004 by Max Shtein.
All rights reserved


ii








To my Mother

iii
Abstract
Weak van der Waals bonding in molecular organic semiconductors allows depositing
them without lattice matching on a variety of substrates, (e.g. glass, steel foil, and
plastic), for low-cost, large-area device applications. As the device performance
improves, lowering fabrication costs becomes important. Organic Vapor Phase
Deposition (OVPD) and Organic Vapor Jet Printing (OVJP) may accomplish this, while
also presenting scientifically interesting mechanisms of thin-film growth.
In OVPD, a hot inert carrier gas picks up molecular organic vapor and transports
it into a hot-wall chamber, where the vapor selectively physisorbs onto a cooled
substrate. The film deposition rate, uniformity, composition and morphology are
controlled through the source and substrate temperatures, carrier gas flow rate, the source
cell and the deposition chamber pressures. The composition and morphology of the
deposited films bear directly on the electrical and optical device performance. Theory,
simulation, and experiments are used to understand the mechanisms governing OVPD
and demonstrate the method's capabilities.
Applications like full-color displays or transistor circuits require lateral patterning
of the active organic thin films. Because organic semiconductors are typically
incompatible with conventional patterning methods (e.g. photolithography), alternative
techniques are employed. In-situ patterning using shadow-masks is studied. For OVPD,
this involves Monte-Carlo modeling of molecular transport in confined geometries,
where the apparatus dimensions are on the order of the molecular mean free path.
Optimum operating conditions (e.g. pressure, mask-substrate separation) and mask
aperture geometry are suggested and verified by patterning experiments.

iv
Using the knowledge thus gained, OVJP is developed. Here, the light carrier gas
mixes with the heavier organic molecules and is rapidly ejected through a collimating
nozzle onto a proximally located cooled substrate. OVJP proceeds entirely in the gas
phase, eliminating the shortcomings associated with liquid-based ink-jet printing,
enabling high-resolution, rapid and direct printing of molecular organic semiconductors.
A theory of the flow is developed and verified by direct simulation Monte-Carlo models
and printing experiments, showing how pressure gradients, nozzle geometry, distance to
the substrate, and choice of carrier gas control the pattern resolution. A high performance
pentacene TFT is printed at ultra-high local deposition rate.


v
"Straight ahead of him, nobody can go very far..."
"The most important things are invisible to the eye."
- Antoine de Saint-Exupery: The Little Prince

Acknowledgements
To my supervisors, Stephen Forrest and Jay Benziger, I am grateful for stimulating
discussions, a balance of ripe opportunity, vigorous motivation and high standards in
research. Their guidance and more than a pinch of patience saw me through equipment
meltdowns and long detours to eventual progress. From my days at Berkeley I have
valued Enrique Iglesia's sound advice and encouragement. Committee members Sandra
Troian and Morton Kostin have suggested improvements to this thesis. Mark Thompson
has frequently offered insights into the chemistry that is at the heart of what we do, while
much of the "real-world" context was thanks to Julie Brown and Teddy Zhou of
Universal Display. Vicki Paulus and Robin Block helped navigate Princeton's
administrative muddy waters.
Paul Burrows, Vladimir Bulovic, and Marc Baldo were my first mentors in the lab
and in the basics of organic semiconductors. I am indebted to Herman Gossenberger for
teaching me the value of a thorough and careful experiment. In the trenches, I learned
alongside and from: Shashank Agashe, Shubo Datta, Brian D'Andrade, Jon Mapel, Vinod
Menon, Gautam Parthasarathy, Barry Rand, Pavel Studenkov, John Thomson,
Hongsheng Wang, Jian Wei, Fengiang Xia, Jiangeng Xue, and others. Many an evening I

vi
spent with Russ Holmes, Fan Yang, Changsoon Kim, Peter Peumans demystifying
remified data, alternating lively scientific discussions with vociferous commiserations
and assuaging words. My greatest education and fun undoubtedly came from the daily
(and nightly) discussions and collaborations with Peter and Changsoon. The benefits of
their insights, sincere critiques, humor and friendship I cannot emphasize enough.
Outside the lab, Changsoon's nuanced taste in music and Peter's companionship
on the bike made my life infinitely more enjoyable. Thanks to Tanya Nigam for her
hospitality, excellent cuisine and encouragement; to Paru Deshpande for a receptive ear
and excellent jinnantonix; to Audrey Lee for the bright colors and excellent cooking; to
Mohsen Moayer for the flagrant disregard of the speed limit; to Bart Pindor for the
perfectly timed mixture of the sublime and the absurd; to Upma Sharma for the sincere
conversation and the code to Lapidus. Along the way, other friends gave me care and
support: Gwen Barriac, Benjie Chen, Ben Fisher, Joelle Frechete, Katherine Gibson, Nate
Gleason, Phaedon (Steve) Koutsourelakis, Julie Lichty, Simon Mui, Joanna Paja, Thomas
Philip, Laura Stark, Lenka Tucek, Mike Weaver, and others. The enduring ameliorative
counsel and friendship of Helen Shvets through the years deserve more gratitude than I
can express here. While a few personalities on occasion expand to fill the available space,
none do so in a manner as beautiful and engaging as hers.
Finally, I thank my family for their support and faith in me and pride in my
accomplishments. In particular, I dedicate this thesis to my Mother, who was and remains
my greatest teacher.

vii
Table of Contents
Abstract .......................................................................................................................... iii
Acknowledgements......................................................................................................... v
Table of Contents.......................................................................................................... vii
List of Figures............................................................................................................... xii
List of Tables ............................................................................................................. xviii
Symbols........................................................................................................................ xix
Chapter 1: Introduction to organic semiconductors and devices................................. 1
1.1 Overview............................................................................................................... 1
1.2 Classification of Solids ......................................................................................... 2
1.3 Advantages and disadvantages of organic electronics.......................................... 4
1.4 Electronic structure and properties of molecular organic semiconductors........... 6
1.4.1 Intramolecular bonding.................................................................................. 7
1.4.2 Intermolecular interactions .......................................................................... 14
1.4.3 Electronic conduction in conjugated organic solids .................................... 17
1.4.4 Electronic excitations in organic solids ....................................................... 24
1.5 Summary............................................................................................................. 33
Chapter 2: Organic device fabrication technology...................................................... 35
2.1 Overview............................................................................................................. 35
2.2 Spin-on film deposition....................................................................................... 38
2.3 Laser induced thermal imaging........................................................................... 41

viii
2.4 Ink-jet printing .................................................................................................... 42
2.5 Vacuum thermal evaporation.............................................................................. 45
2.6 Organic vapor phase deposition.......................................................................... 52
2.7 Organic vapor jet printing................................................................................... 56
Chapter 3: Organic Vapor Phase Deposition............................................................... 58
3.1 Overview............................................................................................................. 58
3.2 OVPD Concept ................................................................................................... 58
3.3 Theory Evaporation ......................................................................................... 60
3.4 Vapor transport ................................................................................................... 64
3.5 Deposition........................................................................................................... 71
Chapter 4: Proof of Concept & Experimental Verification of Theory...................... 75
4.1 Overview............................................................................................................. 75
4.2 Experimental design............................................................................................ 75
4.3 Growth of Alq
3
films .......................................................................................... 77
4.4 Evaporation rate and decomposition temperature............................................... 80
4.5 Thermal decomposition of source materials....................................................... 83
4.6 Deposited film morphology and composition..................................................... 87
4.7 Growth of OLEDs using OVPD......................................................................... 89
4.8 Control of dopant concentration using temperature............................................ 91
4.9 Roll-to-roll deposition......................................................................................... 93
4.10 Summary........................................................................................................... 96
Chapter 5: Growth in confined geometries, application to patterning...................... 97
5.1 Overview............................................................................................................. 97

ix
5.2 The need for patterning OLED example.......................................................... 97
5.3 Ballistic transport in VTE and shadow-masking ................................................ 99
5.4 Shadow-masking and diffusive transport in OVPD.......................................... 100
5.5 Continuum-based analysis is inaccurate for Kn < 10 ....................................... 103
5.6 Modeling gas transport in confined geometries................................................ 105
5.7 Monte-Carlo simulation of OVPD through apertures....................................... 106
5.7.1 Simulation set-up ....................................................................................... 106
5.7.2 First results for patterning of Alq
3
thin films............................................. 108
5.7.3 Effects of chamber pressure on deposit shape ........................................... 110
5.7.4 Effects of mask thickness and separation .................................................. 111
5.7.5 Optimizing mask (aperture) shape............................................................. 113
5.8 Experimental set-up .......................................................................................... 117
5.9 Shadow-masking experiments results and discussion ................................... 120
5.10 Resolution limits and self-aligned contacts by hybrid VTE-OVPD............... 124
5.11 Summary......................................................................................................... 126
Chapter 6: Organic Vapor Jet Printing...................................................................... 128
6.1 Overview........................................................................................................... 128
6.2 OVJP Concept................................................................................................... 128
6.3 OVJP theory...................................................................................................... 130
6.4 Simulation of transitional flow regime ............................................................. 134
6.5 Experimental set-up .......................................................................................... 138
6.6 Direct printing of patterned molecular organic thin films ................................ 141
6.7 OVJP of polycrystalline pentacene films and TFTs ......................................... 148

x
6.8 Summary........................................................................................................... 153
Chapter 7: Growth of pentacene films and thin film transistors ............................. 154
7.1 Overview........................................................................................................... 154
7.2 TFT geometry and operation ............................................................................ 155
7.3 Growth of polycrystalline pentacene thin films from the vapor phase............. 160
7.3.1 Qualitative description of vacuum and vapor phase growth...................... 160
7.3.2 Theory of crystal growth on surfaces......................................................... 164
7.3.3 Role of background carrier gas in crystal growth...................................... 169
7.4 Growth mechanisms for pentacene................................................................... 171
7.5 Evidence for pentacene morphology influencing TFT performance................ 173
7.6 Relative effects of grain size and substrate treatment on device performance . 177
7.7 Effect of surface energy on device performance .............................................. 184
7.8 Further remarks................................................................................................. 186
7.9 Summary........................................................................................................... 187
Chapter 8: Summary, state-of-the-art, challenges and future directions................ 188
8.1 Organic semiconductors and devices................................................................ 188
8.2 Organic semiconductor processing technology ................................................ 189
8.3 Development and application of OVPD........................................................... 190
8.4 Development and application of OVJP............................................................. 192
8.5 Future directions in carrier-assisted deposition, novel devices ........................ 193
8.5.1 Carrier-assisted deposition of metals......................................................... 193
8.5.2 Growth of focal plane arrays using OVJP ................................................. 194
8.5.3 Alternative device form factors, fiber photovoltaics ................................. 196

xi
8.5.4 Large-volume, ultra-purification................................................................ 198
8.5.5 Non-cleanroom device processing............................................................. 199
8.6 Summary........................................................................................................... 200
Chapter 9: References .................................................................................................. 202

xii
List of Figures
Chapter 1:
1-1 Classification of materials 3
1-2 Examples of organic light emitting technology 5
1-3 Flexible organic electronics 6
1-4 Formation of molecular orbitals 7
1-5 Bond length vs. bond strength 9
1-6 MO energy diagram from methane 11
1-7 sp
2
hybridization in ethylene 12
1-8 Energy levels of linear acenes 13
1-9 Van der Waals interactions 15
1-10 Lennard-Jones potential shape 17
1-11 Classical semiconductor energy band diagram 18
1-12 STM image of a Cl-CuPc crystal 20
1-3 Charge traps in anthracene 23
1-14 Frank-Condon shift 25
1-15 Excitons in organic crystals 27
1-16 Energy level diagrams of a heterojunction PV cell and an OLED 28
1-17 Triplet and singlet excitons 29
1-18 CIE diagram of cyclometalated platinum compounds 33


xiii
Chapter 2:
2-1 Cost of flexible electronics and processing temperature 37
2-2 Spin-coating 39
2-3 Laser-induced thermal imaging 42
2-4 Thermal, piezoelectric, and acoustic ink-jet printing 43
2-5 Ink-jet printed full-color display 45
2-6 Schematic of vacuum thermal evaporation (VTE) 46
2-7 Schematic of a full-color display pixel structure 47
2-8 Shortcomings of and remedies for point-source VTE 49
2-9 Linear sources for VTE 50
2-10 Applied Films GMBH linear source VTE system 51
2-11 OVPD Concept 53
2-12 Commercial scale OVPD 55
2-13 OVJP Concept 57

Chapter 3:
3-1 Schematic of OVPD mechanism 59
3-2 Source evaporation cell 60
3-3 Source cell operating regimes 61
3-4 Vapor transport and mixing schematic 67
3-5 Navier-Stokes modeling of mass and heat transfer in OVPD 70
3-6 Velocity, temperature, and concentration profiles in OVPD 72
3-7 Alq
3
deposition rate vs. carrier gas flow (Aixtron OVPD) 74

xiv

Chapter 4:
4-1 OVPD system schematic and temperature profile 76
4-2 Photograph of OVPD system assembly 78
4-3 Alq
3
deposition rate vs. carrier gas flow rate (Princeton OVPD) 80
4-4 Thermogravimetric determination of vaporization enthalpy 81
4-5 DCM2 deposition rate vs. temperature 84
4-6 Digital scanning calorimetry of common OLED materials 86
4-7 AFM images of -NPD deposited on Si at different rates 88
4-8 Photoluminescence spectrum of an OVPD-grown Alq
3
film 89
4-9 Electrical characteristics of an OVPD-grown OLED 90
4-10 Photoluminescence and composition of DCM2-doped Alq
3
films 92
4-11 Clean and efficient growth in OVPD 94
4-12 Roll-to-roll deposition in OVPD 95

Chapter 5:
5-1 Schematic of a pixel in a full-color passive matrix display 98
5-2 Schematic of vacuum thermal evaporation & shadowmasking 100
5-3 Schematic of OVPD and shadowmasking 101
5-4 Geometry definition of the mask aperture and deposit 102
5-5 Pressure-dependence of the patter resolution using Monte-Carlo 109
5-6 Pixel shape factor vs. molecular mean free path 110
5-7 Deposit shape vs. mask thickness 111

xv
5-8 Deposit shape vs. mask-substrate separation 112
5-9 Deposit shape vs. mask aperture shape 114
5-10 Optimization of mask design for pattern sharpness 115
5-11 Schematic of mask & substrate assembly used in the experiment 117
5-12 Micrographs of the masks used in the experiment 119
5-13 Determination of deposit shape using light interference 120
5-14 Experimental deposit profile vs. mask-substrate separation 121
5-15 Comparison of simulated and experimental deposit profiles 122
5-16 Sub-10m feature experimental and simulated deposit profiles 123
5-17 SEM of OVPD-VTE self-aligned hybrid deposition 125

Chapter 6:
6-1 Organic vapor jet printing (OVJP) apparatus schematic 129
6-2 Nozzle eometry and the OVJP mechanism 131
6-3 Qualitative dependence of pattern size on process variables 134
6-4 Direct simulation Monte-Carlo (DSMC) modeling of OVJP 136
6-5 Deposit profiles vs. process parameters obtained by DSMC 137
6-6 Schematic of the OVJP experimental set-up 140
6-7 Micrographs of the nozzles used for OVJP experiments 141
6-8 Alq
3
dots printed on Si vs. nozzle-substrate separation (using N
2
) 142
6-9 Alq
3
dots printed on Si vs. nozzle-substrate separation (using He) 143
6-10 Comparison of pattern resolution achieved with N
2
vs. He 144
6-11 Printed pattern resolution vs. downstream pressure 146

xvi
6-12 Alq
3
bicyclist figure printed on Si showing 1000 dpi resolution 147
6-13 SEMs and x-ray diffraction pattern of pentacene grown by OVJP 150
6-14 Transfer characteristic of an OVJP-grown pentacene TFT 151

Chapter 7:
7-1 Schematic of a thin-film field effect transistor 156
7-2 Energy band diagram, operation, and transfer curve of a TFT 157
7-3 Top- and bottom-contact device geometry 159
7-4 Vacuum vs. vapor phase growth 162
7-5 Modes and mechanism of crystal growth on surfaces 163
7-6 Micrograph of OVPD-grown pentacene on SiO
2
and gold 167
7-7 Influence of carrier gas pressure on crystal growth 170
7-8 X-ray diffraction pattern of OVPD-grown pentacene on SiN
x
172
7-9 Transfer characteristic of an OVPD-grown pentacene TFT 174
7-10 Nucleation in the contact regions for bottom-contact TFTs 175
7-11 Hole mobility vs. deposition conditions for bottom-contact TFTs 176
7-12 Pentacene morphology & mobility on SiO
2
vs. OTS-treated SiO
2
178
7-13 VTE-grown pentacene on OTS / SiO
2
179
7-14 Transfer characteristic of a pentacene TFT on OTS / SiO
2
181
7-15 X-ray diffraction patterns of pentacene on OTS vs. SiO
2
182
7-16 Pentacene TFT performance vs. dielectric surface energy 185



xvii
Chapter 8:
8-1 Flow-through source configuration for a commercial-scale OVPD 191
8-2 Phosphorescent OLEDs grown by a commercial OVPD system 192
8-3 The Petzval condition and retinal implants 195
8-4 Structure of a flexible photovoltaic (solar) fiber 197
8-5 Possible fabrication sequence for the photovoltaic fiber 198

xviii
List of Tables
Table I: Physical property comparison between germanium and anthracene ....................21
Table II: Evaporation enthalpies of typical OLED materials............................................83
Table III: Effects of shadow mask geometry and process conditions on pattern shape..116
Table IV: Summary of electrical characteristics of pentacene TFTs...............................186

xix

Symbols
sticking coefficient; mask aperture wall angle
a nozzle radius
A
e
electron affinity; evaporation area
a
l
lattice constant
c vapor concentration
average dispersion distance
film or boundary layer thickness
d molecular collision diameter
G Gibbs free energy
H
des
enthalpy of desorption
H
vap
evaporation enthalpy
D
i
diffusivity of i
d
m
deposition chamber diameter; mixing dimeter
D
s
surface diffusion coefficient
S
des
entropy of desorption

0
permittivity of vacuum
E
c
, E
v
conduction / valence band energy
E
des
desorption activation energy
E
F
Fermi energy

xx
E
sd
surface diffusion activation energy
h Planck's constant, source-to-mask distance

dep
deposition efficiency
I
DS
drain-source current
I
p
ionizatin potential
j
org
flux of organic molecules to the substrate
k
B
Boltzmann constant
Kn Knudsen number
mean free path
l evaporation source width
L nozzle length
L
m
mixing length
viscosity
m
cg
, m
o
molecular mass of the carrier gas / organic
m
e
*, m
h
* electron / hole effective mass

eff
,
h
effective / hole mobility
M
org
molecular mass of organic species
M
s
spin angular momentum
m
s
spin quantum number
n
0
, p
0
electron / hole concentration
N
c
, N
v
conduction / valence band density of states
P
cell
evaporation source cell pressure
P
coll
probability of molecular collision

xxi
P
dep
deposition chamber pressure
P
dyn
dynamic pressure
P
eq
equilibrium vapor pressure
P
H
, P
L
upstream and downstream pressure
P
org
organic vapor pressure
P
org
s
organic vapor pressure at the substrate surface
Q
D
net deposition rate
r distance
r
cond
condensation rate
r
dep
deposition rate
r
evap
evaporation rate
r
out
outflow rate
r
D
distance traveled by molecule due to diffusion
Re Reynolds number
S subthreshold slope
s mask-substrate or nozzle-substrate separation
T temperature
t mask thickness, time
T
cell
evaporation cell temperature

D
characteristic diffusion time
T
s
surface temperature
U bulk flow velocity
mean molecular thermal velocity; average flow velocity in nozzle

xxii
V

carrier gas volumetric flow rate


V
DS
drain-source voltage
V
T
threshold voltage
angular velocity
w mask aperture width
w(r) interaction potential
electron wavefunction
Z
A,B
nuclear charge

1
Chapter 1: Introduction to organic
semiconductors and devices
1.1 Overview
In this chapter the class of materials termed organic semiconductors is defined. The
molecular structure and the resulting electronic and optical properties of thin films and
crystals are reviewed first, followed by a discussion of bonding interactions in these
films. We then briefly review the operating principles of selected organic-based devices,
where control of molecular structure, film morphology and layer thickness determines
device operation. Optical and electronic device physics, coupled with unique material
properties of organic semiconductors drive the development of alternative device
processing methods. In Ch. 2 the existing fabrication techniques are outlined, including
their advantages and shortcomings with regard to precision and potential cost. Solvent-
based polymer device fabrication is typically favored its potential low cost, while vacuum
growth methods for molecular organic devices have typically resulted in high
performance devices through more precise control of the layer structure. Two novel
methods - organic vapor phase deposition and vapor jet printing - are then introduced as
promising alternatives to vacuum growth for the deposition of small molecular organic
electronics. Chapters 3-7 analyze in greater detail these methods. We summarize the
work in Ch. 8, offering some thoughts on new applications and future research directions.

2
1.2 Classification of Solids
Solids can be classified as covalent, ionic, metallic, or molecular, based on the type of
bonding interaction between the constituent atoms or molecules. (Fig. 1-1) Traditionally,
electronic and optoelectronic devices have been fabricated from covalent materials, such
as silicon, germanium, and Group III-V and II-VI elements (Ga, As, N, In, etc.). Metals
are used as electrical contacts, getters and electrical dopants. In recent years organic
semiconductor materials have been attracting increasing scientific and commercial
interest due to their potential for application in electronic and optoelectronic devices.
As Fig. 1-1 indicates, organic semiconductors belong to the broad class of
molecular solids that can be further subdivided into inorganic and organic. An absolute
definition of a molecular solid is lacking, although Kitaigorodsky (Kitaigorodsky 1973)
proposes one where the distance between atoms within a molecule is smaller than the
distance between atoms in different molecules. And while inorganic molecular solids
exist (e.g. I
2
crystals), Kitaigorodsky points out that owing to the multitude of organic
compounds, molecular solids are essentially all organic. The latter are carbon-rich
compounds whose virtually infinite variety of chemical structure and composition is due
to the versatility of combination of compounds based on the carbon bond.
Compounds containing a large number of resonantly alternating sequences of
single and double bonds are called conjugated, and exhibit ground-state delocalization of
charge over the conjugation length, resulting in increased electrical conduction compared
to saturated hydrocarbons. These highly conjugated compounds constitute organic
semiconductors, which further differentiate into small molecules (molecular weight <
1000 amu) and polymers (with molecular weight, mw, typically >> 1000 amu). A more

3
Figure 1-1: a) General classification of solids into 4 types based on bonding
character. b) Here, molecular materials are of interest, which subdivide into
inorganic and organic. c) Organic materials are typically carbon-rich compounds,
ranging from simple gases (e.g. methane) to complex biological macromolecules
(e.g. DNA). Organic semiconductors belong to a subgroup of organic compounds
which contain a large number of conjugated C-C bonds, which can give rise to
electrical conduction. d) Organic semiconductors further differentiate into small
molecular and polymeric compounds. Small molecular compounds typically
weigh <1000 g/mol and exhibit a well-defined molecular structure, with all
molecules in a purified sample identical to each other. In contrast, polymers are
characterized by the structure of the monomer subunit, while the individual chain
length can vary in a single sample; molecular weights are typically >1000 g/mol.
Organic
Polymeric
Molecular
Covalent Ionic Metallic
Inorganic
E.g.:
Si, Ge, SiO
2
BN, Diamond
E.g.:
NaCl, LiF
AlCl
3
, CuSO
4
E.g.:
Au, Ag
Cu, Al, Mg
E.g.: Ar, Xe, SF
6
, H
2
O, HCl, I
2

PEDOT
PSS
C
60
PPV
a-NPD
Small Molecular
Pentacene
a)
b)
d)
Simple molecules
Saturated HCs
Conjugated HCs
Multi-ring Aromatics
Biological
E.g.: C
6
H
12
, C
2
H
3
Cl
n- & branched alkanes
E.g.: DNA, RNA
Proteins, Lipids
c)
Alq
3
N N
n

4
rigorous distinction between molecular and polymeric compounds is that the constituent
moieties of a purified small molecular weight compound have well-defined, identical
chemical composition and structure; however, the molecular chains of a polymeric
compound can vary significantly in length, without changing the chemical name or the
basic structure. Hence, in this case, the mw is not well defined. We focus on the small
molecular weight materials, which for convenience will be referred to simply as
molecular, in contrast to polymeric. Despite the extended nature of polymer chain
compounds, many of the important electronic properties are shared with molecular
compounds in thin film form, due to the introduction of disorder and coiling of chains in
the course of thin film deposition. Much of the discussion is therefore also applicable to
polymeric semiconductors.
1.3 Advantages and disadvantages of organic electronics
The appeal of organic electronics lies primarily in the potential ease of processing and the
resulting low cost of certain types of devices (e.g. thin film transistors and solar cells), or
improved architectures for emissive devices (e.g. lighting and flat panel displays; see Fig.
1-2). Due to the relatively high permeability of organic thin films to gases, they are also
candidates for chemical sensing.
Organics both polymers and small molecules are called soft materials,
referring to the weak bonds in the molecular solid. This means that films can be deposited
without concern for lattice match with the underlying substrate. (Forrest 1997) The film-
substrate adhesion strength remains approximately the same as adhesion of the film
molecules to each other. Consequently, low-cost, large-area, light-weight and flexible
substrates can be used, such as glass and plastic. (Fig. 1-3) Since no chemical bonds need

5
to be broken and reformed to make the organic films stick to the substrate, they can be
deposited at much lower temperatures than e.g. silicon or GaAs. Thus, organic device
processing typically has a low thermal budget, which can decrease the cost and
complexity of fabrication compared to inorganic materials.
Polymeric semiconductors in many cases possess analogous optical and electronic
properties, and can be solution-processed, either by spin-coating or ink-jet printing (See
Ch. 2), effectively addressing the low-cost processing requirement. However, while
solution-based fabrication may at first seem less troublesome than vaporization methods,
devices employing molecular organics as active layers often exhibit superior electrical
and optical characteristics, in part due to the ability to deposit sophisticated, high-
performance multi-layer structures with atomically sharp interfaces. Such structures are
generally difficult to obtain using solution techniques, limiting the scope of application,
and/or shifting the burden of design onto synthetic chemistry to precisely control phase
separation on the nanometer-scale. The arguments in favor of polymers have traditionally
Figure 1-2: a) 13- inch diagonal active matrix display built by Sony Corp.
using organic LED technolgoy. (From Forrest 2004). b) White emission OLED
illuminating a color strip (inset: white OLED in ambient conditions). (From
Adamovic et al. 2003).
b) a)

6
been based on the observation that molecular-organic processing involves more costly
techniques, such as vacuum evaporation, particularly in the large-area device
applications. The vapor phase techniques presented here potentially lend the large-area
and low-cost advantages to small molecular semiconductors.
1.4 Electronic structure and properties of molecular organic
semiconductors
Most of the electronic properties of organic semiconductors, as well as those dictating the
choice of processing techniques arise from the interplay between the inter- and intra-
molecular forces and molecular symmetry. In Sec. 1.4.1 covalent bonding is reviewed,
emphasizing the origin of -bonds, or electronic conjugation. In Sec. 1.4.2 inter-
molecular interactions are reviewed; together with molecular order and intra-molecular
bonding, they determine electronic and energy transport processes (Sec. 1.4.3 and 1.4.4)
in molecular thin films used in device applications such as OLEDs, PV cells, TFTs. A
discussion of device processing methods is deferred until Ch. 2.
Figure 1-3: a) A passive matrix OLED deposited on plastic (Image courtesy of
Universal Display Corporation). b) A pentacene transistor circuit deposited on
plastic (Image by Jackson group at Penn State University).
a) b)
Figure 1-3: a) A passive matrix OLED deposited on plastic (Image courtesy of
Universal Display Corporation). b) A pentacene transistor circuit deposited on
plastic (Image by Jackson group at Penn State University).
a) b) a) b) a) b)

7

1.4.1 Intramolecular bonding
Atoms combine to form molecules typically by sharing unpaired valence electrons; one
typical representation is an energy level diagram, illustrated in Fig. 1-4. Here, two atomic
orbitals (AOs) combine to form two energy-split molecular orbitals (MOs) (Fig. 1-4a).
The lower energy MOs are called bonding, while the higher energy MOs are antibonding,
(e.g. and *, or and *, respectively). If several AOs combine, an equal number of
MOs are created; increasing the number of AOs results in decreased energy spacing
within the bonding orbital manifold, and similarly for the antibonding manifold (Fig. 1-
4b). The Aufbau rule dictates that lowest energy orbitals are filled first, while the Pauli
Figure 1-4: Atomic orbitals (AOs) (s, p) combine to form molecular orbitals (MOs)
(, ). The MOs with energy lower than the AOs are called bonding, while those
higher in energy are called antibonding (designated by a *). Multiple AOs of
identical or similar energy combine to form degenerate or closely spaced MOs,
forming bands when the energy spacing is significantly smaller than the kT energy.
The greater the number of starting AOs, the smaller the energy spacing. The
electrons supplied by the atoms fill the MOs according to the Pauli exclusion
principle and the Aufbau rule, first populating the bonding orbitals, followed by the
antibonding. Hence, they are also referred to as highest occupied MOs (HOMOs)
and lowest unoccupied MOs (LUMOs).
s, p
s, p
,
,
AO
1
AO
2
HOMO
LUMO
s, p
,
,
AO
HOMO
LUMO
s, p
AO
a)
b)

8
exclusion principle ensures that only up to two electrons can reside in a single orbital,
provided their spins are antiparallel. In a thermodynamically stable molecule, more
electrons occupy the low-energy lying bonding MOs, while the higher energy anti-
bonding MOs remain empty. The highest energy bonding orbital is also called the highest
occupied MO (HOMO), and correspondingly, the lowest energy antibonding MO is
called the lowest unoccupied MO (LUMO). Neutral electronic excitations proceed via a
promotion of an electron from HOMO to LUMO, typically by absorption of light or by
thermal excitation. This will be discussed further in Sec. 1.4.3.
Intra-molecular bond strengths can vary significantly, (e.g. from 1.56 to 11 eV for
I
2
and CO, respectively), depending on the complexities of short-range quantum
mechanical electronic and nuclear interactions. Experimentally determined interatomic
bond lengths correlate with the bond enthalpy, as shown in Fig. 1-5 for diatomic
molecules. (Oxtoby et al. 1990) These results can also be readily generalized to
polyatomic molecules, since it is known that the covalent bond length and strength
between two atoms varies <10% in virtually all of the compounds containing this bond.
(Oxtoby et al. 1990) For example, the C-C bond is ~1.525 0.025 for compounds
ranging from ethane, to diamond, to benzene, to DNA. This can be important when
considering the thermal budget for device processing. For example, a materials
evaporation rate will increase exponentially with its temperature, until the kT energy
exceeds the atomic bond energy, in which case the material can undergo pyrolysis. In
general, however, decomposition will occur before the point of bond pyrolysis, due to
reactions catalyzed by impurities, source container, or the material itself.

9
Figure 1-5: Plot and table of bond strength (as indicated by enthalpy) vs. bond
length for a range of compounds. These typically vary by <10% from compound
to compound. (C=- denotes a triple bond, while C- a conjugated bond as in
benzene. (Source: Oxtoby and Nachtrieb, 1990)
Bond Bond Bond
Molecule Length Enthalpy Enthalpy
(A) (kJ/mol) (eV)
----------------------------------------------------------------
H2 0.751 436 4.52
N2 1.1 945 9.79
O2 1.211 498 5.16
F2 1.417 158 1.64
Cl2 1.991 243 2.52
Br2 2.286 193 2.00
I2 2.669 151 1.56
HF 0.926 568 5.89
HCl 1.284 432 4.48
HBr 1.424 366 3.79
HI 1.62 298 3.09
ClF 1.632 255 2.64
BrF 1.759 285 2.95
BrCl 2.139 219 2.27
Icl 2.324 211 2.19
NO 1.154 632 6.55
CO 1.131 1076 11.2
C-C 1.536 348 3.61
C=C 1.337 615 6.37
C=-C 1.204 812 8.41
C...-C 1.397 505 5.23
0.5 1 5
1
10

B
o
n
d

S
t
r
e
n
g
t
h

(
e
V
)
Bond Length (A)
H
2
HF
I
2 F
2
CO
C-C
(diamond)
0.5 1 5
1
10

B
o
n
d

S
t
r
e
n
g
t
h

(
e
V
)
Bond Length (A)
H
2
HF
I
2 F
2
CO
C-C
(diamond)

10
We now discuss briefly the formation of double C-C bonds, which are responsible
for much of the semiconducting character of conjugated organic compounds. Figure 1-6
is an energy level diagram illustrating the combination of the 2s and 2p electrons of
carbon with 1s electrons of hydrogen in the simplest organic molecule, methane (CH
4
).
Here, carbon forms sp
3
hybrid orbitals with tetrahedral symmetry, resulting in 4 energy
degenerate bonding orbitals (-bonds) and 4 energy degenerate antibonding orbitals (*-
bonds). The electrons from C and H pair up in the lowest energy state, i.e. the -bond. A
similar situation exists for ethane (C
2
H
3
), except one of the hydrogen atoms is substituted
for a methyl group (-CH
3
).
In ethylene (C
2
H
4
), however, the 2p
z
atomic orbital (AO) of carbon remains
unchanged, while the remaining 3 electrons form planar sp
2
hybrid orbitals with a 120
rotational symmetry axis. As Fig. 1-7 shows, the resulting molecular orbital (MO) is a
combination of -bonds between C and H, and - and -bonds between the two C atoms.
Due to the symmetry of the sp
2
hybrid MO, all atoms in C
2
H
4
lie in a plane; this is a
general property of the sp
2
hybrid MO.
The chemical structure of benzene (C
6
H
6
) (Fig. 1-8a) implies an alternating
sequence of single and double bonds. Half of the C-C bonds are -, while half are -
bonds only. However, the vibrational spectrum of benzene has one peak for the C-C
stretch, corresponding to 5.23 eV bond energy and 1.397 bond length, indicating that
all of the C-C bonds are equivalent and intermediate of the C-C and C=C bonds (3.61 eV,
1.536 and 6.37 eV, 1.337 , respectively). To reflect this observation, a resonant
chemical structure is postulated (Fig. 1-8b), where the -electrons are delocalized over
all 6 C atoms. Furthermore, as illustrated in Fig. 1-8c, polyacenes (compounds with two

11
Figure 1-6: a) energy level diagram of the 1s 2s 2p AOs of carbon undergoing
hybridization to form 2sp
3
orbitals. b) The spherical 2s and the dumbell-shaped 2p
orbitals combine to form the asymetric dubmell-shaped 2sp
3
hybrid orbital. c)
Energy level diagram of the carbon-hydrogen bond formation, along with (d) the
resultant tetrahedral structure of methane (CH
4
).
Carbon:
1s
2s
2p
E
Hybridization
1s
sp
3
E
1s
sp
3
Carbon:
1s (x4)
Hydrogen:

+ 4x 4x
Methane
a)
s p
Methane
sp
3
s
or bond
b)
c)
d)

12
Figure 1-7: a) energy level diagram of the 1s 2s 2p AOs of carbon undergoing
hybridization to form two sp
2
and one p orbitals. b) Formation and energy level
diagram of MOs between C and H and and MOs between C and C. c)
Resulting planar structure of ehtylene (C
2
H
4
), with 120 angle between bonds
characteristic of the sp
2
hybridization.
Carbon:
1s
2s
2p
E
Hybridization
1s
2sp
2
a)
2p
b)
C C
H
H
H
H
C C
H
H
H
H
c)
1s (x2) 1s (x2)
,

Hydrogen Hydrogen Carbon Carbon
E


13
or more benzene rings fused together) exhibit even greater electron delocalization, which
is manifested in the decreasing energy spacing between the HOMO and LUMO states,
E
HOMO

LUMO
, decreasing ionization potential, and increasing electron affinity. The
combined effect is to decrease the energy associated with accommodation of excess
charge; that is, a dopant may be ionized either by donating its electron to the host matrix
Figure 1-8: a) Resonant chemical structure of benzene, with an alternating
sequence of single and double bonds. In actuality, all of the carbon-carbon bonds
are equivalent and intermediate of the C-C and C=C types. b) An alternative
representation of benzene, reflecting the equivalency of all carbon-carbon bonds.
c) Chemical structure of several polyacenes (naphthalene, anthracene, tetracene,
and pentacene), where the increasing spatial delocalization of the electron clouds
leads to a decrease in the HOMO LUMO transition energy, E
HOMOLUMO
, a
decrease in the ionization potential, I
p
, and a simultaneous increase in the electron
affinity, A
e
. (Data adapted from Pope & Swenberg, p.204)
a) b)
c)
Compound
E
HOMO
LUMO
5.95 eV
4.34 eV
3.31 eV
2.60 eV
2.14 eV
I
p
A
e
9.2 eV
8.2 eV
7.5 eV
7.0 eV
6.7 eV
--
1.8 eV
2.0 eV
2.4 eV
2.9 eV
a) b)
c)
Compound
E
HOMO
LUMO
5.95 eV
4.34 eV
3.31 eV
2.60 eV
2.14 eV
I
p
A
e
9.2 eV
8.2 eV
7.5 eV
7.0 eV
6.7 eV
--
1.8 eV
2.0 eV
2.4 eV
2.9 eV

14
or taking it away. In the first case A
e
of the host matrix is high, while I
p
of the dopant is
low; the dopant behaves as a donor. In the second case, the I
p
of the host is low, while A
e

of the dopant is high; the dopant is said to be an acceptor.
1.4.2 Intermolecular interactions
Figure 1-9 lists the different types of interatomic and intermolecular attractive
interactions, along with their functional form. Molecular solids are typically comprised of
neutral molecules, which interact via electrostatic and polarization forces, grouped
together as van der Waals interactions. (Gutmann et al. 1981; Israelachvili 1992; Pope et
al. 1982) The interaction potential w(r) is written generally:
( ) =
n m
A B
w r
r r
(0.1)
where r is the separation between the subunits, A and B are constants, n and m are
typically positive integers. The physical origin of first term is the Coulombic repulsion of
molecular electron clouds, while the second term denotes interactions such as the London
dispersion force that the electrons of one molecule feel for the positively charged nuclei
of another. Typically, n > m; the special case of m = 6 and n = 12 is known as the
Lennard-Jones potential, plotted in Fig. 1-10. The interaction energy is infinitely large for
r 0, reflecting the repulsion of molecular electron clouds, and asymptotes to 0 for
infinite separation. The largest negative interaction energy is the point of strongest net
attractive interaction between the molecules, resulting in an equilibrium separation r = r
0
,
termed the van der Waals radius. The constant B is typically comprised of several others,
depending on the number and type of VDW interactions involved, including dipole-

15
Fig. 1-9: Common types of interactions between atoms, ions, and molecules in
vacuum. w(r) is the interaction free energy (in Joules); Q, electric charge (in
Coulombs); u, electric dipole moment (in Coulombmeters); , electric
polarizability (Cm
2
/J); r, distance between interacting moieties (in m); k,
Boltzmann constant; T, temperature (in K); h, Plancks constant; electronic
absorption (ionization frequency) (in s
-1
);
0
, dielectric permittivity of vacuum (in
C
2
/Jm) (Source: Israelachvilli, 1992)
Covalent, Metallic
Charge - charge
Charge - dipole
Dipole - dipole
Charge - non-polar
Dipole non-polar
Non-polar
non-polar
Hydrogen bond
Type of Interaction Interaction energy w(r)
Complicated, short range
(Coulomb energy)
- Q
1
Q
2
/ 4
0
r
- Qu cos / 4
0
r
2
- Q
2
u
2
/ 6 (4
0
)
2
kT r
4
- u
1
u
2
[2 cos
2
cos
2

sin
1
sin
2
cos] / 4
0
r
3
H H
H
2
H
O
H
H
2
O
Q
1
Q
2
r
u Q
2
r
Fixed dipole
u
Q
2
r
Rotating
u
1
r
1
Fixed
u
1
r
Rotating
u
2

2
u
2
Q
r
u

r
Fixed
u

r
Rotating

1

2
r

- u
1
2
u
2
2
/ 3 (4
0
)
2
kT r
6
(Keesomenergy)
- Q
2
/ 2 (4
0
)
2
r
4
- u
2
(1 + 3 cos
2
) / 2 (4
0
)
2
r
6
- u
2
/ (4
0
)
2
r
6
(Debye energy)
- h
2
/ (4
0
)
2
r
6
(London dispersion energy)
Complicated, short range,
Energy - r
-2
H
O
H H
H
O
O
H
H
O
H
H
Covalent, Metallic
Charge - charge
Charge - dipole
Dipole - dipole
Charge - non-polar
Dipole non-polar
Non-polar
non-polar
Hydrogen bond
Type of Interaction Interaction energy w(r)
Complicated, short range
(Coulomb energy)
- Q
1
Q
2
/ 4
0
r
- Qu cos / 4
0
r
2
- Q
2
u
2
/ 6 (4
0
)
2
kT r
4
- u
1
u
2
[2 cos
2
cos
2

sin
1
sin
2
cos] / 4
0
r
3
H H
H
2
H
O
H
H
2
O
Q
1
Q
2
r
u Q
2
r
Fixed dipole
u
Q
2
r
Rotating
u
1
r
1
Fixed
u
1
r
Rotating
u
2

2
u
2
Q
r
u

r
Fixed
u

r
Rotating

1

2
r

- u
1
2
u
2
2
/ 3 (4
0
)
2
kT r
6
(Keesomenergy)
- Q
2
/ 2 (4
0
)
2
r
4
- u
2
(1 + 3 cos
2
) / 2 (4
0
)
2
r
6
- u
2
/ (4
0
)
2
r
6
(Debye energy)
- h
2
/ (4
0
)
2
r
6
(London dispersion energy)
Complicated, short range,
Energy - r
-2
H
O
H H
H
O
O
H
H
O
H
H

16
dipole, dipole-induced dipole, induced dipole-induced dipole, and hydrogen bonding (see
Fig. 1-9).
The energy range for typical VDW bonds is 10
-3
10
-2
eV (Israelachvili 1992;
Silinsh et al. 1994), two to three orders of magnitude lower as compared to
intramolecular bonds. For a neutral, nonpolar molecule such as pentacene (Fig. 1-8)
embedded in a molecular solid (e.g. molecular thin film) only the London Dispersion
force contributes significantly to w(r). The nominal intermolecular separation is r
0
= 3,
and a two-fold increase in r results in an 85% reduction in w(r). The energy required to
remove such a molecule from its matrix into the gas phase, enthalpy of vaporization
(H
vap
) is typically below ~200 kJ/mol, and increases with the molecular size. For
example, for the series of molecules benzene, naphthalene, anthracene, tetracene, and
pentacene (see Fig.1-8), H
vap
= 10.7, 15.7-19.6, 22.8-24.4, 28.1-29.8, 33.5-37.7 kJ/mol,
respectively. (Kitaigorodsky 1973) This is significantly lower than the 600-1000 kJ/mol
lattice energy of typical covalent or ionic solids, such as alkali halides, (Oxtoby et al.
1990; Silinsh et al. 1994) allowing the use of relatively low temperatures for the
evaporation of molecular organic compounds. At the same time, the lower adhesion
energy of VdW bonded materials allows the deposition of ordered thin films on a variety
of substrates, whether or not the film and the substrate are lattice-matched. (Forrest 1997)
While the weaker intermolecular adhesion increases the choice of compatible
substrates, it also can restrict the number of compatible techniques of multi-layer growth.
For example, solvent-processing can be difficult; since the absolute value of interaction
energy is small, the difference in the interaction energy between different types of
molecules and solvents is even smaller. This can lead to unwanted penetration of solvent

17
molecules into the underlying thin films, as well as suboptimum selectivity based on
solubility. Clearly, traditional photolithography relying on solvent-processed photoresist
films is a poor choice for organic semiconductors.
1.4.3 Electronic conduction in conjugated organic solids
In traditional, covalent semiconductors where short-range (~1) order can persist over
large length scales (10s of microns to 10s of centimeters). This order and the equivalence
of interatomic bonds lead to formation of wide conduction and valence bands, CB and
VB, respectively. (See Fig. 1-11) The CB and VB are separated by a band of forbidden
states, denoted by the energy band gap, E
g
. In the absence of ionized impurities,
Fig. 1-10: The Lennard-Jones potential of the form E(r) = A/r
12
B/r
6
, where r is
the intermolecular separation and A and B are material-specific constants. The first
term represents a repulsive interaction due to Coulomb repulsion of the electron
clouds, while the second term denotes an attractive interaction due to the induced
dipole- induced dipole interactions originating from correlated fluctuations of the
molecular electron cloud densities. U is the crystal energy and N is the
coordination number of a molecule in the crystal

E
r
0
E
r
0
r
0
r
0
= (2A/B)
1/6
E = A/r
12
B/r
6
U/N = B
2
/4A
U
c
= H
m
+ H
vap
r

E
r
0
E
r
0
r
0
r
0
= (2A/B)
1/6
E = A/r
12
B/r
6
U/N = B
2
/4A
U
c
= H
m
+ H
vap
rr

18
conductivity is said to be intrinsic, with the concentration of negative charge carriers, n
0
,
given by:
0
exp
(
=
(

CB F
C
B
E E
n N
k T
(1.1)
where E
CB
and E
F
are the CB and Fermi-level energies, respectively, k
B
is the Boltzmann
constant, T is the temperature, and N
C
is the conduction band effective density of states
given by:
3/ 2
*
2
2
2
| |
=
|
\ .
e B
C
m k T
N
h

(1.2)
Fig. 1-11: Energy band diagram for a traditional semiconductor, where the
conduction and valence bands (CB and VB, respectively) are at their respective
energy levels, E
CB
and E
VB
, separated by an energy gap E
g
. The chemical potential
of electrons at thermal equilibrium is the Fermi energy, E
F
. The energy gained by
the addition of an electron is the electron affinity, A
e
, while I
p
is the ionization
potential, or the energy required to remove an electron. All energy levels are
measured with respect to the vacuum level at E = 0.
CB
VB
E
Vacuum Level
E
CB
E=0
A
e
I
p
E
VB
E
F
E
g

19
where m
e
* is the electron effective mass, and h is Plancks constant. Equivalent
expressions can be written for positive charge carriers, or holes:
0
exp
F VB
V
B
E E
p N
k T
(
=
(

(1.3)
3/ 2
*
2
2
2
h B
V
m k T
N
h
| |
=
|
\ .
(1.4)
where E
VB
is the VB energy, while N
v
is the valence band effective density of states, and
m
h
* is the hole effective mass. The electrical conductivity is given by = q (n
0

e
+
p
0

h
); it is n-type if the ratio n
0
/p
0
> 1, and p-type if n
0
/p
0
< 1.
Since (E
CB
-E
F
) > 0, increasing the temperature exponentially increases n
0
and p
0
,
and hence also the electrical conductivity. Note, however, that the carrier mobility,
e,h
,
in traditional semiconductors can also vary with temperature. The increasing lattice
vibrations (phonons) at higher temperatures increase the likelihood of carrier-atom
collisions, lowering the intrinsic carrier mobility via
L
~ T
-3/2
. On the other hand, the
increasing thermal velocity of carriers at higher temperatures causes them to spend less
time in the vicinity of ionized impurities, thereby decreasing the net Coulombic
interaction. This is called ionized impurity scattering, and causes the mobility of carriers
from ionized impurities,
I
to vary as T
3/2
. The effect on the net mobility can be expressed
as:
,
1 1 1
= +
e h I L

(1.5)
The situation in organic semiconductors is often quite different. The weak
intermolecular interactions imply that intermolecular electron sharing in most molecular
solids is weak. Indeed, the electron cloud density typically concentrates on the individual

20
atoms and the intramolecular bonds, dropping to zero in the interstices, (Silinsh et al.
1994) as evidenced in the example of a scanning tunneling micrograph of copper
hexadecachloro-phthalocyanine (Cl-CuPc) (Fig. 1-12). Here, the electron cloud
localization on the individual atoms is apparent (dark regions), with substantial sharing
within intermolecular bonds, but little overlap between the atoms of neighboring
molecules. Such electronic structure of the organic condensed phase determines the
electrical and optical properties of these materials, which retain much of their molecular
nature.
Fig. 1-12: Scanning tunneling electron micrograph of Cl-CuPc (copper
hexadecachlorophthalocyanine) molecules in the ab-plane of the crystal. (Image
from Silinsh and pek, 1994, p. 3) where the dark regions indicate higher electron
cloud density, and bright regions their absence.

21


Several in-depth texts and reviews discuss the structure, optical and electronic
properties in organic solids (primarily crystals) (Bulovic et al. 2000; Dimitrakopoulos et
al. 2001; Forrest 1997; Gutmann et al. 1981; Horowitz 1999; Pope et al. 1982; Silinsh et
al. 1994). However, these authors agree that a comprehensive theory of electrical
conduction in organic semiconductors has not been developed, and the subject remains
the topic of vigorous research. Here we outline some key features.
The conduction and valence band picture is inaccurate in the case of organic
semiconductors, which exhibit strong localization and weak coupling. Charges must
overcome a potential barrier when migrating from molecule to molecule, and as a result,
Property
Atomic weight (g/mol)
Melting point (C)
Density (g/cm
3
)
Density (molec/cm
3
)
Crystal structure
Lattice constant (a, )
Volume compressibility (cm
2
/dyne)
Dielectric constant (static)
Intrinsic ionization energy (eV)
Intrinsic conductivity at 300K (, ohm
-1
cm
-1
)
Electron mobility at 300K (
e
cm
2
/Vs)
Hole mobility at 300 (
h
cm
2
/Vs)
Concentration of intrinsic carriers (cm
-3
)
Thermal expansion coefficient (C
-1
)
Thermal conductivity (k, W/cmC)
Specific heat (cal/gC)
Vacuum ionization energy (eV)
Germanium
72.63
937
5.3
4.4210
22
Diamond
5.66
1.310
-12
16
0.78
0.02
3800
1800
5.210
13
6.110
-6
1
0.074
4.8
Anthracene
178.22
217
1.25
0.4210
22
Monoclinic
6.04-11.16
1010
-12
3.4
3.9
~10
-22
0.88
0.44
~10
-4
14510
-6
10
-3
0.31
5.8
Table I: Comparison of some physical properties of germanium and anthracene,
covalent and molecular semiconductors, respecitvely (from Gutmann & Lyons)
Property
Atomic weight (g/mol)
Melting point (C)
Density (g/cm
3
)
Density (molec/cm
3
)
Crystal structure
Lattice constant (a, )
Volume compressibility (cm
2
/dyne)
Dielectric constant (static)
Intrinsic ionization energy (eV)
Intrinsic conductivity at 300K (, ohm
-1
cm
-1
)
Electron mobility at 300K (
e
cm
2
/Vs)
Hole mobility at 300 (
h
cm
2
/Vs)
Concentration of intrinsic carriers (cm
-3
)
Thermal expansion coefficient (C
-1
)
Thermal conductivity (k, W/cmC)
Specific heat (cal/gC)
Vacuum ionization energy (eV)
Germanium
72.63
937
5.3
4.4210
22
Diamond
5.66
1.310
-12
16
0.78
0.02
3800
1800
5.210
13
6.110
-6
1
0.074
4.8
Anthracene
178.22
217
1.25
0.4210
22
Monoclinic
6.04-11.16
1010
-12
3.4
3.9
~10
-22
0.88
0.44
~10
-4
14510
-6
10
-3
0.31
5.8
Table I: Comparison of some physical properties of germanium and anthracene,
covalent and molecular semiconductors, respecitvely (from Gutmann & Lyons)

22
charge conduction is dominated by the electronic structure of the constituent molecules
and thermally assisted hopping between the localized states (Pope et al. 1982). This is the
situation in virtually all molecular organic solids, except in the case of single crystals.
Table I compares some basic physical properties of an inorganic semiconductor
germanium to those of an archetypal organic semiconductor, anthracene. (Pope et al.
1982) Note the order-of-magnitude lower molecular and densities, three orders of
magnitude lower thermal and dramatically lower intrinsic electrical conductivities for
anthracene.
Consider again the series of polyacenes (naphthalene, anthracene, tetracene,
pentacene) in Fig. 1-8c. Here, -bond conjugation results in the delocalization of
electrons over yet larger space than in the case of benzene, several times the C-C bond
length. By increasing the spatial extent of the electron wavefunction, the electronic
polarizability, , of the molecule is increased, which increases the intermolecular
adhesion energy, w(r), as denoted in Fig. 1-9. The ionization potential also decreases,
while the electron affinity increases. The electrical activation of a dopant molecule, D, in
a host matrix, H, can be written as a chemical oxidation or reduction reaction:
H + D H
-
+ D
+
G = A
e
H
+ I
p
D
(1.6)
H + D H
+
+ D
-
G = I
p
H
+ A
e
D
(1.7)
where G is the Gibbs free energy of the reaction, while I
p
H,D
and A
e
H,D
denote the
ionization potential and the electron affinity, respectively, of the host and dopant.
Keeping in mind the convention that I
p
> 0 and A
e
< 0, it is clear that lower I
p
and larger
negative values of A
e
increase the thermodynamic driving force for the ionization of
dopants.

23
It may be tempting to draw an analogy with traditional semiconductors, to say that
charge conduction in pentacene will thus possess more extrinsic character than in
naphthalene. However, because many dopants form deep-level traps which do not yield
free carriers, in some instances dopants can reduce the conductivity. This has been
observed, for example in anthracene by Karl, as mentioned in (Silinsh et al. 1994)
(p.190), illustrated in Fig. 1-13, where local trapping states are formed by guest
molecules. The above formulae allow the estimation of the depth and character of traps,
but only approximately, due to the changes in effective A
e
and I
p
from local distortion of
Fig. 1-13: Charge carrier traps formed by guest molecules (a) tetracene, (b)
acridine, (c) phenazine, (d) anthraquinone, and (e) phenothiazine in anthracene
host crystal (according to the work of Karl). (Figure reproduced from Silinsh and
pek, 1994, p. 192).

24
the host lattice by the guest molecule. Dipolar molecules may also introduce trapping
effects in proportion to the strength of the dipole via the charge-dipole interaction energy,
as listed in Fig. 1-9. Many of the organic compounds can form oxides and nitrides, such
as ones shown in Fig. 1-13 at ambient conditions, with or without exposure to ultraviolet
(UV) radiation. This can both modify the energy level structure of the molecule, as well
as create molecular dipoles. Therefore care should be taken to avoid exposure to air and
direct intense light during the purification stages, and especially during the growth of thin
films for device applications.
As will be shown in Ch. 7 for pentacene, similarly to traditional semiconductors,
molecular order plays an important role in determining charge conduction in organic
materials. Disorder in the intermolecular spacing will decrease the probability of charge
hopping between molecules due to the variation in intermolecular distance and
orientation. Device processing conditions (e.g. temperature and deposition rate) must be
selected not only with regard to the vapor pressure of the organic material, or the thermal
stability of the substrate, but also for optimal control of the deposited film morphology,
which will in turn affect electrical device performance.
1.4.4 Electronic excitations in organic solids
The structure and dynamic behavior of electronic energy states govern the interaction of
the organic solid with light, and are at the heart of optoelectronic device operation. We
therefore briefly review some features of electronic excitations in organic materials
relevant to device structure and processing.
Optical and thermal excitations in covalent semiconductors result in the
promotion of electrons from CB to VB. However, the excited states are typically very

25
weakly (<0.01eV) bound, having the strong effect of raising the free charge carrier
concentration upon illumination. Strong confinement of the excited state can be achieved
in a structure where a layer of small band-gap material is sandwiched between two layers
of a large band-gap material, forming a quantum well. But organic molecules already
behave as individual quantum wells, retaining many of their properties of the constituent
molecules. Here, electrons are excited from the HOMO to one of the LUMO states
E
e
0
E
e
1
r
0
r
0
*
E
e
r
A B
Absorption
(E
abs
)
Internal
conversion
Emission
(E
em
)
Figure 1-14: a) Molecular configuration energy vs. interatomic distance, r,
diagram for the electronic ground and first excited states, labeled E
e
0
and E
e
1
,
respectively. In each electronic state, a series of vibrationally separated states
exist. Fast (10
-14
s) excitation of a ground state electron into an excited state
promotes the molecule into higher vibrational states of E
e
1
, which then relax via
slower nuclear motion into the lower vibrational modes, which can subsequently
can radiatively relax into one of the vibrational states of E
e
0
.

26
thermally, optically, or by electrical injection, forming strongly bound (>0.05 eV binding
energy) charge pairs, or excitons. We first examine optical excitations.
A molecular energy diagram is shown in Fig. 1-14 for a hypothetical bonding
situation involving electrons shared by two atoms. The electronic ground state level is
designated by the curve E
e
0
, and the first excited state by E
e
1
; vibronic energy spacings
are designated by E
v
0
and E
v
1
for the two electronic states, respectively. The lowest
energy for E
e
0
occurs for an equilibrium interatomic distance, r
0
, and grows when r > r
0
.
When r << r
0
, the electrostatic repulsion energy between the nuclei (e
2
Z
A
Z
B
/4
0
r, where
Z
A,B
is the effective charge of atoms A and B, e is the electron charge, and
0
is the
electrical permittivity of vacuum) grows to infinity, but when r >> r
0
the attractive
interaction between atoms and electrons asymptotes to zero. The energy level spacing for
this molecule in its electronic ground state (E
e
0
) is given by the vibrational energy step
E
v
0
proportional to r
2
for small r. Due to the antibonding character of the E
e
1
state (i.e. the
LUMO), it will possess an equilibrium separation r
0
* > r
0
.
The Born-Oppenheimer approximation states that the much heavier nuclei remain
stationary in the molecular frame of reference on the time scale of electronic motion.
Thus, upon absorption of a photon of energy E
abs
, electrons respond to electromagnetic
radiation much faster than the nuclei (on the order of 10
-14
s), and the optical excitation is
effectively vertical in the E vs. r diagram. However, the probability of transition
between two states and * depends on the overlap integral between and *, which is
greater for the states corresponding to higher vibronic levels of the E
e
1
manifold than
lower ones. Thus, molecule is excited into an energy level slightly above the first excited
one, as shown in Fig. 1-14. Eventually the nuclear motion dissipates the excess energy,

27
and the molecule partially relaxes into the ground vibrational state of E
e
1
. Consequently,
radiative decay of the excited state will occur with an emission photon energy E
em
< E
abs
.
This red-shift of the emission spectrum of a molecule is called the Franck-Condon (or
Stokes) shift, and is characteristic of a majority of molecular organic compounds.
b) c)
d) e)
Ground
State
Frenkel
Exciton
Charge Transfer
Exciton
Wannier-Mott
Exciton
E
Electron (e
-
)
Hole (h
+
)
. . .
r
e
r
e
Frenkel
exciton
Charge Transfer
Exciton
Wannier-Mott
Exciton
a)
Figure 1-15: a) Three types of electronic excitations (or excitons) commonly
observed in molecular crystals, including Frenkel, Charge Transfer (CT), and
Wannier-Mott, listed in order of decreasing binding energy, E
b
, and increasing
effective radius, r
e
. b) Energy level representation of the electronic ground state of
a molecule, c) the Frenkel, d) CT, and e) Wannier-Mott excitons.
b) c)
d) e)
Ground
State
Frenkel
Exciton
Charge Transfer
Exciton
Wannier-Mott
Exciton
E
Electron (e
-
)
Hole (h
+
)
. . .
r
e
r
e
Frenkel
exciton
Charge Transfer
Exciton
Wannier-Mott
Exciton
a)
Figure 1-15: a) Three types of electronic excitations (or excitons) commonly
observed in molecular crystals, including Frenkel, Charge Transfer (CT), and
Wannier-Mott, listed in order of decreasing binding energy, E
b
, and increasing
effective radius, r
e
. b) Energy level representation of the electronic ground state of
a molecule, c) the Frenkel, d) CT, and e) Wannier-Mott excitons.

28
Owing to weak intermolecular interactions, the electronic excitation is typically
localized on a single molecule, which then constitutes an exciton. An alternative picture
involving electron and hole pairs can also be useful in understanding the operation of
(opto)electronic devices. When an electron is promoted into the LUMO, it leaves behind
a positively charged vacancy, or a hole, in the HOMO. For weakly interacting molecules
in a solid, this electron-hole pair is effectively localized on the single molecule, forming a
bound quasi-particle, or a Frenkel exciton (Fig. 1-15). (Pope et al. 1982) Stronger
intermolecular interactions and lower exciton binding energy, E
b
, can increase the exciton
radius, r
e
, relative to the lattice constant, a
l
, forming moderately delocalized charge-
transfer (CT) excitons, or highly delocalized Wannier-Mott excitons. Frenkel excitons
Cathode
Anode
HTL
ETL
LUMO
HOMO
Cathode
Anode
HTL
ETL
LUMO
HOMO
Figure 1-16: a) Principle of operation of an organic photovoltaic (PV) cell, where
electrical current generation proceeds via a four-step process: 1. Absorption of
photons to generate excitons, 2. Diffusion of excitons to the donor-acceptor (DA)
interface, 3. Separation of the electron- hole pair due to energetically favorable
HOMO-LUMO alignment of the D and A, and 4. Extraction of charge to the
electrodes. (Figure courtesy of Peter Peumans). b) Principle of operation of an
organic light emitting diode (OLED), where in contrast to the PV cell, charges are
injected from the electrodes, move through the electron and hole-transporting
layers (ETL and HTL), and recombine in the ETL-HTL interface. Doping this
region with guest molecules having tailored band gap energy, E
g
, allows the control
of electroluminescence wavelength.
a)
b)

29
have large (~1eV) binding energies and have small (<5A) radii, Wannier-Mott excitons
have r
e
>> a
l
. The CT exciton often exists in molecular crystals, having E
b
~10-100 meV,
delocalized over two or more molecules or adjacent polymer chains.
Following their generation, excitons can diffuse throughout the organic solid via
hopping between neighboring molecules (Dexter transfer) or by long-range dipole
interaction (Frster transfer). Exciton and charge dynamics are at the heart of several key
optoelectronic devices. In organic photovoltaic (PV) cells, the photogenerated excitons
are separated into electrons and holes at a donor-acceptor (DA) interface, due to
S = 0
S = 0 S = 1, M
s
= 1, 0, -1
= m
s
= +
= m
s
= +
a)
b)
Figure 1-17: a) Two equivalent representations of the electronic ground state of a
molecule with two spin-antiparallel electrons in the HOMO. b) Four possible spin-
states of the excited state, with the total spin, S = 0 or 1 for the singlet and triplet
excitons, respectively. (Figure courtesy of Peter Peumans).

30
energetically favorable HOMO and LUMO level alignment of the D and A species, as
shown in Fig. 1-16a. Organic light emitting diodes (OLEDs) function in reverse of PV
cells, by injecting electrons and holes from the cathode and anode, respectively, and
recombining them at an interface between the electron- and hole-transporting layers (ETL
and HTL, respectively). The charges combine at this interface to form excitons, which
can then radiatively decay. The wavelength of the emitted light can be controlled by
doping the ETL-HTL interfacial region with molecules having specifically tailored
band-gap energy (E
g
), as illustrated in Fig. 1-16b.
Excitons can be formed in a number of different quantum-mechanical spin states,
having wide-ranging implications for devices like OLEDs. Electrons are Fermions, or
spin particles. The electron spin is denoted as s = . The spin angular momentum can
take on two different values with respect to an arbitrarily defined z-axis, denoted by the
quantum number m
s
= . According to the Pauli exclusion principle, two electrons can
pair up in the same MO, provided their spins are antiparallel, as shown in Fig. 1-17a.
The electronic ground state can be written as
1,2
0
= |+>
0
|->
0
, where the total
electron spin value for both electrons is M
s
= m
s1
+ m
s2
= 0. The exciton, on the other
hand, is a quasiparticle comprised of two Fermions and can take on spin values of S = 0
or 1, where S = M
s
of both the HOMO and the LUMO electrons. The electron in the first
excited state,
1,2
1
, however can take on either m
s
= + or m
s
= values, such that the
new state can have two four possible spin state configurations (Fig. 1-17b):
S = 0, M
s
= 0
1,2
1
= | + >
0
| - >
1
- | - >
0
| + >
1
(1.8)
S = 1, M
s
= +1
1,2
1
= | + >
0
| + >
1
(1.9)
S = 1, M
s
= -1
1,2
1
= | - >
0
| - >
1
(1.10)

31
S = 1, M
s
= 0
1,2
1
= | + >
0
| - >
1
+ | - >
0
| + >
1
(1.11)
Where the S = 0 exciton wavefunction is antisymmetric (changes sign) under exchange of
the two electrons (i.e.
1,2
= -
2,1
=

), while the S = 1 wavefunctions are symmetric


(i.e.
1,2
= -
2,1
=
+
), and are triply degenerate in the absence of a magnetic field.
Accordingly, the former are called singlet excitons, while the latter are called triplet
excitons.
The implication for OLEDs arises from the fact that the probability of radiative
transitions (i.e. photoemission) between the excited and the ground states is proportional
to the magnitude of the electronic transition dipole moment between them, |
1

0
|. The
dipole moment for a two-electron system is given by:
1 2
e r e r = (1.12)
where e is the charge on the electron, while r
1
and r
2
are the coordinates of the two
electrons. The dipole moment for the transition between two states
0
and
1
is:
0 1
1 2
( ) e r r = + (1.13)
If the ground and excited states are both singlets, under exchange of the electrons we
have:
0 1 0 1
1 2 2 1
( ) ( ) e r r e r r

= + = + (1.14)
whereas if the ground state is a singlet, and the excited state is a triplet, we have:
0 1 0 1
1 2 2 1
( ) ( ) e r r e r r
+ +
= + + (1.15)
Since the transition dipole moment cannot depend on the exchange (or reversing the
labeling) of electrons, = 0 in the case of the singlet-to-triplet transition, implying that it
is forbidden and non-radiative. The reverse is also true; optical excitations are from the

32
singlet ground state to the singlet excited state. Thus, photoluminescence typically
examines singlet exciton properties. Electrically generated excitons, on the other hand,
can be created in either the singlet or triplet states, but radiative decay is spin-allowed
only to the singlet exciton. Thus, ordinarily, only 25% of the electrically generated
excitons can give photons, while 75% are wasted. (Baldo et al. 1999) In some molecules,
however, the spin selection rules may be broken via the coupling of the electron spin to
the orbital angular momentum of a heavy metal atom (e.g. Pt in platinum
octaethylporphyryn). The singlet and triplet states then become sufficiently mixed, so that
100% of the electrically generated excitons can decay radiatively. (Adachi et al. 2001)
The allowed radiative transition is typically fast (~10
-9
s excited state lifetime) and is
called fluorescence, while the disallowed transition is therefore typically slow (~10
-6

10
-3
s) and is called phosphorescence. The singlet-triplet dynamics can therefore be
studied by time-resolved spectroscopy in both photo- and electroluminescence (PL and
EL) modes. (Baldo et al. 2000; 2000)
In great part the versatility and appeal of organic semiconductors lay in the
richness of synthetic chemical approaches (Katz et al. 2001) that can be used to tailor the
optoelectronic properties of the organic film to its desired application. In connection to
the singlet / triplet dynamics in organic molecules, for example, heavy metal atom-
chelates (e.g. Pt-and Ir-containing compounds) can be synthesized to increase the EL
efficiency of OLEDs. Furthermore, by substituting electron-rich or electron-withdrawing
ligands in a platinum-containing compound such as shown in Fig. 1-18, enables the
tuning of the compounds luminescence wavelength. (Brooks et al. 2002)

33

1.5 Summary
In this chapter the class of materials organic semiconductors was defined on the basis of
their intra- and intermolecular bonding characteristics. The latter are governed by
relatively weak van der Waals interactions, broadening the choice of compatible
substrates, but also require the development of novel methods of thin film and multilayer
Figure 1-18: CIE (Commission International dEclairaige) diagram of
photoluminescence of phosphorescent cyclo- metalated platinum complexes. (From
Brooks et al. 2002)
Figure 1-18: CIE (Commission International dEclairaige) diagram of
photoluminescence of phosphorescent cyclo- metalated platinum complexes. (From
Brooks et al. 2002)

34
growth. The electronic and optical properties of conjugated organic solids derive from the
electronic energy structure of individual molecules, dominated by the extended -
electron system. Molecular structure can be tuned to control the electrical conductivity, as
well as exciton energy, leading to improvements in device applications, exemplified by
the use of emissive triplet excited states of phosphors to quadruple the efficiency of
OLEDs.


35
Chapter 2: Organic device fabrication
technology
2.1 Overview
Molecular organic electronic devices typically consist of thin (<100nm) films of
molecular semiconductors. The films adhere by van der Waals forces, with the strength
of interaction on the order of only a few kT (at room temperature), decreasing quickly
with intermolecular separation. The bonds between the molecular sub-units in van der
Waals solids are weaker than in covalent, ionic, or metallic solids, resulting in a greater
tendency to form micro-crystalline and amorphous films. Thus, even when deposited
onto highly ordered substrates, relaxation of strain due to lattice mismatch occurs after
only a few monolayers (Forrest 1997), obviating the need for lattice-matching between
the deposited layers and the underlying substrate. This considerably broadens the choice
of substrates in the growth of organic devices, including low-cost glass, metal and plastic
foils.
As the electronic and optoelectronic performance of organic devices improves and
reaches commercialization potential, scale-up of processing technology becomes
important (Bardsley 2004; Forrest 2004), and the cost of materials is one of the major

36
considerations. Here it is important to consider both the deposition of organic thin films,
as well as metal-oxides and metals that complete the device structure.
Figure 2-1a shows the sharp decrease in the cost of devices as production volume
rises, estimated from the cost of materials and prior experience of device manufacturing.
(Bardsley 2003) While in traditional semiconductor manufacturing the cost of materials
is relatively small (e.g. 6% in the case of DRAM (Bardsley 2004)), it can be as high as
45% in LCD manufacture (Bardsley 2004). This is partly due to the device structure and
the nature of materials used. For example, an essential part of both LCD- and OLED-
based displays is the conductive transparent ITO anode, fabricated typically by power-
intensive DC-magnetron sputtering of a costly indium-tin target at elevated substrate
temperatures, typically above 200C. The important figures of merit for ITO-coated glass
in LCD and OLED applications are electrical conductivity and transparency, which have
been found to improve with higher substrate temperature during sputter-deposition (See
Fig. 2-1b). This is an important consideration in manufacture of OLEDs on flexible
substrates, since, as Fig. 2-1c shows, the cost of plastic substrates capable of
withstanding temperatures above 200C can be orders of magnitude higher than the more
fragile plastics.
With regard to the deposition of organic thin films, the same relatively weak
bonding forces that broaden the choice of substrates, also render standard semiconductor
processing techniques (high temperatures and solvents) too damaging of the thin films,
requiring the development of novel device fabrication approaches.(Bardsley 2004;
Blanchet et al. 2003; Forrest 2004; Rogers 2001; Rogers et al. 2002; Sirringhaus et al.
2001) A few particularly promising methods are discussed below. Some of them only

37
Figure 2-1: (a) The cost of flexible displays versus production volume
(US Display Consortium). (b) The cost of plastic substrates with
maximum processing temperature. (c) The resistivity and transmission of
ITO sputtered onto a glass substrate versus the substrate temperature.
(Source: Applied Films) (Source: Hewlett-Packard). (Figures adapted
from Bardsley 2003)
Production (ft
2
/week)
0 40,000 80,000
$

/

f
t
2
0
100
200
300
400
$ 34
$ 75
Active Matrix OLED by Deposition
Passive Matrix OLED by Lamination
Production (ft
2
/week)
0 40,000 80,000
$

/

f
t
2
0
100
200
300
400
$ 34
$ 75
Active Matrix OLED by Deposition
Passive Matrix OLED by Lamination
(a)
(b)
(c)
0 100 200
300 400
100
200
300
400
500
600
20
40
60
80
100
R
e
s
i
s
t
i
v
i
t
y
(
m

/
c
m
2
)
Substrate Temperature (C)
0 100 200
300 400
100
200
300
400
500
600
20
40
60
80
100
R
e
s
i
s
t
i
v
i
t
y
(
m

/
c
m
2
)
Substrate Temperature (C)
100 140 180 220 260
0.1
1
10
100
1000
Temperature (C)
S
u
b
s
t
r
a
t
e

c
o
s
t

(
$
/
m
2
)

(
u
n
c
o
a
t
e
d
)

38
enable the large-area deposition of organic thin films with or without the capability to
deposit heterostructures or pattern the films in the substrate plane. A handful of
approaches permit the simultaneous deposition and patterning of organic thin films, and
are, in principle, more suited to achieving the objective of low-cost of fabrication, from
the standpoint of throughput and less constly materials. (Bardsley 2004)
In many applications (e.g. full-color OLED displays), patterning of organic layers
is a requirement that places additional demands on the fabrication sequence and tooling.
Some of these concerns (e.g. film thickness uniformity across the substrate) are inherent
to the device structure, and thus must be addressed by all fabrication techniques. Other
concerns (e.g. solvent compatibility or mask cleaning) arise from the unique aspects of
each particular method. Several major fabrication techniques are described below, along
with their advantages and disadvantages.
2.2 Spin-on film deposition
A well-established technology already exists in the traditional semiconductor processing
industry for depositing thin films of organic materials spin-on. (Wolf et al. 1999) Spin-
on (See Fig.2-2a) is an essential step in photolithography, where a solution of polymer
photoresist is dispensed onto a wafer, which is then rapidly accelerated (typically at
~20,000rpm/s ramp rate) to a final angular velocity of = 3000-7000 rpm. The
centrifugal force disperses the polymer film across the rotating surface as the solvent
evaporates and the polymer solidifies. The final polymer layer thickness resembles the
profile shown in Fig. 2-2b, while the average coating thickness is given by, ignoring
evaporation,
2 2 1/ 2
4
( ) (1 )
3
o o
t h t

= +

, (Emslie et al. 1958) where


0
is the initial

39
coating thickness, is the polymer solution density, is viscosity, t is the spinning time,
and is the angular velocity. While mathematical formulae have been developed to
predict based on solution properties, surface and spin conditions, most users rely either
on the photoresist manufacturers data specific to the particular photoresist used, or
calibrate their own set of materials and spinners.
Most spin-coating steps are completed within 20-30 seconds, and result in
thickness coating uniformity of 100 /s, if > 6000 rpm. The photoresist spin-on is
most often followed by a baking step to drive off residual solvent and further densify the
film. Exposing the photoresist to ultraviolet (UV) radiation through a patterned optical
z
r

v
r
(z)
Figure 2-2: Coating of substrates using spin-on. A polymer solution is dispensed
onto a substrate, which rotates at an angular velocity = 3000-7000 rpm.
Centrifugal forces spread the solution over the surface, as the solvent evaporates.
Spin-on is often followed by a bake to drive off any residual solvent and further
densify the coating. Better film thickness uniformity across the substrate is
obtained at higher , but is typically limited to 100.
(a)
(b)


40
mask induces chemical changes in the resist, which are then chemically developed,
creating a corresponding pattern of openings. Additional material can be then deposited,
or the underlying material can be selectively etched in the open areas. After removing the
photoresist residue, the process can be repeated to generate complex patterns in the
substrate.
Since the photoresist is used only for patterning, the 100 thickness variation
is adequate for traditional semiconductor manufacture, where the feature size is on the
order of 100nm or larger in the lateral direction. However, since many of the organic
(opto)electronic devices rely on active organic layers as thin as 500 , the coating non-
uniformities are thus an unacceptable 20-40% of the total film thickness.
Furthermore, the photoresist is traditionally spun onto covalently-bonded
substrates (e.g. glass, silicon), which are impervious to the organic or aqueous solvents.
In fabricating organic heterostructures, solvents used for the alternating organic layers
must also follow an alternating organic-aqueous sequence. This severely constrains the
choice of solvents and active organic semiconductor materials, especially in view of low
room-temperature solubility of organic semiconductors in common solvents. An
alternative is to design the polymer chemical structure to either perform different
electronic functions (e.g. a p-n junction contained within a single chain of a di-block co-
polymer), or control the phase segregation of two polymers mixed together (e.g. electron-
donating and electron-accepting polymers). (Katz et al. 2001; Moons 2002) The
drawback of these approaches is the dual burden placed on the polymer chemist to
simultaneously and independently tune the rheological and electronic properties of the

41
compound, to ensure electronic performance (e.g. conductivity, luminescence quantum
efficiency), as well as the proper p-n junction orientation with respect to the electrodes.
2.3 Laser induced thermal imaging
Laser induced thermal imaging (LITI) (Blanchet et al. 2003; Blanchet et al. 2003) is
illustrated in Fig.2-3; it is a variant of the dye sublimation printing technique, where
highly localized heating is used to transfer ink from a donor sheet onto a substrate. For
LITI in particular, a large plastic (donor) sheet is coated with an organic semiconductor
(polymer or small-molecule) and brought into intimate contact with the substrate. A laser
pulse impinges onto the back of the donor sheet, vaporizing the semiconductor off the
sheet and onto the underlying substrate. In this fashion, the solvent compatibility issue is
solved, since two different organic slabs can in principle be transferred sequentially from
two separate donor sheets. In addition, the lateral patterning of the active organic layers is
accomplished without exposure to solvents or UV light.
Both polymer and small molecule (pentacene) TFTs have been fabricated using
this method, on notably large (e.g. 1m x 1m) plastic substrates, with pentacene TFTs
exhibiting field-effect hole mobilities of up to 0.3 cm
2
/Vs. While the precise mechanism
of the transfer is not yet fully understood, the technique holds promise for low-cost
electronics fabrication using much of the existing laser printing equipment infrastructure.
A potential draw-back of LITI is that it still requires the fabrication of the donor
sheet. Spin-on or spray-coating of the donor sheet are well-suited for covering large
areas, but again suffer from poor control of thin-film thickness uniformity. Thus, other
methods of organic layer deposition with better control of film thickness (e.g. vacuum
thermal evaporation (VTE), or organic vapor phase deposition (OVPD)) may be needed

42
to create the donor sheet (they are discussed in Sec. 2.5 and 2.6). Since LITI and
VTE/OVPD do not share the same manufacturing platform, it can be potentially difficult
and costly to integrate these processes. (Bardsley 2004)
2.4 Ink-jet printing
Ink-jet printing is a now-ubiquitous patterning technique, where picoliter droplets of
liquid ink are ejected from microscopic nozzles onto a substrate (e.g. paper or
transparency film). Three variations on this method are illustrated in Fig. 2-4. Thermal
ink-jet (TIJ), developed primarily by Hewlett-Packard (Askeland et al. 1988; Shields
1992), utilizes heat to nucleate and then rapidly expand a bubble in a fluid reservoir,
which then pushes a liquid jet through a tiny nozzle. Another technique developed by
Figure 2-3: a) Illustration of the laser induced thermal imaging (LITI);process. A
donor sheet pre-coated with an organic semiconductor and brought into contact
with the substrate. A laser pulse impinges on the back of the donor sheet,
presumably vaporizing some of the organic material, most of which transfers to
the substrate. The donor sheet is then peeled off; (from G. Blanchett, APL 2003)
b) Example of a 50cm x 50cm array of polymer TFTs printed on plastic using
LITI (from G. Blanchett, NSF Workshop on Organic Electronics, 2003).
(a)
(b)

43
Epson uses a piezoelectric element to rapidly generate a pressure wave inside of a small
container of liquid, which also results in a collimated liquid jet exiting the reservoir
through an orifice. (Percin et al. 2003) In a third variation developed by Xerox PARC,
(Elrod et al. 1989) focused pressure waves from a piezoelectrically actuated acoustic lens
eject droplets as small as 5m in diameter from the surface of a liquid onto a substrate.
By substituting the ink with a solution of conducting or light emitting polymers
(Bharathan et al. 1998; Yang et al. 2000) or small molecular dyes (Hebner et al. 1998),
circuits (Sirringhaus et al. 2001; Sirringhaus et al. 2000), conducting metal films (Huang
et al. 2003) and OLEDs (Bharathan et al. 1998; Yang et al. 2000) can be directly
patterned onto the substrate. Ink jet printing is particularly attractive for organic
electronics fabrication (Calvert 2001), since it is a well-established technology (albeit for
document printing), well suited for large area and flexible substrates. High throughput
can be achieved by utilizing multi-nozzle arrays (Creagh et al. 2003); for example, state-
Heater
Piezoelectric actuator
Piezoelectric
actuator
Heater
Piezoelectric actuator
Piezoelectric
actuator
Figure 2-4: Three variatiants of ink-jet printing: a) Thermal ink-jet (Hewlett
Packard & Canon); b) Piezoelectric actuation (Epson); c) Acoustic lens (Xerox
PARC). (Figure courtesy of Peter Peumans)
(a) (b) (c)

44
of-the-art commercial Hewlett-Packard and Epson printers use arrays of over 500 nozzles
with droplet generation rate in the kHz range. Another significant benefit is that
deposition and patterning are in principle achieved in a single step, with virtually no
wasted material, and only a moderate negative environmental impact.
Several significant challenges are encountered in adapting ink-jet printing to
organic electronics:
1) Since inkjet printing is a solvent-based technique, printing multi-layer
structures suffers from limitations of solubility selectivity, similar to spin-coating.
2) Film thickness uniformity is another major concern (Sirringhaus et al. 2000;
Wang et al. 2004), although on a much smaller length scale. The droplets do not
spread evenly across the substrate, forming beads around the edges and
depressions in the middle as they dry. This problem is addressed by tuning the
wetting angle of the ink on the substrate, typically by patterning confining wells
onto the substrate prior to the ink-jet step. (Sirringhaus et al. 2000) Such wells are
often also necessary to limit the size or edge resolution of the deposited droplet,
since it generally spreads considerably upon impact onto the substrate (Toivakka
2003). This, however, requires the additional step of substrate patterning, which
defeats the direct-deposition advantage of the inkjet.
3) As in the case of spun-on polymer films, inkjet printed films need to be baked
to drive off the residual solvent, which may otherwise lead to degradation of the
device when electrical current is turned on. Thus, substrates need to be able to
withstand the moderate (~100-150C) baking temperatures without warping or
degradation.

45
4) Slow evaporation of the solvent during the printing process may clog the
nozzle (Lee et al. 2003), requiring additional provisions in the printing sequence
as well as the nozzle design.
Despite all of the above challenges, inkjet printing has been successfully used by
a number of groups to print circuits and OLEDs, and by a number of manufacturers (e.g.
Epson, Toshiba) to print full-color OLED displays (See Fig. 2-5). Some of the current
problems encountered in manufacturing such displays include low yield, limited lifetime
(on the order of hours), and high cost of manufacture, which can be potentially reduced
substantially when production volume increases (Bardsley 2003).

2.5 Vacuum thermal evaporation
The small-molecular multi-layer structures have been traditionally deposited by vacuum
thermal evaporation, VTE (Fig. 2-6). In this method the source materials are evaporated
Figure 2-5: Example of an ink-jet printed 2.5 diagonal full- color polymer LED
display (Epson).
Figure 2-5: Example of an ink-jet printed 2.5 diagonal full- color polymer LED
display (Epson).

46
from heated container (boats or Knudsen cells) onto the substrate placed directly above.
A quartz crystal microbalance and a mechanical shutter are used to control the thickness
of the individual layers. Depositing in vacuum carries several significant advantages with
respect to solution techniques (Bharathan et al. 1998; Elrod et al. 1989; Forrest 1997;
2004; Garnier et al. 1994; Hebner et al. 1998; Le 1998; Lee et al. 2003; Paul et al. 2003;
Rogers 2001; Sirringhaus et al. 2000; Yang et al. 2000). Firstly, VTE allows for very low
levels of impurity incorporation into the deposited films, provided the deposition
chamber pressure is low (e.g. < 10
-7
Torr) (Forrest 1997). Additionally, since the
evaporation is carried out in vacuum, the molecular mean free path is long (on the order
of 1m), resulting in rectilinear, direct transport of evaporant molecules from the source to
Figure 2-6: (a) Schematic of vacuum thermal evaporation (VTE). The source
materials are evaporated from individually heated source cells onto a substrate
placed directly above. A quartz crystal monitor (QCM) and a mechanical shutter
are used to control the thickness of individual layers. (b) For film patterning and
contact deposition, a shadow- mask is placed in proximity to the substrate. Due to
the long mean free path () of molecules in vacuum and small mask-substrate
separation (s), the pattern edge resolution can be < 1m.
> 30cm
s
b)
> 30cm
s
b)
Heater
To Pump
Substrate
10
-6
Torr
QCM
Mask
Shutter
a)
Dopant Host
Heater
To Pump
Substrate
10
-6
Torr
QCM
Mask
Shutter
a)
Dopant Host

47
the substrate. Finally, another significant advantage of VTE is being able to evaporate
metals. Since none of the previously discussed methods are capable of depositing
atomically flat metallic thin films on top of the organic layers, they all rely on VTE for
metal contact deposition. And although vacuum fabrication technology is generally
regarded as more costly than, for example, inkjet printing, being able to use the same
platform for the deposition of both organics and metals may reduce the overall cost of
processing, as well as integration with current fabrication facilities (Bardsley 2004).
Figure 2-7: (a) A 13 flat panel display made by Sony Corp. using organic LED
technology. (b) Schematic of a picture element (pixel) of the display (shown as a
passive matrix architecture for simplicity) consisting of a triad of red-, green-, and
blue-emitting OLEDs, each typically ~200m wide, with as much as 20m
separation. (c) Approximate sequence of steps in the fabrication of a full-color
display pixel, wherein different dopants (e.g. btp
2
Ir(acac), ppy
2
Ir(acac), FIrpic) are
used to control the OLED emission wavelength. Depositing distinct pixels with
each dopant thus requires a separate masking step, followed by the deposition of a
metal contact (e.g. LiF/Al).
glass / plastic
ITO
SiN
x
organic
metal
200m
0.3m
20m
glass / plastic
ITO
SiN
x
organic
metal
200m
0.3m
20m
btp
2
Ir(acac)
S
N
I r
2
O
O
CH3
CH3
N
Ir
F
F
O
N
O
2
N
Ir
F
F
O
N
O
2
FIrpic
ppy
2
Ir(acac)
N
Ir
O
O
CH3
CH3
2
LiF/Al
1 2 3 4
b)
c)
a)

48
The patterning of active organic layers by VTE is accomplished through the use
of shadow-masks. Again, full-color displays serve as an illustrative example. A
simplified full-color display architecture and VTE fabrication sequence are illustrated in
Fig. 2-7. The basic element of the display is a picture element, or a pixel, which is a triad
of red-, green-, and blue-emitting (RGB) OLEDs, micropatterned in the substrate plane.
The OLEDs are ~200m wide, with up to 20m gaps between them. The deposition of
the entire display takes place in several steps, each separated by a translation of the
shadow-mask to result in the RGB p`attern (Tian et al. 1999):
Step 1: The hole injection/transport layers can be blanket-deposited for all pixels.
Step 2: A mask is placed over the substrate, covering 2/3 of the pixel locations.
The red-emitting doped layer is then deposited.
Step 3: The mask is translated on the substrate to expose an uncovered 1/3 of the
pixel locations; alternatively, the first mask is replaced by a different second
mask, with openings adjacent to the locations of the previous one. The green-
emitting doped layer is deposited.
Step 4: Step 3 is repeated, this time for a doped layer containing the blue-emitter.
Step 5: The mask from Step 4 is removed, and a different mask with all pixel
locations open is used for depositing the metal cathode.
The last step can be avoided if the substrate contains a pre-fabricated integrated
shadow-mask (Tian et al. 1997).
Since the operating voltage of the OLED depends on the total organic layer
thickness, film thickness non-uniformities across the substrate must be kept less than a
few percent. A point-like evaporation source gives a hemispherical evaporative flux,

49
leading to a thickness profile of the form h = h
0
cos
m
, (m2), as illustrated in Fig. 2-8a.
Thus, to keep the variation in h to < 1% across the entire substrate, the solid angle must
be small, and in most cases this means W/H < 0.01, and most of the source material is
wasted on coating of the chamber walls instead of the substrate. Collimated effusion
sources have been employed in molecular beam epitaxy (MBE), but at the expense of
surface coverage. Typical group III-V devices grown by MBE are small-area devices
used in value-added applications, such as telecommunication lasers and photodetectors,
and the highly collimated, low-coverage sources can still be utilized. On the other hand,
since organic electronic devices are mostly aimed at low-cost, large-area applications
(e.g. flat panel displays, ambient lighting and solar cells), this quickly becomes
Figure 2-8: (a) Schematic of point-source vacuum thermal evaporation (VTE).
The distribution of molecular flux from the source leads to varying film thickness
across the substrate. To keep thickness non- uniformity < 1%, typically the W/H <
0.01, resulting in ~99% of the source material coating the chamber walls. Inset:
empty pocket inside the source forming due to a hot spot and localized
evaporation. (b) Shielded and baffled evaporation containers from R.D. Mathis
Inc., designed to proved a steady flux of organic vapor.
(a)
h

H
W
h

H
W
(b)

50
problematic due to the cost of the wasted source materials, and the time and effort in
periodic cleaning of the deposition chambers.
Furthermore, automatic control of the deposition process is difficult due to the
nature of the source material. Organic sources are typically crystalline powders at
ambient conditions and do not liquefy, but tend to sublime when heated at pressures
below 1 atm. Since the thermal conductivity of these powders is very low (Kitaigorodsky
1973; Silinsh et al. 1994), evaporation takes place mainly where the powder contacts the
source container. After some time, the organic material in immediate contact with the
heated container wall evaporates, leaving pocket of empty space. The weight of the
powder above the empty pockets tends to periodically collapse them, causing abrupt,
uncorrelated changes in the effective heat transfer area, and hence a somewhat erratic
evaporation rate. And although on the time scale of the growth of a single device in the
laboratory, the rates may appear sufficiently stable to for reproducible device
characteristics, in the scaled-up production environment the problem is exacerbated by
A B A B
Figure 2-9: (a) Photograph of several simple evaporation sources (or boats) made
by R.D. Mathis Inc. (b) Schematic representation of a linear evaporation system
with adjacent sources for doping. The substrate is translated across the evaporation
flux.
(a) (b)

51
Figure 2-10: (a) A fully integrated, vertical evaporation system utilizing linear
sources to achieve up to 50% materials use efficiency and uniform coating of
substrates up to 400x500mm
2
in size. (b) A close-up photograph of one of the
evaporation chambers. (Applied Films, GMBH, Germany)
(a)
(b)
Figure 2-10: (a) A fully integrated, vertical evaporation system utilizing linear
sources to achieve up to 50% materials use efficiency and uniform coating of
substrates up to 400x500mm
2
in size. (b) A close-up photograph of one of the
evaporation chambers. (Applied Films, GMBH, Germany)
(a)
(b)

52
the requirement of faster evaporation rates, potentially complicating process automation.
Shielded and baffled boats (Fig. 2-8b) can be used to provide a steady organic vapor flux,
but at decreased net evaporation rate, due to the smaller area available for the vapor to
escape.
To improve the film thickness uniformity over large area substrates, as well as
materials use efficiency, a linear source can be used (Fig. 2-9). In this case, the substrate
and the sources are translated relative to each other to provide a uniform film thickness
across the substrate. A commercial-scale, fully integrated evaporation system (Fig. 2-10)
is available from Applied Films GMBH, capable of coating substrates up to 400 mm x
500 mm. Here, to prevent excessive bowing of the substrates and masks under gravity,
the substrate and the sources are vertically oriented. No details of construction were
available at the time of writing; one of the anticipated technological challenges is
retention of the source material, typically a crystalline powder, in the vertically oriented
evaporation cell, as well as the control of doping, due to potentially poor thermal contact
between the heated evaporation container wall and the powder source in the vertical
arrangement.
2.6 Organic vapor phase deposition
Virtually all of the organic materials used in the thin film devices described thus far have
sufficiently high vapor pressures to be evaporated at temperatures below 400C and then
be transported in the vapor phase by a carrier gas such as argon or nitrogen. This allows
for positioning of evaporation sources outside of the reactor tube (similar to chemical
vapor deposition (CVD)), spatially separating the functions of evaporation and transport,

53
thus leading to more precise control of the deposition process. (Olsen 1982; Stringfellow
1989; Wolf et al. 1999)
Figure 2-11 illustrates the OVPD concept. Here, Alq
3
, which is a crystalline
powder in its purified state at ambient conditions, may be sublimated by heating at a
reduced pressure. The vapors are picked up by an inert carrier gas such as dry nitrogen,
and are transported to a cold substrate. A temperature gradient arises across the
hydrodynamic boundary layer near the substrate surface. When the organic vapor reaches
a condition of supersaturation while diffusing through the cool boundary layer, the
relatively heavy organic molecules nucleate into a solid film on the substrate surface.
Undesirable condensation can be avoided by keeping the walls of the chamber above the
solid-gas equilibrium temperature of the organic, while directing the flow patterns, to
selectively coat the substrate. Using a high purity carrier gas, the incorporation of
impurities into the film is minimized.
Figure 2-11: Schematic of the Organic Vapor Transport Deposition concept. A
hot inert carrier gas (e.g. N
2
) transports organic vapors through a hot-wall reactor
toward a cooled substrate, where the vapor selectively physisorbs.

54
The transport mechanisms, capabilities and commercial implementation of OVPD
will be discussed in more detail in the following chapters. Some of the advantages
include:
1). Selective deposition on the cooled substrate, resulting in > 50% materials use
efficiency;
2). Flow-controlled distribution of the source vapor, enabling uniform coating of
large area substrates by means of appropriately designed gas distributors
3). Precise regulation of the vapor delivery and doping concentration via the flow
rate of a carrier gas saturated with the vapor;
4). Thermally equilibrated source cells by utilizing a pre-heated carrier gas
flowing through the source material;
5). Self-cleaning hot-wall deposition chamber.
The commercially available, fully automated OVPD tool from Aixtron AG (Fig. 2-12)
takes advantage of the above properties of vapor-assisted deposition, enabling the
deposition of high performance OLEDs on large area substrates, with device lifetimes
matching or exceeding those of vacuum-deposited analogs. (Brown 2004)
At the same time, OVPD must overcome several challenges, including:
1). Having to use shadow-masks to pattern the organic layers, where 60% or more
of the source material impinging on the substrate is wasted on coating of the
shadow-mask. Techniques like ink-jet printing have a significant advantage by
virtue of their direct-patterning approach.

55

N
2
N
2
N
2
N
2
T
1
T
2
T
3
Load Port
(connect to
vac. xfer li ne)
Deposition
Chamber
Mixing Unit
Figure 2-12: (a) A mode of OVPD where the carrier gas flows through the source
cells for better vapor pick-up, and onto the substrate through a large-area
distributor for improved film thickness uniformity. (b) Schematic of a
commercial- grade, fully computer-controlled OVPD system available from
Aixtron AG, with multiple-evaporation cell furnaces and a remote deposition
chamber.
(b)
(a)
N
2
N
2
N
2
N
2
T
1
T
2
T
3
N
2
N
2
N
2
N
2
T
1
T
2
T
3
Load Port
(connect to
vac. xfer li ne)
Deposition
Chamber
Mixing Unit
Load Port
(connect to
vac. xfer li ne)
Deposition
Chamber
Mixing Unit
Figure 2-12: (a) A mode of OVPD where the carrier gas flows through the source
cells for better vapor pick-up, and onto the substrate through a large-area
distributor for improved film thickness uniformity. (b) Schematic of a
commercial- grade, fully computer-controlled OVPD system available from
Aixtron AG, with multiple-evaporation cell furnaces and a remote deposition
chamber.
(b)
(a)

56
2). A greater thermal budget relative to VTE, required in continuously heating the
source cells, the chamber walls, and the carrier gas. This also results in additional
heat loads on the substrate, requiring powerful on-board chillers.
3). The current inability of OVPD to deposit atomically flat metal contacts using
carrier gas transport, requiring integration with VTE.
2.7 Organic vapor jet printing
In Ch. 6 organic vapor jet printing (OVJP) is described for the direct patterned deposition
of small molecular organics. This method, illustrated in Fig. 2-9, is at first glance
somewhat of a hybrid between inkjet printing and OVPD. In OVJP, organic molecules
are sublimed into a hot inert carrier gas, which is then expanded through a microscopic
nozzle (or an array of nozzles) forming a highly collimated gas jet. The jet impinges on a
cooled substrate, with organic molecules selectively adsorbing onto the substrate,
forming a well-defined thin film deposit. The deposited pattern size and edge resolution
depend on the nozzle diameter, nozzle-to-substrate separation, and the downstream
ambient pressure.
However, OVJP is distinct from ink jet printing of solution processed polymer
organic semiconductors (Hebner et al. 1998; Paul et al. 2003; Sirringhaus et al. 2000) in
that it uses a hot inert carrier gas, instead of a liquid solvent, to directly print molecular
organic semiconductors. This eliminates meniscus formation, solvent incompatibility,
and other issues limiting inkjet printing. Furthermore, in OVJP no pre-patterning of the
substrate is needed to contain the liquid droplet, whereas in ink jet printing, droplet-
confining wells are required. (Sirringhaus et al. 2001) Unlike OVPD and VTE, OVJP
does not require shadow masks to pattern the organic thin films. A significant

57
disadvantage remains in that atomically flat metal films still cannot be deposited on top
of other organic layers using a carrier transport technique. Despite this, it is anticipated
that OVPD and especially OVJP will prove powerful new methods for highly controlled,
rapid and low-cost organic device growth, especially as it is applied to molecular organic
compounds.


mixing
chamber
N
2
N
2
N
2
M
Al q
~ 500 g/mol
M
N2
~ 28 g/mol
Figure 2-13: Schematic of the organic vapor jet printing (OVJP) concept, where a
hot inert carrier gas picks up organic vapor and ejects it through a microscopic
collimating nozzle, placed in proximity of a cooled substrate. The heavier organic
molecules physisorb onto the substrate surface, forming a well-defined deposit,
whose size and edge dispersion are directly influenced by the deposition
conditions (pressure, temperature, carrier gas molecular mass, distance from the
substrate), and the shape of the nozzle.

58
Chapter 3: Organic Vapor Phase
Deposition
3.1 Overview
In this chapter the method of organic vapor phase deposition (OVPD) is analyzed in
greater detail. Material transport regimes are outlined, and a theory is developed that
allows the prediction of source material evaporation and deposition rates. Formulae
describing the control of doping concentration by source temperature and carrier gas flow
rate are also derived. The theory developed here provides a practical understanding of the
basic mechanisms of OVPD, guiding the design and operation of experimental as well as
commercial-scale OVPD systems.
3.2 OVPD Concept
Vacuum thermal evaporation offers some advantages in the deposition of organic thin
films for device applications, including the ability to preserve source material purity, the
ability to deposit films with monolayer precision, and in some cases up to 50% materials
utilization efficiency. At the same time, some characteristics inherent to VTE make
commercial-scale deposition of organic electronics difficult. Organic vapor phase
deposition (OVPD) was developed in part to overcome the limitations of vacuum thermal
evaporation. It decouples the evaporation and deposition events, using a carrier gas to

59
mediate transport of material from the source to the substrate, thereby achieving greater
control of the deposition process.
The concept of OVPD is illustrated in Fig. 3-1. The source material is heated to
generate molecular vapor, which is picked up and transported by a hot inert carrier gas
toward a cooled substrate, where the organic vapor selectively condenses. Parasitic
condensation of the source material is avoided by cooling the substrate only, while
actively heating the walls of the deposition system. To grow doped films, multiple
component streams are mixed en route to the substrate. The process is carried out at
reduced pressure, increasing gas diffusivity and thereby improving mass transfer rates.
(Bird et al. 1996; Wolf et al. 1999) It is helpful to analyze OVPD as a succession of 3
steps: evaporation and pick-up, vapor transport by convection, and deposition. (Shtein et
Figure 3-1: Organic vapor phase deposition (OVPD) concept. A hot inert carrier
gas (e.g. N
2
) transports hot organic source vapor along a heated-wall reactor
toward a cooled substrate, on which the organic molecules selectively physisorb.

60
al. 2001) Growth of polycrystalline thin films involves a simultaneous annealing of the
growing film by molecular surface diffusion at elevated substrate temperature. (Shtein et
al. 2002)
3.3 Theory Evaporation
The source cell is schematically shown in Fig. 3-2. Organic source material, typically in
the form of a crystalline powder, is heated inside of a container, while a hot inert carrier
gas flows over the organic surface. We are interested in predicting the rate of organic
vapor supply based on the source conditions, such as temperature, pressure, carrier flow
rate, and evaporation surface area.
Qualitatively, for a fixed cell temperature, a very low carrier flow rate (V

), will
result in a vapor-saturated (i.e. vapor-solid equilibrated) cell atmosphere; here, carrier
flow limits how much organic material escapes the cell, which is also said to be in the
Figure 3-2: Schematic of a source cell containing the organic source material,
typically a crystalline powder, maintained at P and T
cell
. The carrier gas enters the
cell picking up the organic vapor, carrying it out at a molar rate r
out
, which is the
balance between evaporation and re-condensation rates, r
evap
and r
cond
, inside the
cell.
V

Solid / Powder
T
cell
P
r
evap
r
cond
carrier gas
inflow
carrier gas
+ vapor
r
out
V

Solid / Powder
T
cell
P
r
evap
r
cond
carrier gas
inflow
carrier gas
+ vapor
r
out

61
flow-controlled (or equilibrium) operating regime. Increasing V

will cause organic


molecules to be swept out of the source cell as soon as they evaporate, thereby decreasing
the organic vapor pressure inside the cell and shifting it away from solid-gas equilibrium.
The cell thus enters a kinetic, or evaporation-controlled regime. These two regimes are
depicted in Fig. 3-3a. The same transition can be accomplished by fixing the carrier flow
rate and changing the cell temperature, T
cell
, as shown in Fig. 3-3b. For low T
cell
,
evaporation rate is low, and the cell is not saturated; it is in the evaporation-controlled
Figure 3-3: a), b) Qualitative depiction of the source cell operating regimes. c)
The resulting rate of organic vapor outflow vs. increasing carrier gas flow rate for
a series of cell temperatures T1, T2, T3.
Evaporation-Controlled
(Kinetic regime)
Flow-controlled
(Equilibrium regime)
Constant T
cell

V
org
eq
P
P
a)
Evaporation-Controlled
(Kinetic regime)
Flow-controlled
(Equilibrium regime)
Constant T
cell

V
org
eq
P
P
a)
Evaporation-Controlled
(Kinetic regime)
Flow-controlled
(Equilibrium regime)
Constant

V
T
cell
org
eq
P
P
b)
Evaporation-Controlled
(Kinetic regime)
Flow-controlled
(Equilibrium regime)
Constant

V
T
cell
org
eq
P
P
b)
Kinetic
Equil.
c)
Kinetic
Equil.
c)

62
(kinetic) regime. Increasing T
cell
leads to vapor saturation, i.e. P
org
/P
eq
1, where P
org
is
the actual vapor pressure of the organic material, while P
eq
is the equilibrium vapor
pressure; here the system is in the flow-controlled (equilibrium) regime. However, in
practice we are concerned with how much organic material escapes the source cell. And
although for higher V

, the organic vapor pressure inside the cell, P


org
, drops, the
molecules are still being swept out of the container. Thus, for a given T
cell
, increasing V


will result in a monotonic increase in the rate of organic vapor outflow. This situation is
depicted qualitatively in Fig. 3-3c.
A quantitative analysis of the source cell begins with a mass balance on the
organic species:
=
evap cond out
r r r (3-1)
where r
evap
and r
cond
are the evaporation and condensation rates, respectively, while r
out
is
the molar rate of organic vapor outflow from the evaporation cell. The condensation rate
is given by the molecular collision rate with the surface and the molecular sticking
probability, :
1
2
=

cond org e
org B
r P A
M k T

(3-2)
where A
e
is the effective surface area of the organic material, M
org
is its molecular mass,
(2M
org
k
B
T)
-1/2
is the surface collision frequency, k
B
is the Boltzmann constant, and T is
the temperature. Similarly, the evaporation rate is:
1
2
=

evap eq e
org B
r P A
M k T
. (3-3)
At equilibrium P
org
= P
eq
, the evaporation and condensation rates balance, giving:

63
=
org eq
P P . (3-4)
The net surface evaporation rate is thus given by:
( )
1
2
=

out eq org e
org B
r P P A
M k T

, (3-5)
The Clausius-Clapeyron equation states that:
0
exp
| |
=
|
\ .
vap
eq
cell
H
P P
RT
(3-6)
where H
vap
is the vaporization enthalpy of the organic material, P
0
varies slowly with
T
cell
, and the ideal gas constant R = k
B
6.0210
23
JK
-1
mol
-1
. The organic vapor pressure
used in this work is typically < 10
-2
Torr (1 Pa), and the carrier gas thus comprises > 99%
of the species present in the deposition chamber. If the vapor and the carrier gas mix
perfectly inside the source cell, the molar outflow of organic material is:
org
out
cell
P
r V
RT

= (3-7)
where V

is the volumetric flow rate of the carrier gas through the source cell. Equations
(3-1) (3-7) combine to give:
0
exp
2
vap
cell
org
org B cell
cell e
H
P
RT
P
M k T
V
RT A

| |
|
\ .
=

+

(3-8)
0
exp
2
vap
i
cell
out
org B cell
cell
e
H
P
RT
r
M k T
RT
A
V



`
)
=

+

(3-9)

64
Since in practice the carrier gas flow in the deposition system is regulated by mass flow
control devices, it will be more convenient to analyze data in terms of the mass flow rate
units, (e.g. standard cubic centimeters per minute, or sccm). The actual volumetric flow
rate of the carrier gas, V

, can be expressed as:


cell std
sccm
std cell
T P
V V
T P

= (3-10)
where sccm V

is the measured flow rate in sccm, while T


std
and P
std
are the standard
temperature and pressure, 298K and 760 Torr, respectively. Equation (3-9) becomes:
0
exp
2


`
)
=

+

vap
cell
out
i B cell std cell
org std
sccm
H
P
RT
r
M k T RT P
A P
V

(3-11)
Figure 3-3c in fact shows the same qualitative dependence of r
out
on the three process
variables, T
cell
and sccm V

, as predicted by Eq. (3-11), assuming an ideal situation where


they do not influence . For constant sccm V

and P
cell
, raising T
cell
will shift the source
closer to the solid-vapor equilibrium (i.e. saturation regime), while at constant T
cell
the
source becomes depleted for higher sccm V

. Equation (3-11) also shows that lower source


cell pressure helps maximize the rate of vapor supply. The cell-loading and material-
specific variables such as H
vap
, , and A
e
, would be determined experimentally; this will
be addressed in Ch. 4.
3.4 Vapor transport

65
Next, we analyze Step 2 in the OVPD mechanism, the transport of vapor from the source
to the substrate. The vapor from individual source cells may be further diluted and
blended with the vapor from other sources. Dilution may be necessary for several
reasons, including prevention of vapor supersaturation en route to the substrate, the need
to establish developed flow and heat transfer in the transport lines, and prevent cross-
contamination between sources. Assuming that the vapor streams become well-mixed en
route to the substrate, the resulting molar vapor concentration, c, is given by:

=
out
tot
r
c
V
(3-12)
where tot V

is the total flow rate of the carrier gas which enters the deposition region.
When two source materials, A and B, are blended,

= + + tot A B dil V V V V , where

dil V is the
dilution flow rate. The resulting concentration of the source vapor of material i in the
deposition chamber is given by:
0,
, ,
,
,
exp
2


`
)
=
(

(
( + +
(

(


vap
i
i
cell
i
i B cell i cell i
std
i dil
i org i std
i sccm
H
P
RT
c
M k T P
RT
V V
A P
V

(3-13)
As will be shown in the following sections, the film deposition rate is limited by diffusion
across a boundary layer. Thus, the final concentration of component i in the film will
scale with the vapor concentration given by Eq. (3-13).
Both the boundary layer thickness and the mixing distance depend on the
characteristics of the flow, including velocity, gas species, temperature, pressure, and
whether the flow is laminar or turbulent. The transition between laminar and turbulent

66
flow typically occurs at a Reynolds number of 2,300 for pipe flow, or on the order of 10
5

for flow past a flat plate. The Reynolds number is:
Re

=
m f
f
U d

(3-14)
where U is the linear flow velocity, d
m
is the diameter of the deposition chamber,
f
and

f
are the density and viscosity of the fluid, respectively. Since the organic vapor is a
minority species in the flow, it can be neglected in calculating Re. For N
2
at 270C and 1
Torr, typical OVPD deposition conditions,
f
= 3.6410
-2
kg/m
3
and
f
= 2.4410
-5
kg/ms.
As will be discussed in Ch. 4, the experimental OVPD system consisted of several
components of varying diameter, ranging from 5mm barrel outlets, a substrate 2cm in
diameter, and a 10 cm diameter glass tube. Carrier gas volumetric flow rate was a
maximum of 100 sccm. Converting to deposition temperature and pressure and using the
largest diameter (10cm) this results in U = 0.32 m/s, and Re = 48, well within the laminar
regime for both flow within the deposition chamber and past the substrate. Since at
typical deposition conditions

tot V <30 sccm, the laminar regime assumption holds


throughout.
Given the laminar nature of the flow, the mixing of the separate source vapor
streams in a simple geometry such as depicted in Fig. 3-4 is dominated by diffusion. The
mixing will be complete and uniform, when the vapors diffuse radially to the point of
filling the entire flow diameter, which occurs in characteristic time
D
:
2
6
=
m
D
i
d
D
(3-15)

67
where D
i
is the gas phase diffusivity of organic vapor of material i. During this time, the
gas molecules entering the main tube will travel a distance L
m
:
2
6

= =
m
m D
i
u d
L U
D
. (3-16)
To first order, using a hard-sphere approximation D
i
is given by the gas kinetic theory:
(Bird et al. 1996)
r
d
L
m
d
m

A
V

B
V
Figure 3-4: Schematic of the source vapors mixing en route to the substrate by
diffusion, provided the flow is laminar, which holds true for typical OVPD
conditions.
Coolant
Source A
Source B
Carrier gas
Pump
Pump
(a)
Substrate
Heater
(b)

68
2
1 1 8
3 3 2

= =

B B
i
dep
k T k T
D u
M P


(3-17)
where M is the reduced molecular weight (M = M
A
M
B
/(M
A
+M
B
), where A and B denote
two different gas species), and are the molecular mean thermal velocity and mean
free path, respectively, is the average molecular diameter ((
A
+
B
)/2), P
dep
is the total
deposition chamber pressure, and T is the gas temperature. For Alq
3
diffusing in N
2
, M
A
=
459 g/mol, and M
B
= 28 g/mol, giving D
Alq
57 cm
2
/s, and L
m
= 9 cm, or L
m
d
m
in the
experimental OVPD system. The reactor designed specifically to take this mixing
mechanism into account; it will be also shown in Ch. 4 that the experimentally observed
mixing distance corresponds to the calculated L
m
.
The carrier gas flow field can be visualized by numerically solving the complete
Navier-Stokes equations for the deposition geometry. The momentum balance is given
by:
( )
2
f f f f
u
u u p u g
t

+ = + +

G
G G G G
(3-18)
where g
G
is the gravitational acceleration, which can become important in accounting for
buoyancy effects inside the chamber at very low flow rates. The continuity equation
0 u =
G
is also included. To obtain the temperature profile of the system, the heat
balance is included:
( )
f f
f P f P
T
c k T c T u Q
t

+ + =

G
, (3-19)
where k is the thermal conductivity, C
pf
is the isobaric heat capacity, and Q is the net heat
flow to the cooled substrate holder.

69
The simultaneous numerical solution of the Navier-Stokes, diffusion and heat
conduction partial differential equations was performed using FEMLab v2.3, using the
damped Newton method. (Deuflhard 1974) Special precautions were taken to ensure
convergence of the non-linear problem. The solver was damped to guard against infinite
Newton iterations. The simulation geometry employed an adaptive mesh, where the
largest grid dimension did not exceed 1/5 of the reactor diameter. No appreciable changes
in the process parameters (deposition efficiency, flowlines, temperature profiles, etc.)
were observed when refining the grid.
For the simulation results shown in Fig. 3-5, the conditions included (in MKS
units) = 1.73e-5(T/300)
1/2
, C
p
=1000 J/K/kg, calculated from ideal gas law at the local
temperature and pressure, k (thermal conductance) = 0.025(T/300)
1/2
W/K/m, D = 0.001
m
2
/s. In Fig. 3-5, a series of flow fields (white arrows) and a color-coded temperature
map are shown for a simplified OVPD geometry. Nitrogen carrier gas enters the
deposition chamber at room temperature through a single, axially positioned barrel and
exits through the pump port positioned downstream. The walls of the main chamber are
maintained at 500 K, while the substrate, situated downstream of the inlet, is maintained
at 290 K. For very low flow rates (Re = 0.05), the carrier gas quickly thermally
equilibrates with the chamber walls and cools rapidly upon contact with the substrate.
Hydrodynamic and thermal boundary layers appear to form in front of the substrate. As
the flow velocity (i.e. Reynolds number) increases, a greater distance is required to heat
the carrier gas. At the same time, the boundary layer becomes more pronounced, while
shrinking in thickness.

70

Re = 0.05
Re = 5
T
T = 500K T = 290K
N
2
inlet
To pump
T = 290K
Figure 3-5: Modeling the flow field and temperature distribution in an OVPD
system using numerical solutions to the Navier-Stokes equations. Unheated N2 is
the carrier gas, entering the deposition chamber requires longer distances from the
inlet to thermally equilibrate for higher Reynolds numbers (i.e. flow velocity). As
expected, the hydrodynamic boundary layer decreases in thickness for higher
velocity flows.
Re = 0.15
Re = 0. 5
Re = 1. 5

71
The organic species is introduced via the mass balance:
( )
i
i i i D
c
D c c u Q
t

+ + =

G
(3-20)
where c
i
is the organic species concentration, and Q
D
is the net deposition rate. The
numerical solutions to Eqs.(3-18) (3-20), along with a set of realistic deposition
conditions (T
wall
= 600 K, T
sub
= 300 K, Re = 10) are plotted in Fig. 3-6. The distribution
of species follows the flow, establishing a concentration boundary layer near the
substrate. The close correspondence between the hydrodynamic, thermal, and the
concentration boundary layers is not surprising, considering the dilute gas conditions.
The deposition of organic molecules on the substrate will proceed by diffusion across this
boundary layer.
3.5 Deposition
As the vapor and carrier gas flow over the cold substrate, a stagnant boundary layer of
thickness develops above the surface (see Figs. 3-1 and 3-6), across which the organic
vapor must diffuse. Since P
eq
depends exponentially on temperature, organic molecules
condense on the cold substrate, giving rise to a concentration driving force across the
boundary layer. The flux of organic molecules to the substrate, j
org
, obeys Ficks law of
diffusion:
gas s
org org
s
org org org org
P P
RT RT
j D c D

= =

(3-21)
where D
org
is the gas phase diffusivity of the organic in the carrier gas, P
org
gas
and P
org
s

are the organic species vapor pressures in the gas phase and at the substrate surface,

72
Figure 3-6: a) Results of numerical solution of the NavierStokes equations for
a simplified OVPD system geometry, where carrier gas is introduced in the
barrel at a Re=10; the flow field is indicated by the streamlines, while the
temperature distribution is shown by the color- map. The flow of the gas around
the substrate creates a stagnant boundary layer, evidenced by the flow field. b)
Concentration distribution color- map of the organic species carried by N
2
. The
concentration boundary layer near the substrate follows the hydrodynamic
boundary layer from (a).
(a)
(b)

73
respectively, evaluated at the deposition chamber and surface temperatures, T and T
s
,
respectively. For low surface temperatures, P
org
s
<< P
org
gas
, Eq. (3-13) thus becomes:
=

org
org org
c
j D (3-22)
where c
org
is the organic vapor concentration in the deposition chamber. Combining Eqs.
(3-13) to (3-22) for the deposition of an organic compound i:
0,
,
, ,
,
,
exp
2


`
)
= =
(

(
( + +
(

(


vap
i
i
dep i cell i
i dep
sub
i cell i cell i
std
i dil
i org i std
i sccm
H
P
r RT D
j
A
mw RT P
RT
V V
A P
V
(3-23)
where r
dep,i
is the deposition rate, A
sub
is the substrate area,
dep
is the deposition
efficiency, which depends on the gas distributor design and the resulting flow pattern
around the substrate. To predict the deposition rate for a given reactor geometry, the flow
equations must be solved to provide information about , while more precise values for
D
i
can be either obtained experimentally, or estimated from correlations such as the
Chapman-Enskog equation (Bird et al. 1996).
It should be noted that if U, P
dep
, and T remain constant, then does not change.
However, since typically varies as U
-1/2
(Bird et al. 1996; Schlichting 1968), and the
boundary layer is expected to thin if U (or

tot V ) is increased. Clearly if the vapor


concentration is maintained, the decrease in can result in an increased deposition rate.
If all flow conditions in the deposition chamber are kept constant, the changes in
r
dep
mirror changes in r
evap
. Figure 3-7 shows data from an experiment where

sccm V was
increased, while

tot V , P
cell
, and T
cell
were kept constant. The deposition rate behaves

74
analogously to the evaporation rate, as depicted in Fig. 3-3c. The plot of 1/r
dep
vs. 1/

sccm V
is linear, as predicted by Eq. (3-15), and can be used to estimate various parameters for
each material source configuration, such as A
e
and the equilibrium vapor pressure.



Figure 3-7: Experimentally determined deposition rate of Alq
3
on Si, matching
the qualitative prediction in Fig. 3-4. Equation 3.11 is confirmed by plotting
1/r
dep
vs. the inverse of the carrier gas flow rate through the source (1/ V) cell,
while keeping the total flow rate constant by means of the make-up N
2
flow.
Figure 3-7: Experimentally determined deposition rate of Alq
3
on Si, matching
the qualitative prediction in Fig. 3-4. Equation 3.11 is confirmed by plotting
1/r
dep
vs. the inverse of the carrier gas flow rate through the source (1/ V) cell,
while keeping the total flow rate constant by means of the make-up N
2
flow.

75
Chapter 4: Proof of Concept &
Experimental Verification of Theory
4.1 Overview
The basic experimental set-up is described in this chapter, along with results validating
the concept, including growth of OLEDs, roll-to-roll deposition of organic thin films, and
precise molecular doping. Data and computer simulations are presented supporting the
theory developed in Ch. 3.
4.2 Experimental design
As Eq. (3-23) suggests, full control of the deposition process requires ability to control
tot V

, cell V

, P
dep
, P
cell
, T
s
, T
cell
. Figure 4-1a is a schematic of the deposition system used
for most of the OVPD experiments discussed here. The main vessel consists of an 11 cm
diameter by 150 cm long Pyrex

cylinder, with flanges on each end for attachment of the


source baseplate and substrate. The center of the main chamber is positioned inside of a
three-zone tube furnace used to establish a temperature gradient along the chamber axis
(See Fig. 4-1b). Each source material resides inside a round glass container, attached to
the end of a 6mm diameter hollow glass pushrod. The round container is closed at the
back to prevent sudden dislodging of the source powder upon pump-down and sudden

76
increase in the carrier gas flow. Each source cell is further separately encased in a 2.5 cm
diameter by 75 cm long glass barrel. The pushrod enters the barrel through an
UltraTorr type Viton o-ring seal, enabling it to be moved in and out of the furnace.
The desired organic source temperature is attained by appropriate positioning of each cell
along the temperature gradient within the reactor (Fig. 4-1b). By this means, two or more
Thickness Monitor Thermocouples
N
2
Inlets Shutter
Furnace
Pump Port
Glass tube
Source barrels
Substrate
holder
Sliding
mount
Thickness Monitor Thermocouples
N
2
Inlets Shutter
Furnace
Pump Port
Glass tube
Source barrels
Substrate
holder
Sliding
mount
Figure 4-1: a) Schematic of the OVPD system described in detail in the text. b)
Measured axial temperature profile in the system, showing how the temperature of
each source may be controlled by its axial position. The temperature profile is
regulated by the three zone temperature settings of the furnace, as shown for two
Zone 3 settings.
0
50
100
150
200
250
300
-10 -5 0 5 10 15 20 25 30 35 40 45
Dopant
Host
Temp
(
o
C)
Distance (inches)
T
3
low
Cooling
water
Substrate
0
50
100
150
200
250
300
-10 -5 0 5 10 15 20 25 30 35 40 45
Dopant
Host
Temp
(
o
C)
Distance (inches)
T
3
low
Cooling
water
Substrate
(a)
(b)
T
3
high

77
compounds with different vapor pressures can be simultaneously evaporated at
comparable rates. Each barrel terminates in a 10 cm long by 0.6 cm diameter snorkel; this
increases the carrier exit velocity, preventing back-diffusion of organic vapors into the
sources.
Carrier gas flow rates are regulated by mass flow controllers, while a combination
of a 40 lpm scroll pump and a butterfly valve was used to regulate the reactor pressure
between 0.025 and 760 Torr. A liquid nitrogen cold trap is also employed to prevent
contamination of the pump oil with residual organic vapor. Evaporation temperature of
each source is measured by thermocouples positioned behind each source boat, while the
temperature profile of the deposition chamber is obtained using an axially positioned
thermocouple probe. Organic vapors condense onto a rotating water-cooled substrate
positioned behind a mechanically operated shutter. Film thickness and growth rate are
monitored by a quartz crystal microbalance. The tooling factor (i.e. the ratio of actual
deposition rate on the substrate to the rate measured by the balance) is calibrated by
measuring the deposited film thickness with an ellipsometer. The real OVPD system is
pictured as constructed in Fig. 4-2.
4.3 Growth of Alq
3
films
The system is first characterized by calibrating the deposition rate for each source for a
set of process conditions. The deposition rate is first determined by depositing a film (e.g.
Alq
3
) on silicon, while monitoring the rate with a quartz crystal microbalance. The actual
deposited film thickness is measured by ellipsometry, providing a tooling factor for a
given set of process conditions, which can be used for deposition on other substrates (e.g.

78
Figure 4-2: Photograph of the OVPD system, including the clam-shell furnace,
open to reveal the 2m long by 10cm diameter glass deposition chamber. Upstream,
a flange is fitted holding four source barrels with individual N2 flow inlets, while
the cooled rotating substrate holder is positioned downstream (right side of photo),
opening into a plexiglass enclosure by means of a sliding bearing mount. Below
the enclosure is a chiller used to regulate the substrate temperature. The four
frames below illustrate more closely some of the key components of the system.
Figure 4-2: Photograph of the OVPD system, including the clam-shell furnace,
open to reveal the 2m long by 10cm diameter glass deposition chamber. Upstream,
a flange is fitted holding four source barrels with individual N2 flow inlets, while
the cooled rotating substrate holder is positioned downstream (right side of photo),
opening into a plexiglass enclosure by means of a sliding bearing mount. Below
the enclosure is a chiller used to regulate the substrate temperature. The four
frames below illustrate more closely some of the key components of the system.

79
ITO-coated glass) on which ellipsometry is difficult. First, the variation of single-
component (in this case, Alq
3
) deposition rate with carrier gas flow rate is investigated.
For the reactor and source cell configuration used, the source operates mainly in
the kinetically limited regime, where the flux of organic vapor into the deposition
chamber is limited by the evaporation rate. Keeping all process conditions constant
except the carrier gas flow rate through the source barrel, the vapor will become diluted
with increasing carrier gas flow rate. Thus, since the film deposition rate is diffusion
limited, Eq. (3-23) simplifies to:
1
2
exp
1
1

| |

|
\ .
=

+
vap
cell
dep
H
k
RT
r
k V
(4-1)
where k
1
and k
2
are constants. Eq. (4-1) can be rewritten as:
3 4
1
exp
vap
dep cell
H
k k V
r RT

| |
| |
= +
| |
\ .
\ .
(4-2)
where k
3
and k
4
are constants which depend on the loading of the source, diffusivity,
boundary layer thickness, and the relative orientation of the source barrel and substrate.
Figure 4-3 shows the dependence of the Alq
3
deposition rate on the carrier gas
flow rate through the Alq
3
source barrel for three different values of T
cell
. The plot
confirms that for these deposition conditions, the organic vapor is apparently diluted by
the carrier gas, leading to the decrease in r
dep
with

V ; this implies that the source is in the


kinetic evaporation regime. The linear fits of the data were performed using H =
162kJ/mol, obtaining k
3
= [1.72 +0.5/-0.32]10
-16
sA
-1
, and k
4
= [6.9 +3.1/-3.58]10-18
sA
-1
sccm
-1
. Similar experiments were performed for several other compounds, including
-NPD, which is frequently used as the hole-transporting material in OLEDs. Subsequent

80
growth of OLEDs in OVPD made use of the deposition conditions determined in the
manner described above, but with in-situ monitoring in place, due to the potential
variations in the source material condition with time or pressure (i.e. k
3
and k
4
).
4.4 Evaporation rate and decomposition temperature
As Eq.(4-2) indicates, the evaporation and hence the deposition rates scale exponentially
with the source temperature, T
cell
. This means that using the same source temperature for
materials having very different characteristic vapor pressures can result in dramatically
different deposition rates. Thus, we are also interested in predicting the evaporation
temperature range for a variety of organic semiconductors. In addition to a general
Figure 4-3: Dependence of the Alq
3
deposition rate (on silicon) on volumetric
carrier gas flow rate for three different source temperatures, T
cell
= 269, 276,
278C.
0 5 10 15 20 25 30 35 40 45 50 55
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
276 C
269 C
278 C


1
/
r
d
e
p

(
1
/
A
)
Barrel Flow (sccm)
0 10 20 30 40 50
0.0
0.5
1.0
1.5
2.0


r
d
e
p

(
A
/
s
)
Barrel Flow (sccm)

81
practical importance, this will also set the temperature gradient (See Fig. 4-1b) to be used
in the OVPD system described above. For convenience, thermogravimetric analysis
(TGA) was used.
1.2 1.3 1.4 1.5 1.6
0.01
0.1
1

H
v
a
p
=
1
7
6

k
J
/
m
o
l
H
v a p
= 1 2 8 k J / m o l


H

v

a

p

=

1

6

6


k

J

/

m

o

l


H

v

a

p

=

1

3

9

k

J

/

m

o

l


H

v

a

p

=

1

6

2


k

J

/

m

o

l
C u P c
I r ( p p y )
P t O E P
- N P D
A l q
3


E

v

a

p

o

r

a

t

i

o

n



R

a

t

e


(

m

g

/

m

i

n

)
1 0 0 0 / T
cell
( K
- 1
)
Figure 4-4: a) Schematic of the thermogravimetric analysis (TGA) apparatus,
where the sample is weighed while being heated in a stream of N
2
or other inert
gas. b) Plot of the evaporation rate versus inverse temperature for several
compounds typically used in OLEDs. The slope of the curves is H
vap
/RT
cell
,
where H
vap
is the enthalpy of vaporization specific to each compound.
(a)
(b)

82
The TGA technique is illustrated in Fig.4-4a; material is placed into a pan and is
weighed while being heated in a stream of N
2
. The mass of the sample and its
temperature are recorded as function of time. Fig.4-4b plots r
evap
vs. 1/T for several
commonly used OLED materials. Here, platinum octaethyl porphyrine (PtOEP), fac-
tris(2-phenylpyridine) iridium (Ir(ppy)
3
) are dopants, Alq
3
and -NPD are the electron
and hole transport materials, respectively, and copper phthalocyanine (CuPc) is the hole
injecting material. Readily apparent is the grouping of materials according to their
evaporation rates and temperature ranges. If a single chamber is to be used to deposit
CuPc, Alq
3
and -NPD, the walls must be kept hot enough to prevent condensation of
CuPc a temperature much higher than the minimum required to prevent parasitic
condensation of Alq
3
and -NPD on the walls.
Table II lists the enthalpies of evaporation of each material, including those of
DCM2 and 4,4-N,N-dicarbazolebiphenyl (CBP), calculated from the thermogravimetry
and OVPD data. Previous determinations of the enthalpy of vaporization of Alq
3
and
CuPc using thermogravimetry and mass spectroscopy (Yase et al. 1995; Yase et al. 1996)
agree with values found here. In addition, the enthalpy of vaporization determined by
OVPD of Alq
3
yields H
vap
= (16276) kJ/mol, also in good agreement with the TGA
results. Thermogravimetry of DCM2 above 250C resulted in decomposition and
formation of non-volatile products, so the TGA results were not representative of DCM2
evaporation. On the other hand, OVPD of DCM2 at temperatures below 230C yielded
H
vap
= (11011) kJ/mol, as shown in Fig. 4-5.

83

4.5 Thermal decomposition of source materials
High rate of device fabrication (e.g. on the scale of m
2
/s) requires a correspondingly high
evaporation rate; r
evap
= Cexp(-H
vap
/RT), where C is a constant that depends on factors
like the surface area of the source material, geometry and flow conditions in the source
container. The evaporation rate increases rapidly with temperature, T, and decreases with
the evaporation enthalpy, H
vap
which is specific to each material, as shown in Sec. 4.2.
a
Measured using OVPD data
b
Doping into Alq
3
c
Thermogravimetry by Yase et al.
d
Mass spectrometry by Yase et al.
TABLE II: Vaporization enthalpies of common OLED materials. All values
obtained by thermogravimetry in this study, except as otherwise indicated
a
Measured using OVPD data
b
Doping into Alq
3
c
Thermogravimetry by Yase et al.
d
Mass spectrometry by Yase et al.
TABLE II: Vaporization enthalpies of common OLED materials. All values
obtained by thermogravimetry in this study, except as otherwise indicated
a
Measured using OVPD data
b
Doping into Alq
3
c
Thermogravimetry by Yase et al.
d
Mass spectrometry by Yase et al.
TABLE II: Vaporization enthalpies of common OLED materials. All values
obtained by thermogravimetry in this study, except as otherwise indicated

84
However, T
cell
cannot be raised arbitrarily, because when the thermal energy exceeds the
bond strengths, or more likely exceed the activation energy of a chemical reaction (e.g.
polymerization of the conjugated organic molecules), the source material can rapidly
degrade. Thus, there is an upper bound on the specific evaporation rate due to the limited
chemical stability of the source. This is illustrated by the evaporation of DCM2, a red
fluorescent laser dye. Plotted in Fig. 4-5 is the deposition rate of this dye vs. 1000/T
cell
,
exhibiting the expected exponential dependence on T
cell
. However, as T
cell
continues to
increase, the slope of the curve changes abruptly, indicating that the material has been
chemically altered, resulting in a different H
vap
. Indeed, examination of the source
1.9 2.0 2.1 2.2 2.3
0.01
0.1
1
10




N
O CH
3
NC CN
DCM2
Figure 4-5: Plot of the deposition rate vs. 1/T
cell
for DCM2, where the
evaporation of DCM2 was done by OVPD, with the deposition rate being moni-
tored by a quartz crystal micro-balance in situ. Increasing the temperature of
DCM2 source resul-ted in decomposition and production of a more volatile
species, as evi-denced by the sharply increased deposition rate for T
cell
> 240C.
D
C
M
2

D
e
p
o
s
i
t
i
o
n

R
a
t
e

(

/
s
)
1000/T
cell
(K
-1
)
1.9 2.0 2.1 2.2 2.3
0.01
0.1
1
10




N
O CH
3
NC CN
DCM2
Figure 4-5: Plot of the deposition rate vs. 1/T
cell
for DCM2, where the
evaporation of DCM2 was done by OVPD, with the deposition rate being moni-
tored by a quartz crystal micro-balance in situ. Increasing the temperature of
DCM2 source resul-ted in decomposition and production of a more volatile
species, as evi-denced by the sharply increased deposition rate for T
cell
> 240C.
D
C
M
2

D
e
p
o
s
i
t
i
o
n

R
a
t
e

(

/
s
)
1000/T
cell
(K
-1
)

85
container after the deposition reveals a hard, frothed, black carbonaceous compound, not
at all resembling the fine powder at the start of the experiment.
Furthermore, as seen from Fig. 4-4, some compounds have similar vaporization
enthalpies, but have dramatically different vapor pressures. If one deposition chamber is
to be used to deposit both CuPc and -NPD, the wall temperature must be sufficiently
high to prevent condensation of CuPc, but low enough to prevent decomposition of -
NPD.
The onset of thermal decomposition can be also ascertained by differential
scanning calorimetry (DSC) (Fig.4-6a). In this method, the sample is heated alongside a
standard such as gold or aluminum, having a constant heat capacity in the temperature
range of interest. Keeping the temperature of the sample and the standard equal (but
raising it progressively with time), and monitoring the difference in the heat flow through
the materials, phase and chemical changes can be determined. Figure 4-6b shows results
for Alq
3
, CBP, -NPD, and CuPc; (samples were loaded into aluminum sample pans
under dry nitrogen). Some materials (Alq
3
and CuPc) exhibit greater thermal stability
than others (CBP and -NPD). By comparing Figs. 4-4b and 4-6b, it becomes apparent
that the high temperatures required to evaporate CuPc at rates comparable to -NPD
would likely result in decomposition of -NPD. The argument applies especially in the
case of co-deposition of two components, as in doping of the films for control of the
emission wavelength.

86

100 200 300 400 500 600 700
-25
-20
-15
-10
-5
0
5
10
15
20
25
30


Q

(
J
)
Temperature (C)
Alq
CBP
CuPc
NPD
Reference Sample
T
ref
= T
s
Q
s
Q
r
Q = Q
s
-Q
r
Al pans sealed
under dry nitrogen
(a)
b)
Figure 4-6: a) Schematic of the differential scanning calorimetry (DSC)
experiment, where the sample is being heated to the same temperature as a
reference (e.g. Al), while monitoring the difference in the heat flow to each
container. b) Plotting Q vs. T
cell
results in a straight line if no phase change is
taking Place (slope ~ heat capacity), a spike if an endothermic phase change is
taking place (i.e. melting), and a permanent change in the slope if the material is
being chemically transformed (i.e. the heat capacity has been permanently
altered.)
(b)
100 200 300 400 500 600 700
-25
-20
-15
-10
-5
0
5
10
15
20
25
30


Q

(
J
)
Temperature (C)
Alq
CBP
CuPc
NPD
Reference Sample
T
ref
= T
s
Q
s
Q
r
Q = Q
s
-Q
r
Al pans sealed
under dry nitrogen
(a)
b)
Figure 4-6: a) Schematic of the differential scanning calorimetry (DSC)
experiment, where the sample is being heated to the same temperature as a
reference (e.g. Al), while monitoring the difference in the heat flow to each
container. b) Plotting Q vs. T
cell
results in a straight line if no phase change is
taking Place (slope ~ heat capacity), a spike if an endothermic phase change is
taking place (i.e. melting), and a permanent change in the slope if the material is
being chemically transformed (i.e. the heat capacity has been permanently
altered.)
(b)

87
4.6 Deposited film morphology and composition
In addition to characterizing the growth rate, several deposited film properties must also
be characterized. Optoelectronic devices of interest (e.g. OLEDs) typically consist of
thin-film multilayers of organic compounds, with low conductivities that result in
operation under high electric fields. The microscopic variation in the device thickness can
lead to shunt current paths (shorts) and premature death of the OLED in operation.
Typically, this requires the film roughness to not exceed ~5% of the total film thickness.
For example, in a single-heterojunction OLED consisting of 500 -NPD and 500 Alq
3

films, individual film roughnesses should be < 25.
For typical growth rates (<5/s), OVPD obtains molecularly smooth films (Fig.
4-7a). However, because of the diffusion-limited deposition characteristic of OVPD,
increase in the deposition rate is via increase in the vapor concentration at the edge of the
boundary layer. As the organic molecules traverse the stagnant boundary layer, they
gradually cool below the saturation limit, potentially leading to homogeneous nucleation
in the gas phase. The formation of organic particles in the vapor phase can lead to a
dramatic increase in the surface roughness, and should be avoided. For example,
depositing -NPD on Si, the transition from molecular film growth to gas-phase
nucleation occurs around 10/s, worsening as the rate increases (See. Fig. 4-7).
The chemical composition of the deposited film can be verified in a number of
ways, including from their photo- and electroluminescence (PL and EL) spectra. For
example, Figure 4-8 shows the PL spectrum of an Alq
3
film grown by OVPD on Si,
exhibiting an intensity maximum at 530nm, similar to what is typically obtained for the
Alq
3
source.

88

Figure 4-7: Atomic force micrographs of 2000- thick -NPD films deposited
on Si by OVPD. Each scan is a 5m square, with the vertical range of 500. The
films were deposited at rates of (top to bottom) 8.4, 10.7, and 11.8 /s, exhibiting
a corresponding rms surface roughness of 8, 36, and 45 .
Figure 4-7: Atomic force micrographs of 2000- thick -NPD films deposited
on Si by OVPD. Each scan is a 5m square, with the vertical range of 500. The
films were deposited at rates of (top to bottom) 8.4, 10.7, and 11.8 /s, exhibiting
a corresponding rms surface roughness of 8, 36, and 45 .

89

4.7 Growth of OLEDs using OVPD
A heterojunction OLED similar to that reported by Tang and van Slyke (Tang et al. 1987)
was grown by OVPD. It consisted of a 500 thick -NPD hole transport and a 500
thick Alq
3
electron transport layer grown by OVPD on top of a pre-cleaned ITO-coated
glass substrate. To deposit the metal cathode, the samples were transferred into a thermal
evaporator (base pressure <10
-6
Torr), where 1000 thick layer of Mg co-evaporated
with Ag was grown in a 25:1 mass ratio. A 500 thick Ag cap was used on top of the
MgAg layer to prevent oxidation of the cathode during device testing. A similar structure
was grown by thermal evaporation of both organic and metal layers, without exposing the
organic layers to the atmosphere prior to deposition of the cathode. The device structure
is illustrated in the inset of Fig. 4-9.
400 500 600 700 800
(nm)


Alq
3

Figure 4-8: Molecular structure (inset) and photo-luminescence spectrum of an
Alq
3
film grown on Si by OVPD
N
o
r
m
a
l
i
z
e
d

P
L

I
n
e
n
s
i
t
y
400 500 600 700 800
(nm)


Alq
3

Figure 4-8: Molecular structure (inset) and photo-luminescence spectrum of an
Alq
3
film grown on Si by OVPD
N
o
r
m
a
l
i
z
e
d

P
L

I
n
e
n
s
i
t
y

90
Figure 4-9 shows the current density vs. voltage (J-V) and quantum efficiency vs.
current density the OVPD-deposited (1mm diameter) OLED. Its characteristics are
similar to those of the vacuum-deposited control device, also plotted on the same graph.
At a current density of 1 A/cm
2
, the voltage for a vacuum deposited device is about 2.5 V
lower than for the OVPD-grown device, yet the leakage current (at V<2V) is an order of
magnitude higher. This suggests a slightly thicker organic stack in the case of the OVPD
device and/or the presence of a thin oxide layer between the organic layer and the
Figure 4-9: Electrical characteristics of the hetero-junction OLED comprised of
~500 thick layers of -NPD and Alq
3
grown by OVPD on ITO-coated glass,
followed by VTE of the 1000 thick Mg:Ag (25:1) ca-thode. The current voltage
response and the quantum efficiency of the vacuum-deposited control device are
also shown.
J


(
m
A
/
c
m
2
)
Voltage (V)
J (mA/cm
2
)

e
x
t
(
%
) Alq
3
NPD
0.1 1 10
O V P D
V T E
10
-5
10
-4
10
-3
10
-2
10
-1
1
10
1
10
2
0.1
0.2
0.3
0.4
0.5
0.6
1.0
0.1 1.0 0.01
Figure 4-9: Electrical characteristics of the hetero-junction OLED comprised of
~500 thick layers of -NPD and Alq
3
grown by OVPD on ITO-coated glass,
followed by VTE of the 1000 thick Mg:Ag (25:1) ca-thode. The current voltage
response and the quantum efficiency of the vacuum-deposited control device are
also shown.
J


(
m
A
/
c
m
2
)
Voltage (V)
J (mA/cm
2
)

e
x
t
(
%
) Alq
3
NPD
0.1 1 10
O V P D
V T E
10
-5
10
-4
10
-3
10
-2
10
-1
1
10
1
10
2
0.1
0.2
0.3
0.4
0.5
0.6
1.0
0.1 1.0 0.01
J


(
m
A
/
c
m
2
)
Voltage (V)
J (mA/cm
2
)

e
x
t
(
%
) Alq
3
NPD
0.1 1 10
O V P D
V T E
10
-5
10
-4
10
-3
10
-2
10
-1
1
10
1
10
2
0.1 1 10 0.1 1 10
O V P D
V T E
10
-5
10
-4
10
-3
10
-2
10
-1
1
10
1
10
2
0.1
0.2
0.3
0.4
0.5
0.6
1.0
0.1 1.0 0.01

91
cathode that may have formed during the transfer step (Burrows et al. 1996). In addition,
the external quantum efficiency,
ext
, of the vacuum evaporated OLED is only marginally
higher than the vapor-deposited OLED. The small difference between these devices
suggests that OVPD preserves the purity of the starting materials and produces high
quality organic thin films, while avoiding air-exposure during processing can help
eliminate the remaining small difference in performance.
4.8 Control of dopant concentration using temperature
For co-deposition of two materials, A and B, the ratio of their respective deposition rates
gives the ratio of their concentration in the deposited film (Shtein et al. 2001):
'
0
'
0
exp
exp
vap
A
A A
A
A
A A A
vap
B B B B B
B
B
B
H k
P k
RT
x T
V
x T H k
P k
RT
V

( | |
+
( |
\ .
(
=
(
| |
( +
|
(
\ .

, (4.3)
where x
A
and x
B
are the mole fractions of A and B, respectively, while k
A
and k
B
are
material- and source-specific constants which vary relatively weakly with T and V

.
Thus, the deposition rate can be simultaneously and independently controlled by both cell
temperature and gas flow rate through each source. The flow is used for fine control,
while the temperature regulation produces greater variation course control of
concentration.
Films of Alq
3
doped with DCM2 (Fig. 4-10a) were grown by OVPD on pre-
cleaned glass and silicon substrates. The concentration of DCM2 was varied by means of
adjusting the DCM2 source temperature for each sample. The photoluminescence (PL)
spectra of the resulting films (Fig. 4.10b) were measured under nitrogen (to prevent

92
Figure 4-10: a) Glass slides with 1000 thick DCM2-doped Alq
3
films grown by
OVPD; b) corresponding photoluminescence spectra of films in (a), exhibiting a
red-shift with increasing DCM2 concentration; c) Plot of the DCM2 concentration
determined from its emission spectrum vs. 1/T
cell
of the DCM2 source, showing
H
vap
= 157 kJ/mol, similar to what was found in Fig.4-5.
2.10 2.15 2.20 2.25
0.1
1
10
H
vap
= 157 +/- 0.15 kJ/mol

400 500 600 700 800
(nm)

10% DCM2
Alq
3
(a)
(b)
(c)
%

D
C
M
2

i
n

A
l
q
3
N
o
r
m
a
l
i
z
e
d

P
L

I
n
e
n
s
i
t
y
1000/T
DCM2

93
potential degradation of PL due to O
2
) using a = 400nm wavelength excitation source
with a = 400nm filter placed at the spectrometer input to prevent detector saturation by
the excitation source. The Alq
3
spectrum, centered at = 526nm, was subtracted from the
doped film spectra prior to determining the DCM2 concentration. The peak wavelengths
of the photoluminescence spectra of DCM2-doped Alq
3
films were then converted to
dopant concentrations using a previously determined correlation between these variables
(Bulovic et al. 1999; Shtein et al. 2001). The primary source of error in calculating the
doping concentration of OVPD grown films stems from the uncertainty in the deposition
rate of the vacuum-deposited films (10%), used to develop the spectrum-concentration
correlation. Figure 4-9c depicts the variation in the DCM2 fraction with DCM2 source
temperature. Temperature values in OVPD evaporation are accurate to within (1)C.
The data can be fitted using H
vap
= (157 36) kJ/mol, close to that determined using a
single DCM2 source in OVPD.
4.9 Roll-to-roll deposition
Figure 4-11 shows the deposition tube with the clamshell furnace open, after several
weeks of deposition experiments and microns of cumulative film growth. The chamber
walls surrounded by the heating elements remained clean, while a small condensation
ring is visible just past the substrate, where the heating elements end. The estimated
materials use efficiency was nearly 50%, or significantly greater than observed with
VTE.
To demonstrate the capability of OVPD to deposit OLED structures on large-area
substrates, a mechanical attachment was built, housing motorized spools and a special
water-cooled susceptor, illustrated in Fig. 4-12a,b. Polyethylene terephthalate (PTFE)

94
film, pre-coated with ITO was cut into long, 5cm wide ribbons and loaded onto the
source spool of the attachment. One end of the film was attached to the take-up spool,
while the center portion was guided into the deposition chamber by the cold susceptor.
The radius of curvature of the susceptor was kept >0.5mm to minimize possible cracking
of the ITO film.
To deposit the -NPD/Alq
3
heterostructure, the PTFE ribbon was first spooled in
and out of the chamber at a rate of ~5cm/min, while the -NPD source was heated to
270C, with 15sccm of N
2
flowing through the source barrel. After the -NPD layer was
completely deposited (~800 from the estimated deposition rate), the Alq
3
source was
Figure 4-11: Photograph showing the deposition tube with the clamshell furnace
open, after several weeks of deposition experiments and microns of cumulative
film growth. The chamber walls surrounded by the heating elements remained
clean, while a small con-densation ring is visible just past the substrate, where the
heating elements end. The estimated materials use efficiency is ~50%, over 100
times greater than what has been observed in practice with VTE.
Figure 4-11: Photograph showing the deposition tube with the clamshell furnace
open, after several weeks of deposition experiments and microns of cumulative
film growth. The chamber walls surrounded by the heating elements remained
clean, while a small con-densation ring is visible just past the substrate, where the
heating elements end. The estimated materials use efficiency is ~50%, over 100
times greater than what has been observed in practice with VTE.

95
30cm
CW
OVPD
chamber
Motor-driven
spools & housing
Cooled
susceptor
ITO-coated
plastic film
(a)
(b)
(c) (d)
Figure 4-12: a) Schematic of the mechanical attachment used to deposit the
heterostructures shown in the panels below. b) A photograph of the attachment
and the downstream end of the deposition chamber in an open configuration prior
to loading of the spools with the plastic film. c) Photograph of the -NPD/Alq
3
heterojunction deposited by OVPD on a 30cm long segment of a 5cm wide ITO-
coated plastic film under UV illumination. d) Photograph of a 2m long by 5cm
wide strip of ITO-coated plastic film with an -NPD/Alq
3
-DCM2 hetero-
structure. Shown under UV illumination, the blue luminescent region is the bare
-NPD, the green region is the Alq
3
-covered -NPD, while the orange-red
regions are DCM2-doped Alq
3
on top of -NPD. The red-shift in the DCM2
luminescence is due to the intentionally increasing concentration of the DCM2
molecule in the Alq
3
film.
30cm
CW
OVPD
chamber
Motor-driven
spools & housing
Cooled
susceptor
ITO-coated
plastic film
(a)
(b)
(c) (d)
Figure 4-12: a) Schematic of the mechanical attachment used to deposit the
heterostructures shown in the panels below. b) A photograph of the attachment
and the downstream end of the deposition chamber in an open configuration prior
to loading of the spools with the plastic film. c) Photograph of the -NPD/Alq
3
heterojunction deposited by OVPD on a 30cm long segment of a 5cm wide ITO-
coated plastic film under UV illumination. d) Photograph of a 2m long by 5cm
wide strip of ITO-coated plastic film with an -NPD/Alq
3
-DCM2 hetero-
structure. Shown under UV illumination, the blue luminescent region is the bare
-NPD, the green region is the Alq
3
-covered -NPD, while the orange-red
regions are DCM2-doped Alq
3
on top of -NPD. The red-shift in the DCM2
luminescence is due to the intentionally increasing concentration of the DCM2
molecule in the Alq
3
film.

96
heated to 275C under 15sccm of N
2
barrel flow, and the spooling direction was reversed,
resulting in an ~800 thick Alq
3
layer. In the case of the DCM2-doped Alq
3
layer, the
DCM2 source was progressively heated to higher temperatures to increase its vapor
concentration in the deposition chamber during the deposition of the Alq
3
layer.
4.10 Summary
This chapter reviewed the design and construction of an experimental OVPD system
based on the theory and computer models developed in Ch. 3. This deposition system
was then used to verify the theory of growth rate and doped film deposition, and
subsequently to deposit molecular organic thin films and hetorostructures for OLEDs.
The limiting film growth regimes were explored, where the thermal decomposition of the
source material can limit the maximum source evaporation rate, and homogeneous
nucleation due to vapor supercooling near the substrate can introduce roughness in the
deposited films. To demonstrate the capability of OVPD to deposit over large areas, an
attachment to the apparatus was constructed, enabling roll-to-roll deposition of the
archetypal -NPD/Alq
3
and DCM2-doped Alq
3
OLED heterostructure on a plastic
ribbon.



97
Chapter 5: Growth in confined
geometries, application to patterning
5.1 Overview
Most practical organic electronic devices require the active organic layers to be patterned
in the substrate plane. As molecular organic semiconductors are deemed incompatible
with traditional semiconductor fabrication methods (e.g. photolithography and plasma
processing), the patterning typically involves the use of shadow-masks. In this chapter we
address the challenge of using OVPD and shadow-masking for thin film patterning in the
substrate plane. Gas phase molecular transport simulations are first presented to outline
process conditions permitting pattern resolution on the order of 1 m in the deposited
organic films, which is adequate for typical full-color display applications. The
experimental results are presented in Ch. 6.
5.2 The need for patterning OLED example
Many thin film devices require the active layers to be patterned in the substrate plane. For
example, full color displays typically consist of millions of pixels, each of which is
composed of three sub-pixels, engineered to emit in the red, green, and blue regions of
the visible spectrum. An example of a passive-matrix full color pixel substructure is
illustrated in Fig. 5-1. Provided the sub-pixels are spaced sufficiently closely, the human

98
Figure 5-1: a) Simplified schematic of a pixel in a full-color passive matrix
display, consisting of three sub-pixels, engineered to emit in the red, green, and
blue part of the spectrum. b) A typical sequence of fabricating the pixel in (a),
involving the deposition in vacuum of a red-emitting sub-pixel through a shadow-
mask which covers ~2/3 of the substrate, followed by the same for a green-emitting
pixel, followed by the blue emitting pixel. The device is completed by depositing a
metal cathode through yet a separate shadow mask. In some cases, the substrate
contains an integral shadow mask prior to the organic deposition sequence,
eliminating the need for step (5).
glass / plastic
ITO
SiN
x
organic
metal
150m
200nm
20m
1)
2)
3)
4)
5)
(a)
(b)

99
eye will average out the components of light emitted from each; by controlling the
proportional intensities of light from the sub-pixels, any color can be produced.
The electroluminescence wavelength in OLEDs is typically controlled by the
dopant used in the EML; (e.g. PtOEP for red, Ir(ppy)
3
for green, FIrpic for blue). Thus,
to avoid unintentional bleeding of color between sub-pixels, the doped emissive layers
must be laterally separated.
5.3 Ballistic transport in VTE and shadow-masking
A typical vacuum deposition geometry is illustrated in Fig. 5-2. The organic material is
evaporated from a resistively heated cell onto a substrate positioned directly above it. In
deposition of doped films, two or more materials are co-evaporated with the sources
separated by a baffle, as shown in the figure. The substrate is placed within a distance
less than the molecular mean free path, , from the source, and the transport of organic
molecules from the source to the substrate is ballistic. For patterned film deposition, a
shadow mask is placed at a distance s in front of the substrate (Fig. 5-2a). For a mask
aperture with straight walls, thickness, t, and using a source of width l centered on the
aperture axis, the shape of the deposit is approximately trapezoidal, with the width of the
edge taper and, hence, the resolution limit, , given by:
( 2 )
2
s t l
h

+
(5.1)
Here, w is the aperture width, and h is the source-to-mask distance, as defined in Fig. 5-2.
The design guidelines suggested by Eq. (5.1) are intuitive: the mask should be as thin as
possible, while the separation, s, should be small, to produce the sharpest edges.
However, the mask also should not bow under its own weight, and it should withstand

100
multiple mount-remount and cleaning cycles. Thus, the actual mask dimensions represent
a compromise between the pattern resolution and hardware robustness requirements, with
typical values for s, t and w about 10, 70, and 100 m, respectively. The typical source
diameter, l 1 cm, and source-to-substrate distance h 50cm, yielding 2 m, which
is adequate for full-color OLED displays.
5.4 Shadow-masking and diffusive transport in OVPD
The OVPD deposition geometry is illustrated in Fig. 5-3a. As shown in Ch.4, the carrier
gas flow creates a hydrodynamic boundary layer at the substrate (typically 1mm < <
5cm), across which the organic species diffuse prior to condensation. The organics are a
minority species (< 1% molar concentration) and collide with the carrier gas molecules
en route to the substrate. The carrier gas background pressure far exceeds 10
-6
Torr,
leading to the condition , where the molecular velocities are completely randomized
pump
substrate
Source A Source B
mask
heater
(a)
(b)
s
w
h
l
substrate
deposit
t
source
mask
Figure 5-2: a) Schematic of the VTE process. b) Schematic of pattern formation in
the case of ballistic transport occurring in VTE, with definition of the relevant
geometrical parameters.

101
near the substrate. This situation is depicted in Fig. 5-3b, equivalent to a source with l
w and with h l (i.e. an infinitely wide source positioned very close to the aperture). This
mode of transport is already dramatically different from vacuum thermal evaporation,
where the molecular source-to-substrate distance, h, is large, transport is ballistic, and
molecular trajectories are all nearly perpendicular to the substrate near the aperture.
Furthermore, if < s, then the organic molecules undergo scattering collisions in the
space between the mask and the substrate, with a potentially non-trivial dependence on
Coolant
Source A
Source B
Carrier gas
Pump
Pump
(a)
(b)
deposit mask
Substrate
Heater
t
s
Figure 5-3: a) Schematic of the OVPD process. b) Illustration of the diffusive
mode of transport near the substrate, resulting in randomized molecular trajectories
in the vicinity of the mask, ultimately leading to more diffuse patterns than in the
case of VTE.

102
the deposition conditions (i.e. P
dep
, s, and t). The resulting deposition pattern is likely to
have substantial tails, significantly different from the trapezoidal pattern typically
obtained by VTE. The essential features of the masking geometry and the shape of the
deposited pixel are depicted in Fig. 5-4.
As a means for quantifying pattern shape, we define the shape factor, , which is
equal to the area bound by -w/2 < x < w/2 (the cross-hatched region in Fig.5-3) divided
by the total deposit cross-sectional area. Then = 1 for a rectangular deposit, decreasing
as the deposit tails extend beyond the edges of the mask aperture. For example, a deposit
shaped as a Gaussian distribution, with half-width w/2, will have 0.67. In analyzing


w
s
t

mask

z
x
deposit
profile
substrate
Figure 5-4: a) Definition of relevant geometrical parameters for the mask aperture
and the deposit formed during OVPD. The shaded region corresponds to the
useful portion of the deposit, defined either by <90% decrease in the deposit
height, or by the projected aperture width, w.


w
s
t

mask

z
x
deposit
profile
substrate
Figure 5-4: a) Definition of relevant geometrical parameters for the mask aperture
and the deposit formed during OVPD. The shaded region corresponds to the
useful portion of the deposit, defined either by <90% decrease in the deposit
height, or by the projected aperture width, w.

103
the effects of aperture geometry on the deposited feature shape, we also consider the
fraction, f, of molecules lost by deposition on the side of the mask facing the substrate,
and on the aperture walls. Then, the deposition efficiency is simply = 1 f.
5.5 Continuum-based analysis is inaccurate for Kn < 10
Assuming simple kinetic theory of a gas and a hard-sphere intermolecular interaction
potential, the molecular mean free path is given by: (Bird et al. 1996)
2
2
B
dep
k T
d P


=

(5.2)
where k
B
is the Boltzmann constant, T is the gas temperature, and d
2
is the effective
collision area for two molecules, each having a diameter d. At 25C and a typical VTE
chamber pressure of 10
-6
Torr, > 50 cm for nitrogen (d = 3.25 ), while in OVPD at
P
dep
= 0.1 10 Torr, ~ 100 1 m.
The Knudsen number, Kn, is frequently used to characterize different transport
regimes in terms of the ratio of to the critical apparatus dimension, L (Kn = /L)
(Stechelmacher 1986) The dimension L is typically chosen such that Kn represents the
ratio of intermolecular to molecule-wall collision frequencies. For gas flow between the
source and the substrate regions, L is the length of the transport chamber or the run lines,
so that Kn << 1; thus, continuum models can be used to understand flow dynamics and
the deposition rate on the scale of centimeters or meters. However, when film patterning
is considered, the apparatus dimension must account for the aperture geometry. In this
case, L is 1 100 m, and 0.1 < Kn < 10 for much of the useful deposition pressure
range. As turns out to be the case, the discrete molecular nature of the gas flow cannot be
ignored. (Bird 1994) This can be illustrated as follows.

104
For the geometry depicted in Fig. 5-4, the mask of thickness t and aperture of
width w is separated from the substrate by distance s. For isotropic diffusion, the width of
the deposit base will increase with s. That is, the longer it takes for a molecule to diffuse
to the substrate, the longer it will take (by the same amount) for it to diffuse laterally. The
mutual cancellation of these rates will result in identical patterns at different pressures,
which, as will be discussed in the following sections, does not match the observed
experimental trend. Moreover, in the continuum assumption, the pattern formation does
not depend on the shape of the aperture above the edge closest to the substrate, which is
again contrary to experiment. Due to the discrete nature of transport at very low pressure,
some collimation of flux through the aperture must occur when the organic molecules
condense on the aperture sidewall.
Thus, while the continuum analysis helps visualize transport on the scale of the
chamber dimension, to predict the OVPD growth of patterns whose size and resolution
are on the order of , the molecular nature of transport near the substrate must be
considered. The following sections employ Monter-Carlo simulations to obtain
quantitative information about the pattern shape, as well as other merit figures of
practical interest, such as efficiency of material transport through the apertures. The
simulations are then compared with experimental observations for a limited number of
conditions. On the basis of the close correspondence between the simulated and
experimentally obtained results, the former technique is used to examine the potential
improvements of new geometries and conditions that can be difficult to achieve
experimentally without a great expenditure of time and resources. The resolution limits of
shadow masking are explored.

105
5.6 Modeling gas transport in confined geometries
Molecular flow through small channels has been previously studied using several
techniques. For example, Guevremont and coworkers (Guevremont et al. 2000)
experimentally analyzed the shape of molecular fluxes from collimated effusive beam
sources and the nature of gas transport within small-diameter capillary arrays. The beam
shape was obtained by translating the beam across a skimmer, behind which a mass
spectrometer collected the molecular flux. Channel conductance was studied for a range
of Kn. The sources consisted of arrays of capillaries, whose individual diameters were as
small as 10m. Although in their experiments ranged from 1 to 10
4
m thereby
covering a wide range of Kn, the overall diameter of the source array, and hence the beam
width, was > 1mm. Furthermore, stagnation of the flow due to the substrate was absent,
in which case the molecular flux is more collimated than when the velocity distribution is
isotropic, as in the case of pure diffusion. Thus, although the beam profiles resemble the
pattern shapes obtained by OVPD, those results are not strictly applicable.
In contrast to the case where collimated beams are employed, transport of species
to the substrate is often diffusive during reactive ion etching and physical vapor
deposition. Semiconductor electronic device processing typically involves etching of, and
deposition in holes and trenches of sub-micron dimensions.(Wolf et al. 1999) The
processing pressure, however, is typically such that Kn 1, similar to OVPD growth of
patterned films. Studies reveal that Monte-Carlo type simulations can accurately predict
deposition and etch rates when coupled with the proper surface reaction models,
(Akiyama et al. 1995; Griffiths et al. 1998) suggesting their applicability to
micropatterning of thin film deposits by OVPD. Models of OVPD growth, however, are

106
greatly simplified by the absence of gas-phase and surface reactions, and re-evaporation
of adsorbed species that must be considered for reactive deposition or etching (Wulu et
al. 1991).
5.7 Monte-Carlo simulation of OVPD through apertures
The computer simulation models the diffusion of organic molecules through the stagnant
carrier gas layer to the substrate. The boundary conditions include a source of organic
molecules at some distance from the substrate, and sinks at the mask and substrate
surfaces. To avoid the shortcomings of the continuum approximation, the molecular
nature of transport in the deposition process is studied using a stochastic (Monte-Carlo)
approach. The objective of the simulations was to examine the trends in deposit shape as
functions of process parameters, such as the deposition pressure, temperature, molecular
diffusivity, mask geometry, and mask-to-substrate spacing.
5.7.1 Simulation set-up
The MC model allows the variation of directly, tracking the random path of a molecule
as it diffuses through a background of carrier gas in the vicinity of the substrate. For
simplicity, the molecular sticking probability on the substrate and mask surfaces is
assumed to be unity, which is usually valid for typical multi-ring aromatic semiconductor
molecules and substrate temperatures 300K. Generally, under OVPD growth at T =
250C, the vapor pressure of the organics is in the 10
-3
Torr range, (Shtein et al. 2001;
Yase et al. 1995; Yase et al. 1996) while the background carrier gas pressure is on the
order of 0.01 to 10 Torr, allowing for a gradient in the organic vapor concentration at a
relatively constant total pressure. In this case, organic-organic collisions are rare except at

107
very high deposition rates where gas phase nucleation can occur. (Shtein et al. 2001)
Hence the model is valid only for low-to-moderate (50 /s) film deposition rates.
The rate of molecular transport in gases due to a concentration gradient is
characterized by the diffusivity, D: (Bird et al. 1996)
1
3
= D u (5.3)
where is the molecular mean thermal velocity. Since the organic vapors consist of
molecules that have collision diameters greater than 10 and are geometrically and
energetically more complex than the hard sphere atoms of kinetic theory, molecular
dynamics simulations should be used for an accurate estimation of the . However, since
this study is concerned with the dependence of trends in pattern evolution on process
parameters, simple kinetic theory equations are sufficient for the analysis.
The Monte-Carlo simulation proceeds as follows. The computational space is
divided into an x-z grid extending infinitely in the y-direction (see Fig. 5-4 for definition
of the coordinate space). The purpose of the grid is to locate the substrate and mask
surfaces, and track changes in the thickness of the deposits. A spherical test organic
molecule is assigned a random initial location (x
0
, y
0
, z
0
) inside the boundary layer and
above the mask. A random initial direction is then chosen, and the molecule travels a
distance r = [(x
1
-x
0
)+ (y
1
-y
0
)+(z
1
-z
0
)]
1/2
, where (x
1
, y
1
, z
1
) is its new location. Molecules
escaping the boundary layer are reflected back toward the substrate. The distance r is the
minimum of the grid size, or /10. The probability of collision, P
coll
, with a carrier gas
molecule is equal to r/, which is checked against a random number, rn, between 0 and
1. If P
coll
< rn, the molecule is again allowed to proceed in the same direction for another
step of distance r. If P
coll
> rn, the particle is assumed to collide with a carrier gas

108
molecule having a velocity chosen randomly from a Maxwell-Boltzmann distribution.
The collision causes the molecule to be deflected with a velocity and an angle consistent
with momentum and energy conservation in the elastic collision of two hard spheres. If
the path of the particle crosses the substrate plane or the aperture wall, the particle is
assumed to stick to the surface with unity efficiency. A single aperture is simulated, with
periodic boundary conditions imposed in the x-direction. Effects of aperture geometry on
deposited pattern formation are modeled by assigning different aperture sidewall angles,
= 45, 90, 135, and = 270.
5.7.2 First results for patterning of Alq
3
thin films
Consider the growth of films of the archetype molecule, tris(8-hydroxyquinoline) (Alq
3
),
having d~10 . The mean free path is calculated using Eq. (5.2), assuming that the
collision diameter for a mixture of Alq
3
in a N
2
carrier gas is the average of the respective
diameters of the two species in self-diffusion viz.: d = (d
Alq3
+d
N2
).
Figure 5-5a shows a gray-scale map of the organic species concentration from a
simulated deposition of 10
5
particles, with s = 7m, t = 3m, and = 135, and =
100m, corresponding to a total deposition pressure of P
dep
~ 0.1 Torr at T = 500K.
While the simulated particle diameter was 10 , it was enlarged in the figure to better
illustrate the deposited profile. The initial particle velocities were assigned from a
random thermal distribution, and are superimposed onto a z-directed velocity vector with
a magnitude of 10m/s. The size of the individual particle has been enlarged to show the
deposited film thickness profile. By decreasing the by a factor of 10, a more diffuse
pattern is obtained, shown in Fig. 5-5b. Note that the films shown here are unrealistically
thick and t is unrealistically thin for conventional OVPD conditions; a film thickness of

109
~1000 and t 50 m are more practical. The figure nevertheless illustrates the
increased parasitic deposition on the mask and aperture walls when is reduced. For
clarity, subsequent simulation results will provide only the thickness profile without the
(a)
(b)
mfp = 100m
mfp = 10m
10m
10m
launching surface
mask deposit
mask
substrate deposit
substrate
log[Concentration], normalized
0 1
Figure 5-5: a) Monte-Carlo simulation results plotted as a gray-scale map of the
organic species, both as a deposit and as gas phase species using 10
5
particles, with
s = 7m, t = 3m, and = 135, and = 100m, corresponding to a total
deposition pressure of P
dep
~ 0.1 Torr at T = 500K. While the simulated particle
diameter was 10 , it was enlarged in the figure to better illustrate the deposited
profile. b) Simulation conditions identical to (a), except = 10m, corresponding
to a ten-fold increase in the deposition pressure, leading to a broadened deposit.
The time-averaged molecular trajectories are also visible, showing the rapid
dispersion of the species past the aperture edge.

110
shadow mask shown. The total number of Alq
3
molecules was also increased to 10
6
to
obtain smoother, more accurate thickness profiles.
5.7.3 Effects of chamber pressure on deposit shape
The effects of deposition pressure on pattern resolution were investigated by varying .
Growth through an aperture with w = 80m, t = 70m, and with the substrate separated
from the lower mask edge by s = 20m was considered. The choices for s and t are
consistent with deposition of high resolution display picture elements (pixels) by VTE,
where the substrate-facing-downward evaporation geometry causes the mask to bow
away from the substrate under gravity, leading to s ~ 20m. To stiffen the mask and
minimize s, usually t > 70m is required for VTE. Considerably thinner masks (e.g. t
50 m) can be used with OVPD due to the possibility of top-down deposition. Figure 5-6
1 10 100 1000
0.60
0.65
0.70
0.75
0.80




P
i
x
e
l

s
h
a
p
e

f
a
c
t
o
r
,

Mean free path, (m)


w = 80 m
s = 20 m
t = 70 m
mask
substrate
Figure 5-6: Variation in the pixel shape factor, , with the mean free path, ,
obtained from the Monte-Carlo simulation. Depositing at lower pressure (or longer
) results in more sharply defined edges (i.e. larger , as defined in the text).
1 10 100 1000
0.60
0.65
0.70
0.75
0.80




P
i
x
e
l

s
h
a
p
e

f
a
c
t
o
r
,

Mean free path, (m)


w = 80 m
s = 20 m
t = 70 m
mask
substrate
Figure 5-6: Variation in the pixel shape factor, , with the mean free path, ,
obtained from the Monte-Carlo simulation. Depositing at lower pressure (or longer
) results in more sharply defined edges (i.e. larger , as defined in the text).

111
shows that the shape factor, , increases only weakly as varies over four orders of
magnitude (1m 1000m, corresponding approximately to 10 Torr P
dep
0.01
Torr), as would be expected for a system approaching the Knudsen transport regime.
However, does not reach unity (corresponding to a rectangular deposit profile) due to
parasitic deposition on the sidewalls of the almost square aperture. As expected, the
highest pattern edge resolution (and hence, highest ) is achieved for the largest , i.e.
the lowest P
dep
. However, as discussed below, the pattern profile exhibits a dome-like
shape due to a relatively large t/w.
5.7.4 Effects of mask thickness and separation
Increasing the aperture thickness can improve collimation of the molecular flux toward
the substrate due to the condensation on the walls of the aperture. However, as Fig. 5-7
0 100 200 300 400 500 600


T
h
i
c
k
n
e
s
s

(
a
.
u
.
)
Position (m)
t = 5,10,20,50m
s = 20m
mfp = 20m
t=50m
mask
substrate
Figure 5-7: Deposit profile as a function of mask thickness, t, obtained from the
Monte-Carlo simulation discussed in the text, using 10
6
Alq
3
molecules in N
2
background carrier gas with mfp = s = 20m. The aperture dimensions are listed on
the plot. The pixel shape factor, , decreases weakly with s for this set of
parameters.
0 100 200 300 400 500 600


T
h
i
c
k
n
e
s
s

(
a
.
u
.
)
Position (m)
t = 5,10,20,50m
s = 20m
mfp = 20m
t=50m
mask
substrate
Figure 5-7: Deposit profile as a function of mask thickness, t, obtained from the
Monte-Carlo simulation discussed in the text, using 10
6
Alq
3
molecules in N
2
background carrier gas with mfp = s = 20m. The aperture dimensions are listed on
the plot. The pixel shape factor, , decreases weakly with s for this set of
parameters.

112
shows, increasing t from 5 to 50m for s = = 20m and w = 300m, has no effect on
the spreading of the lower edge of the deposit. However, as t/w increases, the deposited
pattern becomes significantly more domed due to the scavenging of the organic
molecules by the upper edges of the aperture. The shape factor decreases by less than 5%
over the range 1m < t < 100 m.
The mask separation, s, can vary due to bowing under gravity (in VTE), or
possibly heating during the deposition process. Figure 5-8 shows profiles for depositions
where s = 2, 10, 22, and 50m, and t = 20m, w = 300m, and = 20 m. Appreciable
edge broadening arises due to collisions in the mask-substrate gap for s > . In addition,
for large s the dome in the middle of the deposit is also pronounced and, in contrast to the
0 100 200 300 400 500 600


t = 20m
mfp = 20m
T
h
i
c
k
n
e
s
s

(
a
.
u
.
)



1.0

100
s
0.8
0
Position (m)
mask
substrate
s = 2m
10m
22m
50m
Figure 5-8: Simulated deposit profiles for a cylindrical aperture mask at various
separation distances, s, with dimensions and details of the simulation listed on the
plot. The pattern edge becomes more diffuse, while the deposition efficiency
decreases with increasing t due to parasitic deposition on the aperture sidewall, as
discussed in the text. Inset: The deposit shape factor decreases approximately
linearly with s. (Legend: s = [] 2, [---]10, [] 22, [] 50m.)
0 100 200 300 400 500 600


t = 20m
mfp = 20m
T
h
i
c
k
n
e
s
s

(
a
.
u
.
)



1.0

100
s
0.8
0
Position (m)
mask
substrate
s = 2m
10m
22m
50m
Figure 5-8: Simulated deposit profiles for a cylindrical aperture mask at various
separation distances, s, with dimensions and details of the simulation listed on the
plot. The pattern edge becomes more diffuse, while the deposition efficiency
decreases with increasing t due to parasitic deposition on the aperture sidewall, as
discussed in the text. Inset: The deposit shape factor decreases approximately
linearly with s. (Legend: s = [] 2, [---]10, [] 22, [] 50m.)

113
case of large t/w, the profile is also broadened at the edges. The inset of Fig. 5-8 shows a
linear decrease in with s for this series of depositions. Thus, optimal pattern resolution
leading to a rectangular deposit is achieved for the smallest values of s and t, as expected.
Since OVPD can, in principle, be carried out with the mask positioned above the
substrate, thin masks can be used while s is kept small, contrary to the case in VTE. By
using thinner masks, the pattern deposition efficiency, , is also increased, because less
material will be deposited on the aperture side-wall, as will be shown below. The
drawback of thin masks is that they are more susceptible to thermally and mechanically
induced stresses. Nevertheless, in a top-down configuration of OVPD, smaller values
of s and t are possible, potentially leading to higher resolution patterns than is achievable
using conventional VTE deposition.
5.7.5 Optimizing mask (aperture) shape
It is also possible to minimize t at the aperture edge while keeping the mask thick
elsewhere. The variation in with the aperture shape was investigated by varying the
aperture side-wall angle, = 45, 90, 135, and = 270 (see Fig. 5-4). The pattern
profiles for these aperture shapes are shown in Fig. 5-9. The aperture with = 135
results in the most diffuse edge of the deposit (smallest ) due to the lack of collimation
of the approach angle of the molecules. This occurs to a lesser extent with the biconical
aperture ( = 270), but the sharpest patterns are achieved with = 90 and = 45.
The variation of the shape factor, , versus t is plotted in Fig. 5-10a for all four aperture
geometries. While apertures with = 90 and 45 show a minimal decrease in with t,

114
curves for ' = 270 and = 135 drop significantly with t, consistent with the profiles in
Fig. 5-9.
In addition to allowing deposition of sharper patterns, it is desirable to maximize
the material flux to the substrate. The pattern deposition efficiency, , normalized to that
of the aperture with = 45 and t = 2 m, is plotted vs. t in Fig. 5-10b. For all four
aperture shapes, decreases for larger t, the effect being most pronounced for = 90.
50 100 150 200 250


50 100 150 200 250


50 100 150 200 250


50 100 150 200 250


=90
=45
=270
=135
(a) (b)
(c) (d)
s=10 m; mfp=20 m; t=5-75 m
Increasing t Increasing t
Increasing t
Increasing t
Figure 5-9: a)-d) Effects of aperture wall profile (and mask thickness, t) on deposit
cross-section. In each figure, the mask thickness increases from 5 m (upper curve)
to 75 m (lower curve) in 10 m increments. The mask-to-substrate separation, s,
and mfp are held constant at 10 m and 20 m, respectively.
50 100 150 200 250


50 100 150 200 250


50 100 150 200 250


50 100 150 200 250


=90
=45
=270
=135
(a) (b)
(c) (d)
s=10 m; mfp=20 m; t=5-75 m
Increasing t Increasing t
Increasing t
Increasing t
Figure 5-9: a)-d) Effects of aperture wall profile (and mask thickness, t) on deposit
cross-section. In each figure, the mask thickness increases from 5 m (upper curve)
to 75 m (lower curve) in 10 m increments. The mask-to-substrate separation, s,
and mfp are held constant at 10 m and 20 m, respectively.

115
0.0
0.2
0.4
0.6
0.8
1.0


0.0
0.2
0.4
0.6
0.8


= 90
P
i
x
e
l

s
h
a
p
e

f
a
c
t
o
r
,

Mask thickness (m)


D
e
p
o
s
i
t
i
o
n

e
f
f
i
c
i
e
n
c
y
,

0 20 40 60 80 100
0.0
0.2
0.4
0.6
0.8


= 270
= 135
= 45
(a)
(b)
(c)
Figure 5-10: a) Deposit shape factor, , as a function of the mask thickness, t, for
the simulated depositions in Fig. 9. b) Plot of the deposition efficiency, , vs. t for
various aperture shapes. The efficiency for = 90 is lowest due to parasitic
deposition on the aperture walls. c) The figure of merit, , represents the
combined effects of pattern sharpness and deposition efficiency using a particular
aperture configuration. Apertures with = 45 yield the highest .
0.0
0.2
0.4
0.6
0.8
1.0


0.0
0.2
0.4
0.6
0.8


= 90
P
i
x
e
l

s
h
a
p
e

f
a
c
t
o
r
,

Mask thickness (m)


D
e
p
o
s
i
t
i
o
n

e
f
f
i
c
i
e
n
c
y
,

0 20 40 60 80 100
0.0
0.2
0.4
0.6
0.8


= 270
= 135
= 45
(a)
(b)
(c)
Figure 5-10: a) Deposit shape factor, , as a function of the mask thickness, t, for
the simulated depositions in Fig. 9. b) Plot of the deposition efficiency, , vs. t for
various aperture shapes. The efficiency for = 90 is lowest due to parasitic
deposition on the aperture walls. c) The figure of merit, , represents the
combined effects of pattern sharpness and deposition efficiency using a particular
aperture configuration. Apertures with = 45 yield the highest .

116
By plotting the product vs. t, the cumulative effect of aperture shape on the deposited
pattern can be seen. While the aperture with = 90 yields sharp deposit edges, the drop
in with t for this shape makes it less desirable than the aperture with = 45 for
depositing well-defined patterns with the least amount of material wasted on coating the
mask. However, since includes all material deposited on the substrate, for = 135
is greater than for = 270 and = 90, although the pattern edges are significantly
more spread out. Table III contains a summary of the relative influences of changes in
the deposition conditions and mask aperture geometry on the resulting pattern shape
factor, , pattern deposition efficiency, , and the combined figure of merit. In the
following sections experimental results are shown confirming the models developed thus
Table III: Effects of masking geometry and process parameters on pattern shape.
Parameter
(a)
Pixel Shape
Factor
Deposition
Efficiency
Combined
Figure of Merit
() () ()
s
a
t

(b)

(a)
Here indicates an increase in the parameter, a decrease.
(b)
For a 90, h was insensitive to t.
Table III: Effects of masking geometry and process parameters on pattern shape.
Parameter
(a)
Pixel Shape
Factor
Deposition
Efficiency
Combined
Figure of Merit
() () ()
s
a
t

(b)

(a)
Here indicates an increase in the parameter, a decrease.
(b)
For a 90, h was insensitive to t.

117
far. A novel method for depositing patterned organic films with self-aligned metal
contacts is also demonstrated.
5.8 Experimental set-up
The deposition of thin films of Alq
3
was carried out using the horizontal reactor described
in Ch.4. The deposition profiles of organic thin films obtained using OVPD were
compared with those prepared by VTE. The VTE source-to-substrate distance was
approximately 30 cm and the deposition base pressure was maintained at 10
-6
Torr. The
mask and substrate configuration is shown in Fig. 5-11. The substrates were placed into
recessed sections in a specially machined copper susceptor, the mask was layed on top of
the substrate, and a retainer was clamped over the mask edges by means of several
Figure 5-11: a) Schematic of the substrate and mask assembly used for the
micropatterning experiment. The substrate fits inside of a machined recess in the
copper susceptor, while the Molybdenum mask is placed over it. The mask is
separated from the substrate by a spacer (Type A) or by a Nickel mesh (Type B).
The stack is compressed by a retainer which clamps the edges of the mask &
substrate.
Susceptor
Retainer
Fastening
screw
Machined recess
Substrate
Nickel mesh
Molybdenum mask
Substrate
Spacer
Molybdenum mask
(Type A) (Type B)
Figure 5-11: a) Schematic of the substrate and mask assembly used for the
micropatterning experiment. The substrate fits inside of a machined recess in the
copper susceptor, while the Molybdenum mask is placed over it. The mask is
separated from the substrate by a spacer (Type A) or by a Nickel mesh (Type B).
The stack is compressed by a retainer which clamps the edges of the mask &
substrate.
Susceptor
Retainer
Fastening
screw
Machined recess
Substrate
Nickel mesh
Molybdenum mask
Substrate
Spacer
Molybdenum mask
(Type A) (Type B)

118
screws. The first mask (Type A) was a 60 m thick, 1 cm x 1 cm Mo foil with circular
openings having diameters of 1000, 300, and 100 m. The aperture profile in this mask
was cylindrical ( = 90). A second mask (Type B, Fig. 5-12a) was a 75 m thick Mo
sheet with circular apertures having the same nominal diameters as in Type A. The
openings in this mask had a double-beveled edge, forming a biconical aperture ( =
270). The third mask (Mask C, Fig. 5-12b) was a Ni mesh, 3.5 0.5 m thick, with
nominally 7.5 m and 12.5 m square openings separated by equally wide lines. The
mask-to-substrate separation was controlled by shims of multiple layers of the Ni mesh
placed between the Si substrate and the Mo mask bottom surface. In depositions through
the Ni meshes, 1 cm x 1 cm mesh sheets were fixed to the substrate by sandwiching them
between the substrate and the first or second type of mask, and then were clamped to the
holder by the retainer. Due to the profile of the Ni mesh, the smallest effective separation
was 1.0 0.5 m.
Analysis of the deposited pattern profiles was performed using scanning electron
microscopy (SEM) and atomic force microscopy (AFM) for the smallest pattern sizes,
and interference microscopy (see Fig. 5-13a) for the larger patterns. The latter method
entailed illuminating the substrates with a monochromatic source (with wavelength =
(540 10) nm) and observing the interference fringes formed at the sloping edge of the
bell-shaped deposit (see for example Fig. 5-13b). The thickness profile was extracted
from the digitized pattern image (Fig. 5-13c) by counting the number of fringes from the
edge (Fig. 5-13d) and using H = m / 2n, where H is the pattern thickness, m = 0, 1, 2, 3,
etc. is the fringe order, and n = 1.74 is the refractive index of Alq
3
.

119
Figure 5-12: Scanning electron micrographs (SEMs) of the Molybdenum mask
(Type B) and the Nickel mesh (Type C) used in the experiments. The Ni mesh was
also used as a spacer, being ~4.5m thick and flexible enough to be folded over
several times for spacer thickness to be increased in small increments. Mask Type
A (not shown) was identical to Type B, except the aperture had a cylindrical
profile with straight walls, and not the biconical ones of Type A.
(Type B)
(Type C)
Figure 5-12: Scanning electron micrographs (SEMs) of the Molybdenum mask
(Type B) and the Nickel mesh (Type C) used in the experiments. The Ni mesh was
also used as a spacer, being ~4.5m thick and flexible enough to be folded over
several times for spacer thickness to be increased in small increments. Mask Type
A (not shown) was identical to Type B, except the aperture had a cylindrical
profile with straight walls, and not the biconical ones of Type A.
(Type B) (Type B)
(Type C)

120

5.9 Shadow-masking experiments results and discussion
A micrograph with circular pixels deposited by OVPD through Mask A ( = 90) is
shown in Fig. 5-13b. In Fig. 5-14a, the measured pattern profiles for s = 0m are plotted
(open circles), while those for s = 40m are shown in Fig. 5-14b. The open circles in Fig.
5-14a correspond to films nominally 2m thick, while those in Fig. 5-14b were 1.6m
(a)
Figure 5-13: a) Illustration of the interference microscopy method used to
determine the thick-ness profile of the deposits. b) Micrograph show-ing circular
Alq
3
patterns deposited on Si using OVPD. The interference fringes observed for
each deposit indicate a variation in the thickness near its edge. Here, a cylindrical
mask was used with t = 50 m, s = 0, and w = 100, 300, and 1000 m. c)
Digitized image of a deposit showing the interference fringes near the edge of the
deposit d) Plot of the light intensity along the radius of the deposit, and the
corresponding thickness profile calculated from the interference pattern similar to
the one shown in (b).
d

2d sin = (m+
1
/
2
)/n

Thickness Profile
Light intensity profile
Position (m)
100 200 300 400 500 600 700 0
(d)
1000m
(c)
(b)
(a)
Figure 5-13: a) Illustration of the interference microscopy method used to
determine the thick-ness profile of the deposits. b) Micrograph show-ing circular
Alq
3
patterns deposited on Si using OVPD. The interference fringes observed for
each deposit indicate a variation in the thickness near its edge. Here, a cylindrical
mask was used with t = 50 m, s = 0, and w = 100, 300, and 1000 m. c)
Digitized image of a deposit showing the interference fringes near the edge of the
deposit d) Plot of the light intensity along the radius of the deposit, and the
corresponding thickness profile calculated from the interference pattern similar to
the one shown in (b).
d

2d sin = (m+
1
/
2
)/n

Thickness Profile
Light intensity profile
Position (m)
100 200 300 400 500 600 700 0
(d)
1000m
(c)
(b)

121
thick. Both plots indicate that the pattern deposition efficiency (inferred from the area
under each profile) decreases with the aperture aspect ratio, t/w. The dome in the center
of each pattern, which is particularly pronounced for large t, is similar to that obtained in
the simulation. Also plotted in Fig. 5-14a is the profile for w = 300m and s = 40m
from Fig. 5-14b (filled circles). It is evident that increasing s decreases the efficiency of
pattern deposition (due to condensation on the mask) and edge sharpness. As discussed in
Sec. 5.4, this result is not obtained using the simple continuum diffusion model.
Figure 5-14: Experimental pixel profile for cases with a mask-to-substrate
separation of (a) s = 0 and (b) s = 40m, respectively. The nominal film
thicknesses were (a) 2m and (b) 1.6m. The thickness profile was obtained for
the larger diameter deposits by scanning from the center of the deposit outward,
and reflecting the resulting thickness profile about the origin. For smaller deposits,
the entire deposit was scanned. Both (a) and (b) indicate that the pixel deposition
efficiency decreases with the aperture aspect ratio, t/w, where t and w are the mask
thickness and aperture width, respectively. The profiles from both runs for the w =
300m aperture are scaled by their thicknesses and compared in Fig. 13a (open
and solid circles). The experimental comparison shows that greater mask-to-
substrate distance results in a more diffuse deposit and lower pattern deposition
efficiency.
-1000 -500 0 500 1000
0.0
0.5
1.0
1.5
2.0
Position (m)


-500 0 500 1000


Position (m)
(a) (b)
T
h
i
c
k
n
e
s
s

(

m
)
s = 0 m
scaled by
ratio of
thicknesses
s = 40 m
w = ( ) 1000, ( ) 300, ( ) 100 m w = ( ) 1000, ( ) 300, ( )100 m
Figure 5-14: Experimental pixel profile for cases with a mask-to-substrate
separation of (a) s = 0 and (b) s = 40m, respectively. The nominal film
thicknesses were (a) 2m and (b) 1.6m. The thickness profile was obtained for
the larger diameter deposits by scanning from the center of the deposit outward,
and reflecting the resulting thickness profile about the origin. For smaller deposits,
the entire deposit was scanned. Both (a) and (b) indicate that the pixel deposition
efficiency decreases with the aperture aspect ratio, t/w, where t and w are the mask
thickness and aperture width, respectively. The profiles from both runs for the w =
300m aperture are scaled by their thicknesses and compared in Fig. 13a (open
and solid circles). The experimental comparison shows that greater mask-to-
substrate distance results in a more diffuse deposit and lower pattern deposition
efficiency.
-1000 -500 0 500 1000
0.0
0.5
1.0
1.5
2.0
Position (m)


-500 0 500 1000


Position (m)
(a) (b)
T
h
i
c
k
n
e
s
s

(

m
)
s = 0 m
scaled by
ratio of
thicknesses
s = 40 m
w = ( ) 1000, ( ) 300, ( ) 100 m w = ( ) 1000, ( ) 300, ( )100 m

122
Experimental deposition profiles for Mask B with = 270, w = 100m, t =
75m and s 0 m are compared to simulations in Fig. 5-15. The simulation assumed
= 20m and diffusive deposition (i.e. no bulk flow in the z-direction) at P
dep
0.2 Torr,
as used in the experiment. The simulated aperture geometry matches the experimental
set-up, but s was adjusted to 20m to match the profile of the deposited pattern. The
discrepancy is possibly due to lack of control of mask-to-substrate separation used in the
experiment. Using microscopic analysis we find that the mask with t = 75m and =
270 has surface irregularities on the order of several microns in height such that s > 0,
0 50 100 150 200 250 300
0.0
0.2
0.4
0.6
0.8
1.0

w=100m
t=75m
s=0,20m
=20m
=270
Position (m)
N
o
r
m
a
l
i
z
e
d

t
h
i
c
k
n
e
s
s

(
a
.
u
.
)
Figure 5-15: Thickness profile of an Alq
3
film grown by OVPD through a shadow
mask with an aperture width w = 100m and mask thickness t = 75m. The
simulation (dashed line) was performed with 10
6
Alq
3
molecules assuming s =
20m and mfp = 20m for = 270 mask. Experimental depositions employed
the same type of mask, but without spacers, to obtain Alq
3
patterns having 0.9m
(squares) and 2.1m (circles) maximum thicknesses for w = 100 m. The presence
of pattern dispersion in both runs, and a close resem-blance to the simulated result
indicate inadequate control of the parameter s in the experiment.
0 50 100 150 200 250 300
0.0
0.2
0.4
0.6
0.8
1.0

w=100m
t=75m
s=0,20m
=20m
=270
Position (m)
N
o
r
m
a
l
i
z
e
d

t
h
i
c
k
n
e
s
s

(
a
.
u
.
)
Figure 5-15: Thickness profile of an Alq
3
film grown by OVPD through a shadow
mask with an aperture width w = 100m and mask thickness t = 75m. The
simulation (dashed line) was performed with 10
6
Alq
3
molecules assuming s =
20m and mfp = 20m for = 270 mask. Experimental depositions employed
the same type of mask, but without spacers, to obtain Alq
3
patterns having 0.9m
(squares) and 2.1m (circles) maximum thicknesses for w = 100 m. The presence
of pattern dispersion in both runs, and a close resem-blance to the simulated result
indicate inadequate control of the parameter s in the experiment.

123
contrary to assumptions used in the simulations. Furthermore, uneven clamping of the
mask to the substrate may lead to warping of the mask, and hence to a larger s.
-10 -5 0 5 10
position ( m)
w=6, s=0.5, =45
o
, t=7, mfp=20
w=2, s=0.5, =45
o
, t=3.5, mfp=20
w=2, s=0.5, =45
o
, t=3.5, mfp=40
w=2, s=0.5, =60
o
, t=3.5, mfp=40
w=2, s=0.5, =90
o
, t=3.5, mfp=40


(b)
-10 -5 0 5 10
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4


Position (m)
=270
mfp=20m
s=0.5m
t=3.5m
w=6m
0 30 (m)
Ni mesh
5m 5m
(a)
N
o
r
m
a
l
i
z
e
d

t
h
i
c
k
n
e
s
s

(
a
.
u
.
)
Figure 5-16: a) Comparison of the experimental and simulated pattern profiles of
a patterned Alq
3
film, where the edge sharpness was approximately 1-3 m. Left
inset: An atomic force microscope (AFM) image of a patterned Alq
3
film
deposited by OVPD at 0.1 Torr through a nickel mesh with an aperture width, w =
6m and mask thickness, t = 3.5m, with s <1m. Right inset: SEM side view of
the Ni mesh. b) Simulation results for higher resolution patterns that are predicted
for OVPD at the conditions listed in the figure.
D
e
p
o
s
i
t

h
e
i
g
h
t

(
a
.
u
.
)

124
The highest resolution patterns achieved by OVPD used Mask C with the bottom
of the Ni mesh in close contact with the substrate, resulting in s 1m, due to the profile
of the mesh wires. A patterned Alq
3
film deposited by OVPD at P
dep
= 0.1 Torr was
imaged using atomic force microscopy (AFM) as shown in Fig. 5-16a. The edge
sharpness is ~3m, obtained from the thickness profile extracted from the image in the
left inset of Fig. 5-16a. Simulation of these deposition conditions is also shown as a solid
line. The experimental deposit profile is fit assuming w=6.0m, t=3.5m, s=0.5m,
=20m and =270. The simulated and experimental dimensions are nearly identical,
suggesting that the stochastic simulation accurately describes the patterned deposition
mechanism. Similar patterns have been obtained for depositions at pressures of from 0.1
to 2 Torr, with the lower pressures increasing pattern definition, as expected. Figure 5-
16b shows simulation results for several other deposition conditions. One curve (solid
circles) is for an aperture with = 45 and s = 0.5m, showing improved pattern edge
sharpness compared to that in Fig. 5-16a Rectangular patterns with > 90% can be
achieved for w 2 m and = 90.
5.10 Resolution limits and self-aligned contacts by hybrid VTE-OVPD
Two instances of micron-scale resolution patterning are illustrated by the scanning
electron micrographs in Fig. 5-17: (a) Alq
3
patterns deposited on Si by VTE through Ni
meshes at P
dep
= 10
-6
Torr and s 1m without substrate rotation, and (b) analogous
patterns deposited by OVPD at P
dep
= 2 Torr. The vacuum-deposited patterns show the
trapezoidal profile discussed in Sec. II where < 1m, while OVPD patterns have edge
dispersion on the order of 1-3 m. Simulations in Fig. 5-16b indicate that if apertures

125
10m
10m
10m
(a)
(b)
(c)
VTE growth
OVPD growth
Organic pixel (OVPD)
Metal cathode (VTE)
Hybrid OVPD / VTE growth
Figure 5-17: a) Scanning electron micrograph of an Alq
3
film deposited by
vacuum thermal evaporation on Si through a mesh with an aperture width, w =
7.5m and mask-to-substrate separation, s ~ 0m at 10
-6
Torr. b) Scanning
electron micrograph showing Alq
3
patterns deposited on Si using OVPD at 0.1
Torr through a nickel mesh with t = 3.5m, w = 7.5 m, and s < 1m. c)
Scanning electron micrograph of hybrid OVPD-VTE depo-sition showing Alq
3
patterns deposited on Si by OVPD at 1 Torr, with Ag caps subsequently deposited
by VTE at 10
-6
Torr, without shifting the shadow-mask between depositions.

126
with = 45 and minimal s are used, sub-micron resolution is achievable using this
growth technique.
Finally, fabrication of OLED-based full-color displays entails the deposition of an
array of pixels with one color emissive layer followed by the deposition of a second and
third array of different color pixels, forming the red-green-blue full-color triad.(Tian et
al. 1999; Tian et al. 1997) The deposition of organic layers is followed by metal
deposition to form the cathode contacts to the OLEDs. To confine the metal cathode to
coat only the organic films, and hence prevent shorting of the cathode to the anode
contacts, an integrated polymer shadow mask consisting of photoresist walls is pre-
patterned onto the substrate.(Tian et al. 1997) The integrated shadow mask can be
eliminated by OVPD at moderate pressure through a conventional shadow-mask, yielding
slightly broadened deposits. This is followed by the deposition of a well-defined metal
contact in vacuum without moving the mask between the growth steps. The controlled
dispersion of the organic coating, therefore, prevents electrical shorts around the edge of
the organic film, since VTE of the electrode will automatically yield a smaller size cap
over the organic pixel. An SEM image of the resulting confined electrode structure from
such hybrid deposition is shown in Fig. 5-17c. In the passive-matrix architecture the
confined individual electrode squares are replaced by stripe electrodes. A second set of
pixels may be deposited by simply translating the mask laterally (possibly in-situ),
avoiding the realignment of the mask to a previously deposited pattern.
5.11 Summary
In this chapter we addressed the challenge of in-situ micropatterning the active organic
layers by deposition in the diffusion-limited regime. Since the molecular mean free path

127
of the organic molecules in the vicinity of the substrate is on the order of the apparatus
dimension, (e.g. mask aperture, mask thickness or mask-to-substrate separation),
continuum modeling approaches are inaccurate in describing the deposition. Instead, we
used Monte-Carlo (MC) simulations to more accurately represent the molecular nature of
transport in confined geometries. The quantitative effects of background gas pressure,
aperture geometry, and mask-to-substrate separation were studied. The deposition
experiments confirmed the MC simulations, down to features < 10 m wide. Based on
the close correspondence between the simulation and the experiment, the former were
used to predict that resolutions and features sizes < 1m are achievable at appropriate
deposition conditions and mask geometry, i.e. low background gas pressure, small mask-
to-substrate separations, and a re-entrant aperture profile. Furthermore, the Monte-Carlo
simulations permit the quantitative assessment of the efficiency of molecular transport
through the aperture (quantified by the merit figures and ), which is of practical
importance in device applications.



128
Chapter 6: Organic Vapor Jet
Printing
6.1 Overview
Both OVPD and VTE require the use of shadow-masks to pattern the active organic
layers in the substrate plane. As a consequence, much of the source material is wasted on
coating of the mask. This chapter introduces a novel method of Organic Vapor Jet
Printing (OVJP), which permits the direct patterned deposition (printing) of the organic
semiconductors onto the substrate. Several potential advantages are thus attained the
materials utilization efficiency is nearly 100%, and the masking step is eliminated. In this
chapter the OVJP theory is developed, direct simulation Monte-Carlo models are used to
verify the theory, and printing experiments and results are presented, demonstrating the
patterning and device growth capability of OVJP.
6.2 OVJP Concept
The vapor jet printing concept and apparatus are schematically shown in Fig. 6-1. Carrier
gas enters heated cells containing the molecular organic source material, transporting the
saturated source vapor into a mixing chamber. The gas mixture expands through a
microscopic nozzle as a sub-sonic jet, impinging onto a substrate that can be translated
transverse to the nozzle, thereby generating a patterned deposit. The apparatus also has a

129
dilution channel, allowing for the temperature-independent regulation of the organic
vapor concentration in the jet. Although the carrier gas flow field rapidly diverges due to
the proximity of the substrate to the nozzle outlet, the relatively heavy organic molecules
acquire trajectories substantially more collimated than the carrier gas. The heavier
organic molecules condense on the cooled substrate, while the lighter carrier gas escapes
to the sides.
Although somewhat similar in form, OVJP is nevertheless distinct from
conventional ink jet printing of polymer semiconductors.(Hebner et al. 1998; Paul et al.
2003; Sirringhaus et al. 2000) It uses a hot inert carrier gas, instead of a liquid solvent, to
directly print molecular organic compounds, eliminating meniscus formation, solvent
compatibility issues, and other concerns limiting ink jet printing. Furthermore, in OVJP
no pre-patterning of the substrate is needed, whereas in ink jet printing, droplet-confining
wells are required.
Figure 6-1: OVJP apparatus schematic
A B
N
2
N
2
N
2
Substrate
Source cell
Heater
Nozzle
Jet
Deposit
Figure 6-1: OVJP apparatus schematic
A B
N
2
N
2
N
2
Substrate
Source cell
Heater
Nozzle
Jet
Deposit
Figure 6-1: OVJP apparatus schematic
A B
N
2
N
2
N
2
Substrate
Source cell
Heater
Nozzle
Jet
Deposit

130
6.3 OVJP theory
For application in printing of organic electronic devices such as LEDs or TFTs, the
desired pattern size is on the order of 1-100 m, with the required edge resolution
typically 10% of the pattern width. In the work on thin film patterning using shadow
masks, it was found that the aperture width, 2a, must be on the order of the desired
pattern size, while the aperture-to-substrate separation, s, should be smaller than both the
pattern edge resolution and the molecular mean free path, . For typical deposition
pressures of 0.1-100 Torr used in organic vapor phase deposition (OVPD), is between
500 and 0.5m, respectively. Thus, often s a, resulting in transition regime
transport, between the molecular and continuum flow regimes.(John 1984; Roy et al.
2003)
A similar situation is encountered in OVJP, where the vapor is ejected through
microscopic nozzles placed in proximity to the substrate. Previous work (De La Mora et
al. 1989; Eres 1991; Guevremont et al. 2000; Tison 1993; Vasenkov et al. 1995)
suggested that the freely expanding jet diverges monotonically after exiting the nozzle
and into free space.(Shtein et al. 2001) When a substrate is placed close to the nozzle, the
jet is further dispersed relative to the free-expanding beam due to a stagnation front
immediately above the surface. This occurs because the sudden deceleration of the jet
upon impacting the substrate creates a pressure front, which redirects the flow parallel to
the substrate, causing even greater dispersion of the jet relative to the free expansion
situation. Nevertheless, lateral patterning is still possible, because the heavier organic
species retain most of their momentum normal to the substrate, leading to a deposit that is

131
narrower than the carrier gas plume. The deposition geometry and the axial enrichment of
the jet are illustrated in Fig. 6-2.
The pattern resolution depends upon the nozzle shape, and the interplay of
diffusive and convective processes in the region between the nozzle and substrate.
Important parameters include: nozzle radius (a), nozzle-to-substrate separation (s), the
deposition chamber (or background) pressure (P
L
), and the masses of the carrier gas and
organic molecules (m
cg
, m
p
, respectively). Since no accurate analytical theory exists for
flow in this transition regime, predictions of the pattern shape have to derive primarily
from experiment and direct simulation Monte-Carlo (DSMC) techniques.(Bird 1994)
Nevertheless, a scaling argument can be developed based on simple collision dynamics to
provide guidance as to how these parameters influence the deposit edge resolution.
For a sufficiently long nozzle (L/a > 10, where L is the nozzle length) fully
developed flow is established, where the organic molecules are accelerated by the carrier
u
x

2a
N
2
N
O
Al
3
Figure 6-2: Diagram defining the geo-metry relevant to pattern formation, also
depicting the diverging carrier gas (N2) flow streamlines and the collimated tra-
jectories of heavier organic molecules (in this case, Alq
3
.
u
x

2a
N
2
N
O
Al
3
Figure 6-2: Diagram defining the geo-metry relevant to pattern formation, also
depicting the diverging carrier gas (N2) flow streamlines and the collimated tra-
jectories of heavier organic molecules (in this case, Alq
3
.
u
x

2a
N
2
N
O
Al
3
u
x

2a
N
2
N
O
Al
3
N
O
Al
3
Figure 6-2: Diagram defining the geo-metry relevant to pattern formation, also
depicting the diverging carrier gas (N2) flow streamlines and the collimated tra-
jectories of heavier organic molecules (in this case, Alq
3
.

132
gas to the bulk jet velocity, . If the molecules suffer few collisions en route to the
substrate, they traverse the nozzle-to-substrate gap, s, in:
/ t s u = (7.1)
The organic molecules will be displaced radially outward by the diverging carrier gas
flow as well as by diffusion, by a distance from their original position at the exit of the
nozzle (Fig. 6-2):
x D
t u r = + (7.2)
where u
x
is the radial convective organic molecular velocity, while r
D
is the distance the
molecules travel due to pure diffusion. The radial convective velocity arises from
molecular collisions with the carrier gas:
2
cg
x
x
o
m
s
u u
m
= , (7.3)
where
x
is the carrier gas flow velocity after it is redirected along the substrate plane,
2m
cg
/m
o
accounts for the momentum transferred to the organic molecule in each elastic
organic-carrier molecule collision, and s/ is the total number of collisions occurring
while the molecule is traversing the nozzle-substrate gap. The diffusion travel is:
D
r D t = (7.4)
where D is the diffusivity of the organic molecule in the carrier gas. Assuming
incompressible flow and conservation of mass (i.e.
2
2 x u a u a s = ), and D = 1/3
c , where c is the molecular mean thermal velocity, Eqs. (7.14) combine to give:
2
1
3
cg
o
m
s c s
a m u a


= +

(7.5)

133
The first term in Eq. (7.5) quantifies the pattern dispersion due to horizontal momentum
transfer to the organic molecules from collisions with the diverging carrier gas, while the
second term represents the scaling of the radial diffusion rate to the ballistic transport rate
normal to the substrate.
Although Eq. (7.5) does not predict the deposit shape, it shows the relative
influence of process conditions on the deposited pattern resolution. In particular, using
/ 2
L
kT P = , where is the cross-section of the molecule, the pattern dispersion is
predicted to have a minimum with respect to P
L
, as shown in Fig. 6-3a. This pressure
results in maximum resolution and is in the range of 1-50 Torr for typical OVJP
conditions, (e.g. s 0.2a 20m, m
cg
/m
o
0.05 in a nitrogen carrier, and nozzle
temperature of ~300C). Equation (7.5) also suggests that pattern definition is enhanced
through use of a lighter carrier gas (e.g. He instead of N
2
). Practically, is fixed by the
desired deposition rate via the concentration of the organic vapor. Thus, for a given
nozzle radius a, the remaining adjustable parameters are s and P
L
. The conditions leading
to maximum pattern resolution can be plotted on a process diagram (Fig. 6-3b), where
the operating line dictates values of s for any given P
L
. For example, to maintain high
pattern resolution even at large separation, s, the downstream chamber pressure, P
L
, must
be decreased. The region above the operating line represents printing in the diffusion-
limited regime, while the region below corresponds to convection-limited transport.
This analysis assumes incompressible flow, a situation that is not rigorously true
since both the carrier gas and the organic molecules are first accelerated through the
nozzle, but decelerate upon impact with the substrate. Due to the abrupt change in the
flow direction upon exiting the nozzle and the resulting kinetic impact of the jet on the

134
substrate, the local dynamic pressure in the region between the nozzle and the substrate
generally exceeds P
L
. Since the effective pressure in the flow stream is always higher
than the chamber pressure, P
L
, the negative-slope branch of the curve in Fig. 6-3a is not
experimentally observable, and consequently the pattern dispersion curve should have a
positive slope for all P
L
. Thus, for any chosen value of s, the optimum resolution is
obtained by minimizing P
L
.
6.4 Simulation of transitional flow regime
Since the molecular mean free path is on the order of nozzle to substrate distance, s, or
nozzle diameter, a, Navier-Stkes (continuum flow) equations do not accurately describe
the flow pattern downstream of the nozzle. In this case, a direct simulation Monte-Carlo
Figure 6-3: a) Plot of the expected dependence of the normalized deposit width
(s/a) vs. the downstream pressure (P
L
). The dispersion is minimized at a given
value of P
L
(indicated by the arrow) due to the counterbalance of convective and
diffusive transport rates. b) The conditions for the highest pattern resolution
(minimum dispersion) are plotted to give the optimum operating line. Working
above or below this line will decrease the pattern resolution; high s and P
L
result
in diffusion-controlled transport, while low s and P
L
result in convection-
controlled transport.
s
a
P
L
a

P
L
Ideal
operating
line
Optimum P
L
~P
L
~ 1/P
L
(a) (b)

135
(DSMC) approach is useful to model the flow field. Thousands of molecules are moved
quasi-simultaneously in discretized space, allowing for molecule-molecule and molecule-
wall collisions. Boundary conditions include upstream and downstream pressures (P
H
and
P
L
, respectively), organic molecule concentration (C
o
), nozzle and substrate temperatures
(T
N
and T
S
), and sticking coefficients (0 for the heated nozzle, and 1 for the substrate).
The simulation geometry is shown in Fig. 6-4a. The nozzle was nominally 20m
in diameter by 50m in length, slightly fluted at the downstream aperture, and maintained
at 300C, a temperature sufficiently high to avoid readsorption of the volatilized organics
onto the nozzle walls. The substrate was positioned 25m below the nozzle, and
maintained at 25C. Several downstream pressures were investigated: 0.24, 2.4, 24 and
240 Torr, with the corresponding upstream nitrogen pressure set to P
H
= 400, 600, 800,
and 900 Torr, respectively. A 1% molar upstream concentration of Alq
3
was assumed.
Figure 6-4a is a color-map of the vertical velocity component of the calculated
flow field, with the corresponding trajectories of the carrier gas and the organic
molecules shown in Fig. 6-4b. The velocity map shows that the molecules are accelerated
along the nozzle axis, reaching near sonic velocities of ~200m/s at the orifice.
Immediately above the substrate surface, the gas molecules decelerate, giving rise to a
stagnation front, where the dynamic pressure exceeds the ambient pressure (P
L
) far away
from the nozzle. The heavy organic molecules, however, follow more collimated
trajectories, crossing the carrier gas flow lines. The heavier molecules thus penetrate the
stagnation front and impinge on the substrate in a well-defined adsorbed deposit pattern.
The deposit profiles obtained from DSMC for different printing conditions are plotted in
Figs. 6-5a and b, where the broadening of the deposit due to increasing s and P
L
is

136
(b)
Position (m)
0 50 100
150
Alq
3
trajectories
N
2
flow field
(a)
0
50
Nozzle
Substrate
Vapor jet
D
i
s
t
a
n
c
e

(

m
)
200 100 0 -50
z-component of Velocity (m/s)
Figure 6-4: a) Color-map of the z-component (i.e. perpendicular to the substrate)
of the flow velocity of the carrier gas, showing the collimation of the jet emerging
from the nozzle and the stagnation front just above the substrate surface. b) The
corresponding flow field of the carrier gas (red lines) and the trajectories of
heavier organic Alq
3
molecules (blue lines) are shown. The plots were obtained
from direct simulation Monte-Carlo (DSMC) modeling of OVJP using a nozzle
diameter a = 20 m, nozzle length L = 100 m, nozzle-to-substrate separation s =
30 m, upstream pressure P
H
= 0.24 Torr, and downstream pressure P
L
= 0.24
Torr.

137
-100 -50 0 50 100
0.2
0.4
0.6
0.8
1.0


T
h
i
c
k
n
e
s
s

(
n
o
r
m
a
l
i
z
e
d
)
Position (m)
0.24 Torr
2.4 Torr
24 Torr
240 Torr
Increase
P
L
(a)
-50 0 50 100
0.0
0.2
0.4
0.6
0.8
1.0


T
h
i
c
k
n
e
s
s

(
n
o
r
m
a
l
i
z
e
d
)
Position (m)
25 m
50 m
75 m
100 m
Increase
s
(b)
Figure 6-5: a)The deposit profiles are calculated by DSMC of Alq
3
deposition
with s = 25m, showing a monotonic broadening of the deposit width with
increasing P
L
. This indicates that the dynamic pressure at the nozzle exit
dominates transport, forcing the system away from the dispersion minimum, as
shown in Fig. 1b. b) Deposit profiles are calculated at P
L
= 0.24 Torr for s = 25,
50, 75 and 100 m. The pattern width increases monotonically with s. The solid
lines are drawn as a guide to the eye.

138
evident. Pattern dispersion increases with both s and P
L
. The pattern width is nearly
independent of P
L
when > s, but then increases rapidly with P
L
in situations when < s.
High pattern resolution is achieved when the nozzle is placed within a mean free path
from the substrate, confirming the dynamic pressure effect discussed earlier. The center
of the calculated deposit profile is somewhat domed for all values of s and P
L
, while the
wings extend beyond the 20m nozzle orifice diameter. One means to achieve a deposit
profile with a flat top and a sharp edge, a small-diameter nozzle can be rastered, or
dithered at high frequency over the substrate during the deposition.
6.5 Experimental set-up
The OVJP apparatus consists of several components mounted on a six-way cross, as
shown in Fig. 6-6a. The water-cooled copper substrate holder is positioned in the center
of the cross by means of a computer-controlled XYZ-motion stage. Substrates are first
mounted ex-situ onto a copper plate, which is loaded by hand through the hinged view-
port and attached onto the holder by means of integrated spring clips. The pressure in the
chamber is regulated by a combination of a roughing pump (up to 20lpm pumping
speed), throttled by an electronically controlled butterfly valve, and metered nitrogen
or helium inflow. The source materials are contained within 5 source cells, positioned
inside of a uniformly heated stainless steel cylinder (Fig. 6-6b).
The source cells (Fig. 6-6c) consisted of a specially machined hollow 0.375-inch
diameter by 1-inch long stainless steel cylinder. The cells were attached to one end of a
hollow stainless steel tube, which transports the carrier gas and acts as the rotating shaft
of a hot-valving arrangement, as indicated in Fig. 6-6c. The source cylinder also includes

139
a central dilution channel, a mixing chamber, and a modular nozzle. Nitrogen and helium
are used as the carrier gases.
Two types of nozzles were used, illustrated in Fig. 6-7. The first (Type I) was a
conical cap stainless steel laser-drilled aperture, having a nominal inside diameter of
20m and a channel length of 100m. The second (Type II) nozzle was a polyimide-
coated hollow silica capillary, 4mm long, with 50m and 350m inner and outer
diameters, respectively. The capillary was press-fit into a 350m diameter hole in the
source block. The substrates were mounted onto a water-cooled susceptor attached to a
computer-controlled XYZ-positioning stage. The nozzle-to-substrate distance was
controlled by first moving the stage into contact with the nozzle at three locations,
(detected optically), and subsequently translating the stage from these reference positions.
The substrates used in the patterning experiments consisted of silicon wafers,
precleaned using a procedure described elsewhere.(Shtein et al. 2002) The pattern shapes
were determined by reflected-light interference microscopy.(Shtein et al. 2003) The
substrates used in the printing of the pentacene TFTs consisted of n-type doped (0.001 -
cm) (100) Si wafers with coated with a dry thermally grown, 2100 thick SiO
2
layer
surface treated with octadecyl-trichlorosilane (OTS). The OTS treatment consisted of
sequentially exposing the SiO
2
wafers to UV-ozone for 15 minutes, and saturated OTS
vapor at 20C and 0.1 Torr for 10 minutes immediately prior to printing the pentacene
channel. The 99% pure pentacene powder was further purified twice by vacuum train
sublimation(Forrest 1997) before deposition. The TFTs had gold source and drain
electrodes deposited in vacuum through a shadow mask, while the silicon substrate

140
Source Cylinder
Dilution
channel
Mixing
Chamber
Source cell
Modular
nozzle
Mass flow
controllers
N
2
/ He
N
2
/ He
Roughing
Pump
XYZ-stage
6-way
cross
Hinged Door /
View-/Load-Port
Substrate holder
w/ cooling lines
Set screw
S.S. Tube
Source
Material
(e.g. Alq
3
)
Qrtz Wool
Inlet Channel
Outlet Channel
S.S. Source Cell
N
2
/ He
Rotate for
hot valving
Rotate for
hot valving
Substrate
Throttle
valve
Exhaust
Cold
Trap
(a)
(b)
(c)
Figure 6-6: a) OVJP experimental set-up, with the water-cooled substrate
attached to a computer-controlled xyz-movable stage, a multi-source print-head,
and electronically regulated gas flows and exhaust. b) Schematic of the print-
head, containing several source cells, a mixing chamber, and a modular nozzle. c)
Schematic of one source cell, showing assembly and hot-valving operation.

141
served as the gate electrode. The TFTs were tested in ambient, in the dark, using a
Hewlett-Packard 4155 semiconductor parameter analyzer.
6.6 Direct printing of patterned molecular organic thin films
Figure 6-8a shows a row of Alq
3
dots deposited at P
L
= 0.24 Torr using nitrogen as the
carrier gas along with a Type I nozzle, an Alq
3
source temperature of 270C, and a
substrate temperature of 15C. The distance between the nozzle and the substrate was
varied, starting from approximately s = 0, and increasing in ~28m increments for each
dot. For sufficiently thick deposits, optical interference fringes obtained at a wavelength
Figure 6-7: Nozzles used in the OVJP experiment. a) An example of a conical cap
stainless steel screw with a laser-drilled aperture (here shown with a channel
100m long x 20m diameter). b) Polyimide-coated silica capillary, inserted into
a stainless steel set screw. (Shown here with a 350 and 50m outer and inner
diameters (OD and ID) respectively. The ID of the hole in the set screw is several
microns smaller than the OD of the capillary. The polyimide coating is scraped off
and the capillary is press-fit into the opening. The channel length is defined by
cleaving off appropriate lengths of the capillary.)
50m
Proximity sensor
Au wire
Nozzle
350m
350m
Capillary press-fit
into set screw
Nozzle Type I Nozzle Type II
a)
b)

142
500m
(a)
T
h
i
c
k
n
e
s
s

(

m
)
(b)
Position (m)
s
1
s
2
s
3
s
4
s
5
s
6
-100 -50 0 50 100
0.0
0.5
1.0
1.5
2.0
2.5
3.0


s
2
s
3
s
4
s
5
s
6
Increase
s
Figure 6-8: a) Optical micrograph of Alq
3
dots printed on Si using nitrogen carrier
gas, and a Type II, 20m diameter nozzle at P
H
= 240 Torr and P
L
= 0.24 Torr in a
combinatorial experiment with varying nozzle-to-substrate separation, s = (0, 28,
57, 85, 114, 142 10) m. The deposit thickness profiles are deduced from the
optical interference fringes (see Ref.17). b) Thickness profiles of the Alq
3
dots in
(a) determined from the interference fringes obtained under 540nm wavelength
illumination. The plot of the deposit profiles shows a decrease in pattern resolution
with increasing s.

143
-100 -50 0 50 100
0.0
0.5
1.0
1.5
2.0
2.5
3.0


T
h
i
c
k
n
e
s
s

(

m
)
Position (m)
300 m
s
1
s
2
s
3
s
4
s
5
s
6
s
7
s
8
s
9
s
10
(a)
(b)
Increase s
Figure 6-9: a) Optical micrograph of Alq
3
dots printed on Si using helium
carrier gas, and a Type II nozzle at P
H
= 240 Torr and P
L
= 0.24 Torr in a
combinatorial experiment with varying nozzle-to-substrate separation, s = (0,
15, 30, 45, 60, 75, 90, 105, 120, and 135 10) m. b) Thickness profiles of the
Alq
3
dots in (a) determined from the inter- ference fringes obtained as in Fig. 4.
The plot of the deposit profiles shows a decrease in pattern resolution with
increasing s.

144
of 540nm allow the deposit profile to be determined.(Shtein et al. 2003) The
corresponding thickness profiles in Fig. 6-8b exhibit similar shapes and dimensions to
those obtained by DSMC. This experiment was repeated with helium as the carrier gas,
while all other conditions were kept constant. The resulting Alq
3
deposits are shown in
Fig. 6-9a, with the corresponding thickness profiles plotted in Fig. 6-9b. The shape of the
deposits is similar to that in Fig. 6-8, albeit with visibly higher resolution. One
convenient method of quantifying the printing resolution is through the full-width half-
maximum (FWHM) of the bell-shaped deposit profile.
0 50 100 150 200
0
50
100
150
200

Alq
3
+ N
2
Alq
3
+ He
F
W
H
M

(

m
)
Separation (m)
Figure 6-10: Plot of the FWHM vs. s from Figs. 6-8b and 6-9b, showing that
using a lighter carrier gas (here, He instead of N
2
) can increase the collimation of
the beam and a sharper deposit.

145
Figure 6-10 shows a plot of the FWHM versus s for the deposits in Fig. 6-8. The
data, (open circles), are fit by a line corresponding to Eq. (6.5). Since a, m
cg
/m
o
, and c
(evaluated at the nozzle temperature) are known, the fit employs only two parameters,
and , where is a function of the dynamic pressure (P
dyn
) above the substrate. From the
fit, we obtain = 150 m/s and P
dyn
= 75 Torr (with P
dyn
calculated from evaluated at
145C, corresponding to the average of the nozzle and substrate temperatures). These
values of and P
dyn
are in general agreement with the simulation results. Since the fit is
approximately linear with s, the first term in Eq. (6.5) dominates, meaning that transport
is limited by molecular collisions in the nozzle-to-substrate gap.
The plot of the deposit FWHM versus s from Fig. 6-9 is also shown in Fig. 6-10.
The data exhibit a shallower slope than that obtained from the N
2
carrier gas experiment,
due to the lower m
cg
/m
p
ratio. However, the slope decrease is less than predicted by the
mass ratio change alone. This can be understood in terms of sub-maximal acceleration of
the Alq
3
molecules by the He jet, since the nozzle aspect ratio, L/2a 5, is insufficient to
establish fully developed flow. Indeed, the flow is also not fully developed in the case of
the Alq
3
+N
2
mixture either. Nevertheless, even though the equilibrium thermal velocities
of the Alq
3
and N
2
in the upstream reservoir are closer than in the case of Alq
3
+He, fewer
N
2
-Alq
3
collisions are required for full acceleration of Alq
3
than He-Alq
3
collisions.
Hence, the latter situation results in considerably smaller overall acceleration of the Alq
3

during transport along the nozzle length.
The pressure-dependence of the deposit width was also investigated. Figure 6-
11a shows Alq
3
dots printed on silicon at P
L
= 0.27, 1.0, 3.0, 30, 100, and 760 Torr using
N
2
, at a separation of 25m, using the Type II nozzle. The FWHM for each pressure is

146
0.1 1 10 100 1000
0
50
100
150
P
7
P
6
P
5
P
4
P
3
P
2
P
1
500m
F
W
H
M

(

m
)
(a)
(b)
Figure 6-11: a) Optical micrograph of Alq
3
dots printed on Si using nitrogen
carrier gas, and a Type II nozzle at s = 25 10 m in a combi-natorial
experiment with varying downstream pressure, 0.24 < P
L
< 900 Torr. b) The
FWHMs of the deposits are plotted versus P
L
, showing an increase in the pattern
width with pressure (solid line, filled squares). The DSMC results from Fig. 6-5b
are also re-plotted for comparison (dashed line, open circles). Although the maxi-
mum resolution is somewhat greater at lower pressures in the simulated profiles,
the experi- mental and simulation trends cross at P
L
20 Torr. The flat FWHM at
low P
L
, and the rapid increase at P
L
>200 Torr of the experimental data is
attributed, in part, to the roughness and irregularity of the Type II nozzle used in
the experiment (see text).
P (Torr)

147
plotted in Fig. 6-11b (squares, solid line). The simulation results from Fig. 6-5b are also
provided for comparison (open circles, dashed line). The simulation shows a monotonic
increase in the pattern dispersion with increasing pressure due to the dynamic pressure
a)
c)
1.5mm
b)
Figure 6-12: A 24x32 pixel bitmap image of a bicyclist figure printed by OVJP
using Alq
3
, with the nozzle of 10m diameter by 100m length, nozzle-to-
substrate separation of (20 10) m, a dwell-time of 2s above each pixel
location, and <0.2s time interval for translation between each pixel. The pattern
resolution in this image varies between 500 and 1000 dpi due to the variation in
the nozzle-to-substrate separation across the image. The local Alq
3
deposition
rate was 1300 /s at 270C source temperature, and can be increased to >8000
/s at 300C source temperature, or >18000 /s if He is used instead of N
2
as
the carrier gas. The image was produced by computer-controlled motion of the
susceptor, while the nozzle was held steady.

148
effect, as discussed in Sec. 6-2. The experimental data exhibit a similar behavior,
although with a much sharper increase in FWHM at P
L
100 Torr. It is possible that to
accurately simulate transport at P
L
>100 Torr, the simulation needs to be run for a longer
time period to achieve true steady state representation of the flow profile. Thus, the
simulated P
dyn
is lower than experimental, and the simulation consequently underpredicts
the extent of dispersion.
Figure 6-12 shows a pattern of Alq
3
(an archetypal electron transport material
used in OLEDs) printed by OVJP using a 20m diameter nozzle under process conditions
indicated in the caption. The dwell-time of the nozzle above each pixel was 2s, while the
stage moved between pixels in <0.2s. To simplify the printing process, the source was not
turned off during translation between pixels. A magnified portion of the pattern is shown
(Fig. 6-12c), indicating 500-1000 dots per inch (dpi) resolution, which is similar to that
achieved by IJP of polymer-based TFTs.(Paul et al. 2003) The source temperature was
270C, resulting in a local Alq
3
deposition rate of r
dep
1300 /s. Our result suggests
that a 300C Alq
3
source temperature should result in r
dep
> 8000 /s.(Shtein et al. 2001)
At this growth rate, an array of 800 nozzles can print an SVGA resolution display
(600x800 OLED pixels) in under one minute. This speed is comparable to the current
state-of-the-art inkjet printers, which also use print heads containing in excess of 500
nozzles.
6.7 OVJP of polycrystalline pentacene films and TFTs
Growth of polycrystalline films by OVJP results in an interesting anisotropy which is
closely associated with the non-equilibrium, directional nature of the OVJP process.

149
Molecules condense and diffuse on the substrate surface, nucleating clusters, which then
grow by addition of both surface-diffusing species and molecules arriving at the substrate
from above. The islands do not all grow uniformly at once; since the organic molecules
arrive at the surface at an angle dictated by the flow field, the mutual shadowing of the
islands during growth results in a tilted grain aspect, determined by the incident
molecular velocity vector. This growth process is illustrated in Fig. 6-13a. Figure 6-13b
shows a micrograph of a pentacene pattern printed on SiO
2
using the Type II nozzle at a
local deposition rate >300 /s and s = (35 15)m, showing a continuous line printed by
scanning the nozzle over the substrate during the deposition. Scanning electron
micrograph (SEM) images reveal that indeed the pentacene crystallites to the left and
right of the jet centerline tilt in toward the molecular supply. This effect is not observed
in diffusion-limited growth such as in OVPD,(Shtein et al. 2002) and is in agreement
both with the schematic of Fig. 6-2a as well as the DSMC modeling results in Fig. 6-4b.
Somewhat surprisingly, the very high deposition rates achieved in OVJP can still
result in highly ordered crystalline morphologies. This is indicated by x-ray diffraction
patterns in the case of pentacene printed by OVJP, as shown in Fig. 6-13c. Here, up to
four orders of two main peaks are readily observable for pentacene printed on SiO
2
at
rates >700/s. As will be discussed in Ch. 7, the two peaks correspond to the thin film
and bulk phases of pentacene typically observed in thin films, 15.5 and 14.5 plane
spacing, respectively. However, vacuum-deposited pentacene films are typically
amorphous when grown at rates >10/s.
Enhancement of molecular ordering in seeded molecular beams has been
observed previously. Seeding the organic molecules in a fast-flowing carrier stream

150
(c)
5 10 15 20 25 30
(
0
0
1

)
(
0
0
1
)
(
0
0
2

)
(
0
0
2
)
(
0
0
3

)
(
0
0
3
)
(
0
0
4

)
(
0
0
4
)
2
S
i
g
n
a
l

(
c
o
u
n
t
s
)
Figure 6-13: a) Schematic illustration of polycrystalline film growth by OVJP,
where the angled approach velocity of each molecule, and the mutual shadowing of
growing crystallites can lead to tilted grain growth. b) Micrograph of a pentacene
pattern printed by OVJP on SiO
2
, with scanning electron micrographs below
showing pentacene crystallites tilting in the direction of the carrier gas flow on each
side of the printed line. c) X-ray diffraction pattern of pentacene deposited on OTS-
treated SiO2 at 700/s, showing two peaks indicative of the thin- film and bulk
phases of pentacene, corresponding to 15.5 and 14.5plane spacing, respectively.
(b)
500m
1m 1m
(a)

151
Figure 6-14: a) Scanning electron micrographs of the channel and contact regions
of the pentacene TFT printed by OVJP. The bright layer on top of the
microcrystalline pentacene film is gold. b) Plot of I
DS
versus the drain-source
voltage (V
DS
), showing transistor saturation. c) Drain-source current (I
DS
) versus
gate-source voltage (V
GS
) response of a pentacene channel thin- film transistor
(TFT) printed by OVJP. The characteristic was obtained from the drain-source
current (V
DS
) in the saturation regime (at V
DS
= -40V). The TFT exhibited some
hysteresis in I
DS
, with the threshold voltage, V
T
, shifting from +10 to +17 V in the
forward and reverse V
GS
sweep directions, respectively, as indicated.
Contact region
Channel region
-40 -30 -20 -10 0 10

0
1
2
3
4
5
-40 -30 -20 -10 0
-16
-14
-12
-10
-8
-6
-4
-2
0
2
V
DS
I
D
S
(

A
)
I
D
S
(
A
m
p
s
)
(
I
D
S
)
1
/
2
(
A
m
p
s
)
1
/
2

1
0
3
(a) (b)
(c)
V
GS
(Volts)
10
-11
10
-10
10
-9
10
-8
10
-7
10
-6
10
-5
10
-4

152
allows near- to hyper-thermal velocities (i.e. greater than the mean molecular speed of the
heavier molecules at that temperature) to be reached by the adsorbate. Consequently, the
ability to tune the incident molecular kinetic energy via the carrier gas stream allows to
partially decouples the film crystallization dynamics from surface diffusion effects, likely
leading to highly ordered films, even for relatively cold substrates.(Casalis et al. 2003;
Scoles 1988) This effect can be potentially exploited for improving the performance of
devices, such as polycrystalline channel TFTs.(Shtein et al. 2002)
To demonstrate the utility of the very high local deposition rates characteristic of
OVJP for device applications, the channel region of a pentacene TFT was printed to form
a 6mm x 6mm, uniformly filled square by rastering the narrow jet over the substrate at P
L

= 0.2 Torr, and at a local growth rate of 700/s. The transistor drain-source current (I
DS
)
versus drain-source voltage (V
DS
) characteristic is plotted in Fig. 6-14b, showing that I
DS

saturates, as observed previously for vacuum and OVPD grown pentacene TFTs (Baude
et al. 2003; Gundlach et al. 1997; Shtein et al. 2002). The I
DS
vs. gate-to-source bias
(V
GS
) characteristic is plotted in Fig. 6-14c, indicating an I
DS
on/off ratio of 7x10
5
, and a
channel field-effect hole mobility of
eff
= (0.25 0.05) cm
2
/Vs in the saturation regime.
The channel hole mobility of a vacuum-deposited control TFT was within the
experimental error of the values obtained by OVJP at P
L
= 0.2 Torr.
A pentacene TFT was also printed in nitrogen at atmospheric pressure (P
L
= 760
Torr). The TFT exhibited
eff
= 0.2 cm
2
/Vs but a substantially lower current on/off ratio
of 25, caused mainly by the high off current. The higher off current can be explained by
several factors. The atmospheric pressure-printed pentacene channel was much thicker
(~1m), making the channel region more conductive in the absence of the field effect.

153
Also, higher pressure in the deposition chamber increases conductive and convective heat
transfer from the nozzle assembly to the substrate, which may disrupt the OTS self
assembled monolayer, or the molecular ordering of pentacene on OTS, which has been
previously observed to reduce device performance (Shtein et al. 2002). Chapter 7
discusses in greater detail the relative effects of film morphology and substrate treatment
on the electrical device performance of pentacene TFTs.
6.8 Summary
In this chapter, we introduced a new method for the direct, solvent-free, patterned
deposition of molecular organic semiconductors organic vapor jet printing (OVJP). A
semiquantitative theoretical model was developed to predict the influence of process
conditions (e.g. deposition pressure, gas flow rate), materials (e.g. relative masses of the
carrier gas and the organic), and apparatus geometry (e.g. nozzle shape) on the deposit
shape and resolution. This model was verified using a direct simulation Monte-Carlo
(DSMC) approach, where the trajectories of many molecules are tracked simultaneously
in space as they are expanded through the nozzle and deposit on the substrate.
Furthermore, a first experimental OVJP apparatus was designed and constructed,
allowing for experimental verification of both the model and the computer simulations. A
pentacene thin-film transistor was deposited, where despite the ultra-high local deposition
rates (>700 /s), molecular order was high and the device electrical performance was
comparable to that in TFTs deposited at much lower rates (<1/s) using other methods.



154
Chapter 7: Growth of pentacene films
and thin film transistors
7.1 Overview
Transistors based on organic semiconductors offer a potentially low-cost alternative to
silicon, with greater latitude in the choice of substrates, and adequate device performance
characteristics. In particular, pentacene-channel thin-film transistors (TFTs) have
received considerable attention in recent research in organic electronics with regard to
applications in low-cost electronics (e.g. radio-frequency identification tags (Baude et al.
2003), e-textiles (Boderover et al. 2004)) and LCD-backplane driving circuits (Gundlach
et al. 1999; Kymissis et al. 2001). Pentacene channel TFTs have been demonstrated with
room-temperature field-effect hole mobilities (
eff
) on the order of 1 cm
2
/Vs, and drain
current on-off ratios >10
6
, matching the performance of amorphous silicon transistors
(Gleskova et al. 2001; Wu et al. 2001), though often at a lower substrate temperature
during processing, enabling their deposition on flexible plastic substrates.
(Dimitrakopoulos et al. 2001; Rogers et al. 2002)
It is observed that improved molecular order and reduced density of grain
boundaries can result in improved charge transport in both polymeric (Shaked et al.
2003) and molecular (Shtein et al. 2002) organic semiconductors, similar to what is

155
observed for silicon (Pangal et al. 2001; Wu et al. 2001), where the field effect mobility
is typically on the order of 10
3
, 10-10
2
, and 0.1-1 cm
2
/Vs in single-crystal,
polycrystalline, and amorphous silicon, respectively. Hence, the candidate fabrication
methods for organic TFTs should enable the control of the deposited film morphology
and, thereby, enhance the electrical characteristics of the organic TFTs.
Polycrystalline pentacene channel TFTs exhibiting field-effect mobility up to 2
cm
2
/Vs have been fabricated by a number of techniques, including crystallization from
solution, (Brown et al. 1997) organic molecular beam deposition (OMBD),
(Dimitrakopoulos et al. 1996) vacuum thermal evaporation (VTE), (Gundlach et al.
1997) vapor phase growth of single crystals, (Kloc et al. 1997) and OVPD (Shtein et al.
2002). This chapter focuses on OVPD, in which the growth rate, substrate temperature
and carrier gas pressure can affect the deposited film morphology, thereby allowing the
study of how crystalline film order relates to the device properties.
7.2 TFT geometry and operation
The theory of charge transport in organic materials is still a topic of vigorous research.
Nevertheless, many of the concepts of classical semiconductor theory are still helpful in
understanding the electrical behavior of organic semiconductors. Therefore, it will suffice
to review the basic structure and operating principles of a TFT based on the traditional
energy band model of semiconductors.
A thin-film field effect transistor is shown schematically in Fig. 7-1. An electrical
current, I
DS
, flows in the channel of the thin film semiconductor, under a voltage, V
DS
,
applied between the drain and source electrodes. The electrical conductivity of the
channel is modulated by the voltage V
GS
on the gate electrode. Consider an energy band

156
diagram for a traditional metal-oxide-semiconductor (MOS) junction, as shown in Fig. 7-
2; the semiconductor is p-type. When the three different materials are brought into
contact and thermal equilibrium is established, the Fermi levels (i.e. the chemical
potential of the electrons) align. The conduction and valence bands must bend to
accommodate this, resulting in band offset and a built-in electric field near the
semiconductor-organic interface. Positive charges (holes) therefore accumulate at the
oxide-semiconductor interface (Fig. 7-2a). When a negative voltage is applied to the
metal contact, more holes are drawn to the oxide-semiconductor interface, forming a thin
accumulation region of thickness x
d
(Fig. 7-2b). The extra charge accumulated at the
interface effectively increases the local doping concentration and the number of charge
carriers. Upon applying a voltage perpendicular to the gate-oxide-semiconductor stack,
current can flow along the semiconductor-oxide interface (Fig. 7-2c). The greater the
Source Drain
Gate
Insulator
Semiconductor
L
W
t
ox
V
DS
V
GS
Source Drain
Gate
Insulator
Semiconductor
L
W
t
ox
V
DS
V
GS
Figure 7-1: Schematic of a thin-film field-effect transistor (TFT), where current
between the source and drain electrodes is modulated by the electric field in the
semiconductor channel, which is established by means of a voltage applied to the
gate electrode.

157
applied gate voltage, the greater the charge density in the channel, and the higher the
current; this is known as the field effect.
The I
DS
vs. V
DS
characteristic of the transistor is plotted in Fig. 7-2d. The initial
increase in I
DS
is linear with V
DS
. With current obeying Ohms law, this is known as the
linear regime. Here, Ohms law is (Sze 1969):
| |
=
|
\ .
DS eff DS
W
I q V
L
(7-1)
Figure 7-2: a) Energy band diagram of the gate metal-oxide-semiconductor
junction in a TFT depicted in Fig. 7-1 at zero gate bias (V
GS
). b) The same junction
under V
GS
. c) Perspective of the junction in (b) when a drain-source voltage, V
DS
, is
applied, causing current to flow in the direction indicated by the arrow. d) Typical
current-voltage response of a field-effect transistor.
Metal
Gate
E
F
E
F
V
G
= 0
Oxide
Metal
Gate
E
F
E
F
E
V
E
C
E
F,i
V
G
< 0
Oxide
x
d
Metal
Gate
E
F
Oxide
x
d
V
G
< 0
V
DS
> 0
Source
Drain
V
DS
L
(a) (b)
(c)


Incr. (-V
GS
)
I
D
S
(

A
)
V
DS
0 -40
0
-100
-50
-20
Linear
Regime
Saturation
(d)
Figure 7-2: a) Energy band diagram of the gate metal-oxide-semiconductor
junction in a TFT depicted in Fig. 7-1 at zero gate bias (V
GS
). b) The same junction
under V
GS
. c) Perspective of the junction in (b) when a drain-source voltage, V
DS
, is
applied, causing current to flow in the direction indicated by the arrow. d) Typical
current-voltage response of a field-effect transistor.
Metal
Gate
E
F
E
F
V
G
= 0
Oxide
Metal
Gate
E
F
E
F
E
V
E
C
E
F,i
V
G
< 0
Oxide
x
d
Metal
Gate
E
F
Oxide
x
d
V
G
< 0
V
DS
> 0
Source
Drain
V
DS
L
(a) (b)
(c)


Incr. (-V
GS
)
I
D
S
(

A
)
V
DS
0 -40
0
-100
-50
-20
Linear
Regime
Saturation
(d)

158
where W/L is the channel width-to-length ratio,
eff
is the effective mobility of charges in
the channel (expressed in cm
2
/Vs), and |q| is the magnitude of the charge density in the
channel per unit area.
As V
DS
increases and approaches V
GS
, the magnitude of the electric field near the
drain contact vanishes, and x
d
approaches zero. The conduction channel is said to pinch-
off, limiting I
DS
; this is known as the saturation regime. Transistors are typically
operated in the saturation regime because I
DS
is independent of V
DS
, and is modulated
only by changing V
GS
.
Some of the key performance characteristics of the TFT include the ability to
modulate the drain-source current (I
DS
) by variations in V
GS
(known as the
transconductance, or g
c
= dI
DS
/dV
GS
), the magnitude of I
DS
in the on and off states of
the device (I
DS
on/off ratio), and the effective speed of charge carriers in the channel
(expressed through
eff
). Due to the built-in potentials between the layers comprising the
FET, it may be necessary to overcome a threshold voltage (V
T
) to induce charge
accumulation in the channel. In the saturation regime, the current is given by (Sze 1969):
( )
2 1
2
DS ox eff GS T
W
I C V V
L
= (7-2)
where C
ox
is the capacitance of the oxide, given by C
ox
= 4
0

ox
/ t
ox
, where t
ox
is the
oxide thickness.
As will be shown below,
eff
can depend on the morphology of the semiconductor
channel. This is intuitively evident in view of the discussion in Ch. 1, where molecular
order determines the ease of charge hopping between molecules. Furthermore, grain
boundaries are regions where semiconductor molecules are absent or abruptly change
their relative orientations, thereby disrupting charge transport between grains. Since

159
charge transport in a TFT occurs along the semiconductor-insulator interface, the energy
distribution of charge traps at the interface is also likely to affect I
DS
. The depth and
density of these traps, in turn, depend on the chemistry of the interface, and parameters
such as V
T
and
eff
will depend on the surface treatment of the insulator prior to the
deposition of the organic semiconductor. (See Sec. 7.5 and 7.6).
Figure 7-3 illustrates the two thin film transistor (TFT) architectures studied in
this work. In the top contact geometry, the source/drain contacts are deposited onto the
semiconductor, while in the bottom contact TFT, the semiconductor is deposited onto
the source/drain contacts pre-patterned onto the insulator, followed by deposition of the
semiconductor. The "bottom contact" configuration permits the use of conventional
photolithography to define the channel W/L ratio, thereby potentially allowing the
increase of I
DS
through device geometry modification. However, as will be shown, the
+ + + + + + + + + + + +
- - - - - - - - - - - - - - - -
Source
Drain
Insulator
Gate
Substrate
Channel
Figure 7-3: a) Top-contact geometry of a MOS type FET. b) Bottom-contact
geometry TFT. The material used for each component in this study is labeled.
Si
SiN
x
or SiO
2
Au
Pentacene
(a)
(b)
Pentacene

160
heterogeneous region near the contacts induces morphological irregularities in the
deposited pentacene thin films that become more pronounced for smaller values of L,
partially negating the benefits of shorter channels and masking the influence of the
channel morphology on device performance. After demonstrating these "contact effects,"
we adopt the "top contact" geometry for studying the effects of pentacene morphology
and dielectric surface treatment.
7.3 Growth of polycrystalline pentacene thin films from the vapor phase
7.3.1 Qualitative description of vacuum and vapor phase growth
In OVPD, the growth of polycrystalline pentacene thin films proceeds in the following
sequence:
1) Convective transport of pentacene vapor using a carrier gas (e.g. nitrogen) to
the edge of the boundary layer;
2) Diffusion of pentacene molecules across the boundary layer;
3) Adsorption of pentacene onto the substrate surface;
4) Diffusion of pentacene on the substrate surface;
5) Desorption of pentacene from the substrate back into the boundary layer.
In contrast, deposition in vacuum does not involve gas phase diffusion. Instead, it
proceeds by a simplified process of:
i) Ballistic transport of pentacene molecules from the source to the substrate;
ii) Adsorption of pentacene onto the substrate surface;
iii) Diffusion of pentacene on the substrate surface;
iv) Desorption of pentacene from the surface into vacuum.

161
Convective transport of organic vapor from the source to the substrate in OVPD
was described in Ch. 3. Here, we concentrate on the gas and surface diffusion processes
relevant to polycrystalline film growth from the vapor phase. Figure 7-4 illustrates the
temperature profiles and the corresponding pentacene concentration profiles for growth in
vacuum and in OVPD. In case of vacuum growth, the cooling of the evaporant molecules
is abrupt, where the adsorbed species is forced to equilibrate with the surface
immediately upon contact. In contrast, the gas boundary layer present in OVPD allows
the molecules to cool gradually, dramatically reducing the temperature gradient. Since
the organic vapor was generated at a an evaporation temperature T
evap
exceeding the
substrate temperature, T
sub
, constitutional supercooling of the vapor will occur near the
substrate, reducing the equilibrium vapor pressure, P
eq
, of the organic (See Fig. 7-4c).
This provides the driving force for the condensation and growth of crystals.
Crystal growth itself proceeds by a combination of processes, including the
diffusion of molecules on the bare substrate, formation of stable nuclei, growth of the
nuclei by addition of diffusing molecules on the surface, and coalescence of individual
grains when their edges meet. The parameters governing these four processes include the
flux of admolecules to the surface, the critical (minimum stable) nucleus size, the surface
diffusivity (D
s
), and the surface tension of the crystallite,
s
. The molecular flux is given
by the deposition rate, the critical nucleus for pentacene on SiO
2
is four molecules (Ruiz
et al. 2003), while D
s
and

are more difficult to determine quantitatively, and depend on


the molecular structure and the chemistry and topology of the surface.
At very low deposition rates, the density of nucleation sites is very low, and
surface diffusion will dominate crystallization. This is known as diffusion-limited

162
Figure 7-4: a) Comparison of the molecular temperature profiles in vacuum and
vapor phase deposition scenarios; b) Corresponding profiles of the organic vapor
pressure, which is a function of temperature, P(T). c) Plot of the vapor pressure vs.
temperature, showing the actual vapor pressure P
org
(T
evap
) of the organic species
generated at the source evaporation temperature T
evap
. Upon cooling of the vapor
from T
evap
to T
sub
, P
org
exceeds the equilibrium vapor pressure P
eq
(T
sub
), driving the
condensation and crystal growth on the substrate.
Deposition in Vacuum
Vapor Phase Growth
T
T
Deposition in Vacuum
Vapor Phase Growth
T
T
Deposition in Vacuum
Vapor Phase Growth
Porg
P
eq
P
org
Driving force
Deposition in Vacuum
Vapor Phase Growth
Porg
P
eq
P
org
Deposition in Vacuum
Vapor Phase Growth
Porg
P
eq
P
org
Driving force
T
evap
T
sub
P
eq
(T
evap
)
P
org
(T
evap
)
P
eq
(T
sub
)
Cooling
Condensation
Driving
Force
Equilibrium Vapor
Pressure Curve
Temperature Profile Vapor Pressure Profile
Distance (x) Distance (x)
(a) (b)
(c)

163
aggregation (DLA) (Witten et al. 1981; 1983), and produces highly branched, fractal
crystals (Halsey 2000). In electronic device fabrication, however, high deposition rates
are often required, resulting in multi-nucleation processes, where the size and shape of
the crystals is not governed by the kinetics of surface diffusion alone, but also by the
thermodynamics of the thin-film condensed phase.
Three possible modes of crystal growth on surfaces have been identified,
illustrated in Fig. 7.5a (Venables et al. 1984). In the island (Volmer-Weber) growth
Figure 7-5: (a) Illustration of the three modes of crystal growth on surfaces: (i)
layer-by-layer (Frank-van der Merwe), (ii) layer-plus-island (Stransky-Krastanov)
and (iii) island (Volmer_weber). (From Venables, et al., 1984). (b) Illustration of
the mechanisms governing crystal growth from the vapor phase. (From Irisawa,
2003)
(i) (ii)
(iii)
< 1 ML
1 < < 1 ML
> 2 ML
(a)
(b)
= coverage

164
mode, clusters of the adsorbate nucleate on the surface and grow into islands. In this case,
the adsorbate-adsorbate interaction exceeds the adsorbate-surface interaction. In layer
(Frank-van der Merwe or epitaxial) mode, the surface interaction is large, and the
admolecules condense into a complete surface-bound monolayer, followed by more
weakly bound monolayers. A common intermediate growth (Stransky-Krastanov) mode
occurs, in which one or few surface-bound, strained adsorbate monolayers are formed,
and become covered with islands of the condensed phase as the growth proceeds.
The structure of the initial monolayer is strongly influenced by the underlying
substrate, while the island density can vary over nine orders of magnitude with substrate
temperature (Venables et al. 1984). For molecular organic quasi-epitaxy (Forrest 1997),
the adsorption energy is typically on the order of 1-2 eV, the surface diffusion barrier is
on the order of 50-100 meV (Casalis et al. 2003; Forrest 1997; Heringdorf et al. 2001;
Heringdorf et al. 2004; Israelachvili 1992; Kitaigorodsky 1973; Krause et al. 2002;
Verlaak et al. 2003). In comparison, the room temperature thermal energy of a molecule
is 26 meV.
7.3.2 Theory of crystal growth on surfaces
Figure 7-5b illustrates the key steps in the growth of crystals on the substrate, once a
stable grain has nucleated (Irisawa 2003). Here, the net flux of molecules at the surface,
j
v
, is given by the difference between the adsorption and desorption rates. The larger the
energy barrier to desorption, and the lower the surface temperature, the longer the
admolecule will spend on the surface. The average residence time of a molecule on the
surface is inversely proportional to the desorption rate constant. Since desorption is an
activated process, the desorption time,
s
, is:

165
1
exp
des
s B s
E
k T

| |
=
|
\ .
(7-3)
where E
des
is the energy of desorption, T
s
is the surface temperature, and is
characteristic of the frequency of desorption "attempts", given by the effective surface
vibration frequency (10
11
-10
13
s
-1
) (Venables et al. 1984). Note that for endothermic
desorption, E
des
> 0, and increasing T
s
should decrease
s
. During the time
s
, the
adsorbed molecules will tend to diffuse along the surface, until finding a kink (labeled as
K in Fig. 7-5) in the crystal facet, where the coordination number is highest and the
energy is minimized with respect to other locations on the surface. Since surface
diffusion is also an activated process, the diffusion coefficient D
s
is:
2
exp
sd
s
B
E
D a
k T

| |
=
|
\ .
(7-4)
where a is the lattice constant and E
sd
is the surface diffusion activation energy; E
sd
> 0,
and increasing T should increase D
s
. An effective surface diffusion distance,
s
, is
defined following the Einstein relation
s
= (D
s

s
)
1/2
:
exp
2
des sd
s
B
E E
a
k T

| |
=
|
\ .
(7-5)
where it was assumed that for
s
and D
s
is the same. If the admolecule adsorbs within
this distance
s
of the site K, it will contribute to the growth of a pre-existing crystal via
surface diffusion, otherwise, it will desorb. Note that since E
des
> E
sd
> 0, increasing T
results in a decrease in
s
, because the Boltzmann factor favors desorption events more
than surface diffusion.

166
From the above expressions and the net flux of pentacene molecules to the
surface, the rate of lateral crystallite growth, can be determined. This rate is referred to as
"step velocity" v

, and can be expressed as a product of the characteristic area of the


adsorption site (a
2
), a characteristic length scale (e.g.
s
), and the molecular surface flux.
Using Eq. (7-5) (Irisawa 2003):
3
2 exp
2 2
eq
org org
des sd
B org B
P P
E E
v a
k T mw k T

| |
=
|

\ .
(7-6)
where P
org
and P
org
eq
are the actual and the equilibrium vapor pressures of the organic
species (in this case, pentacene), while mw
org
is the molecular weight (278g/mol for
pentacene). Here, (P
org
- P
org
eq
) is the net driving force for providing fresh admolecules
for the growth of crystallites, and
( )
2
eq
org org org B
P P mw k T is the net molecular flux
at the surface. Since E
sd
can vary significantly between different types of surfaces, Eq. 7-
6 predicts different crystallite growth rates and sizes on different materials.
It is tempting to use this mechanism to explain the morphology of OVPD grown
pentacene on SiO
2
and gold, as shown in Fig. 7-6. However, a number of uncharacterized
variables and processes (e.g. surface defects, cluster mobility, molecular isomerization)
can affect nucleation and surface diffusion (Venables et al. 1984). Thus, when
experiments are not performed in ultra-high vacuum on cleaved crystalline surfaces, the
theories developed to describe them will contain fitting parameters, which mask the true
mechanism. In Fig. 7-6, for example, the surface roughness of electron-beam sputtered
gold may also enhance nucleation of the pentacene crystals, and thus result in finer
granularity of the grown film. Relief of quasi-epitaxial (Forrest 1997) strain by grain

167
boundary creation is an unlikely mechanism, since the underlying SiO
2
surface is not
crystalline.
Furthermore, using Eq. 7-6 for quantitative predictions of crystal growth rates
depends critically on accurate values of E
des
and E
sd
, which can be difficult to obtain. The
desorption free energy change is given by E
des
= H
des
- TS
des
, where H
des
is the
desorption enthalpy change and S
des
is the desorption entropy change. When the surface
is a crystal or a film comprised of the same species as the admolecule, H
des
= H
vap
and
S
des
= S
vap
, where H
vap
and S
vap
are the vaporization enthalpy and entropy changes,
Figure 7-6: A scanning electron micrograph of a 700 thick pentacene grown by
OVPD at 3 /s on a SiO
2
substrate at Ts = 25C, with 500 thick gold
source/drain contacts deposited prior to pentacene growth. The micrograph shows
the difference in pentacene grain size between the two surfaces.
Pentacene
on Gold
Pentacene
on Gold
Pentacene
on SiO
2
p
+
Si
SiO
2
Gold
Pentacene

168
respectively. For pentacene, H
vap
= 145-180 kJ/mol (1.5-1.8 eV) at 25C (Verlaak et al.
2003), and S
vap
= 560 J/molK. However, E
sd
values are not readily available. A number
of studies have been published on the dynamics of pentacene nucleation and growth, for a
variety of substrates and techniques, including MBE (Heringdorf et al. 2001; Heringdorf
et al. 2004), VTE (Ruiz et al. 2003), seeded noble gas beams (Casalis et al. 2003), as
well as models of nucleation based on first principles and basic thermodynamic data for
pentacene (Verlaak et al. 2003). However, the activation energy for surface diffusion was
not directly addressed. Verlaak et al. (Verlaak et al. 2003) examine the onset of 2- and 3-
dimensional modes of nucleation, finding that the transition from 2-D to 3-D nucleation
requires a jump in the chemical potential of 72 meV or 6.9 kJ/mol for pentacene on SiO
2
,
yielding
s
= 9.3 m for a = 15 at 25C. Although the 72 meV activation energy is
consistent with Van der Waals bonding interactions of multi-ring aromatics on a variety
of surfaces (Forrest 1997; Israelachvili 1992; Kitaigorodsky 1973), the calculation of
s
is
clearly overlooking some physical mechanism, considering the observed grain size for
these conditions (Fig. 7-6). The presence of surface defects can significantly alter the
surface energy potential, increasing E
sd
, while the mechanism of desorption from SiO
2
is
a complex sequence of steps, varying with molecular shape. A rigid flat molecule such as
pentacene may have to re-orient itself in a step-wise fashion to have fewer and fewer
atoms in contact with the substrate. As it the molecule decreases the number of contact
points with the substrate (e.g. flat-, to edge-, to end-bound states), its binding energy
decreases in a step-wise fashion, while the barrier for surface diffusion decreases with
each re-orientation. Thus, as T
s
increases, the actual E
des
and E
sd
in Eq. (7-5) also change.
The binding energy of different molecule-surface configurations can be estimated using

169
the (computationally intensive) atom-atom potential methods (Forrest 1997;
Kitaigorodsky 1973). The accurate prediction of crystal growth would thus depend on
these numbers, coupled with molecular dynamics or Monte-Carlo simulations to model
the surface diffusion.
7.3.3 Role of background carrier gas in crystal growth
The above discussion of growth made little distinction between vacuum and vapor phase
deposition. An important difference arises when the k
B
T energy of the molecule diffusing
on the surface approaches E
des
increasing the probability of its desorption. In vacuum the
desorbed molecules are essentially "lost" from the surface. In contrast, the presence of a
diffusion-limitation of the boundary layer in OVPD allows the molecule to "resample"
the surface via gas phase transport near the substrate surface. This process is illustrated in
Fig. 7-7a, and is essentially the same mechanism as in the depression of vapor pressure
by a background inert gas surrounding the evaporating crystal, as shown by (Kloc et al.
1998; Kloc et al. 1997) for alpha-hexithiophene (-6T) (Fig. 7-7b). Qualitatively Eq. (7-
6) predicts that raising the surface temperature will favor desorption over diffusion. With
carrier gas blanketing the substrate, this may in fact aid in the growth of crystallites, by
transporting the admolecules in the gas phase directly above the surface to distant
crystallites, instead of waiting for surface diffusion to complete. Since the velocity of
molecules in the gas phase is greater than on the surface, this gas phase-mediated
transport should enhance crystal growth on the surface. In addition, it is possible to add
energy to the organic species on the surface by increasing their impact energy. This can
be achieved using seeded gas beam methods (Casalis et al. 2003), and is a likely
additional difference between OVJP and vacuum or vapor growth.

170

Vacuum Vapor Phase
(a)
(b)
Figure 7-7: a) Illustration of the mechanism of crystal growth in vacuum, where if
a molecule desorbs from the surface before it can find a favorable site at a crystal
facet, it is lost and cannot resample the surface. In contrast, deposition from the
vapor in the presence of carrier gas molecules (circles) can enhance crystal growth.
Molecules which desorb are not lost, but have a higher chance of re-sampling the
surface due to the diffusion barrier presented by the surrounding inert gas
background. b) Plot of evaporation rate of alpha-hexithiophene (-6T) vs. the
pressure of helium surrounding the -6T crystals, clearly showing the effect of
diffusion limitation on the ability of molecules to escape the surface; (from Kloc et
al, 2003).
He
-6T
Vacuum Vapor Phase
(a)
(b)
Figure 7-7: a) Illustration of the mechanism of crystal growth in vacuum, where if
a molecule desorbs from the surface before it can find a favorable site at a crystal
facet, it is lost and cannot resample the surface. In contrast, deposition from the
vapor in the presence of carrier gas molecules (circles) can enhance crystal growth.
Molecules which desorb are not lost, but have a higher chance of re-sampling the
surface due to the diffusion barrier presented by the surrounding inert gas
background. b) Plot of evaporation rate of alpha-hexithiophene (-6T) vs. the
pressure of helium surrounding the -6T crystals, clearly showing the effect of
diffusion limitation on the ability of molecules to escape the surface; (from Kloc et
al, 2003).
He
-6T

171
7.4 Growth mechanisms for pentacene
To summarize the discussion thus far, the two extremes of film growth are adsorption-
limited and surface diffusion-limited. When the net rate of arrival of species is low
relative to the surface diffusion rate, larger crystallites are obtained, provided the surface
temperature is low enough for equilibrium to favor the condensed phase. In depositing
molecular organic thin films for electronic device applications, there is a competition
between the need to deposit the films quickly for production cost reasons, and the need to
grow high quality crystalline films to improve device performance. The grain size can be
increased by post-deposition annealing, or by appropriately adjusting the deposition
conditions (e.g. temperature, rate, and pressure). Increasing T
s
during growth can improve
the diffusion of admolecules on the surface, but not indefinitely, since the probability of
desorption also increases with T
s
.
During our work on pentacene transistors, it was found that adsorption-limited
growth of pentacene in OVPD typically occurs at high (>10
4
Pa) chamber pressures, low
(<1 /s) deposition rate, and T
s
> 40C substrate temperature. When the deposition
pressure and the substrate temperature are both minimized, organic species adsorb on the
substrate upon arrival, typically resulting in small crystals or amorphous films. Vacuum
thermal evaporation at >4 /s and T
s
< 40C is one example of surface diffusion-limited
growth. Thus, by changing the deposition conditions, such as deposition rate and
substrate temperature, it should be possible to explore both growth regimes. In addition to
temperature and rate, the background nitrogen pressure can be used to a limited extent in
controlling crystallite growth, as discussed in Sec. 7.3.3.

172
Even at relatively low substrate temperature, T
s
= 10C, the natural tendency of
pentacene molecules to form ordered films coupled with the diffusion-limited transport
result in polycrystalline growth, as indicated by Figs. 7-6 and 7-8a for pentacene grown
at 0.25 Torr and 3 /s on PE-CVD SiN
x
. The x-ray diffraction pattern of the film (Fig. 7-
8b) shows several orders of the peak doublet corresponding to the (001) crystal
orientation having 15.48 and 14.44 plane spacing, respectively, indicating that
pentacene molecules "stand up" on the surface. This has been observed by others
(Bouchoms et al. 1999; Dimitrakopoulos et al. 1996; Gundlach et al. 1999), who found
that the 15.5 spacing is dominant in thin (<500 ) films, while the proportion of the
14.44 increases with film thickness, as well as substrate temperature during growth.
Similar diffraction patterns were obtained for pentacene on SiO
2
(shown later in the text).
Figure 7-8: a) SEM of a polycrystalline pentacene thin film grown by OVPD on
PE-CVD SiN
x
at P
dep
= 0.25Torr, T
sub
= 10C, T
src
= 240C, r
dep
= 3/s. b) X-ray
diffraction pattern of the 2 scan of the film in (a), showing 5 orders of the (001)
and (001) (thin-film and bulk) peak doublet, corresponding to 15.48 and
14.44 plane spacing, respectively. Similar patterns were obtained for pentacene
on SiO
2
, shown later in the text.
Pentacene on SiN
x
(a)
5 10 15 20 25 30
0.05 0.100.15 0.20 0.250.30
0
1
2
3
4
5
6
Pentacene (001)
d
1
=15.48
d
2
=14.44
sin()
p
e
a
k

o
r
d
e
r
2
C
o
u
n
t
s
(b)
1m
Figure 7-8: a) SEM of a polycrystalline pentacene thin film grown by OVPD on
PE-CVD SiN
x
at P
dep
= 0.25Torr, T
sub
= 10C, T
src
= 240C, r
dep
= 3/s. b) X-ray
diffraction pattern of the 2 scan of the film in (a), showing 5 orders of the (001)
and (001) (thin-film and bulk) peak doublet, corresponding to 15.48 and
14.44 plane spacing, respectively. Similar patterns were obtained for pentacene
on SiO
2
, shown later in the text.
Pentacene on SiN
x
(a)
5 10 15 20 25 30
0.05 0.100.15 0.20 0.250.30
0
1
2
3
4
5
6
Pentacene (001)
d
1
=15.48
d
2
=14.44
sin()
p
e
a
k

o
r
d
e
r
2
C
o
u
n
t
s
(b)
5 10 15 20 25 30
0.05 0.100.15 0.20 0.250.30
0
1
2
3
4
5
6
Pentacene (001)
d
1
=15.48
d
2
=14.44
sin()
p
e
a
k

o
r
d
e
r
2
C
o
u
n
t
s
(b)
1m

173
7.5 Evidence for pentacene morphology influencing TFT performance
Figure 7-9 shows the I
DS
vs. V
DS
and V
GS
response of a bottom-contact pentacene TFT,
where the 700 thick pentacene channel was deposited by OVPD, using N
2
as the carrier
gas, at a deposition rate r
dep
= 3/s, substrate temperature T
s
= 10C and deposition
pressure P
dep
= 0.25 Torr. The substrates were comprised of n-type doped (0.001 -cm)
(100) Si wafers (which also served as substrates and large-area gate contacts), with
2000 thick SiN
x
gate dielectric grown by plasma-enhanced chemical vapor deposition
(PE-CVD). For these devices, the source and drain contacts were pre-patterned onto the
SiN
x
layer using conventional photolithography. Prior to deposition, pentacene source
material was twice purified by vacuum train sublimation (Forrest 1997) to yield 1 mm
crystals. Additional purification (i.e. growth from the vapor phase) was done under a flow
of ultrahigh purity N
2
at 800 Torr (Kloc et al. 1997) to yield pentacene crystals several
millimeters in diameter, which were used as source material for the OVPD of TFT
channels.
In characterizing pentacene TFTs, it is important to differentiate between factors
arising from charge conduction in the pentacene crystal itself, and factors arising from
device geometry. Figure 7-9a shows the electrical characteristics of a pentacene TFT
with W/L = 100m/30m, exhibiting p-type (hole) conduction, with I
DS
saturating at V
DS

< -15V. According to Eq. 7-2, the mobility in the saturation regime is obtained either
from the slope of the (I
DS
)
1/2
vs. V
GS
curve (Fig. 7-9b), or using (dI
DS
/dV
GS
) / (C
ox
/2W/L).
Using this latter method,
eff
= 0.03 cm
2
/Vs, while I
DS
on/off ratio = 10
5
, and V
T
= -4V. If

eff
represented the true mobility of holes in pentacene, changing the device geometry
(via W/L) should not have altered the measured
eff
, because W/L is factored out of the

174
Figure 7-9: a) I
DS
vs. V
DS
behavior of a bottom-contact pentacene TFT with Au
S/D electrodes, SiN
x
gate dielectric and W/L = 100m/30m, deposited at the
same conditions as in Fig. 7-4. b) I
DS
and I
DS
1/2
vs. V
GS
characteristics of the TFT,
exhibiting I
DS
on/off ratio of 10
5
, V
T
= -4V and
eff
= 0.03cm
2
/Vs. Inset: It is
found that
eff
increases with the channel length L.
(a)
-30 -25 -20 -15 -10 -5 0
-1.0
-0.8
-0.6
-0.4
-0.2
0.0
0.2

V
ds
(Volts)
I
D
S
(

A
)
V
GS
= -30
V
GS
= +4
(b)
-40 -30 -20 -10 0 10


V
GS
(Volts)
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5

0 5 10 15 20 25 30
0.00
0.02
0.04


L (m)

e
f
f
(
c
m
2
/
V

s
)
[
I
D
S
(

A
)

]
1
/
2
10
-10
10
-9
10
-8
10
-7
10
-6
10
-5
I
D
S
(
A
m
p
s
)

175
expression for
eff
. However, as the Inset of Fig.7-9b shows, decreasing L results in
lower values of
eff
, even for the devices deposited measured on the same substrate. This
can be understood by examining the pentacene channel morphology.
Figure 7-10a is an SEM of the TFT channel region for W/L = 100/6. The
pentacene film is polycrystalline, with small grains and a high density of grain boundaries
at the source/drain contacts. Evidently, the gold contact pads act as nucleation sites for
the pentacene vapor, resulting in smaller crystals and a higher grain boundary density per
unit surface area. Thus, when decreasing L (Fig.7-10b) in the bottom-contact geometry,
the pentacene in the channel region contains a higher grain boundary density. Since
eff

varies inversely with L, one concludes that grain boundaries impede charge conduction.
To approach the true mobility of holes in single-crystal pentacene the average pentacene
grain size should be increased relative to L. Larger grains may be obtained by increasing
both the substrate temperature and the deposition pressure. Higher substrate temperature
1m
1m
(b)
(a)
1m
1m
(b)
(a)
Figure 7-10: a) SEM of a bottom-contact pentacene TFT with SiN
x
gate dielectric,
with L = 6 m. b) SEM of another bottom-contact TFT on the same substrate, but
with L = 2.3 m, both deposited at the same condition as the device in Fig. 5a. In
both cases pentacene growth appears to nucleate at the Au S/D contacts, leading to
finer grains and higher grain boundary density near the current injection region.
This is what ultimately limits
eff
for smaller channels, explaining the trend shown
in the Inset of Fig. 5b.

176
provides the thermal energy to anneal defects out of the deposited film, while higher
pressure in the deposition chamber decreases the likelihood of desorption of pentacene
molecules from the surface at the elevated temperature by impeding transport through the
boundary layer.
Using T
s
= 65C and P
dep
= 8 Torr, pentacene crystallites are obtained on the
order of 1 m or greater in size (Fig. 7-11c). However, even for L ~ 1 m, the observed
Figure 7-11: Plot of
eff
vs. deposition pressure, P
dep
for bottom-contact
pentacene TFTs deposited onto SiN
x
with pre-patterned Au S/D contacts;
pentacene was grown at several substrate temperatures. For T
s
= 20C, there is a
peak in
eff
at ~1 Torr. For T
s
= 10C, increasing the channel length, L, results in
an increase in
eff
, similar to what is observed for T
s
= 20C (see Fig. 7-5) and for
T
s
= 50C. At T
s
=65C,
eff
drops to 210
-3
cm
2
/Vs, explained by de-wetting of
pentacene in the S/D region (see SEM in (c)). For pentacene deposited on
unpatterned SiN
x
at 65C, with top-deposited S/D contacts, pentacene
crystallization is unaffected by the contact, and
eff
= 0.6 cm
2
/Vs
0.1 1 10
10
-3
10
-2
10
-1
10
0

incr L
10C L=30,10,6,2 m
Pressure (Torr)

1m 1m
1m 1m
Pentacene on SiN
x
; Top Contact
Pentacene on SiN
x
; Bott. Contact

e
f
f
(
c
m
2
/
V

s
)
T
s
= 10C T
s
= 10C
T
s
= 20C T
s
= 20C
T
s
= 50C T
s
= 50C
T
s
= 65C T
s
= 65C
T
s
= 65C (Top contact) T
s
= 65C (Top contact)
(a)
(b)
(c)

177

eff
drops by almost an order of magnitude from that in a TFT deposited at T
s
= 10 or
20C. The answer again lies in the pentacene morphology. As shown in Fig. 7-11c, at
elevated T
s
pentacene de-wets the gold source/drain contacts, resulting in large voids and
poor overlap between the semiconductor and the contacts. One solution is to grow
pentacene on blank SiN
x
substrates and deposit the source/drain contacts on top. This
requires the use of shadow-masks, which typically limits L to lengths >10m.
Nevertheless, using such a shadow-mask,
eff
= 0.6 cm
2
/Vs was obtained for pentacene
TFTs having >5m average pentacene crystallite size. These findings are summarized in
Fig. 7-11a, where
eff
is plotted vs. the deposition pressure for various substrate
temperatures and device geometries.
7.6 Relative effects of grain size and substrate treatment on device
performance
As PE-CVD SiN
x
is generally rougher and less homogeneous than dry thermally-grown
SiO
2
, the latter was chosen for the next set of experiments, to decrease substrate-induced
grain nucleation. Top-deposited 500 thick gold source/drain contacts were used. As
Fig.7-12a-c shows, the simultaneous increase in T
s
and P
dep
, and a decrease in r
dep
result
in larger average grain size for 700-1000 thick pentacene films grown by OVPD on
SiO
2
. For pentacene TFTs made from the samples shown in Fig.7-12a-c having W/L =
1000m / 45m,
eff
= 0.06, 0.12 and 0.58 cm
2
/Vs, increasing with the average
pentacene grain size. Here, pure island growth dominates, and the grain boundaries
penetrate down to the pentacene-insulator interface (see Fig. 7-5a). Since charge
transport in the TFT likely takes place in the first few monolayers of pentacene

178
(Pratontep et al. 2004), the increasing mobility is a logical outcome of the decreasing
density of grain boundaries at the pentacene-SiO
2
interface.
Still, the grain size (~510 m) does not exceed L (45m), and grain boundaries
in the channel arguably result in
eff
<
crystal
, where
crystal
is the mobility of holes in
single-crystal pentacene. At the time, several studies (Gundlach et al. 1999; Gundlach et
al. 1997; Kymissis et al. 2001; Lin et al. 1997) reported improved device characteristics
1 m
(a) 10C, 0.25 Torr, 3.0 /s

eff
=0.06cm
2
/Vs
1 m
(b) 40C, 6.0 Torr, 1.0 /s

eff
=0.12cm
2
/Vs
(c) 65C, 10.5 Torr, 0.3 /s
2 m

eff
=0.58cm
2
/Vs
(d) 10C, 0.25 Torr, 3.0 /s
1 m

eff
=1.2cm
2
/Vs
(e) 40C, 6.0 Torr, 1.0 /s
1 m

eff
=1.2cm
2
/Vs
1 m
(f) 65C, 10.5 Torr, 0.3 /s
0.6 <
eff
< 0.9cm
2
/Vs
Figure 7-12: (a)-(c) SEMs of 1000 thick pentacene films grown by OVPD on SiO
2
for top-contact TFTs at the deposition conditions indicated below each frame. The
average crystal size increases as T
s
and P
dep
increase, with concurrently decreasing
r
dep
. The top-contact TFTs using these films exhibited increasing
eff
as the
pentacene film grain size increased, indicating that grain boundaries indeed increase
the overall channel resistance. (d)-(f) SEMs of pentacene grown by OVPD on OTS-
treated SiO
2
at the deposition conditions matching those in (a)-(c). Unlike films
deposited on SiO2, pentacene on OTS has finer grain size, relatively unaffected by
growth conditions, except when T
s
exceeds 60C, possibly due to degradation of
OTS at high temperature. The
eff
is thus relatively unaffected by growth conditions,
but is much greater than in devices with untreated SiO
2
; this is somewhat
counterintuitive, due to the much higher grain boundary density in OTS-treated
devices.

179
when chemically treating the substrate to result in a hydrophobic surface prior to the
deposition of pentacene. Octadecyl-trichlorosilane (OTS) was one such surface treatment
of choice. The polar SiCl
3
head groups chemically react with the underlying SiO
2

substrate and neighboring SiCl
3
molecules in the presence of water to form a self-
assembled monolayer chemically bound to the substrate, while the saturated hydrocarbon
backbone results in an ordered hydrophobic surface. It was thought that the non-polar
pentacene molecules would form a better-ordered film on this type of surface, thus
explaining the observed increase in
eff
and I
DS
on/off ratios. Thus, to investigate the
Figure 7-13: SEMs of 1000 thick pentacene grown by VTE on OTS-treated
SiO
2
, where the deposition rate was < 2/s. The micrographs shows a
polycrystalline layer of pentacene having grains on the order of 0.5 m in
diameter, covered by twig-like islands embossing the grain boundaries in the
wetting layer. The polycrystalline layer is too thick to be the wetting layer
typically observed in Stransky-Krastanov growth mode, while the rod-like crystals
may rise from the underlying film under the pressure of the merging grain
boundaries. The measured
eff
for TFTs comprised of vacuum-deposited
pentacene on OTS-treated SiO
2
were also approximately 1 cm
2
/Vs. (Figure
courtesy of Changsoon Kim).

180
effects of surface modification on pentacene crystallization and TFT performance, the
substrates in Fig. 7-7a-c were also treated with OTS. The OTS treatment consisted of a
15min exposure to O
2
-Ar plasma, followed by a 24 hour soak in DI H
2
O, followed by a
3-hour or longer soak in a 700 M solution of octadecyltrichlorosilane (OTS) in
chloroform and hexane (3:7 by volume).
Scanning electron micrographs of 700-1000 thick pentacene thin films grown on
the OTS-treated substrates are shown in Fig.7-12d-f. Surprisingly, pentacene formed
substantially smaller grains on the OTS-treated SiO
2
compared to the untreated
substrates. More surprisingly, in view of the smaller pentacene grain size, the
eff

measured for these TFTs exceeds 1.0 0.1 cm
2
/Vs. Figure 7-14 shows the I
DS
vs. V
GS

and V
DS
characteristics of a typical OVPD-grown pentacene TFT with an OTS-treated
SiO
2
gate dielectric. The saturation-regime
eff
= 1.4 cm2/Vs, I
DS
on/off ratio = 10
8
-10
9
,
in reverse and forward V
GS
sweeps, respectively, while V
T
= 5V.
The surprising observations may be reconciled by a different surface growth
mode for pentacene on OTS. If a layer-plus-island (Stransky-Krastanov) mode is
assumed, the thin wetting layer of pentacene on OTS participates in charge conduction,
while the grain boundaries visible in the top-most layers does not penetrate down to the
pentacene-OTS interface. The wetting typically being only one or a few monolayers,
direct observation of it can be done using atomic force microscopy (AFM). Here, its
existence is inferred from the mobility behavior of the TFTs. Provided that charge
transport is indeed confined to the few pentacene monolayers immediately adjacent to the
gate dielectric, further verification of the Stransky-Krastanov growth mode can

181
potentially be done by varying the channel thickness. If the
eff
is invariant with thickness
for > 2 or 4 monolayers, the wetting layer is indeed present.
For vacuum-deposited 1000 thick pentacene films on OTS (Fig. 7-13), a
polycrystalline but relatively flat pentacene layer grows on top of OTS, but becomes
covered in smaller twig-like crystallites embossing the grain boundaries. The measured

eff
for the VTE-grown TFTs is also on the order of 1 cm
2
/Vs, indicating once again that
the channel mobility is largely unrelated to the top film morphology when the OTS
-80 -60 -40 -20 0
10
-13
10
-12
10
-11
10
-10
10
-9
10
-8
10
-7
10
-6
10
-5
10
-4
10
-3

I
D
S

(
A
m
p
s
)
V
GS
(Volts)


-40 -30 -20 -10 0
-20
-10
0
V
DS
(Volts)


-40 = V
GS
-30
-20
W / L = 1000 / 45
V
DS
= -40 V
I
D
S
(

A
m
p
s
)
On SiO
2
On OTS/SiO
2
Figure 7-14: I
DS
vs. V
GS
characteristic of an OVPD-grown 1000 thick
pentacene channel TFT with W/L = 1000m/45m, obtained in the saturation
regime at V
DS
= -40V. The device contained an OTS-treated SiO
2
gate dielectric;
it is compared to a TFT with no OTS treatment. The OTS-treated device
exhibited
eff
= 1.4 cm
2
/Vs, I
DS
on/off ratio 10
8
-10
9
(forward and reverse V
GS
sweeps, as indicated), and V
T
= -5V. Inset: I
DS
vs. V
DS
characteristic of the
device, showing classical current saturation behavior. The TFT without the OTS
exhibited lower I
DS
on/off ratio and
eff
.

182
treatment of SiO
2
is present. However, at the higher growth temperature, T
s
, the adhesion
between pentacene and OTS can be disrupted in favor of pentacene-pentacene
interaction, resulting in island growth and larger grains (Fig. 7-12f). Here, the grain
boundaries may penetrate down to the OTS surface, resulting in the lower measured
eff
.
Figure 7-15 shows the x-ray diffraction pattern for a 1000 thick pentacene layer
deposited onto SiO
2
at 1-2 /s, a chamber pressure of 6.0 Torr, and a substrate
temperature of 40C. The peaks arise from the previously observed "bulk" and "thin-film"
Figure 7-15: X-ray diffraction data for pentacene films grown by OVPD on SiO
2
(bottom curve) and OTS-treated SiO
2
(top curve). Both samples exhibit the peaks
corresponding to thin-film and bulk phases of pentacene (15.5 and 14.5 (001)
plane spacing, respectively). The OTS-containing sample also shows a pronounced
peak at 2 18, corresponding to ~6 (111, 110) plane spacing; a similar
feature can be seen in the curve for the untreated substrate, albeit at much lower
intensity. The smaller plane spacing corresponds more with the short axis of the
pentacene unit cell. It is thus likely that a larger fraction of pentacene molecules on
the OTS-treated SiO
2
are oriented with the long molecular axis parallel to the
substrate plane, compared to those on non-treated SiO
2
.
2
5 10 15 20 25 30
Pentacene / SiO
2
Pentacene / OTS / SiO
2
l
o
g
(
I
n
t
e
n
s
i
t
y
)

(
a
r
b
.

u
n
i
t
s
)
(002)
(002)
(003)
(003)
(
0
0
4
)
(004)
(005)
(005)
(111,110)
Pentacene
1
5

0
.
5


183
phases of pentacene.(Gundlach et al. 1999) The thin-film phase has an interplanar
spacing of 15.5 0.1 , while the bulk-phase, characteristic of thicker films and higher
substrate temperatures, has an interplanar spacing of 14.4 0.1 . The x-ray diffraction
pattern for a 700 thick pentacene film on OTS-treated SiO
2
is also shown in Fig. 7-15.
The pentacene film on the OTS-treated SiO
2
substrate exhibits a Bragg reflection at 19
0.25, which is less pronounced in the case of untreated SiO
2
. This peak is likely an
unresolved doublet arising from (110) and (111) reflections of pentacene (Campbell et
al. 1961), with an interplanar spacing of approximately 3 . Observation of these features
in the 2- scan indicates that a significant fraction of the pentacene molecules are
oriented with the long molecular axis parallel to the substrate. Pentacene in this
orientation has also been observed when deposited onto copper (Lukas et al. 2002;
Schuerlein et al. 1995), where the surface morphology is similar to that observed for
pentacene on OTS. However, the depth distribution of this pentacene phase is not yet
clear, but would be a useful measurement.
For the devices in Fig. 7-12, the subthreshold slope, S, for TFTs on SiO
2
and
OTS-treated SiO
2
was found to be S = (2.1 0.1) and (0.6 0.1) Volts/decade,
respectively. The
eff
= (0.12 0.02) cm
2
/Vs and (0.6 0.1) cm
2
/Vs, while the peak I
DS

on/off ratios were 5x10
5
and 5x10
7
for SiO
2
and OTS-treated SiO
2
substrates,
respectively. In one possibility, the I
DS
at high V
GS
is increased in the case of OTS may be
attributed to the tighter packing of pentacene crystallites in the channel, or, alternatively,
to the increased strain of pentacene on OTS. In the first instance, the improved physical
contact between pentacene grains increases the overall electrical conductivity of the
channel during charge accumulation in the on-state. In the second instance, the

184
decreased intermolecular distance can improve the - electron overlap. The increased
density of grain boundaries does not seem to affect I
DS
in the "off"-state, which can be
seen from the smaller S values for OTS-treated TFTs. This is particularly pronounced in
the case of samples grown under conditions used in Figs. 7-12a and d, where
eff

increases by a factor of 20, but the average grain size decreases by a factor of two.
However, the smaller S values may also be due to the lower energy and number of
charge traps present in the regions between the grain boundaries, directly underneath the
crystals. In this case, the lowering of the dielectric surface defect or dipole concentration
may contribute.
7.7 Effect of surface energy on device performance
Another possible explanation for the improvement of various device performance
parameters can be based on changes to the surface energy of the dielectric with the
various treatments. Figure 7-16b shows the I
DS
vs. V
GS
response for variously treated
SiO
2
substrates. The treatments included: i) O
2
plasma exposure, ii) baking in air at
150C, iii) OTS treatment following the bake, and iv) OTS treatment following the O
2

plasma exposure. The contact angle of water on these substrates (Fig.7-16a) increases
from (i) to (iv), indicating a decreasing hydrophilic character of the surface, i.e. a
lowering of the surface energy of the oxide. Concomitantly, the TFTs exhibited a
simultaneous increase in the I
DS
on/off ratio, increase in the I
DS
-V
GS
subthreshold slope, S,
a shift from V
T
> 0 to V
T
< 0, and a less-pronounced hysteresis in the I
DS
vs. V
GS
behavior.
While the exact physics of this is not yet understood, it is possible that the lowering of the
dipole strength at the dielectric-pentacene interface (evidenced by increasing water
contact angle and decreasing V
T
) may in fact contribute to decreased charge trapping,

185
both in the depth of traps (evidenced by a sharper sub-threshold I
DS
rise) as well as their
number (evidenced by milder I
DS
-V
GS
hysteresis).
-70 -60 -50 -40 -30 -20 -10 0 10 20
10
-12
10
-11
10
-10
10
-9
10
-8
10
-7
10
-6
10
-5
10
-4
10
-3
W=45m
L=1000m
top Au contacts
2100A SiO
2
with surface treatments
Non-Polar
Surface
Polar
Surface
Non-Polar
Surface
Polar
Surface
O
2
plasma
bake
bake + OTS
plasma + OTS

I
D
S

(
A
m
p
s
)
V
GS
@ V
DS
= - 40 V


(a)
(b)
(i) (ii) (iii) (iv)
(i)
(ii)
(iii)
(iv)
Figure 7-16: (a) Photograph showing water droplets on 4 different SiO
2
substrates,
where each substrate received a different treatment: (i) 10 min in O
2
plasma, (ii)
same as i, followed by 15 min bake at 150C in air, (iii) same as iii, followed by an
OTS treatment, and (iv) same as i, followed by an OTS treatment. The substrates are
shown in order of increasing hydrophobicity, as evidenced by the increasing wetting
angle of the droplet. (b) I
DS
vs. V
GS
saturation-regime characteristics of pentacene
TFTs grown on the substrates used in (a), having top-deposited contacts and the
identical channel dimensions. The data show that I
DS
on/off ratio,
eff
, V
T
, and the
subthreshold slope all increase with the dielectric surface hydrophobicity.

186
7.8 Further remarks
Despite a large body of work focusing on charge transport in organic semiconductors,
there can be a lack of reproducibility of results from laboratory to laboratory, and even
within the same laboratory from one time to another. In the course of this study, some of
these patterns have been observed. For example, using freshly purified pentacene to
deposit channels at T
s
= 40C, P
dep
= 6.0 Torr, and r
dep
= 1 /s,
eff
= (1.5 0.2) cm
2
/Vs
was obtained. However, similar conditions more commonly yielded
eff
= (0.9 0.3)
cm
2
/Vs for pentacene stored under nitrogen for longer than 24 hours. Devices with
pentacene deposited on identically treated substrates in the same run showed substrate-to-
substrate variations in
eff
on the order of 15%, although freshly-purified source material
almost always yielded the higher
eff
values. Table IV lists some observed trends with
respect to TFT electrical characteristics and process conditions, which should be minded
in the preparation of devices and specified in reports.

Table IV Table IV

187
7.9 Summary
Pentacene thin films with controlled morphology were grown on a variety of substrates
by OVPD. The main objective was to use vapor phase growth to control pentacene
morphology and to understand the relative effects of morphology and the gate dielectric
surface treatment upon the electrical characteristics of the TFTs.
To decouple the influence of the heterogeneous source/drain contact region on
pentacene morphology, top-deposited contacts were used. For pentacene grown on SiO
2

larger crystallites were obtained at lower deposition rates and higher the substrate
temperatures. The average pentacene grain size increased by a factor of ~10, causing the
field-effect hole mobility,
eff
, to increase by a similar factor. Up to a 20-fold
improvement in
eff
was obtained by pre-treating the SiO
2
surface with OTS a
surfactant that forms a chemically bound self-assembled monolayer (SAM) on the SiO
2

surface. Somewhat surprisingly, the improvement in
eff
occurred despite the higher
density of pentacene grain boundaries in the channel region of OTS-treated devices.
Furthermore, several TFT performance parameters improved with hydrophobicity of
SiO
2
. We conclude that surface treatment had a much stronger effect on the electrical
characteristics of the TFT than grain morphology, and that the observed surface
morphology may not provide a clear picture of the molecular order of the buried
pentacene layers closest to the dielectric surface. Grazing-incidence neutron scattering
(Scoles 1988) or scanning probe techniques (Mccarty et al. 1999) can be used to study
the packing of pentacene not exceeding several monolayers of coverage on OTS.


188
Chapter 8: Summary, state-of-the-art,
challenges and future directions
8.1 Organic semiconductors and devices
This thesis began with an introduction to some basic properties of a class of compounds
called organic semiconductors. These are highly conjugated organic small molecules and
polymers, where excess positive or negative charge is readily distributed over much of
the molecule due to the extended nature of the molecular orbital (Kitaigorodsky 1973;
Pope et al. 1982). However, due to the weak intermolecular interactions dominated by
van der Waals forces, electron and hole mobilities in organic compounds rarely exceeds
~1 cm
2
/Vs, orders of magnitude lower than in their inorganic, covalently bonded
cousins, e.g. silicon, germanium, gallium-arsenide, etc (Pope et al. 1982). In addition, or
perhaps as a consequence, electronic excitations (excitons) in organic solids are typically
contained on a single molecule, or shared by no more than the nearest neighbor
molecules (Pope et al. 1982; Silinsh et al. 1994). Thus, much of the optoelectronic
behavior of organic solids is dictated by their molecular structure, thereby opening up a
powerful toolbox of synthetic chemistry for precise control of light-matter interactions in
these materials.

189
As the understanding of the optical and electronic processes in organic solids
grows, practical device applications have emerged, including light emitting diodes, solar
cells, and transistors. And although charge mobility in organic solids is orders of
magnitude lower than in inorganic semiconductors, by using extremely thin (on the order
of 10
1
-10
3
) layers, certain key performance characteristics could be improved (e.g.
power efficiency of OLEDs and solar cells, speed of photodetectors, etc.). At the same
time, it has been realized that the weak intermolecular forces enable the deposition of
organic semiconductors on a wide variety of substrates (e.g. glass, plastic), without
attempting to lattice-match the substrate and the active layers (Forrest 1997). This has
opened up a broad range of device applications, such as light-weight, flexible, wearable,
and low-cost electronics.
8.2 Organic semiconductor processing technology
Since most of the existing methods of semiconductor device fabrication are geared
toward inorganic materials, and are costly, considerable research has been thus directed
toward developing novel methods of organic semiconductor device fabrication (Forrest
2004). Low-cost alternatives such as ink-jet printing (Hebner et al. 1998; Sirringhaus et
al. 2000), dye sublimation (Blanchet et al. 2003), and vacuum evaporation (Forrest 1997;
Tang 1986; Tang et al. 1987) emerged. The polymer community favors solvent-based
methods, while the small-molecular community prefers molecular beam (or vacuum
evaporation) techniques. The room-temperature solubility of small molecules in most
solvents is very poor, while the thermal energy required to evaporate polymers or
molecules with >1000 molecular weight generally exceeds their inter-atomic bond
strengths. Thus, small-molecular weight devices possess an advantage over the polymer-

190
based devices in the greater control of thickness, doping and sophistication of multi-layer
device structures, while being unable to efficiently scale up for low cost manufacturing
by vacuum evaporation. Organic vapor phase deposition has thus been developed to
address some of these issues for small molecular weight semiconductors.
8.3 Development and application of OVPD
In the course of recent research in OVPD, our understanding of the physical mechanisms
governing OVPD increased (Shtein et al. 2001), enabling the controlled growth of
OLEDs (Baldo et al. 1998; Baldo et al. 1997; Shtein et al. 2001), TFTs (Shtein et al.
2002), and more recently solar cells. The specific requirements of organic thin-film
patterning were also addressed (Shtein et al. 2003), thus setting the stage for
commercialization of the OVPD technology. Figure 8-1 shows a schematic of a
commercial-grade OVPD system built by a semiconductor equipment manufacturer,
Aixtron AG (Aachen, Germany) in collaboration with Princeton University (Princeton,
New Jersey) and Universal Display Corporation (Ewing, New Jersey). At least four such
systems have been built, with systems installed and operating in Ewing and Princeton,
while OVPD is becoming recognized as a key enabling technology in the low-cost
fabrication of organic electronics. High performance phosphorescent OLEDs have been
deposited on substrates up to 6-inch x 6-inch, with film thickness uniformity >95%,
excellent quantum efficiency and operation lifetime characteristics (Fig. 8-2). Current
research in OVPD focuses on utilizing the unique aspects of OVPD to control nucleation
and growth of bulk heterojunctions (e.g. CuPc/PTCBI) for efficient solar cell
applications.

191
Load Port
Deposition
Chamber
Control
Module
Heated
Source
Module
Mixing Unit
N
2
N
2
N
2
N
2
T
1
T
2
T
3
N
2
N
2
N
2
N
2
T
1
T
2
T
3
N
2
N
2
N
2
N
2
T
1
T
2
T
3
Figure 8-1: (Top) Schematic of OVPD more suited to commercial-scale
production, where carrier gas flows through the source containers for more
efficient pick-up, and is subsequently evenly spread over the substrate using
an appropriately engineered disbributer. (Bottom) Illustration of a
commercial-grade OVPD system built by Aixtron AG in collaboration with
Princeton and Universal Display Corporation.

192

8.4 Development and application of OVJP
Despite these achievements, it was realized that using shadow-masks to pattern the
organic layers posed an inherent limitation. While >50% efficiency for delivering the
source material to the substrate can be achieved in OVPD, up 60% of that can be wasted
on coating of the mask in any given step of depositing a full-color OLED display.
Furthermore, shadow masks can only be used ~10 times before coatings become thick
enough to flake off and cause particulate contamination, and coated masks could undergo
only ~10 cleaning cycles before needing to be replaced. (Bardsley 2004) To address this
problem, organic vapor jet printing was developed. (Shtein et al. 2004) In OVJP, the
organic molecules are printed directly onto the substrate, by seeding a hot inert carrier
gas (e.g. N
2
or He), and expanding the gaseous mixture through microscopic collimating
Figure 8-2: Photograph of a 6x6 glass substrate with an array of green-
emitting phosphorescent OLEDs grown on it using the OVPD system
illustrated in Fig. 8-1. The uniform brightness of the devices under
operation implies uniform film thickness across the entire substrate.

193
nozzles at near-sonic average velocity. When the seeded gas jet impinges on a cooled
substrate, the organic molecules selectively physisorb onto the substrate, forming a well-
defined deposit. This highly non-equilibrium mode of film growth and patterning enables
the printing of high-performance polycrystalline pentacene TFTs at ultra-fast (>700A/s)
local deposition rates. Current research on OVJP focuses on the printing of organic
heterostructures (e.g. -NPD/Alq
3
) for OLED applications.
8.5 Future directions in carrier-assisted deposition, novel devices
8.5.1 Carrier-assisted deposition of metals
As mentioned in Ch. 2, both OVPD and OVJP cannot be used to deposit metals using the
carrier gas-assisted mechanism. Metals typically require very high temperatures
(>800C) to evaporate; upon contact with inert gases such as Ar or He, the rapidly
increasing probability of 3-body collisions leads to the formation of a metal aerosol. The
aerosol particle size can vary substantially, depending on the process conditions, but
generally exceeds 10nm, making it difficult to deposit atomically flat metal contacts.
An alternative technique may involve starting with metal-organic precursors,
transporting them using a carrier gas toward the substrate, and driving off the ligands in
the vapor phase just before the molecules strike the surface. The scission of ligands can
be initiated by intense (laser) light, a plasma discharge, or gas-phase chemistry. The
drawback of these approaches is that if the precursor breaks up, degradation of the pre-
deposited organics can occur. A detailed study of the relative bond strengths of several
classes of compounds may be helpful. Furthermore, to resolve this problem research is

194
needed to understand the nature of the metal-organic charge injecting contact to fully
understand and optimize the range of physical properties leading good electrical contacts.
8.5.2 Growth of focal plane arrays using OVJP
Curved focal plane photodetector arrays (Wang et al. 2003) offer potential advantages
over flat imaging arrays, which are currently used for cameras and sensors. The focal
surface of a simple thin lens is curved, whose shape is given by the Petzval condition.
The curvature leads to out-of-focus regions when flat imaging arrays are used. Figure 8-3
(top) shows that either the edges or the center of an image can be in focus at any given
time. Correcting for this type of aberration requires complex objective optics, which
increase the bulk and cost of the imaging apparatus. In contrast, the human eye solves
this imaging problem by using a curved focal plane the retina as shown in Fig. 8-3
(bottom).
One potential application is in portable (e.g. cell-phone) cameras. If a curved
focal-plane array can be deposited onto an inexpensive plastic lens, the image sharpness
may improve. Alternatively, curved very large-area arrays for use in telescopes can
substantially improve image quality and reduce the complexity of the optics. Even
replacing the retina itself with prosthetic photodetector implants is a potential application,
as has been shown with silicon photovoltaic arrays, which are forced to be small, due to
the high curvature of the eye.
Unfortunately, existing methods of inorganic semiconductor fabrication and
patterning cannot easily produce such an array, since a curved surface cannot be created
out of a flat sheet without creasing. One approach involves assembling and joining large
(~1cm) arrays of silicon on a curved backing (Wang et al. 2003). In another method,

195
Figure 8-3: (Top) Illustration of the Petzval condition, whereby a simple
thin lens always produces a curved focal surface, which is impossible to
image perfectly using flat imaging arrays, requiring the use of complex and
bulky corrective optics in the objective. (Bottom) Illustration of the human
eye, where the retina is natures own curved photodetector, matching
perfectly the Petzval surface produced by the lens. To the right is illustrated
the placement of a prosthetic retinal implant consisting of an array of
photovoltaic elements on a silicon chip, designed to partially restore vision
in certain cases of blindness (e.g. retinitis pigmentosa and macular
degeneration), where only the retina has lost its photodetection capability,
while the rest of the visual apparatus remains intact.

196
photoresist is deposited onto a flat substrate, which is subsequently plastically deformed,
while metal deposition and lift-off are used to pattern flexible metal interconnects
between stiff semiconductor islands on the curved substrate (Hsu et al. 2002).
Alternatively, OVJP can be used to deposit the photodetectors and other components onto
a pre-curved substrate. This approach would employ an OVJP nozzle mounted on a
computer-controlled goniometer. Contact deposition may potentially limit this approach,
although metal sputtering can be used for conformally coating curved surfaces. For
retinal implants, understanding the behavior of substrates and organic materials in
biological media is an obvious practical concern.
8.5.3 Alternative device form factors, fiber photovoltaics
As OLED flat panel display technology becomes commercially viable, novel applications
and yet-lower price-points become important for both OLEDs and other organic devices.
Electricity generation using photovoltaic cells is one example. Silicon solar cells are
simply too expensive to be used in all but the most specialized of applications, given that
conventional methods of electricity generation, and even wind-power, can provide
electricity at $0.02 or $0.04/kWhr. With an average solar flux of ~1kW/m
2
in favorable
locations, it can be shown that to be competitive, solar cells must achieve a $30/m
2
price
point, while also substantially reducing the cost of the solar module installation. The first
can be difficult, considering that ITO-coated glass which is used for organic solar cell
fabrication already costs ~$30/m
2
. There is little hope to decrease the cost of this
substrate, since the ITO deposition process has been optimized for several decades,
concurrent with already tremendously high production volumes. The second is difficult

197
due to the fact that both Si and glass are both relatively fragile and heavy. Unfortunately,
ITO-coated light-weight plastic substrates are even more expensive.
An alternative may involve fabrication of organic PV cells on fibers, (Fig. 8-4 and
8-5) and weaving them into cloth. In this way, installation costs can be drastically
reduced, while the range of application/deployment can also increase (e.g. backpacks and
Figure 8-4: a) Structure of the proposed solar fiber, consisting of a nylon strand
(radius r
1
) metallized (thickness d
1
) and coated with the photogenerating
polymer blend (radius r
2
). The primary photogenerating core contacts a metal
anode wire (diameter d
2
), and both are encased in a protective nylon sheath
(radius r
3
).
MDMO-PPV
PCBM
PEDOT
PSS
CuPc
C
60
h
Li/Al
Nylon core
Inner conductor (e.g. Al)
Outer / Auxiliary conductor (e.g. Al)
Barrier coating
Photoactive layer
Figure 8-4: a) Structure of the proposed solar fiber, consisting of a nylon strand
(radius r
1
) metallized (thickness d
1
) and coated with the photogenerating
polymer blend (radius r
2
). The primary photogenerating core contacts a metal
anode wire (diameter d
2
), and both are encased in a protective nylon sheath
(radius r
3
).
MDMO-PPV
PCBM
PEDOT
PSS
CuPc
C
60
h
Li/Al
Nylon core
Inner conductor (e.g. Al)
Outer / Auxiliary conductor (e.g. Al)
Barrier coating
Photoactive layer

198
tents made from such solar cell cloth). The cost of manufacture can be further
decreased by virtue of adapting existing low-cost fiber-drawing technology to a new set
of photoactive materials. It is anticipated that liquid-processable materials may be most
suited for this application.
8.5.4 Large-volume, ultra-purification
With the increasing volume of device production, the need for larger quantities of
purified source material also grows. Current methods of organic material purification
(e.g. solvent fractionation for polymers, or train sublimation for small molecules) are
(a) (b) (c) (d) (e)
(g)
(h)
(i)
(j)
(f)
(k)
Process flow
Unspool
nylon wire
Al coat
Coat with
active layers
Co-wind
Aux. anode
Barrier
coating
Spool up
and ship
Figure 8-5: A possible process flow diagram for manufacturing a solar cell
on a continuous fiber, which can be subsequently woven into solar cell
cloth with a wide range of applications.

199
simply inadequate with regard to their slow speed and small quantities of purified
material obtained. For example, it can take a week or more to complete 3 purification
stages for relatively small quantities (~3g) of pentacene.
An inherent limitation exists for small molecular weight organics. Due to the
weak intermolecular bonding, the energy required to disrupt the crystalline order is
relatively small, allowing for a high density of defects and impurities to be incorporated
(Kitaigorodsky 1973; Silinsh et al. 1994). To achieve high quality crystals, growth in an
equilibrium-limited regime may be preferred, for example, employing small temperature
gradients and boundary layer diffusion (Kloc et al. 1997). Since the transport from the
vapor to the solid phase is driven primarily by the temperature gradient across the
boundary layer, rapid growth of high quality crystals is thus difficult.
Another solution may employ nucleation and growth of crystals from a
supercritical fluid. Supercritical CO
2
has been used extensively in the chemical, food, and
pharmaceutical industries for extraction, owing to the chemical inertness and high
solubility of the constituent organics in it. Thus, high concentrations of organic molecules
in CO
2
can in principle be achieved. Furthermore, this fluid is interesting from a mass
transport point of view, since compressibility, viscosity, mass and thermal diffusivity all
tend to be much higher in the supercritical phase, or diverge at the critical point. Thus,
large nucleation driving forces, as well as rapid molecular transport are possible upon the
transition from the super- to the sub-critical phases, all at relatively low temperatures,
albeit by means of pressure change at very high (>70atm) pressures.
8.5.5 Non-cleanroom device processing

200
Keeping with the theme of low-cost device processing, it becomes obvious that
traditional approaches to semiconductor manufacturing involving cleanrooms have
limited applicability. The high costs of cleanroom floor space and highly skilled labor are
prohibitive when trying to shift device fabrication from the high-value-added, low-yield
to the commodity-scale modes. Therefore, it is necessary to develop materials, device
structures, and processing methods whereby dust or ambient contamination does not
affect device performance. Alternatively, fully automated processes must be designed
whereby the samples are fully contained within a controlled, inert environment, from
loading of the substrates to their emergence in the final packaged state.
8.6 Summary
This thesis described the novel methods of organic vapor phase deposition (OVPD) and
organic vapor jet printing (OVJP) for the deposition and patterning of molecular organic
thin film electronics. The range of applications presently includes organic LEDs, thin-
film transistors, lasers, and photovoltaic cells on a variety of substrates and form factors,
including light-weight flexible plastics and fibers.
Theory, computer simulations, and OVPD experiments were reviewed, outlining
the material transport regimes in OVPD, as well as the equations predicting deposition
rate and doping concentration. The presence of a hydrodynamic boundary layer at the
substrate leads to diffusion-limited deposition, requiring a stochastic approach in
modeling the molecular nature of transport in confined geometries used for in-situ
patterning of active organic thin films. Monte-Carlo simulations were performed and
confirmed using shadow-masking experiments in OVPD.

201
Based on the knowledge of gas-phase dynamics in OVPD, a novel method
OVJP was developed for the direct patterned deposition of molecular organics. This
method potentially enables high-resolution printing of molecular organic electronics, at
nearly 100% efficient use of the source material. It avoids the use of shadow-masks for
organic layer deposition, which may greatly simplify the fabrication sequence and bill of
materials, lowering the overall cost of organic electronic devices.
After developing a semi-quantitative theory of OVJP, direct-simulation Monte-
Carlo simulations were used to predict the deposited pattern shape and resolution. An
experimental OVJP system was constructed and used to demonstrate the concept, by
printing high (>1000 dots per inch) resolution patterns of Alq
3
, -NPD, and pentacene at
ultra-high local deposition rates. A pentacene TFT was printed at >700/s local
deposition rate, having device performance characteristics similar to TFTs with highly
crystalline pentacene layers deposited at rates orders of magnitude lower.
As the optical and electrical performance of organic-based devices improves, and
their complexity increases, the development of low-cost fabrication methods becomes
important. Many novel processing methods have been developed, geared specifically to
organic semiconductors. The two carrier-assisted vapor-phase techniques described here
show significant promise, with OVPD already entering the commercial market, after
nearly 10 years of its scientific development. While it is hoped that OVJP will meet a
similar fate, many exciting scientific and technological challenges still await. There is
plenty of room for improvements, innovations, and radical new concepts in device
processing.


202
Chapter 9: References
Adachi, C., Baldo, M.A., Thompson, M.E. and Forrest, S.R. (2001). "Nearly 100%
internal phosphorescence efficiency in an organic light-emitting device." J. Appl.
Phys. 90(10): 5048.
Akiyama, Y. and Imaishi, N. (1995). "Applicability of one-dimensional diffusion-model
for step coverage analysis - comparison with a simple Monte-Carlo method."
Appl. Phys. Lett. 67(5): 620.
Askeland, R.A., W.D., C. and Sperry, W.R. (1988). "The 2-nd Generation Thermal Inkjet
Structure." Hewlett-Packard Journal 39(4): 28.
Baldo, M.A., Adachi, C. and Forrest, S.R. (2000). "Transient analysis of energy transfer
in organic electrophosphorescence I." Phys. Rev. B 62(16): 10958.
Baldo, M.A., Adachi, C. and Forrest, S.R. (2000). "Transient analysis of energy transfer
in organic electrophosphorescence II." Phys. Rev. B 62(16): 10967.
Baldo, M.A., Deutsch, M., Burrows, P.E., Gossenberger, H.F., Gerstenberg, M., Ban,
V.S. and Forrest, S.R. (1998). "Alq3 undoped OLED by OVPD." Adv. Mater.
10(1505).
Baldo, M.A., Kozlov, V.G., Burrows, P.E., Forrest, S.R., Ban, V.S., Koene, B. and
Thompson, M.E. (1997). "Low pressure organic vapor phase deposition of small
molecular weight organic light emitting device structures." Appl. Phys. Lett. 71:
3033.

203
Baldo, M.A., O'Brien, D.F., Thompson, M.E. and Forrest, S.R. (1999). "The excitonic
singlet-triplet ratio in a semiconducting organic thin film." Phys. Rev. B
Bardsley, J.N. (2003). Roadmap toward flexible displays. Vale of Glamorgan, U.S.
Display Consortium.
Bardsley, J.N. (2004). "International OLED Technology Roadmap." IEEE J. Sel. Top.
Quant. 10(1): 3.
Baude, P.F., Ender, D.A., Haase, M.A., Kelley, T.W., Muyres, D.V. and Theiss, S.D.
(2003). "Pentacene-based radio-frequency identification circuitry." Appl. Phys.
Lett. 82(22): 3964.
Bharathan, J. and Yang, Y. (1998). "Polymer electroluminescent devices processed by
inkjet printing: I. Polymer light-emitting logo." Appl. Phys. Lett. 72: 2660.
Bird, G.A. (1994). Molecular Gas Dynamics and the Direct Simulation of Gas Flows,
Oxford University Press.
Bird, R., Stewart, W. and Lightfoot, E. (1996). Momentum, Heat and Mass Transfer,
John Wiley & Sons.
Blanchet, G.B., Loo, Y.L., Rogers, J.A., Gao, F. and Fincher, C.R. (2003). "Large area,
high resolution, dry printing of conducting polymers for organic electronics."
Appl. Phys. Lett. 82(3): 463.
Blanchet, G.B. and Rogers, J.A. (2003). "Printing techniques for plastic electronics."
Journal of Imaging Science and Technology 47(4): 296.
Boderover, E. and Wagner, S. (2004). "A woven inverter circuit for e-textile
applications." IEEE El. Dev. Lett. 25(5): 295.

204
Bouchoms, I.P.M., Schoonveld, W.A., Vrijmoeth, J. and Klapwijk, T.M. (1999).
"Morphology identification of the thin film phases of vacuum evaporated
pentacene on SiO2 substrates."
Brooks, J., Babayan, Y., Lamansky, S., Djurovich, P.I., Tsyba, I., Bau, R. and Thompson,
M.E. (2002). "Synthesis and characterization of phosphorescent cyclometalated
platinum complexes." Inorganic Chemistry 41: 3055.
Brown, A., Jarrett, C., deLeeuw, D. and Matters, M. (1997). "Field-effect transistors
made from solution-processed organic semiconductors." Syn. Met. 88(1): 37.
Brown, J.J. (2004). Vice President Technology Development for Universal Display
Corporation.
Bulovic, V., Deshpande, R., Thompson, M.E. and Forrest, S.R. (1999). "Tuning the color
emission of thin film molecular organic light emitting devices by the solid state
solvation effect." Chemical Physics Letters 308: 317.
Bulovic, V. and Forrest, S.R. (2000). Polymeric and molecular organic light emitting
devices: A comparison. Electroluminescence II. 65: 1.
Burrows, P.E., Shen, Z., Bulovic, V., McCarty, D.M., Forrest, S.R., Cronin, J.A. and
Thompson, M.E. (1996). "Relationship between electroluminescence and current
transport in organic heterojunction light-emitting devices." J. Appl. Phys. 79:
7991.
Calvert, P. (2001). "Inkjet Printing for Materials and Devices." Chemistry of Materials
13: 3299.
Campbell, R.B., Trotter, J. and Robertson, J.M. (1961). "Crystal and molecular structure
of pentacene." Acta Crystallographica 17(7): 705.

205
Casalis, L., Danisman, M.F., Nickel, B., Bracco, G., Toccoli, T., Ianotta, S. and Scoles,
G. (2003). "Hyperthermal molecular beam deposition of highly ordered organic
thin films." Phys. Rev. Lett. 90(20): 206101.
Creagh, L.T. and McDonald, M. (2003). "Design and performance of inkjet print heads
for non-graphic-arts applications." MRS Bull. 28(11): 807.
De la Mora, J.F. and Rosell-Llompart, J. (1989). "Aerodynamic focusing of heavy
molecules in seeded supersonic jets." Journal of Chemical Physics 91(4): 2603.
Deuflhard, P. (1974). "A modified Newton method for the solution of ill-conditioned
systems of non-linear equations with application to multiple shooting." Numerical
Mathematics 22: 289.
Dimitrakopoulos, C., Brown, A. and Pomp, A. (1996). "Molecular beam deposited thin
films of pentacene for organic field effect transistor applications." J. Appl. Phys.
80(4): 2501.
Dimitrakopoulos, C.D. and Mascaro, D.J. (2001). "Organic thin-film transistors: A
review of recent advances." IBM J. Res. Dev. 45(1): 11.
Elrod, S.A., Hadimioglu, B., Khuri-Yakub, B.T., Rawson, E.G., Richley, E., Quate, C.F.,
Mansour, N.N. and Lundgren, T.S. (1989). "5 micron ink jet droplets." J. Appl.
Phys. 65: 3441.
Emslie, A.G., Bonner, F.T. and Peck, L.G. (1958). "Flow of a Viscous Liquid on a
Rotating Disk." J. Appl. Phys. 29(5): 858.
Eres, D. (1991). "High-speed epitaxy using supersonic molecular jets." Materials
Research Society Symposium Proceedings 201: 11.

206
Forrest, S.R. (1997). "Ultrathin organic films grown by organic molecular beam
deposition and related techniques." Chem. Rev. 97(6): 1793.
Forrest, S.R. (2004). "The path to ubiquitous and low-cost organic electronic appliances
on plastic." Nature 428(April): 911.
Garnier, F., Hajlaoui, R., Yassar, A. and Srivastava, P. (1994). "All-polymer field-effect
transistor realized by printing techniques." Science 265(5179): 1684.
Gleskova, H., Wagner, S., Gasparik, V. and Kovac, P. (2001). "150 degrees C amorphous
silicon thin-film transistor technology for polyimide substrates." J. Electrochem.
Soc. 148(7): G370.
Griffiths, S.K. and Nilson, R.H. (1998). "Optimum conditions for composites fiber
coating by chemical vapor infiltration." J. Electrochem. Soc. 145(4): 1263.
Guevremont, J.M., Sheldon, S. and Zaera, F. (2000). "Design and characterization of
collimated effusive gas beam sources: Effect of source dimensions and backing
pressure on total flow and beam profile." Rev. Sci. Instr. 71(10).
Gundlach, D., Jackson, T., Schlom, D. and Nelson, S. (1999). "Solvent-induced phase
transition in thermally evaporated pentacene films." Appl. Phys. Lett. 74(22):
3302.
Gundlach, D., Kuo, C., Nelson, S. and Jackson, T. (1999). Organic Thin Film Transistors
with Field Effect Mobility > 2 cm/Vs. 57th Device Research Conference Digest.
Gundlach, D., Lin, Y., Jackson, T., Nelson, S. and Schlom, D. (1997). "Pentacene organic
thin-film transistors - Molecular ordering and mobility." IEEE El. Dev. Lett.
18(3): 87.

207
Gutmann, F. and Lyons, L.E. (1981). Organic Semiconductors. Malabar, Florida, Robert
E. Krieger Publishing Company.
Halsey, T.C. (2000). "Diffusion-limited aggregation: A model for pattern formation."
Physics Today 53(11): 36.
Hebner, T.R., Wu, C.C., Marcy, D., Lu, M.H. and Sturm, J.C. (1998). "Ink-jet printing of
doped polymers for organic light emitting devices." Appl. Phys. Lett. 72(5): 519.
Heringdorf, F.M.z., Reuter, M.C. and Tromp, R.M. (2001). "Growth dynamics of
pentacene thin films." Nature 412(6846): 517.
Heringdorf, F.M.z., Reuter, M.C. and Tromp, R.M. (2004). "The nucleation of pentacene
thin films." Applied Physics a-Materials Science & Processing 78(6): 787.
Horowitz, G. (1999). "organic TFT review." J. Mater. Chem. 9: 2021.
Hsu, P.I., Bhattacharya, R., Gleskova, H., Huang, M., Xi, Z., Suo, Z., Wagner, S. and
Sturm, J.C. (2002). "Thin-film transistor circuits on large-area spherical surfaces."
Appl. Phys. Lett. 81(9): 1723.
Huang, D., Liao, F., Molesa, S., Redinger, D. and Subramanian, V. (2003). "Plastic-
compatible low resistance printable gold nanoparticle conductors for flexible
electronics." J. Electrochem. Soc. 150(7): G412.
Irisawa, T. (2003). Crystal Growth Technology, William Andrew Publishing/Noyes.
Israelachvili, J. (1992). Intermolecular and surface forces. San Diego, Academic Press
Ltd.
John, J.E.A. (1984). Gas Dynamics. Boston, Allyn and Bacon, Inc.

208
Katz, H.E., Bao, Z. and Gilat, S.L. (2001). "Synthetic chemistry for ultrapure,
processable, and high-mobility organic transistor semiconductors." Accounts of
Chemical Research 34: 359.
Kitaigorodsky, A.I. (1973). Molecular Crystals and Molecules. New York and London,
Academic Press.
Kloc, C. and Laudise, R.A. (1998). "Vapor pressures of organic semiconductors: alpha-
hexathiophene and alpha-quaterthiophene." J. Cryst. Growth 193(4): 563.
Kloc, C., Simpkins, P., Siegrist, T. and Laudise, R. (1997). "Physical vapor growth of
centimeter-sized crystals of alpha-hexathiophene." J. Cryst. Growth 182(3-4):
416.
Krause, B., Durr, A.C., Ritley, K., Schreiber, F., Dosch, H. and Smilgies, D. (2002).
"Structure and growth morphology of an archetypal system for organic epitaxy:
PTCDA on Ag(111)." Phys. Rev. B 66: 235404.
Kymissis, I., Dimitrakopoulos, C. and Purushothaman, S. (2001). "High-performance
bottom electrode organic thin-film transistors." IEEE Transaction on Electron
Devices 48(6): 1060.
Le, H.P. (1998). "Progress and trends in ink-jet printing technology." Journal of Imaging
Science and Technology 42(1): 49.
Lee, S., Lee, J., Kim, J.H. and Chung, H.S. (2003). "Effect of post-detachment cleaning
on the hydrophilic nature of nickel thin film nozzle plates for piezo ink-jet printer
heads." Plating and Surface Finishing 90(9): 58.

209
Lin, Y., Gundlach, D., Nelson, S. and Jackson, T. (1997). "Stacked pentacene layer
organic thin-film transistors with improved characteristics." IEEE El. Dev. Lett.
18(12): 606.
Lukas, S., Witte, G. and Woll, C. (2002). "Novel mechanism for molecular self-assembly
on metal substrates: Unidirectional rows of pentacene on Cu(110) produced by a
substrate-mediated repulsion." Phys. Rev. Lett. 88(2).
McCarty, G.S. and Weiss, P.S. (1999). "Scanning probe studies of single nanostructures."
Chem. Rev. 99(7): 1983.
Moons, E. (2002). "Conjugated polymer blends: linking film morphology to performance
of light emitting diodes and photodiodes." Journal of Physics: Condensed Matter
14: 12235.
Olsen, G.H. (1982). Vapor-phase epitaxy of GaInAsP. GaInAsP. T. P. Pearsall. New
York, John Wiley & Sons Ltd.
Oxtoby, D.W. and Nachtrieb, N.H. (1990). Principles of Modern Chemistry. Orlando,
Saunders College Publishing.
Pangal, K., Sturm, J.C. and Wagner, S. (2001). "Integrated Amorphous and
Polycrystalline Silicon Thin-Film Transistors in a Single Silicon Layer." IEEE
Transaction on Electron Devices 48(4): 707.
Paul, K.E., Wong, W.S., Ready, S.E. and Street, R.A. (2003). "Additive jet printing of
polymer thin-film transistors." Appl. Phys. Lett. 83(10): 270.
Percin, G. and Khuri-Yakub, B.T. (2003). "Piezoelectric droplet ejector for ink-jet
printing of fluids and solid particles." Rev. Sci. Instr. 74(2): 1120.

210
Pope, M. and Swenberg, C.E. (1982). Electronic Processes in Organic Crystals. Oxford,
New York, Clarendon Press, Oxford University Press.
Pratontep, S., Brinkmann, M., Nuesch, F. and Zuppiroli, L. (2004). "Correlated growth in
ultrathin pentacene films on silicon oxide: Effect of deposition rate." Phys. Rev. B
69(16).
Rogers, J.A. (2001). "Rubber stamping for plastic electronics and fiber optics." MRS
Bull. 26(7): 530.
Rogers, J.A. and Bao, Z. (2002). "Printed plastic electronics and paperlike displays."
Journal of Polymer Science Part A: Polymer Chemistry 40: 3327.
Roy, S. and Raju, R. (2003). "Modeling gas flow through microchannels and nanopores."
J. Appl. Phys. 93(8): 4870.
Ruiz, R., Nickel, B., Koch, N., Feldman, L.C., Haglund, R.F., Kahn, A., Family, F. and
Scoles, G. (2003). "Dynamic scaling, island size distribution, and morphology in
the aggregation regime of submonolayer pentacene films." Phys. Rev. Lett.
91(13).
Schlichting, H. (1968). Boundary-Layer Theory, McGraw-Hill.
Schuerlein, T., Schmidt, A., Lee, P., Nebesny, K. and Armstrong, N. (1995). "Large
molecule epitaxy on single-crystal metals, insulators and single-crystals and
MBE-grown layered semiconductors." Japanese Journal of Aapplied Physics Part
1-Regular Papers Short Notes & Review Papers 34(7B): 3837.
Scoles, G., Ed. (1988). Atomic and Molecular Beam Methods, Oxford.

211
Shaked, S., Tal, S., Roichman, Y., Razin, A., Xiao, S., Eichen, Y. and Tessler, N. (2003).
"Charge density and film morphology dependence of charge mobility in polymer
field-effect transistors." Adv. Mater. 15(11): 913.
Shields, J.P. (1992). "Thermal Inkjet Review, or How Do Dots Get From The Pen to The
Page." Hewlett-Packard Journal 43(4): 67.
Shtein, M., Gossenberger, H.F., Benziger, J.B. and Forrest, S.R. (2001). "Material
transport regimes and mechanisms for growth of molecular organic thin films
using low-pressure organic vapor phase deposition." J. Appl. Phys. 89(2): 1470.
Shtein, M., Mapel, J., Benziger, J.B. and S.R., F. (2002). "Effects of film morphology
and gate dielectric surface preparation on the electrical characteristics of organic-
vapor-phase-deposited pentacene thin-film transistors." Appl. Phys. Lett. 81(2):
268.
Shtein, M., Peumans, P., Benziger, J.B. and Forrest, S.R. (2003). "Micropatterning of
small molecular weight organic semiconductor thin films using organic vapor
phase deposition." J. Appl. Phys. 93(7): 4005.
Shtein, M., Peumans, P., Benziger, J.B. and Forrest, S.R. (2004). "Direct mask- and
solvent-free printing of molecular organic semiconductors." Adv. Mater.
(Accepted April 2004).
Silinsh, E.A. and Capek, V. (1994). Organic Molecular Crystals. New York, American
Institute of Physics.
Sirringhaus, H., Kawase, T. and Friend, R.H. (2001). "High-resolution ink-jet printing of
all-polymer transistor circuits." MRS Bull. 26(7): 539.

212
Sirringhaus, H., Kawase, T., Friend, R.H., Shimoda, T., Inbasekaran, M., Wu, W. and
Woo, E.P. (2000). "High-resolution inkjet printing of all-polymer transistor
circuits." Science 290: 2123.
Stechelmacher, W. (1986). "Knudsen flow 75 years on: the current state of the art for
flow of rarefied gases in tubes and systems." Rep. Prog. Phys. 49: 1083.
Stringfellow, G.B. (1989). Organometallic Vapor-Phase Epitaxy. London, Academic.
Sze, S. (1969). Physics of semiconductor devices. New York, Wiley-Interscience.
Tang, C.W. (1986). "2-Layer organic photovoltaic cell." Appl. Phys. Lett. 48(2): 183.
Tang, C.W. and VanSlyke, S.A. (1987). "Heterostructure OLED." Appl. Phys. Lett. 51:
913.
Tian, P.F., Bulovic, V., Burrows, P.E., Gu, G., Forrest, S.R. and Zhou, T.X. (1999).
"Precise, scalable shadow mask patterning of vacuum-deposited organic light
emitting devices." J. Vac. Sci. Techn. A 17(5): 2975.
Tian, P.F., Burrows, P.E. and Forrest, S.R. (1997). "Photolithographic patterning of
vacuum-deposited organic light emitting devices." Appl. Phys. Lett. 71(22): 3197.
Tison, S.A. (1993). "Experimental data and theoretical modeling of gas flows through
metal capillary leaks." Vacuum 44(11-12): 1171.
Toivakka, M. (2003). Numerical Investigation of Droplet Impact Spreading in Spray
Coating of Paper. 2003 TAPPI 8th Advanced Coating Fundamentals Symposium,
Chicago, TAPPi Press.
Vasenkov, A.V., Belikov, A.E., Sharafutdinov, R.G. and Kuznetsov, O.V. (1995). "Flow-
field properties under deposition of films from low-density jets." J. Appl. Phys.
77(9): 4757.

213
Venables, J.A., Spiller, G.D.T. and Hanbucken, M. (1984). "Nucleation and growth of
thin films." Reports on Progress in Physics 47: 399.
Verlaak, S., Steudel, S., Heremans, P., Janssen, D. and Deleuze, M.S. (2003).
"Nucleation of organic semiconductors on inerg substrates." Phys. Rev. B 68:
195409.
Wang, J.Z., Zheng, Z.H., Li, H.W., Huck, W.T.S. and Sirringhaus, H. (2004). "Dewetting
of conducting polymer inkjet droplets on patterned surfaces." Nature Materials
3(3): 171.
Wang, L., Ast, D.G., Bhargava, P. and Zehnder, A.T. (2003). "Curved Silicon
Electronics." Materials Research Society Symposium Proceedings 769: H2.8.1.
Witten, T.A. and Sander, L.M. (1981). "Diffusion-Limited Aggregation, a Kinetic
Critical Phenomenon." Phys. Rev. Lett. 47(19): 1400.
Witten, T.A. and Sander, L.M. (1983). "Diffusion-Limited Aggregation." Phys. Rev. B
27(9): 5686.
Wolf, S. and Tauber, R.N. (1999). Silicon Processing for the VLSI Era : Process
Technology, Lattice Press.
Wu, M. and Wagner, S. (2001). "Amorphous silicon crystallization and polysilicon thin
film transistors on SiO2 passivated steel foil substrates." Applied Surface Science
175: 753.
Wulu, H.C., Saraswat, K.C. and McVittie, J.P. (1991). "Simulation of mass-transport for
deposition in via holes and trenches." J. Electrochem. Soc. 138(6): 1831.

214
Yang, Y., Chang, S.C., Bharathan, J. and Liu, J. (2000). "Organic/polymeric
electroluminescent devices processed by hybrid ink-jet printing." J. Mater. Sci.-
Mater. Electron. 11(2): 89.
Yase, K., Takahashi, Y., Ara-Kato, N. and Kwazu, A. (1995). "vapor pressure of organics
(CuPc, etc.)." Jap. J. Appl. Phys. Part 1 34: 636.
Yase, K., Yoshida, Y., Uno, T. and Okui, N. (1996). "Vapor pressure of organics." J.
Cryst. Growth 166: 942.

Você também pode gostar