SOLUTION PROCESSED ORGANIC SEMICONDUCTOR THIN-FILM TRANSISTORS FOR FLEXIBLE ELECTRONICS: DEVICE PHYSICS, DEVICE MODELING, FABRICATION TECHNOLOGY, AND INTERFACE ENGINEERING A DISSERTATION SUBMITTED TO THE DEPARTMENT OF ELECTRICAL ENGINEERING AND THE COMMITTEE ON GRADUATE STUDIES OF STANFORD UNIVERSITY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY Zihong (Bill) Liu November 2009 [Defense Date: September 10 th , 2009]
166
Embed
SOLUTION PROCESSED ORGANIC SEMICONDUCTOR THIN-FILM ...zc714wk4534/Stanford PhD... · semiconductor based thin-film transistors ... performance organic semiconductor materials and
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
SOLUTION PROCESSED ORGANIC SEMICONDUCTOR
THIN-FILM TRANSISTORS FOR FLEXIBLE ELECTRONICS:
DEVICE PHYSICS, DEVICE MODELING, FABRICATION
TECHNOLOGY, AND INTERFACE ENGINEERING
A DISSERTATION
SUBMITTED TO THE DEPARTMENT OF ELECTRICAL ENGINEERING
AND THE COMMITTEE ON GRADUATE STUDIES
OF STANFORD UNIVERSITY
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
Zihong (Bill) Liu
November 2009
[Defense Date: September 10th, 2009]
http://creativecommons.org/licenses/by-nc/3.0/us/
This dissertation is online at: http://purl.stanford.edu/zc714wk4534
I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.
Yoshio Nishi, Primary Adviser
I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.
Zhenan Bao
I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.
Krishna Saraswat
Approved for the Stanford University Committee on Graduate Studies.
Patricia J. Gumport, Vice Provost Graduate Education
This signature page was generated electronically upon submission of this dissertation in electronic format. An original signed hard copy of the signature page is on file inUniversity Archives.
iii
iv
Abstract
Organic or carbon electronics has been a fast-growing field in recent years covering a
broad range from nanoelectronic devices to macroelectronic systems. Besides the single-
graphene or single-carbon nanotube transistor toward extending the scaling limit of
traditional silicon metal-oxide-semiconductor field-effect transistor (MOSFET), organic
semiconductor based thin-film transistors have been actively investigated due to their
promise in large-area electronics fabricated on flexible substrates using low-cost
unconventional means, such as low/room-temperature printing and roll-to-roll
processing. This dissertation focuses on the study of device physics, device modeling,
fabrication technology, and interface engineering for solution-processed organic field-
effect transistors (SPOFET) for flexible electronics applications. There are primarily
four parts of contributions originated from this dissertation work.
The first part introduces the design and demonstration of high-performance, low-
voltage flexible SPOFETs fabricated on plastic substrates with a carrier mobility over
0.2 cm2/Vs, a turn-on voltage of near 0 V, and a record low subthreshold slope of ~80
mV/dec in ambient conditions. These exceptional characteristics are achieved by novel
device architecture design, 3-D statistical modeling for solution-shearing process
optimization, and phenyl-terminated self-assembled monolayer (SAM) based interface
engineering.
In the second part, SAM relevant physical effects and chemistry effects at the
organic semiconductor-dielectric interface are systematically investigated. Through
careful selection of a group of phenyl-terminated SAMs, we elucidate how the
performance and reliability of organic transistors are controlled by the critical
semiconductor-dielectric interfacial SAMs. In addition, we briefly introduce a spin-
v
coating process for depositing high-quality phenyl-terminated SAMs for organic
electronics applications.
The third part focuses on the device physics and device modeling of organic
transistors. In this dissertation work, we have proposed and developed a universal
physical model for organic transistors by incorporating both the charge injection effects
and charge transport properties, and successfully applied it to resolve many elusive
physical phenomena observed so far, such as the peculiar mobility scaling behavior with
respect to the channel length, the contact resistance effect, and the mysterious surface
potential profiles of organic transistors which have been experimentally probed yet
poorly understood. Of particular importance is that we discover an overshoot region in
the mobility scaling behavior and identified the existence of a critical channel length for
the peak field-effect mobility.
In the last part, we investigate novel contact engineering for organic transistors
toward lowering charge injection barrier and reducing the interfacial disorder width or
localization states. We have explored and demonstrated Fermi-level depinning at the
metal-organic interface for low-resistance Ohmic contacts by inserting an ultrathin
interfacial Si3N4 insulator in between. The contact behavior is successfully tuned from
rectifying to quasi-Ohmic and to tunneling by varying the Si3N4 thickness within 0-6 nm.
Detailed physical mechanisms of Fermi-level pinning/depinning responsible for the
metal-organic semiconductor contact behavior are clarified based on a proposed
lumped-dipole model.
vi
vii
To My Family Who Raised Me Up in the World of Humanity
To My Teachers Who Raised Me Up in the World of Science
viii
ix
Acknowledgments
This journey would not have been possible without the support, guidance, inspiration
and encouragement from many people to whom I am deeply and forever indebted.
First of all, I would like to extend my heartfelt gratitude to my supervisors, Professor
Yoshio Nishi in the Department of Electrical Engineering and Professor Zhenan Bao in
the Department of Chemical Engineering, who both have been rigorously advising,
strongly supporting, and generously helping me throughout my graduate studies and
research. Imagining working in both electrical engineering and chemical engineering as
handily as I do today would never come true without extraordinary guidance and
support from both of my supervisors in the two distinctive departments. Professor
Yoshio Nishi has over forty years of research and management experience in the
semiconductor industry and academia. His broad intelligence and unique insight to the
semiconductor technology have been a constant source of inspiration and
encouragement to my graduate research as well as future career. In addition, his well-
respected leadership and personality deeply influence me and set an example for us
students. Professor Zhenan Bao is a well-known expert in the field of organic
electronics and has made revolutionary contributions to the development of high
performance organic semiconductor materials and devices. Her keen sense of
experiments, grand ambitiousness of scientific research and enduring patience provided
me invaluable help to thrive in the emerging field of nano-/macro-electronics. I also
appreciate both supervisors for guiding me how to develop to a true research scientist.
I am also grateful to Professor Alberto Salleo and Professor Krishna Saraswat for
their valuable suggestions on the research and for their precious time and efforts as
members of my reading and defense committee.
x
Over the past three years of my PhD studies and research at Stanford University,
there have been constant valuable interactions and collaboration with members from
various groups focusing on semiconductor devices, organic electronics, and
nanotechnology, including Dr. Hector Becerril, Dr. Mark Roberts, Dr. Peng Wei, Dr.
Liangbing Hu, Dr. Eric Verploegen, Ajay Virkar, Dr. Toshifumi Irisawa, Dr. Joon Hak
Oh, Dr. Blanka Magyari-Kope, and Dr. Masaharu Kobayashi. I appreciate their
tremendous help and great collaboration that made this work possible. In addition, I
would like to thank Professor Yoshio Nishi’s entire group and Professor Zhenan Bao’s
entire group where numerous colleagues offered support and help along with my
research in the labs.
Special thanks go to the Toshiba Inc. who has been sponsoring this project through
Stanford CIS-FMA program. Particularly, Dr. Bipul Paul and Dr. Masaki Okajima have
been providing suggestions and discussions as the CIS-FMA project mentors, and I
appreciate their support and contributions to the success of this project.
Also, I wish to take this opportunity to extend my sincere thanks to all of my
teachers from kindergarten to graduate school who have truly raised me up in the world
of science, step by step and word by word.
Research is tough. Failures and frustrations always come along the way to the
ultimate success. I can not sustain it without mental well-being and spiritual support,
most importantly, from my family. I am indebted to my parents, my three elder sisters,
and my girlfriend for their never-fading and purely selfless love, support, encouragement
and prayers throughout my whole life. I will try my best to spend the rest of my life
making them as happy as they have made me.
xi
Contents
Abstract..................................................................................................... iv
Acknowledgments .................................................................................... ix
List of Tables........................................................................................... xv
List of Figures ....................................................................................... xvii
force, and (f) exchange/Pauli repulsion. An interface dipole is always created
upon the M/O junction formation........................................................................... 110
Figure 5.9 M/O interface energy band diagrams based on the proposed lumped
interface dipole model: (a) before M/O interface formation; (b) upon M/O
interface formation, Fermi-level pinning arises from various interface dipole
elements; (c) after inserting an ultrathin Si3N4 insulator, the Fermi-level
depinning takes effect by blocking the physisorption/chemisorption................ 112
Figure 5.10 Simulation results based on the proposed simple depinning model
showing the dominant mechanisms responsible for the contact resistance
behavior after insertion of the ultrathin insulating layer at the M/O interface. 114
1
Chapter 1
Introduction
We choose to go to the moon. We choose to go to the moon in this decade and do the other
things, not because they are easy, but because they are hard, because that goal will serve to
organize and measure the best of our energies and skills, because that challenge is one that we
are willing to accept, one we are unwilling to postpone, and one which we intend to win.
- John F. Kennedy, 1962
Imagine a world filled with flexible personal digital assistants (PDA) or cellphones,
large-area rollable displays, bendable scanners, wearable electronic clothes, as well as
many other electronic devices with versatile functions, fantastic appearance and
mechanical flexibility. An intuitive question would be if there is any possibility to
integrate the well-established semiconductor device technology traditionally based on
silicon hard materials into these novel applications. The prospect seems optimistic, yet,
significant challenges remain both in scientific understanding of the newly emerging
phenomena and in engineering of the reliable devices.
2 CHAPTER 1. INTRODUCTION
1.1 Historical Overview of Semiconductor Electronics
It is natural to learn from history since history always guides future. Along with the
development of semiconductor electronics over the past few decades, people have been
broadly exploring two intriguing fields: How to make the devices extremely small?
And, how to make the systems extremely large?
Specifically, the first one is toward the limit of scaling down each component,
namely for “nanoelectronics” nowadays, and the other one is toward the limit of scaling
up the system, namely for “macroelectronics”. Identifying their specific applications and
understanding their unique characteristics are required for the success in engineering
devices and systems toward both extremes. To be accurate while without loss of
generality, we’ll focus on discussing a most representative and impotant applications for
each area, i.e. very large scale integration (VLSI) circuit chips and large area displays,
respectively. Figure 1.1 shows the historical development of traditional silicon metal-
oxide-semiconductor field-effect transistors (MOSFET) which is the central component
of VLSI chips. The primary focuses in such nanoelectronics area include reducing the
cost per function or cost per transistor, scaling down of the device to smaller feature
size, packing higher integration density, increasing circuit operation speed, and
minimizing chip power dissipation. The future, however, is challenged by the scaling
bottleneck beyond current 15/22 nm feature size due to fundamental physical effects
arising from materials, processes, and the device architectures [1].
In contrast, for macroelectronic applications as represented by displays, efforts are
centered on reducing the cost per area in addition to expanding the system to larger size
with increased resolution, higher mechanical flexibility, enhanced robustness, and lighter
weight. The requirements of the dimensional scale and operation frequency on its core
component for the backplane, typically hydrogenated amorphous silicon (α-H:Si) or
polycrystalline silicon thin-film transistors (TFT), are less stringent than the MOSFETs
used for high-performance VLSI circuits. The future of macroelectronic applications
relies more heavily on the cost per area and ubiquity limit opposed by traditional
materials, processes, and devices employed in the large-area electronics.
CHAPTER 1. INTRODUCTION 3
1970 1980 1990 2000 2010 202010-9
10-7
10-5
10-3
10-1
101
Year
Cos
t per
Tra
nsis
tor (
$)
100
101
102
103
104 Minim
um Feature S
ize (nm)
Figure 1.1 The historical development of traditional silicon metal-oxide-semiconductor field-effect transistors (MOSFET) used for VLSI circuits. Source data: Intel and ITRS.1
1.2 Carbon Materials for Nano- and Macro-Electronics
In order to circumvent the aforementioned limits, there has been extensive research
work in recent years on exploring novel semiconductor materials and their relevant
processes and devices. Among many others, carbon (organic) electronic materials are
found very promising in both nanoelectronics and macroelectronics areas. Carbon
electronics has emerged as a fast-growing field covering a broad range from
nanoelectronic devices to macroelectronic systems.
On one hand, individual carbon nanotube, graphene, organic molecule, or carbon
nanoribbon was reported having superior electrical properties with extraordinary
intrinsic carrier mobility over 100,000 cm2/Vs, up to 1,000 times higher than in bulk
silicon [2-3]. Consequently, single-graphene [4] or single-carbon nanotube transistor [5]
provides a possible route toward extending the scaling limit of traditional silicon
1 Intel and Dataquest reports (December 2002), see Gordon E. Moore, “Our Revolution”, http://www.sia-online.org/galleries/default-file/Moore.pdf; International Technology Roadmap for Semiconductors (ITRS), http://www.itrs.net/
4 CHAPTER 1. INTRODUCTION
MOSFET. The main hurdle at present is the poor reproducibility of identical devices
required for large scale integrated circuits. How to precisely control the device
placement and integration at will is a major challenge on pushing forward the carbon
nanomaterials into high performance nanoelectronic applications.
On the other hand, carbon nanotube, small-molecule and polymer organic
semiconductor thin-film1 transistors can be readily fabricated with large-scale control of
positioning and integration, though the electrical performance and dimensional scaling
capability is more confined. Typically, the carrier mobility in these thin films is on the
order of 0.1-10 cm2/Vs [6], two to three orders lower than that in single crystal silicon,
thus being inappropriate for high speed computing where traditional single crystal
silicon still dominates. Fortunately, such performance still approaches, if not surpasses,
that of hydrogenated amorphous silicon (α-Si:H) TFTs which are widely used in display
backplanes [7]. Moreover, their unique possibility of low-temperature deposition and
excellent mechanical flexibility enable us to fabricate large-area electronics on flexible,
light-weight substrates using low-cost unconventional means, such as room-temperature
printing and roll-to-roll processing [8]. Therefore, carbon semiconductor based thin film
devices are highly promising for novel and practical macroelectronic applications where
relatively low mobility 2 and medium device feature size satisfy the requirements.
Foreseeable examples include electronic paper, flexible displays, ubiquitous electronic
walls, organic radio-frequency identification (RFID) tags, organic solar cell and light-
emitting diode arrays, chemical sensors, and e-clothes.
For this dissertation work, we focus on studying the organic semiconductor based
thin-film transistors for the aforementioned macroelectronics and flexible electronics
applications.
1 Thin film typically consists of arrays or a number of individual molecules or nanotubes that form a thin layer of film. 2 In comparison to single crystal silicon or individual carbon nanotube, graphene, and nanoribbon for high performance nanoelectronic devices.
CHAPTER 1. INTRODUCTION 5
1.3 Motivations and Objectives
As discussed in Section 1.2, organic semiconductor based thin-film transistors have
attracted wide attention in recent years due to their promise in low-cost, large-area,
flexible electronics and the compatibility with scalable fabrication process. Figure 1.2
illustrates the typical structure of a bottom gate, top contact and bottom gate, bottom
contact organic TFT. Besides conducting materials for the source (S), drain (D), and
gate (G), dielectrics for isolating the channel from the gate and organic semiconductor
thin films acting as the active channel are also crucial to the organic TFT operation.
GDielectric
Organic Semiconductor Thin FilmS D
GDielectric
S Organic Semiconductor D
Figure 1.2 Schematic structure of a typical (a) bottom-gate, top-contact and (b)bottom-gate, bottom-contact organic TFT. S: source, D: drain, G: gate.
Of particular interest in the device fabrication is how to deposit the active channel
materials. In general, organic semiconductor thin films can be either vapor processed or
solution processed, depending on their physical and chemical properties. Vacuum
evaporation as a mature fabrication technology is widely used for most organic small-
molecule semiconductors to grow highly uniform films on the substrate. This is
particularly important for large area circuits and systems. However, vapor process
suffers from relatively high temperature in evaporation, high fabrication cost from
expensive setup and material waste, and low throughput as confined by the equipment
space.
An alternative to the vacuum evaporation deposition method is solution processing.
The concept is simple: when the materials are soluble to some specific solvents, one can
6 CHAPTER 1. INTRODUCTION
deposit the film by drop casting, spin-coating, or other ways from the material’s solution
phase [9]. Advantages of solution processing of organic semiconductors include:
(1) It is compatible with printing techniques toward printed electronics1. Since
the devices can be fabricated with screen printing [10], inkjet printing [11-13],
or microcontact printing [14], they possess benefits from all these well-
developed existing processes.
(2) It is possible to manufacture the devices on flexible substrates using a fast,
roll-to-roll process. Figure 1.3 shows a conceptual roll-to-roll manufacturing
process for solution processed organic transistors. The flexible substrate
moves between two end rollers; modules (1)-(n) serve different layer
depositions. Configurations for each module include three substrate-
supporting transfer rollers, a shearing roller clamping the substrate with
designed pressure, and an organic solution tank close to the shearing roller.
Figure 1.3 A proposed roll-to-roll manufacturing process for solution processed organic thin-film transistors.
1 See http://flextech.org/.
CHAPTER 1. INTRODUCTION 7
(3) Low temperature processing becomes available. In solution phase deposition,
the temperature is typically below 100ºC or even down to room temperature,
being significantly lower than that required in vapor process [15].
(4) There is no throughput limit against evaporation setup space.
(5) Self-assembly in solution phase provides a novel method for versatile
patterning of the devices [16-17].
(6) The fabrication cost is considerably reduced based on the above superiority.
Despite the great potential of solution processed organic thin-film field-effect
transistors (SPOFET), their electrical performance as show in Figure 1.4 is still lagging
behind the counterparts based on vapor process due to unfavorable film microstructures
and poor uniformity. A number of issues remain for SPOFET although the
performance has been improved dramatically in the past decade.
displays [14, 23-24], and RFID tags [13, 25-26]. Compared to their counterparts on rigid
substrates like on glass or silicon wafer, flexible SPOFETs normally exhibit lower
performance. Typically, the charge carrier mobility is on the order of 10-2-10-1 cm2/Vs
and the subthreshold slope (SS), as illustrated in Figure 2.1, is either over 1 V/dec or
being overlooked although it is a critical parameter in determining the device switching
speed.
1log( ) ln10DS
GS
I kTSSV e
−∂= > ⋅
∂
Figure 2.1 Illustration of the subthreshold slope (SS), as defined by ∂(logIDS)/∂VGS, which reflects how fast the transistor is switched between the on and off state.
More recently, flexible SPOFETs with a mobility over 0.1 cm2/Vs [13, 24-30] or a
rarely low subthreshold slope down to 100 mV/dec [29] have been demonstrated by
virtue of improved processing methods, a series of newly discovered high-mobility and
soluble organic semiconductors, and novel dielectric materials [29]. Nevertheless,
obstacles in achieving both specifications simultaneously still appear [31], especially for
flexible SPOFETs based on common polymeric dielectrics such as photoresist [20],
hr to promote the cross-linking reaction. The second layer was spin-coated from a lower
concentration of PVP:EAD solution of 63 mg/mL at 5 krpm to better tune the surface
morphology and eliminate pinholes in the dielectric, and again followed by annealing.
Bilayer PVP-EADPEDOT:PSSITO
500 nm
ITO
PET(Polyethylene-Terephthalate)
Bilayer Cross-linked PVP (PVP-EAD)
Au
PEDOT:PSS
Au
SS S
S Si
175 μm
4T-TMSPBTS-SAM
N
O
O
n
OHO
N
O
O
n
OOH
SiO O
O
SiO
O
SiO
O
SiO
O
(a)
(b) (c)
Figure 2.2 (a) Device structure and (b) photograph of our flexible 4T-TMS SPOFETs. (c) Scanning electron microscopy (SEM) image of the bilayer PVP-EAD dielectric that possesses apparent single-layer morphology. SEM was performed at 3 keV in a tilt angle of 45º on an deliberately peeled-off area to examine both the surface and the cross section.
As shown in Figure 2.2 (c), our bilayer PVP-EAD dielectric of ~180 nm possesses
apparent single-layer morphology without distinguishable traces in between or on the
top surface, as revealed by scanning electron microscopy (SEM) [FEI XL30 Sirion]. The
device yield based on this bilayer PVP-EAD was found 50-60% higher than that of
Oxidation/Reduction peak of reference molecule with HOMO=-4.80eV
(a)
(b)
(c)
4.6 4.8 5.0 5.2 5.4 5.6 5.8 6.0 6.2
0
1000
2000
3000
4000
5000
Yie
ld1/
2 (cps
1/2 )
Yield (cps) Yield1/2 (cps1/2)
Energy (eV)
Yie
ld (c
ps)
0
10
20
30
40
50
60
70
Figure 2.3 (a) UV-vis absorption spectra of 4T-TMS in a solution/film phase. (b) Cyclic voltammety (CV) diagram of 4T-TMS with a reference compound of HOMO=-4.80 eV in 1,2-dichlorobenzene. (c) Photoelectron spectroscopy measurement results for 4T-TMS films.
Figure 2.5 Surface structure, water contact angle and cross-polarized microscope image of 4T-TMS film deposited by solution shearing process on (a) PTS treated substrate and (b) OTS treated substrate. Large crystalline domain films can be successfully deposited on PTS substrate, while nonuniform or little film is consistently observed on OTS substrate due to its dewettability to the 4T-TMS organic semiconductor solutions.
4T-TMS solutions are prepared using different solvents, including chlorobenzene, p-
xylene and 1,2-dichlorobenzene, at various concentrations (5 mg/mL-12 mg/mL). Next,
we used spin-coating, drop casting, and solution-shearing [15, 36-39, 43] methods to
deposit 4T-TMS thin films. For the solution-shearing method depicted in Figure 2.6
(experimental setup placed in a fume hood), briefly, a few drops of semiconductor
effects [45], and the change of metal work function due to contamination at the
interface of Au and 4T-TMS, we estimate a Schottky barrier height of ~0.2 eV which is
adequate for charge injection from the source to the channel.
(a)
(b)
E
-5.10eV
Vacuum Level
φm φm
4T-TMS
-5.10eV
LUMO: -2.58eV
HOMO: -5.27eV
Au Au
E χ
Ev=-5.17eV
EfEi
Ec=-4.05eV Eg=
8~9eV
0.95eV
SiO2
Vacuum Level
4T-TMS
LUMO: -2.58eV
HOMO: -5.27eV
PTSSi
Figure 2.7 Simplified energy band diagrams without considering interface dipole effect for the top-contact 4T-TMS SPOFET on silicon substrate along (a) source-to-drain channel direction and (b) gate-to-film direction. The HOMO/LUMO levels are based on the measurement as described in Section 2.1.1.
Figure 2.8 Cross-polarized optical micrographs of the 4T-TMS thin films deposited by (a)-(b) spin-coating method, from 5 mg/mL solution in chlorobenzene at (a) room temperature or (b) elevated temperature of ~80ºC; (c)-(d) drop casting method (c) with or (d) without solvent annealing; (e)-(f) solution shearing method with (e) 4T-TMS solution concentration: 6 mg/mL in xylene, deposition temperature: 112 ºC, shearing speed: 0.10 mm/s and (f) 4T-TMS solution concentration: 6 mg/mL in chlorobenzene, deposition temperature: 78.6 ºC, shearing speed: 0.10 mm/s. (g) AFM tapping-mode height image of a local film shown in (e).
Figure 2.9 Compiled statistical analysis of 162 SPOFET samples in “BL-1”, which shows the frequency distribution of the average mobility (μsat) with respect to different (a) solvents; (b) solution concentrations; (c) shearing speeds; and (d) deposition temperatures as listed in Table I. Spline interpolation algorithm is employed to generate the plots to facilitate comparison
Figure 2.10 Compiled statistical analysis of 162 SPOFET samples in “BL-1”, which shows the frequency distribution of the average Ion/Ioff for different (a) solvents; (b) solution concentrations; (c) shearing speeds; and (d) deposition temperatures as listed in Table I. Spline interpolation algorithm is employed to generate the plots to facilitate comparison.
Figure 2.11 (a) The frequency distribution of μsat for the three different solvents used in “BL-2”. Spline interpolation algorithm is employed to generate the curves to facilitate comparison. (b)-(d) Three-dimensional surface and contour plots of the average mobility (μsat) as a function of different shearing speeds, solution concentrations, and deposition temperatures. Here we applied Renka-Cline interpolation gridding algorithm and experimental data points of 172 SPOFETs in “BL-2” for the plotting.
Figure 2.12 (a)-(b) Solution concentration effect, (c)-(d) shearing speed effect; and (e)-(f) deposition temperature effect in the solution shearing process, as revealed by the bright-field or cross-polarized microscopy images.
Figure 2.14 Electrical characteristics of solution-shearing processed 4T-TMS SPOFETs with two different channel lengths (LA=35 μm and LB=16 μm), both of which were prepared on the same wafer under the following conditions: 4T-TMS concentration C=8 mg/mL in xylene, deposition temperature T=84 ºC, shearing speed R=0.10 mm/s. (a) Device A shows effective mobility μsat ~0.2 cm2/Vs, Ion/Ioff ~106. (b) Device B shows μsat ~0.3 cm2/Vs, Ion/Ioff ~7×105; the relatively poor subthreshold slope may be caused by contamination from the short-channel shadow mask during the gold deposition.
To understand our SPOFET performance uniformity under a specific processing
condition, we plot in Figure 2.15 the mobility and Ion/Ioff distribution of 18 measured
samples on the same wafer as devices A and B. These devices, with channel lengths of
13-50 μm and channel widths of 960-1000 μm, show an overall μsat=0.20±0.06 cm2/Vs
and an average Ion/Ioff>106.
0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.400
15
30
45
60 Gaussian Fit
Freq
uenc
y (%
)
Effective Mobility µsat (cm2/Vs)
0
20
40
60
80 Gaussian Fit
1010103
Freq
uenc
y (%
)
Ion/Ioff102 105104 109108107106
Figure 2.15 The effective saturation mobility (μsat) and Ion/Ioff distribution of 18 measured samples fabricated on the same wafer as for devices A and B shown in Figure 2.14. The red line is based on the Gaussian fitting model.
start to degrade shortly after exposure to air [51], these 4T-TMS SPOFETs exhibit
favorable environmental stability.
(a)
0 20 40 60 80 100 120 140 160 180 2000
20
40
60
80
100
120
Cur
rent
On/
Off
Rat
io
Time (Day)
Nor
mal
ized
Mob
ility
(%)
Normalized Mobility, Device-1 Normalized Mobility, Device-2 Current On/Off Ratio, Device-1 Current On/Off Ratio, Device-2
103
106
109
(b)
0 5 10 15 20 25 30 35 400
20
40
60
80
100
120
Normalized Mobility Current On/Off
Time (Day)
Nor
mal
ized
Mob
ility
(%)
103
106
109
Cur
rent
On/
Off
Rat
io
Ambient Air & Occasional Light Continuous Exposure to Fluorescent Light (~40 W)
Figure 2.16 (a) Stability measurement of solution-shearing processed 4T-TMS SPOFETs stored in ambient air with occasional exposure to room light. (b) Stability measurement of five 4T-TMS SPOFETs fabricated on the same substrate, stored in ambient air with occasional exposure to room light for 23 days, and later continuously exposed to a fluorescent light (~40 Watts). For both plots, the average mobility measured at the first time is normalized to 100%; the zero time corresponds to the date of device fabrication.
estimated. This result suggests that solution-sheared 4T-TMS films have highly
favorable molecular orientations for charge carrier transport along the channel [52-53].
0 5 10 15 20 25 300.0
5.0x104
1.0x105
1.5x105
2.0x105
2.5x105
3.0x105C
ount
s/s
2θ (degrees)
(001)
(002)(003) (006)
Au E
lect
rode
Arr
ay Shea
ring
Figure 2.17 Out-of-plane X-ray diffraction (XRD) spectra of solution-sheared 4T-TMS thin film. The inset is a cross-polarized optical microscopy image of the highly crystalline film on plastic PET substrate.
The electrical characteristics of our flexible 4T-TMS SPOFETs were measured and
analyzed in the same way as for the devices on silicon substrate [see Section 2.1.4],
except that here Ci≈22 nF/cm2 and W/L=1000 μm/50 μm. The flexible SPOFETs are
found showing excellent transistor behavior at a bias down to -1.5 V [see Figure 2.18 (a)
and (b) for the transfer and output characteristics, respectively] with a near zero turn-on
voltage and a low VT of 0.5-0.7 V. Quantitative analysis of the turn-on/threshold
voltage is beyond the scope of this work; instead, we propose a simplified energy band
diagram shown in Figure 2.18 (c) which suggests the ideal turn-on/threshold voltage is
attributed to an excellent alignment between the work function of PEDOT:PSS, the
HOMO and Fermi level of 4T-TMS, and the dipole moment of PBTS SAM at the
semiconductor-dielectric interface. The PEDOT:PSS film not only acts as a buffer layer
avoiding detrimental indium diffusion from ITO to the polymeric dielectrics in our
SPOFETs, but also reduces the threshold voltage of the transistors by ∆Vt=0.4-0.5 V
per its higher work function (5.2-5.3 eV) [54] than that of ITO (4.8 eV).
0.5 0.0 -0.5 -1.0 -1.5-13
-12
-11
-10
-9
-8
-7
sqrt
(I ds) (
A0.5)
Gate Voltage Vgs (V)
Log (
I ds) (
A) S ~80mV/dec
0.0000
0.0001
0.0002
0.0 -0.5 -1.0 -1.5
Vgs
-0.6 V-0.9 VDr
ain C
urre
nt I ds
(nA)
Drain Voltage Vds (V)
-1.2 V
0
-20
-40
0 2 4 6 8 10 120
200
400
600
Operating Voltage (V)
Subth
resh
old S
lope (
mV/de
c)
1.0E-9
1.0E-8
1.0E-7
1.0E-6
1.0E-5
Ion (A)
101102103104105106
on/of
f rati
o
Figure 2.18 (a) Transfer [Vds=-1.5 V] and (b) output characteristics of the flexible 4T-TMS SPOFETs. (c) A simplified energy band diagram clarifying the device physics responsible for the ideal turn-on voltage. (d) Device performance with respect to different operating voltages.
based on the PBTS-modified PVP-EAD dielectric. Additionally, they also suggest a very
low trap density in the solution-sheared 4T-TMS films.
0.1 1 10
10-2
10-1
100
H. Sirringhaus, et al., Science (2000) J. Rogers, et al., IEEE EDL (2000) J. Rogers, et al., PNAS (2001) G. Gelinck, et al., Nat. Mater. (2004) R. A. Street, et al., Mater. Today (2006) D. Gundlach, et al., Nat. Mater. (2008) K. Myny, et al., ISSCC (2009) M. Bohm, et al., ISSCC (2006) S. Lee, et al., Org. Electron. (2008) H. Yan, et al., Nature (2009) J. H. Cho, et al., Nat. Mater. (2008) H. Klauk, et al., Nature (2007) This work
Mob
ility
(cm
2 /Vs)
(Subthreshold Slope)-1 (dec/V)
Figure 2.19 A review on the representative flexible SPOFET performances (except [54] which is based on vapor process) reported in literature in the past decade. Our work contributes to one of the highest performances, with a record-low subthreshold slope of ~80 mV/dec.
For higher driving current applications, we further investigated the electrical
characteristics at different operating voltages (i.e. initial Vds for Ids-Vgs test). It is not
surprising that the subthreshold slope increases slightly and the on/off ratio degrades
with the operating voltage as shown in Figure 2.18, primarily due to the use of a
common gate structure [see Figure 2.2 (a)]. A relatively large overlap capacitor between
the drain/source and gate contributes leakage to the off current which increases
exponentially with the applied electric field, while the on current changes approximately
according to a square law. We anticipate an improved operating-bias-dependence of the
subthreshold characteristics by patterning the gate and further optimization of the
It doesn't matter how beautiful your theory is, it doesn't matter how smart you are. If it
doesn't agree with experiment, it's wrong.
- Richard Feynman
As briefly mentioned in Chapter 2, organic transistor performance is closely associated
with the surface condition of the dielectrics, or, equivalently, the interface property
between the semiconductor and the dielectrics. This chapter focuses on studying the
interface effects and relevant interface engineering by self-assembled monolayers (SAM).
3.1 Background and Motivation
Looking back to history, even since the first transistor invented by John Bardeen, Walter
Brattain, and William Shockley at Bell Labs in 1947, people have been realizing the
crucial role of the surface states in determining the field-effect transistor behavior [58].
46 CHAPTER 3. INTERFACE ENGINEERING BY SELF-ASSEMBLED MONOLAYER
Numerous scientists and engineers have practically spent their whole life in eliminating
the defects to improve the semiconductor-dielectric interface for traditional MOSFET.
Indeed, this interface is equally important for organic thin-film field-effect transistor,
simply based on the fact that most of the charge carriers induced by the gate voltage in
the field-effect transistor, regardless of inorganic or organic, are accumulating in a very
thin channel close to the semiconductor-dielectric interface, as illustrated in Figure 3.1.
Any defects, trap states, or imperfections are anticipated to influence charge transport
along the interface.
Figure 3.1 Schematic illustration of the gate voltage induced charge carrier layer, which is close to the semiconductor-dielectric interface.
A widely adopted measure to modify the dielectric surface for organic transistors is
based on self-assembled monolayers (SAM) as first proposed for vapor processed
pentacene transistor [41, 59]. The basic concept is illustrated in Figure 3.2: organosilane
molecules with a silane headgroup react with the hydroxyl-terminated dielectric surface,
spontaneously assemble and polymerize in coverage of the entire dielectric surface to
remove originally unfavorable trapping sites or defects. It has been shown that [15, 36,
41-42, 59-65], by adequate employment of SAMs with desired molecular structure, one
is able to modify the interface states distribution, charge carrier density, surface energy,
and film growth mode and crystallinity, all of which may contribute to improved device
performance and reliability.
As earlier representative attempts to understanding the SAM effects on the organic
transistor, Salleo et al. [65] investigated solution-processed poly-9,9’ dioctyl-fluorene-co-
bithiophene (F8T2) thin-film transistors with the dielectric surface modified by different
CHAPTER 3. INTERFACE ENGINEERING BY SELF-ASSEMBLED MONOLAYER 47
SAMs, and suggested that SAM induced mobility enhancement may be involved with
molecular interactions between the polymer and the SAM. This work also pointed out
the possibility of using dipolar SAM molecule to control the threshold voltage for the
transistor. Later, Kobayashi et al. [64] and Pernstich et al. [63] did a more systematic
study on the SAM-dipole effects on the charge carrier density and the threshold voltage.
These works provide pioneering insights into general physical and chemical effects
collectively arising from the SAM molecules. However, the SAMs being used in each
particular study possess different functional groups which are in direct interaction with
the organic semiconductor. Depending on the chemical properties, it is not impossible
for the functional groups to distort the apparently observed dipole moment relevant
physical effects. Furthermore, the SAM quality as deposited even from the same process
can vary considerably for the SAMs with different functional groups. This issue used to
be ignored and raises further complications in interpreting the experimental
observations on the organic transistor’s electrical characteristics. Other works [60-62]
have also explored the SAM effects based on the molecules with the same functional
end group, typically –CH3, but with different intra alkyl-chain lengths. These studies
exclude the chemistry effects of the functional groups from the physical effects,
however, no dipole relevant physical effects were separately investigated. It is thus still
unclear how the specific functional group’s chemistry effect and the dipole relevant
physical effect may each contribute to the interface engineering for organic transistors.
SiO
OO
Si
OO
Si
OO
Si
OO
CFF
F
Figure 3.2 Schematic illustration of the concept of dielectric surface modification by self-assembled monolayers.
48 CHAPTER 3. INTERFACE ENGINEERING BY SELF-ASSEMBLED MONOLAYER
In this work, through careful selection of a group of phenyl-terminated SAMs, we
manage to decouple the dipole moment relevant physical effect from the surface
chemistry effect of the functional groups, to elucidate how the performance and
reliability of organic transistors are controlled by interfacial SAMs [66-67].
The chemical structures of the phenyl-SAM molecules used for this study are shown
in Figure 3.3. The major distinctions of this study are highlighted below:
PTS PETS PBTS PHTS PAPTS
Si O
OO
NH
Si ClCl
ClSi Cl
Cl
Cl Si
Cl
Cl Cl SiClCl
Cl
Figure 3.3 A group of phenyl-terminated SAM molecules used for the interface engineering study in this work. The intra alkyl chain length increases from PTS, PETS, PBTS, to PHTS, PAPTS is comparable to PBTS except that an –NH- group replaces a –CH2- in the intra part of PBTS. (PTS: phenyltrichlorosilane, PETS: phenethyltrichlorosilane, PBTS: 4-phenylbutyltrichlorosilane, PHTS: 6-phenylhexyltrichlorosilane, PAPTS: N-phenylaminopropyltricholorosilane)
• As introduced in Chapter 2 [see Figure 2.5], phenyl-terminated SAMs are
particularly favorable for SPOFETs and can be superior to other SAMs including
the methyl-terminated OTS which is widely used for vapor processed organic
transistors. Therefore, comprehensive understanding of the phenyl-terminated
SAM effects will enable us to further improve the SPOFET performance;
CHAPTER 3. INTERFACE ENGINEERING BY SELF-ASSEMBLED MONOLAYER 49
• With the same functional aromatic end group and the consistent intra alkyl chain
from PTS to PHTS, we are able to exclude the interference of the SAM functional
group’s chemistry effects to the SAM dipole physical effects. This differs from
previous studies and will enable us to decouple these two effects for individual
identification;
• With the only difference of an intra-chain atom between PBTS and PAPTS, we
are able to testify the chemical effects per specific SAM molecular structures and
provide direct evidence of SAM chemistry effects on the SAM quality and the
relevant transistor performance and reliability.
• Ambient atmosphere including H2O and O2 has been shown to have strong
influences on the transistor performance [68], bringing additional artifact sources
in the observed SAM effects. Therefore, we pay particular attention in the device
fabrication and measurement, all in N2 atmosphere or vacuum without exposure
to air during the processing, thus excluding any possible intervence of the ambient
environment in analyzing the SAM effects.
• Consistent quality of the phenyl-SAMs being investigated in this work is ensured
by a dedicated new deposition process, thus eliminating the artificial effects
resulted from different interfacial SAM quality.
3.2 Phenyl-SAM Processing and Characterization
3.2.1 Spin-coating Process and Mechanism
In general, organosilanes can be either vapor deposited or solution deposited onto the
substrate surface to render the SAM structure. Vapor process regularly yields a smoother
[6] but low-density and amorphous SAM layer [69]. For the latter case, a typical protocol
is immersing the substrate in a solution of organosilanes and solvents for at least a few
hours. This method, though being easier in processing, is prone to rendering a rough
50 CHAPTER 3. INTERFACE ENGINEERING BY SELF-ASSEMBLED MONOLAYER
multilayer structure due to the high susceptibility of silanes to the presence of water [6].
In reality, the silane soaking process also suffers from contamination most likely from
the backside of the wafer. Additionally, for flexible electronics applications, immersing
the flexible substrate in the silane solution faces more challenges.
Recently, Nie et al. proposed a simple spin coating process using nonpolar solvents
such as trichloroethylene (TCE) and chloroform with a dielectric constant of ~4 to
deliver full-coverage octadecylphosphonic acid (OPA) SAM on oxide surfaces [70].
Later, Ito et al. extended this approach to deposit crystalline ultrasmooth alkylsilane
SAM such as OTS on silicon oxide surface, giving rise to very high mobility for vapor
processed pentacene and C60 transistors [71]. The same mechanism involved in both
processes is that a raft of hydrophilic headgroups of OPA and OTS in the nonpolar
solvent like TCE spontaneously aggregate and align on the liquid drop surface,
providing an ideal environment to fast react with the hydroxyl-terminated oxide surface
during the spin coating process [70]. However, both OPA and OTS are methyl-
terminated long alkyl chain molecules and their based SAMs are more favorable in vapor
processed organic transistors instead of in SPOFETs as discussed in the background
section. In this work, we extend the method by Nie et al. [70] and Ito et al. [71] to
develop a simple spin coating process for the deposition of high-quality phenyl-
terminated SAMs which play particularly important roles in SPOFETs [72].
The basic process and its relevant mechanism are illustrated in Figure 3.4. We
emphasize that the conditions including the solvent, concentration and spin coating
rate/time are critical to the success of the SAM deposition, as revealed by the
preliminary film screening from microscopy images shown in Figure 3.5. Of particular
importance is the application of anhydrous toluene as the solvent for the phenyl-
terminated silanes here. Although TCE, chloroform and toluene are all nonpolar
solvents, the former two proven to be ideal for spin-coating of OPA and OTS SAMs
[70-71] are found being inappropriate for phenyl-terminated silanes. This observation
indicates the importance of the aromatic group of toluene in interacting with the
functional endgroup of the phenyl-silanes and thus further facilitating the
orientation/alignment of the trichlorosilane headgroup on the liquid surface.
CHAPTER 3. INTERFACE ENGINEERING BY SELF-ASSEMBLED MONOLAYER 51
Figure 3.4 Schematic illustration of the spin-coating process for depositing phenyl-terminated SAMs from the solution of phenyl-trichlorosilane in anhydrous toluene. The trichlorosilane headgroup spontaneously aggregate and align on the liquid drop surface, providing an ideal environment for the fast reaction of the trichlorosilane with the hydroxyl group on the substrate surface [70-71]. The anhydrous toluene solvent being selected here is critical to the process. The conjugated aromatic group of toluene is supposed to interact with the functional endgroup of the phenyl-silane and facilitate the alignment of the phenyl-silane on the liquid drop surface.
Figure 3.5 A preliminary screening of different film morphology as deposited by different spin-coating processes for PBTS. Images were taken under bright-field microscope. The solvent, concentration and spin rate/time are all critical to high quality phenyl-SAM.
52 CHAPTER 3. INTERFACE ENGINEERING BY SELF-ASSEMBLED MONOLAYER
The aligned tricholorsilane headgroup on the liquid surface reacts fast with the
hydroxyl group on the substrate surface and the excessive solution is removed during
the spin process. Subsequent post-processing by treating the spin-coated substrates in
hydrochloric acid vapor for overnight is necessary to allow further hydrolysis and
siloxane polymerization on the surface. The optimized process details are given in the
Appendix section in this dissertation.
3.2.2 Spin-coated Phenyl-SAM Characterization
The phenyl-terminated SAMs as deposited by the same optimized spin-coating process
are characterized by techniques including water contact angle (CA), atomic force
microscopy (AFM), ellipsometry, and grazing incidence X-ray diffraction (GIXD).
Figure 3.6 shows the measured water contact angle [Edmund Scientific goniometer] on
each phenyl-SAM modified SiO2 surface. For PTS, PETS, PBTS, and PHTS with similar
molecular structure, the water contact angle increases monotonically with the intra-alkyl-
chain length from ~70º to ~90º, indicating relatively low surface energy and high density
of the SAMs as deposited by the spin coating process. For PAPTS which contains a -
NH- in the intra-chain, a lower contact angle is observed over the comparable PBTS
where only -CH2- appears in the chain. This effect may arise from the SAM morphology
difference as induced by the molecular structures during the spin coating process, and
the stronger dipolar interaction between the PAPTS SAM and the polar water molecule.
As shown in Figure 3.7, AFM measurement [Veeco Nanoscope] reveals that the
phenyl-terminated SAMs as deposited by the optimized spin-coating process are
generally showing remarkably high degree of smoothness, with typical roughness of 0.1-
0.2 nm (rms) for PTS/PETS/PBTS/PHTS and 0.2-0.3 nm (rms) for PAPTS. No
appreciable difference among the morphology of PTS/PETS/PBTS/PHTS SAMs is
observed. However, a relatively rougher film surface is found for PAPTS, indicating the
SAM molecular structure with specific chemical functional groups in affecting the SAM
quality even based on the same process. This underscores the importance of selecting
appropriate SAM structures in studying the SAM effects for organic electronic devices.
CHAPTER 3. INTERFACE ENGINEERING BY SELF-ASSEMBLED MONOLAYER 53
PTS PETS PBTS PHTS PAPTS50
60
70
80
90
100
Mol
ecul
ar L
engt
h, b
y M
OPA
C9
(Å)
Con
tact
Ang
le (0 )
Si
Cl
Cl Cl Si
Cl
ClCl
SiO
O
O
NH
SiCl
Cl
Cl
SiCl
Cl
Cl
Exception for N-included chain
0
2
4
6
8
10
12
14
16
Figure 3.6 Measured water contact angle on phenyl-SAM modified SiO2 surfaces. The molecular length of each SAM molecule is computed by PM3/MOPAC9 model and shown for comparison purpose here.
Si
Cl
Cl Cl
Si O
OO
NH
Figure 3.7 Representative phenyl-SAM film morphology for (a) PTS/PETS/PBTS/PHTS and (b) PAPTS, as deposited by the optimized spin-coating process. For PTS-series SAM, the typical roughness is 0.1-0.2 nm (rms); for PAPTS SAM, the roughness is 0.2-0.3 nm (rms). This indicates the spin coating process yields excellent phenyl-SAM quality.
54 CHAPTER 3. INTERFACE ENGINEERING BY SELF-ASSEMBLED MONOLAYER
Long alkylsilanes such as OTS (~2 nm) have been reported to form a well packed,
crystalline SAM on the amorphous SiO2 surface under appropriate processing
conditions [71]. However, this phenomenon is not observed for the relatively shorter
phenyl-silane based SAMs as investigated in this work. GIXD experiments performed at
Stanford Synchrotron Radiation Lightsource (SSRL) on these phenyl-SAM films
deposited on native SiO2 do not show crystalline features in the diffraction pattern [see
Figure 3.8], indicating the phenyl-SAMs are amorphous or disordered, probably due to
the weaker van der Waals interactions between neighboring molecules for the short
phenyl-SAM molecules (<1.4 nm).
SiCl
Cl
Cl
Figure 3.8 GIXD image of the PHTS SAM on native silicon oxide surface as deposited by the spin coating process. By courtesy: Dr. Eric Verploegen.
The phenyl-SAM layer thickness is measured using ellipsometry [Sopra Bois-
Columbes] and shown in Figure 3.9. In addition, we compute the SAM molecule length
with PM3/MOPAC9 model and the Gaussian Package 03’, B3LYP/6-31G(d) model.
Both simulations yield similar results on the molecular length. Therefore, it is possible to
estimate the average tilt angle for the SAM molecule on the surface by
cos /d Lθ = (3.1)
CHAPTER 3. INTERFACE ENGINEERING BY SELF-ASSEMBLED MONOLAYER 55
where θ is the tilt angle to the surface normal direction, d is the measured film thickness,
and L is the molecular length, for which the Gaussian simulation is used here. As shown
in Figure 3.9, the average tilt angle for PTS/PETS/PBTS/PHTS is ~40º-55º.
Interestingly, the PAPTS is found giving a considerably larger SAM thickness and
negligible tilt angle compared to the phenyl-SAMs without the nitrogen atom.
PTS PETS PAPTS PBTS PHTS --0
5
10
15
20
25
30
35
40
-100
-75
-50
-25
0
25
50
75
100
Mol
ecul
ar L
engt
h or
SA
M T
hick
ness
( Å) Measured SAM Thickness, d
Simulated Molecule Length, L (PM3/MOPAC9) Simulated Molecule Length, L (Gaussian)
Tilt
Angl
e, θ
(0 )
cos (θ)= d/L
Figure 3.9 Measured phenyl-SAM thickness (d) using ellipsometry and the simulated molecule length (L) based on PM3/MOPAC9 model or Gaussian Package 03’, B3LYP/6-31G(d) model. The average title angle of the SAM molecule with respect to surface normal direction, θ, is estimated by cos (θ)=d/L, where L is based on the Gaussian simulation here.
3.3 Phenyl-SAM Effects in Organic Transistors
3.3.1 PBTTT Transistor Structure and Device Fabrication
To systematically investigate the SAM effects on the organic transistor performance and
reliability, we fabricated poly(2,5-bis(3-alkylthiophen-2-yl)thieno[3,2-b]thiophene)
56 CHAPTER 3. INTERFACE ENGINEERING BY SELF-ASSEMBLED MONOLAYER
(PBTTT) transistors with the structure shown in Figure 3.10, which incorporate
different phenyl-SAMs introduced above at the dielectrics-semiconductor interface.
PBTTT is a high-mobility, solution-processable p-type material, exhibiting liquid-
crystalline phase under thermal annealing conditions [73]. PBTTT film is spin-coated
onto the phenyl-SAM modified 300 nm SiO2 following the procedures reported before
[73], rendering relatively high uniformity for its transistor performance. All of the device
fabrication and measurement were performed in N2 or vacuum environment without
exposure to ambient O2 and H2O, thus excluding the environment induced artifacts [68]
in the observed SAM effects.
Thermally grown SiO2
N++ Si
4T-TMS Active LayerD S
PBTTTD S
PTS PETS PBTS PHTS PAPTS
Si O
OO
NH
Si ClCl
ClSi Cl
Cl
Cl Si
Cl
Cl Cl SiClCl
Cl
PBTTT
Figure 3.10 Device structure of PBTTT transistor incorporating different phenyl-SAMs at the dielectric-semiconductor interface. Courtesy: PBTTT was provided by Dr. Iain McCulloch and Dr. Martin Heeney (formerly in Merck).
3.3.2 Device Electrical Characteristics
In general, PBTTT transistors based on the spin-coated phenyl-SAMs exhibit excellent
transistor behavior with satisfactory transfer and output characteristics [see Figure 3.12
CHAPTER 3. INTERFACE ENGINEERING BY SELF-ASSEMBLED MONOLAYER 57
for an example]. Relatively high mobility (>0.4 cm2/Vs) and on/off ratio (>105) are
obtained for PBTS and PHTS based devices.
20 0 -20 -40 -60 -8010-9
10-8
10-7
10-6
10-5
10-4
Gate Voltage VGS (V)
Dra
in C
urre
nt I D
S (A
)
0.000
0.003
0.006
0.009
0.012
0.015
I DS0.
5 (A0.
5 )
VDS=-80V
0 -20 -40 -60 -800
-20
-40
-60
-80
-100
-120
-140
D
rain
Cur
rent
I DS
(μA
)
Drain Voltage VDS (V)
VGS=-80V
VGS=-60V
VGS=-40V
VGS=-20V
Figure 3.11 Representative (a) transfer and (b) output curves for the PBTTT transistors based on the spin-coated phenyl-SAMs. The results shown here are specifically based on PBTS.
Figure 3.12 shows the overall average mobility and threshold voltage for the PBTTT
transistors incorporated with different phenyl-SAMs or without any SAM at the
dielectric-semiconductor interface. It is found that,
• For PTS/PETS/PBTS/PHTS with alkyl intra chain only, they all improve the
transistor performance compared to that without SAM modification on the
dielectric surface. In this series of phenyl-SAMs, longer intra chain gives rise to
higher mobility, in consistence with the water contact angle and surface energy
measurement results as shown in Figure 3.6;
• PAPTS, despite having comparable molecular structure to the other phenyl-SAMs
where only a –CH2- is replaced by –NH-, yields significantly lower mobility. This
phenomenon clearly indicates the chemical effects resulted from the specific SAM
structure, and reiterates the field-effect mobility is not simply determined by the
surface energy, otherwise the mobility for PAPTS should be similar to that for
PTS as judged from Figure 3.6;
58 CHAPTER 3. INTERFACE ENGINEERING BY SELF-ASSEMBLED MONOLAYER
• The threshold voltages of the PBTTT transistors with phenyl-SAMs are in
opposite region to that without SAM, and they are apparently dependent on the
particular SAM structure.
PAPTS SiO2 PTS PETS PBTS PHTS0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
Mob
ility
(cm
2 /Vs)
no-SAM
PAPTS SiO2 PTS PETS PBTS PHTS-15
-10
-5
0
5
10
15
20
Thre
shol
d Vo
ltage
(V)
no SAM
20 0 -20 -40 -60 -8010-9
10-8
10-7
10-6
10-5
10-4
VGS (V)
Dra
in C
urre
nt I D
S (A
)
0.000
0.003
0.006
0.009
0.012
0.015
I DS0.
5 (A
0.5 )
Vds=-80V
Figure 3.12 (a) Mobility and (b) threshold voltage of PBTTT transistors incorporated with different phenyl-terminated SAMs at the dielectric-semiconductor interface.
CHAPTER 3. INTERFACE ENGINEERING BY SELF-ASSEMBLED MONOLAYER 59
It should be emphasized that hysteresis measurement provides ample information
on the interface property. The general mechanisms and the corresponding effect on the
hysteresis of the transfer curve are illustrated in Figure 3.13.
VGS (V)
Dra
in C
urre
nt I D
S (A
)
0(1)
VGS (V)
Dra
in C
urre
nt I D
S (A
)
0(2)(3)(4)
Figure 3.13 General mechanisms responsible for the hysteresis effect in the organic transistor. Only (1) the charge injection from semiconductor to the interface induces counterclockwise hysteresis in the transfer curve. The other mechanisms including (2) charge injection from the gate electrode, (3) residual dipole caused by slow polarization in the bulk dielectric, and (4) mobile ions in the dielectric contribute to the counterclockwise hysteresis in the transfer curve.
The measurement results on the hysteresis of our devices with different phenyl-
SAMs are shown in Figure 3.14. For all the devices, we only observed counterclockwise
hysteresis depicted in Figure 3.13 (note: the reversed VGS axis in Figure 3.14), which
indicates that charge injection from the semiconductor to the interface dominates in
these devices. PTS/PETS/PBTS/PHTS are found rendering similar negligible hysteresis
effect in the transfer curve for PBTTT transistors. Moderately large hysteresis can be
seen in devices without SAM modification. However, substantially larger hysteresis is
60 CHAPTER 3. INTERFACE ENGINEERING BY SELF-ASSEMBLED MONOLAYER
found for the PAPTS modified PBTTT transistors. We thus argue that the surface
modification by spin-coated PTS/PETS/PBTS/PHTS yields improved, high-quality
interface conditions, while spin-coated PAPTS SAM tends to give rise to more traps and
defects at the interface. This argument is further supported by bias stress measurement
as shown in Figure 3.15. Considerably larger bias stress effect (threshold voltage shift) is
observed for PAPTS modified PBTTT transistors, suggesting more traps and defects
exist at their dielectric surface, while no appreciable difference on the bias stress effects
for the PTS/PETS/PBTS/PHTS is observed. Therefore, the spin-coated PTS series
SAMs are believed to exhibit similar interface quality.
20 0 -20 -40 -60 -8010-10
10-9
10-8
10-7
10-6
10-5
10-4
VGS (V)
Dra
in C
urre
nt I D
S (A
)
SiCl
Cl
Cl
SAM: PBTS
Vds=-80V
20 0 -20 -40 -60 -8010-11
10-10
10-9
10-8
10-7
10-6
10-5
SiO
O
O
NH
VGS (V)
Dra
in C
urre
nt I D
S (A
)
SAM: PAPTS
Vds=-80V
60 40 20 0 -20 -40 -60 -8010-10
10-9
10-8
10-7
10-6
10-5
10-4
VGS (V)
Dra
in C
urre
nt I D
S (A
)
No SAM
Vds=-80V
Figure 3.14 General hysteresis effect for PBTTT transistors modified with different phenyl-SAMs. PTS/PETS/PBTS/PHTS show no appreciable difference in the hysteresis, therefore only a representative I-V hysteresis curve is given here as adopted from PBTS based transistors.
CHAPTER 3. INTERFACE ENGINEERING BY SELF-ASSEMBLED MONOLAYER 61
1 10 100 1000 100000.01
0.1
1
10
100
Thre
shol
d Vo
ltage
Shi
ft, -d
Vth
(V)
Stress Tiime (s)
PTS PETS PBTS SiO2 PAPTS PHTS
VGS=-60 VTrecovery=200 s
Si O
OO
NH
Figure 3.15 Bias stress effect on the PBTTT transistors with different phenyl-SAMs. Considerably larger bias stress effect is observed for the PAPTS modified interface. The applied gate bias for the bias stress measurement is VGS=-60V. For each I-V measurement during the bias stress test, a recovery time of 200 seconds is given to allow the fast reversible traps recovered.
3.3.3 Threshold Voltage Control by SAM Dipole Moment
The threshold voltage of organic transistors is apparently related to the specific SAM
molecular structure. To understand how the threshold voltage is controlled by the SAM,
we show in Figure 3.16 the energy band diagram of the PBTTT transistors with SAM
modification layer at the dielectric-semiconductor interface. The dipolar SAM creates an
additional electric field at the interface, changes the accumulation/depletion regime
when the gate voltage is zero, or equivalently, bends the energy alignment, and thus
tunes the threshold voltage. A simple model to resolve this effect is shown in Figure
3.17. The induced surface charge density (QS) by the addition of the SAM dipole at the
dielectric interface is described as
62 CHAPTER 3. INTERFACE ENGINEERING BY SELF-ASSEMBLED MONOLAYER
Figure 3.16 Simplified energy band diagram of the PBTTT transistor incorporating the SAM modification layer at the dielectric-semiconductor interface. The SAM dipole creates an additional electric field normal to the interface, bending the energy alignment and thus tuning the threshold voltage.
Figure 3.17 A simple model to describe the dipole moment effect in tuning the surface charge density and transistor’s threshold voltage. The SAM dipole (p) on the dielectric surface induces additional surface charge (QS). Θ is the tilt angle of the dipole moment.
CHAPTER 3. INTERFACE ENGINEERING BY SELF-ASSEMBLED MONOLAYER 63
cos SAMs
SAM
p NQL
⋅ Θ⋅= (3.2)
where p is the SAM dipole moment, Θ is the tile angle of the dipole relative to the
surface normal, NSAM is the SAM density, and LSAM is the SAM molecular length. The
corresponding shift of the threshold voltage (ΔVt) is found to be
0
Tox
ox
ox
ox
xdxe Qs
TVtC
ρ+
Δ =−
∫i
(3.3)
where ρox is the mobile charge density in the dielectric, Tox is the dielectric thickness, Cox
is the dielectric capacitance per area, e is the elementary charge.
We emphasize that the threshold voltage can be only compared in this model for the
devices with similar interface SAM quality, otherwise the distinct trapping effects would
affect the observed threshold voltage and make the comparison unfair. Therefore, based
on our observations and discussions on the PAPTS of which the chemical effects
significantly influence its SAM property and distinguish it from the other alkyl-chain
only phenyl-SAMs, we exclude this molecule in the following analysis.
The dipole moment for each molecule (p0), SAM molecule density (NSAM), and SAM
molecular length (LSAM) are simulated using Gaussian package and PM3/MOPAC9
model as listed in Table 3.1, and the tilt angle Θ is estimated to be 30º-60º based on the
SAM characterization shown in Figure 3.9 as well as the dipole simulation. One should
note that the dipole moment in the monolayer form (p) is attenuated over the isolated
molecule due to Coulomb interaction of parallel dipoles [74]. Therefore, an attenuation
factor of 5 is assumed here [74], i.e. p=p0/5.
64 CHAPTER 3. INTERFACE ENGINEERING BY SELF-ASSEMBLED MONOLAYER
Table 3.1 SAM molecule dipole, molecule length, and SAM density as simulated using Gaussian package and PM3/MOPAC9 model.
SAM
molecule
Molecule dipole
(p0) [debye]
Molecule length
(L) [Å]
SAM density (NSAM)
[Å2/molecule]
PTS 1.31 7.01 19.46
PETS 1.67 8.93 25.32
PBTS 1.69 11.41 21.80
PHTS 1.67 14.48 26.68
Figure 3.18 shows the measurement results and the simulation results for the
threshold voltage shift with respect to the SAM dipole and SAM induced surface charge
density. In general, experimental results on both p-type PBTTT transistors and n-type
amorphous phenyl-C61-butyric acid methyl ester (PCBM) transistors show that the
threshold voltage shift is linear to the SAM dipole-induced surface charge density, in
excellent agreement with the theoretical model shown in equation (3.3).
However, discrepancy between the experiment fitting line and the theoretical line for
the slope is still noticeable. We emphasize that an important mechanism has not been
considered yet to date, i.e. strong external interfacial electric field as induced by the gate
voltage would distort the inherent dipole moment of the SAM molecule as illustrated in
Figure 3.19, thus influence the model fitting in (3.3). This effect, namely electric
shielding tensors effect, must be given careful consideration in future studies. From the
experiment point of view, in-situ and real-time measurement of the SAM dipole on an
oxide surface under the applied external electric field would give direct evidence of the
electric shielding tensors effect. In the meantime, it is also possible to model this effect
using Ab-initio simulation in the future.
CHAPTER 3. INTERFACE ENGINEERING BY SELF-ASSEMBLED MONOLAYER 65
0 1x1012 2x1012 3x1012 4x101230
25
20
15
10
5
0
-5
-10
-15
0 1x1012 2x1012 3x1012 4x101230
25
20
15
10
5
0
-5
-10
-15
0 1x1012 2x1012 3x1012 4x101230
25
20
15
10
5
0
-5
-10
-15
PTSPETS
PBTSPHTS
Theoretical Line (Slope=-e/Cox)
Experimental Fitting Line
PTS
PETSPBTS
Experimental Fitting Line
Thre
shol
d Vo
ltage
(V)
Surface Charge Density (cm-2)
Theoretical Line (Slope=-e/Cox)
PHTS
PBTTT
PCBM
Figure 3.18 Measurement results vs modeling results for the threshold voltage shift with respect to the surface SAM dipole and SAM induced surface charge density. Both p-type PBTTT transistors and n-type PCBM transistors are measured in this work.
66 CHAPTER 3. INTERFACE ENGINEERING BY SELF-ASSEMBLED MONOLAYER
Figure 3.19 Schematic illustration of the proposed electric shielding tensors effect of the SAM at the dielectric-semiconductor interface: strong external interfacial electric field as induced by the gate voltage would distort the inherent dipole moment of the SAM molecule, thus influencing the performance and reliability of the organic transistors.
3.4 Summary
We have systematically investigated dipole moment related physical effects and
chemistry effects of the SAM at the organic semiconductor-dielectric interface. Through
careful selection of a group of phenyl-terminated SAMs, we elucidate how the
performance and reliability of organic transistors are controlled by interfacial SAMs. In
addition, we briefly introduce a spin-coating process for depositing high-quality phenyl-
terminated SAMs for organic electronics applications.
67
Chapter 4
Device Physics and Universal
Modeling of Organic Transistor
In physics, your solution should convince a reasonable person. In math, you have to convince a
person who's trying to make trouble. Ultimately, in physics, you're hoping to convince Nature.
And I've found Nature to be pretty reasonable.
- Frank Wilczek
So far we have presented the experimental demonstration of high-performance flexible
organic thin-film transistors based on solution process and systematically investigated
the electrical performance and reliability controlled by the semiconductor-dielectric
interfacial monolayer. In this chapter, we’ll study the device physics of organic
transistors with focus on charge injection effects at the metal-organic interface and
charge transport properties in the organic semiconductor film. A universal device model
capturing many of the physical effects observed for organic transistors will be
introduced [75].
68 CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR
4.1 Background and Motivation
As an important figure of merit for transistors, the transition frequency, defined by
fT ∝ μFETV/2πL2, reflects circuit switching speed/bandwidth and strongly depends on
the extrinsic field-effect mobility (μFET)1 as well as the channel length (L). Therefore,
appropriate downscaling of the organic thin-film transistor, resembling that of
traditional silicon MOSFET which has been substantially benefited from the
sophisticated device scaling theories [1], is also crucial for the development of organic
electronic circuits and systems. Earlier works have explored organic transistors with
channel lengths from hundreds of micrometers down to sub-100 nm [76-84].
Experimental results show that the extrinsic field-effect mobility of these transistors is
channel-length variant, typically, with a lower value for sub-micrometer-channel devices.
As a consequence, the advantages originated from dimensional scaling to organic
circuits are severely compromised. Even for organic transistors with channel lengths
above micrometers, it is still not uncommon that the mobility decreases with shortening
of the channel length [81, 85-90]. Such channel-length dependence of the extrinsic
mobility has been suggested qualitatively to contact resistance effects; nevertheless,
analytical and quantitative studies remain in lack.
On the other hand, an inverse trend to the above mobility scaling behavior has been
reported for organic transistors more recently [15, 27, 43, 86, 91-92]. As briefly
mentioned in Chapter 2, for instance, we found an extrinsic field-effect mobility
decreasing with channel length for solution-processed, top-contact, polycrystalline
small-molecule organic TFTs [15, 43]. Hamadani et al. [91] and Verlaak et al. [92]
observed an higher effective mobility in the saturation regime for shorter channel
polythiophene-/pentacene-based TFTs and they attributed this to the electric field
dependence of charge transport. Gundlach et al. [27] and Smits et al. [86] found similar
mobility scaling behavior for contact-controlled polycrystalline acene-based TFTs and
1 In literature, it is also termed “apparent” or “effective” mobility, the value of which is extracted from the transfer curve in linear or saturation regime
CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR 69
respectively. However, the origins are interpreted differently. For Gundlach et al.’s
devices with contact-induced polycrystalline active layer, decreased disorder and reduced
density of grain boundaries in the shorter channel device is believed in lowering the
activation energy for charge transport and accordingly enhancing the intrinsic mobility
[27]. While for Smits et al.’s SAMFET work, they argued that the film coverage and two-
dimensional percolation are responsible for the mobility scaling characteristics [86].
The apparent paradox arising from a variety of literature reports demands that the
mobility scaling behavior of organic transistors should be resolved in a more complete
way, both analytically and quantitatively. Also, fundamental understanding of the elusive
mobility scaling behavior is essential for technological improvement of the device
performance and crucial for scaled organic circuits design.
In this work, we propose and develop a universal physical model for organic
transistors toward resolving the mobility scaling behavior, clearly elucidating their
underlying physical mechanisms by taking into account both contact effects at the
metal-organic interface and charge carrier transport properties in the active organic
semiconductor layer. We introduce an extended polaronic mobility-edge (EPME) model
which includes the electric field dependence of the charge carrier mobility, namely,
Poole-Frenkel (PF) like exponential dependence of the carrier mobility on the electric
field. Contact effects are described in a system of diffusion-limited charge injection into
a disordered organic film with Gaussian density of states (DOS) at the metal-organic
semiconductor interface. This universal model then combines both space charge limited
conduction (SCLC) and gate voltage induced charge accumulation to describe the
transistor operation. It successfully enables us to account in a single formula for most of
the significant physical phenomena observed so far for organic semiconductors. Further
analysis reveals that, depending on the physical origins including carrier injection barrier
height, localized trap states energy distribution, electric field dependence of carrier
mobility, and channel length dependence of film crystallinity, the mobility scaling curve
with respect to the channel length ranging from a few nanometers to hundreds of
micrometers can be monotonically upwards, or slowly get saturated, or, surprisingly,
exhibit an overshoot region.
70 CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR
We stress that, beyond resolving the mobility scaling issue, our device model is also
versatile in explaining many other elusive physical phenomena for organic thin-film
transistors, such as the contact resistance effect and the surface potential profiles along
the channel which have been experimentally probed but yet poorly understood. Since
there is no commercially available device simulator dedicated for organic transistors
(unlike well-developed industry standard device simulator, e.g. Sentaurus, for traditional
silicon MOSFET), we implement the universal model into a 1-D device simulator for
organic TFTs. The simulation results of various electrical characteristics are in excellent
agreement with experimental observations.
4.2 Universal Modeling of Organic Transistors
There are a variety of structures available for standard organic TFTs [6], which can be
broadly categorized as bottom-contact and top-contact devices shown in Figure 1.2. A
compact model to describe the electrical characteristics of organic TFTs seemingly relies
on their particular structure [93-94], yet, the underlying physical mechanisms dominating
the device operation are consistent, and, are possible to describe in a generalized 1-D
manner as depicted in Figure 4.1.
Figure 4.1 A generalized 1-D representation of the physical process dominating the device operation for organic transistors.
CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR 71
Charge carriers are first injected from the source electrode to the organic film
through the metal-organic semiconductor interface, then cross a transition region in the
vicinity of the source electrode before reaching the active channel layer. Eventually, the
charge carriers are extracted at the drain electrode through the metal-organic interface
immediately after traversing the other transition zone close to the drain. In bottom-
contact devices, the transition regions arise from enhanced structural disorder of the
organic semiconductor film deposited along the electrode edge as revealed by atomic
force microscopy (AFM) [95], and they tend to give rise to multifold effects, including:
• Charged traps enriched in the depletion zone screen out the gate voltage [96-97]
and lead to considerably lower free carrier concentration than that in the channel;
• Enhanced disorder and lower carrier concentration in the amorphous or
microcrystalline transition region reduce the effective carrier mobility;
• Disordered film structure induces inhomogeneities and energy level fluctuations at
the metal-organic semiconductor interface [97-104].
These effects must be carefully considered and taken into account for the device
modeling. Note that similar transition regions and relevant effects can be also identified
in top-contact devices as confirmed by both experimental observations [105] and
simulation results [93, 106-107]. The transition width (d(tr)), however, differs from in top-
contact and bottom-contact structures since the former is more heavily related to the
organic film thickness due to the fact that most charge carriers need to cross the entire
film to reach the interfacial channel. While for top-contact devices, metal evaporation
and organic film surface contamination are anticipated to cause conjugation break and
traps creation [105-106] in the neighborhood of the contacts, thus contribute to the
transition width.
Based on the above analysis, our universal device model incorporates contact effects
[88, 90, 93-97, 108-115] at the metal-organic interface as well as in its neighboring
transition region, and charge transport effects from the transition region to the active
channel. Since the contact resistance is normally found less pronounced at the drain for
charge extraction [112, 114], for simplicity, we only consider the lumped contact effects
72 CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR
at the source where charge injection occurs. For consistency and generality, we focus on
p-type organic transistors throughout discussions in this chapter. The results apply to n-
type devices by the same token.
4.2.1 Charge Injection at the Metal-Organic Interface
Although charge injection at the metal-organic semiconductor interface has been
investigated extensively with diversified theories [116], finding an appropriate model
suitable for organic transistors is not an easy task. One of the most classical models to
treat injection from metal to semiconductor or insulator is based on Richardson-
Schottky (RS) thermionic emission theory [117], in which the current density (J) is
expressed as
3*
022 3
/4exp exp
2b
RSe FeemJ kT
kT kTπεφ
π= ⋅ − ⋅ (4.1)
where e is the elementary charge, m* the carrier effective mass, ħ the reduced Planck
constant, k the Boltzmann constant, T the temperature, φb the real barrier height
between the metal and the semiconductor or insulator, which can deviate from the ideal
Schottky barrier due to Fermi-level pinning effect [45] as discussed in Chapter 5, F0 the
electric field at the interface, and ε the dielectric constant of the semiconductor or
insulator material. The second exponential term in equation (4.1) represents the barrier
lowering by ∆φb=(eF0/4πε)1/2 at a distance of x0=(e/16πεF0)1/2 away from the interface
due to external electric field and the image force as illustrated in Figure 4.2. The RS
model reflects the maximum charge injection current and doesn’t take into account
carrier diffusion or inelastic scattering induced backflow current inside low-mobility
materials, including most organic semiconductors as discussed in this work.
CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR 73
Metal Fermi level
Metal-Organic Interface
b∆ b
External electric field:-eF0x
HOMO0 x
Image force: -e2/16πεx
Gaussian localized States Mean electrostatic potentialU(x)=e B-eF0x-e2/16πεx
x0
Interface dipoles:Fermi-level pinning
Figure 4.2 The energy level diagram of the metal-organic semiconductor interface.
We therefore invoke diffusion-limited thermionic emission model for charge
injection at the metal-organic interface [116, 118-122]. In this framework, charge
injection has been either treated by considering an extra drift-diffusion process [118-
120], or, alternatively, in conjunction with surface recombination in an image potential
[116, 121-122]. Indeed, the model can be described in the following general form,
3
00 0
/4' , exp expbtr
e FeJ eN F TkT kT
πεφμ β= ⋅ ⋅ − ⋅ (4.2)
with N0 being the effective density of states (DOS) for the organic semiconductor, μ’(tr)
the carrier mobility in the transition region and close to the contact interface where
electric field is equal to F0. Note that the mobility here is dependent on the electric field,
temperature, carrier concentration as well as the film microstructure [52-53, 123], and
differs from that in the active channel region. We’ll come back to this point in next
section. β(F0, T) is a function in terms of the electric field and temperature, and its exact
74 CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR
formula depends on the particular derivation approach. A more accurate and unified
description at both low [118-119] and high [119] electric field can be deduced from
Scott et al.’s injection-recombination model [121-122] as follows:
Figure 4.3 The calculated boundary condition for diffusion-limited charge injection at metal-organic semiconductor interface.
So far, we have quantitatively justified the applicability of diffusion-limited charge
injection theory to model organic transistors from the perspective of pristine metal-
organic interface. However, the disordered nature of organic semiconductor films close
to the contact interface has not yet been considered. As pointed out in literature before
[97, 109, 112, 128], neglect of the disorder feature raises contradictory argument on the
activation energy of the injection current (eφJ). Experimental results showed that eφJ is
lower than the activation energy of carrier mobility (eφμ) [109, 112, 128], apparently in
disagreement with prediction from equation (4.2), where φJ=φμ+φb-∆φb≥φμ satisfies. A
well accepted description to the disordered nature of organic semiconductor films is
based on Gaussian density of localized states [97-104, 116] as illustrated in Figure 4.2. In
this circumstance, the injection current density is found to be
76 CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR
30
0 0
2
2
320
0 0
/4' , exp exp
1 exp2 / 2
/4/2' , exp exp
tr
b
btr
e FeJ eN F TkT kT
e e de
e Fe kTeN F TkT kT
πεφμ β
φ φ φπσ σ
πεφ σμ β
+∞
−∞
⎡ ⎤⎢ ⎥= ⋅ ⋅ − ⋅ ⋅⎢ ⎥⎢ ⎥⎣ ⎦
⎡ ⎤− −⎢ ⎥⋅ ⎢ ⎥⎣ ⎦
−= ⋅ ⋅ − ⋅
∫
(4.5)
where σ is the Gaussian width of the localized states energy distribution. After taking
into account the Gaussian distribution of the localized states at the interface, we find the
activation energy of injection current is lowered by ∆σ=σ2/2kT≈50-200 meV for
T=300 K, σ=50-100 meV [97, 100-104, 109], thus being consistent with experimental
observations.
We emphasize that it is reasonable to treat charge injection at the metal-organic
interface separately from charge transport in the transition region beyond the interface
[97, 119]. Space charge effect is ignored in dealing with charge injection due to the fact
that the rapid variation of the electrostatic potential takes place within a short distance
from the electrode, typically x0=(e/16πεF0)1/2<10 Å, and the electric field F0 may be
regarded as being constant at the interface [119]. For charge transport in the transition
region with a typical width of d(tr)=20-200 nm [97, 109, 112], space charge effect
becomes pronounced since the field strength (F) in this region no longer satisfies the
condition of F>>2ned(tr)/ε~107 V/m, where n is the carrier concentration [119].
Therefore, we conclude that space charge limited conduction (SCLC) dominates the
charge transport in the organic transition region while diffusion limited charge injection
dominates at the metal-organic interface. As discussed in the following sections, they
naturally converge in the device operation through a self-consistent interfacial electric
field, F0. It has been suggested by Koehler et al. [97] recently that the commonly
observed yet poorly understood contact voltage/resistance for organic transistors may
originate from a combining effect of charge injection at the interface and SCLC in the
transition region. However, the feasibility of either mechanism for organic transistors is
CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR 77
not justified and the mobility difference in different regions is not considered in their
work.
We now proceed to discuss and model the charge transport process in the active
organic semiconductor layer, including that in the transition region close to the contact
and the major channel region, as shown in Figure 4.4 (a).
Figure 4.4 (a) Schematic illustration of the transition region close to the contact and the channel region in the organic semiconductor film; (b) The DOS profiles of the transition region and the major channel based on the mobility edge model; (c) An equivalent representation of (b) with the critical energy for both regions set to be identical to simplify the mathematical derivation.
78 CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR
4.2.2 Charge Transport in the Active Layer: EPME Model
Here we discuss charge transport in the active semiconductor layer which is obviously
not a pure organic single crystal. Instead, it is an amorphous or polycrystalline organic
film coupled with disorders, traps and defects. These nonidealities give rise to a broad
distribution of Anderson localized states [129], and, accordingly, band only transport
description is unrealistic.
In general, there are two families of classical models available in dealing with charge
transport process in the disordered small-molecule or polymer films. One is based on
the hopping theory, including Miller-Abrahams’ fixed range hopping model [130] and
Mott’s variable range hopping (VRH) model [131], which calculate thermally activated
tunneling events between localized states. In this theoretical framework, carriers may
either hop to a nearest site with large activation energy or a distant site with low
activation energy. Vissenberg and Matters [132] have successfully applied this concept in
conjunction with a percolation criterion for the film conduction to the carrier mobility
model for amorphous organic TFTs. The other one is based on the mobility edge (ME)
concept [131, 133-134] as illustrated in Figure 4.4, or, equivalently, multiple trapping and
release (MTR) concept [135-136], which defines a critical energy (E0, the “mobility
edge”) separating immobile localized states (μ=0) with a DOS tail in the forbidden
bandgap from the extended band of mobile states (μ=μ0). Trapped carriers become
mobile by thermal excitation into the delocalized band and thereby contribute to the
conduction. This model has been successfully applied to study semicrystalline polymer
transistors by Salleo et al. [133] and later refined by Chang et al. [137] by incorporating
the polaronic effect.
For the work presented here, we choose the ME theoretical framework since
hopping transport theory such as the VRH model doesn’t take into account the nature
of supramolecular ordering and mesoscale microstructure in the semi- or poly-crystalline
organic semiconductor films. Also, ME model has been shown from studies of the
microstructure effect and polaronic effect on the carrier mobility that it provides an
improved fitting to the experimental data over VRH based model [133, 137]. It should
CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR 79
be noted that, however, previous ME models adopted for organic transistors haven’t yet
considered the electric field effect on the mobility. Since the electric filed dependence of
the mobility becomes pronounced at low temperature or high field, which is the
situation typically found in the transition region as discussed in last section, we therefore
first develop an extended polaronic mobility edge (EPME) model here by taking into
account the field dependence of the carrier mobility. This EPME model for charge
transport is then integrated into our universal device model for describing the organic
transistor operation.
As shown in Figure 4.4 (b), the DOS profiles of the organic semiconductor films are
represented by a delocalization band with an exponential tail of donor-like localized
states decaying into the bandgap [133]. The exponential tail here is an approximation of
the Gaussian profile generally accepted for disordered organic semiconductors. First
principle calculations on the electronic structure of polycrystalline polymers [138] have
explicitly shown that the carrier delocalization in ordered lamellas increases the
bandwidths for both the highest occupied molecular orbital (HOMO) and the lowest
unoccupied molecular orbital (LUMO), consequently, giving rise to a bandgap reduction.
Therefore, considering the different disorder levels in the transition region and in the
channel region as discussed at the beginning of Section 4.2, we propose that the critical
energy defining the mobility edge (E0) is shifted by ∆E0=E0(Ch)-E0(Tr), roughly 0.2-0.4 eV,
between these two regions. The DOS profiles shown in Figure 4.4 (b) are described as
0 0
0 0
1 sgnexp2
1 sgn12
tail
tail tail
tail
tail c
N E E E EDOS EE E
N E E E EE E
− − + −= ⋅ ⋅Δ Δ
− + −+ ⋅ − ⋅Δ
(4.6)
where sgn(x) is the sign function, Ec is a parameter to tailor the DOS profile for the band,
Ntail is the concentration of tail states, ∆Etail=Etail-E0 is the tail width of the localized
states. The DOS for the delocalization band is simply represented with a form for 3-D
free hole gas. The above parameters accompanied by a footnote, “tr” or “ch”, in Figure
80 CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR
4.4 (b), denote those for the transition region or the channel region, respectively. One
should note that the tail states concentration and band mobility (μ=μ0) are assumed
identical for both regions since the major distinction induced by structural modifications
is the localized tail width [133].
When a gate voltage (VGS) applies, charges tend to accumulate along the organic
semiconductor-dielectric interface. The total charge density, including that of immobile
localized trap states and mobile free carriers, can be approximated as
, i GS ontot GS
C V VN V Th−= (4.7)
where Ci is the dielectric capacitance per area, Von is the turn-on voltage, or, equivalently,
the flat band threshold voltage when charge start to accumulate, which ideally should be
around 0 V [36], h is the first-order approximation of the charge layer thickness (~1 nm)
at the dielectric-semiconductor interface.
On the other hand, Figure 4.4 indicates that the total charge density, Ntot(VGS, T),
and the mobile carrier concentration, Nmob(VGS, T), are represented as follows:
, 1tot GSN V T DOS E F E dE+∞
−∞= ⋅ −∫ (4.8)
0
, 1E
mob GSN V T DOS E F E dE−∞
= ⋅ −∫ (4.9)
where
/1
1 fE E kTF Ee −=+
(4.10)
is the Fermi-Dirac distribution. Combine equations (4.6)-(4.10), we can readily find the
solution to Fermi energy, Ef., and the results apply to both the transition region (“tr”)
and the channel region (“ch”). Figure 4.5 shows the calculated relationship between the
CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR 81
tail width of localized states (∆Etail or σ), the relative Fermi energy level (Ef-E0), and the
applied gate voltage (VGS) for a typical organic transistor at room temperate. It is noted
that the Fermi level position moves to deeper bandgap with the increased localization
tail width, in agreement with the observation that the effective mobility decreases with
the broadening of the localized states. Also, the applied gate voltage systematically tunes
the Fermi energy level position for every single tail width. Higher gate voltage induces
more free charge carriers in the organic semiconductor films, and, equivalently, shifts
the Fermi energy level closer to the mobility edge. Another important feature as revealed
in Figure 4.5 is that, for the transition region where the tail width of localized states is
larger than that the channel region, e.g. ∆Etail(tr)=100 meV and ∆Etail(tr)=30 meV, its
relative Fermi level tends to be 0.2-0.5 eV deeper. Consider the difference in the critical
energy by ∆E0=E0(Ch)-E0(Tr),~0.2-0.4 eV as mentioned before, the absolute Fermi energy
positions in both regions thus align consistently.
0.0 0.2 0.4 0.6 0.8 1.00.0
0.1
0.2
0.3
0.4
ΔEtail(ch)
ΔEtail(tr)-1 V
-10 V-20 V
-40 V-60 V-80 V
Tail
Wid
th (e
V)
Relative Fermi Position, Ef-E0 (eV)
VGS=-100 V
0.0
0.1
0.2
0.3
0.4
Figure 4.5 The relationship between the tail width of localized states (∆Etail), the relative Fermi energy level (Ef-E0), and the applied gate voltage (VGS). The following conditions are employed for the calculation here: 300 nm SiO2 gate dielectric, Von=0 V, h=1 nm, Ntail=1021/cm3, Ec=40 meV, T=300 K.
82 CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR
Without incorporating polaronic effect, the overall effective carrier mobility, μeff, is
simply defined as μeff=μ0×(Nmob/Ntot) [133]. In reality, strong electron-phonon coupling
existing in the conjugated organic semiconductor molecules or polymers leads to
polaron formation and thus contributes to the temperature activation of the carrier
mobility. A refined description of the effective mobility has been suggested [137] as
0,, exp,
mob GS aeff GS
tot GS
N V T EV TN V T kT
μ μ= ⋅ ⋅ − (4.11)
where the polaronic effect is represented by an extra exponential term, based on the
assumption that the mobility beyond the mobility edge is thermally activated with a
polaron activation energy of Ea.
Another important factor in determining the carrier mobility behavior is the electric
field dependence, which has been an active research subject for disordered organic
materials. It is now well recognized that this dependence strongly relies on the
temperature and field range as well as the film’s structural properties, such as energetic
or positional disorder levels [139-146]. A routine finding on organic semiconductors,
from both Monte Carlo simulations [139-142] and experiments [143-146], is that within
an intermediate field range, typically 104-106 V/cm, the field-dependent mobility exhibits
the Poole-Frenkel like behavior and can be described in a general form as
1exp n FT
μ γ∝ ⋅ (4.12)
where γ(1/Tn) is an empirical parameter relying on temperature (T), n=1-2 [139, 141-
146]. Note that the lateral field strengths for the transition region and the channel region
shown in Figure 4.4 (a) can be considerably different. Therefore, a rough estimation of
the average field by F=VDS/L (L is the channel length) as frequently seen in literature is
not necessarily correct. We estimate the field strengths here from the scanning probe
potentiometry profiles [112, 114] and find that the field dependence as described in
(4.12) is appropriate in both regions. However, it should be stressed that, due to their
CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR 83
appreciable difference in the film microstructure and thus energetic/spatial disorder,
γ(1/Tn) for the transition region is expected larger than the less disordered channel
region: γ(tr)(1/Tn)>γ(ch)(1/Tn). This can be inferred from Bässler’s formula [139],
2 20
1n C
T kTσγ = ⋅ −∑ (4.13)
where σ is the energetic width of the Gaussian DOS, ∑ represents the off-diagonal
(spatial) disorder. For disordered organic materials, γ(1/Tn) is normally 0.001-0.005
cm1/2V-1/2 at room temperature. With the increasingly ordered structure (or smaller σ)
and higher temperature, γ(1/Tn) eventually vanishes or even turns negative [140, 143].
The extended polaronic mobility edge (EPME) model proposed in this work
incorporates the effective mobility derived in (4.11) as the prefactor for the field-
dependent mobility described in (4.12). The carrier concentration, temperature, and
electric field-dependent effective mobility is expressed collectively as below:
0, 1, , exp exp,
mob GS aeff GS n
tot GS
N V T EV T F FN V T kT T
μ μ γ= ⋅ ⋅ − ⋅ ⋅ (4.14)
Equation (4.14) is a general form adequate for both the transition region and the
channel region. Here we assume the EPME effective mobility derived for charge carriers
located at the dielectric-semiconductor interface applies to the entire film thickness. This
approximation is judged from the nature of mobility edge model which inherently
relates the film microstructure to the effective mobility. For clarity, we reiterated the
mobility in both regions as follows:
0
, 1, , exp exp,
mob tr GS atr GS tr n
tot GS
N V T EV T F FN V T kT T
μ μ γ= ⋅ ⋅ − ⋅ ⋅ (4.15)
84 CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR
0
, 1, , exp exp,
mob ch GS ach GS ch n
tot GS
N V T EV T F FN V T kT T
μ μ γ= ⋅ ⋅ − ⋅ ⋅ (4.16)
We emphasize that, in circumstances such as in contact-induced polycrystalline
organic thin film transistors [27], solution-shearing processed polycrystalline organic
transistors [15, 38, 43], etc., the overall film crystallinity is correlated to the channel
length. The channel length dependent crystallinity will thus affect the μ(ch) in (4.16) by an
additional function of channel length. We’ll return to this point when discussing the
mobility scaling behavior in Section 4.3.3.
4.2.3 Unified Model for Transistor Operations
Based on the discussion and analysis in Sections 4.2.1 and 4.2.2, we are now able to
combine all of the effects to describe the transistor operation. Since SCLC dominates in
some local regions, it is necessary to include in the overall drain current (IDS) both the
SCLC current arising from the drain-source voltage (VDS) and the charge accumulation
current arising from the gate-source voltage (VGS). Accordingly, we have
DS SCLC tr accumulation tr SCLC ch accumulation chI I I I I= + = + (4.17)
For SCLC component with negligible diffusion current, it follows Gauss’s law
eF ρε
∇⋅ = (4.18)
and the current transport equation
, , , , xSCLC SCLC SCLC x GS x x GS x
dFI J A WHe F V T F WHF V T Fdx
ρ μ μ ε= ⋅ ≅ = ⋅ ⋅ ⋅ (4.19)
CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR 85
where ρ is the charge carrier density, x denotes the channel direction from source (x=0)
to drain (x=L) as shown in Figure 4.4 (a), ASCLC is the effective cross sectional area for
the SCLC current. In a first-order approximation, ASCLC =W×H, where W is the
transistor’s channel width, H is the organic semiconductor film thickness.
For charge accumulation induced by the gate voltage at the dielectric-semiconductor
interface, it contributes to the drain current at position x by
, ,accumulation GS x x i GS on xI V T F F C W V V Vμ= ⋅ ⋅ ⋅ ⋅ − − (4.20)
where Vx is the electrical potential at position x resulted from the drain-source voltage,
and it relates to the electric field (Fx) by
x xF V=−∇⋅ (4.21)
and
0 0
L L
x x DSF dx V dx V= −∇⋅ =−∫ ∫ (4.22)
Plug equations (4.15), (4.16), and (4.19)-(4.22) into equation (4.17), we have
0
'0
, 1exp exp,
'
mob GS aDS x xn
tot GS
xx
i GS on x
N V T EI WF FN V T kT T
dFH C V V F dxdx
μ γ
ε
= ⋅ ⋅ − ⋅ ⋅
× ⋅ ⋅ + ⋅ − +∫
(4.23)
Integration of equation (4.23) from x=0 to x=L on both sides yields the drain
current equation for the transistor as follows:
86 CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR
0 0
'0
,1exp exp,
'
La mob GS
DS x xntot GS
xx
i GS on x
E N V TWI F FL kT T N V T
dFH C V V F dx dxdx
μ γ
ε
⎧⎪⎪= − ⋅ ⋅ ⋅⎨⎪⎪⎩
⋅ ⋅ ⋅ + ⋅ − +
∫
∫
(4.24)
Alternatively, equation (4.24) can be formulized with the transition region and the
channel region described by separate terms,
0 0
'0
0
,1exp exp,
'
1exp exp
trd mob tr GSaDS x tr xn
tot GS
xx
i GS on x
ax ch
N V TEWI F FL kT T N V T
dFH C V V F dx dxdx
EW FL kT T
μ γ
ε
μ γ
⎧⎪⎪= − ⋅ ⋅ ⋅⎨⎪⎪⎩
⋅ ⋅ ⋅ + ⋅ − +
+ − ⋅ ⋅
∫
∫
'0
,,
'
tr
L mob ch GSxnd tot GS
xx
i GS on x
N V TF
N V T
dFH C V V F dx dxdx
ε
⎧⎪⎪ ⋅⎨⎪⎪⎩
⋅ ⋅ ⋅ + ⋅ − +
∫
∫
(4.25)
where d(tr) is the width of the transition region, typically 20-200 nm.
Equation (4.23)-(4.25) are the derived current equation for organic transistors based
on our universal model incorporating both the charge injection effects and charge
transport properties. The model is essentially physics-oriented since each variable
involved has real and particular physical meanings. This model successfully enables us to
account in a single formula for most of the significant physical phenomena observed so
far for organic semiconductor films and their based transistors.
In order to solve the above differential equations to fully simulate the transistor’s
behavior, it is necessary to obtain the boundary conditions. Therefore, we invoke the
injection current density at the source contact, JSD, as given by equation (4.5) in Section
4.2.1. Substitute (4.15) for the mobility μ(tr) in (4.5), we obtain:
CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR 87
3
0 0 0
2
0
/41exp exp
' /2, exp
aSD tr n
mob tr b
tot tr
eEJ eN FkT T kT
N e kTF TN kT
πεμ γ
φ σβ
= − ⋅ + ⋅
−⋅ ⋅ ⋅ −
(4.26)
Note that we distinguish μ’(tr) in equation (4.5) and μ(tr) in (4.15) by assuming that the
carrier mobility for charge injection at the metal-organic interface, i.e. μ’(tr) in (4.5), is
correlated to the film disorder or localized states tail width by N’mob(tr)/Ntot(tr)∝exp(-σ/kT),
and, is independent of the applied gate voltage. While for the charge transport mobility
in the film, i.e. μ(tr) in (4.15), it is apparently gate voltage dependent since Nmob(tr)/Ntot(tr) is
defined by equations (4.7)-(4.10).
The charge injection current equation (4.26) provides the following boundary
conditions for the transistor’s ordinary differential current equation (4.23),
DS SD inj SDI J A J WHα= ⋅ = ⋅ (4.27)
0 ' 00' 0
x
x xV F dx == − =∫ (4.28)
0 0x xF F == (4.29)
where Ainj is the effective injection area, α is a prefactor determined by the transistor
geometry. In addition, to resolve the full transistor behavior, another similar boundary
condition for F(x=d(tr)) and V(x= d(tr)) should also be employed at the transition and
channel region connection position, x=d(tr).
So far, we have analytically derived the universal device model by incorporating both
charge injection effects and charge transport properties for organic thin-film transistors.
This model has been implemented in C/Mathematica programs to render a full 1-D
88 CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR
device simulator with most of the important physical effects represented by particular
physics-oriented parameters or variables. In next section, we will discuss the simulation
results based on this model and compare them to our or previously reported
experimental results, and hence to testify the theories for the device modeling and
resolve a few elusive physical phenomena for organic transistors.
4.3 Simulation vs. Experimental Results
For the simulation purpose here, we apply typical values as identified from experiments
for the representative parameters in the model (see Table 4.1). The definitions and the
value settings of these parameters have been discussed and justified in Section 4.2.
Table 4.1 Representative parameters, corresponding physical meaning and typical value used for the simulation based on the universal device model in this work
Parameter Physical Meaning Value
W Transistor channel width 1000 µm
L Transistor length Specified in simulation
H Organic semiconductor film thickness 50 nm
d(tr) Transition region width 100 nm
h Charge layer thickness 1 nm
µ0 Intrinsic mobility 1 cm2/Vs
Ci Gate dielectric capacitance 11.5 nF/cm2 (300 nm SiO2)
T Temperature 300 K
Von Onset voltage 0 V
VGS,VDS Gate-source voltage, Drain-source voltage Specified in simulation
Ntail Concentration of tail states 1021 /cm3
Ec DOS profile tailoring parameter 40 meV
Ea Polaron activation energy 15 meV
CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR 89
ΔEtail(tr) Localized trap states tail width, in transition ~σ, 50-100 meV
ΔEtail(ch) Localized trap states tail width, in channel 30 meV
γ(1/Tn) Prefactor of mobility’s PF field dependence Specified in simulation
ε Organic semiconductor dielectric constant 3 ε0
φb Metal-organic injection barrier height Specified in simulation
σ Gaussian width of film’s localized states ~ ΔEtail(tr), 50-100 meV
4.3.1 Transfer and Output Characteristics
The basic electrical characterization of organic transistors includes the transfer and
output curves as shown frequently in the previous chapters. Based on the universal
device model introduced in Section 4.2, we simulate the output [Figure 4.6 (a)] and
transfer [Figure 4.6 (b)] characteristics for a typical organic transistor with the parametric
values listed in Table 4.1. It can be found that the device model excellently reproduces
the typical IDS-VGS and IDS-VDS curves for organic transistors operating in both the linear
region and the saturation region.
Particularly, with a relatively large injection barrier height at the metal-organic
semiconductor interface (φb=0.7 eV) and localized trap states distribution (σ=100 meV),
a superlinear, concave starting region is clearly reflected in the output characteristics as
shown in Figure 4.6 (a). This phenomenon has been routinely observed in experiments
for organic transistors with poor contacts or significant contact resistance [95, 109, 147].
To gain a better fundamental understanding on the contact-induced peculiarity, we also
simulate the same device with a smaller injection barrier height of φb=0.2 eV. The
superlinearity in the output curves, though does not entirely disappear, is considerably
alleviated per the injection barrier height reduction as seen from Figure 4.6 (a) inset. In
addition, the driving current is enhanced by ~20% over the device with an injection
barrier height of 0.7 eV. This clearly shows the importance of controlling the metal-
organic interface for low injection barrier height to optimize the device performance.
90 CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR
The residual superlinearity for even low injection barrier height can be attributed to the
other factor, localized trap states distribution width (σ), as discussed in next section.
Compare the above simulation results to our 4T-TMS SPOFET performance as
presented in Chapter 2, we reaffirm the 4T-TMS SPOFET possesses favorable contacts
at the source/drain with a low contact injection barrier height and a narrow localized
trap states distribution in the organic film.
CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR 91
0 -20 -40 -60 -80 -100
0
50
100
150
VGS=-20 V
VGS=-60 V
VGS=-80 V
VGS=-100 V
I DS (μ
A)
VDS (V)
φb=0.7 eV, σ=0.1 eV
φb=0.2 eV, σ=0.1 eV
VGS=-100 V
0 -10 -20
0
10
20
30
40
I DS (μ
A)
VDS
(V)
(a)
0 -20 -40 -60 -80 -100
0
20
40
60
80
100
120
140
VGS (V)
I DS (μ
A)
10-13
10-11
10-9
10-7
10-5
(b)
I DS (A
)
VDS=-100 V
Figure 4.6 Simulation results of the (a) output and (b) transfer characteristics, based on the universal device model introduced in Section 4.2. In addition to parameters given in Table 4.1, the followings are included: L=10 µm, γ(tr)(1/Tn)= 0.001 cm1/2V-1/2, γ(ch)(1/Tn)=0, (a) φb=0.7 or 0.2 eV, σ=0.1 eV, and (b) φb=0.7 eV, σ=0.1 eV. Note: for the transfer curve simulation, an Ioff
(IDS@VGS=0)=1 pA is assumed per practical situations.
92 CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR
4.3.2 Surface Potential Profile and Contact Resistance
Surface potential profile along the dielectric-semiconductor interface has been well
studied and understood in traditional silicon MOSFETs [117]. It was until recently that
this topic was recognized to be critical in studying organic thin-film transistors. In fact,
the surface potential profile contains ample information on the transistor operation
status and the contact effects. A few early works have probed this potential profile for
organic transistors using high-resolution conductive probe atomic force microscopy
(AFM) [114], electrostatic force microscopy (EFM) [148], or scanning Kelvin probe
microscopy (SKPM) [112, 149], thus providing a foundation for justifying the electric
field distribution and estimating the contact resistance. However, the fundamental
physics and mechanisms responsible for these experimental observations are still poorly
understood.
Based on our universal model which takes into account both the charge injection
effects at the metal-organic interface and charge transport properties in the active film
layer, we are able to simulate the surface potential profile for the organic thin-film
transistors [see Figure 4.7 (a)]. Devices operating from linear regime (-VGS>-(VDS+VTH))
to saturation regime ((-VGS<-(VDS+VTH))) are explicitly revealed by the surface potential
profile change upon increase of the applied drain voltage. For –VDS>80 V, the channel
area close to the drain is gradually getting to depletion mode for saturation operation.
The channel regions which are in accumulation mode show a seemingly quasi-linear
potential profile and only consume little voltage drop along the channel, in distinct
contrast to the transition regions where a significant voltage drop occurs in the case of
φb=0.7 eV and σ=100 meV. This dramatic voltage consumption in a short distance close
to the source, i.e. the transition region as depicted in Figure 4.4 (a), contributes to the
extrinsically large contact resistance as regularly found in devices with poor contacts. For
direct comparison, Figure 4.7 (b) invokes a representative surface potential profile of the
P3HT-polymer transistor as recorded in experiments by noncontact potentiometry using
scanning Kelvin probe microscope by Bürgi et al. [112]. It can be found that the
CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR 93
universal model developed in this work excellently reproduces surface potential profiles
94 CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR
To understand the peculiar surface potential drop in the vicinity of source and the
contact resistance, we recall in Section 4.2 the physics of charge injection and charge
transport as incorporated in the universal model. Apparently, both charge injection right
at the metal-organic interface and the SCLC in the short transition region cause the
substantial voltage consumption and thus significant contact resistance, as revealed by
Figure 4.7 (a) inset. We further perform simulation on the surface potential profiles for
the same transistor only with different injection barrier height (φb=0.7 eV or 0.3 eV) and
localized trap states tail width (σ=100 meV or 50 meV). Figure 4.8 shows that,
• Both charge injection barrier height and the localized trap states distribution
contributes to the surface potential profiles and thus contact resistance for organic
thin film transistors;
0 2 4 6 8 10
-80
-60
-40
-20
0
0 100 200
-40
-20
0
Pote
ntia
l (V)
Channel Position (μm)
VGS=-100 VVDS=-80 V
φb=0.7 eV, σ=0.1 eV
φb=0.3 eV, σ=0.1 eV
φb=0.3 eV, σ=0.05 eV
Pot
entia
l (V
)
Channel Position (nm)
Figure 4.8 Simulated surface potential profiles along the channel direction for the same transistor only with different injection barrier height (φb=0.7 eV or 0.3 eV) and localized trap states tail width (σ=100 meV or 50 meV). The inset is a magnified view in the 200 nm region close to the source contact.
CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR 95
• For the same localized states distribution in the organic film, lower charge
injection barrier height at the metal-organic interface mitigates voltage drop at the
transition region and thus reduces the contact resistance. This is consistent with
the observation on output curves as shown in Figure 4.6 (a);
• Surprisingly, for the same charge injection barrier height at the metal-organic
interface, narrower localized trap states distribution, or, equivalently, smaller
Gaussian energy width of the organic film’s localized states close to the metal-
organic contacts, leads to substantially lower voltage drop in the transition region,
accordingly, giving rise to significantly reduced contact resistance. This effect is
not expected at first glance, but indeed plays a critical role in determining the
organic transistor performance as found in this study.
4.3.3 Mobility Scaling Behavior
As introduced at the beginning of this chapter, a critical issue remaining for organic
transistors is to resolve their mobility scaling behavior. By virtue of our universal
physical model in Section 4.2, we are now able to elucidate the apparently contradictory
mobility scaling features as found in experiments and reported in literature.
Figure 4.9 shows the simulation results of the mobility scaling behavior with respect
to different physical origins including carrier injection barrier height (φb), localized trap
states energy distribution (σ), Poole-Frenkel like field dependence of carrier mobility
(γ(1/Tn)), and channel length dependence of film crystallinity (γc). It is revealed that,
• Depending on the weight of the above physical effects, mobility scaling curve
with respect to the channel length (L) ranging from 100 nm to 50 µm can be
monotonically upwards, or slowly get saturated, or, surprisingly, exhibit an
overshoot region;
• Lower charge injection barrier (e.g. φb=0.3 eV vs. φb=0.7 eV), smaller localized
states distribution tail width in the transition region (e.g. σ=50 meV vs. σ=100
meV), and stronger field dependence of carrier mobility in the channel region (e.g.
96 CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR
γ(ch)(1/Tn))=0 vs. γ(ch)(1/Tn))=0.01 cm1/2V-1/2) all give rise to the increase of the
extrinsic field-effect mobility (μFET).
• Importantly, large modulation of the extrinsic mobility can be realized by 1-2
orders of magnitude through engineering the charge injection barrier and the
localized states distribution at the metal-organic semiconductor interface;
• The organic transistors show an extrinsic field-effect mobility increasing
monotonically with the channel length when a large injection barrier height and a
wide disordered localization states distribution occur at the contact region, which
leads to significant extrinsic contact resistance as discussed in Section 4.3.2. Such
phenomenon is consistently observed for organic transistors with the so-called
“poor contacts” in literature [76-90], and adversely compromises the benefits
from device scaling theory as introduced at the beginning of this chapter;
• It is possible to maintain a steady extrinsic field-effect mobility at a normal
channel length above the transition region width, typically, on the level of
micrometers, given that an adequate combination of charge injection barrier
height at the metal-organic interface and localized states energy distribution in the
film is established. For instance, the simulation here shows the extrinsic mobility
is almost constant for a channel length of 5-50 µm, with an injection barrier height
of 0.3 eV and a localized trap states energy distribution width of 50 meV;
• Significantly, we discover an overshoot region that may emerge in the mobility
scaling curve for organic thin film transistors within normally fabricated device
size range, provided that the electric field dependence of carrier mobility
(γ(ch)(1/Tn)) and the channel length dependence of film crystallinity (γc) in the
channel region take strong effect. This finding is understood as follows: For short
channel devices, e.g. L=100 nm to 5 µm for the simulation shown in Figure 4.9,
the extrinsic field-effect mobility is dominated by charge injection at the contact
interface and charge transport in its neighboring transition region with a low
mobility. Therefore, shrinking the channel length equivalently enhances this
dominance and leads to a lower extrinsic mobility. However, the scenario
CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR 97
becomes different when the channel length is far greater than the transition width,
e.g. L>5 µm for the simulation here: First, the weight of charge transport in the
channel region becomes important or even dominant; Secondly, under the same
operating voltage, shortening of the channel length increases the longitudinal
electric field strength, thus rendering a higher effective mobility in the channel
region, and as a consequence, boosting the extrinsic mobility of the device;
Additionally, for polycrystalline film transistors, the overall film crystallinity may
be enhanced for shorter channel devices [15, 27, 43], thus further increasing the
effective mobility. This effect can be quantitatively described as an extra term in
Equation (4.16) to modify the portion of free mobile states over the total states in
the channel region, Nmob(ch)(VGS, T, L)/Ntot. For simplicity, the simulation
performed here lumps together both the electric field dependence effect and the
film crystallinity effect as shown in Figure 4.9. Our finding in this study also
successfully explains the experimental results for the mobility scaling behavior of
solution-shearing processed polycrystalline 4T-TMS SPOFETs, as shown in
Figure 4.10. Note that the mobility scaling trend here is apparently opposite to
other reports [76-90], but still in agreement with our finding based on the
universal model. We emphasize that for these devices in top-contact structure,
shorter channel lengths below 10 µm are unavailable yet due to shadow mask
limitation. Future fabrication of shorter channel devices will be able to testify the
overshoot region as predicted in the model.
• The critical channel length for the peak extrinsic mobility in the overshoot region,
Lpeak, depends on the weight of the electric field dependence of carrier mobility
(γ(ch)(1/Tn)) and the channel length dependence of film crystallinity (γc) in the
channel region, the carrier injection barrier height (φb), as well as the localized trap
states energy distribution (σ). Lower injection barrier height and smaller localized
trap states energy tail width gives rises to shorter channel length where peak
extrinsic mobility appears. This suggests that for normal bottom contact devices
where the contact regions are showing poor morphology than top contact ones,
the critical channel length Lpeak can be considerably large, usually beyond the
98 CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR
typical device channel length as we fabricate in lab. This explains why it is rare to
find the mobility scaling trend similar to that shown in Figure 4.10 for most
bottom contact devices.
• Indeed, the critical channel length for the peak mobility serves an indication of the
Figure 4.9 Mobility scaling behavior with respect to different physical origins including carrier injection barrier height (φb), localized trap states energy distribution (σ), Poole-Frenkel like field dependence of carrier mobility (γ(1/Tn)), and channel length dependent film crystallinity (γc). For simplicity of the simulation here, the film crystallinity effect is lumped into the mobility’s electric field effect.
CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR 99
0 10 20 30 40 500.0
0.1
0.2
0.3
0.4
0.5
Mea
sure
d E
xtrin
sic
Mob
ility
μ FET
(cm
2 /Vs)
Channel Length, L (μm)
Lpeak
Figure 4.10 Measurement results of the extrinsic field-effect mobility with respect to different channels for the solution-shearing processed polycrystalline 4T-TMS SPOFETs. The device fabrication and measurement are detailed in Chapter 2. All of these devices are fabricated on the same wafer with the same process. The dashed line is for visual guidance and refers to the modeling and simulation results shown in Figure 4.9. It is found that the finding based on our model is in excellent agreement with the experimental results.
4.4 Summary
We have developed a universal physical model for organic transistors by
incorporating both charge injection effects at the metal-organic interface and charge
transport properties in the organic semiconductor film, and successfully applied the
device model to resolve many elusive physical phenomena observed so far, such as the
peculiar mobility scaling behavior, the contact resistance effect, and the mysterious
surface potential profiles of organic transistors which have been experimentally probed
yet poorly understood. We also discovered in this study that an overshoot region may
100 CHAPTER 4. DEVICE PHYSICS AND MODELING OF ORGANIC TRANSISTOR
emerge in the mobility scaling curve with respect to the channel length. This work
provides fresh insight into understanding of the fundamental physics of organic
transistors as well as new engineering methods for improving the organic transistor
performance.
101
Chapter 5
Contact Engineering for
Organic Electronic Device
The important thing in science is not so much to obtain new facts as to discover new ways of
thinking about them.
- William Bragg
5.1 Background
In Chapter 4, we have discussed the physics of organic thin film transistors with focus
on charge injection effects and charge transport properties. Through device modeling
and simulation in conjunction with experimental corroboration, we explicitly showed the
significant role of the contacts in determining the organic transistor performance,
including the electrical transfer and output characteristics, the surface potential profile
102 CHAPTER 5. CONTACT ENGINEERING FOR ORGANIC ELECTRONIC DEVICE
along the channel direction, the contact resistance, the extrinsic mobility and its scaling
behavior. As outlined in Chapter 4, it is crucial to lower the charge injection barrier
height and minimize the localized states energy distribution width at the metal-organic
interface to achieve optimized, high-performance devices.
Achieving this goal turns out not an easy task. It is well recognized from both
theoretical [150-152] and experimental [153-155] studies that a high Schottky injection
barrier height can emerge at the metal-organic interface irrespective of the metal work
function, a phenomenon known as Fermi-level pinning, which consequently gives rise to
poor contact properties and unfavorable organic semiconductor device performance.
In this chapter, we present the first direct evidence and demonstration of Fermi-
level depinning at metal-organic semiconductor (M/O) interfaces by inserting an
ultrathin interfacial Si3N4 insulator in between [45]. The contact behavior is successfully
tuned from rectifying to quasi-Ohmic and to tunneling by varying the Si3N4 thickness
within 0-6 nm. Detailed physical mechanisms of Fermi-level pinning/depinning
responsible for the M/O contact behavior are clarified based on a proposed lumped-
dipole model. Experimental results are in good agreement with the theory and the
model.
5.2 Experimental Results on Fermi-Level Depinning
We chose p- and n-type organic semiconductors, pentacene and 3,4,9,10-perylene-
tetracarboxylic-dianhydride (PTCDA), respectively, for this study. Both materials have
been found to exhibit relatively strong Fermi-level pinning at their M/O interfaces [156],
with the pinning factor (S=dφb/dψmetal) of S~0.4 for metal/pentacene and S~0 for
metal/PTCDA. Si3N4 was chosen as the interfacial layer because (1) less-oxygen ambient
suppresses unintentional oxidation of the underlying material, and (2) Si3N4 has a
moderately large bandgap with its conduction/valence band relatively far away from the
LUMO/HOMO level of the aforementioned organic semiconductors, thus excluding
the possibility of charge injection assisted by a close transition energy level as
CHAPTER 5. CONTACT ENGINEERING FOR ORGANIC ELECTRONIC DEVICE 103
traditionally suggested for organic LEDs or solar cells. Si3N4 with different thicknesses
of 0-6 nm (measured from dummy silicon wafers by ellipsometry) was deposited at
room temperature by a precisely-controlled high-vacuum LSI sputtering system.
Figure 5.1 illustrates the device structure and the process flow for our
Au/Si3N4/pentacene diodes. Atomic force microscope (AFM) images of the pentacene
film as shown in Figure 5.2 indicate that there is no damage caused by the Si3N4
sputtering process on the underlying organic semiconductor layer. A large Au electrode
is used as the common anode, and the cathode electrodes with a much smaller area are
either Au pads deposited by thermal evaporation or Au bonding wires. We found that
the diodes with/without Si3N4 based on the evaporated Au pads are regularly shorted
due to Au atoms penetration, leading to an artifactual Ohmic contact with very low
resistance as shown in Figure 5.3.
Figure 5.1 Device structure and process flow of the Au/Si3N4/pentacene diodes. Pentacene was chosen for this study since the pinning factor (S=dφb/dψmetal) S~0.4 is relatively small for the metal/pentacene interface.
104 CHAPTER 5. CONTACT ENGINEERING FOR ORGANIC ELECTRONIC DEVICE
Figure 5.2 AFM images of 200 nm pentacene films deposited on Au at room temperature (a) before and (b) after Si3N4 sputtering deposition. Height scale: 300 nm.
-0.6 -0.4 -0.2 0.0 0.2 0.4 0.6
-200
-100
0
100
200
Curre
nt d
ensit
y (A/
cm2 )
Voltage (V)
t(SiN)=4.1 nm t(SiN)=2.9 nm t(SiN)=1.1 nm
Figure 5.3 I-V measurement results showing the diodes with/without Si3N4 based on evaporated Au pads are regularly shorted due to metal atoms penetration along the grain boundaries or the roughness-induced vacancies.
To resolve the shorting issues and eliminate the electrical artifacts resulting from the
gold atom penetration, we introduced the gold wire based diode configurations for this
study as depicted in Figure 5.1. Figure 5.4 shows the measured I-V characteristics of the
Au/Si3N4/pentacene diodes with different Si3N4 thicknesses, providing direct evidence
CHAPTER 5. CONTACT ENGINEERING FOR ORGANIC ELECTRONIC DEVICE 105
that the metal-organic semiconductor diodes are successfully tuned from rectifying to
quasi-Ohmic and to tunneling ones with increase of the Si3N4 thickness in between, thus
in corroboration with the Fermi-level pinning/depinning theory and model as
introduced in following section. Quasi-Ohmic contact behavior for Au/pentacene is
observed when the Si3N4 thickness is nominally ~4 nm. Figure 5.5 shows the
normalized dynamic resistance (RAC=∂V/∂I) and static resistance (RDC=V/I) of the
Au/Si3N4/pentacene diodes with respect to different Si3N4 thickness. Both resistances
reach the minimum at an optimal Si3N4 thickness of ~4.1 nm, a further evidence of
effective Fermi-level depinning and contact resistance reduction by the insertion of the
Si3N4 interfacial layer. At first glance it is surprising that a nominal Si3N4 thickness of ~4
nm leads to quasi-Ohmic contacts from the tunneling theory point of view. After taking
into account surface roughness of the underlying layer which lowers the effective Si3N4
thickness, we argue that such an experimental observation is still expectable, and in no
doubt is particularly important for practical applications.
106 CHAPTER 5. CONTACT ENGINEERING FOR ORGANIC ELECTRONIC DEVICE
-1.0 -0.5 0.0 0.5 1.0-150-100-50
050
100150
t(SiN)=4.1 nmAu/SiN/Pentacene
Voltage (V)
Curre
nt (n
A)
-1.0 -0.5 0.0 0.5 1.0-10
0
10
20
30
40
Curre
nt (n
A)
Voltage (V)
no SiN:Au/Pentacene
-1.0 -0.5 0.0 0.5 1.0-0.6-0.4-0.20.00.20.40.6
t(SiN)=6.1 nmAu/SiN/Pentacene
Voltage (V)
Curre
nt (n
A)
Figure 5.4 I-V characteristics of the Au/Si3N4/pentacene diodes with different Si3N4 thicknesses (with Au wire as the top cathode electrode), providing direct evidence that the Au/pentacene diode has been successfully tuned to rectifying, quasi-Ohmic, and symmetric tunneling behavior by modulating the Si3N4 thickness. (The effective contact area between Au wire and Si3N4/pentacene is only a fraction of the wire cross section due to the surface roughness of Si3N4/pentacene layer, and may vary from device to device. One should note that the I-V shape giving the diode property is the essence here.)
CHAPTER 5. CONTACT ENGINEERING FOR ORGANIC ELECTRONIC DEVICE 107
Figure 5.5 Normalized dynamic resistance (RAC=∂V/∂I) and static resistance (RDC=V/I) of the Au/Si3N4/pentacene diodes with respect to the Si3N4 thickness, as calculated from their respective I-V measurement curves. Normalization is based on the fact that the maximum forward-biased diode current (here at V=+1 V) is less affected by the Si3N4 thickness. Note: The device with a thick Si3N4 layer (t=6.1 nm) was not normalized here as tunneling dominates therein.
108 CHAPTER 5. CONTACT ENGINEERING FOR ORGANIC ELECTRONIC DEVICE
We also fabricated Ag/Si3N4/PTCDA diode array as shown in Figure 5.6 for this
study. Instead of measuring an individual diode’s I-V curve, here we evaluated the
contact resistance of Ag/Si3N4/PTCDA by measuring two back-to-back diodes in series
connection with a low-conductivity PTCDA bulk film [see Figure 5.7 (a)]. Figure 5.7 (b)
shows explicitly a minimum contact resistance observed for the Ag/PTCDA interface
with an optimal sandwiched Si3N4 thickness of ~2.5 nm, reaffirming the Fermi-level
depinning at M/O interfaces by inserting the ultrathin Si3N4 interfacial layer. The
apparent difference of the optimal Si3N4 thickness for Au/pentacene and Ag/PTCDA
interfaces can be attributed to their different interface roughness (and thus different
effective interfacial layer thickness). This also underscores that one should consider the
particular device structure and materials therein when applying the Fermi-level
depinning technology to high performance organic devices.
(A) (A)(B)
(C)
(D)
(A) (A)
(A) Ag electrode array with thickness=30 nm and spacing=250 μm acting as bottom electrodes; (B) Si3N4 layer with different thicknesses of 0.6 nm, 1.0 nm, 1.9 nm, 2.4 nm, 3.2 nm, 3.9 nm, 6.1 nm; (C) 130 nm PTCDA deposited at room temperature by thermal evaporation;(D) 300 nm SiO2 wafer as the device substrate (holder).
O
OO
O
OO
PTCDA
(A) (B) (C)
1mm
Figure 5.6 Device structure, microscopy image, and process flow of the Ag/Si3N4/PTCDA diodes. PTCDA was chosen here since the pinning factor (S=dφb/dψmetal) is S~0 for metal/PTCDA interface, indicating very strong Fermi-level pinning effect at the metal/PTCDA interface.
CHAPTER 5. CONTACT ENGINEERING FOR ORGANIC ELECTRONIC DEVICE 109
0 ( )
0
0
( )
0( )
,
.
;
;
total c sh film
c
cc
film sh film
cctotal film sh film
R R R l W
R Const
RR Area W LW L
lR RL
R lR R R RW L L
W L α
α
α
⇓⋅ ≈ + ⋅ ⋅ ⋅
≈ +
= = ⋅⋅
= ⋅
≈ + = + ⋅⋅
⋅
0 1 2 3 4
108
109
-15
-10
-5
0
To
tal r
esist
ance
(Ω cm
2 )
Si3N4 Thickness (nm)
Norm
alize
d Re
lative
Co
ntac
t Res
istan
ce Rc0
(arb
. uni
t)
Figure 5.7 (a) A quick and straightforward extraction method for the relative contact resistance of the diodes with different Si3N4 thicknesses, through measurement of the I-V characteristics on two adjacent electrodes; (b) A minimum contact resistance is observed at the Ag/PTCDA interface with a sandwiched Si3N4 thickness of ~2.5 nm, reaffirming the Fermi-level depinning at M/O interfaces by inserting the ultrathin interfacial Si3N4 layer.
5.3 Theory and Proposed Model
Intuitively, introducing an ultrathin Si3N4 layer in between the metal-organic
semiconductor interface may give rise to a few effects. One is modulation of the organic
film’s localized states distribution close to the interface, i.e. σ as discussed in Chapter 4,
since the Si3N4 layer either protects the film from physical interference of metal
110 CHAPTER 5. CONTACT ENGINEERING FOR ORGANIC ELECTRONIC DEVICE
deposition for top-contact structure (Figure 5.1) or facilitates better film growth on the
metal for bottom-contact structure (Figure 5.6). The other one is modulation of the
charge injection barrier height, i.e. φb. We now proceed to discuss the mechanisms
responsible for this effect, Fermi-level depinning, as briefly mentioned at the beginning
of this chapter.
Figure 5.8 Different mechanisms of Fermi-level pinning at the metal/organic semiconductor interface, including (a) MIGS and intrinsic surface states, (b) charge transfer, (c) covalent bonding, (d) permanent molecular dipole, (e) image force, and (f) exchange/Pauli repulsion. An interface dipole is always created upon the M/O junction formation.
CHAPTER 5. CONTACT ENGINEERING FOR ORGANIC ELECTRONIC DEVICE 111
Classical metal-induced gap states (MIGS) theory [157] has been applied to explain
Fermi-level pinning/depinning at traditional metal-inorganic semiconductor (e.g. Si [158]
or Ge [159]) interface. The concept is simply illustrated in Figure 5.8 (a): free-electron
wave function penetrates into the semiconductor bandgap, giving rise to a large amount
of MIGS which pin the Fermi energy close to the charge neutrality level (CNL) and
form a large Schottky barrier for charge injection. However, the scenario for M/O
interfaces is found more complicated, and a variety of mechanisms as depicted in Figure
5.8 (a)-(f) have been suggested to understand the M/O interfacial electronic structures
[154-156]. The overall effect is an interface dipole (Δ) created upon the formation of the
M/O junction. Note that there can be disorder-induced energy level fluctuations and
Anderson localized states at the M/O interface as discussed in Chapter 4, which are
shown only leading to an effective barrier lowering by σ2/2kT. For simplicity and clarity,
effective HOMO/LUMO levels are denoted in this chapter.
Figure 5.9 (a)-(b) shows a generalized picture of the M/O interface energy band
diagram, indicating that the Fermi-level pinning manifests itself as formation of a
lumped interface dipole with its magnitude depending on the metal. The measured
injection barrier height φb(real) can be considerably larger than the ideal φb(ideal) following
Schottky–Mott limit due to the interface dipole induced energy shift ∆0, and can be
modeled by
0
00
cos
b real b ideal metal metal
i i
i r
U HOMO HOMO
N P dπ
φ φ ψ ψ
θ θ θ θε ε
= + Δ = − ⋅ −
⋅ ⋅+∑∫ (5.1)
where U(x) is the unit step function, θ is the alignment angle between the dipole element
and the interface normal, Ni and Pi are, respectively, the density and moment of the
dipole element type i, corresponding to different dipole origins as shown in Figure 5.8.
Since large charge injection barrier at the source/drain of organic TFTs leads to
significant contact resistance, deteriorating TFT electrical performance and their
112 CHAPTER 5. CONTACT ENGINEERING FOR ORGANIC ELECTRONIC DEVICE
mobility scaling behavior as discussed in Chapter 4, it is crucial to mitigate Fermi-level
pinning effect at such M/O interfaces, and the dipole elements from all aforementioned
mechanisms, Ni and Pi, should be eliminated.
00
( ) ( ) cos /( )i i ri
N P dπ
θ θ θ θ ε εΔ = ∑∫
Figure 5.9 M/O interface energy band diagrams based on the proposed lumped interface dipole model: (a) before M/O interface formation; (b) upon M/O interface formation, Fermi-level pinning arises from various interface dipole elements; (c) after inserting an ultrathin Si3N4 insulator, the Fermi-level depinning takes effect by blocking the physisorption/chemisorption.
CHAPTER 5. CONTACT ENGINEERING FOR ORGANIC ELECTRONIC DEVICE 113
As illustrated in Figure 5.9 (c), an ideal insulator shields
physisorption/chemisorption interaction at the M/O interface and thus releases the
Fermi level pinning. A quasi-zero Schottky barrier or Ohmic contact is therefore
expected under depinning circumstances.
To more quantitatively understand the Fermi-level depinning effect, we propose a
simple depinning model based on the fact that there are two competing mechanisms
responsible for the charge injection current or contact resistance Rc after inserting the
thin insulator. One is direct or Fowler-Nordheim (FN) tunneling through the insulator
with its probability exponentially decreasing with insulator thickness t. The other is
charge injection over the effective Schottky barrier φb(real) which decreases with the
insulator thickness (t) due to the depinning effect. Therefore, assuming roughly that
( ) ( ) 0 exp( / )nb real b ideal ct tφ φ= + Δ ⋅ − (5.2)
where n>1, tc is a critical thickness parameter, the contact resistance thus can be
described as
0 1 ( ) 0 2exp[ ( exp( / ) )] exp( )nc c b ideal cR R t t tβ φ β≅ ⋅ ⋅ + Δ ⋅ − ⋅ ⋅ (5.3)
where β1 and β2 are thermionic emission relevant coefficient and tunneling relevant
coefficient, respectively.
Figure 5.10 shows an exemplified simulation plot for Rc-t based on this depinning
model, which is in good agreement with our experimental results as presented in last
section [see Figure 5.5 and Figure 5.7]. It is also found that the insulator thickness must
be optimized carefully: initial increase of the insulator thickness reduces contact
resistance due to Schottky barrier modulation; once beyond the optimal thickness,
tunneling resistance through the insulator becomes dominated rapidly.
114 CHAPTER 5. CONTACT ENGINEERING FOR ORGANIC ELECTRONIC DEVICE
Figure 5.10 Simulation results based on the proposed simple depinning model showing the dominant mechanisms responsible for the contact resistance behavior after insertion of the ultrathin insulating layer at the M/O interface.
5.4 Summary
We have successfully demonstrated Fermi-level depinning at two different M/O
interfaces by inserting a precisely-controlled ultrathin interfacial Si3N4 layer. The contact
behavior is tuned from rectifying to quasi-Ohmic and to tunneling by modulation of the
CHAPTER 5. CONTACT ENGINEERING FOR ORGANIC ELECTRONIC DEVICE 115
Si3N4 thickness within 0-6 nm. We also proposed a lumped dipole model to clarify the
detailed physical mechanisms of Fermi-level pinning/depinning at the M/O interface.
Experimental results are in good agreement with the theory and the model. This work
represents a significant step toward the fundamental understanding of M/O interface
properties and technological advancement of achieving low-resistance Ohmic contacts
for organic electronic device applications, and can be particularly useful for optimization
of the organic transistor performance through source/drain contact engineering.
116 CHAPTER 5. CONTACT ENGINEERING FOR ORGANIC ELECTRONIC DEVICE
117
Chapter 6
Conclusions and Outlook
Give me a fulcrum, and I shall move the world.
- Archimedes of Syracuse (Ἀρχιμήδης)
In this dissertation, we have systematically studied carbon or organic semiconductor