Polymer Field-Effect Transistors
by
Stamatis Georgakopoulos
Submitted for the degree o f doctor of philosophy
Faculty of Engineering and Physical Sciences
University of Surrey
September 2012
© Stamatis Georgakopoulos 2012
ProQuest N um ber: 27558549
All rights reserved
INFORMATION TO ALL USERS The quality of this reproduction is d e p e n d e n t upon the quality of the copy subm itted.
In the unlikely e v e n t that the author did not send a c o m p le te manuscript and there are missing p a g e s , these will be n oted . Also, if material had to be rem oved,
a n o te will ind icate the deletion .
uestProQuest 27558549
Published by ProQuest LLO (2019). Copyright of the Dissertation is held by the Author.
All rights reserved.This work is protected against unauthorized copying under Title 17, United States C o d e
Microform Edition © ProQuest LLO.
ProQuest LLO.789 East Eisenhower Parkway
P.Q. Box 1346 Ann Arbor, Ml 4 8 1 0 6 - 1346
Contents
Abstract............................................................................................................................................4
1. Introduction.............................................................................................................................. 7
1.1 Conventional Electronics.................................... 7
1.2 Printed Electronics......................................................................................................... 8
1.3 Scope and structure of the thesis................................................................................. 10
2 Conduction in organic semiconductors................................................................................ 14
2.1 Basic electronic system................................................................................................ 14
2.2 Band structure and conduction.....................................................................................16
2.3 Polarisation effects........................................................................................................19
2.3.1 Polarisation energy and inj ection........................................................................20
2.3.2 Molecular polarisation energy and charge transport......................................... 21
2.4 Hopping transport.........................................................................................................23
2.4.1 Gaussian Disorder Model.....................................................................................23
2.4.2 Correlated Disorder M odel..................................................................................24
2.5 Microstructure and electrical performance................................................................ 25
2.6 Ambient stability of organic semiconductors............................................................ 28
3. Organic field-effect transistors.............................................................................................. 33
3.1 Metal-semiconductor contacts.....................................................................................33
3.1.1 Inj ection and Schottky barriers......................................................................... 34
3.1.2 Fermi-Level Pinning............................................................................................ 35
3.1.3 Metal-semiconductor surface dipole manipulation............................................36
3.1.4 Ohmic contacts in OFETs.............................................................. 37
3.2 Field-Effect Transistors...............................................................................................39
3.2.1 Semiconductor-insulator interface...................................................................... 41
3.3.2 V oltage gain................................................................................ 42
3.3 Electrode configurations.............................................................................................. 43
4. Transistor Fabrication......................................................................................................47
4.1 Introduction....................................................................................................................47
4.2 Substrate cleaning......................................................................................................... 47
4.3 Thickness and surface characterisation of polymer thin films.................................. 48
4.4 Metal thin film deposition.............................................................................................48
4.5 Metal patterning............................................................................................................ 49
4.6 SAM deposition.............................................................................................................51
4.7 Kelvin Probe studies.................................................................................................... 52
4.8 Schottky barrier engineering........................................................................................55
4.9 Semiconductor.............................................................................................................. 55
4.10 Insulator.................................................................................................................... 58
5 Stability and performance of Top- and Bottom-Gate FETs.............................................. 61
5.1 Introduction................................................................................................................... 61
5.2 Fabrication and methods.............................................................................................. 61
5.3 Performance comparison: TG and BG PFETs........................................................... 62
5.4 Dual-Gate PFETs..........................................................................................................64
5.5 DG PFET Ambient Stability........................................................................................68
5.6 Conclusions..................................................................................................................70
6. Morphology and charge transport of amorphous conjugated polymers...........................73
6.1 Introduction................................................................................................................... 73
6.1 Fabrication and Methods............................................................................................. 74
6.2 Microstructure assessment.......................................................................................... 75
6.3 FET Characteristics, contact quality and resistance...................................................78
6.4 Gate voltage dependence of mobility..........................................................................82
6.5 Field and Temperature dependence of mobility..........................................................84
6.6 Gaussian Disorder Model fitting................................................................................. 86
6.7 Conclusions...................................................................................................................90
7. Polymer Source-Gated Transistors..................................................................................... 94
7.1 Limitations of the Field-Effect Transistor................................................................... 94
7.2 Source-Gated Transistors............................................................................................. 95
7.3 Polymer Source-Gated Transistor fabrication and methods......................................97
7.4 Polymer Source-Gated Transistor performance..........................................................98
7.5 Conclusions....................... 106
8. Conclusions..........................................................................................................................109
8.1 Summary......................................................................................................................109
8.2 Future work................................................................................................................. 115
8.3 List of Publications......................................................................................................116
8.4 List of oral and poster presentations......................................................................... 117
Abstract
High Ionisation Potential (IP) amorphous conjugated polymers are very practical
semiconductors and promising candidates for printing applications as they exhibit 1) high
air-stability due to the high IP, and 2) reproducible electrical performance due to the
uniformity of amorphous morphology. However they generally exhibit low mobilities on the
order of 10' cm W s and below. This work is based mainly on two high-IP amorphous
conjugated polymers poly(indenofluorene-triarylamine) (PIFTAA) and poly(indenofluorene-
phenanthrene) (PIFPA).
The long term ambient stability of PIFTAA and PIFPA with IPs of 5.45 eV and 5.79
eV respectively is characterised in Field-Effect Transistors (FETs) over a period of 4 and 2
months respectively. FET parameters such as the tum-on voltage and subthreshold slope are
found to be generally stable, and the charge carrier mobility is found to degrade at an
approximate rate of 10% per month, which is amongst the lowest reported values for organic
semiconductors.
PIFTAA and particularly PIFPA exhibit high field-effect saturation mobilities of
0.03 - 0.04 cm W s and 0.2 - 0.3 cm W s respectively, which are unusually high for
amorphous conjugated polymers. The morphologies are examined by atomic force
microscopy, grazing incidence wide angle x-ray scattering, and differential scanning
calorimetry, and no evidence of crystallinity is detected, suggesting that the conjugated
polymers are indeed amorphous.
To investigate charge transport in PIFTAA and PIFPA, FETs of multiple channel
lengths are fabricated, providing mobility data for multiple electric fields, and measured over
a range of temperatures. In addition to PIFTAA and PIFPA, the measurements are performed
on typical amorphous conjugated polymers poly(triarylamine) (PTAA) and
poly(indenofluorene-triarylamine-triarylamine) (PIFTAATAA), with mobilities of 0.003
cm W s and 0.004 cm W s respectively. The gate voltage dependence of the mobility
extracted from FET measurements, as well as the 1/T fit of the mobility with temperature is
consistent with a Gaussian Density of States. The indenofluorene copolymers PIFTAA,
PIFTAATAA, and PIFPA exhibit clear negative electric field dependence of the mobility,
signature of high spatial disorder in the polymer films. The temperature dependence of the
mobility is fed into the Gaussian Disorder Model, which indicates that the source of the high
mobility for PIFPA is mainly strong intermolecular coupling indicated by the high pre-factor
mobility as well as low energetic disorder along the path of charge flow. These results
challenge the widely accepted concept that high crystallinity is a requirement for mobility
exceeding 0.1 cm W s in organic semiconductors.
Finally, a new type of contact-limited transistor is demonstrated with conjugated
polymers. Source-Gated Transistors (SGTs) have a similar structure to FETs, and the main
difference is the Schottky source/drain-semiconductor contact, which results in depletion of
the near-source region of the semiconductor. Consequently, the behaviour of the transistor
changes significantly as compared to FETs. SGTs are demonstrated with several electrode
conjugated polymer combinations. SGTs saturate at significantly lower voltages than FETs,
and saturation is not lost for short channels and thick insulators. Also, evidence of
independence of the current from channel length is observed, which is consistent with
contact-based modulation as opposed to FET channel-based modulation. These advantages
come at a cost of output current of at least one order of magnitude, while the intrinsic voltage
gain is mostly maintained.
DeclarationThis thesis is submitted for the degree o f Doctor o f Philosophy at the University o f
Surrey. This thesis and the work to which it refers are a result o f m y own efforts.
Any ideas, data images or text resulting from the work o f others (whether published
or unpublished) are fully identified as such within the work and attributed to their
originator in the text, bibliography or in footnotes. This thesis has not been submitted
in whole or in part for any other academic degree or professional qualification.
Stamatis Georgakopoulos
September 2012
1. Introduction
1.1 Conventional Electronics
Since the invention of the first integrated circuit (IC) in 1958 by Jack Kilby, the
fabrication of electronic devices has been largely based on silicon wafer chips. Electronic
devices require three types of materials; semiconductors, insulators, and conductors. The
high charge carrier mobility of the silicon semiconductor, the ease of formation of high
quality silicon oxide insulator by thermal annealing, and the abundance of silicon in the
earth’s crust, fuelled the explosive expansion of the silicon IC industry.
The production of IC chips is based on a number of sophisticated techniques. A large
silicon ingot is crystallized from a melt in the Czochralski process, followed by cutting into
wafers and polishing. Ion implantation doping utilizes ion beams to dope specific areas on
the wafer chip with electron donors/acceptors, and fonn electron or hole conductors, or shift
the Fermi level in a semiconductor, and requires high temperature processing up to 1000°C
to activate the dopants. The formation of silicon oxide insulator requires thermal annealing at
very high temperatures of up to 1200°C. Theraial evaporation or sputtering of metals are
vacuum-based techniques that are utilized along with high resolution photolithography or
electron-beam lithography to pattern component contacts or interconnects. Chemical vapour
deposition is another vacuum-based technique used to deposit additional semiconducting
layers.
The production of silicon ICs is based on complicated high-cost processes that
require vacuum processing, high temperatures, and sophisticated tools, giving rise to a large
entry barrier into the IC production industry. In turn, after several decades of miniaturisation,
the performance of ICs has reached very high levels and allowed the creation of numerous
powerful devices including computers and smart phones.
Figure 1.1. Intel Xeon processor dies on a silicon wafer [1].
1.2 Printed Electronics
Over the last 15 years, the concept of printed plastic electronics based on
straightforward fabrication processes for the production of electronics has been gaining
momentum. Considering production processes outside the field of electronics, one of the
most convenient techniques for the additive deposition of components is printing. The
production of newspapers is a low-cost, quick, large-area, very high yield process that allows
the end product to be sold at low prices. Regardless of the different types of printing
techniques, the main feature is the deposition of inks. Therefore, to combine printing with
electronics, functional electronic inks are required. Such inks can indeed be produced mainly
from dissolvable organic polymers, nanoparticles, nanowires, and nanotubes.
Electronic inks have the potential to be deposited and patterned by a variety of
techniques with the potential of low-cost, large-area, low-temperature, ambient air based
processing, on a variety of substrates including flexible plastic substrates. Such techniques
include gravure printing [2], inkjet printing [3], screen printing [4], self-assembly [5], dip-
coating [6], laser writing [7], and imprint lithography [8,9] (Fig. 1.2).
Figure 1.2. Imprint lithography based flexible display printed by Arizona state university and Hewlett Packard [10].
The most important figure of merit is the charge carrier mobility, which is a measure
of the velocity of charge carriers in the material. The charge caiTier mobility of
polycrystalline silicon can be up to few hundreds of cm'/Vs [11] and is partly responsible for
its widespread use. The mobility of solution-processable semiconductors tends to be
significantly lower, with high perfonnance polymers exhibiting maximum mobilities of
approximately 1 - 3 cim/Vs [12,13,14]. Higher mobilities can be attained by sacrificing
convenience, such as high-temperature spray-cast zinc oxide based films with mobilities of
up to 50 cm^/Vs [15], or vacuum grown rubrene crystals with mobilities of up to 40 cmVVs
[16].
Although it is clear that solution-processable electronics will not compete with
conventional single crystal wafer semiconductor based electronics in terms of performance,
not all electronics applications require high mobilities. Such applications include light-
emitting diodes [17], light-emitting transistors [18], photovoltaics [19], RFID tags [20],
chemical [21] and pressure sensors [22] (Fig. 1.3), and transistor pixel drivers for
electrophoretic [23] and active matrix displays [24].
Figure 1.3. Artificial skin (pressure sensor) from [22].
The advantage of solution-processable electronics is the practicality of device
fabrication. The most suitable candidates in tenns of semiconductor deposition are
conjugated polymers. Deposition can occur from solution, at room temperature, in ambient
conditions, over large area substrates, flexible substrates, the polymer coatings can be
transparent and/or be light-emitting. Although their mobility is not as high as for small
molecules and nanomaterials, 0.1 cm^/Vs is considered the benchmark for numerous
applications [25], and as such, various commercial exploitations are possible.
1.3 Scope and structure of the thesis
Electronic conduction along conjugated polymer chains is based on delocalised n electron clouds, and the bulk conductivity is determined by intermolecular n-n stacking.
High charge carrier mobility polymers usually exhibit extensive n-n stacking in the form of
crystal domains.
Conjugated polymers have the potential for practical fabrication. However, their
commercial applications are limited by two issues. Firstly, the electrical performance in
ambient atmosphere is unstable due to moisture and oxygen, which is mostly resolved by
utilising hydrophobic insulators in FET structures and by using semiconductors having high
Ionisation Potential (IP). Secondly, the morphology and thus electrical performance of
polycrystalline conjugated polymers is affected by a variety of processing conditions,
causing low reproducibility. Amorphous conjugated polymers have more uniform
morphologies and thus more reproducible performance, but lower charge carrier mobilities.
To address the above issues, this thesis focuses on Field-Effect Transistors (FETs)
based on a class of conjugated polymers that have both high IP and amorphous
morphologies, and are therefore excellent candidates for printed electronics, even though the
FET structures in this work are not produced by printing. Charge carrier mobility of 0.3
cm^A/s is demonstrated with amorphous high IP indenofluorene-phenanthrene copolymer,
that is comparable to performance of amorphous silicon transistors widely used in liquid
crystal displays, and uncommonly high for a conjugated polymer with amorphous
morphology.
Chapter 2 gives a description of electronic conduction in organic semiconductors,
the models that are used to characterise charge transport in disordered conjugated polymers,
the correlation between crystallinity and electrical performance and reproducibility, and the
stability of organic semiconductors in ambient atmosphere.
Chapter 3 describes organic FET structure and operation principles, and focuses on
their two main constituents: electrode-semiconductor contacts, and the transistor channel
along the semiconductor-insulator interface. FET parameters and extraction methods are also
included.
Chapter 4 describes the experimental steps and materials involved in transistor
fabrication, and includes a study of electrode workfunctions based on Kelvin probe
measurements.
Chapter 5 presents a study of FET performance and long term ambient stability of
top gate and bottom gate FET configurations fabricated with high IP amorphous polymers.
10
The study is based on fabricated dual-gate FET structures characterised over a period of 1 to
4 months.
Chapter 6 focuses on morphology and charge transport of four amorphous
conjugated polymers, including the indenofluorene-phenanthrene copolymer which is of
particular interest. The morphology study is based on atomic force microscopy, grazing
incidence wide angle x-ray scattering, and differential scanning calorimetry. The charge
transport study is based on the gate voltage, electric field, and temperature dependence of the
charge carrier mobility measured in FETs. The FET data is fed into the Gaussian disorder
model to gain insights on the correlation between mobility and morphology.
Chapter 7 demonstrates a type of contact limited transistor based on Schottky
electrode-semiconductor contacts. This type of transistor opens a window of exploitation for
very high IP conjugated polymers that present challenge for establishing Ohmic contacts.
Most importantly, Source-Gated Transistors (SGTs) behave distinctively differently than
FETs, exhibiting improvement in certain aspects of transistor performance as compared to
FETs. SGTs are demonstrated with several different conjugated polymer semiconductors and
electrodes.
Chapter 8 presents the thesis summary and conclusions, as well as discussion for
11
* URL: http://www.intel.com/pressroom/archive/releases/2010/2010033Ocomp_sm.htm [Accessed 26/09/2012] M. Hambseh, K. Renter, M. Stanel, G. Schmidt, H. Kempa, U. Fugmann, U. Hahn, A.C. Rubier.
“Uniformity of folly gravure printed organic field-effect transistors”. Mat. Sci. Eng. B 170, 93 (2010) H. Sirringhaus, T. Kawase, R.H. Friend, T. Shimoda, M. Inbasekaran, W. Wu, E.P. Woo. “High-
resolution inkjet printing of all-polymer transistor eircuits”, Sci. 290,2123 (2000)D.A. Pardo, G.E. Jabbour, N. Peyghambarian. “Application of screen printing in the fabrication of
organic light-emitting devices”. Adv. Mater. 12, 1249 (2000) J. Collet, O. Tharaud, A. Chapoton, D. Vuillaume. “Low-voltage, 30 nm channel length, organie
transistors with a self-assembled monolayer as gate insulating films”, Appl. Phys. Lett. 76, 1941 (2000)
J. Ghim, K.J. Baeg, Y.Y. Nob, S.J. Kang, J. Jo, D.Y. Kim, S. Cho, J. Yuen, K. Lee, A.J. Heeger. “Perfiuorocyclobutane containing polymeric gate dielectric for organic thin film transistors with high on/off ratio”, Appl. Phys. Lett. 89, 202516 (2006) M. Mas-Torrent, E.E. Laukhina, V. Laukin, C.M. Creely, D.V. Petrov, C. Rovira, J. Veciana. “Direct
micro-patterning of TTF-based organic conductors on flexible substrates”, J. Mater. Chem. 16, 543(2006) J.Z. Wang, J. Gu, F. Zenhausem, H. Sirringhaus. “Low-cost fabrication of submicron all polymer
field effect transistors”, Appl. Phys. Lett. 88, 133502 (2006) S.H. Ahn, L.J. Guo. “High-speed roll-to-roll nanoimprint lithography on flexible plastic substrates”.
Adv. Mater. 20,2044 (2008)URL: http://www.engadget.eom/2008/12/08/hp-and-asu-demo-bendable-unbreakable-electronic-
displays/ [Accessed 26/09/2012]H.S. Lee. “The field effect electron mobility of laser-annealed polycrystalline silicon MOSFETs”,
Solid State Electron. 24, 1059 (1981)S.K. Park, T.N. Jackson, J.E. Anthony, D.A. Mourey. “High mobility solution processed 6,13-
bis(triisopropyl-silylethynyl) pentacene organic thin film transistors”, Appl. Phys. Lett. 91, 063514 (2007)
D.J. Gundlach, I. McCulloch, M. Heeney. “Undoped polythiophene field-effect transistors with mobility of 1 cmV'^s"'”, Appl. Phys. Lett. 91, 243512 (2007)
T. Lei, J. H. Don, J. Pei. “Influence of alkyl chain branching positions on the hole mobilities of polymer thin-film transistors”. Adv. Mater. DOI: 10.1002/adma.201202689 (2012)
G. Adamopoulos, A. Bashir, S. Thomas, W.P. Gillin, S. Georgakopoulos, M. Shkunov, M.A. Baklar, N. Stingelin, R.C. Maher, L.F. Cohen, D.D.C. Bradley, T.D. Anthopoulos. “Spray-deposited Li-doped ZnO transistors with electron mobility exceeding 50 cm2/Vs”, Adv. Mater. 22, 4764 (2010)
T. Hasegawa, J. Takeya. “Organic field-effect transistors using single crystals”, Sci. Technol. Adv. Mater. 10, 024314 (2009)
B. Geffroy, P. Le Roy, C. Prat. “Organic light-emitting diode (OLED) technology: materials, devices and display technologies”, Polym. Int. 55, 572 (2006)
12
M. Muccini. “A bright future for organic field-effect transistors”, Nat. Mater. 5, 605 (2006)D. Wohrle, D. Meissner. “Organic solar cells”. Adv. Mater. 3, 129 (2004)
Subramanian, J.M.J. Frechet, P.O. Chang, D.C. Huang, J.B. Lee, S.E. Molesa, A.R. Murphy,D.R. Redinger, S.K. Volkman. “Progress toward development of all-printed RFID tags: Materials, processes, and deviees”, IEEE Proe. 93, 1330 (2005)
J.T. Mabeck, G.G. Malliaras. “Chemical and biological sensors based on organic thin-film transistors”. Anal. Bioanal. Chem. 384, 343 (2006)^ T. Someya, T. Sekitani, S. Iba, Y. Kato. “A large-area, flexible pressure sensor matrix with organic field-effect transistors for artificial skin applications”, Proc. Nat. Acad. Sci. 101, 9966 (2004)^ G.H. Gelinck, H.E.A. Huitema, E. Van Veenendaal, E. Cantatore, L. Schrijnemakers, J.B.P.H. van der Putten, T.C.T. Geuns, M. Beenhakkers, J.B. Giesbers, B.H. Huisman, E.J. Meijer, E.M. Benito, F.J. Touwslager, A.W. Marsman, B.J.E. van Rens, D.M. de Leeuw. “Flexible active-matrix displays and shift registers based on solution-processed organic transistors”, Nat. Mater. 3, 106 (2004)
L. Zhou, A. Wanga, S.C. Wu, J. Sun, S. Park, T.N. Jackson. “All-organic active matrix flexible display”, Appl. Phys. Lett 88, 083502 (2006)
H. Sirringhaus, “Materials and applieations for solution-processed organic field effect transistors”, IEEE Proc. 97, 1570 (2009)
13
2 Conduction in organic semiconductors
2.1 Basic electronic system
This section is partly adapted from [26,27].
Electrons orbiting nuclei occupy discrete energetic states. These states are classified
into shells (K, L, M), which are subdivided to subshells (s, p, d), which consist of one or
multiple orbitals according to the angular momentum of electrons (Is, 2s, 2px, 2py, 2pz).
According to Pauli’s principle each orbital can be filled with a maximum of two electrons of
opposing spins.
The electronic and chemicals properties of atoms are primarily determined by the
outer shell. The outer shell of the carbon atom consists of 2s2px2py2pz orbitals which are
occupied by 4 electrons (Fig. 2.1). When carbon forms chemical bonds, it undergoes a
process called hybridisation, resulting in the formation of a number of hybrid orbitals of
energy between the original 2s and 2p. Hybridisation in carbon can occur between the 2s
orbital and one to three 2p orbitals (sp, sp , sp^). We will focus on sp as it is the only useful
scheme for efficient electronic conduction. Sp hybridisation is common in organic
molecules and results in the formation of three sp orbitals, while the pz orbital remains
unaffected (Fig. 2.1).
Ground state carbon sp carbon
>-S?c
LU
± ± _ 25 vl't 2Py 2Pz
K Is K
f 't' f _Lsp2 sp2 sp2
Figure 2.1. Shell and subshell configuration of ground state carbon and sp hybridised carbon that occurs upon bonding. Other types of hybridisation are possible (sp, sp ) but not considered.
The sp orbitals of neighbouring carbon atoms overlap and interact to form o-bonds,
in which the electrons are shared between the carbon atoms and are strongly bound on the
intemuclear plane (dashed lines in Fig. 2.2). The pz orbitals form Ti-bonds, the axis of which
is vertical to the intemuclear plane (Fig. 2.2). Two energy levels are generated, one below
the initial energy {n bonding orbital), and one above the initial energy (ti* anti-bonding
orbital) (Fig. 2.3). These molecular orbitals originate from in-phase (bonding) and out-of-
phase (anti-bonding) interactions of the atomic orbitals.
14
H H
C C
H
H
H
Figure 2.2. Illustration of bonding and anti-bonding molecular orbital formation in sp carbon atoms (left) and electron density for n and n* in ethylene (right).
The electron concentration in 71-bonds lies further from the nucleus as compared to
o-bonds, hence Tc-bonds are weaker than o-bonds. Double bonds consist of one o- and one 71-
bond. In 7i:-conjugated molecules, namely systems of alternating single and double bonds, k- electrons are weakly bound and can easily move from one site to the other. This mechanism
allows charge to délocalisé over large parts of the molecule and is the basis of conduction for
organic semiconductors and conductors.
When multiple molecules are in close proximity, bonding and anti-bonding
interactions cause splitting of the molecular orbitals [28] (Fig. 2.3). Charge conduction
occurs at the Lowest Unoccupied Molecular Orbital (LUMO) level for electrons (negative
charge) and at the Highest Occupied Molecular Orbital (HOMO) for holes (positive charge).
15
LtlMO
HOMO
Figure 2.3. Bonding-antibonding interactions between two ethylene molecules (left) and molecular orbital splitting for multiple stacked ethylene molecules (right) from [28].
2.2 Band structure and conduction
This section is partly adapted from [29].
The situation is different in classical inorganic semiconductors, where the molecular
orbitals form energetically continuous bands, thus electron conduction occurs along the
conduction band and hole conduction occurs along the valence band. The general principles
of inorganic semiconductors apply to some extent to organic semiconductors.
In the ground state, intrinsic (undoped) semiconductors at zero temperature, the
valence band is fully occupied with electrons and the conduction band fully depleted of
electrons (fully occupied by holes) (Fig. 2.4b). The energetic difference between the
conduction band edge and the valence band edge (or LUMO and HOMO in organic
semiconductors respectively) is the band gap Eq, which contains no states to be occupied by
charge carriers. For charge carriers to flow there must be empty states to move into, and
since the valence (conduction) band is full of electrons (holes), electron (hole) flow is only
possible following excitation over the band gap and into the conduction (valence) band.
16
(b) Intrinsic (c) Intrinsic Sem iconductor Sem iconductor (d) n -doped
(a) Conductor y_QK y>oK Sem iconductor (e) Insulator Vacuum Level
a ô D o p a n t Level
EAC onduction Band
IP
- - V alence Band
Core Levels
Figure 2.4. Band structure for different types of electronic materials. EA is the electron affinity (vacuum level to LUMO or conduction band onset). IP is the ionisation potential (vacuum level to HOMO or valence band onset). Eg is the band gap and Ep is the Fermi level (red dashed lines).
As such, the band gap largely determines the electronic properties of materials. In
conductors there is no band gap (conduction and valence band overlap. Fig. 2.4a). In
insulators the band gap is too large to allow the exeitation of charge earriers (Fig. 2.4e). In
semiconductors the band gap has moderate values (Fig. 2.4b) and excitation can be aehieved
thermally (Fig. 2.4e), or optieally (by absorbing a photon of energy equal or higher to the
band gap - not shown). Alternatively, electrons (holes) can be injeeted straight into the
conduction (valence) band from a material with higher eoncentrations (metal or doped
semiconductor), given appropriate energy level alignment (covered in Chapter 3). Intrinsic
classical semiconductors have relatively small band gaps, translating into relatively high
carrier concentrations (i-Si 1.12 eV, i-Ge 0.68 eV, i-GaAs 1.42 eV [29,30]), while
conjugated polymers have band gaps typically in the range of 2-3 eV [31,32].
Additionally, the eleetronic properties ean be altered by doping. Doped
semiconductors are semieonductors that eontain impurities which may be unintentional,
resulting from lack of control during processing, or intentional, added on purpose to provide
free carriers. Each silicon atom in a silicon lattice contains 4 outer shell electrons which are
shared with four neighbouring silicon atoms (covalent bonding), energetically these are
valenee band electrons. Silicon is an intrinsic (undoped) semiconductor and the
concentration of electrons in the conduction band equals the coneentration of holes in the
valence band. In the presenee of phosphorus atoms (5 outer shell eleetrons per atom) in the
silicon lattice, 4 of the outer shell electrons of the phosphorus atom will be used up in
bonding (valence band electrons), while the fifth electron will remain unbound and highly
energetic (near-conduction band electron energy). If the energy of the unbound electrons is
within thermal energy kT of the conduction band edge, the dopant atoms are ionised and the
unbound electrons populate the conduction band (Fig. 2.4d). Phosphorus is an electron donor
and phosphorus-doped silieon has a higher number of electrons in the conduction band than
17
holes in the valence band (Fermi level near conduction band), and is referred to as n-type
silicon. Similarly, implanting boron atoms (3 outer shell electrons per atom) into silieon will
populate the valence band with holes (missing eleetrons from each silicon-boron bond).
Boron is an electron acceptor and boron-doped silicon has a higher number of holes in the
valence band than electrons in the conduction band (Fermi level near valenee band), and is
referred to as p-type silicon. Quantitative description of doping is beyond the seope of this
work. More information ean be found in [29].
Intentional doping is not utilised with organic semiconductors in transistor
applications, which are generally intrinsic. However, unintentional doping can be a concern
with organic semiconductors and is caused by atmospheric species, mainly oxygen, which
causes the removal of electrons from organic molecules or polymer ehains (p-doping) [33]
(see Section 2.6). The susceptibility to oxidation generally decreases with increases
ionisation potential of the organie semiconductor (Section 5.5).
At non-zero temperatures electrons behave somewhat like boiling water, i.e. there is
a temperature dependent probability that some electrons will be shot upwards and overcome
the potential barrier (such as the band gap). Charge carrier energetic distribution is
represented by the Fermi level Ep, whieh is the level at which carrier occupation is equal to
0.5 (50% of states are full). The probability that a charge carrier will overeome a potential
barrier E-Ep given thermal energy kT (where T is the absolute temperature and k is
Boltzmann’s constant) is given by:
F(E) = ---------------------------------- (Eq. 2.1)l + eq)[(E-Ep)/kT]
The amount of electrons at an energy E is given by the product of the density of the
Density Of States (DOS) N(E) and the occupation probability F(E). The density of eleetrons
n or holes p in the eonduetion and valence band respeetively is obtained by integrating for all
band energies (where Ec and Ey are the band onsets):
(•00
n = f Nc(E)F(E)dE (Eq. 2.2)JBc
p = N V (E)F(E)dE (Eq. 2.3)
When the bands are populated, current flow is triggered by two mechanisms: drift
under an applied electric field, and diffusion between regions of different eharge densities.
Under an applied voltage V, eharge flow between two electrodes with spacing L in transit
time t will occur with a drift velocity Vd=L/t at an electric field F=V/L. As the drift velocity
18
is generally proportional to the applied field, the velocity of charge carriers is charaeterised
by the mobility parameter p:
V(JH = f = - (Eq.2.4)
The net current density J consists of the sum of the eontributions from both eleetrons
and holes, J = Jn+Jp, which include both the drift eurrent (first term) and diffusion current
(second term):
Jn = qftnnF + qD„ ^ (Eq. 2.5)dx
Jp = qftpPF- qDp — (Eq. 2.6)dx
In the above equations q is the elementary charge, pn and pp are the electron and hole
mobilities, D„ and Dp are the Einstein diffusion constants for electrons and holes
respectively.
In summary, electrieal conduetivity is strongly affeeted by the charge carrier
mobilities pn and pp and the density of earriers whieh is proportional to the DOS N(E). N(E)
applies over the entire band, however only the bottom of the conduction band (LUMO in
organics) and top of the valence band (HOMO) are occupied, and the DOS in those energy
ranges is the effective DOS. While in classieal semiconductors approximations have been
developed for the effeetive DOS relating it to the earrier effective mass and other
semiconductor properties [29], the picture in organics is not as clear and the effective DOS is
determined semi-empirically.
2.3 Polarisation effects
This section is partly adapted from [34,35].
The time required for a charge carrier to polarise its surroundings upon arrival to
a site, is a function of the band gap and does not differ significantly between inorganie and
organie semiconductors. However, the carrier residenee time at a particular site, is an inverse
funetion of bandwidth, i.e. energetic overlap between neighbouring sites (adjaeent atoms in a
silicon lattice or moleeules in a conjugated organic solid). While the bandwidths are on the
order of several eV for classical inorganic semiconductors, the typieal bandwidth for organic
semiconductors are on the order of 0.2 eV [28], resulting in a significant difference between
19
the two. While in inorganic semiconductors the residence time of charge earriers is too short
for polarisation effeets to occur, in organic semiconductors the long residenee time results in
significant polarisation effects.
There are different sources of polarisation. Molecular polarisation concerns the
displacement of the nuclei on the molecule where the eharge resides. Lattice polarisation
involves movements of the entire lattice, however its effect is negligible [36]. Electronic
polarisation stems from the distortion of the delocalised electron clouds of nearby molecules.
2.3.1 Polarisation energy and injection
The energy dissipated to polarisation upon injection of charge into an organic
semiconductor includes both the eleetronie and moleeular polarisation energies. This is
illustrated for an arrangement of benzene rings in Fig. 2.5. Suppose that a Ti-electron is
withdrawn from the eentral molecule (a hole is injeeted into the bulk) (Fig. 2.5b). With
respect to the vacuum energy level, if these moleeules initially had HOMO energy of Ehomo,
the recently injected hole will have energy of Ehomo - Ep+ (Fig. 2.5e), where the second term
is the positive polarisation energy. Similarly an injected electron would have energy of
Elumo + Ep.. The polarisation energy ineludes eontributions from both the molecular and
electronic contributions Epm+ and Epe+ respectively, so the injected hole energy would be
Ehomo - EpM+ - EpE+. Ep+ values for conjugated polymers have been reported in the range of
0 .4 -0 .7 eV (Section 3.1.2).
/ X iu \ /M V acuum Level
, ^ , 0 0 0 0 0 ------------------
Ü 0 O O Oo < D # o aQ Q Q Q QQ Q Q Q Q HOMO
Figure 2.5. (a) An arrangement of benzene rings and their delocalised n-electrons (rings), and(b) the resulting polarisation upon injection of a hole, (c) Energy band diagram showing the energy loss of holes and electrons due to polarisation.
20
2.3.2 Molecular polarisation energy and charge transport
EpM+ has been estimated by quantum chemical calculations for oligoacenes and other
oligomers: 145 meV for TPD, 93 meV for naphthalene (2 fused benzene rings), 68 meV for
anthracene (3 rings), 56 meV for tetracene (4 rings), 48 meV for pentacene (5 rings) [37].
Calculations for long molecules (polymers) are computationally intensive and data is not
available. The molecular polarisation energy in oligomers calculated from theory exhibits
inverse proportionality to molecular length, and saturates beyond the localisation length of
the polarisation, however results vary significantly depending on which model is used [38].
Intermolecular charge transfer should not be appreciably affected by the electronic
polarisation energy and should be dominated by the molecular polarisation energy. Consider
two adjacent molecules with some overlap of the 7i-electron clouds of energetic width t (the
intermolecular charge transfer integral) (Fig. 2.6b). The energy variation during charge
transfer would be t - Epm+. In theory, the parameter t - Epm+ is crucial and defines the regime
of charge transport in the semiconductor [34]. For t > Epm+, the polaron is delocalised and
will flow in band-like behaviour with thermally deactivated mobility (lattice phonon
scattering limited). Such behaviour has been experimentally observed for pentacene single
crystals, polycrystalline thin films, and TIPS pentacene thin films at low temperatures
[39,40]. For t < Fpm+ the polaron is localised and transport occurs by phonon-assisted
hopping between the localised states, which occurs in most organic semiconductors.
(b)a
PM +
Figure 2.6. (a) Variation in hole energy due to molecular polarisation during intermolecular transfer, (b) partial overlap of Tr-electron clouds of neighbouring molecules and charge transfer integral t.
An extensive review can be found in [37] on efforts to calculate t and Fpm+ from
theory, however there is no direct experimental technique for the measurement of these
parameters and thus are used in theoretical context.
21
The intermolecular charge transfer integral t is often estimated as a fraction of the
effective HOMO/LUMO splitting W, typically -0.1 eV for oligomers, depending on the
model used. For two co-facial molecules W = 2t, while for an infinite one dimensional stack
of molecules W = 4t [37]. The charge transfer integral t varies significantly with
intermolecular packing parameters such as distance and relative displacement of the
molecules (Fig. 2.7) [28,37], and thus depends on the packing scheme of every different
molecular system.
|— HOMO -D-LUMO
— 0.15
I 0.10W 0.4
3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0 Distance (A)
6 8 10 12 14 16 18 20Shift (A)
Figure 2.7. Left: Calculated electronic splitting for HOMO/LUMO of two sexithiophene conjugated oligomers as a function of intermolecular distance. Right: Splitting as a function of shift for one of the two molecules from [28].
The intenuolecular electron transfer rate is often calculated with Marcus theory
that was initially developed to describe electron transfer from a donor to an acceptor in
solution (polarisable medium) [41]:
X / 4 - tk^T = Aexp
kT(Eq. 2.7)
Where X is the reorganisation energy which is defined as twice the molecular
polarisation energy, and A is a constant dependent on the frequency of an electron
overcoming the energy barrier that was later defined to account for tunnelling effects at low
temperatures [42]:
.22n t(Eq. 2.8)
A connection between kgy and the charge carrier mobility can be established by [43]:
2(Eq. 2.9)
_ q akT
Where a is the lattice constant of the material.
22
Marcus theory approaches are ideal for describing charge transfer on the molecular
level and establishing correlations between electronic conduction and intermolecular
configurations, but spatial and energetic disorder are not accounted for and thus the predicted
thermal activation behaviour of the mobility significantly deviates from experimental data
for disordered conjugated polymers [44]. Hybrid approaches have been taken to incorporate
disorder but are beyond to scope of this work [45,46].
2.4 Hopping transport
An alternate approach is based on tunnelling-like transfer between localised sites
(from i to j), and predicts a hopping rate yÿ with an attempt-to-escape frequency yo, inter-site
distance R, charge delocalisation radius Ro, and energies Ej and Ej respectively:
_ m E j - E ;
Yijy e kT forEj>Ei
2R (Eq.2 .10)
Yoe f o r E j < E i
This is the Miller-Abrahams formula [47]. This model accounts for two crucial
parameters: inter-site distance (relative to the delocalisation radius), and energetic difference
(which should describe localised polaron hopping as well).
2.4.1 Gaussian Disorder Model
Charge transport in conjugated organics mainly occurs by phonon-assisted hopping
between localised states. On the molecular level, the hopping rate depends on inter-site
distance and energy difference (Miller-Abrahams). On a higher level, the Gaussian Disorder
Model (GDM) proposed by Bassler [48] assumes a Gaussian distribution of site energies
g(E) given by:
g (E ) = - 7 = e x p ■\l27Jn V 2 a T
(Eq. 2.11)
The effective DOS g(E) is characterised by the width parameter o, which takes
experimental values in the range of 50-150 meV [46] and is related to the dipole moment and
polarisability of the semiconductor [49] (or of the semiconductor-insulator interface in
23
transistors). Additionally, the GDM assumes a distribution of intersite distances that is
characterised by the dimensionless parameter Z (positional disorder). This parameter is not
physically well-founded and cannot be directly related to structural of molecular properties
[46]. The pre-factor mobility po is an indicator of intermolecular coupling [50].
The GDM also accounts for the electric field dependence of the mobility by
assuming that an appliedelectric field will tilt the DOS, lowering the activation energy for
charge transport. Poole-Frenkel dependence of the mobility p on electric field F is commonly
observed in organic semiconductors in the following form:
h l|A «V F (Eq.2.12)
In the GDM, the mobility as a fimction of temperature and field p(T,F) is given by:
exp<
p(T,F) = po exp3kT
cVr
exp cVf
kT- r forE>L5
a -2 .2 5 forZ<1.5
(Eq. 2.13)Where C is a constant dependent on inter-site separation and for a typical hopping
distance of 0.6 nm takes a value of C = 2.9-10"^ cmV^^ . An important weakness of the GDM
is that the field dependence of the mobility is not reproduced at low electric fields on the
order of 10 Vcm"\ which are on the high end of experimentally obtainable fields from
transistor data. However, even without the last term of the equation that describes the field
dependence, the main parameters (po, a) can be still be extracted [51], while the polarity of
the field dependence can be used as a qualitative indicator.
2.4.2 Correlated Disorder Model
In the GDM, the energies of transport sites are assumed to be uncorrelated. The
model was later modified on the basis that a site will influence the energies of neighbouring
sites, leading to an extended range for the mobility field dependence down to 10' Vcm'^
[52,53]. The Correlated Disorder Model (CDM) gives the mobility as:
3
p(T,F) = poexp5kT
exp< C o J ^V a
- r (Eq. 2.14)
Where R is the hopping distance and F is a parameter similar to S in the GDM.
24
In the CDM, it is also possible to correlate the energetic disorder with microscopic properties
such as the dipole moment [52].
2.5 Microstructure and electrical performance
In most organic semiconductors, 7i-stacking results in the formation of periodic
structures (crystal domains). As Ti-stacking occurs along one dimension the conductivity can
be anisotropic, with maximum conductivity along the Ti-stacking direction as demonstrated
with oligophenylenevinylene (Fig. 2.8) [54].
10'Drain
■710'
S o u rce
■910'Drain
WWWSource
■20 -40 •60 ■80 1000G a te v o lta g e (V)
Figure 2.8. Transistor current for parallel and perpendicular Tr-stacking directions with respect to electric field for oligophenylenevinylene. Adapted from [54].
The degree of crystallisation and thus mobility is significantly affected by numerous
processing conditions such as thennal annealing [55] (Fig. 2.9a), substrate surface properties
[56,57] (Fig. 2.10), solvent evaporation dynamics (solvent boiling point [58] (Fig. 2.9b) and
deposition technique [59]).
25
dL Chloroform X ylene
- C H B TCB
-60 -50 -40 -30 -20 -10 0 V (V)
10 20
Figure 2.9. (a) Effect of annealing on crystal domain size of pBTTT. AFM images of spin-coated film before (top set) and after (bottom set) annealing from [55]. (b) Effect of different solvents on transistor current with P3HT from [58].
To center
To source . or drain ▼
PentaceneSource
SiOGate
Î.
Figure 2.10. Pentacene morphology in different parts of a transistor channel from [57]. The crystallinity is affected by the underlying surface.
High mobility crystalline solution processable organic semiconductors assemble into
highly ordered microstructures with large crystal domains and generally exhibit hole
mobilities in the range of 0.1-1 cm W s with some widely used materials being P3HT (up to
0.1 cm^/Vs [59]), pBTTT (up to 0.6 cm^/Vs [60]), TIPS pentacene (up to 1 cnf/Vs [61]).
Highly purified oligomers such as pentacene and rubrene can give significantly higher
mobilities of up to 35 and 40 cm^/Vs respectively [62,63], but the lack of solubility is a
major setback for a class of semiconductor whose advantage is the ease of processing.
Amorphous polymers that show no preferential Tc-stacking direction have been found
to exhibit very low mobilities [64], the most widely used being polytriarylamine (PTAA) (up
to 0.005 cm^/Vs [65]) and PTPA (up to 0.001 cm'/Vs [2]). Recently, higher mobility
amorphous polymers have been discovered, such as PTVTF with mobility of 0.02 cm^/Vs
26
[66], indenofluorene-triarylamine co-polymer (PIFTAA) exhibiting field-effect mobilities of
0.03-0.04 cmVVs [67,68,69], and indenofluorene-phenanthrene (PIFPA) with claimed field-
effect mobility of 0.5 cm^/Vs [70] though no transistor data is shown. PIFTAA and PIFPA
have not undergone microstructural characterisation in literature to assess crystallinity (aside
from AFM images for PIFTAA [68]), while grazing incidence x-ray diffraction
measurements for PTVTF did not yield any peaks, supporting the lack of crystallisation [66].
PIFTAA and PTPA FETs exhibit highly reproducible performance, and the mobility
shows narrow spread (Fig. 2.11), in contrast to crystalline organic semiconductors that tend
to exhibit wider spreads (Fig. 2.12).
(a) 2 5E-03
« 2.0E-03
1.5E-03
I 1.0E-03
5.0E-04
0 OE+008 10 12
transistor nr.20
20
15
10
5
00.01 0.02 0.03 0.04 0.05 0.06 0.07Saturation mobility (c m W .s)
Figure 2.11. (a) Mobility variation for gravure printed PTPA FETs. Adapted from [2]. (b) Mobility variation for approximately 1500 PIFTAA FETs. The PIFTAA film was deposited by spin-coating, gravure printing, and screen printing. Adapted from [681.
(a)
05 - A -
2 106 84Device Number
0.01
IE-3
IE-4
IE-5DTETEDT BEDT ETTDM BET TTDMETT
TTF derivative
Figure 2. variation
12. (a) Mobility variation of TIPS-pentacene FETs adapted from [71]. (b) Mobility of different TTF derivatives (drop-cast single crystals) adapted from [72].
27
2.6 Ambient stability of organic semiconductors
Degradation in organic transistors is indicated by the degradation of the charge
carrier mobility (reduction in on-state current), shifts of the threshold voltage, decay of
subthreshold slope, increased doping density (increased off-state current), and increased
hysteresis. The main factors that contribute to degradation are oxygen and moisture in
ambient air.
In pentacene single crystals, exposure to oxygen has been found to result in p-doping
(Fig. 2.13a), a process that is enhanced by light, while exposure to moisture was found to
result in the formation of hole trap states and reduction of current (Fig. 2.13b) [73]. Oxygen
has been found to result in shallow and deep trap state formation in rubrene [74]. Oxygen has
also been found to result in subtlireshold slope decay in pentacene [75].
Ü 1 5
50 100 150 200 250
f//7?e [min]
ambient air
o 0.7
0.6
50 100 150 200 250
time [min]
Figure 2.13. Effects of exposure of pentacene single crystals to (a) dry air and (b) ambient air from [73]. I/Ivacc is the measured current in dry/ambient air over the current measured in vacuum prior to exposure.
In a study of 6 polymer insulators, the degree of hysteresis in polymer transistors
was found to be proportional to the hydrophilicity of the insulator. Hysteresis in affected
devices is eliminated by drying the transistors (annealing in vacuum) and returns upon
exposure to ambient atmosphere [ 76 ]. In an extensive review of threshold voltage
instabilities in organic transistor, moisture has been identified as an important factor [77].
The use of hydrophobic insulators mostly resolves moisture related problems, but
highly permeable oxygen cannot be blocked and the semiconductor must be made stable to
oxidation by molecular design. It has been suggested that the oxidative stability is
28
proportional to the IP of the semiconductor [78]. The mobility degradation rates for several
organic semiconductors as a function of IP are summarised at the end of Chapter 5.
P.W. Atkins. Atkins’ Physical Chemistry (7 ed.). Oxford University Press (2002)T. N. Sorrell. Organic Chemistry (2” ed.). University Science Books (2006)J. L. Bredas, J. P. Calbert, D. A. da Silva Filho, J. Comil. “Organic semiconductors: A theoretical
characterization of the basic parameters governing charge transport”, Proc. Nat. Acad. Sci. 99, 5804 (2002)
S. M. Sze, K. K. Ng. Physics of semiconductor devices ed.). John Wiley & Sons (2007)P. J. Collings. “Simple measurement of the band gap in silicon and germanium”. Am. J. Phys. 48,
197 (1980)L. Qian, Y. Zheng, K. R. Choudhury, D. Bera, F. So, J. Xue, P. H. Holloway. “Electroluminescence
from light-emitting polymer/ZnO nanoparticle heterojunctions at sub-bandgap voltages”. Nano Today 5,384(2010)
J. L. Bredas, J. Comil, A. J. Heeger. “The exciton binding energy in luminescent conjugated polymers”, Adv. Mater. 8, 447 (1996)
C.K. Lu, H. F. Meng. “Hole doping by molecular oxygen in organic semiconductors: Band- stmcture calculations”, Phys. Rev. B 75,235206 (2007)
Z. Bao, J. Locklin. Organic Field-Effect Transistors. CRC Press (2007)M. Schwoerer, H. C. Wolf. Organic Molecular Solids. Wiley-VCH (2007)E. A. Silinish, V. Capek. Organic molecular crystals: Interaction, localization, and transport
phenomena. AIP Press (1994)J. L. Bredas, D. Beljonne, V. Coropceanu, J. Comil. “Charge-transfer and energy-transfer processes
in 7i-conjugated oligomers and polymers: a molecular picture”, Chem. Rev. 104, 4971 (2004)S. S. Zade, N. Zamoshchik, M. Bendikov. “From short conjugated oligomers to conjugated
polymers. Lessons from studies on long conjugated oligomers”. Accounts Chem. Res. 44, 14 (2011)O. D. Jurchescu, J. Baas, T. T. M. Palstra. “Effect of impurities on the mobility of single crystal
pentacene”, Appl. Phys. Lett. 84, 3061 (2004)O. Ostroverkhova, D. G. Cooke, S. Shcherbyna, R. F. Egerton, F. A. Hegmann, R. R. Tykwinski, J.
E. Anthony. “Bandlike transport in pentacene and ftinctionalized pentacene thin films revealed by subpicosecond transient photoconductivity measurements”, Phys. Rev. B 71, 035204 (2005)
R. A. Marcus. “Chemical and electrochemical electron-transfer theory”, Ann. Rev. Phys. Chem. 15, 155 (1964)
R. A. Marcus. “Electron transfer reactions in chemistry. Theory and experiment”. Rev. Mod. Phys. 65, 599(1993)
M. Pope, C. E. Swenberg. Electronic processes in organic crystals and polymers (2" ed.). Oxford University Press (1999)
29
s. Athanasopoulos, J. Kirlq)atrick, D. Martinez, J. M. Frost, C. M. Foden, A. B, Walker, J. Nelson. “Predictive study of charge transport in disordered semiconducting polymers”. Nano Lett. 7, 1785(2007)
I. I. Fishchuk, A. Kadashchuk, H. Bassler, S. Nespurek. “Nondispersive polaron transport in disordered organic solids”, Phys. Rev. B 67, 224303 (2003)
H. Bassler, P. M. Borsenberger, R. J. Perry. “Charge transport in poly(methylphenylsilane): The case for superimposed disorder and polaron effects”, J. Polym. Sci. Part B 32, 1677 (1994)
A. Miller, E. Abrahams, “Impurity conduction at low concentrations”, Phys. Rev. 120, 745 (1960)H. Bassler. “Charge transport in disordered organic photoconductors: A monte carlo simulation
study”, Phys. Status Solidi B 175,15 (1993)A. Dieckmann, H. Bassler, P. M. Borsenberger. “An assessment of the role of dipoles on the density
of states function of disordered molecular solids”, J. Chem. Phys. 99, 8136 (1993)P. M. Borsenberger. “Hole transport in tri-p-tolylamine-doped polymers”, J. Appl. Phys. 68, 5188
(1990)J. Veres, S. D. Ogier, S. W. Teeming, D. C. Cupertino, S. M. Khaffaf. “Low-k insulators as the
choice of dielectrics in organic field-effect transistors”. Adv. Funct. Mater. 13, 199 (2003)Y. N. Gartstein, E. M. Conwell. “High-field hopping mobility in molecular systems with spatially
correlated energetic disorder”, Chem. Phys. Lett. 245, 351 (1995)D. H. Dunlap, P. E. Parris, V. M. Kenkre. “Charge-dipole model for the universal field dependence
of mobilities in molecularly doped polymers”, Phys. Rev. Lett. 77, 542 (1996)'*T. Yasuda, M. Saito, H. Nakamura, T. Tsutsui. “Anisotropic carrier transport properties of highly
aligned oligophenylenevinylenes in organic field-effect transistors”, Appl. Phys. A - Mater. Sci. Process. 95, 179 (2009)
I. McCulloch, M. Heeney, C. Bailey, K. Genevicius, I. Macdonald, M. Shkunov, D. Sparrowe, S. Tierney, R. Wagner, W.M. Zhang, M.L. Chabinyc, R.J. Kline, M.D. McGehee, M.F. Toney, “Liquid- crystalline semiconducting polymers with high charge-carrier mobility”, Nat. Mater. 5, 328. (2006)
Y. Hu. Q. Qi, C. Jiang. “Influenece of different dielectrics on the first layer grain sizes and its effect on the mobility of pentacene-based thin-film transistors”, Appl. Phys. Lett. 96, 133311 (2010)
D. Gupta, M. Katiyar, D. Gupta. “An analysis of the difference in behaviour of top and bottom contact organic thin film transistors using device simulation”. Organic Electronics 10, 775 (2009)
J.F. Chang, B. Sun, D.W. Breiby, M.M. Nielsen, T.I. Soiling, M. Giles, I. McCulloch, H. Sirringhaus. “Enhanced mobility of poly(3 -hexylthiophene) transistors by spin-coating from high- boiling-point solvents”, Chem. Mater. 16,4772 (2004)
H. Sirringhaus, P.J. Brown, R.H. Friend, M.M. Nielsen, K. Bechgaard, B.M.W. Langeveld-Voss, A.J.H. Spiering, R.A.J. Janssen, E.W. Meijer. “Microstructure-mobility correlation in self-organised conjugated polymer field-effect transistors”, Synth. Met. 111-112, 129 (2000)^ I. McCulloch, M. Heeney, M.L. Chabinyc, D. DeLongchamp, R.J. Kline, M. Colle, W. Duffy, D. Fischer, D. Gundlach, B. Hamadani, R. Hamilton, L. Richter, A. Salleo, M. Shkunov, D. Sparrowe, S.
30
Tierney, W. Zhang. “Semiconducting thienothiophene copolymers: design, synthesis, morphology, and performance in thin-film organic transistors”. Adv. Mater. 21, 1091. (2009)
M. M. Payne, S. R. Parkin, J. E. Anthony, C. C. Kuo, T. N. Jackson. “Functionalized higher acenes: hexacene and heptacene”, J. Am. Chem. Soc. 127,4986 (2005)
O.D. Jurchescu, J. Baas, T.T.M. Palstra. “Effect of impurities on the mobility of single crystal pentacene”, Appl. Phys. Lett. 84, 3061 (2004)
T. Hasegawa, J. Takeya. “Organic field-effect transistors using single crystals”, Sci. Technol. Adv. Mater. 10, 024314 (2009)^ H. Sirringhaus. “Devices physics of solution-processed organic field-effect transistors”. Adv. Mater. 17, 2411 (2005)
J. Veres, S. Ogier, G. Lloyd, D. de Leeuw. “Gate insulators in organic field-effect transistors”, Chem. Mater. 16, 4543 (2004)
D. S. Chung, S. J. Lee, J. W. Park, D. B. Choi, D. H. Lee, J. won Park, S. C. Shin, Y. H. Kim, S. K. Kwon, C. E. Park. “High performance amorphous polymeric thin-film transistors based on poly[(l,2- bis-(2’-thienyl)vinyl-5’,5” -diyl)-alt-(9,9-dioctylfluorene-2,7-diyl] semiconductors”, Chem. Mater. 20, 3450 (2008)
W. Zhang, J. Smith, R. Hamilton, M. Heeney, J. Kirkpatrick, K. Song, S. E. Watkins, T. Anthopoulos, I. McCulloch. “Systematic improvement in charge carrier mobility of air stable triarylamine copolymers”, J. Am. Chem. Soc. 131, 10814 (2009)
J. M. Verilhac, M. Benwadih, A. L. Seiler, S. Jacob, C. Bory, J. Bablet, M. Heitzman, J. Tallal, L. Barbut, P. Frere, G. Sicard, R. Gwoziecki, I. Chartier, R. Coppard, C. Serbutoviez. “Step toward robust and reliable amorphous polymer field-effect transistors and logic fiinctions made by the use of roll to roll compatible printing processes”. Organ. Electron. 11,456 (2010)
J. M. Verilhac, M. Benwadih, S. Altazin, S. Jacob, R. Gwoziecki, R. Coppard, C. Serbutoviez. “Effects of amorphous semiconductor thickness on top gate staggered organic field-effect transistors”, Appl. Phys. Lett. 94, 143301 (2009)
H. Kim, N. Schulte, G. Zhou, K. Mullen, F. Laquai. “A high gain and high charge carrier mobility indenofluorene-phenanthrene copolymer for light amplification and organic lasing”. Adv. Mater. 23, 894(2011)
Z. He, K. Xiao, W. Durant, D. K. Hensley, J. E. Anthony, K. Hong, S. M. Kilbey II, J. Chen, D. Li. “Enhanced performance consistency in nanoparticle/TIPS pentacene based organic thin film transistors”. Adv. Funct. Mater. 21, 3617 (2011)
M. Mas-Torrent, P. Hadley, S. T. Bromley, X. Ribas, J. Tarres, M. Mas, E. Molis, J. Veciana, C. Rovira. “Correlation between crystal structure and mobility in organic field-effect transistors based on single crystals of tetrathiafulvalene derivatives”, J. Am. Chem. Soc. 126, 8546 (2004)
O. D. Jurchescu, J. Baas, T. T. M. Palstra. “Electronic transport properties of pentacene single crystals upon exposure to air”, Appl. Phys. Lett. 87, 052102 (2005)
31
c. Krellner, S. Haas, C. Goldmann, K. P. Pemstich, D. J. Gundlach, B. Batlogg. “Density of bulk trap states in organic semiconductor crystals: discrete levels induced by oxygen in rubrene”, Phys. Rev. B 75, 245115 (2007)
D. Knipp, T. Muck, A. Benor, V. Wagner. “Environmental stability and electronic transport of pentacene thin film transistors”, J. Non-Cryst. Solids 352,1774 (2006)
Y.H. Noh, S.Y. Park, S.M. Seo, H.H. Lee. “Root cause of hysteresis in organic thin film transistor with polymer dielectric”. Organic Electronics 7, 271 (2006)
H. Sirringhaus. “Reliability of organic field-effect transistors”. Adv. Mater. 21, 3859 (2009)D. M. De Leeuw, M. M. J. Simenon, A. R. Brown, R. E. F. Einerhand. “Stability of n-type doped
conducting polymers and consequences for polymeric microelectronic devices”, Synth. Met. 87, 53 (1997)
32
3. Organic field-effect transistors
3.1 Metal-semiconductor contacts
This section is partly adapted from [79].
As explained earlier, metals have no bandgap and thus an abundance of free
electrons and holes. Semieonductors have a bandgap and the majority earrier is detenuined
by the relative position of the Fermi level to the LUMO or HOMO (p-type for HOMO
proximity as in Fig. 3.1a). In Fig. 3.1a the work funetion is denoted by (Pm for the metal and
cps for the semiconductor and is the amount of energy required for an electron to escape the
material (or a hole to enter it). Eea is the electron affinity and Ejp the ionisation potential.
When a low workfunction metal is brought into eontact with a p-type semiconductor, in the
vicinity of the contact some holes from the semiconductor flow into the metal, Fenui levels
align, and a negatively charged depletion region of width Wd and depth qvpei is fonned in the
vicinity of the contact (Fig. 3.1b). \ | / b i is the built in potential that quantifies band bending in
such a process.
(a) Metal
op(Uc
LU_0JoX
VL
4>r
SemiconductorVL VL'
VL
4>s
■EA■ LUMO
■Ep• HOMO
4>r
(c)
VL —
LUMO 4)m
Wn
HOMO 4>e
/Figure 3.1. Energy band diagrams, (a) Metal and p-type semiconductor separated by a gap. b) Metal and semiconductor in contact and resulting band bending, c) Hole barrier decrease in the presence of an interface dipole (negative pole towards semiconductor). VL is the vacuum level, Ep is the Eermi level, ipai H the potential due to band bending, A is the potential due to an interface dipole.
Hole flow from the metal to the semiconductor (MS) must overcome an energetic
barrier (pe which is equal to the energetic difference between the HOMO level and the metal
workfunction:
98 — —(PM (Eq. 3.1)
In a more realistic system, interface dipoles may be present and may offset the bands
in the semiconductor. Dipoles with the negative pole towards the semiconductor (Fig. 3.1c)
reduce the hole injection barrier by creating additional potential A across the interfacial
layer.
33
9 b - Ejp “ 9 m (Eq. 3.2)
3.1.1 Injection and Schottky barriers
The application of voltage bias across the contact will break up the Fermi level
alignment as voltage drop occurs only in the semiconductor and not the metal. For the same
reason the MS barrier cpB will be unaffected by any voltage.
For a forward bias (Fig. 3.2a), i.e. a bias that will lower the semiconductor to metal
(SM) barrier, the barrier encountered by the hole flow will be vi/bi-Vf, and the net hole
cuiTent will flow from the semiconduetor into the metal. For a reverse bias (Fig. 3.2b), the
barrier encountered by the hole flow will increase to \ | / b i+ V r , and significantly smaller hole
current will flow from the metal to the semiconductor as described by Eq. 3.4.
This behaviour gives rise to the rectifying characteristic of Schottky diodes: high
current ean only flow from the semiconductor to the metal in forward bias (Fig. 3.2c). In
reverse bias, the dominant current is the metal to semiconductor current, the magnitude of
which is determined by the height of %. The reverse-bias current saturates (Fig. 3.2c) as the
widened depletion region absorbs all additional voltage after a certain small saturation
voltage.
(a) (b)
VL- VL VL-
■ LUMO
•EpHOMO
VL (c)
LUMOV;
*.xr HOMO
V
Figure 3.2. Semiconductor band shift under applied (a) forward and (b) reverse voltage bias (c) I-V curve for a Schottky diode. VL is the vacuum level.
The channel of a field-effect transistor with Schottky metal-semiconductor contacts
essentially consist of two back to back diodes, one at the source-semiconductor and one at
the drain-semiconductor interface. If a negative source-drain voltage is applied to a p-type
FET, the source-semiconductor contact will be reverse biased and the drain-semiconductor
contact will be forward biased, as such the current through the device will be dominated by
the metal to semiconductor hole flow at the reverse biased source-semiconductor contact.
The current through a Schottky diode (Fig. 3.2c) at applied voltage V is given by
[79]:
q V
(Eq. 3.3)I = I s ( e k T - l )
34
Where Is is the saturated current in reverse bias:
I s = S A r t e kT (Eq. 3.4)
Where A is the area and S is Richardson’s constant.
The Schottky barrier tpe in reality undergoes an effective lowering Acp as a function
of electric field F. The image force lowering arises from the induction of a positive charge on
the metal surface by an approaching electron and is given by:
(Eq. 3.5)4ks,
fPBeff =<Pb -A 'P (Eq. 3.6)
Where Sg is the permittivity of the semiconductor. As a result, the effective MS
barrier height becomes slightly voltage dependent: decreases with reverse bias and increases
with forward bias (though still is smaller than the uncorrected barrier height (pe).
3.1.2 Fermi-Level Pinning
In figure 3.1 we considered the contact between a low work function metal and a p-
type semiconductor, in which case the barrier height can be estimated by assuming vacuum
level alignment between the metal and semiconductor and subtracting the metal work
function from the semiconductor IP (eq. 3.1). In a more general case, where metal work
function can also be larger than the semiconductor work function, a different regime of band
alignment can occur, referred to as Fermi level pinning. This regime can be observed when
the metal work function exceeds the organic semiconductor polarisation level Ep+, at which
point the Fermi level will become pinned and any further increase in cpM will cause
misalignment of the VLs of the metal and semiconductor, band bending, and positive space
charge (p-doping in the semiconductor in the vicinity of the contact).
Fermi-pinning has been experimentally studied with Ultraviolet Photoelectron
Spectroscopy (UPS) and was used to estimate the polarisation energy of several conjugated
polymers [80].The surface potential of the polymers was measured for a number of
underlying metals of different workfunctions, and the Fermi pinning level was interpolated at
the point where the polymer surface potential stopped being proportional to the metal work
function. Finally the interpolated value was subtracted from the measured polymer IP, and
the polarisation energy was estimated (P3HT 0.5 eV, TFB 0.7 eV, PI OAF 0.4 eV, PFO 0.6
eV) [80]. This measurement may contain errors since UPS is a surface probing technique and
surface polarons possess lower energy than bulk polarons [81,82,83]. This difference has
35
been experimentally measured to be 0.3 eV for anthracene by comparing energies of
photoelectrons emitted in parallel to the film surface (presumed to come from surface
polarons) and photoelectrons emitted perpendicularly to the film surface (presumed to come
from the bulk) [82].
Fermi piiming has been utilised in Kelvin Probe (KP) measurements to investigate
band bending and polarisation energies [84]. For band bending, varying thicknesses of
conjugated polymer thin films were prepared on metals of different workfunctions and the
surface potentials were measured and plotted as a function of thickness (Fig. 3.3a). For
polarisation energies, the Fenni pinning levels were interpolated similarly to [80] (Fig. 3.3b)
and subtracted from UPS measured IPs to estimate Ep+ (F8BT 0.6 eV, CN-ether-PPV 0.6 eV,
PFBBTT 0.4 eV, N2200 0.4 eV) [84].
★ Sm
HI
-©- 5.0
5.5
6.0
6.5
7.0
O Au ▼ Ag A Clevios "AI 4083
Cievios'^HIL 1.3 X MoO,(70nm)□ MoOj(IOOnm)
(b)3.5-
4.0
— 4.5(jO“©■
5.0
5.5-3 5 6 74
* IV I (eV)
0 20 40 60 80 100 120Distance (nm)
Figure 3.3. (a) Semiconductor surface potential as a function of polymer thickness on the metal underlying layer, and (b) semiconductor surface potential versus metal workfunction for F8BT, modified from [84]. Plots for three additional polymers can be found in this reference, all of which exhibit band bending to a similar degree.
3.1.3 M etal-semiconductor surface dipole manipulation
Surface dipoles of varying polarity and magnitude can be engineered on metal
surfaces by treating them with Self-Assembled-Monolayers (SAMs). SAMs consist of
primarily three components: a binder that adsorbs on target surface, a spacer, and the polarity
36
and magnitude determining end group. SAMs are typically deposited by sample immersion
into SAM solution. For commonly used metals in p-type organic transistors such as Ag and
Au, thiol-binder SAMs are often used to increase the workftinction. The magnitude of the
dipole is mainly determined by the length of the molecule and the electronegativity of the
terminal group.
In work with Ag, SAM treatment was performed with linear thiol molecules
HS(CH2)9CH3, HS(CH2)ioNH2, and HS(CH2)2(CF)?CF3, with calculated dipoles towards the
semiconductor of +2.24 D, -1.77 D, and -1.69 D respectively, resulting in workfunction
shifts as compared to pristine Ag of -0.70 eV, -0.45 eV, and 0.85 eV respectively, also
measured by electroabsorption measurements on Ag/MEH-PPV/Ca Schottky diodes and
found to be -0.60 eV, -0.45 eV, and +0.50 eV respectively [85]. Similar studies are common
[86,87,88].
Long chain thiols as mentioned above are insulating and impede injection after a
certain length (effective mobility as a ftmction of alkyl thiol length starts to drop for SAMs
longer than 4 carbon atoms [88]), thus a commonly used class of thiols (in the present work
also) has short aromatic spacers of various terminations, such as nitrobenzenethiol [89],
fluorobenzenethiol and pentafluorobenzenethiol [90].
Considering the SAM induced dipole A, total positive polarisation energy Ep+, and
electrode workftinction cpM, the hole barrier in the VL alignment regime should be given by:
(pB =Epp -(pM — A —Ep+ forEBA +Ep_ <(pM + A < E ]p -Ep+(Eq. 3.7)
3.1.4 Ohmic contacts in OFETs
A contact is ohmic when the current density J increases linearly with applied voltage
V according to Ohm’s law [91]:
Johm = eN oIi^ (Eq. 3.8)
Where L is the electrode spacing, p is the mobility and No is the number of free
carriers per unit volume. Practically, a contact is ohmic when the current is limited by the
transport process in the semiconductor and not the contact, which is determined by the
relative resistances of the semiconductor and contacts. Consider a transistor channel with
source and drain contacts (Fig. 3.4) of resistance Rg and respectively, as well as channel
resistance Rchan- The total contact resistance is Rc=Rs+Rd- If Rchan»Rc then the IVd
37
relationship (output characteristic) will be linear, as in the case of Ag-P3HT transistors (Fig.
3.4c) [92], for which the channel resistance is roughly 400 kQcm and Rg is 15 kHcm. For
high barrier contacts Cr-P3HT contacts, the Rg is roughly 300 kQcm. As a result, the output
characteristic is not linear at low Vd. Additionally, since a large portion (almost half) of Vd
is dropped on the contact, the current is reduced. For the transfer characteristic in Fig. 3.4b,
although the Ag-P3HT FET is operating at an effective drain voltage of -5 V, the Cr-P3HT is
operating at half the effective drain voltage because of the high contact-to-channel
resistance. Note that the channel resistance scales with channel length (roughly 20 kQcm/pm
for P3HT FETs in [92]), and as such contact effects become significant at short channels.
(a)Source Sem ico n d u c to r Drain
Chan
(C150
100
-50
80 0 20 -40 -60
Source
300 KK J 4 5 KV
190
300 KP3HTCr
-2 -1 0 1 2 3 4 5 6 7D istance from source (fim)
Figure 3.4. (a) Representation of source and drain contact resistance R§ and Rd, and channel resistance Rchan in a FET channel, (b) transfer characteristic of P3HT FET (channel length 22 pm) with Ag, Cr, A1 source/drain contacts and (c) output characteristic, (d) non-contact potentiometry measurement of P3HT FET (channel length 5pm) with Cr electrodes and voltage drop on the contacts and channel at different temperatures. (b,c,d) from |92j.
In organic FET experiments, barriers heights of less than 0.3 eV have been found to
be near-ohmic, while output characteristics lose linearity for 0.3 and higher bamers [92,93].
Ohmic contacts in organic transistors of typical channel lengths 5 to 100 pm generally have
38
contact resistance of less than SOkQcm, while the contact resistance for Schottky contacts
can be significantly higher [92, 93, 94]. In [92] the contact resistance was found to be similar
for the source and drain for ohmic contacts, while it was dominated by the source resistance
for Schottky contacts, consistent with reverse-biased source Schottky diode and forward-
biased drain Schottky diode (section 3.1.1).
Also, non-linear injection has been observed as a result of positive space charge
build-up, which was found to be alleviated by using semiconductors of higher IP [95].
3.2 Field-Effect Transistors
FETs consist of two components: the back to back source-semiconductor and
semiconductor-drain diodes (source-semiconductor-drain channel), and the Semiconductor-
Insulator-Gate capacitor. FETs are three tenuinal devices (source, drain, gate) and
fundamental electronic components. In digital electronics they are used as voltage controlled
switches that allow or inhibit current flow depending on the applied voltages, enabling
straightforward implementation of binary logic. In analog electronics they are used as current
drivers and voltage amplifiers.
In p-type FETs, charge carriers enter the conductive channel at the source-
semiconductor contact and exit at the drain-semiconductor contact. The metal-semiconductor
contacts are assumed to be ohmic. Accumulation mode Field-Effect Transistors (FETs) are
natural AND gates since they require application of two voltages for current to flow, the gate
voltage that induces charge along the channel, and the drain voltage that causes charge to
flow from the source to the drain (Fig. 3.5).
Gate
VSG
Insulator
SemiconductorSource
AyDrain
Figure 3.5. Structure of thin film field-effect transistor and representation of charge accumulation and flow upon application of gate and drain voltages.
39
For a parallel plate capacitor the capacitance per unit area Ci is given by:
(Eq.3.9)d
Where A is the plate area set to Icm^ (lO'^m), Sq is the permittivity of free space, Sr is
the relative permittivity of the insulator, and d is the thickness of the insulator.
Upon application of a voltage V between the two plates, the capacitor is charged and
equal and opposite charge Q=CV is induced on the two opposing surfaces. The situation is
slightly different in a semiconductor-insulator-gate capacitor as the voltage across the
semiconductor channel is not constant for a non-zero source-drain voltage Vd. The charge
per unit area Qi is given by:
VryQ i = Q j ( V G - V c - V T ) d V c = Q ( V G - V T - ^ ) (Eq.3.10)
0Where Vq is the source-gate voltage, Vt is the threshold voltage required to bring
the semiconductor to a conducting state, and Vc is the voltage in the semiconductor channel
that varies from 0 V at the source to Vd V at the drain.
Equation 3.8 is written in a different form for an FET :
^ = (Eq.3.11)W L
The source-drain current per unit channel width IsdAV can be found from QipF
where the electric field F is given by the drain voltage and channel length L. Substituting eq.
3.10 into eq. 3.11 gives;
ISD = — (Y q - V p (Eq. 3.12)
Equation 3.12 is only valid for approximately 0<Vd<Vg, the so-called linear regime
of operation. When the drain voltage Vd exceeds a certain voltage Vsat, charge carriers are
depleted near the drain. Any additional increase in Vd will be dissipated across the depletion
region and will not contribute to current flow.
Vp) = V sat = ^ ~ Vq - Vx for Y j ^ > Y q - Y j (Eq. 3.13)
Where K is a constant relating to insulator thickness [79]. Replacing Vd with Vq-Vt
in equation 3.12 gives the saturation regime current equation:
40
IgD = “ Cin(VG - V j (Eq. 3.14)
The mobility in the linear and saturation regime is given by:
dIcD L
|xSat =, d V o ,
L(Eq. 3.16)
WCiThe mobility model in the saturation regime is less susceptible to contact resistance
as it does not rely on drain voltage. As a result the saturation mobility is always higher than
the linear mobility and their difference is an indicator of contact quality.
Correction of the linear mobility, as well as the separation of contact and channel
resistance is achievable by exploiting the dependence of the chaimel resistance on chaimel
length. The Transmission Line Method (TLM) can only be applied in the linear regime as the
uniformity of resistance in the channel is crucial.
^total ~ ‘ ^chan (Eq. 3,17)
For TLM, FETs of different channel lengths are fabricated, the output and transfer
characteristics are measured, and the total output resistance is calculated from the inverse
slope of the output characteristic. The output resistances from FETs of different Ls are then
plotted as a function of L. The contact resistance Rc is extrapolated with a linear fit at L=0,
while the contact resistance corrected mobility pcor can be calculated from the slope of the
plot which is Rchan-
3.2.1 Semiconductor-insulator interface
Although the transistor current (Eq. 3.12, 3.14) is proportional to the insulator
capacitance and thus the insulator permittivity (Eq. 3.9), it has been shown that the high
insulator permittivity causes increase of energetic disorder at the semiconductor-insulator
interface, resulting in significant drop of mobility for amorphous polymer semiconductors
(Fig. 3.6b) [51]. This behaviour has been attributed to static dipolar disorder that decays
strongly with distance away from the insulator surface (Fig. 3.6a) [96].
41
(a)
p=2D
COoQ Increasing xj
-0.5 0.5Energy / eV
(b) 10-
10-3-
10-
10
PTAA1 top gate PTAA1 bottom gate PTAA2 top gate PTAA2 bottom gate
10 15
Figure 3.6. (a) Calculated DOS broadening due to static dipolar disorder in the dielectric with increasing distance into the semiconductor. Each line corresponds to a step of 1 A from x = 0 A up to X = 5 A. Figure from [96]. (h) PTAA FET mobility as a function of insulator permittivity from [51].
3.3.2 Voltage gain
The tranconductance gm is extracted from transfer of output characteristics from the
derivative of the drain current Id with respect to gate voltage Vq. The output conductance gd
is extracted from output characteristics from the derivative of the drain current Id with
respect to drain voltage Vd. The intrinsic voltage gain A is given by [29]:
ê m
gd(Eq. 3.18)
42
3.3 Electrode configurations
(a) Gate (b)
Insulator
jSemiconductor
Substrate
(c) (d)
Figure 3.7. Thin Film Transistor configurations: (a) Bottom-Contact (BC) Top-Gate (TG) (staggered), (b) BC Bottom-Gate (BG) (co-planar), (c) BC TG (co-planar), (d) Top-Contact (TC) BG (staggered)
There are four different electrode configurations for Thin Film Transistors (TFTs).
The source and drain electrodes can be deposited either before or after the semiconductor,
called Bottom Contact (BC, Fig. 3.7a,b) and Top Contact (TC, Fig. 3.7c,d) respectively.
Similarly, the gate electrode before or after the other layers, called Bottom Gate (BG, Fig.
3.7b,d) or Top Gate (TG, Fig. 3.7a,c) respectively. In addition, an electrode configuration is
classified as staggered if the source/drain and gate are on opposite sides of the
semiconductor (Fig. 3.7a,d), and co-planar if the source/drain and gate are on the same side
of the semiconductor (Fig. 3.7b,c). These configurations each have distinct advantages and
disadvantages originating from electrode geometry, deposition processes, and environmental
factors. The features of each configuration are difficult to analyse since charge injection and
transport depend on multiple parameters which are practically inseparable, such as electrode
semiconductor contact area, microstructure of semiconductor in the contact vicinity, dipoles
and resulting energetic barriers, presence of semiconductor in source-gate capacitor, metal
deposition process, interface states.
TC exhibit less contact resistance than BC as a consequence of fabrication, since the
deposition of the conductor on top of the semiconductor leads to a better contact due to the
relationship between semiconductor crystallinity and underlying surface properties (Fig.
43
2.10). However BC is more practieal as it allows harsh patterning processes to be utilised
(such as photolithography) without damaging the sensitive organic semiconductor.
The staggered configuration has an additional component as compared to the co-
planar configuration, the semiconductor is in series with the insulator as part of the source-
gate capacitor, and charge carriers must traverse a resistive region in the semiconductor to
reach the channel at the insulator interface (access resistance). As such, the sum of contact
and access resistance, the source resistance, is higher for staggered than co-planar
configurations [97]. Additionally, the source resistance in a staggered configuration is a
proportional to the semiconductor thickness, while it is constant for a co-planar configuration
[97]. Finally, in the field effect saturation regime the semiconductor is depleted in the
vicinity of the drain contact, and this capacitor in series with the insulator capacitor results in
a decrease of total source-gate capacitance.
For the staggered configuration, the increase in source resistance results in decreased
transconductance (Fig. 3.8a), and the decrease in total gate capacitance results in higher
operating frequency (Fig. 3.8b) and intrinsic gain [97].
( b ) 300.7
0.6
BG BC (co-planar) BG TC (staggered)
0.4
O)0.3 BG TC (staggered) BG BC (co-planar)
0.22500 15050 100 200150 200 2500 50 100
d (n m ) d (n m )
Figure 3.8. Variation of (a) transconductance gn, in FET saturation regime at constant current and (b) unity gain frequency fx versus semiconductor thickness d for simulated staggered and co-planar configurations, modified from [97].
Additionally, the staggered configuration has been shown to be less susceptible (in
tenus of current decrease) to Schottky bamers than the co-planar configuration, both in
experiment and simulation with pentacene [93]. The staggered configuration is relatively
unaffected by Schottky barriers up to 0.3 eV [93].
No direct comparisons could be found between TG and BG, which is investigated in chapter 5.
44
s. M. Sze, K. K. Ng. Physics of semiconductor devices (3 éd.). John Wiley & Sons (2007)C. Tengstedt, W. Osikowicz, W. R. Salaneck, I. D, Parker, C.H. Hsu, M. Fahlman. “Fermi-level
pinning at conjugated polymer interfaces”, Appl. Phys. Lett. 88 (2006) 053502D. R. T. Zahn, G. N. Gavrilla, M. Gorgoi. “The transport gap of organic semiconductors studied
using the combination of direct and inverse photoemission”, Chem. Phys. 325, 99 (2006)W. R. Salaneck. “Intermolecular relaxation energies in anthracene”, Phys. Rev. Lett. 40, 60 (1978)Z. G. Soos, E. V. Tsiper, R. A. Pascal Jr. “Charge redistribution and electronic polarization in
organic molecules crystals”, Chem. Phys. Lett. 342, 652 (2001)^ I. Lange, J.C. Blakesley, J. Frisch, A. Vollmer, N. Koch, D. Neher. “Band bending in conjugated polymer layers”, Phys. Rev. Lett. 106, 216402 (2011)
I. H. Campbell, S. Rubin, T. A. Zawodzinski, J. D. Kress, R. L. Martin, D. L. Smith. “Mobility and diffusivity in a generalised Frenkel-Kontorova model”, Phys. Rev. B 54, 321 (1996)
D. M. Alloway, M. Hofmann, D. L. Smith, N. E. Gruhn, A. L. Graham, R. Colorado Jr, V. H. Wysocki, T. R. Lee, P. A. Lee, N. R. Armstrong. “Interface dipoles arising from self-assembled monolayers on gold: UV photoemission studies of alkanethiols and partially fluorinated alkanethiols”, J. Phys. Chem. B 107,11690 (2003)
P. Marmont, N. Battaglini, P. Lang, G. Horowitz, J. Hwang, A. Kahn, C. Amato, P. Calas. “Improving charge injection in organic thin-film transistors with thiol-based self-assembled monolayers”. Organ. Electron. 9,419 (2008)
P. Stoliar, R. Kshirsagar, M. Massi, P. Annibale, C. Albonetti, D. M. De Leeuw, F. Biscarini. “Charge injection across self-assembly monolayers in organic field-effect transistors: odd-even effects”, J. Am. Chem. Soc. 129, 6477 (2007)
T. J. Ha, D. Sparrowe, A. Dodabalapur. “Device architectures for improved amorphous polymer semiconductor thin-film transistors”. Organ. Electron. 12, 1846 (2011)^ J. P. Hong, A. Y. Park, S. Lee, J. Kang, N. Shin, D. Y. Yoon. “Tuning of Ah work functions by self-assembled monolayers of aromatic thiols for an efficient hole injection for solution processed triisopropylsilylethynyl pentacene organic thin film transistors”, Appl. Phys. Lett. 92, 143311 (2008)
Y. Shen, A. R. Hosseini, M. H. Wong, G. G. Malliaras. “How to make Ohmic contacts to organic semiconductors”, Chem. Phys. Chem. 5, 16 (2004)
L. Burgi, T. J. Richards, R. H. Friend, H. Sirringhaus. “Close look at charge carrier injection in polymer field-effect transistors”, J. Appl. Phys. 94, 6129 (2003)
D. J. Gundlach, L. Zhou, J. A. Nichols, T. N. Jackson, P. V. Necliudov. “An experimental study of contact effects in organic thin film transistors”, J. Appl. Phys. 100, 024509 (2006)^ B. H. Hamadani, D. Natelson. “Nonlinear charge injection in organic field-effect transistors”, J. Appl. Phys. 97, 064508 (2005)
T. Hirose, T. Nagase, T. Kobayashi, R. Ueda, A. Otomo, H. Naito. “Device characteristics of short- channel polymer field-effect transistors”, Appl. Phys. Lett. 97, 083301 (2010)
45
^ T. Richards, M. Bird, H. Sirringhaus. “A quantitative analytical model for static dipolar disorder broadening of the density of states at organic heterointerfaces”, J. Chem. Phys. 128, 234905 (2008)
M. N. Islam, B. Mazhari. “Comparative analysis of unity gain frequency of top and bottom-contact organic thin film transistors”. Solid State Electron. 53, 1067 (2009)
46
4. Transistor Fabrication
4.1 Introduction
This chapter describes the fabrication steps for transistors and variations for different
structures. The materials used in this work are presented (metals, thiols, insulator,
semiconductors). The workfunctions of various electrodes and processing parameters are
examined by Kelvin Probe (KP).
Chapter 5 is based on Bottom-Gate (BG), Top-Gate (TG), and Dual-Gate (DG)
Field-Effect Transistors (FETs) with Au/thiol source/drain electrodes. Chapter 6 is based on
multi-channel TG structures. Chapter 7 is based on TG with various source/drain electrode
combinations. All the different structures can be thought of as subsets of DG transistors (Fig.
4.1).
Top Gate
Insulator
Sem iconductor
Insulator
Substrate
Figure 4.1. Illustration of structure of Dual-Gate transistors. Dimensions are not to scale.
4.2 Substrate cleaning
Coming 1737 alkaline earth boro-aluminosilicate glass substrates were cut in Ixlcin
squares and cleaned in a sonic bath by immersion into light decon 90:H2O solution,
deionised (Dl) H2O (3 times), acetone, and isopropanol (approximately 3 minutes per step).
Finally the samples were dried with pressurised dry nitrogen. The surface roughness
measured by profilometry was found to be approximately 5 nm.
47
4.3 Thickness and surface characterisation of polymer thin films
Thin film thicknesses for metal and polymer layers were measured by scratching the
film and imaging the trench by tapping mode Atomic Force Microscopy (AFM, Veeco
Dimension 3000) or profilometry (Tencor Alphastep 200 profilometer). The same techniques
were used to characterise roughness and AFM was used to image surfaces with PPP-NCH
tips (8nm tip radius).
4.4 Metal thin film deposition
The metals mainly used in this work were Au and Cr, whereas Ni and Ti were used
in limited cases. Gates for all devices were made with evaporated Au. Source/drain
electrodes for FETs were made with evaporated or sputtered Au, with or without Cr or Ti
adhesion layers. Source/drain electrodes for SGTs were made with sputtered Cr or Ni, in
addition to Au electrode variations listed in section 4.6.
The use of adhesion layers significantly improved the scratch resistance of Au
electrodes (critical for establishing contact with microprobes), although at the cost of slightly
reduced workfunction as shown later on. The presence of an adhesion layer did not make any
noticeable difference for fabricated FETs with conjugated polymers except for PIFPA, for
which the lack of adhesion layer gave near-ohmic contacts with Au/PFBT (Chapter 5), and
non-ohmic contacts with Ni/Au/PFBT and Cr/Au/PFBT (Chapter 6 and 7 respectively). This
is discussed in chapter 7.
Au, Cr, Ti, and Ni thin films were deposited by DC magnetron sputtering in argon
with typical deposition rate of approximately 5 nm / min, in thicknesses of 30-40 nm.
Sputtering was generally used for metal deposition onto glass substrates, whereas thermal
evaporation was used for deposition onto more sensitive polymer surfaces as in the case of
top gates and source/drain electrodes on top of the polymer insulator for dual gate transistor
fabrication. Evaporation was performed in 10' mbar pressure with rates of 0.6 to 2 nm / min
(ramped up with thickness). Film thickness was monitored with dual Quartz Crystal
Microbalance (QCM) monitors, occasionally calibrated by measuring the actual film
thickness.
48
4.5 Metal patterning
Source/drain electrodes were produced by photolithography and etching of metal
thin films. Gates were produced mainly by evaporation through a shadow mask (FET
structure, Fig. 4.2b), except for TLM structures (Fig. 4.3) where photolithography and
etching was used.
For photolithography, samples were initially spin-coated with hexamethyldisilazane
(HMDS) (4000 RPM, 30 s) which promotes the adhesion of photoresist and fixes the surface
tension to allow for the reproduction of photoresist thickness. Photoresist thickness must be
reproduced precisely as the relationship between exposure dose and photoresist thickness is
non-trivial (sinusoidal [98]). As long as the thickness is reproducible, the value is irrelevant
(for etching, not for lift-off) and was not measured.
S I805 positive photoresist (Microposit) was spin-coated at 4000 RPM for 30s and
annealed at 105°C for 3 min. The optimal exposure dose for the production of 2.5 pm
features was found to be 50±3 mJ/cm^. The UV lamp intensity was always measured and the
exposure time adjusted to obtain the desired total dose.
The main source/drain pattern used in this work was an interdigitated structure of
effective channel width W = 1 cm and channel lengths of L = 2.5, 5, 10 pm (Fig. 4.2). In
some cases a similar structure was used with an additional L = 20 pm. For Transmission
Line Method (TLM) structures a different mask was used with W = 1 mm and L = 2.5, 5, 10,
20,40, 60, 80,100,120, 140,160,180,200 pm (Fig. 4.3).
49
SOOum
500|im 500|im
Figure 4.2. Photomask pattern for (a) interdigitated source/drain electrodes and (b) shadow mask pattern for gate electrode evaporation (from CAD file), (c) Microscope image of structure with bottom gate, insulator, and patterned An source/drain electrodes, (d) Same structure following semiconductor, top insulator and top gate deposition.
(a) (b)
SOOpimSOO^im
Figure 4.3. (a) Photomask pattern for TLM source/drain electrodes (yellow) and overlapping gate (green) from CAD file, (b) Microscope image of patterned An source/drain electrodes.
50
Following exposure the samples were immersed in MF319 (Rohm & Haas)
photoresist developer for 30 s and then annealed at 105°C for 3 min.
An was etched with 4:1:100 I:KI:H20 with an approximate etching rate of 1 nm / s.
Etching was initially isotropic, and the etching rate on the xy plane dropped noticeably as the
film thickness approached zero, regardless of the underlying surface. As such, samples were
typically immersed for twice the initial estimated time (60 s for 30 nm films). Although this
overetching should not result in appreciable undercutting, unexplained undercutting of up to
2 jam was occasionally observed.
Cr was etched with perchloric acid - cerium ammonium nitrate (MS 8, Gower
Chemicals) with highly isotropic etching rate of approximately 1 nm / s. Ni was etched with
3:1 HChHNOs with variable etching rate in the range of 3-6 nm / s. Ti was etched with 1:2
NH4 0 H:H2 0 2 with etching rate of 2 nm / s.
Following etching, samples were placed under a stream of DI water for 1 min, dried
with nitrogen, immersed into acetone multiple times to remove the photoresist, immersed
into isopropanol to remove acetone residue, and dried with nitrogen. Finally, only for An
electrodes, samples were exposed to O2 plasma (50 W, 30s, 30 ml/min) to remove iodine
species from the An surface as explained in section 4.7.
4.6 SAM deposition
Following source/drain electrode cleaning, samples were treated with thiol SAMs.
As described in section 3.1.3, thiol molecules consist of a headgroup (thiol) and a
spacer/end-group. In this work we have mainly used thiols with pentafluorobenzene (PFBT,
Fig. 4.4a) and fluorobenzene (FBT, Fig. 4.4b) terminations (Sigma Aldrich). The fluorinated
terminations form a strong negative pole towards the semiconductor, increasing the effective
workfunction of the electrode and reducing the hole injection barrier. Samples were
immersed in ImMol thiol solution in 100ml of ethanol for 15 minutes. Samples were then
rinsed with ethanol and dried with pressurised nitrogen.
(b),SH
. X T '
Figure 4.4. Main thiol SAMs used in this work, (a) pentafluorobenzenethiol (b) fluorobenzenethiol.
51
4.7 Kelvin Probe studies
KP measures the contact potential difference (CPD) between a vibrating reference
electrode and a sample surface. The reference electrode and sample are placed in parallel
capacitor configuration with a small spacing of approximately 1mm (Fig. 4.5). The contact
potential between the sample and reference electrode is Vcpd= -((Pi-(p2)/e, where (pi and cp2
are the respective workfunctions. When the two plates have different workfimctions, a
periodic vibration with frequency co gives rise to a current i(t) = ( V dc + V c pd ) coACcos(cût),
where AC is the variation of capacitance over time and Vdc is the applied bias [34]. The
applied bias is changed until the zero point of the current is reached, at which point the
applied potential is the workfunction difference between the sample and tip. KP only returns
the difference in workfunction between two samples, therefore for evaluation of absolute
values a reference sample with a known workfunction must be used. The samples were
fabricated by evaporating or sputtering thin metal layers on glass substrates and in some
cases further surface processing was performed.
actuatorA
sample # V DC
Figure 4.5. Illustration of Kelvin Probe setup.
Results noted as KPl to KP3 were performed at Merck Chemicals with a McAlister
KP6500. KP4 was performed at the London Centre for Nanotechnology (University College
London) with assistance from M. Lazzerini. All measurements were performed in ambient
conditions.
Reference samples used: The workfunction of Au is highly sensitive to the adsorption of
atmospheric contaminants and exhibits high values of 5.2 eV for clean surfaces and values as
low as 4.7 eV for samples after prolonged storage in ambient conditions [99] (reduction with
52
time was also observed in our measurements, as can be seen from KPl-1, K Pl-2). Thus it
cannot be used as a reliable reference for absolute workfunction conversions. For
experiments KPl and KP2 only relative workfunction values are given. For KP3, Cr is used
as a reference with a workfunction of 4.5 eV [100]. For KP4, highly oriented pyrolytic
graphite is used as a reference with a workfunction of 4.6 eV [101]. The surface of the
graphite was peeled off prior to the measurement, providing a reliable clean surface
reference for absolute workfunction conversion.
K Pl Sample Description Relative workfunction (eV) ±0.1 eV
1 1-hour-old Au Reference2 1-day-old Au -0.13 Au + PFBT (1-min treatment) +0.94 Au + PFBT (10-min treatment) +1.15 Au + PFBT (90-min treatment) +1.16 Au + I/KI etch + s. c. +0.57 Au + I/KI etch + s. c. + PFBT +0.5
Table 4.1. Relative workfunctions for 40nm evaporated Au samples on glass substrates with different surface processing. Measurement 1 (KPl). Solvent cleaning is denoted by s. c.
KP2 Sample Description Relative workfunction (eV) ±0.1 eV
1 Au Reference2 Au + Pentafluorobenzene-Thiol +0.83 Au + Fluorobenzene-Thiol +0.64 Au + I/KI etch + s. c. +0.25 Au + I/KI etch + s. c. + PFB-Thiol +0.2
Table 4.2. Relative workfunctions for 40nm evaporated An samples on plastic substrates with different surface processing. Measurement 2 (KP2). Solvent cleaning is denoted by s. c.
The effect of immersion time of evaporated Au thin films on glass into PFBT SAM
solution was examined. It was found that majority thiol adsorption is achieved within 1 min
(KPl-3) to 10 min (KPl-4), and longer immersion had no effect (KPl-5).
The thiol SAMs were expected to significantly increase the workfunction of Au and
result in linear injection as seen in transistor output characteristics. This was not the case
initially, and thus it was decided to perform Kelvin Probe studies and assess the effectiveness
of the SAM treatments.
The Au surface workfunction was measured for a variety of conditions found in the
transistor fabrication process. Most importantly, it was found that following iodine-based
etching and solvent cleaning (H2O, acetone, isopropanol) the thiol had no effect on the
53
workfunction (KPl-6, K Pl-7, KP2-4, KP2-5), while a large increase was measured for
PFBT treated Au without prior etching of 1.1 eV and 0.8 eV (K Pl-4, KP2-2) as compared to
evaporated Au reference. The only difference between KPl and KP2 is the substrate, which
is glass for KPl (also KP3 and KP4) and plastic for KP2 which may be responsible for the
difference, although it falls within the typical workfunction variation of the Au surface.
KP3 Sample Description Absolute Workfunction (eV) ±0.1 eV
1 An as deposited 4.72 An, I:KI etched, s. c., O2 plasma 5.23 Au, etch, s. c., plasma, PFBT 5.64 Au, etch, s. c., plasma, FBT 5.55 Cr as deposited 4.5 (Reference [100])6 Cr, MS 8 etched 4.57 Ni as deposited 4.78 Ni, HChHNOs etched 4.7
Table 4.3. Absolute workfunctions for 40nm evaporated An samples on glass substrates with different surface processing. Measurement 4 (KP4). The workfunction of Cr is assumed to be 4.5 eV [100] to convert to absolute values. Solvent cleaning is denoted by s. c.
KP4A Sample Description Absolute Workfunction (eV) ±0.1 eV
1 Graphite 4.6 (Reference [101])2 Au 4.93 Cr/Au 4.84 Au/FBT 5.35 Au/PFBT 5.7
Table 4.4. Absolute workfunctions for 40nm evaporated An samples on glass substrates with different surface processing. Measurement 5 (KP5). The workfunction of graphite is assumed to be 4.6 eV [101] to convert to absolute values. For fields with two values the second value was measured approximately 10 min later.
The iodine contamination problem was overcome by exposing samples to O2 plasma
before SAM treatment (50W, Im, 30 ml/min) (KP3-2), giving effective workfunctions of 5.6
eV and 5.5 eV for PFBT and FBT treated samples respectively (KP3-3, KP3-4). The
workfunctions of Cr and Ni were found to be unaffected by etching (KP3-5,6,7,8). The
utilisation of a Cr adhesion layer for Au was found to decrease the workfunction by
approximately 0.1 eV as compared to bare Au (KP4A-3, KP4A-2). The difference in
workfunction between PFBT and FBT treated Au was found to be 0.1-0.4 eV (KP3-4, KP2-
3, KP4A-4) and assumed to be on average 0.2 eV. Although the measurement error is only
several meV, there is a significant reproducibility error of ±0.1 eV as the measurements were
performed in ambient conditions and contaminant adsorption could not be avoided.
54
4.8 Schottky barrier engineering
For near-ohmic contacts in FETs, An was treated with PFBT SAM. To fabricate
Schottky source/drain-semicondnctor contacts for SGTs (Chapter 7) we utilised the
workfunction differences between PFBT and FBT SAMs, Au and Cr/Au electrodes, and Au
and Cr electrodes as summarised in Fig. 4.6.
VL-
Cr-
Cr/Au/FBT_
Cr/Au/PFBT - Au/PFBT"
4.5
■4.9 <■5.1■5.3■5.5■5.7
Figure 4.6. Effective workfunctions of different electrodes used for organic SGTs in Chapter 7. Estimates are based on KP measurements, with error ± 0.1 eV.
4.9 Semiconductor
Two semiconductors were mainly used in this work, an indenofluorene-triarylamine
co-polymer (PIFTAA, Fig. 4.7a), and an indenofluorene-phenanthrene co-polymer (PIFPA,
Fig. 4.7b). Two additional materials were used for the charge transport study in Chapter 6,
poly(triarylamine) (PTAA, Fig. 4.7c) and poly(indenofluorene-triarylamine-triarylamine)
(PIFTAATAA, Fig. 4.7d). One additional material was used for Source-Gated Transistors in
Chapter 7 (ADS250BE, Fig. 4.7e). All five conjugated polymers are amorphous and air-
stable (microstructure examined in section 6.3) with field-effect saturation mobilities of
approximately 0.03 cm^/Vs for PIFTAA and 0.3 cm^/Vs for PIFPA, 3-10'^ cm^/Vs for
PTAA, 4-10'^ cm W s for PIFTAATAA, and 1 ■ 10' cm W s for ADS250BE.
PIFTAA, PIFPA, and PTAA were obtained from Merck Chemicals (D. Sparrowe, F.
Meyer). PIFTAATAA was obtained from Imperial College London (S. R. Ashraf, I.
McCullouch, W. Zhang). ADS250BE was purchased from American Dye Source. The
materials were used as received without further purification.
55
To deposit the polymers, the fibres were dissolved in anhydrous toluene in
concentrations of 5-lOmg/ml and solutions were spin-coated at 1000-3000 RPM, followed
by annealing at 100°C for 10 min. Solution preparation was performed under dry nitrogen
and the remaining steps (spin-coating and annealing) were performed in ambient conditions.
Figure 4.7. Chemical structures of (a) poly (indenofluorene-triarylamine) (PIFTAA), (b) poly(indenofluorene-phenanthrene) (PIFPA), (c) poly(triarylamine) (PTAA), (d) poly(indenofluorene-triarylamine-triarylamine) (PIFTAATAA), (e) poly[(9,9-dihexylfluorenyl- 2,7-diyl)-co-(N,N’ bis{p-butylphenyl}-l,4diaminophenylene)] (ADS250BE)
56
The HOMO and LUMO levels are E h o m o = 5 . 4 5 eV and E l u m o = 3 . 1 7 eV for PIFTAA
and E h o m o = 5 . 7 9 eV and E l u m o = 2 . 9 5 eV for PIFPA as reported from Merck Chemicals,
measured by Cyclic Voltammetry (CV). As CV is performed in solution (significantly lower
polymer chain density than in solid state), the measured energy should have significantly
reduced contributions from electronic polarisation energy, but it should have full
contributions from molecular polarisation energy (typical value 0.1 eV, section 2.3.2), thus
Ecv = Ehomo - EpM+. Technical details on the CV measurement as reported by Merck
Chemicals are given in [102]. The reported error was ±0 . 1 eV. The HOMO level for
PIFTAATAA was reported from Imperial College London (no method information) as 5.47
eV. For PTAA the HOMO has been reported at 5.2 eV [103].
As part of KP4, the workfunction of PIFTAA and PIFPA was measured. The
samples consisted of thin films of the conjugated polymers (30 nm) spin-coated on top of
PFBT treated Au and then annealed at 100°C for a few minutes. The Au/PFBT workfunction
is 5.6 eV, which is higher than the measured workfunction for both PIFTAA and PIFPA
(Table 4.5), thus the contact is in Fermi pinning regime (section 3.1.2), and the measured
potential should correspond to the positive polarisation level with electronic and molecular
contributions ( E k p = E h o m o - E p e + - E p m + ) . Therefore the positive electronic polarisation
energy is given by Epe+ = Ecv - Ekp and turns out to be 0.15 eV for PIFTAA and 0.49 eV for
PIFPA with error ± 0.2 eV. These values are on the low end of typical electronic polarisation
energies measured for other conjugated polymers (section 3.1.2). Since the electrode
workfunction is 0.3 eV higher than the workfunction of both conjugated polymers, there may
be downwards band bending which has been shown to cause space charge accumulation in
higher-than-HOMO-workfunction contacts [84,104].
KP4B Sample description Absolute workfunction (eV) ±0.1 eV
1 Graphite 4.6 (Reference [101])2 PIFTAA 5.33 PIFPA 5.3
Table 4.5. Surface workfunction values for Au/PFBT/Semiconductor (30 nm) samples obtained from KP measurements.
57
4.10 Insulator
The insulator mainly used in this work is the fluoropolymer Cytop™ grade
CTL809M (Asahi Glass). It was deposited by spin-coating typically at approximately 2000
RPM with the exact value being slightly tuned depending on the underlying surface to give
thickness of 1 pm, followed by annealing at 100°C for 10 min. The characteristics of Cytop
are perfectly suited to organic transistors: The low relative permittivity ( S r = 2 . 1 ) prevents
energetic variance broadening along the semiconductor-insulator interface, maximising
mobility (section 3 . 2 . 1 ) . Also, Cytop is highly hydrophobic, which is desirable as insulator
hydrophobicity effectively eliminating moisture related degradation: threshold shifts under
prolonged voltage bias and hysteresis (section 2.6). Specifically the combination of PIFTAA
and Cytop has been shown to be stable under prolonged bias [68].
The highly hydrophobic nature of Cytop may be desirable for stability but is
problematic when solution processing must be performed on top, as in the case of bottom-
gate transistors, where the semiconductor solution does not adhere at all on the Cytop
surface. This issue was actually overcome with the same step that resolved the iodine
contamination problem of Au electrodes: exposure to light oxygen plasma, which was not
found to damage the insulator structurally (Fig. 4.8), although evidence on some deep trap
formation and resulting positive shift of the tum-on voltage are found in Chapter 5.
58
(a) 8nm
(e)
t e m m r n
2[im4° d
2|im
8nm
6° (f)2|im
8nm
2|im 2 |im
Figure 4.8. Tapping mode AFM images (left: phase, right: height) of (a,b) as-spun/annealed Cytop surface (rms roughness 0.41 nm), (c,d) surface after Au evaporation, etching, and solvent cleaning (rms roughnes 0.49 nm), (e,f) surface after oxygen plasma ashing at 50W for 30s (rms roughness 0.45 nm).
59
Microposit S1800 G2 series photoresists data sheet ^ W. N. Hansen, K. B. Johnson. “Work function measurements in gas ambient”. Surf. Soi. 316, 373 (1994)
D. E. Eastman. “Photoelectric work functions of transition, rare-earth, and noble metals”, Phys. Rev. B 2 (1970) 1
M. M. Beerbom, B. Lagel, A. J. Cascio, B. V. Doran, R. Schlaf. “Direct comparison of photoemission spectroscopy and in site Kelvin probe work function measurements on indium tin oxide films”, J. Electron Spectrosc. 152, 12 (2006)
Cyclic voltammograms were recorded on a Princeton Applied Research VersaSTAT 4 Potentiostat/Galvanostat using platinum electrodes at a scan rate of 50 mV/s and a Ag/Ag+ (0.1 M of Ag NO3 in acetonitrile) reference electrode in an anhydrous and argon-saturated solution of 0.1 M of tetrabutylammonium tetrafluoborate (BU4NBF4) in acetonitrile. In these conditions the oxidation potential Eoxi/2 of ferrocene was 0.1 V versus Ag/Ag+ whereas the Eoxi/2 of ferrocene was 0.41 V versus SCE. The HOMO and LUMO energy levels were determined from the oxidation and reduction onset of the second scan from CV data taking into account the SCE level at -4.7 eV (A. J. Bard, L. R. Faulkner. Electrochemical methods: Fundamentals and Applications, 2”* ed., John Wiley and Sons Inc., New York (2001))
K. E. Lilja, H. S. Majumdar, K. Lahtonen, P. Heljo, S. Tuukkanen, T. Joutsenoja, M. Valden, R. Osterbacka, D. Lupo. “Effect of dielectric barrier on rectification, injection and transport properties of printed organic diodes”, J. Phys. D: Appl. Phys. 44,295301 (2011)
H. Hwang, E. G. Kim, J. Liu, J. L. Bredas, A. Duggal, A. Kahn. “Photoelectron spectroscopic study of the electronic band structure of polyfluorene and fluorine-arylamine copolymers at interfaces”, J. Phys. Chem. C 111, 1378 (2007)
60
5 Stability and performance of Top- and Bottom-Gate FETs
5.1 Introduction
The main objectives of this work were to compare transistor operating parameters
and long term ambient shelf stability between the Bottom-Contact (BC) Bottom-Gate (BG)
and BC Top-Gate (TG) Field-Effect Transistor (FET) configurations, and determine any
significant effects that are gate geometry specific. Initially, individual BG and TG FETs
were demonstrated, and BG FET performance was significantly inferior to that of TG FET,
presumably because the BG configuration lacked the natural passivation of the top insulator
and gate.
A Dual-Gate (DG) structure was developed utilising the same bottom and top
insulators (to our knowledge this has not been previously demonstrated), which was
achieved by slightly reducing the hydrophobicity of the bottom insulators by exposure to 0%
plasma (section 4.10) to allow for spin-coating of the semiconductor solution. Using the
same top and bottom insulators allows investigation without concern for different charge
transport parameters (permittivity and its effect on PFET mobility (section 3.2). In
combination with the amorphous morphology of the semiconductors PIFTAA and PIFPA
(lack of crystallinity induced variations - section 2.5, the morphology of PIFTAA and PIFPA
are examined in section 6.3), the top and bottom insulator-semiconductor interfaces should
be very similar, allowing for a close investigation of the differences between BG and TG.
DG PFETs were operated exclusively in single gate mode whilst the other gate was left to
float. Individual BG or TG devices are referred to as BG or TG respectively while the BG or
TG components of DG devices are referred to as BG-mode or TG-mode respectively.
5.2 Fabrication and methods
BC TG PFETs were fabricated by depositing an Au layer, patterning source/drain
electrodes by photolithography and etching, cleaning the electrodes and treating with PFBT,
followed by spin-coating of the semiconductor and insulator (Cytop™), and finally
depositing an Au TG. More details on each step can be found in Chapter 4.
BC BG PFETs were fabricated by depositing an Au gate, spin-coating the insulator,
depositing a metal layer, patterning source/drain electrodes by photolithography and etching,
cleaning the electrode and insulator surface and treating with a thiol SAM, and drop-casting
the semiconductor.
61
BC DG PFETs (Fig. 5.1) were made by combining the above, except that the
semiconductor was spin-coated. Also, for TG and BG PFETs, the source/drain electrode
patterning process did not include oxygen plasma ashing of the electrodes and this was found
to have a significant effect in perfonnance as etchant based contamination prevented the
adsorption of thiol SAMs (explained in section 4.7).
Top Gate
Insulator
Sem iconductor
Insulator
Substrate
Figure 5.1. Illustration of structure of Dual-Gate transistors. Dimensions are not to scale.
Electrical measurements were perfonned with a Keithley SCS-4200 semiconductor
parameter analyser and Karl Suss microprobes. All fabrication, storage, and characterisation
steps were perfonued in ambient conditions.
5.3 Performance comparison: TG and BG PFETs
The PIFTAA TG PFETs performed well, with near zero Vqn, ON/OFF ratio of 10\
and field-effect saturation mobility of 0.03 cm^/Vs, calculated with eq. 3.16 from transfer
characteristics (Fig. 5.2a). The output characteristic was non-linear at low drain voltages
(Fig. 5.2c). The TG PFETs were highly stable in ambient conditions, and the same devices
measured 20 days later showed no hysteresis and small mobility degradation (down to 0.024
cnf/Vs), thus demonstrating overall stable performance (Fig. 5.2b,d).
62
10-10'
< 1 0 '-? 1 0 '
10'10"'10"'
— V Q—-80V • —Vq—-10 V
TG Day 1 nSAi= 0.030 cm^/Vs
(C)
-80 -60 -40 -20 0 20 40Vg (V)
(b) l o SlO'i10'"i10'1
< 10'1_Q lo ' i
lO-'ilo ' i
10"'iin"'
— V Q—-60 V— V Q—-10 V
TG Day 20 nSAT= 0.024 cm"A/s
-60 -40
80 70 60
^ 5 0 a 40
_9 30 20 10 0
(d)
-20 0Vg (V)
20 40
-60VVg OV to -60V (-10V step )
TG Day 1
80 70 60 50
1 40 3 30
20 10 0
Vg OV to -60V (-1 OV step)-60V
TG Day 20
-60 -50 -40 -30 -20 -10 0Vd (V)
-60 -50 -40 -30 -20 -10 0Vd (V)
Figure 5.2. (a) Transfer (forward scan only) and (c) output characteristics of PIFTAA TG
PFETs 1 day after fabrication and (b) transfer characteristic (forward and reverse) and (d)
output characteristic 20 days after fabrication. Channel length 10pm, width 1cm, insulator
thickness 1pm, PIFTAA thickness approximately 100 nm. Source/drain electrodes were treated
with PFBT but treatment was ineffective as indicated by non-linearity of output characteristics
at low drain voltages.
BG PFETs on Cytop™ bottom insulator were fabricated also without exposing the
source/drain electrodes to O2 plasma. The insulator surface was highly hydrophobic,
preventing spin coating of the semiconductor which was instead deposited by drop casting.
Performance was found to be significantly different than TG, with ON/OFF ratio of 10 ,
negative shift of V q n , reduced mobility by a factor of 1 0 (0.003 cm^/Vs), high subthreshold
swing, and high hysteresis (Fig. 5.3a). The mobility measured 12 days later had degraded to
8 -10'" cm'/Vs (Fig. 5.3b) (although a different device was measured). The BG PFETs show
more pronounced non-linearity in output characteristics at low drain voltage as compared to
TG, consistent with reports of higher susceptibility of co-planar configurations to Schottky
barriers as compared to staggered configurations (section 3.3). Between the fabricated TG
and BG PFETs, the main difference is the lack of natural passivation offered by the top
insulator and gate as in the TG configuration.
63
V Q—-GOV ' Vp—-1 OV
BG Day 1
|isAT=0-003 cm A/s
— Vd=-50V — Vd=-15V
BG Day 12
HsAT^O.0008 cm /Vs
-40 -20Vg (V) (d)
Vg OVto -60V (-10V step)0.8 -60V
0.6BG Day 1
a 0.4-
0.2
0.060 -50 -40 -30 -20 -10 0
1Q
0.3
0.2
0.1
0.0
-60 -40 -20 0 20Vg (V)
Vg OV to -50V (-10V step)
BG Day 12
-50 -40Vd (V)
-30 -20Vd (V)
-10
Figure 5.3. (a) Transfer and (c) output characteristics of PIFTAA BG PFETs 1 day after
fabrication and (b) (d) 12 days after fabrication. Channel length (a) (c) 20pm, (b) (d) lOpm,
width 1cm, insulator thickness Ipm. PIFTAA thickness not measured (presumably a few
hundred nanometers). Source/drain electrodes were treated with PFBT but treatment was
ineffective.
5.4 Dual-Gate PFETs
At this point two issues were addressed. Firstly, KP studies were undertaken to
improve injection from source/drain electrodes. The KP studies (Section 4.7) indicated
contamination from the PKPHiO etchant and exposure to O2 plasma was found to resolve
the issue without measurably damaging the surface of the Cytop™ bottom insulator (Section
4.10). Secondly, the BG and TG structures were combined into a DG structure to resolve the
lack of passivation for BG.
DG PFETs were fabricated with PIFTAA and PIFPA, with channel length 10pm,
width 1cm, insulator thickness 1pm and semiconductor thickness approximately 100 nm.
PIFTAA DGs performed similarly for both TG mode and BG mode. The exposure of
64
source/drain electrodes to O2 plasma has resolved the Schottky barrier issue that resulted
from ineffectiveness of the SAM treatment (Fig. 5.4a,c). Although Vqn for TG mode was
similar to TG PFETs (approximately V o n = -2 V ), for BG mode a positive shift was observed
(V o n = 5 V ) (Fig. 5.4b,d). As oxygen is known to p-dope organic semiconductors (section
2.6), the oxygen plasma treatment and residual species on the bottom insulator are most
likely to be responsible. The subthreshold swing for TG mode is approximately 0.3 V/decade
while for BG mode it is 1 V/decade. This difference cannot be attributed solely to gate
geometry as the presence of oxygen has been shown to increase the subthreshold swing in
pentacene FETs (section 2.6). The mobilities were similar for both gate modes, with 0.03-
0.04 em^/Vs for TG mode and 0.04-0.05 cm^/Vs for BG mode. The lower mobility of TG
mode is likely an underestimation caused by access resistance. The mobility was calculated
in the linear and saturation regime, and the mobility profiles with gate voltage (Fig. 5.4b,d)
are in the corresponding regime for all gate voltages. The mobilities as a function of gate
voltage fit a clear parabolic profile for BG mode, while somewhat weaker gate voltage
dependence is observed from TG mode. The gate voltage dependence of the mobility is
characteristic of the shape of the Density of States (DOS) and is thus consistent with a
Gaussian DOS (discussed in more detail in chapter 7).
Id (Vq=-30V)Vg OV to -30V (-10V s tep )
-30V-4 j — Id (Vd=-2V)
T 10 PIFTAA TG m ode
PIFTAA TG m ode
0.02 ^
-30 -25
(c)-20 -15 -10
Vd (V)-20 -10
Vg (V)
50
40
30
20
10
0
Vg OV to -30V (-1 OV step )
-30 V
PIFTAA BG m ode
-30 -25 -20 -15 -10Vd (V)
-5
Id (Vq=-30V) — Iq (Vd=“2V) — H lin
- PIFTAA m o d e
0.03 EÜ
-20 -10Vg (V)
Figure 5.4. Output and transfer characteristics of PIFTAA DG PFET on (a) (b) TG mode and
(c) (d) BG mode. In transfer characteristics the drain current is represented by thick lines (left
65
y-axis) and the mobility by thin lines (right y-axis). In transfer characteristics red lines indicate
saturation regime and black lines indicate linear regime.
For PIFPA DG PFETs (Fig. 5.5), Vqn did not undergo positive shift for BG mode.
This may be due to the increased IP (5.79 eV for PIFPA as compared to 5.45 eV for
PIFTAA) resulting in improved oxidation stability. For the subthreshold swing two distinct
regimes are seen. For low gate voltages the value is 0.3 V/dec for TG mode and 0.7 V/dec
for BG mode, and for higher gate voltages the value is approximately 1.8 V/dec for TG and
2V/dec for BG. As the lack of positive shift for BG mode Vqn indicates higher resilience to
residual oxygen species, the subthreshold swing is likely to be inherently lower in staggered
configurations such as BC TG. The high subthreshold swing at higher gate voltages for
PIFPA may be indicative of an effective density of states tail extending into the band gap.
The OFF current is higher for PIFPA DG than for PIFTAA DG by approximately a factor of
10. If the semiconductors were intrinsic, the opposite would be expected, since PIFPA has a
significantly higher band gap (2.84 eV) than PIFTAA (2.28 eV, section 4.9). The mobilities
exhibit the same parabolic profile and identical values for both gate modes of 0.22 cm^/Vs
(peak field-effect saturation mobility).
D (Vd=-30V) In (Vn=-2V)Vq OV to -30V (-1 OV step)
-30VPIFPA TG m ode
PIFPA TG m ode
< 60
•30 "25 "20 "15 "10 “5V n (V )
Vg OV to -30V (-1 OV step )
-30V
-20 -10 V g (V )
— Id (Vq=-30V)— Id (Vq—"2V)
PIFPA BG m ode
0 10
— p s a t— p lin 0.3
PIFPABG m ode 0 .2 ^
Eu
0.1 -
-30 -25 -20 -15 -10 -5V d (V )
20 -10V g ( V )
Figure 5.5. Output and transfer characteristics of PIFPA DG PFET on (a) (b) TG mode and (c)
(d) BG mode. In transfer characteristics the drain current is represented by thick lines (left y-
axis) and the mobility by thin lines (right y-axis). In transfer characteristics red lines indicate
saturation regime and black lines indicate linear regime.
66
PIFPA TG mode and BG mode output charaeteristics exhibit slightly non-linear
eurrent-voltage behaviour at low drain voltages (Fig. 5.5a,c). Also, the linear mobility (0.1
cm^/Vs) is noticeably lower than the saturation mobility (0.22 em^/Vs). The difference
between the linear and saturation mobility is higher than for PIFTAA (approximately 0.035
cm^/Vs and 0.045 em^/Vs respectively), and PIFTAA output characteristics show no non-
linearity at low drain voltages.
No hysteresis was found for BG or TG mode devices (Fig. 5.6). The hysteresis
characteristics were taken on day 0 (fabrication day) and the OFF current for PIFTAA BG
was unusually high (Fig. 5.6a) but was found to drop to similar levels to other FETs from the
next day onwards (as seen on Fig. 5.4d).
PIFTAA TG m ode— Vp—-30V— Vp=.2V
PIFTAA BG m ode
< 10
10Vp=-30V 10PIFPA BG m ode P FPA TG m odeVp=-2V
^ 10
Vp—-30V — Vp=-2V
-20 -10Vg (V)
-20 -10Vg (V)
Figure 5.6. Forward and reverse PFET transfer characteristics for (a) (b) PIFTAA and (c) (d)
PIFPA, for (a) (c) BG mode and (b) (d) TG mode.
67
5.5 DG PFET Ambient Stability
Fig. 5.7 illustrates changes of Vqn for TG and BG mode in the linear and saturation
regimes over a characterisation period of 143 days (PIFTAA) and 62 days (PIFPA). No
appreciable change over time was observed.
{a)8 6 4
> 2 | o
-2 -4
PIFTAA
« A A BG
I A TG
0 20 40 60 80 100120140160 Day
(b) 2
1
0> -1>°-2
-3
-4
PIFPA
BG
TG
0 20 40 60 80 100120140 160 Day
Figure 5.7. Progression of turn-on voltage Vqn for (a) PIFTAA and (b) PIFTA over time for BG
mode (black symbols) and TG mode (blue symbols).
The linear and saturation mobilities for PIFTAA and PIFPA TG and BG mode were
extracted over the respective time periods. The degradation over time is similar for saturation
and linear mobilities, which indicates that the source/drain-semiconductor contacts are
stable. The pronounced initial degradation for PIFTAA TG mode (Fig. 5.8b) may suggest
that the source of the degradation is the bulk of the semiconductor. All devices undergo
increased degradation for the first two weeks and somewhat stabilise afterwards. This may
be related to residual solvent evaporation from the semiconductor bulk as the PFETs were
not annealed for particularly long (-10-15 min). The mobility degradation rates for both gate
modes after the initial degradation seem to be similar as opposed to individual TG and BG
PFETs.
68
(a)0.05
_ 0.04n>g 0.03o
0.02
0.01
PIFTAA BG (b)0.040
X X Saturation 0.035 • . g 0.030
i ' s 0.025 Linear r o.020
■ X0.015
PIFTAA TG
: A A
Saturation
Linear
(c),0.25
0.20
0.15
0.10
0.05
0 20 40 60 80 100120140160Day (d)
PIFPA BG Saturation
Linear
0 20 40 60 80 100120140160 Day
0.25
0.20
0.15
0.10
0.05
0 20 40 60 80 100120140160 Day
PIFPA TG• »% . . Saturation
Linear
0 20 40 60 80 100120 140 160 Day
Figure 5.8. Progression of linear (black symbols) and saturation (red symbols) mobility over
time for (a) (b) PIFTAA and (c) (d) PIFPA in (a) (c) BG mode and (b) (d) TG mode.
Fig. 5.9 displays the approximate degradation rate of the mobility per month (1-
Pfinai/pinitiai)/month for pentacene [105] (IP 4.9 eV [106]), oligothiophenes (5.2 eV for
sexithiophenes 6T and 5.4 eV for quaterthiophene 4T) [107], polythienothiophenes (pBTCT
5.3 eV, pATBT 5.1 eV, pBTTT 5.05 eV) [108], PTPA [109] (5.36 eV [110]), F8T2 [109]
(5.46 eV [111]), PIFTAA (5.45 eV) and PIFPA (5.79 eV) as a function of approximate IP.
The transistor data from literature was obtained for devices exposed to ambient conditions
without passivation and most of them experienced threshold shifts of several volts. Although
the trend is not clear, degradation rates seem to be lower for higher IPs. PIFTAA and PIFPA
show very low degradation rates, especially considering their high mobility values as
compared to another stable polymer PTPA with field-effect mobility on the order of 10’
cnf/Vs).
69
cE 100:
CII 1“0)■o
.QO
■ 6T •pBTTT
■ pATBT • Dec-4T-Dec ■ Pentacene ■ F8T2
■ pBTCT
Eth-6T-Ethî , PIFTAADec-6T-DeclHex-6T-Hex ■ PTPA
■ PIFPA
' I ....' I > 1.... '..... I • I ' I5.0 5.2 5.4 5.6 5.8 6.0
Ionization Potential (eV)
Figure 5.9. Mobility degradation rate (%) per month as a function of ionisation potential for above references and PIFTAA and PIFPA.
5.6 Conclusions
Amorphous conjugated polymer DG PFETs with the same top and bottom insulators
were successfully fabricated, allowing for a close comparison of configuration effects
without concern for different charge transport parameters (permittivity and its effect on
PFET mobility (section 3.2.1) or microstructure variations (section 2.5).
The use of oxygen plasma ashing on the bottom insulator and electrodes was found
to affect PIFTAA BG mode in terms of Vqn positive shift (TG mode Vqn = - 2 V, BG mode
Von = +5 V) indicating the presence of residual oxygen species on the bottom insulator
surface and resulting deep trap formation. PIFPA BG mode did not exhibit Vqn positive shift
( V q n = -1 V for both gate modes), which may result from the increased IP. V q n for both
semiconductors and gate modes were stable without any measurable shift over time. Also, no
appreciable hysteresis is observed for any FET, indicating semiconductor-insulator interfaces
free of mobile ions.
The subthreshold swing is found to be lower by approximately a factor of 2 to 3 for
TG mode of both materials (PIFTAA TG mode 0.3 V / dec, BG mode 1 V/ dec, PIFPA TG
mode 0.3 V / dec, BG mode 0.7 V / dec) and may be related to the larger electrode
semiconductor contact area for the staggered configuration. At higher gate voltages the
subthreshold swing for PIFPA is similar for both gate modes (TG mode 1.8 V / dec, BG
mode 2 V / dec) which may be the result of a shallow trap level tail extending into the
bandgap, inherent to the conjugated polymer. Additionally, the intrinsic voltage gain is not
found to differ between gate modes and conjugated polymers.
70
PIFTAA DG FETs exhibit linear eurrent-voltage relationship in output
characteristics at low drain voltages, and the linear and saturation mobilities are similar
(0.035 cm W s and 0.045 cm W s respectively). PIFPA DG FETs exhibit slightly non-linear
relationship, and the linear and saturation mobilities are more different (0.10 cm W s and
0.22 cm W s respectively), indicating an electrode-semiconductor contact issue.
The mobility degradation rates were slightly pronounced for PIFTAA TG mode but
overall were found to be very similar between gate modes, and between linear and saturation
regimes, indicating the degradation originates in the semiconductor bulk or insulator-
semiconductor chaimel and not the source/drain-semiconductor contacts. The long term
mobility degradation rate in ambient conditions is amongst the lowest reported for organic
transistors and seems to exhibit an inverse trend with IP (Fig. 5.9).
In general, performance and long term ambient stability between BG mode and TG
mode examined for PIFTAA and PIFPA was demonstrated to be similar as opposed to
individual TG and BG PFETs, indicating that the top passivation for TGs by the top insulator
and gate are the main source of long term stability. PIFTAA and PIFPA in combination with
Cytop™ are amongst the most air stable organic semiconductor hole transporting systems.
71
s. Cipolloni, L. Mariucci, A. Valletta, D. Simeone, F. De angelis, G. Fortunato. “Aging effects and electrical stability in pentacene thin film transistors”. Thin Solid Films 515, 7546 (2007)
F. Li, P. Graziosi, Q. Tang, Y. Zhan, X. Liu, V. Dediu, M. Fahlman. “Electronic structure and molecular orientation of pentacene thin films on ferromagnetic Lao.ySrojMnOs”, Phys. Rev. B 81, 205415 (2010)
Q. Xia, M. Burkhardt, M. Halik. “Oligothiophenes in organic thin film transistors - Morphology, stability and temperature operation”, Org. Electron. 9, 1061 (2008)
I. McCulloch, M. Heeney, M.L. Chabinyc, D. DeLongchamp, R.J. Kline, M. Colle, W. Duffy, D. Fischer, D. Gundlach, B. Hamadani, R. Hamilton, L. Richter, A. Salleo, M. Shkunov, D. Sparrowe, S. Tierney, W. Zhang. “Semiconducting thienothiophene copolymers: design, synthesis, morphology, and performance in thin-film organic transistors”. Adv. Mater. 21, 1091. (2009)
H. Kempa, K. Reuter, M. Bartzsch, U. Hahn, A. C. Huebler, D. Zielke, M. Forster, U. Scherf. Stability study of all-polymer field-effect transistors. IEEE Polytronic Conference (2005)
K.M. Yeh, C.C. Lee, Y. Chen. “Poly(4-vinyltriphenylamine): Optical, electrochemical properties and its new application as a host material of green phosphorescent Ir(ppy)3 dopant”, Synth. Met. 158, 565 (2008)
K. Yonezawa, M. Ito, H. Kamioka, T. Yasuda, L. Han, Y. Morimoto. Advances in Optical Technologies, ID 316045 (2012)
72
6. Morphology and charge transport of amorphous conjugated
polymers
6.1 Introduction
This chapter investigates Field-Effect Transistor (FET) performance for four
different amorphous polymer semiconductors: poly(triarylamine) (PTAA),
poly(indenofluorene-triarylamine) (PIFTAA), poly(indenofluorene-triarylamine-
triarylamine) (PIFTAATAA),and poly(indenofluorene-phenanthrene) (PIFPA). The
chemical structures can be found in section 4.9. Amorphous microstructures are practically
advantageous for the low-cost, straightforward solution-déposition of conjugated polymer
thin films (Section 2.5). The mobilities of PTAA and PIFTAATAA are on the order of 10'
cm W s which is typical for amorphous polymers, while the mobilities for PIFTAA and
PIFPA are on the order of 10' cm W s and 10' cm^/Vs respectively and are comparable to
polycrystalline conjugated polymers such as P3HT and pBTTT. Such high mobilities for
amorphous conjugated polymers are unusual and cannot be explained by simple structure-
performance relationship, and thus require detailed investigation.
In the first part of this chapter the microstructure of the aforementioned polymers is
examined by Atomic Force Microscopy (AFM), Grazing Incidence Wide Angle X-ray
Scattering (GIWAXS), and Differential Scanning Calorimetry (DSC) (only for PIFPA).
AFM was performed at the University of Surrey. The DSC data was provided by D.
Sparrowe (Merck Chemicals). For GIWAXS measurements, initially a PIFPA sample was
fabricated and sent to the Technical University of Denmark, Riso, (DTU) where
characterisation was performed by Y. Gu and M. M. Nielsen. A second batch with samples
of all four polymers was sent to the Norwegian University of Science and Technology,
Trondheim, (NTNU) where characterisation was performed by H. Becker, K. Hoydalsvik,
and D. W. Breiby. At the time of writing of this chapter the final report from the second
batch had not yet been delivered, and only preliminary results were available.
The second part of this chapter is based on the characterisation of the charge density,
electric field, and temperature dependence of the FET mobility, and fitting to the Gaussian
Disorder Model (GDM) (Section 2.4.1). In addition, the contact and channel resistances are
extracted by utilising the Transmission Line Method (TLM) on multiple channels, and the
effects of contact quality on the aforementioned dependences is discussed. Variable electric
field characterisation was achieved by fabricating FETs of different channel lengths (Section
45, Fig. 4.3). The FETs were measured in vacuum over a range of temperatures. Variable
73
temperature measurements were performed at the physics department of the Queen Mary
University of London with technical assistance from G. Adamopoulos.
The Gaussian Disorder Model (GDM) (Section 2,4.1) returns four parameters which
can be used to build a picture of charge transport in disordered materials: Firstly, the pre-
factor mobility p o ( F —>0,T—»oo) which is the mobility in the complete absence of disorder and
is an indicator of wave function overlap between neighbouring sites, exhibiting highest
values of 1 cm W s for pure molecular crystals [50]. Secondly, the energetic variance o of the
Gaussian DOS which characterises energetic disorder along the semiconductor-insulator
interface, with smaller values leading to more efficient intersite hopping and thus mobilities.
Thirdly, the variance E characterises positional disorder, although it is dimensionless and
does not correspond to any physical quantity and can thus be used only comparatively.
Finally, C is a numerical constant that characterises intersite distance and takes value of
2.9 10" cnW^^^ for a distance of 0.6 nm (typical distance in molecular crystals such as
pentacene) [112].
6.1 Fabrication and Methods
For tapping mode AFM (Section 4.2), thin films of the conjugated polymers with
approximately 40 nm thickness were spin-coated on glass substrates from anhydrous toluene
solution and annealed at 100°C for 10 min.
For the DTU GIWAXS measurement, 100 nm thick PIFPA was spin-coated on top
of a Si/Si02 substrate and annealed at 100°C for 10 min.
For the NTNU GIWAXS measurement, thin films of 30-40 nm thickness of the
conjugated polymers PTAA, PIFTAA, PIFTAATAA, and PIFPA were spin-coated on
Si/Si02/Cytop™ substrates. The Cytop™ layer was approximately 500 nm thick and was
utilised because the native Si0 2 surface is known to significantly degrade the performance of
organic semiconductors due to reduced crystallisation and other effects, unless it is
passivated typically with Self-Assembled Monolayers (SAMs) or thin films of hydrophobic
organic insulators [113]. The Cytop™ layer was exposed to light O2 plasma to reduce the
hydrophobicity of the surface and allow for spin-coating of the conjugated polymers (Section
4.10). No evidence of crystallinity were found for Cytop™ by AFM (Section 4.10). A
reference sample of Si/Si02/Cytop™ was provided for the NTNU measurement to account
for any diffraction peaks from Cytop™.
For GIWAXS, scattering from the Si02 substrate was suppressed by setting the
sample surface at an angle of 0.18° to the X-ray beam, smaller than the critical angle
(approx. 0.22° for Si/Si02) for total external reflection from the sample substrate. The
74
scattered intensity was recorded on 2D X-ray photon-sensitive image plates. The results were
obtained with Cukh radiation corresponding to 1.54Â wavelength. For the PIFPA DTU
measurement the sample was exposed for -14 hours to enable the detection of very weak
signals. For the NTNU measurements samples were exposed for 2 - 3 hours. The signal was
mapped into reciprocal space coordinates relative to the sample substrate orientation, and
signal intensity was integrated azimuthally to give signal intensity versus scattering vector
plots.
FETs were fabricated in the bottom-contact (BC) top-gate (TG) configuration with
Cr/Au source/drain electrodes (1 nm, 30 nm respectively) for PTAA, PIFTAA, and
PIFTAATAA, and Ti/Au source/drain electrodes (1 nm, 30 nm respectively) for PIFPA. The
performance for PIFPA FETs with Ti/Au was found to be very similar to Cr/Au electrodes
(discussed later on). The polymer semiconductors and Cytop™ insulator were spin-coated
(thickness 40 nm and 1pm respectively). An gates (thickness 30 nm) were deposited and
patterned by evaporation, photolithography, and etching. More details can be found in
chapter 4. All FETs have a channel width W = 1 mm. The measured channel lengths in pm
were: PTAA (10, 20, 40, 60, 80, 100, 120, 140, 160, ISO}, PIFTAA and PIFTAATAA {5,
10, 20, 40, 60, 80, 100, 120, 140, 160, 180, 200}, PIFPA (3, 5, 10, 20, 40, 80, 180}. The
measured temperatures in Kelvin were: PTAA, PIFTAA, PIFTAATAA {200, 230, 260, 290,
310, 330, 350}, PIFPA {100,130,160,190, 220, 250,280, 310, 340}.
Variable temperature measurements were performed in a LakeShore probe station
with a Keithley SCS-4200 semiconductor parameter analyser. The temperature on the
sample holders was monitored with a LakeShore temperature controller. The measurements
were performed in vacuum (10^ mbar) in dark. The sample holder was cooled to the lowest
temperature and when the temperature reached ± 1°K of the desired temperature (accounting
for overshoot) the system was left to reach thermal equilibrium for approximately 1 hour
before starting the electrical measurements. For subsequent measurements with increasing
temperature, 10-15 mins were allowed for thermal equilibration counting from the time at
which ± 1°K was reached (again accounting for overshoot). The electrical measurement time
for a set of transistors was approximately 30 - 40 mins for one temperature point (not
including thermal equilibration time).
6.2 Microstructure assessment
During measurements of the conjugated polymer surface with tapping mode AFM,
some coalescence was observed for PIFPA (Fig. 6.1a), while the surfaces for PIFTAA,
75
PTAA, and PIFTAATAA were relatively featureless (Fig. 6.1b-d). The RMS roughness of
all samples was very low (height scale 3 nm).
PIFPA 3nm (b) PIFTAA 3nm
2^m
PIFTAATAA 3nm (d) PTAA 3nm
2^m 2pm
Figure 6.1. Tapping mode AFM height images of surfaces of conjugated polymers (a) PIFPA, (b) PIFTAA, (c) PIFTAATAA, (d) PTAA.
GIWAXS measurements on the Si/Si02/Cytop™ reference show mainly highly
diffuse scattering as well as several sharp peaks (highlighted with red circles, Fig. 6.2a).
These sharp peaks appear for the rest of the samples and since they originate from the
reference they are ignored. For the samples coated with the conjugated polymers, the profiles
show a broad ring centred at Q = 1.1 À"' (Fig. 6.2) which corresponds to a distance of 5.7 A,
which is in the typical range of ti-ti stacking in organic molecular crystals. This first ring is
slightly more intense towards the vertical scattering direction with respect to the substrate
surface. A low intensity outer ring is also observed for all conjugated polymers at Q = 2.68
A ' which corresponds to a distance of 2.34 A, the origin of which is not know. In general,
while the measurement range does not allow us to rule out the existence of small (< 5 nm)
crystalline domains, the lack of any distinct diffraction peaks strongly suggests that the
76
conjugated polymers can be considered as amorphous. The DTU GIWAXS measurement for
PIFPA is in slightly different range. While the Qxy = 2.68 Â"’ ring is outside the
measurement range, another ring is observed at Q = 0.4 Â (Fig. 6.3a,b) corresponding to a
distance of 15.7 Â, however it is too weak and broad to be taken as evidence of crystallinity.
The lack of crystallinity for PIFPA is also supported by the lack of detectable heat flow
peaks in the DSC characteristic (Fig. 6.3c).
(a)3
0^ 2 tH
O l
0-3 -2
Substrate
-1 0 .1 Q xy (1/A)
N (c)P FTAA P FRA
2 - 1 0 .1 Qxy (1/A)Qxy (1/A)
PTAAPIFTAATAA
0.03 og“2
0.02
1 0 . 1 Q xy (1/A)
1 0 . 1 Q xy (1/A)
Figure 6.2. GIWAXS signal remapped into cylindrical reciprocal space coordinates for (a)Si/Si02/Cytop substrate, and substrates coated with (b) PIFTAA, (c) PIFPA, (d) PIFTAATAA, (e) PTAA.
77
-2.5 -2
(b)
-0.5 C 0.5Qxy(1/Â)
(c)
IIU_= 0.5 0.0-sr
0.0 -0.50.0 0.5 1.0 1.5 2.0 2.5 50 0 50 100 150 200 250
Q(Â- Temperature (C)
Figure 6.3. (a) GIWAXS signal remapped into cylindrical reciprocal space coordinates for Si/SiOi/PIFPA. (b) GIWAXS signal intensity versus scattering vector, (c) DSC characteristic of PIFPA
6.3 FET Characteristics, contact quality and resistance
The transfer characteristics of FETs with PTAA, PIFTAATAA, PIFTAA, and
PIFPA can be seen in Fig. 6.4. The FETs exhibit lower current than PIFTAA and PIFPA
FETS in chapter 5 due to the reduced channel width and thus ON current (W = 1 cm in
chapter 5, W = 1 mm here). Also, for FETs in this chapter the top gate was patterned by
evaporation, photolithography, and etching, and the harsher processing resulted in increased
gate leakage. FETs are characterised in the linear and saturation regimes. The gate leakage in
the saturation regime is negligible. Errors for evaluating mobilities due to gate leakage in the
linear regime are presented where appropriate. Also, PTAA FETs exhibit high OFF cuiTent
even in vacuum conditions, possibly due to irreversible oxidation in ambient conditions or
unintentional residual impurities from synthesis. All the data presented in this chapter was
78
measured in vacuum in dark. It is noted that the mobility for all materials dropped by
approximately 10% as compared to measurements in ambient conditions.
10
10
< 1 0
“ 10"
10"
10"
PTAAT=290KL=10pm
— Id (Vd“"38V)— Iq (Vd“ “6V)— Iq (Vd=-38V)— Ig (Vd“"6V)
I
(b)10
10"
io"
< 10^,-1010
10-12
-40 -30 -20 -10 Vg (V)
10 2010
— Iq (Vq—-38V)— Id (Vd” "6V)— IG (Vd=-38V)— IG (V D=-6V)
PIFTAATAA T=290K L=10pm
-40 -30 -20 -10 0Vq (V)
10 20
(C)10
10
10'
— Id (Vq=-38V) — Iq (V0=-4OV)— Iq (Vd="6V)
io" 1 — Iq (Vd“"10V)— Ig (Vd““38V) — Ig (V0=-4OV)— Ig (Vd”"3V) 10"] — Ig (VD—"1OV)
< 10
10
10-12PIFTAAT=290KL=10pm
10-40 -30 -20 -10 0
Vg (V)10 20
< 10 “ 10"
10"
10"
PIFPAT=280KL=10pm
“50 -40 -30 -20 -10 0 Vg (V)
10 20
Figure 6.4. Transfer characteristics for FETs with W=lcm, L=10fim, at T=290K (280K for PIFPA), measured in vacuum in dark.
The output characteristics of FETs can be seen in Fig. 6.5. Ohmic contacts are
observed for PTAA, PIFTAA, and PIFTAATAA, while the current for PIFPA does not
increase until a drain voltage of -5 V. The effective workfunction of Cr/Au/PFBT (used for
PTAA, PIFTAA and PIFTAATAA) was measured as 5.5 eV (section 4.7). While the
effective workfunction of Ti/Au/PFBT has not been measured, a significant deviation is
unlikely. The HOMO energy minus the positive polarisation energy for PIFPA was
measured as 5.3 eV (Section 4.9). Since the electrode effective workfunction is higher than
Ehomo-Ep+ and the contact is not ohmic, the vacuum levels (VL) between the electrode and
semiconductor must be misaligned. A possible explanation would be the fonuation of a trap
level along the Ti/PIFPA interface at hole energy above Ehomo-Ep+ that results in a Schottky
barrier for hole injection. Similar behaviour for PIFPA with delayed onset in output
characteristics has been regularly observed with Cr adhesion layer (not shown). To date, the
only successful near-ohmic contacts with PIFPA have been established by not using an
79
adhesion layer (Chapter 5, Fig. 5.5), although a slight non-linearity is present which indicates
a barrier due to the fonuation of a dipole between PFBT and PIFPA.
(a)0.12
< 0 .0 8
Ip_P 0.04
0.00
PTAA T=290K L=10pmVg OV to -40V
(-5V step )
(b)0.20-
0.15
?0.10
_p0.05
0.00
PIFTAATAA T=290K L=10^nu
-40V Vg OV to -40V(-10V step )
-40 -30 -20 -10Vd (V)
(c) (d)2.5 _ P I F T A A T=290K L=10pm
2.0 - 4 0 V " ' \ ^ VG OV to -40V8
? 1.5-(-10V step)
_o 1.0 __p 4-_p Q
0.5 2
0.0 0
-40 -30 -20 -10 0Vd (V)
PIFPA T=280K L=10nm
-40V" ^ VG OV to -40V -10 V step )
-40 -30 -20 -10V d (V)
-40 -30 -20 -10Vd (V)
Figure 6.5. Output characteristics for FETs with W =l, L=10, at T=290K (280K for PIFPA), measured in vacuum in dark.
The output characteristics were measured for multiple channel lengths and the total
resistance Rtotai was extracted at a drain voltage in the linear regime V d = -4 V for PTAA,
PIFTAATAA, and PIFTAA, and V d = -1 2 V for PIFPA. The total resistance values were then
plotted against channel length L (Fig. 6.6), for gate voltages corrected for threshold voltages
extrapolated from the currents in Fig. 6.7. A linear fit was applied to Rtotai over L, and was
found to fit the data points well which indicates that the contact resistance is not field
dependent. The contact resistance Rc was extrapolated at L = 0 pm and channel resistance
Rchan was extracted from the slope. Rc and Rchan were plotted against Vg-Vt (Fig. 6.6e and f
respectively). Surprisingly, PIFPA FETs exhibit the lowest contact resistance along with
PIFTAA FETs of approximately 2 Mflcm at Vg-Vt= -30V, while PTAA and PIFTAATAA
exhibit almost one order of magnitude higher values. The contact resistance decreases with
the magnitude of the gate voltage and the dependence seems to be similar for all conjugated
polymer FETs. The channel resistance is lowest for PIFPA at approximately 300knciu/pm at
Vg-Vt = -30 V, 600 kflcm/pm for PIFTAA, and approximately one order of magnitude
80
higher for PTAA and PIFTAATAA, fitting an inverse trend with the FET mobilities of the
conjugated polymers as expected.
(a)6x10® 1
5x10®
c 4x10®
I 3x10®
2x10®OK 1x10®
0-
(c)2.0x10"
^ 1 .5 x 1 0 ■
âi.oxio"
0 5.0x10® on
0.0
(e)
10
10
I , . 'é ’
10®
10'
PTAA T=290K V q -Vj =-12V
50 100 150L (nm)
200
50 100 150L (nm)
200
PIFTAAPTAAPIFTAATAAPIFPA
1
(b) ,
3 .0 x 1 0
2.5x10®
s ' 2.0x10® o2 1.5x10®
■§ 1.0x10®
^ 5 .0 x 1 0"
0.0
PIFTAATAA T=290K VQ -Vy=-15V
(f)
Ea
10'
10
10"
-30 -20 "10 0Vg -Vt (V)
10
O10-
10®
50 100L (^im)
150 200
(d )PIFTAA T=290K Vq -Vj =-18V 8x10®i
- 7x10®6x10®
■ ?U 5x10®■ 2 4x10®
o 3x10®■ 2 2x10®
.■ 1x10®0
PIFPA T=280K Vq -V-p=-16V
50 100LC^m)
150 200
PIFTAAPTAAPIFTAATAAPIFPA
“30 -20 -10 0V g - V t ( V )
10
Figure 6.6. (a-d) Total output resistance Rtotai extracted from the linear regime of output characteristics versus channel length L at Vd=-4V (a-c) or Vd=-12V (d). The gate voltage has been corrected for threshold voltage extracted from Fig. 6.7. (e) Contact resistance values extrapolated from (a-d) at L = 0 pm. (f) Channel resistance per unit channel length extracted from the slopes of (a-d). All values in (a-1) were extracted from FETs with channel width W = 1 mm and multiplied by a factor of 0.1 to convert to flcm. The errors of (e-f) are due to data point variance from the linear fit.
81
The extracted contact resistance values are very high as compared to P3HT and
pentacene ohmic contacts on the order of 10"* Qcm [114, 115 ]. For PTAA, PIFTAATAA,
and PIFTAA, the channel resistance is high enough to absorb the majority of the applied
drain voltage (only barely for PIFTAA, a 10pm long FET would have a channel resistance of
6MQcm and contact resistance 2 MQcm, consequently 6/8 of the applied drain voltage
would generate drift current in the channel). For PIFPA, a significant amount of drain
voltage is lost on the contacts due to the reduced channel resistance. The effect of drain
voltage loss on contacts is reflected on the difference between the mobility in the saturation
(Fig. 6.7 left set) and linear regime (Fig. 6.7 right set). For PTAA, PIFTAATAA, and
PIFTAA, the mobilities are almost identical in both regimes, while for PIFPA the linear
mobility is approximately 4 times lower than the saturation mobility. Drain voltage loss due
to high contact resistance relatively to channel resistance is not the cause of the non-linear
output characteristic of PIFPA, as the onset voltage of V d = -5 V was not found to scale at all
with channel length.
6.4 Gate voltage dependence of mobility
Fig. 6.7 shows the linear and saturation mobilities for conjugated polymer FETs as a
function of gate voltage. The shape of the gate voltage dependence of the mobility depends
on the shape of the density of states (DOS). In the presence of disorder the p(Vo) profile is
exponential [116] or more correctly parabolic as expected from a Gaussian DOS [117], while
the mobility in single crystal FETs does not exhibit gate voltage dependence [118].
The mobility of PIFTAA and PIFTAATAA are parabolic (Fig. 6.7c-f), as is
probably the mobility of PTAA, although not enough gate voltage is applied to probe the
shape sufficiently. In contrast, the mobility profile of PEFPA is parabolic for low gate
voltages, with an additional superimposed negative linear component for higher gate
voltages (Fig. 6.7g,h). This mobility profile is narrower than the profile measured for PIFPA
FETs without adhesion layers (Chapter 5, Fig. 5.5). It is noted that the negative linear
component of the mobility profile for PIFPA FETs with Ti adhesion layer increasingly
overtakes the parabolic component (the transition occurs at lower gate voltage magnitudes)
in inverse proportion to the applied electric field in the channel.
82
PTAA T=290K L=10nm
X .
PTAAT=290K L=10um
0 -2 .5 Vq —“38 VT 2.0
5 1 .0o- 0.6
(C)
-40 -30 -20 -10 0 10 20Vg (V)
PIFTAATAA T=290K L=10jimVn=-38V
-40 -30 -20 -10 0 10 20Vq (V)
PIFTAATAA T=290K L=10nm
’■ % V d=-6V
-40 -30 -20 -10 0 10 20V n N )
40 -30 -20 -10 0 10 20Vg (V)
-40 -30 -20 -10 0Vg (V)
10 20
P FPA< 3.5 - ■. T=280K L=10nm& 3 0W 2.5
■Vq —“40 V
/0.5
1.6(h) 2 .5 ,
2.01 .2^
1.51 .0 ^0.8 9 O0.6 2 n
1.0
0.4 ^ 0.50 .2 50.0 0 0
2.5
2.0-^cz
1 .5 aok
1.0 I
O.S g
PIFTAA T=290K L=10nm
Tj/D="®V
PIFTAAT=290K L=10^im 2.052.0?C 1.2
O'<0 1.0V q —“38V
O 0.8^ 0.6 ê o .4o- 0.2 ^ 0.0
-40 -30 -20 -10 0 10 20Vg (V)
PIFPAT=280K L=10nm
■■,Vq =-io v
2 |
i l *
-50 -40 -30 -20 -10 0 10 20Vg (V)
-50 -40 -30 -20 -10 0 10 20Vg (V)
Figure 6.7. Left: Threshold voltage (gray lines) extrapolated from Sqrt(Io) (black lines) and mobility in the saturation regime (blue symbols). Right: Threshold voltage extrapolated from Iq and mobility in the linear regime for FETs with W =l, L=10, at T=290K (280K for PIFPA), measured in vacuum in dark.
83
6.5 Field and Temperature dependence of mobility
The mobility in a system dominated by intersite hopping generally exhibits positive
field dependence in accordance with the Poole Frenkel model [112]. However, in
semiconductors with high spatial disorder, the electric field may force charge carriers into
energetically unfavourable sites. The GDM model mathematically allows for negative field
dependence, which occurs when the spatial disorder parameter S overtakes the energetic
disorder parameter (a/kT)^. Negative field dependence has been observed in polythiophenes
at low fields [119].
In this work the field and temperature dependence were measured for both the linear
and saturation regimes. The dependence of ln(p) over can be seen in Fig. 6.8 for PTAA,
PIFTAA, PIFTAATAA, and PEFPA, for extraction in the linear regime (right set) and
saturation regime (left set). PTAA shows weak field dependence for both linear and
saturation mobilities. For PIFTAATAA, PIFTAA, and PEFPA, there is a noticeable
difference in the field dependence between the linear and saturation regimes. Saturation
mobilities show very weak negative field dependence, except PEFPA at low temperatures.
Linear mobilities show significantly stronger negative field dependence. As we have ruled
out the dependence of contact resistance on the electric field (indicated by the linear Rtotai
over L fit), the linear mobility should be much more reliable than the saturation mobility. In
the saturation regime, part of the field is dropped across the depletion region and thus
characterisation of the field dependence of the semiconductor would be in high error.
Figure 6.9 shows the 1/T dependence of the saturation mobilities for PTAA,
PIFTAATAA, PIFTAA, and PEFPA. The temperature dependence of the linear mobilities is
similar (not shown). Deviation from a linear fit is found for the lowest temperature point
(200 K) for PTAA, PEFTAATAA, and PIFTAA, as well as the three lowest temperature
points (100 K, 130 K, 160 K) for PIFPA. It is possible that this deviation comes from
experimental error, but unlikely as the probe station was allowed thermal equilibration time
of 30 mins to 1 hour from the time when the temperature sensor displayed a temperature
within ± 1 K of the desired lowest temperature value. Weakly temperature dependent
behaviour of the current is consistent with tunnelling current which is reasonable at low
temperatures when intermolecular barriers are high. The data points in deviation are ignored
(1 lowest for PTAA, PIFTAATAA, and PIFTAA, 3 lowest for PIFPA) and linear fits are
applied to the rest.
84
(a) -5.5
:= -6.0
g -6.5
& -7.0
I -’’•5 # -8.0
PTAA (b)
(C)-5.0
|: - 5 .5 Ü -6 .0
r -6.5 ^-7 .0 1 -7 .5 = -8.0
-8.5
(e)-3.0
-3.5Eic% -4.0
1 .4.5
-5.0
(g)
«.♦♦♦♦ ♦,►►►► ► . * ■*T ▼
♦ 350K» 330K4 31 OKT 290KA 260K. 230K. 200K
50 100 150 200Sqrt(F) (Sqrt(V/cm))
250
A
PIFTAATAA
: : Î ♦ 350K► 330K4 310K
* A A T 290KA 260K
• • # . 230K. 200K
0 50 100 150 200 250 300 350Sqrt(F) (Sqrt(V/cm))
PIFTAA350K 330K 31 OK 290K 260K 230K 200K
50 100 150 200 250 300 350 Sqrt(F) (Sqrt{V/cm))
-1.0^ • • •.-.-1 .5 - • • •E -2.0 ♦ ♦ ♦
$ -2.5S -3.0T -3.5 ▼ ▼ ▼
« -4.0 :r -4.5
-5.0-
PIFPA . 340K• 31 OK♦ 280K.. 250K4 220K
* * T 190KA 160K. 130K- 100K
50 100 150 200 250 300 350 400 Sqrt(F) (Sqrt(V/cm))
Eo>5
a.r
( d )
IC
(f)
E
Ic
-5.5
- 6.0
-6.5
-7.0
-7.5
- 8.0
-5.0-5.5- 6.0-6.5-7.0-7.5- 8.0-8.5
-3.0
-3.5
-4.0
-4.5
-5.0
TTt t
}.■
i i
PTAA
A 350K» 330K4 310K
290KA 260K. 230K. 200K
50 100 150 200Sqrt(F) (Sqrt(V/cm))
PIFTAATAA
250
* 350K330K
4 310Kr 290KA 260K• 230K■ 200K
50 100 150 200 250 300 350 Sqrt(F) (Sqrt(V/cm))
£:
I -
PIFTAA
350K 330K 310K 290 K 260 K 230K 200 K
( h)
50 100 150 200 250 300 350 Sqrt(F) (Sqrt(V/cm))
E
Ï
- 1.0-1.5- 2.0-2.5-3.0-3.5-4.0-4.5-5.0
PIFPA • 340K• 310K♦ 280K► 250K4 220KT 190KA 160K. 130K. 100K
50 100 150 200 250 300 350 400 Sqrt(F) (Sqrt(V/cm))
Figure 6.8. Dependence of ln(p) as a function of Sqrt(F) for different temperatures. Left: Saturation mobility, Right: linear mobility. Error bars for the linear mobilities are due to gate leakage.
85
(a)
- 6.0
J -6.6
â - 7 .2
I -7.8
-8.4
(C)- 2.8
^ -3.2 E
-3.6
-4.0<^ -4.4
-4.8
PTAA
PIFTAA
• I• I
200pm180pm160pm140pm120pm100pm
160pm140pm120pm100pm80pm60pm40pm20pm10pm
(b)-4.8
Ü -5.6
5 -6.4
I -7.2
- 8.0
1.0 1.5 2.0 2.5 3.010-5/t2 (10-5/K2)
( d )
80pm60pm40pm20pm10pm5pm
- 1.0-1.5
I -2.0^ -2.5 X-3.C I -3.5
S -4.0
1.0 1.5 2.0 2.510-5/t2 (10'5/k2)
3.0
PIFTAATAA
Î
200pm180 pm160pm ♦ 80pm140 pm * 60pm120pm # 40pm100pm * 20pm
. 10pm■ 5pm
1.0 1.5 2.0 2.5 3.010"5/t2 (10-5/K2)
PIFPA180pm80pm40pm20pm10pm5 pm3 pm
3 6 9 1210-®/t2 (10'5/k2)
15
Figure 6.9. Dependence of ln(p) in saturation regime polymers and temperatures.
over 1/T for different conjugated
6.6 Gaussian Disorder Model fitting
To fit to the GDM, firstly linear fits were applied to the ln(p) vs 1/T plots. The
slope dln(p)/d(l/T^) and the x-axis intercept ln(p(T^oo)) were extracted and plotted against
for the linear (Fig. 6.10) and saturation mobilities (Fig. 6.11). For the linear mobilities,
no reliable fit could be obtained for PTAA and PIFTAA. For the linear mobilities of
PIFTAATAA and PIFPA, as well as the saturation mobilities of all four conjugated
polymers, a was extracted from the x-axis intercept and C was extracted from the slope of
plots 6.10c,e and 6.1 la,c,e,g, while I was extracted from the slope and po from the x-axis
intercept of plots 6.10d,f and 6.1 lb,d,f,h. The results are summarised in table 6.1.
86
(a) -2.3
c < r ^ -2.4
E ^5 (X -2.5
I fin -2.6
o o
-2.7
(b)PTAA ? -3-2
5 -3.3
(C)
-2.4
20 30 4 0 50 60 70 80Sqrt(F) (Sqrt(V/cm))
PIFTAATAA
-3.4
-3.5
-3.6
-3.7
-3.8
PTAA
HI =-2-3in m o o
( d ) _
Ç 20 30 4 0 50 60 70Sqrt(F) (Sqrt(V/cm))
80
'«-3-0
-3.1
1 -3 .2
g -3.3
rJ -3.4
3 -3 .5
(e)20 40 60 80 100
Sqrt(F) (Sqrt(V/cm))
IIin in o o
(g)
■1.4
■1.5
1.6
■1.1
120
(f)
PIFTAATAA
5 20 40 60 80 100Sqrt(F) (Sqrt{V/cm))
120
PIFTAA -1.7
iCM
EÜc
- 1.8
-1.9
- 2.0
- 2.1
- 2.2
20
c e < i-1 .2
I t ^Ï #o O
40 60 80 100Sqrt(F) (Sqrt(V/cm))
PIFPA
120 f - ' - '
( 'Ç .O .Sg -0.8
- 1.0
3 - 1.2 & -1.4 § -1.6
- 1.8
1 PIFTAA
- 2.0X -2 .2
50 100 150Sqrt(F) (Sqrt(V/cm))
^ 0 -
20 40 60 80 100Sqrt(F) (Sqrt(V/cm))
i PIFPA
120
50 100 150Sqrt(F) (Sqrt(V /cm ))
200
Figure 6.10. Slopes (left) and x-axis intercepts (right) extracted from the ln(p) vs 1/T plots for the linear mobilities of four conjugated polymers. Error bars represent the variance of data points from the linear fit in ln(p) vs 1/T plots.
87
(a)- 2.1
CM
£ I -2.2
11 I# -2-3O O
(c)
^ CM CM ^
II% %
-2.4
- 2.2
-2.3
-2.4
^ -2.5
(e)
..-s CM CM ^
Iflo in o o
(g)
c<r ^
1 1in m o o
- 1.2
-1.3
-1.4
-1.5
- 1.6
-0.7
- 0.8
-0.9
- 1.0
- 1.1
- 1.2
(b)PTAA
50 100 150 200Sqrt(F) (Sqrt(V/cm))
PIFTAATAA
_ - 3 .8
^ -3.9Ü& -4.0
A^ -4-1
^-4 .2
(d)
CMEoc
50 100 150 200 250 300 -Sqrt(F) (Sqrt(V/cm)) j
PIFTAA
I50 100 150 200 250
Sqrt(F) (Sqrt(V/cm))
PIFPA
(0
CMEu
Ë
C300 -
(h)
I 1
IE3c
I0 50 100 150 200 250 300 350 400
Sqrt(F) (Sqrt(V/cm))
- 2.8-2.9-3.0-3.1-3.2-3.3-3.4-3.5-3.6
- 1.6
-1.7- 1.8
-1.9- 2.0- 2.1
- 2.2-2.3
- 0.2-0.3-0.4-0.5- 0.6-0.7- 0.8-0.9- 1.0- 1.1
PTAA
50 100 150 200Sqrt(F) (Sqrt(V/cm))
PIFTAATAA
i
50 100 150 200 250 300Sqrt(F) (Sqrt(V/cm))
PIFTAA
50 100 150 200 250 300Sqrt(F) (Sqrt(V/cm))
PIFPA
50 100 150 200 250 300 350 400 Sqrt(F) (Sqrt(V/cm))
Figure 6.11. Slopes (left) and x-axis intercepts (right) extracted from the ln(p) vs 1/T plots for the saturation mobilities of four conjugated polymers. Error bars represent the variance of data points from the linear fit in ln(p) vs 1/T plots.
Material/Regime Model/Measurement o(meV)
]io(cm /Vs)
c(cmfV^/^)
Z Ref
PTAA/Sat GDM/FET 62 0.02 1.2 10- 3.1 ThisworkPIFTAATAA/Sat GDM/FET 65 0.05 1.6 10" 3.8
PIFTAATAA/Lin GDM/FET 65 0.05 2.2-10"^ 4.8PIFTAA/Sat GDM/FET 51 0.2 2.5-10-^ 2.8PIFPA/Sat GDM/FET 45 0.75 4.6 10" 2.0PIFPA/Lin GDM/FET 48 0.5 4.2-10-^ 4.2
PIFPA SGDM/TOF 42 120PTAA SGDM/FET 68-85 0.04-0.1 121PTAA SGDM/TOF 57 0.2 121
Polythiophenes GDM/CELIV 58-74 0.0027-0.022
4 10- - 9 10"*
0.5-9.5
119
Polyfluorenes GDM/TOF 85-122 2.M0-^-0.018
2.1 10"*- 2.810"*
1.5-2.7
122
PFO GDM/TOF 66-95 0.005-0.045
2.4 10"*- 3.2 10"*
1.8-2.7
123
P3HT GDM/TOF/CELIV 61-75 0.004-0.01
1.4 10"*- 3.6 10"*
3-3.9
124
TPD GDM/TOF 76-103 0.005-0.07
1 10"*- 6 10"*
2.3-2.7
125
Table 6.1. GDM model parameters extracted in this chapter from linear and saturation regime FET mobilities, as well as reported literature values for different extraction techniques (Time Of Flight, Charge Extraction by Linearly Increasing Voltage). SGDM denotes simplified GDM model where the second term (field dependence term) of eq. 2.13 is ignored. The approximate error for extracted values is 2% for a and po, 5% for C, and 15% for Z. Errors estimated by considering variance of fits to fig. 6.10 and 6.11.
The linear and saturation mobility fits to the GDM model give similar values for o,
po, and C, while Z is significantly higher for linear mobilities, as expeeted since they exhibit
stronger negative field dependence, which is mathematically accounted for by the Z
parameter. The spatial disorder parameter Z accounts for the negative electric field
dependence of the mobility and takes values of 4.2 and 4.8 for PIFPA and PIFTAATAA.
Although a reliable fit could not be obtained for PIFTAA linear mobilities, a similarly high
value of Z would be expected to account for the observed negative field dependenee. The
negative field dependence seems to be inherent to eo-polymers with indenofluorene units,
and no field dependenee could be measured for PTAA.
For PTAA and PIFTAATAA, the extracted values are within a wide range of
reported values for disordered organic semiconductors. PIFTAA and especially PIFPA
exhibit very low energetic disorder of 51 meV and 45-48 meV respectively. The energetic
disorder for PIFPA is in elose agreement to reported value of 42 meV for Time Of Flight
(TOF) measurement, whieh is a measurement of bulk energetic disorder and should always
89
be lower than the disorder along an interfaee (such as the semiconductor-insulator interfaee
for a FET).
The C constant which is a characteristic of mean intersite distanee is similar for
PTAA, PIFTAATAA, and PIFTAA (1.2 10"*-2.510"*, while it is higher for PIFPA (4.2-10"*-
4.6-10"*). Perhaps this variation of C ean be associated with the GIWAXS characteristie
rings corresponding to a distance of 5.7 Â when the final report fi*om NTNU becomes
available.
The pre-faetor mobility po which is the mobility in absence of energetie disorder and
is proportional to the wave function overlap between neighbouring sites is surprisingly high
for PIFPA (0.5-0.75 em W s) considering its amorphous microstructure. In Marcus-Hush
theory based simulations for the intermoleeular hole transfer rate, it has been reported that
lateral and torsional disorder in poly(9,9-dioctylfluorene) ean result in inereased mobility as
eompared to highly ordered eases. The authors suggest that disorder results in a larger
distribution of intermolecular packing parameters and thus charge transfer rates, resulting in
the formation of an efficient transport network based on interehain hops at locations with
high transfer rates [126]. The portion of high transfer rate sites does not need to be high, as
charge transport pathways in conjugated polymer thin films form through the sites with
optimal transfer parameters, and the remainder of the thin film has little influence, as
reported for P3HT-insulator blends that retain high mobility for up to 90 wt% of insulating
component [127]. In another example, dithienopyrrole-thiophene semicrystalline eonjugated
copolymers exhibit high mobilities of up to 0.2 cm^/Vs in as-cast low-order thin films that
drop by a factor of 2 to 20 upon annealing-induced crystallisation [128].
Given the laek of order in PIFPA, the above rationale can reasonably explain the
strong intermolecular coupling indicated by the high prefaetor mobility, and in eombination
with low energetie disorder, provides a basis for the high mobility of this amorphous
conjugated copolymer.
6.7 Conclusions
Conjugated polymers PTAA, PIFTAATAA, PIFTAA, and PIFPA were
investigated. AFM and GIWAXS measurements (as well as DSC for PIFPA) did not result in
the detection of any crystalline features, suggesting that the conjugated polymers are
amorphous. The FET charge carrier mobility was examined for different gate voltages
(charge densities), different chaimel lengths (eleetrie fields) and temperatures.
90
Even though high effective workfunction electrodes were used for FET souree/drain-
semiconductor eontaets, the eurrent-voltage charaeteristies were ohmie for PTAA,
PIFTAATAA, and PIFTAA, but not for PIFPA. Strangely, the contact resistance measured
by TLM was relatively low (2 MQcm at Vq-Vt = -30V for PIFTAA and PIFPA), however
the lower channel resistance of PIFPA as compared to PIFTAA (300 kUcm/pm and 600
kHcm/pm respectively) results in higher suseeptibility to contact resistance and may be the
origin of the non-linear eurrent-voltage behaviour.
All conjugated polymers exhibited parabolic gate voltage dependence eonsistent
with a Gaussian DOS, with a relatively narrow width and peeuliar shape for PIFPA. As a
function of gate voltage, the contact resistance for PIFPA is found to deerease more rapidly
than the channel resistance, therefore the negative gate voltage dependence of the mobility
should not be an artefact from contact issues (sueh an artefaet would result in higher positive
measured gate voltage dependence of the mobility, not negative).
Mobilities extracted over variable temperatures were found to fit well to a Gaussian
profile characterised by 1/T dependence of ln(mobility), except for PIFPA in the
temperature range of 100 K to 160 K, whieh may or may not be due to experimental setup
errors. The weak temperature dependence for PIFPA observed at low temperatures indicates
a transition from phonon-assisted hopping over the intermoleeular energy barriers to
temperature independent tunnelling through the barriers.
No electric field dependence of the mobility was detected for PTAA, while all three
indenofluorene containing copolymers exhibited elear negative field dependenee whieh is a
signature of high spatial disorder (eharacterised by high Z of 4.8 and 4.2 for PIFTAATAA
and PIFPA respeetively).
The dipolar energetie disorder values for PIFTAA and PIFPA were found to be
amongst the lowest reported values (51 meV and 48 meV respectively).
The high pre-factor mobility for PIFTAA (0.2 cm W s) and PIFPA (0.5 - 0.75
cm^A/s) indicate strong coupling between hopping sites and was explained on the basis that
in disordered semiconduetors the molecules may take on more efficient eonfigurations for
charge transport which may be forbidden in a erystal structure. In sueh a ease, the portion of
sites with optimum eharge transfer integrals would be limited as compared to a conjugated
polymer with more uniform charge transfer integral distributions, which may explain the
redueed width of the DOS indieated by the gate voltage dependence of the mobility for
PIFPA.
91
D. Hertel, H, Bassler. “Photoconducion in Amorphous Organic Solids”. Chem. Phys. Chem. 9, 666(2008)
J. Veres, S. Ogier, G. Lloyd. “Gate Insulators in Organie Field-Effeet Transistors”. Chem. Mater. 16,4543 (2004)**'*L. Burgi, T. J. Richards, R. H. Friend, H. Sirringhaus. “Close look at charge carrier injection in polymer field-effect transistors”, J. Appl. Phys. 94, 6129 (2003)
D. J. Gundlach, L. Zhou, J. A. Nichols, T. N. Jackson, P. V. Necliudov. “An experimental study of contaet effects in organic thin film transistors”, J. Appl. Phys. 100, 024509 (2006)
A. R. Volkel, R. A. Street, D. Knipp. “Carrier transport and density of state distributions in pentacene transistors”, Phys. Rev. B, 66, 195336 (2002)
N. Tessler, Y. Roichman. “Amorphous organic molecule/polymer diodes and transistors - Comparison between predictions based on Gaussian or exponential density of states”. Organ. Electron. 6,200 (2005)*** A. F. Stassen, R. W. I. de Boer, N. N. losad, A. F. Morpurgo. “Influence of the gate dielectric on the mobility of rubrene single-crystal field-effect transistors”, Appl. Phys. Lett. 85, 3899 (2004)
V. Kazukauskas, M. Pranaitis, L. Sicot, F. Kajzar. “Negative mobility dependence on electric field in poly(3-alkylthiophene)s”. Mater. Sci. 12, 187 (2006)
H. Kim, N. Sehulte, G. Zhou, K. Mullen, F. Laquai. “A high gain and high charge carrier mobility indenofluorene-phenanthrene copolymer for light amplification and organic lasing”. Adv. Mater. 23, 894(2011)
J. Veres, S. D. Ogier, S. W. Leeming, D. C. Cupertino, S. M. Khaffaf. “Low-k insulators as the choice of dielectics in organic field-effect transistors”. Adv. Funct. Mater. 13, 199 (2003)
R. U. A. Khan, D. Poplavskyy, T. Kreouzis, D. D. C. Bradley. “Hole mobility within arylamine- containing polyfluorene copolymers: A time-of-flight transient-photocurrent study”, Phys. Rev. B 75, 035215 (2007)
T. Kreouzis, D. Poplavskyy, S. M. Tuladhar, M. Campoy-Quiles, J. Nelson, A. J. Campbell, D. D.C. Bradley. “Temperature and field dependence of hole mobility in poIy(9,9-dioctylfluorene)”, Phys. Rev. B 73, 235201 (2006)
A. J. Mozer, N. S. Sarieiftci, A. Pivrikas, R. Osterbacka, G. Juska, L. Brassat, H. Bassler. “Charge earrier mobility in regioregular poly(3-hexylthiophene) probed by transient conductivity techniques: A eomparative study”, Phys. Rev. B 71, 035214 (2005)
H. H. Fong, K. C. Lun, S. K. So. “Hole transport in molecularly doped triphenylamine derivative”, Chem. Phys, Lett. 353, 407 (2002)
S. Athanasopoulos, J. Kirkpatrick, D. Martinez, J. M. Frost, C. M. Foden, A. B. Walker, J. Nelson. “Predietive study of charge transport in disordered semicondueting polymers”. Nano Lett. 7, 1785 (2007)
92
A. Kumar, M. A. Baklar, K. Scott, T. Kreouzis, N. Stingelin-Stutzmaim. “Efficient, stable bulk charge transport in crystalline/erystalline semiconductor-insulator blends”. Adv. Mater. 21, 4447(2009)
J. Liu, R. Zhang, G. Sauve, T. Kowalewski, R. D. McCullough. “Highly disordered polymer field effect transistors: n-aUcyl Dithieno[3,2-b:2’,3’-d]pyrrole-based copolymers with surprisingly high charge carrier mobilities”, J. Am. Chem. Soc. 130,13167 (2008)
93
7. Polymer Source-Gated Transistors
This chapter demonstrates a new type of organic contact limited transistor that is
based on Schottky contacts of source/drain electrodes with conjugated polymers. This type
of transistor was initially reported with amorphous silicon semiconductor, and exhibits
certain distinct differences as compared to classical field-effect transistors, some of which
may be particularly interesting for plastic electronics.
7.1 Limitations of the Field-Effect Transistor
The Field-Effect Transistor (FET) is a widely used transistor type in organic
electronics that relies on ohmic contacts. An efficient p-type FET requires the source
electrode workfunction to be close enough to the Ionisation Potential (IP) of the
semiconductor to allow the injection of holes. In the case of insufficiently high
workfunction, a Schottky barrier is formed, reducing the transistor output current and
causing non-linear eurrent-voltage characteristics with increasing barrier height (section
3.1.4) [129,130,131,132,133]. According to the field-enhanced thermionic emission model,
the current through a reverse biased Schottky barrier decreases exponentially with the
effective height of the barrier (Eq. 3.4) and the effective height of the barrier decreases with
the square root of the applied field (Eq. 3.5, 3.6).
The electrical behaviour of a typical FET is largely determined by the geometry of
the conduction channel. The FET output current is inversely proportional to the channel
length (Eq. 3.12, 3.14). The current through an FET is modulated by changing the resistivity
of the transistor channel, which is achieved by biasing the gate-semiconductor capacitor and
accumulating charge along semiconductor-insulator interface. Current saturation for a
particular gate voltage Vq occurs above a certain drain voltage Vd>Vsat, given by Vsat^Vq-
V t , where V j is the threshold voltage. At Vsat, the drain end of the channel is depleted of
charge. For strong saturation, i.e. “flat” output characteristics above Vsat? the insulator
thickness must be significantly smaller than the channel length (Fig. 7.1). Although organic
FETs with ultrathin insulators have been demonstrated, practical polymer dielectrics usually
require higher thicknesses to avoid gate leakage [134,135], limiting the usable channel
length range. Strong saturation (high output impedance) is desired as the intrinsic voltage
gain is proportional to the output impedance (Eq. 3.18).
94
[a)
î '3 4
(b)W=1 mm, L=2.5pm, djns=1 0.4 Vq OV to -40V (-5V step)
PIFTAA-Cr/Au/PFBT 0.3
<3 0.2
0.1
0.0
W=1mm, L=100^m, d jns= V rnOV to -40V (-5V step)
PIFTAA-Cr/Au/PFBT-40V
-40 -30 -20 -10V d (V )
-40 -30 -20 -10Vd (V)
Figure 7.1. (a) Loss of saturation when the channel length L (2.5 pm) approaches the thickness of the insulator dins (1pm), and (b) flat saturation for a long channel PIFTAA FET.
7.2 Source-Gated Transistors
The concept of Source-Gated Transistors (SGTs) was introduced by Shannon and
Gerstner based on amorphous silicon [136]. SGTs structurally resemble FETs, but rely on
two conditions: 1) a source-semiconductor barrier, 2) the semiconductor layer must be
sandwiched between the source and the insulator/gate, as in the case of a bottom-contact top-
gate FET with the gate overlapping with the source electrode. According to the field-
enhanced thennionic emission model, the current through a reverse biased Schottky barrier
decreases exponentially with increasing effective barrier height (Eq. 3.4) and the effective
barrier height decreases with the square root of the applied field (Eq. 3.5, 3.6). The source-
semiconductor-insulator-gate stack in an SGT allows part of the gate voltage to drop across
the semiconductor and effectively pull down the source barrier by image force lowering,
allowing the injection of charge by field-enhanced thermionic emission, and the current
increases proportionally with the gate voltage [137]. Consequently, the transition from a
channel-modulated operating regime in FETs, to a source-modulated regime in SGTs,
considerably changes transistor behaviour and gives rise to useful features as discussed later
on. Shannon and Balon have suggested that source gating should be achievable with
different kinds of bamers including Schottky barrier, unipolar barrier, metal-insulator-
semiconductor barrier and space charge limited barrier [138].
Under no external bias, the semiconductor area near the electrodes is partly depleted
of charge due to the presence of Schottky barriers (Fig. 7.2a). Under a negative source-drain
voltage, the drain Schottky barrier is forward-biased and its contribution assumed to be
negligible (Eq. 3.3). The source Schottky barrier is reverse-biased, and an increasingly
negative drain voltage forces the depletion to extend towards the semiconductor-insulator
95
interface (Fig. 7.2b), eventually pinching off the conduction channel and saturating current
flow. As the semiconductor near the source is depleted, it can be treated as a dielectric in
series with the insulator (Fig.7.2b) to predict the change of potential across the source
depletion region per unit Vq. As any additional potential must be compensated by the drain
voltage before reaching saturation, this model was proposed to predict the rate of change of
saturation voltage per unit volt applied on the gate (dVsAr/dVo). The dielectric model
derived by Shannon and Gerstner for amorphous silicon SGTs states that the source-drain
voltage at which current saturates ( V sa t) can be expressed as [139]:
VgAT = - Vx ) + K (Eq. 7.1)(C j + C s )
Where C; and C; are the capacitance per unit area of the insulator and depleted
semiconductor respectively, and K is a parameter related to the voltage required to deplete
the semiconductor.
As a consequence of the dielectric model, the source saturation voltage in an SGT
should be a fraction of the drain saturation voltage in an FET, determined by C; and Cg. The
current at source saturation should be given by the field-enhanced thermionic emission
model (Eq. 3.3-3.5) where F is the electric field across the source depletion region. As such,
the output conductance of an SGT will be determined by how the field across the source
depletion region is affected by the total source-drain field [136]. Amorphous silicon SGTs
can exhibit higher voltage gain [136,138] and lower power consumption [138] than FETs of
equivalent dimensions, can saturate strongly even with thick insulators at short channel
lengths[140], and may exhibit current that is independent of the channel length [141], since
current modulation occurs at the source contact.
96
(a Gate
Insulator
S ource ' D rain
(b)
Vr
?
- Cs
6
Gate
insulator
D e p le t io n
S ource
E•-vac
^ H o le-------
Ep ■—, . $ ■
B arrie r------- E,p
a
D rain O
Vd
Figure 7.2. Schematic of thin film source-gated transistor (a) under no bias, and (b) under drain and gate bias. A more accurate (simulated) picture of charge and electric field distributions in the SGT channel can be found in [138].
7.3 Polymer Source-Gated Transistor fabrication and methods
Bottom-contact Top-Gate thin film transistor structures (Fig. 7.2) were fabricated
with semiconductors PIFTAA, PIFPA, and ADS250BE. Different source/drain electrodes
were used to fonn varying Schottky barrier heights which were estimated from Kelvin Probe
(KP) measurements (Section 4.7, 4.8). All gates used are Au. More infonnation on
fabrication can be found in chapter 4.
Source-gating was observed in the following semiconductor-electrode combinations:
ADS250BE-Cr, PIFTAA-Cr, PIFTAA-Ni, PIFPA-Cr, PIFPA-Cr/Au/FBT, PIFPA-
Cr/Au/PFBT. The estimated barrier heights are given in table 7.1. For the semiconductors,
the Ionisation Potential (IP) values from Cyclic Voltammetry (CV) are used. Although KP
measurements on samples of PIFTAA and PIFPA Fenni-pinned at the Ehomo-Ep+ level
indicated that the IP of both materials is 5.3 eV, this is in strong disagreement with observed
source-gating from PIFPA-Cr/Au/FBT and PIFPA-Cr/Au/PFBT contacts. There is a
possibility that the KP data is correct and that the barrier in said SGTs is based on a different
mechanism.
97
Contact E h o m o (CV) (eV) (Pm (KP) (eV) Ehom o-<Pm (eV)ADS250BE-Cr n /a
PIFTAA-Cr 5.5 4.5* 1 .0PIFTAA-Ni 5.5 4.7 0.8PIFPA-Cr 5.8 4.5* 1.3
PIFPA-Cr/Au/FBT 5.8 5.3 0.5PIFPA-Cr/Au/PFBT 5.8 5.5 0.3
Table 7.1. Estimated Schottky barrier heights for holes from CV measurements for the conjugated polymer HOMO and KP measurements for the electrodes. The value for Cr was assumed from literature and used as a reference in some KP measurements (details in section 4.5).
The nitrogen and temperature measurements for PIFTAA-Cr were performed in a
Linkam LTS420 probe stage with a Keithley SCS-4200 semiconductor parameter analyser.
All other measurements were performed with the same analyser and Karl Suss microprobes.
7.4 Polymer Source-Gated Transistor performance
One of the first instances of source-gating was observed in transistors with PIFTAA-
Cr contacts. The output characteristics have are dramatically different than for FETs, and the
current saturates at a drain voltage of -2 V to -5 V, even for gate voltage of up to -80 V (Fig.
7.3a), which would be approximately 15 times lower than for a PIFTAA FET at V g= -8 0 V .
The current level is significantly reduced. For a PIFTAA FET of 2,5 pm channel length, the
drain current per unit channel width at V g= -4 0 V and V d= -4 0 V is ITG* A/cm (Fig. 7.1a),
while for the PIFTAA-Cr SGT of identical channel length the current at the respective
voltages is 3 10' A/cm, a current reduction of 3T0^. The intrinsic voltage gain A of a
PIFTAA FET with channel length L = 2.5 pm calculated at V d= -9 V and V g= -1 0 V according
to eq. 3.18 and was found to be A = 12 (not shown - calculated from PIFTAA FETs in
chapter 6 at T = 290 K). For the PIFTAA-Cr SGT, the gain was calculated at V d= - 5 V and
V g= -1 0 V from the slope of the output characteristic in Fig. 7.3a and transfer characteristic in
Fig. 7.3b, and found to be 1.5. The SGT has lower output conductance but also significantly
lower transconductance as compared to the FET. The PIFTAA-Cr SGT was found to be
highly sensitive to atmospheric conditions and temperature (Fig. 7.3c). Upon heating from a
temperature of 300 K to 325 K, the source saturation was masked by a more typical FET
drain current for all gate voltages, while retaining most of the current reduction. A similar
effect was observed when the sample was exposed to nitrogen at room temperature, but the
source depletion vanished (Fig. 7.3c).
98
(a) 80
60
< 40
20
Vp OV to -80V (-10V step)
L=2.5^m
(C)140 120
100
< 80^ 60Q
4020
0
-30 -20 -10Vd (V)
L=2.5^m V =-50V
PIFTAA-Cr(b) 10'
10' 10
- 1 0 ' ^ 1 0 ' “ 1 0 "
1 0 "
1 0 "
10"
L=2.5^m
— Id (Vd”"5V)— Id (Vq=-30V)— Ig (Vd=-5V)— I g (Vd=-30V)
-80 -60
(d)25
— N2 T=300K 20— Air T=375K— AirT=350K ^ 1 5
10— Air T=325K S— Air T=300K -P
-40 “20Vg (V)
Vq=-50V
20
L—10^m— L=5nm— L=2.5nm
-30 -25 -20 -15 -10 -5Vd (V)
-25 -20 -15 -10Vd (V)
-5
Figure 7.3. Channel width W=lcm, insulator thickness d;=lpm. (a) Output and (b) transfer characteristics of PIFTAA-Cr SGT. (c) Output characteristic at V g= - 5 0 V at different temperatures measured in ambient air, and in nitrogen, (d) Output characteristic at Vg—50V for three different channel lengths.
Further heating to 350 K and 375 K resulted in smaller eurrent increase. The
observed thennal activation of the current is too weak to be ascribed to a Schottky barrier of
estimated height of 1.0 eV (Table 7.1). Activation energy measurements in transistors with
non-ohmic contacts and disordered organic semiconductors in literature have yielded values
that are lower than the activation energy of the barrier height [130, 132]. Injection in such
systems has been explained by considering hopping into a broadened Gaussian Density of
States (DOS) along the electrode-semiconductor interface, resulting only in partial control of
the Schottky barrier height by the electrode workfunction [132, 133, 142 ]. Such
considerations are not included in this work. Finally, it is noted that for PIFTAA-Cr SGTs,
perfonnance was generally not consistent and exhibited variation for different channel
lengths (fig. 7.3d), between devices of the same channel length, and for the same devices
over time (no consistent trends observed).
In addition to PIFTAA-Cr, source-gating was observed from PIFTAA-Ni transistors
(Fig. 7.4), however perfonnance was highly inconsistent amongst measured devices and the
current exhibited strong instability with applied drain voltage (note that the cun'cnts in output
and transfer characteristics do not match). No further characterisation was perfonned.
99
P IF T A A -N i
( a ) 1500Vq—-1 oov
1000
Q 500
V g —- 5 0 V
50 -40 -30 -20 -10 0
lO'i
L=5pm
10 ' 1
- 10 1
-10
-11
-100 -80 -60 -40 -20V g ( V )V d (V )
Figure 7.4 W=1 cm, d|=l jum. (a) Output and (b) transfer characteristic of PIFTAA-Ni SGTs.
SGTs were also fabricated with ADS250BE-Cr (Fig. 7.5). These devices exhibited
relatively consistent performance (Fig. 7.5a) however characterisation was not continued and
instead focused on higher mobility semiconductors PIFTAA and PIFPA.
ADS250BE-Cr(b)
200 n
150
< 1 0 0
-20 -15 -10 ■5 0
10V n ~ “6 0 V
< 10
-10
-11
V d ( V )
Figure 7.5. W=lcm, dj=lpm. (a) Output characteristics of ADS250BE-Cr SGTs at V g= - 6 0 V for a collection of different channel lengths (Ix 2.5pm, 4x 5pm, 4x 10pm). (b) Transfer characteristic for L=2.5pm.
SGTs were fabricated with PIFPA-Cr with estimated bamer height 1.3 eV. The
current for L=2.5pm at V g= - 3 0 V and V d= - 1 0 V is 2 On A (Fig. 7.6a). A PIFPA FET with
L=10pm at the respective voltages gives 50pA (Fig. 5.5a). Correcting for the chamiel length,
the current reduction for the SGT is 10 . The voltage gain at V g= - 4 2 V and V d= - 4 V is 3,
while the voltage gain A for a PIFPA FET of L=5pm (not shown - calculated from
characteristics in chapter 6) at Vg = -10 V and Vd = -10 V is A = 59.
100
To test the validity of the dielectric model (Eq. 7.1) and source depletion
mechanism, PIFPA-Cr SGTs were fabricated with different thicknesses of semiconductor
and insulator layers summarised in table 7.2. The capacitances per unit area of the respective
layers were calculated with Eq. 3.9 to find the respective Ci/(Ci+Cs) ratios. The dVsAx/dVo
values were calculated from output characteristics by measuring the voltage at the onset of
saturation Vsat and dividing with respective Vq values (also shown in Fig. 7.6c). The
calculated C;/(Ci+Cs) ratios and measured dVsAr/dVo values are plotted in Fig. 7.6d and
represented by solid lines and symbols respectively. Measurements and model predictions
are found to be in good agreement. In addition to the SGT in Fig. 7.6a, another SGT of
different dVsAr/dVo is shown in Fig. 7.6e with saturation voltages below -2 V for gate
voltages of up to -72 V. Assuming an FET would saturate at approximately -70 V for V g= -
70, the SGT saturation voltage is an improvement of a factor of 35. PIFPA-Cr SGTs, as most
SGTs in this work, exhibit severe current instability with prolonged drain bias. Fig. 7.6f
shows the output characteristic of the same SGT in Fig. 7.6e after approximately one hour of
constant bias at V g= - 5 0 V and V d= - 3 0 V . The current has increased by approximately a
factor of 200 and the voltage gain at V g= - 4 0 V and V d = - 5 V is 25, however this state is not
stable and fully reversed within several tens of minutes. This effect may be related to trap
filling at the electrode-semiconductor interface and indicates that injection cannot be
adequately explained solely by considering Schottky barrier dynamics.
101
(a) 40
30
I 2 0Q
10
0
(C)
-1
-2
S -3^ -4
-5
-6
.72V Vg OV to -72V (-6V step) L=2.5nm
ds=100nmj=1600nm
PIFPA-Cr
(b) 1 0
l o' i 10
“10 “8 “6 -4 -2Vd (V)
. ds d, I.d|60/168060/90030/900100/160030/660
“80 “70 “60 “50 “40 -30 -20 -10Vg (V)
^ 1 0r i o " “
1 0 "
1010
— Id (Vd~'1 V)— Ig (Vd "*! V)— ID (Vd”“10V)— Ig (V d=-10V)
“60 “50 “40 “30 “20 “10 0 10 20 30
0.10
0.08
0.06
< 0.04
0.02
0.001.0 1.5 2.0 2.5 3.0
Cj (nF/cm)
(e)
<c
72VVq OV to -72V (“6V step)
dj=60nm d„=1680nm
=2.5nm
■10 “8 “6 “4 “2Vd (V)
(f)
1000
<cq 500
-50V
L=2.5nmVq OV to “50V (“5V step)
■10 “8 “6 “4 “2Vd (V)
Figure 7.6. W=lcm. (a) Output and (b) transfer characteristic of PIFPA-Cr SGT with ds=100nm and di=1600nm. (c) Vsat values measured for PIFPA-Cr SGTs of different layer thicknesses (Table 7.2) at different gate voltages, (d) Measured values for dYsAx/dVc (symbols) against insulator capacitance (x-axis) and semiconductor capacitance (solid lines), (e) Output characteristic of SGT with dj=60nm and d;-1680nm. (f) Same SGT after prolonged constant gate and drain bias stress.
102
ds (nm) di (nm) Cs (nF/cm^) Ci (nF/cm^) Ci/(C;+CJ dVsAx/dVo(measured)
60 1680 44.2 TO'^ 1.13-10" 0.025 0.01960 900 44.2-10'" 2.11-10" 0.046 0.05130 900 88.5-10'" 2.11-10'^ 0.023 0.024100 1600 26.5-10" 1.19-10" 0.043 0.04930 660 88.5-10" 2.88-10" 0.032 0.033
Table 7.2. Summary of layer thicknesses and respective capacitances per unit area used to fabricated PIFPA-Cr SGTs and test the dielectric model. C;/(Ci+CJ are the ratios predicted by Eq. 7.1, and dVsAr/dVc are values measured from the output characteristics of fabricated devices.
In an attempt to fabricate higher current SGTs, a Self-Assembled Monolayer (SAM)
of moderate dipole strength was used (FluoroBenzeneThiol - FBT) to create a Schottky
barrier in combination with Au electrodes. With an estimated barrier height of 0.5 eV,
PIFPA-Cr/Au/FBT SGTs show higher currents than PIFPA-Cr SGTs. Correcting for a high
V t= - 2 0 V , the current at VG-Vf=-30V and V d = - 1 0 V is approximately O.SpA, which is a
current reduction of 2T0^ as compared to respective voltages in a PIFPA FET (corrected for
L) (Fig. 5.7d). The voltage gain at V g= - 6 0 V ( V g - V t= - 4 0 V ) and V d = - 6 V is 59 which the
same with PIFPA FETs of L = 5 pm at V g= - 1 0 V and V d= - 1 0 V (calculated from
characteristics in chapter 6 at T=280 K, not shown) with A = 59. PIFPA-Cr/Au/FBT SGTs
also showed some degree of current instability, although not placed under prolonged
constant bias, after approximately 10 measurement cycles the current has doubled (Fig.
7.7c). The performance of these SGTs was fairly consistent. Measurements during the first
days showed no source saturation and lower currents, but when measured several months
later multiple SGTs on the same substrate showed very similar output characteristic shapes
and current levels. Such behaviour has been observed on all the different types of fabricated
SGTs in this work to different extents: there seems to be a time requirement for the transistor
to start exhibit source-gating. Processes on such timescales should be of chemical nature,
suggesting an electrode-semiconductor interaction. However, on chapter 5, evidence was
presented that contacts in Au/PFBT FETs are stable over periods of months. The only
remaining variable may be Cr - the adhesion layer, which could be the origin of chemical
changes and trap formation in the semiconductor in the vicinity of the contact over time.
103
(a)
PIFPA-Cr/Au/FBT(b)
10^
10®
10 ®
10"®
10 " '
10
L=5^m 1.5
— Id (Vo=-10V)— Ig (Vd= - 1 0 V ) 5 ' 1 - 0
- 0 . 5
0.0
Vg OV to -70V (-5V step) -70V^ — L=5nm
-80 -60 -40 -20Vg (V)
(C)
20 -20 ■15 -10Vd (V)
7 0 V Vg OV to -70V (-5V step)
L=5|im
-20 -15 -10Vd (V)
Figure 7.7. W -lcm , di=l^im. (a) Transfer and (b) output characteristic of PIFPA-Cr/Au/FBT SGT. (c) Output characteristic of same SGT after approximately 10 cycles of measurements.
Finally, source gating was observed in PIFPA-Cr/Au/PFBT transistors with
estimated barrier height of 0 .3 eV. The difference from FETs is the adhesion layer whieh
was measured by KP to lower the workfunction of Au by approximately 0 .1 eV. This contact
does not consistently exhibit source-gating. For batch 1 , behaviour was very consistent
across all measured devices on the substrate. The current at V g - V t= - 3 0 V and V d = - 1 0 V is
lOOnA/cm (Fig. 7.8a), with a current reduction of approximately 10 as compared to PIFPA
FETs. A multichannel structure was used for this batch, and perfonuance was found to be
very similar for all measured devices. The current extracted at the onset of source saturation
for different channel lengths can be seen in Fig. 7.Be, and it is almost identical for all except
the 4 shortest channels.
104
(a L=20|amVg OV to -40V (-5V step)40-
-40V30
10
■2 0■6 ■4
PIFPA-Cr/Au/PFBT(b)
10 10 10''
^ 1 0 '
S id ' -910"
10' 1 0 '
1 0 " '
1 0 "
— Id (Vd=-6V)— Ig (Vd““OV)
Vn (V)-40 -30 -20 -10 0 10 20
Vg (V)
— L=2.5nm — L—5fim—— L—10|j.m— L=20nm
L=40pm
Vq —“6 0 V
— L=60nm— L=80|im— L=100 im— L=120 171— L=140nm— L“160p.m— L=180 171— L=200nm
M L-
Figure 7.8. W=lmm, di„s=lpm. (a) Output and (b) transfer characteristic of PIFPA- Cr/Au/PFBT SGT. (c) Output characteristics at Vg—40V from SGTs of different channel lengths.
For batch 2 of PIFPA-Cr/Au/PFBT transistors, the current at V g - V t= - 3 0 V and V d = -
lOV is 20pA/cm (Fig. 7.8a), with a current reduction of only approximately a factor of 10 as
compared to PIFPA FETs. The main difference between batches 1 and 2 is the electrode
deposition method (evaporation and sputtering respectively). The evaporated Au for batch 1
might have been contaminated leading to ineffectiveness of the PFBT treatment, thus it is
likely that the actual barrier height is higher than the estimated barrier height.
105
(a) 80 70 60 50
< 40 3 30
Q- 20
10 0
-10
PIFPA-Cr/Au/PFBT (2)
-60V Vg OV to -60V (-5V step)
-10 -8 -6 -4Vd (V)
-2
V D“"1 OV
-50 -40 -30 -20 -10Vg (V)
10
Figure 7.9 W=lcm, dins=lpm, L=2.5nm. (a) Output and (b) transfer characteristic of PIFPA- Cr/Au/PFBT SGT (Batch 2).
The current reductions from SGTs in this work are show in Fig. 7.10. A trend seems
to be followed for the barrier height for the moderate and high barriers.
10000
1000
PIFTAA-Cr
100
3 10
PIFPA- PIFPA-CR Cr/Au/PFBT
PIFPA-Cr/Au/FBT
■PIFPA- Cr/Au/PFBT (2)
0.0Barrier Height (eV)
Figure 7.10. Current reduction for different types of SGT in this chapter
7.5 Conclusions
Conjugated polymer based Source-Gated Transistors have been demonstrated with a
variety of semiconductors and electrodes of different barrier heights (PIFTAA-Cr 1.0 eV,
PIFTAA-Ni 0.8 eV, ADS250BE-Cr, PlFPA-Cr 1.3 eV, PlFPA-Cr/Au/FBT 0.5 eV, PIFPA-
Cr/Au/PFBT 0.3 eV).
All SGTs saturate at significantly lower voltages than FETs. While the saturation
voltage per unit gate voltage for FETs increases linearly with a coefficient of 1, in fabricated
conjugated polymer SGTs the coefficient can be as low as 0.02. The dielectric model that
predicts saturation behaviour was tested by fabricating SGTs with varying insulator and
106
semiconductor layer thicknesses and found to be valid. As a result of the very low
dVsAi/dVo ratio, the saturation voltage in SGTs remains at low drain voltages even for high
gate voltages (saturation at Vd = -1.5 V at Vg = -72 V for PIFPA-Cr SGT in Fig. 7.6e).
Output conductance is lower than for equivalent FETs, however for most SGTs large
reductions in current result in significantly reduced transconductance and the intrinsic
voltage gain A is at lower levels than FETs (A = 12 for PIFTAA FET with L = 2.5pm, A = 3
for PIFTAA-Cr SGT of the same channel length). The highest current SGT fabricated was
PIFPA-Cr/Au/PFBT (2), with a current reduction of a factor of 10 as compared to equivalent
PIFPA FET.
Since the current in SGTs is mostly modulated by the source-semiconductor contact
and not the channel, the current can be independent from the channel length. The invariance
of the current on channel length was demonstrated by PIFPA-Cr/Au/PFBT SGTs, for which
the current was similar in the range of 30 - 40 nA for channel lengths of 2.5 to 200 pm.
In general, polymer SGTs were found to be unstable with temperature, atmospheric
conditions, and time, and much work remains to be done before they can find practical use
(barrier optimisation to minimise current loss, drain field relief to prevent the drain voltage
from lowering the barrier and increasing the current and thus output conductance), however
several useful properties have been demonstrated and may be well fitting to organic
electronics.
Transistors can be fabricated with non-Ohmic contacts, allowing the utilisation of
very high IP air-stable conjugated polymers without the need for high work function
electrodes. Output resistance in SGTs is high, even when the insulator thickness is similar to
the channel length, allowing SGTs to saturate with thick printed insulators and short
channels, and at very low drain voltages. The independence of the current on channel length
in SGTs can be utilised with low resolution printing techniques such as ink-jet to provide
consistent current over variable channel lengths.
107
A. Bolognesi, A. Di Carlo, P. Lugli. “Influence of carrier mobility and contact barrier height on the electrical characteristics of organic transistors”, Appl. Phys. Lett. 81,4646 (2002)
L. Burgi,T. J. Richards, R. H. Friend, H. Sirringhaus. “Close look at charge carrier injection in polymer field-effect transistors”, J. Appl. Phys. 94, 6129 (2003)
D. J. Gundlach, L. Zhou, J. A. Nichols, T. N. Jackson, P. V. Necliudov. “An experimental study of contact effects in organic thin film transistors”, J. Appl. Phys. 100, 024509 (2006)
B. H. Hamadani, D. Natelson. “Nonlinear charge injection in organic field-effect transistors”, J. Appl. Phys. 97, 064508 (2005)
S. Alborghetti, J. M. D. Coey, P. Stamenov. “Dependence of charge carrier injection on the interface energy barrier in short-channel polymeric field effect transistors”, Appl. Phys. Lett. 100, 143301 (2012)
M. Hambsch, K. Reuter, M. Stanel, G. Schmidt, H. Kempa, U. Fugmann, U. Hahn, A. C. Hubler. “Uniformity of fully gravure printed organic field-effect transistors”. Mat. Sci. Eng. B 170, 93 (2010) ^ R. P. Ortiz, A. Facchetti, T. J. Marks. “High-k organic, inorganic, and hybrid dielectrics for low- voltage organic field-effect transistors”, Chem. Rev. 110 (2010) 205
M. Shannon, E. G. Gerstner. “Source-gated thin-film transistors”, IEEE Electron Devic. Lett. 24 (2003) 405^ F. Balon, J. M. Shannon. “Analysis of schottky barrier source-gated transistors in a-Si:H”, Solid
State Electron. 50 (2006) 378J. M. Shannon, F. Balon. “Source-gated thin-film transistors”. Solid State Electron. 52, 449 (2008) J. M. Shannon, E. G. Gerstner. “Source-gated transistors in hydrogenated amorphous silicon”.
Solid State Electron. 48, 1155 (2004)J. M. Shannon, F. Balon. “High-performance thin-film transistors in disordered and poor-quality
semiconductors”, IEEE Trans. Electron. Devic. 54, 354 (2007)F. Balon, J. M. Shannon, B. J. Sealy. “Modelling of high-current source-gated transistors in
amorphous silicon”, Appl. Phys. Lett. 86, 073503 (2005)M. A. Baldo, S. R. Forrest. “Interface-limited injection in amorphous organic semiconductors”,
Phys. Rev. B 64, 085201 (2001)
108
8. Conclusions
8.1 Summary
High Ionisation Potential (IP) amorphous conjugated polymers are promising
candidates for printed organic electronics as they overcome two important issues. Firstly,
Field-Effect Transistors (FETs) fabricated with hydrophobic insulators and high IP
conjugated polymers are highly stable in ambient conditions. Secondly, the consistent
amorphous microstructure reduces the spread of FET operating parameters and their
dependence on various processing conditions. The combination of the above can allow for
the production of FETs of reproducible performance by a variety of printing techniques in
ambient conditions.
A concerning issue with amorphous conjugated polymers is the low charge carrier
mobility which is typically on the order of 10' em W s for widely used materials (PTAA,
PTPA). It is widely believed that mobility of conjugated polymers increases with the
crystallinity of the sample, with typical polycrystalline polymers exhibiting mobilities up to
1 em^/Vs (TIPS-pentacene, pBTTT).
This thesis is based on the fabrication of FETs with hydrophobic insulator and
amorphous high IP conjugated polymers. The main semiconductors used were PIFTAA
(field-effect saturation mobility P fe - s a t = 0.03-0.04 cm W s) and PIFPA ( p f e - s a t = 0.2-0.3
cmW s), and to a lesser degree, PTAA ( p f e - s a t = 0.003 cmW s), PIFTAATAA ( p f e - s a t =
0.004 cmW s), and ADS250BE ( p f e - s a t = ~10'^ em^/Vs). PIFTAA and PIFPA are excellent
candidates for printing applications as they exhibit stability in ambient conditions (high IP),
reproducible electrical performance (amorphous morphology), and high mobility. Electrodes
for FETs were fabricated by metal evaporation/sputtering, photolithography, and etching.
The semiconductors and insulator were spin-eoated. More information can be found in
chapter 4. Investigations on various topics are summarised:
109
Chapter 4
> Work function engineering for FET electrodes utilising different metals and thiol-based
Self-Assembled Monolayers (SAMs):
Kelvin Probe (KP) studies revealed an issue with Au electrodes. The iodine-based
etchant that was used to pattern the electrodes passivated the Au surface, rendering the
thiol SAM treatment ineffective, as indicated by lack of work function shift for etched,
cleaned, and thiol-treated samples as compared to etched and cleaned samples. The issue
was resolved by exposure to light 0% plasma.
Several different electrode and thiol SAM (PentaFluoroBenzeneThiol PFBT,
FluoroBenzeneThiol FBT) combinations were characterised by KP and their
workftinctions were determined including: Au/PFBT 5.6 eV (used for FETs),
Cr/Au/PFBT 5.5 eV (FETs and SGTs), Cr/Au/FBT 5.3 eV (SGTs), Ni 4.7 eV (SGTs), Cr
4.5 eV (from literature, SGTs)
> Dual-Gate (DG) FET engineering utilising the same top and bottom insulator:
The key to utilisation of highly hydrophobic insulator Cytop™ as a bottom insulator
was slight reduction of hydrophobicity by exposure to light O2 plasma, enabling spin-
coating of the semiconductor solution on the surface. DG organic FETs with the same
top and bottom insulator to our knowledge have not been previously reported.
Chapter 5
> Long term ambient stability characterisation for PIFTAA and PIFPA:
The long term ambient stability of FETs with PIFTAA and PIFPA was characterised
over a period of 4 and 2 months respectively and the mobilities of both materials were found
to degrade at a rate of approximately 10% per month. The ambient stability of PIFTAA and
PIFPA is amongst the highest reported for conjugated polymers, only surpassed by PTPA,
the mobility of which is too low for commercial applications. A positive trend of ambient
stability with organic semiconductor IP is observed.
110
> Comparison of performance and stability between Top-Gate (TG) and Bottom-Gate
(BG) FET configurations with PIFTAA and PIFPA:
In a study relating to FET configurations, the long term ambient stability and general
performance between two widely used FET configurations TG and BG was examined.
Individual BG FETs performed inferiorly to individual TG FETs (higher subthreshold swing,
hysteresis, negative tum-on voltage shift, lower stability). A close comparative study was
achieved with DG FETs. By utilising the same material for top and bottom insulators.
Density of States (DOS) broadening due to dipolar disorder from the insulator surface should
be the same for the top and bottom transistor conduction channels. Also, the amorphous
semiconductors used remove the concern for morphology variation at the top and bottom
transistor channels, which is a significant issue for crystalline conjugated polymers.
Both gate modes were found to perform similarly in the DG structure, with no
hysteresis, on/off ratio in excess of 10 , and near-zero tum-on voltage. Some minor
differences were observed. For PIFTAA the mobility in TG mode of the DG FET was
slightly lower than for BG mode (-20%) whieh may be caused by access resistance of the
relatively thick semiconductor film (100 nm). For PIFPA the mobility was the same for TG
mode and BG mode, which may be due to reduced effect of access resistance resulting from
the higher conductivity of PIFPA. The subthreshold swing was lower in TG mode of the DG
than in BG mode, by a factor of 2 to 3 for both PIFTAA and PIFPA. Additionally it was
found that PIFTAA is somewhat affected by residual oxygen species on the surface of the
insulator (O2 plasma induced), indicated by a positive shift of turn on voltage (7 V shift)
suggesting deep trap formation. PIFPA was not affected, possibly due to the higher IP and
thus stability against oxidation. The mobility degradation rate of both gate modes was
similar. The degradation rate between the mobility in the linear and saturation regime was
similar, suggesting that the souree/drain-semiconductor contacts are stable and degradation
originates either from the semiconductor thin film bulk or the semiconductor-insulator
interface.
Chapter 6
> Morphology evaluation for amorphous conjugated polymers
Atomic force microscopy revealed no crystal domains for PTAA, PIFTAATAA,
PIFTAA, and PIFPA. Slight coalescence was observed for PIFPA. Grazing incidence wide
angle x-ray scattering measurements revealed only broad rings for the aforementioned
111
polymers in the range of 2 to 16 Â. Differential scanning calorimetry of PIFPA showed no
heat flow peaks thus no physical phase transitions. The evidence suggests that the
morphologies of PTAA, PIFTAATAA, PIFTAA, and PIFPA are amorphous.
> Charge transport charaeterisation of amorphous conjugated polymers
Charge transport in disordered conjugated polymers is described with the Gaussian
Disorder Model (GDM), which is based on phonon-assisted hopping between localised
states, and assumes a Gaussian distribution of energies (charaeterised by o, energetic
disorder), a Gaussian distribution of the dimensionless ratio of inter-site distance to charge
localisation radius (characterised by Z, spatial disorder), and mobility in the absence of
disorder (po, characteristic of intermolecular coupling). To fit data to the GDM, mobility
values are required as a function of temperature and electric field. These values are
commonly obtained by Time of Flight (TOF) measurements on diode structures, which are
simpler to measure and less error prone than FETs. Fitting FET-extracted data to the GDM is
challenging and no reports have been found in literature aside from fittings to a simplified
GDM model (ignoring the field dependence part of the model). In this work, FETs of
different channel lengths (ranging from 2.5 pm to 200 pm) were fabricated to obtain data on
different electric fields, and they were measured over a range of temperatures (200 K to 350
K for PTAA, PIFTAA, PIFTAATAA, 100 K to 340 K for PIFPA). Mobility data from both
the linear and saturation regime were examined and fitted to the model. It was found that the
temperature dependence of both the linear and saturation mobilities was very similar, and
that in terms of field dependence the saturation mobilities deviated significantly from linear
mobilities. The latter were assumed to be correct as the electric field distribution is not
complicated by near-drain charge depletion as in the saturation regime. Also, the multi
channel data was fit to Transmission Line Method (TLM) to evaluate contact and channel
resistances. No contact related effects were observed for PTAA, PIFTAA, and
PIFTAATAA, supporting the validity of the behaviour of the linear mobilities. Non-Ohmie
current-voltage behaviour was observed for PIFPA, which may or may not have affected the
observed field dependence.
All conjugated polymers exhibited parabolic dependence of the mobility with gate
voltage, consistent with a Gaussian Density of States (DOS). The parabolic mobility profile
for PIFPA was noticeably narrower than for the other three polymers. The temperature
dependence for all polymers was a good fit with the mobility over 1/T , consistent with a
Gaussian DOS, with one exception: The mobility for PIFPA at low temperatures (100 K to
160 K) became weakly dependent on temperature, suggesting a transition from phonon-
112
assisted hopping over the intermolecular energy barriers to temperature independent
tunnelling through the barriers. For PTAA, no field dependence of the mobility could be
measured. The indenofiuorene based conjugated polymers PIFTAA, PIFTAATAA, and
PIFPA, exhibited clear negative field dependence, which is a signature of high spatial
disorder. In cases of high spatial disorder, an increasing electric field may force charge
carriers to move through energetically unfavourable sites, thus the field dependence shows as
negative, even though the Poole-Frenkel mechanism of field-assisted intermolecular barrier
lowering is still in effect. Finally, the mobility in the absence of disorder po was found to be
relatively low for PTAA and PIFTAATAA, on the order of 10' em W s, while it was 0.2
cm W s for PIFTAA and 0.5-0.75 cm W s for PIFPA, the latter value being similar to
reported values for molecular crystals. This was explained on the basis that spatial disorder
allows for a wider distribution of intermolecular stacking schemes and thus charge transfer
integrals, therefore increasing the probability of formation of efficient intermolecular
couplings that may be forbidden in a periodic lattice. This hypothesis may be consistent with
the narrow DOS observed from the gate voltage dependence, as the portion of efficiently
coupled sites could be very small as compared to a periodic lattice. Finally, the energetic
disorder characteristic parameter o for PIFTAA and PIFPA (51 and 48 meV respectively)
was found to be amongst the lowest reported values for organic semiconductors, however it
is not sufficient to explain the high mobility of PIFPA which is ascribed to the above
hypothesis of disorder-enhanced intermolecular couplings.
Chapter 7
> Demonstration of source-gating with different semiconductors/electrodes
A type of contact limited transistor, the Source-Gated Transistor (SGT) based on
Schottky source/drain-semiconductor contacts was demonstrated for the first time with
organic semiconductors. In SGTs the current is modulated by the source-semiconduetor
reverse-biased Schottky barrier, as opposed to FETs in whieh the current is modulated by the
resistivity of the semiconductor-insulator interface. Current saturation in SGTs occurs due
near-source depletion as opposed to FETs in whieh current saturation occurs due to near
drain depletion.
SGTs were demonstrated with conjugated polymers and electrode combinations of
estimated Schottky barrier heights: PIFTAA-Cr 1 eV, PIFTAA-Ni 0.8 eV, ADS250BE-Cr,
PIFPA-Cr 1.3 eV, PIFPA-Cr/Au/FBT 0.5 eV, PIFPA-Cr/Au/PFBT 0.3 eV. All measured
113
SGTs exhibited very low saturation voltages with weak gate voltage dependence of the
saturation voltage as predieted by the dieleetrie model proposed for amorphous silieon SGTs.
This model suggests that the dependence of the saturation voltage on gate voltage is
determined by the relative depleted semiconductor/insulator capaeitanees per unit area. The
model was tested by fabricating SGTs of different semiconductor/insulator layer thicknesses
and measured values for dVsAx/dVo were in agreement with model predicted values, with
values measured as low as 0.02. For an FET, this eoeffieient would be 1, i.e. at a gate voltage
of - 70 V, the current would saturate for a drain voltage of approximately - 70 V. For an
PIFPA-Cr SGTs, at a gate voltage of - 70 V, the current saturates at a drain voltage of - 2 V,
and at similar drain voltages for all other SGTs. Additionally, for flat saturation in FETs, the
insulator thickness must be significantly smaller than the channel length. This rule does not
apply to SGTs, whieh show flat saturation for a channel length of 2.5 pm and insulator
thickness of 1 pm, thus SGTs may be utilised in applications with thick insulators typically
deposited by printing techniques. Also, since SGTs are not based on Ohmic contacts, they
can be fabricated with very high IP air-stable conjugated polymers that Ohmie contacts
cannot be easily established to.
SGTs suffer from an inherent current decrease stemming from the large series
resistance of the near-source depletion region. The SGT fabricated with the smallest current
decrease was PIFPA-Cr/Au/PFBT with estimated barrier height of 0.3 eV and current
decrease by a factor of 10.
As the current in SGTs is modulated by the source contact, the current should exhibit
weaker dependence on chaimel length as opposed to FETs in whieh the dependence is
inversely linear. For PIFPA-Cr/Au/PFBT SGTs, the current was found to be almost
independent of the channel length from 2.5 pm to 200 pm, however this was generally not
always the case for all SGTs.
Fabricated SGTs were generally found to be highly unstable with temperature,
atmosphere, and prolonged drain bias. Even though the concept is demonstrated, significant
work must be done to obtain a better understanding of SGT mechanisms with organic
semiconductors and more stable contacts must be formed.
114
8.2 Future work
The source of degradation in ambient conditions of the measured FETs was traced to
either the conjugated polymer bulk or the insulator-semieonductor interface, and
source/drain-semiconductor contact degradation was not detected. The degradation could
originate from residual solvent evaporation, oxygen absorption, or moisture absorption.
Insights to the origin of degradation can be gained by characterising performance over time
in different atmospheres such as dry air and moist nitrogen to separate the contributions.
Additionally, Quartz Crystal Microbalance (QCM) measurements would be useful in
determining net weight gain/loss over time. The stability of the conjugated polymers
probably cannot be improved unless by chemical design or purification.
The hypothesis that disorder may enhance intermolecular couplings is of particular
interest. Improvement in the performance of conjugated polymers may be gained by
inducing disorder, either by blending crystalline polymers with amorphous polymers, or by
chemical design. Gains can probably be predicted to some degree by quantum chemical
calculations to evaluate whether the adopted intermolecular stacking scheme of a given
conjugated polymer corresponds to the highest possible intermolecular charge transfer
integral. If not, the conjugated polymer would likely benefit from disorder.
For conjugated polymer FETs, the mechanisms governing charge injection at the
souree/drain-semiconductor interface are not well understood. Especially for contacts with
disordered conjugated polymers, it is crucial that a method is adopted that can characterise
the DOS broadening, trap density, and trap levels along the contact interface, as well as the
dependence of the Schottky barrier height on mentioned parameters. If the mechanisms
governing charge injection are understood, then Ohmic contacts may be realisable for a
wider range of conjugated polymers, electrodes, and channel lengths. Also, SGTs with lower
current loss and higher stability should be possible.
At the current state of high IP amorphous conjugated polymers, PIFPA is a
promising material for printing applications, however an electrode material that reliably
forms Ohmic contacts is required. It is possible that the high dipole induced by the
fluorinated thiol SAMS has a negative impact on the DOS in the vicinity of the contacts.
This hypothesis can be tested by fabricating contacts with a lower dipole SAM such as
nitrobenzenethiol, although a higher work function underlying electrode would be required
to prevent formation of a Schottky barrier.
Although PIFTAA exhibits somewhat low mobility of 0.03 - 0.04 cm^/Vs, Ohmie
contacts are significantly less challenging to form, ambient stability is high, and
115
reproducibility is exceptional. PIFTAA would be a good candidate for commercial
applications if high currents or switching speeds were not required.
8.3 List of Publications
> S. Georgakopoulos, D. Sparrowe, F. Meyer, M. Shkunov. “Stability of top- and
bottom-gate amorphous polymer field-effect transistors”, Appl. Phys. Lett. 97,
243507 (2010)
> G. Adamopoulos, A. Bashir, S. Thomas, W. P. Gillin, S. Georgakopoulos, M.
Shkunov, M. A. Baklar, N. Stingelin, R. C. Maher, L. F. Cohen, D. D. C. Bradley, T.
D. Anthopoulos. “Spray-deposited Li-doped ZnO transistors with electron mobility
exceeding 50 cm(2)A^s”, Adv. Mater. 22,4764 (2010)
> G. Adamopoulos, A. Bashir, W. P. Gillin, S. Georgakopoulos, M. Shkunov, M. A.
Baklar, N. Stingelin, D. D. C. Bradley, T. D. Anthopoulos. “Structural and electrical
characterization of ZnO films grown by spray pyrolysis and their application in thin-
film transistors”, Adv. Funct. Mater. 21, 525 (2011)
> S. Georgakopoulos, Y. Gu, M. M. Nielsen, M. Shkunov. “Air-stable Ti-eonjugated
amorphous copolymer field-effect transistors with high mobility of 0.3cm2/Vs”,
Appl. Phys. Lett. 101, 213305 (2012)
> S. Georgakopoulos, D. Sparrowe, F. Meyer, M. Shkunov. “Polymer source-gated
transistors”, Org. Electron. [In press (submitted 07/2012), under review]
> S. Georgakopoulos, H. Becker, K. Hoydalsvik, D. W. Breiby, D. Sparrowe, F.
Meyer, M. Shkunov. “Charge transport in field-effect transistors with amorphous
indenofiuorene 7t-conjugated copolymers analyzed with the Gaussian disorder model
“, J. Appl. Phys. [to be submitted]
116
8.4 List of oral and poster presentations
> “Organic field-effect transistors”, Oral presentation, Nano-Electronies Centre
meeting, 2009, Guildford, UK
> “Air-stable amorphous polymer semiconductor field-effect transistors”. Oral
presentation. Prototyping in the Micro and Nano Scale workshop, 2009, Haifa, Israel
> “Electrical performance of air-stable amorphous polymer semiconductors in top-gate
and bottom-gate organic field-effect transistor architectures”. Oral presentation,
Merck CASE conference, 2009, Southampton, UK
> “Dual-gate amorphous polymer field-effect transistors”. Poster presentation, One-P
summer school: Organic nanomaterials for electronics and photonics, 2010, Erice,
Italy
> “Top-Gate and Bottom-Gate amorphous polymer field-effeet transistors”. Oral
presentation, Merck CASE conference, 2010, Southampton, UK (unable to attend
due to voleano/ash cloud disruption - given by M. Shkunov)
> “Performance of bottom-gate and top-gate amorphous polymer field-effect
transistors”. Poster presentation. Technology for Polymer Electronics, 2010,
Rudolstadt, Germany (unable to attend due to voleano/ash cloud disruption -
abstract published in proceedings)
> “Organic source-gated transistors”. Oral presentation, Nano-Electronics Centre
meeting, 2010, Guildford, UK
> “Organic source-gated transistors”. Poster presentation. Materials Research Society
Fall, 2010, Boston, USA.
> “Polymer Schottky-barrier transistors”. Oral presentation, Merck CASE conference,
2011, Southampton, UK
> “Polymer source-gated transistors”. Poster presentation, European Conference on
Molecular Electronics, 2011, Barcelona, Spain
117
> “Polymer Sehottky-barrier transistors”, Oral presentation. Technology for Polymer
Electronics, 2012, Rudolstadt, Germany
> “High ionization potential amorphous jr-eonjugated polymer field-effeet transistors
with high mobility of 0.3 em2A/s”, Oral Presentation, International Symposium on
Flexible Organic Electronics, 2012, Thessaloniki, Greece
> “Polymer Schottky-barrier transistors”. Oral Presentation, International Symposium
on Flexible Organic Electronics, 2012, Thessaloniki, Greece
> Quarterly meetings with Merck Chemicals for the first three years with oral
presentations
118
Reproduced with permission of copyright owner. Further reproduction prohibited without permission.