Organic Lateral Heterojunction Devices for

Vapor-phase Chemical Detection

by
John C. Ho
Submitted to the Department of Electrical Engineering and Computer
Science
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy in Electrical Engineering and Computer Science
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
June 2009
©2009, Massachusetts Institute of Technology. All rights reserved.

Author . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Department of Electrical Engineering and Computer Science
May 21, 2009

Certified by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Vladimir Bulović
Professor
Thesis Supervisor

Accepted by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Terry P. Orlando
Chairman, Department Committee on Graduate Students
2
 Organic Lateral Heterojunction Devices for Vapor-phase
Chemical Detection
by
John C. Ho

Submitted to the Department of Electrical Engineering and Computer Science

on May 21, 2009, in partial fulfillment of the
requirements for the degree of
Doctor of Philosophy in Electrical Engineering and Computer Science

Abstract
As the U.S. is engaged in battle overseas, there is an urgent need for the develop-
ment of sensors for early warning and protection of military forces against potential
attacks. On the battlefields, improvised explosive devices (IEDs) have resulted in
54% of all coalition deaths in Iraq and 59% in Afghanistan. The U.S. military has re-
sponded with an intensive program of technology development, spending $12.4 billion
over the past three years on counter-IED equipment, technology R&D, training, and
other measures through the Joint Improvised Explosive Device Defeat Organization
(JIEDDO). “Sniffing”technology, based on fluorescent polymers, has emerged as one
of the most sensitive tools in the military’s explosives-detecting arsenal. Fluorescent
polymer sensors have demonstrated effective ultra-trace, vapor-phase detection of
trinitrotoluene (TNT) at security checkpoints and during vehicular sweeps, achieving
sensitivity levels comparable to the canine nose. In this sensor scheme, a large contri-
bution to the noise signal is the inefficient transduction of the polymer’s chemically
sensitive, photoluminescent signal into an electrical signal. The poor optical coupling
of the polymer’s photoluminescence into a photodetector, reduces the signal-to-noise
ratio (SNR) and overall sensor performance. In this work we have developed a novel
transduction mechanism and device structure that enable more efficient photon-to-
electron conversion resulting in direct transduction of the chemical signature into an
electrical signal. Device models detailing the physical processes of device operation
will be presented along with supporting experimental evidence. In addition, detec-
tion of as little as 1 picogram of TNT has been demonstrated. Ultimately, this work
results in a solid-state sensor platform that can be readily engineered to accept any
chemosensitive fluorescent polymer for a variety of ultra-trace sensing applications
with potential uses in the medical diagnostic, industrial processing, environmental,
and defense industries.

Thesis Supervisor: Vladimir Bulović

Title: Professor

3
4
Acknowledgments

First and foremost, I want to thank my “adventure buddy” Sara for being by my side
from day one of my Ph.D. I have and will always maintain that she is my “lucky”
charm, and as luck would have it I met her just as I was starting on the sensor project.
I am most grateful for her unconditional patience and tireless support over the last
several years as I navigated the tricky, winding road towards a degree. Getting my
degree is that much more rewarding knowing that I get to share my achievement with
her. She was there through all the highs and the lows and still found a way to love
me anyway. I look forward to this next chapter in our lives, and I can’t wait to get
started.

To my family, thank you for providing me with a support network outside of MIT.
I was lucky to have all of them so close to Cambridge during my time at MIT. It’s a
blessing that few other students at MIT can have.

I also have to thank my advisor and friend, Vladimir, for allowing me to be a part
of his lab over the last seven years. His genuine enthusiasm for science and creativity
are second to none, and I am grateful to have worked with him. It was his trust in
my work that allowed me to grow and thrive as a scientist.

To my mentor and officemate, Alexi, I owe a great debt. When I was a clueless
student with no research experience, Alexi was an excellent role model, tutor, teacher,
and friend. We went through MIT on parallel tracks, and we helped each other to stay
on course. I will never forget his limitless patience and excellent scientific intuition,
which I relied on throughout our time together. I can honestly say that my experience
at MIT has been better for knowing and working with him.

I can’t forget to mention Jill and Tim for their curiosity and youthful energy,
which enabled me to see my research in new ways. Jill’s been a great help to me as
soon as she joined our group, and I will always appreciate her warmhearted nature.
She will make a great professor someday. Tim’s been a very generous and caring
office mate, and I will miss his inquisitive mind and probing “what if” scenarios. I
look forward to great things from him.

5
 To Aimee, thank you for sharing this important research with me. Without her
chemistry skills and tireless heroics on my behalf, none of this would be possible. I
sincerely hope we get the chance to work together in the future.
To my “brother from another” lab, Ivan. Thanks for always providing the comic
relief that is so hard to find at a graduate school like MIT. I will miss our day-to-day
interactions, but I know that we’ll find ways to stay connected. After all, how many
Martin Lawrence fans can there be?
Finally, to the rest of the members of LOOE (former and present), who have
contributed to this work in countless ways, thank you. We have one of the great
research groups at MIT, and that’s a direct product of all of the wonderful people
in our group. It was my honor to work with all of you, and I can’t express enough
gratitude for that.

6
Contents

1 Introduction 29
1.1 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.1.1 Engineering Motivations . . . . . . . . . . . . . . . . . . . . . . . 35
1.1.2 Scientific Motivations . . . . . . . . . . . . . . . . . . . . . . . . . 36
1.2 Structure of Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

2 Introduction to Organic Semiconductor Properties and Applications 41

2.1 Concepts from Quantum Mechanics and Solid-state Physics . . . . . . 42
2.1.1 Atomic Orbitals: Particle in a Box . . . . . . . . . . . . . . . . . 42
2.1.2 Molecular Orbital Theory . . . . . . . . . . . . . . . . . . . . . . 46
2.1.3 Spectroscopic Transitions . . . . . . . . . . . . . . . . . . . . . . 48
2.1.4 Excitons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2.2 Organic Thin Film Properties . . . . . . . . . . . . . . . . . . . . . . . . 53
2.3 Overview of Organic Electronics Industry . . . . . . . . . . . . . . . . . 54
2.3.1 Manufacturing Processes: Printing is the Future . . . . . . . . . 55
2.3.2 Organic Electronic Products: OLEDs Light the Way for Adoption 61
2.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

3 Vapor-phase Chemical Detection 73

3.1 Survey of State-of-the-art Vapor-phase Solid-state Chemosensing Or-
ganic Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.1.1 Electrical Odor Sensors . . . . . . . . . . . . . . . . . . . . . . . . 75
3.1.2 Optical Odor Sensors . . . . . . . . . . . . . . . . . . . . . . . . . 87

7
 3.1.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

4 Bilayer Heterojunction Photoconductors 111

4.1 Organic Heterojunction Theory . . . . . . . . . . . . . . . . . . . . . . . 111
4.1.1 Excitons at Heterojunction Interfaces . . . . . . . . . . . . . . . 113
4.1.2 Energy Band Alignment . . . . . . . . . . . . . . . . . . . . . . . 114
4.2 Photoconductor Device Theory . . . . . . . . . . . . . . . . . . . . . . . 122
4.2.1 Carrier Generation . . . . . . . . . . . . . . . . . . . . . . . . . . 123
4.2.2 Carrier Transport . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
4.2.3 Figures of Merit . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
4.3 Bilayer Heterojunction Photoconductor . . . . . . . . . . . . . . . . . . 134
4.3.1 Device Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
4.4 Model System: TPD/PTCBI Bilayer Heterojunction Photoconductors 140
4.4.1 Current-Voltage Characteristics . . . . . . . . . . . . . . . . . . . 142
4.4.2 Spectral Response . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
4.4.3 Intensity Measurements . . . . . . . . . . . . . . . . . . . . . . . 147
4.4.4 Transient Measurements . . . . . . . . . . . . . . . . . . . . . . . 151
4.5 Other Material Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
4.5.1 CdSe Quantum Dot/Spiro-TPD Heterojunction Photoconductor 155
4.5.2 TDBC J-aggregate dye/Zinc Indium Oxide Heterojunction Pho-
toconductor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
4.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168

5 Multilayer Heterojunction Photoconductors 177

5.1 Quantum Well in an Electric Field . . . . . . . . . . . . . . . . . . . . . 177
5.2 n-i-p-i Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
5.3 Trilayer Photoconductor Model . . . . . . . . . . . . . . . . . . . . . . . 180
5.4 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
5.4.1 Cascading Energy Bands . . . . . . . . . . . . . . . . . . . . . . . 183
5.4.2 Alternate Energy Band Alignments . . . . . . . . . . . . . . . . 192
5.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206

8
6 Bilayer Heterojunction Sensors for Vapor-phase Explosives Detec-
tion 213
6.1 Device Model and Theory of Operation . . . . . . . . . . . . . . . . . . 214
6.2 Material Sets and Fabrication . . . . . . . . . . . . . . . . . . . . . . . . 216
6.2.1 Fluorescent Conjugated Polymers . . . . . . . . . . . . . . . . . 216
6.2.2 Metal Oxides . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
6.3 Proof of Concept: ZZ-TOP and SnO2 . . . . . . . . . . . . . . . . . . . 226
6.3.1 Developing a Sensor Platform Technology . . . . . . . . . . . . 239
6.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243

7 Concluding Remarks and Future Areas of Research 249

7.1 Research Highlights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
7.2 Research Opportunities . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250

9
10
List of Figures

1-1 (a) Chemical structure of 2,4,6-trinitrotoluene (TNT) and (b) a fluo-

rescent polymer with pentiptycene molecules incorporated in backbone. 30

1-2 (a) Fluorescent conjugated polymer solvated in toluene. (b) Digital

image depicting a thin film of fluorescent polymer before and after
exposure to TNT vapor. . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

1-3 (a) An illustration detailing a state-of-the art sensing scheme utilizing

a chemosensitive fluorescent polymer. (b) Image of commercially avail-
able Fido XT— explosives detector manufactured by ICx Technologies. 32

2-1 Schematic of 1-D particle in a box. Spatial confinement leads to dis-

crete energy levels and wavefunctions (Ψ) following the inset equations,
where E is energy, V is potential energy, m is mass, L is the length of
the box, and h is Plancks constant. . . . . . . . . . . . . . . . . . . . . . 44

2-2 Table of 3-D visualizations of hydrogen atomic orbitals with varying

quantum numbers [2]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

2-3 Schematic of energy level splitting that occurs in molecular orbitals

of diatomic Li. σ (in-line) and π (out of plane) bonding orbitals are
differentiated by the way the atomic orbitals overlap, * indicates anti-
bonding orbitals [4]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

11
2-4 Illustration of the Franck-Condon shift observed in electronic transi-
tions. Vgnucl and Venucl are the nuclear potential energy functions of the
ground and excited electronic states respectively. ∆EgF C and ∆EeF C are
the Franck-Condon relaxation energies of the ground state and excited
state respectively. E↑el and E↓el are the dominant electronic excitation
and relaxation transitions respectively [6]. . . . . . . . . . . . . . . . . . 51

2-5 Illustration of the different types of excitons. After [7]. . . . . . . . . . 52

2-6 A typical Jablonski diagram that illustrates many of the possible elec-
tronic transitions associated with a molecule [8]. . . . . . . . . . . . . . 53

2-7 Schematic illustrations of a) flatbed screen printing and b) rotary

screen printing [21]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

2-8 Using a laboratory printing machine, printed electronics, such as RFID

tags, are being developed by PolyIC. By using a roll-to-roll flexography
printing process it is possible to produce large volumes at very low cost.
Source: PolyIC press picture . . . . . . . . . . . . . . . . . . . . . . . . . 58

2-9 (a) Diagram of rotogravure printing process (b) Multi-color rotogravure

printing machine. Source: R.H. Technocrates Center . . . . . . . . . . 59

2-10 Schematic of thermal inkjet printing process [23]. . . . . . . . . . . . . 60

2-11 (a) Cross-sectional view of piezoelectric ink chamber (b) Dimatix’s

benchtop DMP 2800 piezoelectric inkjet printer. Source: Dimatix . . 60

2-12 Cross sectional schematic view of the molecular jet print head and the
steps involved in deposition [23]. . . . . . . . . . . . . . . . . . . . . . . . 61

2-13 (a) Revenue projections for OE markets ($ Billions) (b) Revenue pro-
jections for OE by application, 2014 ($ Millions). Source: Nanomarkets
LC (2007) [26] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

2-14 Prototype 40” OLED display (worlds largest). Source: Samsung . . . 63

2-15 White, small molecule OLED measuring 4 x 12 cm2 . Source: Philips

Lighting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

2-16 Organic rectifier printed on 6” PEN-foil. Source: IMEC . . . . . . . . 65

12
2-17 Demonstration of flexible organic photovoltaic technology. Source:
Konarka . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

2-18 Roadmap for organic electronics applications. Source: OE-A [32]. . . 67

3-1 Classification tree of different chemical sensor technologies. Lighter

boxes will not be covered in this chapter. . . . . . . . . . . . . . . . . . 76

3-2 (a) Schematic of a typical chemiresistor device using interdigitated elec-

trodes. There are two steady state conditions during operation: 1) no
analyte present in the sampling environment, and 2) analyte present
in the sampling environment resulting in a change in resistivity of the
film. (b) Three-dimensional pattern depicting representative carbon
black chemiresistor responses to seven test analytes [8]. . . . . . . . . . 77

3-3 (a) Schematic of a typical chemicapacitor device using micro-machined,

parallel plate electrodes. In this scheme, swelling of the film causes
changes in both the dielectric constant () and the voltage (V) which
contribute to changes in capacitance. (b) Sample chemicapacitor re-
sponse of the polymer, siloxanefluoro alcohol (SXFA), to dimethyl
methylphosphonate (DMMP) in a background of dry nitrogen and ni-
trogen with 40% relative humidity [33]. . . . . . . . . . . . . . . . . . . 83

3-4 (a) Schematic of a typical OTFT gas sensor with a patterned gate elec-
trode. The simultaneous measurement of multiple device properties,
yields more information and makes OTFT’s a more attractive sensor
platform. (b) Fractional changes in drain current (Ion ) for eight semi-
conductors upon exposure to four odorants: toluene, eugenol, vanillin,
and l-carvone [45]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

3-5 Kretschmann configuration of SPR gas sensing method. The graph

indicates a real-time measurement of intensity at a given wavelength
and angle of reflected light. . . . . . . . . . . . . . . . . . . . . . . . . . . 90

13
3-6 (a) Extrinsic optical fiber sensor, where fibers are used to guide input
and output signals (b) Intrinsic optical fiber sensor, where the sensing
film is incorporated into fiber structure as a cladding material. . . . . 92

3-7 Schematic of a typical configuration of a PL gas sensor. The LED

is used to excite the sensing film, while the filter acts to prevent the
excitation from reaching the photodetector. Upon introduction of a
gaseous analyte, the sensing film’s luminescence is quenched, producing
a reduction in the PL intensity measured by the photodetector. . . . . 95

4-1 (a) Type-I (Straddling) heterojunction (b) Type-II (Staggered) hetero-

junction (c) Type-III (Broken-gap) heterojunction. . . . . . . . . . . . 112

4-2 (a) Cross section and (b) energy band diagram of a typical OLED
device consisting of a transparent indium tin oxide (ITO) anode, a
hole transport layer (HTL), an electron transport layer (ETL), and a
metal cathode. q is the charge of an electron and VF is the forward
voltage bias. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

4-3 (a) Cross section and (b) energy band diagram of a typical photovoltaic
device consisting of a transparent indium tin oxide (ITO) anode, an
acceptor material (A), a donor material (D), and a metal cathode. q
is the charge of an electron and VR is the reverse voltage bias. . . . . 115

4-4 Schematic of relevant energy band parameters: ionization energy (I),

electron affinity (A), and work function (ΦF ). After [7]. . . . . . . . . . 116

4-5 (a) Electron density in metal with tailing at the surface to form a dipole
layer. (b) Potential energy [V(x)] plot vs. distance of an electron a
distance x from a dipole layer. Fermi level (EF ) is uniform throughout
solid. Φm 1 and Φm 2 are work functions of surfaces with different crystal
orientation. After [7] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

14
4-6 (a) Potential surface of a metal and organic which are not in contact.
(b) Contact of the metal and a thin film of organic causes the common
potential barrier to drop. (c) Schematic representing common VL at
interface. ΦB n /ΦB p denote the injection barriers for electrons/holes.
(d) Energy diagram showing shift of VL (∆) due to dipole formation
at the interface [7]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
4-7 Schematic diagrams of organic-organic heterojunctions, where ΦF is
work function, EF is Fermi level, A is electron affinity, I is ionization
potential, and Eg is the bandgap. Subscripts denotes which material
the properties belong to. (a) depicts a situation where ΦFI <AII before
and after contact. (b) depicts ΦFI >III before and after contact. (c)
depicts ΦFI >AII before and after contact. After [13]. . . . . . . . . . . 121
4-8 Simulated exciton carrier concentration profile across a thin film of
organic semiconductor with L = 50 nm, α = 5x105 cm-1 , D = 1x10-3
cm2 s-1 [19], 100 mW/cm2 of 2.25 eV photons, and various boundary
conditions. SL =S0 =0 (violet), SL =S0 =1020 (green), SL =0 and S0 =1020
(blue), SL =1020 and S0 =0 (red). . . . . . . . . . . . . . . . . . . . . . . . 126
4-9 Representative current-voltage characteristics for a blocking contact
(green), an ohmic contact (red), and an injecting contact (blue). . . . 128
4-10 The interdigitated gold fingers form 60 photoconductive channels each
(1000 ± 1) µm long and (10 ± 1) µm wide yielding an effective device
area of (0.6 ± 0.06) mm2 . (a) An optical microscope image of a set of
device electrodes. (b) Energy band diagram of a lateral bi-layer het-
erojunction photoconductor consisting of an exciton generation layer
(EGL) and a charge transport layer (CTL). (c) A cross-sectional view
of the same bi-layer device is also shown. Both illustrations depict the
physical processes involved in steady-state device operation: (1) Light
absorption in EGL, (2) Exciton diffusion through EGL, (3) Exciton
dissociation at EGL/CTL interface and charge transfer to CTL, and
(4) Charge transport in the photoconductive channel. . . . . . . . . . . 137

15
4-11 Absorption spectra (50 nm thick films), chemical structure, and pro-
posed energy band levels of the materials used in the bi-layer hetero-
junction photoconductor: N,N’-bis(3-methylphenyl)-N,N’-diphenyl-1,1’-
biphenyl-4,4’-diamine (TPD) as CTL and 3,4,9,10-perylenetetracarboxylic
bis-benzimidazole (PTCBI) as EGL. Also shown is the emission pro-
file of the green LED (dotted line, λpeak = 532 nm) operated with a
forward current of 2 A and an intensity of 40 mW/cm2 . Energy band
levels represent experimental values for the HOMO/LUMO based on
photoemission spectroscopy data [6, 14]. As shown, the energy bands
correspond to transport levels that provide upper and lower limits for
the optical bandgap and emphasize the Type-II heterojunction between
TPD/PTCBI. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

4-12 Semi-logarithmic plot of the current-voltage characteristics from a se-

ries of lateral photoconductor devices using TPD and PTCBI. Current-
voltage sweeps are taken in the dark (dashed) and under illumina-
tion (solid) through the glass substrate from a 40 mW/cm2 green
LED (λpeak ∼ 532 nm). The inset cross-sectional diagrams describe
the thicknesses and device structures. Both heterojunction devices
(Au/PTCBI/TPD [green] and Au/TPD/PTCBI [red]) exhibit steady-
state photoresponse. The Au/TPD device (black) exhibits no photore-
sponse to the green LED, consistent with its absorption spectrum. The
Au/PTCBI device (blue) exhibits a steady-state photoresponse with
lower currents than heterojunction devices. . . . . . . . . . . . . . . . . 144

4-13 (a) Log-log plot of current-voltage characteristics for a series of TPD/PTCBI

devices under illumination. The dashed lines are provided to show the
transition between ohmic and SCLC in the blue curve (500 nm TPD/50
nm PTCBI). Inset: a circuit model of conduction pathways in the bi-
layer devices, where the relative size of the resistor is indicated. (b) A
schematic drawing depicting the non-linear sidewall growth that can
occur with thin films evaporated on bottom contacts. . . . . . . . . . . 146

16
4-14 Semi-logarithmic plot of external quantum efficiency (EQE) versus
wavelength, for a series of device structures illustrated in the inset.
Electric fields applied during the measurement are indicated to the
left of each curve. The Au/TPD device yields no measurable pho-
toresponse. Both heterojunction devices (solid and dashed lines) ex-
hibit over 1000-fold improvement in the EQE over the Au/PTCBI de-
vice (dotted line), at 2-fold lower bias voltages. At λ = 552 nm, the
Au/PTCBI/TPD device and Au/TPD/PTCBI device yield EQEs of
(8 ± 1)% and (12 ± 1)%, respectively. Accounting for the absorption
of λ = 552 nm light in the PTCBI/TPD device, we find an internal
quantum efficiency (IQE) of (140 ± 2)%. This result demonstrates that
lateral bi-layer heterojunctions are capable of quantum gain. . . . . . 148

4-15 Log-log I-V characteristics of 50 nm TPD/50 nm PTCBI device under

varying levels of green LED illumination. Inset: Relative photocurrent
as a function of relative intensity reveals sublinear dependence. . . . . 150

4-16 Simulated variation of photocurrent with light intensity for the Tabak-
Warter model, where L is carrier diffusion length and α is absorption
coefficient. Breakpoints between different regimes of dependence are
indicated by dashed red lines. After [34]. . . . . . . . . . . . . . . . . . 151

4-17 (a) Semi-logarithmic plot of normalized photocurrent decay under var-

ious intensity illumination and an applied electric field of 15 V/µm.
Light is turned off at time t = 0. Dark current is subtracted from all
curves before fitting double exponentials over the range 2.8 µs to 12
µs. (b) Fitted time constants plotted versus intensity. . . . . . . . . . . 153

17
4-18 (a) Section view of the device structure. The thickness of the Spiro-
TPD layer is 200 nm. For each device there are 60 sets of Au electrodes
arranged in an interdigitated array forming a serpentine channel (not
shown). The channel length is 10 µm and the electrode width is 1000
µm. (b) Chemical structure of the hole-transporting material, Spiro-
TPD. (c) Schematic of CdSe nanocrystal (∼ 5-7 nm) passivated with
TOPO ligands [42]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

4-19 (a) Current-voltage characteristics measured for Spiro-TPD (control)

and Spiro-TPD/QD (QD) devices. Dashed lines represent dark cur-
rents. Solid lines show the device response under illumination from a
green LED with an intensity of 74 mW/cm2 and with emission cen-
tered at λ = 521 nm. The inset shows the energy band diagram for
the QD device. The energy levels for the Spiro-TPD were taken from
[43]. Energy levels for the CdSe QDs were calculated following the ap-
proach reviewed in [44] and using bulk CdSe parameters obtained from
[45]. (b) Absorption spectra of Spiro-TPD (dashed line) and CdSe QDs
(solid line). Emission from the LED (dotted line) is absorbed primarily
by the QD film [42]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157

4-20 Log-log current-voltage characteristics of various Spiro-TPD/QD de-

vices. Dotted lines represent single films of Spiro-TPD, while solid
lines represent the QD-stamped bilayer devices. The magenta dotted
line is a guide to the eye that represents the functional form of SCLC. 160

4-21 EQE spectra for a variety of biases. Inset shows EQE at λ = 590
nm as a function of bias. Photocurrent follows a nearly V 2 power law
indicative of space charge limited conduction in Spiro-TPD. . . . . . . 161

4-22 Absorption spectra of TDBC dye in monomer phase (blue) and J-

aggregate phase (red). Inset: Chemical structure of TDBC dye. . . . . 163

4-23 Absorption spectra of ZIO and TDBC films. Emission spectrum of

amber LED centered at λ = 595 nm is also shown. . . . . . . . . . . . 165

18
4-24 Log-log current-voltage characteristics for a series of devices contain-
ing ZIO and TDBC. I-V sweeps are collected under both light (solid)
and dark (dashed) conditions. Illumination is provided by an amber
LED (λpeak = 595 nm). Inset cross-sectional diagrams detail device
structures and film thicknesses. . . . . . . . . . . . . . . . . . . . . . . . 166

4-25 Plot of photocurrent decay for a 50.0 ± 0.5 nm ZIO/5.0 ± 0.5 nm TDBC
bilayer heterojunction photoconductor. Illumination by an amber LED
(λpeak = 595 nm) is turned off at time t = 0 s. . . . . . . . . . . . . . . . 167

4-26 Spectral response data for a 50.0 ± 0.5 nm ZIO device and a 50.0 ±
0.5 nm ZIO/5.0 ± 0.5 nm TDBC bilayer device. Inset cross-sectional
diagrams detail device structure and film thicknesses. An electric field
of 1 x 106 V/µm is applied to both devices. . . . . . . . . . . . . . . . . 168

5-1 Schematic of energy bands and wavefunctions of electrons and holes in

a quantum well under an applied electric field. . . . . . . . . . . . . . . 178

5-2 Schematic of a n-i-p-i doping superlattice. Space charge regions be-

tween the doped slices creates insulating zones. The conduction and
valence band edges modulate parabolically with spatially separated
electron and hole wavefunctions in their ground states. . . . . . . . . . 181

5-3 Energy band diagrams of (a) a bilayer device consisting of CTL/EGL

and (b) a trilayer device consisting of CTL/CSL/EGL. Relevant inter-
facial recombination rates and carrier diffusion rates are depicted. . . 183

5-4 Chemical structures of materials used in trilayer devices. Clockwise

from top-left corner: N,N’-bis(3-methylphenyl)-N,N’-diphenyl-1,1’-biphenyl-
4,4’-diamine (TPD), 3,4,9,10-perylenetetracarboxylic bis-benzimidazole
(PTCBI), N,N’-diphenyl-N,N’-bis(1-naphthyl)-1,1’-biphenyl-4,4”-diamine
(α-NPD), oleic acid-capped lead selenide quantum dot (4.7 nm PbSe
QD), p-bis(triphenylsilyl)benzene (UGH2), tris (8-hydroxyquinolinato)
aluminum(III) (Alq3 ). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184

19
5-5 Absorption spectra of TPD, Alq3 , and PTCBI films (50 nm thick).
The green LED emission spectrum (λpeak = 532 nm), operated at an
intensity of 40 mW/cm2 , is included for reference (dashed line). Inset:
Suggested energy bands of TPD [14], Alq3 [15], and PTCBI [16] based
on photoemission spectroscopy data and optical bandgaps and digital
images of materials in purified, powder form. . . . . . . . . . . . . . . . 185

5-6 Set of AFM images of various films consisting of (a) 50 ± 0.5 nm TPD
(b) 50 ± 0.5 nm of TPD/6.0 ± 0.5 nm Alq3 (c) 50 ± 0.5 nm TPD/6.0 ±
0.5 nm Alq3 /50 ± 0.5 nm PTCBI. All images have same height range of
5 nm. Low RMS roughness in all cases suggests good wetting properties
of deposited films. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186

5-7 Scatter plot of nine separate trilayer device trials consisting of 50 ±

0.5 nm TPD/Alq3 /50 ± 0.5 nm PTCBI, with varying Alq3 thickness.
Quantum efficiency is plotted versus the thickness of Alq3 and suggests
no definite trend. All quantum efficiencies are obtained with the same
applied field and wavelength and normalized to the control (bilayer
device) of that particular device trial. . . . . . . . . . . . . . . . . . . . 187

5-8 Scatter plot of eight separate trilayer device trials consisting of 50 ±

0.5 nm TPD/Alq3 /50 ± 0.5 nm PTCBI, with varying Alq3 thickness.
Dark current is plotted versus Alq3 thickness and suggests an overall
increase due to the presence of Alq3 . All current levels are obtained at
the same applied field and are plotted relative to the current levels of
the bilayer (control) device for that particular device trial. . . . . . . . 188

5-9 Absorption spectra of TPD, NPD, and PTCBI (50 nm thick). The
green LED emission spectrum (λpeak = 532 nm), operated at an in-
tensity of 40 mW/cm2 , is included for reference (dashed line). Inset:
Suggested energy bands of TPD [14], NPD [15], and PTCBI [16] based
on photoemission spectroscopy data and optical bandgaps. . . . . . . . 190

20
5-10 Log-log current-voltage characteristics of 50 ± 0.5 nm TPD/NPD/50
± 0.5 nm PTCBI, with 0, 3.5, and 4 ± 0.5 nm of NPD. Arrows indicate
increasing NPD thickness and dashed red lines are guides that depict
ohmic and trap-limited conduction. . . . . . . . . . . . . . . . . . . . . . 191

5-11 Scatter plot of three separate trilayer device trials consisting of 50 ±

0.5 nm TPD/NPD/50 ± 0.5 nm PTCBI, with varying NPD thickness.
Light current (▽) is plotted versus NPD thickness and suggests an
overall increase due to the presence of NPD up to 4 nm. All current
levels are obtained at the same applied field and are normalized to the
current levels of the bilayer (control) device for that particular device
trial. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193

5-12 Scatter plot of three separate trilayer device trials consisting of 50 ±

0.5 nm TPD/NPD/50 ± 0.5 nm PTCBI, with varying NPD thickness.
External quantum efficiency is plotted versus NPD thickness at λ = 368
nm (∎) and at λ = 532 nm (◻), suggesting an increase in photoresponse
due to the presence of NPD. All measurements are obtained at the
intensity and are normalized to the photocurrent levels of the bilayer
(control) device for that particular device trial. Larger electric fields
result in larger improvements. . . . . . . . . . . . . . . . . . . . . . . . . 194

5-13 Absorption spectra of TPD, UGH2, and PTCBI (50 nm thick). The
green LED emission spectrum (λpeak = 532 nm), operated at an in-
tensity of 40 mW/cm2 , is included for reference (dashed line). Inset:
Suggested energy bands of TPD [14], UGH2 [30], and PTCBI [16] based
on photoemission spectroscopy data and optical bandgaps. . . . . . . . 196

5-14 Log-log current-voltage characteristics of a TPD/PTCBI bilayer device

(black), and two TPD/UGH2 (1.8 nm, 6.9 nm)/PTCBI trilayer devices
(red and green respectively). . . . . . . . . . . . . . . . . . . . . . . . . . 197

21
5-15 Set of AFM images of various films consisting of (a) 50 ± 0.5 nm
TPD/4 ± 0.5 nm UGH2 (b) 50 ± 0.5 nm of TPD/50 ± 0.5 nm UGH2.
All images have same height range of 100 nm. Island growth suggests
poor wetting properties of UGH2 deposited on TPD. . . . . . . . . . . 198

5-16 External quantum efficiency of TPD/UGH2/PTCBI trilayer device

versus UGH2 thickness at λ = 368 nm (∎) and at λ = 532 nm (∎).
Various applied electric fields from 10 V/µm to 30 V/µm are indicated
in the legend. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199

5-17 Absorption spectra of TPD, PTCBI (each 50 nm thick) and PbSe (10
nm). The green LED emission spectrum (λpeak = 532 nm), operated
at an intensity of 40 mW/cm2 , is included for reference (dashed line).
Inset: Suggested energy bands of TPD [14], PTCBI [16], and oleic
acid-capped PbSe QDs [39] based on photoemission spectroscopy data,
charge transport data, and optical bandgaps. . . . . . . . . . . . . . . . 202

5-18 Log-log current-voltage characteristics of a TPD/PTCBI bilayer device

(blue) and a TPD/PTCBI/PbSe QD trilayer device (red). Inset device
cross sections indicate relevant film thicknesses. . . . . . . . . . . . . . 203

5-19 Absorption spectra of stamped film of PbSe QD on glass (black), on

TPD (red), on TPD/PTCBI (blue). . . . . . . . . . . . . . . . . . . . . 204

5-20 Spectral photoresponse of a TPD/PTCBI bilayer device (blue) and a

TPD/PTCBI/PbSe QD trilayer device (red) with an applied electric
field of 10 V/µm. Inset device cross sections indicate relevant film
thicknesses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205

6-1 Illustration depicting cross-sectional views and energy band diagrams

of a chemosensitive bilayer heterojunction photoconductor. These il-
lustrations highlight the impact of the analyte on carrier concentrations
at the EGL/CTL interface. The sample graph of photocurrent vs. time
demonstrates the effect of the analyte on device photoconductivity. . 217

22
6-2 Schematic of a 4-level system representative of amplified stimulated
emission in a conjugated polymer. Ex are the energies, Nx are the
electron populations, and Rxy are the rates of transition between the
different levels of the 4-level system. After [4]. . . . . . . . . . . . . . . 221

6-3 An example sketch of the lasing threshold increase after exposure of

the CP film to a quenching analyte, such as TNT. The change in the
output emission intensity is largest at pump energies near the lasing
threshold of analyte-exposed films. After [19]. . . . . . . . . . . . . . . 225

6-4 Emission spectra of a 17 ± 1 nm thin film ZZ-TOP before (light green)

and after (dark green) exposure to saturated TNT vapor. Inset: Digital
image of ZZ-TOP thin film spin-coated on a glass substrate and excited
by UV illumination. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227

6-5 (a)Helium I UPS data from ITO sample (100 Ω/◻) and ITO coated
with 10.0 ± 0.5 nm film of ZZ-TOP polymer with no bias relative to
reference electrode. (b)Energy level schematic of UPS spectra super-
imposed over the energy bands of the sample materials. The work
function of ITO (Φm ) can be extracted from the onset of photoemit-
ted electron collection. Note the small vacuum level shift (∆) and
the large energy difference between the ZZ-TOP HOMO and the ITO
Fermi level (εFHOM O ). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228

6-6 Normalized absorption spectra of a 17.0 ± 0.5 nm thick film of ZZ-TOP

(green), a 100.0 ± 0.5 nm thick film of SnO2 (grey), and the emission
spectra of a blue (λpeak = 460 nm) and UV LED (λpeak = 385 nm).
Inset: Suggested energy bands of ZZ-TOP and SnO2 [25] based on
photoemission spectroscopy data and optical band gaps. . . . . . . . . 230

6-7 AFM images of various fluorescent polymers spin-coated onto glass

substrates and SnO2 films coated with those same polymers. All images
have the same height range of 20 nm. . . . . . . . . . . . . . . . . . . . 232

23
6-8 Log-log plot of the current-voltage characteristics from a series of lat-
eral photoconductor devices using SnO2 and ZZ-TOP. Current-voltage
sweeps are taken in the dark (dashed) and under illumination (solid)
through the glass substrate from a 30 mW/cm2 Blue LED (λpeak = 532
nm). The dashed red line is a guide for the eye and follows Ohm’s law.
The inset cross-sectional diagrams describe the thicknesses and device
structures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233

6-9 Illustration of photocurrent measurement setup incorporating a TNT

vapor delivery mechanism. . . . . . . . . . . . . . . . . . . . . . . . . . . 236

6-10 Semi-log spectral response characteristics for devices containing SnO2

and ZZ-TOP. All devices have same applied electric field (5 V/µm).
100.0 ± 0.5 nm SnO2 film (black) is less quantum efficient than 20.0 ±
0.5 nm SnO2 film (grey). 100.0 ± 0.5 nm SnO2 /17.0 ± 0.5 nm ZZ-TOP
before TNT exposure (green) is plotted against same device after TNT
exposure (dark green). Inset: Real-time plot of photocurrent decay at
a fixed wavelength of λ = 456 nm upon TNT vapor exposure. TNT
vapor is introduced at time t = 0 s. . . . . . . . . . . . . . . . . . . . . . 238

6-11 (a) Normalized emission spectra of various chemosensitive fluorescent

polymers. Inset: photograph of UV-excited polymers in solution and
thin film. (b) Chemical structures of various chemosensitive fluorescent
polymers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239

6-12 (a) Normalized absorption spectra of a 100.0 ± 0.5 nm SnO2 film (grey),
5.0 ± 0.5 nm HW film (orange), 17.0 ± 0.5 nm ED4 film (magenta),
18.0 ± 0.5 nm ZZ film (blue), and 17.0 ± 0.5 nm ZZ-TOP film (green).
Spectral response at 5 V/µm of (b) 20.0 ± 0.5 nm SnO2 /18.0 ± 0.5
nm ZZ (c) 100.0 ± 0.5 nm SnO2 /17.0 ± 0.5 nm ED4 (d) 20.0 ± 0.5
nm SnO2 /5.0 ± 0.5 nm HW before and after exposure to 300 cc/min
saturated TNT vapor. Insets: Photocurrent vs. time plot at (b) λ =
456 nm and (c)/(d) at λ = 500 nm and TNT vapor on at t = 0 seconds.241

24
6-13 Normalized photocurrent vs. time plot of various chemosensitive flu-
orescent polymer/metal oxide devices. Inset: Time constants from
exponential fits vs. flow rate of 20.0 ± 0.5 nm SnO2 /5.0 ± 0.5 nm HW
device. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242

25
26
List of Tables

2.1 Comparison of properties of molecular and covalent crystals. After [10]. 54

3.1 Commercially available and research-grade solid-state organic vapor-

phase chemical sensors and e-nose instruments as of May 2007. Abbre-
viations: Organic Conjugated Polymer (OCP), Phthalocyanine (Pc),
Metallophthalocyanine (MPc), Carbon Nanotube (CNT), Volatile Or-
ganic Compound (VOC), Chemical Warfare Agents (CWA), Toxic In-
dustrial Compound (TIC), Organic Thin Film Transistor (OTFT),
Surface Plasmon Resonance (SPR), and Photoluminescent (PL). . . . 97

27
28
Chapter 1

Introduction

A journey of a thousand miles must

begin with a single step.

LAO TZU
Chinese Philosopher

Devices incorporating organic (carbon-containing) semiconducting films have

steadily achieved commercialization starting with photoreceptors in xerography [1],
then progressing to organic light emitting diode (OLED) displays [2] and thin film
photovoltaics [3, 4], and finally to chemical sensors [5]. Chemical sensor research rep-
resents a rich field of study, utilizing a diverse array of technologies to accomplish the
difficult task of transducing sensory information into a quantifiable chemical, optical,
or electrical signals. The need to extract increasing amounts of sensory information
from our environment, fuels the development of sensors that are easier to manufac-
ture, capable of miniaturization, and more sensitive to vapors of interest. Vapor-phase
sensors have the advantage of employing a non-destructive, non-invasive method of
detection with the potential for high degrees of sensitivity. Depending on the vapor
pressure of the analyte (the chemical species being detected), some vapor-phase sen-
sors are capable of detecting as little as a femtogram of material (e.g. several thousand
molecules) [6]. A promising class of materials for ultra-trace, vapor-phase chemical
sensing is the fluorescent conjugated polymer [7]. Pioneering research by the Swager
Group in the Department of Chemistry at MIT, has led to the development of a poly-

29
Figure 1-1: (a) Chemical structure of 2,4,6-trinitrotoluene (TNT) and (b) a fluores-
cent polymer with pentiptycene molecules incorporated in backbone.

mer that retains its photoluminescent efficiency in thin films. Attached pentiptycene
molecules create scaffolding to allow the polymer strands to maintain adequate sep-
aration, preventing self-quenching processes in the solid state [Fig. 1-1]. In addition
to its favorable optical properties, this polymer is capable of exhibiting luminescence
quenching upon the introduction of TNT vapor [Fig. 1-2]. Once TNT molecules bind
to the polymer, excitons generated by photoluminescent excitation can diffuse along
the polymer backbone to transfer charge to the TNT molecules. The large exciton
diffusion lengths in conjugated polymers allow the TNT molecules to quench excitons
that are generated many monomer units away from the binding site. In this manner,
these conjugated fluorescent polymers can exhibit a large differential luminescence
signal with very small amounts of TNT.

Organic materials, such as fluorescent conjugated polymers, have the advantages

of being both synthetically flexible [8] and amenable to room temperature processing,
which allows for the development of an infinite library of chemical sensors at lower
costs. While using sensors to detect explosives or as medical diagnostics tools may
seem straightforward, each application presents a unique set of challenges and specifi-
cations. What may work for the detection of land mines underfoot may not necessarily
be useful for detecting disease within a human body. In fact, much of the bottleneck

30
Figure 1-2: (a) Fluorescent conjugated polymer solvated in toluene. (b) Digital image
depicting a thin film of fluorescent polymer before and after exposure to TNT vapor.

in commercializing chemical sensor research has been framing the right problem to
be solved with the given sensor technology, all while minimizing production costs.

1.1 Problem Statement

State-of-the-art sensors utilizing fluorescent polymers have been commercialized, with

success, in the area of ultra-trace vapor-phase explosives detection. Figure 1-3(a),
details a typical fluorescent polymer sensing scheme. Land mine and improvised
explosive device (IED) detection has played an increasingly important, life-saving role
in the war theaters of Iraq and Afghanistan. Developed by ICx Technologies, the Fido
XT— portable explosives detector [Fig. 1-3(b)] can be combined with autonomous
robots to safely find and disarm hidden explosives. The combined approach of using
the Fido— sensor in conjunction with the conventional method of canine explosives
detection, has proven effective in vehicular sweeps and at security checkpoints.
While the Fido— detector represents a technological milestone in explosives de-
tection, there is room for improvement and innovation. Despite representing the
world’s best TNT vapor detector with parts-per-quadrillion sensitivity, Fido— cannot
always adequately sense plastic explosive compounds (i.e. RDX) which have low va-
por pressures. In order to improve upon this chemical sensor, an important quantity

31
Figure 1-3: (a) An illustration detailing a state-of-the art sensing scheme utiliz-
ing a chemosensitive fluorescent polymer. (b) Image of commercially available Fido
XT— explosives detector manufactured by ICx Technologies.

to consider is the signal-to-noise ratio (SNR). There are two routes to improving
SNR, increasing the signal intensity or reducing the noise signal. Signal noise can
be attributed to both the optical and electrical components of the sensor. Thus, one
strategy is to investigate new sensor architectures that can simplify the sensor design
and reduce sources of electrical and optical noise.
For inspiration on ways to improve solid state fluorescent polymer chemical sen-
sors, we can investigate two related device structures - photoconductors and chemire-
sistors. Photoconductors consist of an absorptive material sandwiched between two
electrodes. When light is absorbed by the active material, excitons are generated and
dissociated within the bulk of the film and the photogenerated charge is measured as
an increase in conductivity between the two contacts. One useful feature of photo-
conductors is the ability to exhibit gain. Gain is the unitless ratio between the carrier
lifetime and the carrier transit time, reaching values above unity when the photogen-
erated carriers take longer to recombine than to traverse the distance between the
two contacts. Typical inorganic photoconductors have gains up to 106 . For higher
levels of gain, the fundamental tradeoff between device sensitivity and response time
becomes more pronounced as the photoconductor can take milliseconds to respond
to light at high gain levels [9].
Just as a photoconductor changes its conductivity with incident light intensity,
the chemiresistor responds with an analogous change in conductivity when exposed to

32
chemical analytes in the ambient environment. The chemiresistor consists of a chem-
ically sensitive film deposited between two electrical contacts, where analytes (the
species being detected) can alter the chemosensing film conductivity in two ways.
The analyte can either be absorbed to alter film density or be adsorbed to the surface
of the film. During analyte absorption, molecules diffuse through the film, increasing
the distance between neighboring conduction centers which reduces the overall con-
ductivity of the film. In this case the chemosensing film can be engineered to operate
at a percolation threshold that can result in large changes in conductivity for small
amounts of analyte [10]. During adsorption the analyte can either chemically react
with the film (chemisorption) by transferring charge, or the analyte can physically
adhere to the surface of the film through weak intermolecular forces (physisorption).

Chemiresistors offer a simple, direct transduction mechanism of a chemical signal

into an electrical signal. However, the lack of chemical specificity and relatively low
sensitivity levels limit the applications for this type of sensor. Chemical specificity
can be added to the active material through functionalization chemistry, however
this has a direct effect on film conductivity. Because a single chemoresistive film has
difficulty distinguishing specific analytes, oftentimes these sensors are incorporated
into sensor arrays that yield chemical distinction capabilities, but this comes with
the added complexity of calibration routines and algorithms needed to interpret the
aggregate signals.

In order to design a chemical sensor that can utilize the simple transduction mech-
anism of the chemiresistor while exhibiting electrical signal amplification similar to a
photoconductor, we can extend the paradigm of the bi-layer photoreceptor drums used
in xerography. The photoreceptor drum is comprised of two films, each supporting a
different physical function. The first film is referred to as the charge generation layer
(CGL) and typically consists of dye molecules suspended in a solid matrix. The dye
molecules act as absorption sites and generate excitons (bound electron/hole pairs)
in response to light excitation. The other film is the charge transport layer (CTL), an
electrically conductive film that transports the photogenerated charge from the CGL
to the surface where they can recombine with oppositely charged carriers deposited

33
by a corona field [1]. In addition to playing a key role in separating the functions
of charge generation and charge transport, the bi-layer heterojunction allows each
film in the photoreceptor to be optimized independently to change overall device
performance.

Observing that a chemosensitive fluorescent conjugated polymer can generate ex-

citons much like the CGL of a photoreceptor, we can imagine pairing the polymer
with a CTL to create a heterojunction that can dissociate excitons into charge car-
riers. Consider what happens if we incorporate a CTL and a fluorescent conjugated
polymer into a photoconductor/chemiresistor device structure. By virtue of utilizing
lateral electrodes, one surface of the device will be exposed to the environment, pro-
viding an opportunity to utilize the fluorescent conjugated polymer as the topmost
film. Additionally, the chemically active film will be relieved from having to transport
charge by the presence of an underlying CTL that can efficiently move charge through
the device.

In a chemosensitive fluorescent polymer the exciton generation rate of the film

not only depends on the intensity of incident light excitation, but also the presence
of analytes in the local environment [7]. The photoluminescence efficiency, η, of
chemosensitive fluorescent polymers can be strongly modulated by a chemical signal,
which is a signature of a change in the population of excitons that can radiatively
decay. The change in exciton population corresponds to a change in charge carrier
density at the heterojunction interface, which is measured as a change in photocur-
rent. Because the fluorescent polymer introduces excited charge carriers from its
photogenerated excitons, there remains the possibility that the carrier lifetimes can
exceed the transit times, amplifying the transduced chemical signal. The photo-
conductive gain is related to the incident light intensity, implying that the sensor’s
sensitivity can be tuned by changing the incident light signal, giving the sensor a large
dynamic range. In addition, the chemosensitive lateral heterojunction photoconduc-
tor can be driven with a modulated light signal, allowing for locked-in measurements
of the photogenerated, transduced chemical signal. These AC measurements remove
the contributions from the background DC current flowing through the CTL, remov-

34
ing the constraint of high signal-to-noise ratios that are a requirement for extracting
useful measurements from chemiresistors. The development of this novel solid-state,
vapor-phase chemical sensor platform will reflect the interdisciplinary nature of the
maturing Organic Optoelectronic field by utilizing thin film characterization and fab-
rication techniques (Materials Science), device characterization and microfabrication
(Electrical Engineering), and device modeling (Physics). As such, there are both
scientific and engineering motivations that underlie this work.

1.1.1 Engineering Motivations

Ideally, a chemical sensor technology would reliably detect a specific chemical analyte,
in real-time, with sensitivities ranging from a single molecule up to a few grams
enabling sensor usage in a wide range of applications. However, reality dictates that
chemical sensors must be engineered to satisfy the requirements of the specific sensing
application. Thus, it is useful to evaluate sensor technologies with the following
metrics: 1) Sensitivity 2) Specificity 3) Dynamic range 4) Response time and 5)
Durability. While a number of these attributes are determined by the chemosensitive
film alone, the key to overall sensor performance lies in the transduction mechanism
of the chemical signal into a measured quantity.
In a rugged, portable chemical sensor an electrical signal is favored over an op-
tical signal because precise alignment of optical components is not a requirement.
Fluorescent polymer sensing schemes utilize a photodetector to accomplish the trans-
duction from an optical signal into an electrical one. However, approximately 1% of
the luminescence signature is coupled into the photodetector, due to the lambertian
emission profile of the polymer and the physical distance between the film and the
photodetector. By increasing this coupling efficiency there stands to be a concomi-
tant increase in the SNR. Additionally, integrating the transduction mechanism into
the heterojunction structure itself, removes the need for filters and photodetectors,
which ultimately helps to reduce components and sensor complexity.
By separating the sensor functionality into a chemosensitive exciton generation
layer (EGL) and a CTL, we have essentially created a modular sensor platform with

35
interchangeable layers that can be engineered for optimum sensor performance. Using
a library of developed organic synthesis techniques, fluorescent conjugated polymers
can be tailored to detect particular analytes with high sensitivity, while maintaining
photoluminescent efficiency in the solid state. In addition, once the energy bands of
the polymer are determined, the polymer can be paired with an underlying CTL to
create a Type-II heterojunction that can effectively dissociate excitons. Film thick-
nesses and electrode configurations can also be varied to alter sensor sensitivity and
response times. Once the specifications for a sensing application are known, these
parameters become the engineering tools for solid-state sensor design.

1.1.2 Scientific Motivations

While organic heterojunctions play a key role in most organic optoelectronics (i.e.
photovoltaics, photoreceptors, and light emitting devices), the physical processes at
the interface between two dissimilar materials are not completely understood. Con-
troversy stems from the difficulty of directly measuring carrier concentrations and
the recombination rates in active device structures. Nearly all two terminal, organic
bilayer heterojunction devices are fabricated in a vertical stack, with the heterojunc-
tion sandwiched on the top and bottom by the two electrodes. In this configuration,
carrier recombination becomes difficult to distinguish from other field-dependent pro-
cesses such as carrier transport and exciton dissociation. With the electrodes in a
lateral, planar arrangement, the carrier transport now flows orthogonally to the car-
rier generation and recombination processes. This suggests that it may be possible
to experimentally observe the recombination lifetime at an organic heterointerface
with relatively simple measurements. Such measurements would fill a gap in our
understanding of organic semiconducting materials and would enable better designs
and simulations of organic devices, with the ultimate goal being the development of
simulation/design tools similar to those used in the inorganic semiconductor industry.
Another method to explore the recombination lifetime at a heterointerface is
through the fabrication of a multi-layer heterojunction device. The simplest version
of this structure consists of an EGL, a CTL, and a charge spacer layer (CSL) that

36
is placed between the CTL and the EGL to form a cascading energy band structure.
The thin CSL spatially separates the dissociated carriers to reduce the likelihood of
bimolecular recombination across the EGL/CTL interface. A thickness study of the
spacer layer can experimentally demonstrate the dependence of the carrier lifetime and
photoresponse efficiency on the spatial separation of dissociated charge. While there
are exotic experimental methods (i.e. Photo-CELIV [11], etc.) that have been used
to determine bimolecular recombination rates, the potential benefits to organic device
design that arise from having simple measurements and models to extract difficult-
to-measure properties (i.e. carrier lifetimes) form the scientific underpinnings of this
work.

1.2 Structure of Thesis

This work is organized into seven chapters. In Chapter 2, we begin by introducing
the field of organic optoelectronics and discussing some of the material properties
that make organic films unique. Chapter 3 follows with a review of state-of-the-art
solid-state vapor phase chemical sensors in order to build some context for the appli-
cations of this research. After this, the operational theory of bilayer heterojunction
photoconductors will be discussed, in Chapter 4, along with experimental demonstra-
tions. The bilayer heterojunction photoconductor is an important proof-of-concept
device that represents the first step towards the development of vapor-phase chemical
sensors. Chapter 5 focuses on the extension of the heterojunction photoconductor ar-
chitecture to include multiple layers and the consequent changes in device operation.
Chapter 6 details a proof-of-concept bilayer chemosensor platform that specifically
detects TNT vapor. The final chapter will conclude the thesis with a look at some of
the possible extensions of this research.

37
38
Chapter 1 References

[1] P. M Borsenberger and D. S. Weiss. Organic photoreceptors for xerography.

Optical Engineering. Marcel Dekker, New York, 1998.

[2] L. Zhou, A. Wanga, S. Wu, J Sun, S. Park, and T. N Jackson. All-organic active
matrix flexible display. Applied Physics Letters, 88(8):083502, 2006.

[3] P. Peumans, A. Yakimov, and S. R. Forrest. Small molecular weight organic thin-
film photodetectors and solar cells. Journal of Applied Physics, 93(7):3693–3723,
2003.

[4] B. P Rand, J. Genoe, P. Heremans, and J. Poortmans. Solar cells utilizing small
molecular weight organic semiconductors. Progress in Photovoltaics: Research
and Applications, 15(8):659–676, 2007.

[5] H. Chen, M. Josowicz, and J. Janatax. Chemical effects in organic electronics.

Chemistry of Materials, 16(23):4728–4735, 2004.

[6] C. J Cumming, C. Aker, M. Fisher, M. Fox, M. J la Grone, D. Reust, M. G

Rockley, T. M Swager, E. Towers, and V. Williams. Using novel fluorescent
polymers as sensory materials for above-ground sensing of chemical signature
compounds emanating from buried landmines. IEEE Transactions on Geoscience
and Remote Sensing, 39(6):1119–1128, 2001.

[7] S. W Thomas, G. D Joly, and T. M Swager. Chemical sensors based on amplifying

fluorescent conjugated polymers. Chemical Reviews, 107:1339–1386, 2007.

39
 [8] H. M Huang, K. Wang, W. H Tan, D. An, X. H Yang, S. S Huang, Q. Zhai,
L. Zhou, and Y. Jin. Design of a modular-based fluorescent conjugated polymer
for selective sensing. Angewandte Chemie-International Edition, 43(42):5635–
5638, 2004.

[9] S. M. Sze and K. K. Ng. Physics of Semiconductor Devices. John Wiley & Sons,
Hoboken, third edition, 2007.

[10] B. C Sisk and N. S Lewis. Vapor sensing using polymer/carbon black composites
in the percolative conduction regime. Langmuir, 22(18):7928–7935, 2006.

[11] A. J. Mozer, N. S. Sariciftci, L. Lutsen, D. Vanderzande, R. Osterbacka, M. West-

erling, and G. Juska. Charge transport and recombination in bulk heterojunction
solar cells studied by the photoinduced charge extraction in linearly increasing
voltage technique. Applied Physics Letters, 86(11):112104–3, March 2005.

40
Chapter 2

Introduction to Organic
Semiconductor Properties and
Applications

Education is a progressive discovery

of our own ignorance.

WILLIAM J. DURANT
Professor/Writer

Organic nanostructured optoelectronics encompasses the application of carbon-

based or molecular-sized material sets for the development of optical and electrical
devices, supplanting traditional semiconductor materials such as silicon. The key
properties of organic nanostructured materials that sets them apart from bulk in-
organic semiconductors are 1) Strong optical absorption/emission across the visi-
ble wavelengths (λ = 300 - 800 nm), 2) Low melting points, which allows for low-
temperature processing and deposition of thin films over large areas on a wide variety
of substrates, 3) Chemical flexibility, enabling synthesis of a vast library of novel
materials from which to experiment, and 4) Weak intermolecular interactions, which
yield hopping charge transport.
In this chapter, the focus will be on the building block of virtually all organic

41
electronic devices - the organic semiconductor. Organic semiconductors encompass
a wide range of chemical compounds, each with unique material properties. Rather
than attempting to profile and categorize all of the available materials, we will at-
tempt to identify and present the common physical principles underlying organic
semiconductor materials. In addition, we provide the reader with a brief look at the
organic electronics (OE) industry’s present and future.

2.1 Concepts from Quantum Mechanics and Solid-

state Physics

In order to build a common knowledge and language base with the reader, this section
will act as an introduction to the fundamentals of quantum mechanics and solid-state
physics. Many of the concepts presented here will provide the necessary framework
and tools to progress through the later chapters of this work. It is not the goal of
this section to be comprehensive of all topics covered or mathematically rigorous, but
merely to introduce the reader to some of the key concepts that play a role in organic
and nanostructured materials.

2.1.1 Atomic Orbitals: Particle in a Box

Atoms are the basic building blocks of matter, consisting of three types of sub-atomic
particles: the proton, the neutron, and the electron. A basic atomic model consists of
negatively charged electrons orbiting a positively charged nucleus (made of protons
and neutrons). Classical mechanics and electromagnetism cannot explain the stability
of atoms, predicting the eventual collision of the orbiting electron and the oppositely
charged nucleus. Thus, the push for more accurate models of the atom helped to
motivate the initial development of quantum mechanics.
Quantum mechanics is a tool that is used to explain experimental results, and its
ability to accurately predict and explain data regarding atomic properties (i.e. bond
energy, spectrum of emitted light, etc.) drives its widespread acceptance. Using

42
the mathematical formalism developed in quantum mechanics, the state of any atom
at a given time can be described by a complex wave function. This mathematical
abstraction in conjunction with the Heisenberg uncertainty principle enables the con-
ceptualization of an electron “cloud”around each atom’s nucleus, which represents
the probability of finding the electrons somewhere within that region of space but
with the exact positions unknown.

Despite the uncertainty, useful results regarding atomic energy levels can be de-
rived. Consider the situation where a single particle is located within an infinitely
deep potential well. This situation is known as the “particle in a box”, and in physical
terms, the particle is defined as a single point enclosed in a box where it experiences
no forces (i.e. zero potential energy). The impenetrable walls of the box are formed
by potentials that rise to infinity. This simple model can serve as a useful analogy to
an electron orbiting a nucleus: inside the atomic orbital (or the box) the electron is
allowed, outside it is forbidden [Fig. 2-1].

In classical mechanics, the equations of motion governing the particle within the
energy well are trivial: the particle moves in a straight line at a constant velocity until
it is reflected off of the boundaries of the well. In quantum mechanics, when solving
for the electron wavefunction in Schrodinger’s equation and applying infinite potential
boundary conditions, the results are not as intuitive. The particle’s confinement leads
to three quantum results that agree with experimental observation but contrast with
classical mechanics: 1) discrete energy levels 2) a nonzero lowest energy level (zero-
point energy) and 3) spatial nodes at each energy level that imply positions where the
particle can never be found. While, the “particle-in-a-box”approximation accurately
describes these results and the trend of increasing difference in energy levels with
decreasing box dimensions, it does not accurately describe the simplest atom with a
single electron (hydrogen).

A more rigorous analytical model which uses an isotropic coulomb potential is used
in the Hamiltonian that describes the electron wavefunction of a one electron atom [1].
The atomic orbitals can be identified by a unique set of four quantum numbers: the
principal (n), the azimuthal (l), the magnetic (m), and the spin projection (s) [Fig.

43
Figure 2-1: Schematic of 1-D particle in a box. Spatial confinement leads to discrete
energy levels and wavefunctions (Ψ) following the inset equations, where E is energy,
V is potential energy, m is mass, L is the length of the box, and h is Plancks constant.

44
Figure 2-2: Table of 3-D visualizations of hydrogen atomic orbitals with varying
quantum numbers [2].

2-2]. The principal quantum number describes the energy level of a given atomic
orbital, which is also referred to as an “electron shell”. The azimuthal quantum
number describes angular momentum of the orbital, which specifies the shape of the
orbital (s, p, d, f, etc.) and influences chemical bonds and bond angles. The magnetic
quantum number describes the projection of the angular momentum of the orbital
along a specified axis. The spin angular momentum describes the spin of the electron.
Spectroscopic experiments indicate that each orbital can contain no more than
two electrons. In addition, two electrons can never have the exact same quantum
state (i.e. set of quantum numbers) according to Pauli’s exclusion principle. The set
of rules restricting the values of the quantum numbers can be used to explain not
only the experimental spectroscopic results (i.e. Stark and Zeeman effects), but also
the electron configuration of atoms and the periodic table.
The hydrogen atom can be solved analytically, because the wavefunction describes
the probability for one electron and one nucleus. For atoms with two or more elec-
trons, analytical solutions become much more difficult to achieve due to electron-
electron and electron-nucleus interactions, and the governing equations require nu-

45
Figure 2-3: Schematic of energy level splitting that occurs in molecular orbitals of
diatomic Li. σ (in-line) and π (out of plane) bonding orbitals are differentiated by
the way the atomic orbitals overlap, * indicates anti-bonding orbitals [4].

merical simulations to reach approximate solutions.

2.1.2 Molecular Orbital Theory

The optical and electrical properties of covalently bonded organic small molecules
result from the overlap of the atomic orbitals of the constituent atoms. According
to molecular orbital theory [3], linear combinations of atomic orbitals (LCAO) can
be used to construct molecular orbitals (MOs). These MOs can be categorized as
bonding, anti-bonding, or non-bonding orbitals. A bonding orbital is a type of orbital
where the electrons in that state have a higher probability of being between nuclei than
elsewhere, and it will tend to hold nuclei together (hence the name). If the electrons
tend to spend more time elsewhere than between the nuclei, the orbital will function
as an anti-bonding orbital and weaken the bond. Electrons in non-bonding orbitals
are usually in low energy (nearly atomic) orbitals that associate almost entirely with
one nucleus or the other, and thus spend equal time between nuclei or not with no
effect on bond strength [Fig. 2-3].
In diatomic molecules, a lower-energy bonding orbital is created by the in-phase
addition (constructive interference) of two atomic orbitals while a higher-energy anti-
bonding orbital is formed by the out-of-phase addition (destructive interference) of
the two atomic orbitals. The amount of splitting of the energy levels associated with

46
the molecular orbitals is determined by the interaction between the atoms. A definite
energy is associated with each orbital (whether atomic or molecular), and it can be
interpreted as the energy required to remove the electron from the orbital into free
space. In the ground state (unexcited) of an atom or molecule, electrons will settle
into the lowest-energy orbitals available. The available shared electrons then populate
each molecular orbital according to the Pauli exclusion principle (no more than two
electrons in each orbital, and if there are two the spins must be opposite) to yield
a highest occupied molecular orbital (HOMO) and a lowest unoccupied molecular
orbital (LUMO). The energy gap between these two molecular orbitals determines
the transition energies and consequently the absorption and emission spectrum of a
single molecule.

In general, the calculations of energy levels and wavefunctions for a polyatomic

molecule can be a difficult task that is made simpler by the approximation that
electronic transitions occur instantaneously in the time frame of the nuclear system.
This approximation is known as the Born-Oppenheimer (BO) approximation and
can be thought of as heavy nuclei moving more slowly than light electrons. The
success of the BO approximation is due to the high ratio between nuclear and electron
masses. The mathematical consequence of the BO approximation is that it allows the
wavefunction of a molecule to be separated into its electronic and nuclear components:

He ∣ψe (r; R)⟩ = Ee (R) ∣ψe (r; R)⟩ (2.1.1)

̵2 N
h
He = − ∑ ∇2 + Vee + Ven (2.1.2)
2me i=1 i
Hn ∣ψn ⟩ = E ∣ψn ⟩ (2.1.3)
̵ 2 M ∇2j
h
Hn = − ∑ + Ee (R) (2.1.4)
2 j=1 mj

Here r and R are the positions, me and mj are the masses, ψe and ψn are the wave-
functions of the electrons and nuclei respectively. The first term in the electron
Hamiltonian represents the kinetic energy of the electrons, the second term (Vee )
is the electron coulomb potential, and the third term (Ven ) is the electron-nucleus

47
coulomb potential. The first term in the nuclear Hamiltonian represents the kinetic
energy of the nuclei, while the second term represents the electron energy potential.
Thus, Schrodinger’s equation can be solved using the electronic wavefunction inde-
pendent of the nuclear motion, as if the nuclei are fixed in a certain configuration.
Then, Schrodinger’s equation can be solved again for the nuclear wavefunction where
the effective potential is defined by the electronic energy solved previously. Compu-
tational techniques such as the Hartree-Fock self consistent field method and density
functional theory enable the calculation of near exact energies (within 0.05 eV) [5].

2.1.3 Spectroscopic Transitions

Molecular spectra are more complex than atomic spectra, which arises from the com-
plicated structure of the molecule. The spectra of atoms is due only to the electronic
transitions, whereas the spectra of molecules in the gas phase reflect the complex in-
teraction between electronic, vibrational, and rotational transitions. Thus, molecular
spectroscopic data can yield a great deal of information regarding the shape and size
of molecules or the strength and stiffness of bonds.
The energy associated with electronic transitions (∼1 eV) is usually greater than
the energy of vibrational transitions (∼100 meV), which is usually greater than the
energy of rotational transitions (∼10 meV). Thus, while it is possible to observe pure
rotational transitions, a vibrational transition is normally accompanied by rotational
transitions. Likewise, electronic transitions are accompanied by both vibrational and
rotational transitions making them more complicated to interpret. In the solid-state,
molecules do not have the rotational degree of freedom, thus only electronic and
vibrational transitions can be observed in thin films.

Absorption and Emission

The spectrum of atoms arises from permitted transitions between its energy levels,
and the difference in energy between those states is transferred to a photon of energy
hν. Not all transitions between states are allowed, and the rules that specify the

48
permitted transitions are based on the transition dipole moment between the two
states. The transition dipole moment between states ∣i⟩ and ∣f ⟩ can be defined as

µif = ⟨i∣ µ̂ ∣f ⟩ (2.1.5)

where µ̂ = −er̂ is the electric dipole operator. The transition dipole moment can be
thought of as a measure of the degree to which the electron perturbs the electromag-
netic field when it makes a transition between the states. The wavelength of light is
related to its energy through de Broglie’s relation, which codifies the wave-particle
duality of matter:
hλ
E= (2.1.6)
c
where h is Planck’s constant, λ is wavelength, and c is the speed of light. A simple
rule of thumb for converting wavelength to energy is to use the relation E = 1254
nm⋅eV/λ.

Macroscopically, when light illuminates a piece of semiconductor (organic/inorganic)

a number of different events can occur. A photon having little mass can reflect (spec-
ular and diffuse) off the surface of the material, transmit through the material, or
get absorbed by the material. Different wavelengths of light will experience differing
amounts of absorption, since the energy of the incoming photon must be similar to an
allowed electronic transition. This results in the characteristic absorption spectrum
of the material.

The absorption coefficient (α) is one way to describe the amount of light (at a given
wavelength) absorbed by a material, and the coefficient is employed in the empirical
relation known as Beer’s Law: I = Io e−αL . This relationship can be used to quantify
the absorption of light by a material as long as the following conditions apply: 1) the
material has homogeneously distributed and independent absorption centers that do
not scatter light, 2) the light is at normal incidence, and 3) the light is of low enough
intensity that it does not optically saturate or pump the material to cause stimulated
emission. The absorption of a photon results in an excited state of matter, where
the energy of the photon transfers to the material as an electric potential energy.

49
There are several mechanisms for the energy to relax back to its ground state which
can be categorized as radiative or non-radiative pathways: 1) emission of a photon
(fluorescence), 2) dissipation into vibrational modes in the material (i.e. heat), and 3)
transition to a state with a different spin angular momentum (intersystem crossing)
before relaxing from that triplet state.
The fluorescence spectrum of emitted photons is always shifted towards lower
energies (red-shifted) relative to the absorption spectrum; the difference in energy
is known as the Franck-Condon shift. The origin of the shift can be explained by
considering the time evolution of a molecule following the absorption of a photon.
Invoking the BO approximation, the electronic transition to an excited state will occur
within a fixed nuclear framework. At this point the nuclear wavefunction evolves into
its final state to accommodate the excited phonon mode. This occurs on the time scale
of a nuclear oscillation (∼10-13 s) and results in a new nuclear configuration. Internal
conversion happens on the same time scale to allow the excited state to couple to
lower vibrational modes, lowering the energy. Finally, the electron relaxes back to
its ground state, which emits a photon and triggers another nuclear reconfiguration
to accommodate the electronic ground state in addition to internal conversion to the
lowest vibrational mode [Fig. 2-4].

2.1.4 Excitons

In general terms, an exciton can be defined as an excited state consisting of a bound

electron-hole (hole = absence of electron) pair. Exciton formation occurs when a
photon interacts with a semiconductor to excite an electron into a higher energy state.
The promotion of the electron leaves a positively charged environment behind, which
can be treated as a quasi-particle known as a hole. As a result of the coulombic
attraction between the electron and its hole, there is a slight reduction in energy
(with respect to free carriers) resulting in a bound excited state. It is useful to treat
the bound excited state as a neutral quasiparticle, which can either diffuse through
the molecular crystal, transfer its energy to another exciton, recombine to produce
luminescence, or dissociate to form free carriers.

50
Figure 2-4: Illustration of the Franck-Condon shift observed in electronic transitions.
Vgnucl and Venucl are the nuclear potential energy functions of the ground and excited
electronic states respectively. ∆EgF C and ∆EeF C are the Franck-Condon relaxation
energies of the ground state and excited state respectively. E↑el and E↓el are the
dominant electronic excitation and relaxation transitions respectively [6].

51
 Figure 2-5: Illustration of the different types of excitons. After [7].

There are three types of excitons: 1) Frenkel excitons, 2) Wannier-Mott excitons,

and 3) charge transfer excitons. In organic materials the dielectric constant tends
to be small, which allows a strong coulomb interaction to occur between an electron
and its hole. This type of exciton is known as a Frenkel exciton, where the binding
energy is on the order of 1 eV and large enough to restrict the electron and hole
to the same molecule. A Wannier-Mott exciton has a radius that is much larger
than the lattice spacing. This type of exciton is more commonly found in covalently
bonded crystals that have large dielectric constants that tend to screen the Coulombic
interaction between electrons and holes. Typically, the binding energy of a Wannier-
Mott exciton is on the order of 0.1 eV. An intermediate case occurs when the electron
and hole occupy adjacent molecules. This situation is known as a charge-transfer
exciton and is the precursor to the generation of free carriers [Fig. 2-5].
Aside from the spatial differences in the types of excitons, the total spin angular
momentum of exciton states can be used to classify them as either singlet or triplet
excitons. As previously noted, any given molecular orbital can only contain at most
two electrons. According to Pauli’s exclusion principle the spins of the electrons in
the same orbital cannot be the same. Therefore, one electron must be “spin up”(s =
+1/2) while the other is “spin down”(s = -1/2). Excitons do not share this restriction,
consisting of an electron in one orbital and a hole in a lower orbital. This leads to the
slightly more complex situation where there are four possible spin combinations of
the electron-hole pair. Constructing the set of spin wavefunctions yields one exciton
state with a total spin of 0 (singlet) and three degenerate exciton states with a total

52
Figure 2-6: A typical Jablonski diagram that illustrates many of the possible elec-
tronic transitions associated with a molecule [8].

spin of 1 (triplet). The distinction between singlets and triplets will prove important
in electronic transitions, because angular momentum must be conserved. Therefore,
a triplet excited state is forbidden from relaxing to a singlet ground state unless there
is a process available that can change the total spin of the system from 1 to 0 (spin
orbit coupling) [Fig. 2-6].

2.2 Organic Thin Film Properties

Organic semiconductor films are generally assemblies of molecule or polymer chains.

Regardless of the orientation of the molecules in a thin film, organic materials (crys-
talline or amorphous) are held together by weak intermolecular forces called Van der
Waals forces [9]. These Van der Waals forces preclude covalent and ionic bonding and
consists of repulsive and attractive components. The repulsive component prevents
the collapse of molecules as entities are placed closer together. The attractive com-
ponent itself has three distinct contributions: 1) the electrostatic interaction between
charges and permanent multipoles 2) the polarization effect, which is the interaction

53
 Molecular crystals Covalent crystals
Lattice interaction Van der Waals (∼ 10−3 -10−1 eV) Covalent (∼ 2 - 4 eV)
Carrier mobility Low (<1 cm2 V−1 s−1 ) High
Mean free path Short (order of lattice constant) Long (102 - 103 times lattice constant)
Carrier effective mass Large (102 - 103 times electron mass) Small (less than electron mass)
Charge transport type Hopping Band
Exciton type Frenkel excitons (0.3 - 1.0 eV binding) Wannier excitons (0.01 - 0.1 eV binding)
Melting point Low High
Mechanical strength Low High
Compressibility High Low

Table 2.1: Comparison of properties of molecular and covalent crystals. After [10].

between a permanent mulitpole on one molecule and an induced multipole on another,

and 3) the transient (fluctuating) multipole effect which is temporarily induced by
the probabilistic movement of the electron density cloud.

The weak intermolecular forces result in narrow electronic bandwidths and lo-
calized excitations, restricting conduction to hopping transport. Organic semicon-
ductors are often highly disordered (both spatially and energetically), hence charge
carrier mobility is smaller relative to their covalently-bonded counterparts. Due to
poor electrical conductivity, organic materials are employed in applications that ex-
ploit their strong optical properties or their potential for large-area fabrication. A
comparison of some of the major differences between molecular crystals and covalent
crystals is provided in the table above.

2.3 Overview of Organic Electronics Industry

Organic electronics (OE) covers a wide array of consumer electronics and photonics
products enabled by the use of organic materials. These materials may be insulating
or conducting and light emitting or absorbing. A number of reports state that the
organic electronics industry, which reported less than $400M in revenue in 2007,
would likely grow to $15B in 2015 [11, 12]. The key drivers of this explosive growth
rate (CAGR >50%) will be the adoption of more cost effective processing techniques
(primarily printing) and the shift in focus on applications that utilize simple, large-
area electronics rather than the more impressive display technologies.

54
2.3.1 Manufacturing Processes: Printing is the Future

Although OE products are created using a variety of techniques including evapora-

tion and thermal transfer, there is considerable interest in using solution processing
because of its potential as a cost effective manufacturing approach. Printing is fun-
damental to OE technology’s primary attraction - the ability to produce low-cost
electronics on large-area, flexible substrates [13]. Roll-to-roll capital equipment costs
are relatively small, with virtually no masking requirements, and throughputs are
suitable for many applications [14]. The promise of cost savings in printed OE is
so great that several manufacturers are attempting to turn their existing organic
materials into “inks”for compatibility with future printing equipment.

Almost every known printing technique has been used to print electronics, but
the three most important are direct contact printing (including screen printing [15],
flexography [16], and gravure [17]), ink-jet printing [18, 19], and organic vapor jet
printing [20].

Screen printing has been used for decades to make EMI shielding, capacitors,
membrane switches and several other applications, and it has been used with a variety
of conductive polymers, such as PPV:PCBM (poly [p-phenylene vinylene] and P3HT
(poly-3-hexylthiophene). In the traditional screen printing process a fine mesh screen
(made of steel, nylon, or polyester) is suspended above a patterned stencil resting on
the substrate, with the resolution of the print determined by the mesh size. A flood
bar is used to fill the mesh with ink, after which a squeegee is used to push the mesh
in contact with the substrate, transferring the ink where the stencil is not blocking
[Fig. 2-7(a)]. The thickness of the print is a controllable parameter, and the use of
viscous inks enables thicker films than can be achieved with other printing techniques.

Rotary screen presses are designed for continuous, high-speed screen printing.
Seamless metal mesh cylinders act as the screens. The open-ended cylinders are filled
with ink from one end of the cylinder so that a fresh supply is constantly maintained.
The squeegee is a free floating steel bar inside the cylinder and the squeegee pressure
is maintained and adjusted by magnets mounted under the substrate [Fig. 2-7(b)].

55
Figure 2-7: Schematic illustrations of a) flatbed screen printing and b) rotary screen
printing [21].

56
Rotary screen presses are most often used for printing textiles, wallpaper, and other
products requiring unbroken continuous patterns [21].

Flexography is the approach most often cited as the way to ramp up to high volume
printed electronics and is a printing technique where the printing plate is flexible
[Fig. 2-8]. The patterned stamp is inked with the material to be deposited and then
brought into contact with the substrate to transfer the ink. In general, the lower print
quality and resolution relegates flexography to applications with large tolerances,
currently being used to print RFID antennas on labels using silver inks, batteries,
organic circuits, photovoltaics, and low-cost lighting. Soft lithography is a related
laboratory technique where patterned PDMS is used as the flexible stamp. While
it has been suggested that soft lithography is compatible with roll-to-roll printing
processes [22], both flexography and soft lithography suffer from stamp distortions
preventing accurate pattern registration over large areas.

Gravure printing is a high-throughput process limited to print runs in excess

of a million copies due to the large fixed costs involved. This technique is based
on engraving a metal cylinder with a pattern of cells of variable size and depth.
A metal blade is used to wipe excess ink off of non-printing areas while leaving
ink within the cells (a.k.a. doctor blading). When a flexible substrate is brought
into contact with the engraved cylinder the ink transfers and is able to achieve low
film thicknesses due to the use of lower viscosity inks [Fig. 2-9]. There are many
factors influencing gravure print quality including substrate properties (smoothness,
compressibility, porosity, ink receptivity, wettability, etc.) and ink properties (ink
chemistry, viscosity, drying, etc.). Furthermore, process parameters, such as doctor
blade angle and doctor blade pressure, impression pressure, and printing speed have a
tremendous effect on the quality of the printed ink film. Despite the large parameter
space, the robustness of the metal cylinder carrying the image recommends the process
for large-scale manufacturing lines. The engraved cylinder will have good printing
stability over time, whereas the chlorinated solvents used in flexography tends to
degrade the photopolymer print plates.

Inkjet printing has already been widely used with organic materials and can print

57
Figure 2-8: Using a laboratory printing machine, printed electronics, such as RFID
tags, are being developed by PolyIC. By using a roll-to-roll flexography printing
process it is possible to produce large volumes at very low cost. Source: PolyIC press
picture

58
Figure 2-9: (a) Diagram of rotogravure printing process (b) Multi-color rotogravure
printing machine. Source: R.H. Technocrates Center

layers with thin ink film, which not only saves on material costs, but is often critical
for large-area applications such as displays, lighting panels and sensor arrays. A
traditional drawback of inkjet printing is its lower throughput capability relative to
gravure or flexography. The two main types of inkjet printing technologies are thermal
and piezoelectric. In thermal inkjet printing droplets are generated by heating the
wall of the ink chamber causing formation of vapor bubbles and ejection of droplets
through the nozzle [Fig. 2-10]. In piezoelectric printing a pressure wave (driven by an
applied voltage waveform) in the ink chamber results in formation of droplets at the
nozzle [Fig. 2-11]. Both types of inkjet printing are inexpensive to implement, reliable,
and compatible with “drop on demand” (DoD) applications. One major difference
in the technologies is that drop size in thermal inkjet printing is determined by the
chamber size, while the drop size in piezoelectric inkjet printing can be changed by
applying different voltage waveforms. The other key difference is that piezoelectric
inkjet printing enables the use of heat-sensitive inks as it does not rely on heating to
generate droplets.

A simple extension of inkjet printing enables an entirely new printing process,

with the potential for high throughput. Kateva, a start-up company based in CA, is
commercializing molecular jet printing [24, 25] technology by developing print heads

59
 Figure 2-10: Schematic of thermal inkjet printing process [23].

Figure 2-11: (a) Cross-sectional view of piezoelectric ink chamber (b) Dimatix’s
benchtop DMP 2800 piezoelectric inkjet printer. Source: Dimatix

60
Figure 2-12: Cross sectional schematic view of the molecular jet print head and the
steps involved in deposition [23].

that can be retrofitted onto existing inkjet printing capital equipment lines used for
display technologies. The thermal inkjet nozzles load porous membranes with solvated
organic material, after which heating is applied to eject the evaporated material
onto a nearby substrate [Fig. 2-12]. High throughput is achieved by arraying the
nozzles for simultaneous printing. This deposition technique should prove useful for
small-molecule organic materials, whose electronic and optical properties and film
uniformity are significantly degraded when undergoing solvent processing.

2.3.2 Organic Electronic Products: OLEDs Light the Way

for Adoption

New applications for OE are constantly being invented. While these new applica-
tions are almost always intellectually interesting, few are likely to become revenue
generating products within the next few years. One example is organic memory for
which various formulations and architectures have been available for several years.
While many important firms have patents on various organic memory products, it
seems unlikely that this kind of memory will find its way into a commercial product

61
Figure 2-13: (a) Revenue projections for OE markets ($ Billions) (b) Revenue pro-
jections for OE by application, 2014 ($ Millions). Source: Nanomarkets LC (2007)
[26]

until organic radio frequency identification (RFID) tags and other disposable prod-
ucts start to take on a much higher degree of sophistication. OLED technology is
the biggest consumer of organic semiconductor materials and the largest source of
revenue in the OE industry, accounting for $308.5M in 2008 and potential to reach
$6B by 2015 [11].

Organic Light Emitting Diode (OLED) Displays

The most commercially significant market opportunity today is OLED (Organic

Light-Emitting Diode) technology [Fig. 2-14]. OLED displays have the following
advantages over the current industry leader liquid crystal display (LCD) technology:
1) emissive technology with no backlight requirement 2) better color saturation 3)
better contrast ratio 4) smaller power requirements. Only in the last several years
have OLED displays started to generate significant revenue, with the market pro-
jected to grow to $4.5B in 2010 [27]. Today 25 to 30 firms produce OLED displays,

62
 Figure 2-14: Prototype 40” OLED display (worlds largest). Source: Samsung

primarily using small-molecule fluorescent and phosphorescent materials. The key

players in this space are Samsung, LG Electronics, Univision, Sony, Epson-Seiko, Pi-
oneer, and TDK. A few firms are actively producing polymer OLED displays, with
OSRAM and Delta Optoelectronics actively shipping products.

OLED Lighting

While displays dominate the present OE market, OLED lighting technology [Fig. 2-
15] is expected to account for an increasing share and is getting attention from a select
group of OLED makers. Lighting places special demands on materials with regard to
lifetimes, color, and power consumption. For example, the brightness of the OLED
lamp is inversely proportional to its lifetime, while the most power-efficient lamp may
not provide the most aesthetically pleasing light. Color stability is a major issue for
OLED lighting that is exacerbated by differential aging of the red, green, and blue
materials used to create white light. Despite these challenges, OLED lighting’s high
power efficiency and thin, flexible form factor is driving adoption among architects
and designers as viable substitutes for more traditional lighting such as incandescent,
fluorescent, and high intensity discharge (HID) lamps. In the future, the OLED

63
Figure 2-15: White, small molecule OLED measuring 4 x 12 cm2 . Source: Philips
Lighting

lighting industry is projected to reach $6B in 2018 [28] with Philips, GE, Konica
Minolta, Lumiotec and OSRAM entering mass production in 2011.
Over the next several years, the markets for radio-frequency identification (RFID)
tags and display backplanes will drive the production of OTFTs. OTFTs have steadily
achieved market traction as device performance matches or exceeds that of some
silicon TFTs and as significant capital investments have been made in production
facilities. The market for organic RFID tags is presently quite small reaching sales of
$13.7M in 2007 and forecasted revenues of $11.6B by 2015 [11]. PolyIC is one of the
key players in this market.
OTFT backplanes are finding their way into electronic ink (e-ink) displays, which
helped them to garner $2.1M in 2008 with projections of $3.3B of revenue in 2015
[29]. Plastic Logic is one such startup firm that manufactures flexible e-ink displays
by using an OTFT backplane. Polyera is another startup that has developed an n-
type polymer material with high mobility (∼0.5 cm2 V−1 s−1 ) and good solubility. The
company aims to ship flexible displays based on complementary OTFT backplanes
within 3 years.
It is now being widely suggested that these types of simple electronics could help

64
 Figure 2-16: Organic rectifier printed on 6” PEN-foil. Source: IMEC

jump start the entire OE industry. Thus, some of the firms that ultimately aim at
offering multifunctional organic RFIDs will first produce low function tags that can be
used for simple point-of-sale or marketing applications. The focus now is on pushing
products out into the marketplace, even if they do not dazzle the customer in the
way that a printed, flexible OLED display complete with printed organic backplane
might.

Organic Photovoltaics (OPVs)

With the recent push for renewable energy, there has been a widespread prolifer-
ation of photovoltaic companies and technologies. The lower power efficiencies of
organic photovoltaics (OPV) has kept the market lagging behind its inorganic coun-
terpart. However, OPV’s potential will lie in its high throughput production ca-
pabilities, which will meet increasing demand for solar over the next few decades
as inorganic materials suppliers reach production or material resource limitations.
Polymers, small-molecule organics, hybrid organic/inorganic, and nanoparticle-based
inks have all been used as active materials to create commercially viable photovoltaic
devices with power efficiencies ranging from 5%-11% [30]. The dollars/watt metric

65
Figure 2-17: Demonstration of flexible organic photovoltaic technology. Source:
Konarka

must reach $0.50/watt to be competitive with conventional energy generation meth-

ods. This threshold may eventually be surpassed with better materials, new cell
designs, and solution-processed manufacturing. Currently, the market is a miniscule
$500K (2008) but will reach $223M in 2015 [31]. Some interesting developments in
the consumer electronics market are already beginning to emerge. G24i and Konarka
will release OPVs targeting the cell phone market, while SWATCH plans to have
dye-sensitized solar cell watches in 2009.

2.4 Summary
As nanotechnology enters our communal lexicon and gets tossed around as a market-
ing buzzword, it’s clear that research on the atomic scale has arrived. The formation
of an international industry association specific to the organic electronics industry
(OE-A) is just another sign that the market for consumer electronics utilizing organic
materials is maturing rapidly. Just as the OE industry is maturing and organic de-
vice performance is beginning to rival that of its inorganic counterparts, the vision of
low-cost, large-area electronics is steadily becoming a reality. With organic materials
enabling a wide array of novel technologies, the future of the industry will rely largely
on our ability to imagine new, game-changing applications.

66
Figure 2-18: Roadmap for organic electronics applications. Source: OE-A [32].

67
68
Chapter 2 References

[1] D. J. Griffiths. Introduction to Quantum Mechanics. Prentice Hall, Upper Saddle

River, NJ, 1995.

[2] S. Tokita, T. Sugiyama, F. Noguchi, H. Fujii, and Hidehiko Kobayashi. An

attempt to construct an isosurface having symmetry elements. The Journal of
Computer Chemistry Japan, 5(3), 2006.

[3] R. S. Mulliken. Spectroscopy, molecular orbitals, and chemical bonding. Nobel

Lecture, December 1966.

[4] F. Buth. Organic electronics. Walther Schottky Institut, TU Munchen,

2008. http://www14.in.tum.de/konferenzen/Jass08/courses/4/buth/
slides buth.pdf.

[5] P. W Atkins and R. S Friedman. Molecular Quantum Mechanics. Oxford Uni-

versity Press, New York, 4th ed edition, 2005.

[6] C. F. Madigan. Theory and simulation of amorphous organic electronic devices.

Ph.D. thesis, MIT, 2006.

[7] J. D. Wright. Molecular Crystals. Cambridge University Press, Cambridge; New

York, 1995.

[8] Olympus microscopy resource center. http://www.olympusmicro.com.

[9] V. A. Parsegian. Van der Waals forces. Cambridge University Press, Cambridge,
2006.

69
[10] E. A. Silinsh and C. Vladislav. Organic Molecular Crystals: Interaction, Local-
ization and Transport Phenomena. American Institute of Physics, New York,
1994.

[11] Opportunities in organic electronic markets. White paper, Nanomarkets, March

2008. http://www.printedelectronicsnow.com/whitepapers/download/7/
NMOEMaterials 032408.pdf.

[12] P. Harrop. Organic electronics forecasts, players and op-

portunities. White paper, IDTechEx Ltd., 2005.
http://www.idtechex.com/research/reports/organic electronics
forecasts players and opportunities 000117.asp.

[13] R. Parashkov, E. Becker, T. Riedl, H.-H. Johannes, and W. Kowalsky. Large area
electronics using printing methods. Proceedings of the IEEE, 93(7):1321–1329,
2005.

[14] H. Klauk. Organic Electronics. Wiley-VCH, Weinheim, 2006.

[15] Z. Bao, Y. Feng, A. Dodabalapur, V. R. Raju, and A. J. Lovinger. High-

Performance plastic transistors fabricated by printing techniques. Chemistry
of Materials, 9(6):1299–1301, June 1997.

[16] J. Kenny. Printing electronics: Flexo and screen. Printed Electronics, February
2009.

[17] E. Hrehorova and R. Kattumenu. Evaluation of gravure print forms for printed
electronics. Gravurezine, March 2007.

[18] T. R. Hebner, C. C. Wu, D. Marcy, M. H. Lu, and J. C. Sturm. Ink-jet printing

of doped polymers for organic light emitting devices. Applied Physics Letters,
72(5):519–521, February 1998.

[19] S. Chang, J. L., J. Bharathan, Y. Yang, J. Onohara, and J. Kido. Multicolor

organic Light-Emitting diodes processed by hybrid inkjet printing. Advanced
Materials, 11(9):734–737, 1999.

70
[20] M. S. Arnold, G. J. McGraw, S. R. Forrest, and R. R. Lunt. Direct vapor jet
printing of three color segment organic light emitting devices for white light
illumination. Applied Physics Letters, 92(5):053301–3, February 2008.

[21] Print process descriptions: Printing industry overview: Screen print-

ing. Technical report, Printers’ National Environmental Assistance Center.
http://www.pneac.org/printprocesses/screen/.

[22] P. Cosseddu, E. Orgiu, and A. Bonfiglio. Soft lithography fabrication of fully

flexible and transparent all organic FETs for large area applications. In Organic
Electronics – Materials, Devices and Applications, volume 965, Boston, 2007.

[23] J. Chen. Novel Patterning Techniques for Manufacturing Organic and Nanos-
tructured Electronics. Ph.D thesis, MIT, June 2007.

[24] M. Shtein, P. Peumans, J. B. Benziger, and S. R. Forrest. Direct mask-free

patterning of molecular organic semiconductors using organic vapor jet printing.
Journal of Applied Physics, 96(8):4500–4507, October 2004.

[25] J. Chen, V. Leblanc, S. H. Kang, P. J. Benning, D. Schut, M. A. Baldo, M. A.

Schmidt, and V. Bulović. High definition digital fabrication of active organic
devices by molecular jet printing. Advanced Functional Materials, 17(15):2722–
2727, 2007.

[26] Organic electronics: A market & technology assessment. White paper, Nanomar-
kets, 2007.

[27] Organic light emitting diode (OLED) displays: A global strategic

business report. Technical report, Global Industry Analysts, 2008.
http://www.laserfocusworld.com/display article/322472/12/none/none/
INDUS/OLED-display-market-to-top-$4.5-billion-by-2010.

[28] OLED lighting in 2009 and beyond: The bright future. Technical report, Dis-
playSearch, 2009. http://www.displaysearch.com/.

71
[29] Next chips: Organic transistors and memories and the applications they will
enable. White paper, Nanomarkets, 2008.

[30] Organic photovoltaic markets. White paper, Nanomarkets, 2008.

[31] Materials market for thin-film and organic photovoltaics. White paper, Nanomar-
kets, 2008.

[32] Organic electronics. White paper, Organic Electronics Association,

2007. http://www.vdma.org/wps/wcm/resources/file/eb697e48f1969d3/
OE-A Brochure%202007 5MB 72dpi.pdf.

72
Chapter 3

Vapor-phase Chemical Detection

Smell is a potent wizard that

transports you across thousands of
miles and all the years you have
lived.

HELEN KELLER
Author/Educator

Before we survey the wide range of technologies available for chemical sensing, a
distinction must be made between chemical sensors and electronic noses (e-noses).
Most chemosensing applications favor one of two functionalities when sensing ana-
lytes: detection or distinction. For detection, the information regarding a specific
chemical’s presence in the environment is of utmost importance. For distinction,
the relevant information is the identification of the chemical constituents that make
up the odor. Typically, individual chemical sensors are implemented for chemical
detection, while an array of such devices coupled with signal processing algorithms
comprise an e-nose capable of distinction. Clearly, both types of devices will share
some of the same technological challenges: reliability, portability, and sensitivity.
Chemical sensors used for detection are designed to be specific, with the ability to
exclusively respond to one particular analyte. However, e-noses use arrays of sensors
that are (ideally) orthogonal in odor space while also covering a wide range in odor
space. In odor space, one can imagine molecules being spatially located based on

73
their similarity to other odors, as measured by human noses. In other words, two
odors judged to have similar odor quality are positioned closer together in the mul-
tidimensional odor space, than two different smelling odors. The dimensions of odor
space are related to the physical and chemical attributes of the molecules themselves:
molecular weight, number of double bonds, number of carbon atoms, etc. Thus, the
big challenge for e-noses is to make an array that incorporates enough sensors to
cover the application-specific odor space while maintaining orthogonality among the
various sensors to minimize the complexity of the data processing algorithms used.
Another issue for sensor arrays is the variation in the rate of degradation between
different sensors, necessitating complex calibration routines [1]. Algorithms and sig-
nal processing techniques are again the main approaches to solving this issue and
can be difficult to implement depending on the sensor technology. Another challenge
for electronic noses is the creation of an accurate odor space itself. While there are
numerous statistical odor classification schemes that weight each physicochemical at-
tribute differently, to date there is no general relationship that can accurately predict
the odor quality of molecules from their structure alone [2].
Because electronic noses are composed of an array of chemical sensors, the per-
formance of the sensor array is limited by the properties (sensitivity, selectivity,
reproducibility, reversibility, etc.) of the individual sensors. Therefore, this chap-
ter will focus on the individual chemical sensing approaches and their transduction
mechanisms, rather than the aggregate performance of the array coupled with signal
processing techniques.

3.1 Survey of State-of-the-art Vapor-phase Solid-

state Chemosensing Organic Devices

The basic function of chemical sensors is to transduce a chemical signal into a mea-
sured entity. Within this broad class of devices are vapor-phase chemosensors, which
convert the presence of specific volatile organic molecules in the gas phase into a

74
measurable signal. Solid-state organic devices have the additional requirement of
using carbon-based materials that are deposited as thin films. Despite these con-
straints, solid-state vapor-phase organic chemosensors encompass a diverse array of
materials and device architectures. One simple sensor classification scheme uses the
method of transduction to differentiate the technologies: electrical, optical, calorimet-
ric (thermal) or gravimetric (mass). The electrical group can be further subdivided
by device structure into chemiresistors, chemicapacitors, and organic thin-film tran-
sistors (OTFT). The optical group consists of chemoresponsive dyes, optical fiber
sensors, surface plasmon resonance (SPR) sensors, and luminescent sensing schemes.
The gravimetric group consists of surface acoustic wave (SAW) sensors, quartz crys-
tal microbalances (QCM), and microcantilever sensors [Fig. 3-1]. This survey will be
limited in scope to cover the most recent advances in solid-state vapor-phase organic
chemosensors, and as such, will not be a comprehensive survey of all existing vapor-
phase chemosensors. The scope of this chapter restricts the discussion to research
performed within the last seven years and will reference older research, when neces-
sary, to illustrate fundamental concepts. The focus of this survey also requires the
organic (small-molecule, polymer, or carbon nanotube) materials to play a key role
in the transduction mechanism by utilizing their electrical, optical, piezoelectric, or
thermoelectric properties, rather than functionalizing an underlying inorganic sensor.
Thus, the reader is encouraged to consult other sources to learn more about SAW
[3], QCM [4, 5], and microcantilever [6, 7] sensing schemes that incorporate organic
materials as functional groups to enhance selectivity.

3.1.1 Electrical Odor Sensors

These chemosensors represent the simplest of gas sensors in that their chemical reac-
tivity is directly transduced into an electrical signal. Changes in resistance, capaci-
tance, voltage, or current indicate the presence of a particular analyte. These types
of sensors are widely used for gas and odor measurements in commercial electronic
nose sensor arrays.

75
Figure 3-1: Classification tree of different chemical sensor technologies. Lighter boxes
will not be covered in this chapter.

Chemiresistor

Chemiresistors represent the most common and simple type of sensor that incorpo-
rate organic materials. Generally speaking, the chemiresistor consists of a pair of
electrodes that form contacts to the chemosensing material, which is deposited onto
an insulating mechanical support [Fig. 3-2].
Normal operation involves applying a constant current through the sensing ma-
terial while measuring the resulting potential difference at the electrodes. Thus, a
change in resistivity would result in a change in the potential difference measured at
the contacts. Inherently, this class of sensors requires less overhead electronics than
more complex transduction schemes, which makes them suitable for miniaturization
and portable applications. Some weaknesses include low signal to noise ratios and the
variability in response repeatability to the same analyte, making them less useful when
analyte concentrations need to be quantified. In addition, scalability is a major factor
limiting the practical application of these sensors because bulk resistance changes are
smaller with decreasing device size. Within this class of sensors, there are several

76
Figure 3-2: (a) Schematic of a typical chemiresistor device using interdigitated elec-
trodes. There are two steady state conditions during operation: 1) no analyte present
in the sampling environment, and 2) analyte present in the sampling environment re-
sulting in a change in resistivity of the film. (b) Three-dimensional pattern depicting
representative carbon black chemiresistor responses to seven test analytes [8].

classes of materials that are used for the sensing layer: organic conjugated polymers
and polymer composites, carbon-black/organic composites, phthalocyanines, and car-
bon nanotubes. All of these materials exhibit an inherent change in conductivity due
to the interaction with gases and can be used in different odor sensor schemes. For
clarity, they are introduced here in the simplest sensor architecture.

Organic Conjugated Polymers and Polymer Composites

The alternating single and double bonds in organic conjugated polymers (CPs) allow
for the delocalization of carriers along the polymer backbone, which gives the polymer
the ability to conduct electricity. When this ability is coupled with the introduction
of excess charges through doping, the polymer can then exhibit conductance from
that of an insulator to that approaching a metallic conductor [9].
Recently organic conducting polymers have become the focus of much of the ma-
terials research in chemosensing devices. Synthetic flexibility allows the chemical and
physical properties of polymers to be tailored over a broad range of values for any

77
given application. In addition, polymers exhibit tunable specificity to volatile organic
compounds, which makes them ideal candidates for replacing canonical sensor materi-
als such as metal oxide semiconductors. In a conducting polymer film, when a gaseous
analyte is introduced it acts as either an electron donor or an electron acceptor. Most
organic conducting polymers are p-type materials with holes as the majority charge
carrier, determining the thin film conductance. If the conducting polymer donates
an electron to the gas phase analyte, then its hole population and, consequently, its
conductivity will increase. Alternatively, if the same conducting polymer receives an
electron from the gas, then the hole population will decrease, reducing conductivity.
Secondary effects include the slow diffusion of the gas into the conducting polymer
itself, which can cause conformational changes in the polymer backbone [10]. Analyte
gas diffusion generally results in a slow sensor response with more signal hysteresis.

Current trends in organic conducting polymer chemiresistor research involve the

fabrication of nanostructures such as nanowires, nanotubes, or nanofibers made out of
conducting polymers. The nanostructured conducting polymers have demonstrated
improved sensor performance by increasing the exposed surface area as well as re-
ducing the diffusion depth required for the penetration of gas molecules (due to high
surface area to volume ratios) [11–14]. In addition, research into the incorporation of
metal oxide nanostructures into the polymer films or incorporation of other polymers
has yielded improvements in sensitivity, aging, and the ability to detect non-polar
analytes [15–17]. The main advantages to using organic conducting polymer chemire-
sistors are their ease of fabrication, where polymers can be deposited under room
temperature conditions. In addition, their compatibility with microfabricated struc-
tures makes them particularly suitable for low-cost high throughput manufacturing.
Another advantage organic conducting polymer chemiresistors have is their ability
to operate at room temperature. Metal oxide semiconductor chemiresistors are typ-
ically heated to between 200 ○ C and 500 ○ C for normal operation, which dissipates
much more power. These polymers also typically have a sensor response time that
is proportional to analyte concentration and the rapid adsorption/desorption kinet-
ics as determined by van-der Waals bonding at the surface, allowing the sensor to

78
regenerate quickly. At the lowest analyte concentrations, the 10-100 second diffusion-
limited response times [18] remain comparable to existing metal-oxide sensors that
have response times in the range of a few seconds [19], depending on factors such
as operating temperature and analyte concentration. Other weaknesses of organic
conducting polymer chemiresistors are their variability from sensor to sensor and
sensitivity to moisture.

Carbon-black Polymer Composite

Sensors using carbon-black polymer composites utilize a polymeric insulating matrix

to separate and fix conductive carbon-black nanoparticles. The relative concentra-
tions of the carbon-black particles to the polymer and the form factor of the active
sensor determine the conductivity of the composite and its signal to noise ratio [1].
With a high enough concentration of carbon black, the composite exhibits a simple,
linear response to analyte concentration. Alternately, low concentrations of carbon
black enable the composite to operate near the percolation threshold, yielding im-
provements in sensitivity with higher signal to noise ratios, at the cost of having less
reproducibility, less stability, and a more complex, non-linear response to analyte
concentration [20].
The carbon black composite is usually spray-deposited onto interdigitated elec-
trodes, where vapors that are introduced to the composite interact with the poly-
mer and cause it to swell, thereby inducing a conformational change in the distance
between neighboring carbon-black particles. Swelling results in a decrease in film
conductivity. The sensitivity of sorption-based sensors is based on the interactions
between the gaseous analyte and the polymer. By incorporating different functional
groups into the polymer to target specific functional groups on the analyte, the se-
lectivity of the sensor can be tuned to specific volatile organic compounds [21]. The
composite’s sensitivity is determined by the amount of vapor that the sensor can
absorb. This interaction between gaseous analytes and the polymer matrix depends
strongly on the density of functional groups within the matrix available to interact
with the analyte.

79
 Recent work being done in this class of sensors has incorporated non-polymeric
materials to bind the carbon black particles, such as monomeric, low vapor pressure
organic molecules with functional groups attached [8]. The benefit of using smaller
organic molecules in the matrix is an increased density of functional groups present in
the composite, producing an increase in sensitivity. Also, the random orientation of
the organic molecule matrix should create a more permeable structure for enhanced
vapor diffusion. Finally, the use of non-polymeric materials allows for new synthetic
techniques to add chemical functionality and physical properties to the sensors that
are not readily accessible in polymeric materials. Interestingly, these non-polymeric
carbon-black composite sensors also exhibit some degree of selectivity to various an-
alytes, making them practical for array-based sensing.
In general, carbon-black composites demonstrate fast response times, good re-
versibility, reproducibility, and stability. However, they lack the ability to react ex-
clusively to many different gaseous analytes, making them better suited for a sensor
array application where pattern recognition algorithms can aid in the identification
of analytes.

Phthalocyanines

Phthalocyanines (PCs) are a robust class of semi-conducting organic small molecules

that are used in a wide range of applications: photovoltaics, non-linear optics, xerog-
raphy, and chemical sensing [22]. Chemical sensors based on metallophthalocyanine
(MPc) thin films for the detection of donor and acceptor compounds (NH3 [23, 24],
NOx [25], Halogens) have been developed in the form of chemiresistors for over a
decade. The bases for using these molecules as gas sensing materials are their ease of
deposition as high quality, thin films (increasing the exposed surface area); their abil-
ity to achieve a desired molecular functionality by changing the central metal atom
or adding substituents to the phthalocyanine ring; and their thermal and chemical
stability.
The conductivity of MPc chemiresistors is sensitive to the presence of oxidizing
and reducing gases because a change in carrier concentration in the film occurs when

80
charge transfer states arise from the adsorption of electron-withdrawing or donating
gases. Previous studies have shown that there are many factors that can affect the
MPc’s ability to sense gases: film morphology, film thickness, operating tempera-
ture, and post-deposition annealing [26]. The Pc films can be deposited on electrodes
through a number of different methods such as Langmuir-Blodgett, spin-coating, ther-
mal evaporation, or self-assembly.

Carbon Nanotubes

Discovered in 1991 by Sumio Iijima as a byproduct of an electrical discharge between

graphite electrodes in an argon atmosphere [27], carbon nanotubes are chemical com-
pounds that consist of concentric cylinders of covalently bonded hexagonal carbon
rings. Carbon nanotubes range in size from a few nanometers in diameter and up to
hundreds of micrometers in length. Single-walled carbon nanotubes (SWNT) consist
of one concentric cylinder, while multiple-walled carbon nanotubes (MWNT) consist
of several concentric cylinders. Carbon nanotubes have several unique mechanical
and electrical properties that make them interesting from a sensing perspective: 1)
carbon nanotubes have a large surface area to volume ratio [28], 2) they can be eas-
ily functionalized without disturbing their electronic structure, and 3) their electrical
properties are extremely sensitive to the effects of charge transfer and chemical doping
by various molecules.
Because of the difficulty in manipulating single nanotubes, carbon nanotube chemire-
sistors are based on films of spin-casted nanotubes on top of interdigitated electrodes.
The interpenetrating network of carbon nanotubes has two sensing mechanisms. The
first transduction mechanism is charge transfer from physically adsorbed analyte
molecules, which modulate the Fermi level of the nanotube to cause a conductivity
change. This intratube modulation is only significant if the nanotube is semiconduct-
ing in nature. The second transduction mechanism occurs when analyte molecules
adsorb in the spaces between nanotubes, creating hopping conduction pathways be-
tween nanotubes that change the film’s conductivity. This intertube modulation oc-
curs for all types of molecules and for both metallic and semiconducting nanotubes,

81
similar to the interaction between conjugated polymers and adsorbed molecules. The
intratube modulation dominates the sensor response at lower concentrations with a
superlinear dependence on concentration and a high sensitivity, while the intertube
modulation shows a linear dependence and lower sensitivity over a broad range of
concentrations [29].
In principle, carbon nanotubes are not sensitive to many organic vapors, espe-
cially nonpolar molecules that will not transfer charge. Thus, polymer coatings and
sputtered Pd films have been used to functionalize the nanotubes to specifically de-
tect Cl2 and H2 , respectively [30]. Advantages of using carbon nanotube films include
detection levels on the order of ppm at room temperature, good reversibility, as well
as low variation between sensor response (less than 6%) making them suitable for use
in arrays. However, nanosensors based on this mechanism face sizeable challenges for
commercialization, such as the inability to selectively grow semiconducting nanotubes
and the complex techniques required to handle single nanotubes.

Chemicapacitor

Chemicapacitors are structurally the same as chemiresistors in that they consist of a

pair of electrodes and a sensing material. The sensing layer can be either sandwiched
between the two contacts [31] or deposited on top of a pair of planar interdigitated
electrodes on an inert mechanical support [32, 33]. However, instead of relying on
the changes in conductivity to detect the presence of analytes, this class of sensors
measures the change in capacitance that develops across the chemosensitive film as
the film absorbs the gaseous analyte. Absorption of analyte molecules changes the
dielectric constant and other physical properties (e.g. volume) to produce deviations
from the baseline capacitance [Fig. 3-3]. The chemicapacitors are usually excited
with an AC voltage excitation (ranging from a few kHz up to 500 kHz) while the
capacitances are measured using various readout circuits (frequency counters, output
voltage, Sigma-Delta-modulators, etc.). Further details on the sensing mechanisms
in chemicapacitor sensors can be found in [34].
The simple device design and low power consumption makes this platform attrac-

82
Figure 3-3: (a) Schematic of a typical chemicapacitor device using micro-machined,
parallel plate electrodes. In this scheme, swelling of the film causes changes in both the
dielectric constant () and the voltage (V) which contribute to changes in capacitance.
(b) Sample chemicapacitor response of the polymer, siloxanefluoro alcohol (SXFA), to
dimethyl methylphosphonate (DMMP) in a background of dry nitrogen and nitrogen
with 40% relative humidity [33].

83
tive for a wide range of gas sensing applications. In addition, the compatibility of the
device structure with traditional semiconductor fabrication techniques allows for easy
integration with complementary metal oxide semiconductor (CMOS) circuits and re-
duces the manufacturing costs needed to miniaturize the technology for portable gas
sensing applications. While the sensor itself is rugged (no moving parts) and allows
for a wide array of polymeric materials to be used as the sensor layer, the reliance on
changes in the dielectric constant of the polymer makes these sensors susceptible to
the high dielectric constant of moisture in the sampling environment. Humidity leads
to large capacitive changes, which initially led to polyimide chemicapacitors being
used as humidity sensors [35, 36].

More recent work in the chemicapacitor sensor field includes volatile organic com-
pound detection using a variety of polymers [37] and liquid crystals [38]. However,
most of the organic polymers that are used as sensitive coatings are only partially
selective to most volatile organic compounds. To achieve finer selectivity of analytes,
these sensors are usually placed in an array and combined with pattern recognition
algorithms. Another method for enhancing the selectivity in chemicapacitive sensors
is the use of configurable electrode configurations to provide two capacitive measure-
ments with the same sensor. For simple, planar interdigitated structures, the electric
field is mostly contained in the space above the electrodes to within a distance of
half the electrode periodicity [39]. The interdigitated electrode periodicity can be
changed by means of CMOS switches on the sensor chip, therefore the effective ratio
between sensing layer thickness and electrode periodicity changes. For thinner films,
the capacitance change will depend more on the total polarizable material absorbed
and less on the dielectric constant. The opposite is true for thicker films, making the
differential measurement more robust to ambient humidity [40].

Typically, chemicapacitors have sufficient sensitivity to detect common solvents

to below 100 ppm, and below 1 ppm for explosives and chemical warfare agents.
However, sensitivity will be temperature dependent as heat increases solvent volatil-
ity, decreasing the amount of absorbed analyte. The signal-to-noise ratio generally
increases with increasing polymer and analyte polarity. With dedicated signal con-

84
Figure 3-4: (a) Schematic of a typical OTFT gas sensor with a patterned gate elec-
trode. The simultaneous measurement of multiple device properties, yields more
information and makes OTFT’s a more attractive sensor platform. (b) Fractional
changes in drain current (Ion ) for eight semiconductors upon exposure to four odor-
ants: toluene, eugenol, vanillin, and l-carvone [45].

ditioning circuits the chemicapacitor is capable of resolving changes in capacitance

as small as .07 fF [41]. The response times (time to reach 90% of full signal), which
range from tens of seconds to minutes, are generally limited by the rates of diffusion
of analyte molecules through the polymer [33].

Organic Thin-Film Transistor (OTFT)

Organic thin-film transistor (OTFT) technology is still considered a relatively new

research area, with the first working OTFT fabricated out of polythiophene over two
decades ago [42]. One argument in favor of using OTFTs as gas sensors is that they
can provide more information about a particular analyte than an equivalent chemire-
sistor by monitoring changes in bulk conductivity, threshold voltage, field-induced
conductivity, and field effect mobility [43]. The other advantage OTFT sensors have
over chemiresistors is the signal amplification inherent in transistor device structures,
which produces gains in sensitivity and signal-to-noise ratio [44].
The basic structure of an OTFT gas sensor is shown in [Fig. 3-4]. Typically, a

85
conductive mechanical support is used as the gate electrode, upon which an insulating
dielectric coating is deposited. Next, a thermally evaporated or spun-coated thin film
of active organic semiconductor is added. Finally, the source and drain electrodes are
deposited over the semiconductor and are patterned to allow exposure of the active
region to the environment. The organic field effect transistor is biased so that a chan-
nel of field-induced charge is created at the interface between the organic layer and
the dielectric. When a bias is applied across the source and drain contacts, current
flows through the device. The transduction of the chemical signal into an electrical
signal occurs when the introduction of an analyte changes the current flowing through
the sensor. The major difference from a chemiresistor is that the gate voltage allows
for a modulation of the carrier density in the active sensing layer, which can enhance
sensitivity. Because morphology and attached end/side groups on the organic semi-
conductor molecules provide sites for analyte interaction, both can be adjusted to
provide additional improvements in sensitivity and selectivity [46]. There are several
materials that are used for the sensing layer in this class of gas sensors: conducting
polymers (CP) [47], phthalocyanines [48], and carbon nanotubes (CNT) [49]. The
transduction mechanism for each of these materials remains unchanged from the case
when the same materials are used for chemiresistors as discussed above.

Generally, there are three different areas where an analyte can modify the proper-
ties of the OTFT sensor: the bulk of the organic semiconductor, the interface between
the organic layer and the dielectric material, and the interface between the organic
layer and the electrical contacts [50]. Just as with chemiresistors, the conductivity in
the bulk organic semiconductor can change with the transfer of charge between the
gaseous analyte and the sensing film. The other bulk effect is a change in mobility,
which occurs when the interaction between the sensing layer and the analyte causes
swelling of the organic semiconductor. Because the active channel of current flow in
the OTFT is mainly confined to the first few molecular layers at the interface be-
tween the organic layer and the insulating dielectric material, amphiprotic molecules
(molecules that act as either an acid or base) such as alcohols, organic acids, water,
etc., can react on the surface of the dielectric to increase the device conductivity.

86
This additional current is typically considered a parasitic conductive pathway and
competes with the measured current signal from the organic semiconductor. Lastly,
chemical modulation of the interface between the electrical contacts and the organic
semiconductor can modulate the contact resistance and be responsible for the ob-
served sensing effect.
Recent work has focused on enhancing the sensitivity of OTFT sensors. Crone et
al. have reported that OTFT sensors are easily integrated into sensing circuits that
can enhance the performance of the sensor. Ring oscillators and adaptive amplifiers,
incorporating OTFT sensors, exhibit improved recovery time, sensitivity, and selec-
tivity [45]. In an effort to increase the incremental change in conductivity from each
analyte molecule by reducing the active area, Wang et al. have fabricated organic
transistors with channel lengths on the nanometer scale. While the results did not
yield the expected gains in sensitivity, the studies have elucidated the dependence of
the sensor response on the relative size of the channel length to the grain size [51, 52].
Because of their inherent simplicity and ease of integration with traditional silicon
devices, OTFTs will continue to be an active area of vapor-phase chemical sensor
research.

3.1.2 Optical Odor Sensors

For this class of chemosensors, changes in the optical properties of the organic sens-
ing layer are used to detect gaseous analytes. Many organic materials have high
absorption coefficients, high photoluminescent efficiencies, and absorption and emis-
sion profiles that span the visible spectrum (300 nm to 800 nm). For these reasons,
organic materials inherently provide a straightforward transduction mechanism to
convert chemical signals into visual outputs, which can then be detected by our
eyes, photodetectors, or CCD arrays for further processing. Optical odor sensors
are versatile in that they can use a range of measurements (absorbance, reflectance,
fluorescence, refractive index, polarization, colorimetry, interference, or scattering) to
simultaneously collect information regarding intensity and wavelength.
Generally, optical sensors consist of four basic components: a mechanism or a

87
material that transduces a chemical signal into a change in optical properties, an ex-
citation source to probe the sensor, optics for channeling light to and from the sensor,
and a detector for converting optical signals into electrical signals. This class of sen-
sors can be broadly organized into fluorescent and non-fluorescent technologies. Flu-
orescent technology is represented by amplified fluorescent polymers, non-fluorescent
technology by chemoresponsive dyes and surface plasmon resonance (SPR), while
optical fiber sensors encompass both types of sensor schemes.

Chemoresponsive Dyes

Colorimetric sensors using chemoresponsive dyes are the simplest of the optical odor
sensors and are analogous to simple pH indicator strips that change color in the
presence of an acid or base. Colorimetric gas sensors work in much the same way
by utilizing dyes that can change color in the presence of vapor-phase analytes. Not
only do we acquire qualitative information about the presence of certain gaseous
analytes through color changes that can be perceived by the human eye and brain,
but if scanned with a CCD, that information can then be processed in a quantitative
fashion to determine concentrations. While this type of sensor can use many different
instruments to convert the optical signal into an electrical response (the human eye
[53], flatbed scanner [54–56], photodetectors [57]), the sensors rely upon the same
fundamental shifts in the absorption spectra of the chemoresponsive chromophore to
indicate the presence of analyte.
When designing a chemoresponsive dye two principles must be followed: 1) each
dye must contain an interaction center that reacts strongly with analytes and 2) each
interaction center must be strongly coupled to an efficient chromophore. The first
principle requires that the analytes chemically interact with the dyes rather than
merely physically adsorbing to the dye molecules. The importance of the strong
dye-analyte interaction is highlighted by recent data that suggests that mammalian
olfactory receptors are metalloproteins, and that chemical interactions with the metal
center are intrinsic to the mechanism of smelling [58].
The metalloporphyrin family of organic molecules has found widespread use in

88
colorimetric sensors due to their intense coloration, their large spectral shifts upon
binding with foreign molecules, and the suite of organic chemistry techniques devel-
oped for these compounds. Many different chemoresponsive dyes can be synthesized
by combining different metals with the porphyrin structure to form stable metallo-
porphyrin complexes. In these complexes, the central metal atom plays a primary
role in the sensitivity and selectivity properties of the dye. Many volatile organic
compounds are excellent ligands for metal ions and readily bind to the central metal
atom to provide large color changes in the dye. In addition, the ability to add a
wide range of side groups to peripheral positions on the porphyrin ring can enable
enhanced selectivity.
Sensor arrays made from these chemoresponsive dyes are sensitive (110 ppb to 2
ppm) to volatile organic compounds and with proper choice of dye and mechanical
support are not affected by ambient humidity. Also, the response time of colorimetric
sensors is relatively slow (2 min. at 1 ppm analyte concentration) limited by the
flow rate of the gas delivery to the sensor or equilibration of low vapor concentra-
tions over large surface areas [59]. The simple, low-cost fabrication techniques for
depositing chemoresponsive dye sensor arrays makes them ideal for general-purpose
vapor dosimeters or disposable analyte-specific detectors. However, implementation
of a gas monitoring system using this technology becomes prohibitively expensive to
install and maintain.

Surface Plasmon Resonance

Surface plasmon resonance (SPR) is an optical phenomenon where the conditions

for total internal reflection at the interface between materials with different indices
of refraction are satisfied, allowing an exponentially decaying evanescent wave to
propagate in the medium with the lower index of refraction. When that interface
is coated with a thin film of conductive metal (i.e. gold or silver), the evanescent
wave can excite surface plasmon waves at the interface that propagate parallel to the
interface. The coupling between the incident light with the surface plasmon waves
(SPW) in the thin film causes the reflected light to assume a particular wavelength,

89
Figure 3-5: Kretschmann configuration of SPR gas sensing method. The graph indi-
cates a real-time measurement of intensity at a given wavelength and angle of reflected
light.

angle, intensity, or phase. Thus, any changes in the local refractive index at the
interface will change the optical properties of the reflected light from the interface
and produce a signal [Fig. 3-5].
SPR chemical sensors consist of an optical system, a transducing medium, and an
electronic system for data processing. The transducing medium converts changes in
the quantity of detected analyte into changes in refractive index, enabling the use of
SPR to probe the variation. Typically, thin films of polymers [60] and small-molecule
organics [61] are used as the transducing medium. As with most organic materials,
the films can be deposited using spin coating, dip coating, or Langmuir-Blodgett tech-
niques. SPR techniques can even be combined with optical fiber sensors to miniatur-
ize the device design. While comparable sensitivities can be achieved, careful control
over the measurement environment (mechanical, temperature, polarization) must be
maintained [62, 63].
Sensor sensitivity, stability, and resolution depend upon the combined properties
of the optical system and the sensing film. However, the selectivity, reversibility,
stability, and response time are primarily determined by the properties of the sensing
film. Because the SPR signals are strongly influenced by the optical properties of the

90
sensing film, SPR sensors are extremely sensitive to small changes in the refractive
index (down to 5 x 10-7 RIU) corresponding to better than 1 ppm levels of detection.
Despite these high sensitivities, SPR will only detect adsorbates that have a molecular
weight > 500 Da, which can efficiently change refractive indices [64]. Some recent
work has demonstrated the use of nematic liquid crystals [65] and cavitands (a class
of supramolecule) [66] for SPR gas sensors. These materials exhibit relatively fast
response times (< 1 s) enabling this class of SPR sensors to be used for real-time
monitoring of some volatile organic compounds.

Optical Fiber Sensors

Fiber optic technology first developed for the communications industry has been
finding its way into chemical sensors. There are many technological reasons that
make fiber optics sensors an attractive proposition. Optical fiber sensors are passive,
requiring no electricity for operation. This can be important when trying to detect
analytes in highly combustible environments or if the analyte itself is flammable at
low vapor concentrations. For the same reason, fiber optic sensors are resistant to
electromagnetic interference (EMI), making them useful for detecting analytes in
environments with high and variable electric fields such as the power industry. In
addition, fiber optics are capable of transmitting optical signals for long distances
with low losses, enabling a receiver to be placed some distance away from the point of
detection for remote sensing capabilities. By using existing optical instrumentation,
optical sensing can be readily multiplexed to carry different measurement signals
simultaneously. For most applications, however, fiber optic sensors are far more
costly to implement than other gas sensing approaches [67].
Recent work has investigated conductive polymers [68, 69], sol-gels [70] and chemore-
sponsive dyes [71] for use as the gas-sensitive transduction medium in fiber optic
sensors. In general, fiber optic sensors can be divided into two different sensing ar-
chitectures: extrinsic and intrinsic. If the transduction of the chemical signal into an
optical signal occurs within the fiber itself, it is known as an intrinsic sensor. Ex-
trinsic sensors perform the transduction outside of the fiber [Fig. 3-6]. In extrinsic

91
Figure 3-6: (a) Extrinsic optical fiber sensor, where fibers are used to guide input and
output signals (b) Intrinsic optical fiber sensor, where the sensing film is incorporated
into fiber structure as a cladding material.

sensors the fiber acts as a transmission line to bring light to a detection cell and to
guide the light output to a photodetector or a spectrometer [72]. The detection cell
contains the organic materials that provide selective gas detection. However, most
fiber optic sensor research focuses on intrinsic sensors, with the most common being
evanescent wave sensors.

In evanescent wave sensors, a small section of the passive cladding of the fiber is
replaced with a chemically active sensing material. Thus, any interaction between a
gaseous analyte and the sensitive material will manifest in a change in the effective
index of refraction in the fiber, changing its transmission properties [68]. The chem-
ically active material can be deposited onto the fiber in a number of ways once the
silica cladding is removed through polishing or etching: dip coating, sol-gel deposi-
tion, or Langmuir-Blodgett [73]. For more information regarding the operation and
theory of evanescent wave fiber optic sensors the reader is directed to literature [74].

92
 A major limitation of the sensitivity of the intrinsic fiber sensor is the ability to
transfer optical power into the evanescent field. Recent work has incorporated new
types of fibers to increase the sensitivity of measurements. Newly developed hollow
core fibers, which guide light mainly in the cladding, can be used to couple more
power into the chemosensitive materials [75]. If the hollow core fiber is used as the
active sensing area between two silica core fibers, then the transitions between the
fibers will force the light to travel from the core into the cladding then back into the
core. Another type of fiber is the microstructure fiber or photonic crystal fiber, which
incorporates air holes into the silica cladding to enhance the interaction between the
evanescent field and the gaseous analyte. By adjusting the size and placement of the
air holes, light propagation in the fiber can be controlled [76].
Other advances in intrinsic fiber optic sensors include Bragg grating fiber sensors
[77, 78], active fiber core [70], optodes [73, 79], fluorescent fibers [80, 81], and some
hybrid combination of the previous sensor schemes [63]. In general, the selectivity
of the fiber optic sensor is governed by the light source wavelength and the chemi-
cal and physical properties of the sensing material. Special input signal modulation
techniques can be used to enhance the selectivity of the sensors by reducing the cross-
reactivity of other gases. Sensitivities are competitive with other sensor technologies
(low ppm) and response times are relatively fast (∼10 s), allowing for real-time moni-
toring of gaseous analytes such as ammonia [82]. However, prohibitive manufacturing
costs, careful calibration requirements, and large amounts of overhead electronics and
optics make fiber optic sensor systems less robust and more difficult to integrate into
existing silicon circuit architectures. In addition, low dynamic range (as compared to
other sensing schemes) and unsustainable reversibility limit the number of suitable
applications.

Photoluminescent Sensors

Photoluminescence (PL) techniques are at the heart of many optical sensing schemes
for a wide range of applications and can be extremely sensitive, enabling the detection
of single molecules [83]. In general, PL is generated by fluorophores that absorb light

93
at wavelengths within their absorption spectrum and emit light at longer wavelengths,
corresponding to the emission spectrum. The energy difference in emission and ab-
sorption spectra is referred to as a Stoke shift and represents vibrational relaxation
and other energy loss mechanisms following absorption of a photon by a molecule.
The PL sensor scheme is surprisingly simple and can yield high sensitivities [Fig. 3-7].
A thin film of the chemoresponsive material is excited by an excitation source (LED,
laser, etc.) and a photodetector is used to collect the emission of the transducing
material. The introduction of an analyte vapor will result in a change in optical
properties yielding a differential measured signal. An increase in emission results in
a “turn-on” signal, while emission quenching results in a “turn-off” signal. “Turn-
on” schemes require that the photoluminescent material be highly quenched. As a
result, the distance between quenching sites will be less than the exciton diffusion
length. In this situation, the sensitivity of the “turn-on” response will be decreased
relative to the “turn-off” response, which relies on long exciton diffusion lengths to
yield the high sensitivities necessary for ultra-trace detection. Additionally, the mea-
sured quantity in both cases is a differential intensity signal, making the lower limit
of the signal-to-noise ratio (SNR) the determining factor of sensitivity. Thus, the
only case where the “turn-on” response is easier to observe is if the new signal is
produced against a completely dark background, so that the detection is no longer
based on a ratio of signals. This is usually not the case, as background fluorescence
often negates the sensitivity advantage of the “turn-on” response with regard to SNR.
However, for trace detection in complex environments, “turn-on” detection schemes
do have an advantage in selectivity as there are many more “false-positives” caused
by adventitious quenching as opposed to fluorescence enhancement [84].

With the ability to measure a number of different parameters (decay time, energy
transfer, polarization, quenching efficiency, intensity) and simple instrumentation,
luminescent sensor techniques have become standard practice [85]. The ubiquity of
luminescent sensing schemes and low-cost, high quality electronics, has even made
possible the use of a computer screen as a programmable light source and a web
camera as a photodetector in a chemical sensing technique known as the computer

94
Figure 3-7: Schematic of a typical configuration of a PL gas sensor. The LED is
used to excite the sensing film, while the filter acts to prevent the excitation from
reaching the photodetector. Upon introduction of a gaseous analyte, the sensing
film’s luminescence is quenched, producing a reduction in the PL intensity measured
by the photodetector.

screen photo assisted technique (CSPT) [86]. As luminescent sensing methods have
matured, much of the ongoing research in this field focuses on the synthesis of new
luminescent chemoresponsive materials to enhance analyte sensitivity and selectivity.
Organic PL chemoresponsive materials span a wide range of compounds including
metalloporphyrin dye molecules [87], supramolecules, conjugated organic polymers
[88], conjugated inorganic polymers [89], and phosphorescent molecules [90]. These
luminescent compounds are either deposited as thin films [91], used as coatings for
particles [92], or incorporated into inert, permeable matrices [93].

3.1.3 Summary

While much of the surveyed research exhibits promising vapor-phase sensing perfor-
mance, many of the technologies remain experimental and bound to a laboratory
setting [Table 3.1]. Most of the commercial gas sensors available today utilize older,
more mature technologies such as electrochemical cells, catalytic beads, photoion-
ization detectors (PID), SAW, metal oxide semiconductors (MOS), and QCM. The
dearth of viable organic solid-state vapor-phase chemosensors indicates that there is

95
much work still to be done (in terms of material stability, selectivity, etc.) before
commercialization becomes commonplace for organic sensors.

96
 Chemosensor Type Material classes Analytes Applications Companies
Chemiresistor OCP VOCs Medical diagnosis, Food/Beverage, Process control, Security Scensive Technologies
OCP VOCs Pharmaceutical, Quality control Alpha M.O.S.
OCP VOCs Breath analysis Abtech Scientific Inc.
Pc/MPc NOx , NH3 , O2 , Halogens Environmental monitoring (Research only)
Carbon black composite VOCs Petrochemical, Food/Beverage, Packaging, Plastics, Pet food Smiths Detection (Cyranose)
CNT H2 , CO2 Industrial gas detection, Breath analysis Nanomix
Chemicapacitor OCP VOCs, CWAs, TICs Environmental monitoring, Security, Process control Seacoast Science Inc.
OCP H2 O Indoor air quality (Relative Humidity) Jackson Systems
OTFT OCP VOCs Environmental monitoring (Research only)
CNT H2 S, NOx , COx , NH3 , N2 Environmental monitoring (Research only)
Chemoresponsive Dyes Metalloporphyrins Toxic Gas Toxic gas dosimeter Honeywell Analytics
Metalloporphyrins VOCs Bacteria identification, Breath analysis ChemSensing Inc.
Metalloporphyrins Hydrocarbon Meat spoilage FQSI
Metalloporphyrins H2 S, Cl2 , CO, N2 H4 , O3 Toxic gas dosimeter American Gas and Chemical Co.
SPR Metalloporphyrins O2 Oxygen sensing Sentronic
Optical Fiber Dye/Polymer matrix Hydrocarbons Environmental monitoring PetroSense

97
Dye/Polymer matrix O2 Oxygen sensing Sentronic
Dye/Polymer matrix VOCs, O2 Environmental monitoring, Process control Interlab (Optosen)
Dye/Polymer matrix O2 Oxygen sensing Ocean Optics
PL Sensors OCP Explosives, CWAs Security, Life Sciences ICx Technologies
OCP VOCs Air quality, Security, Process control CogniScent
OCP VOCs (In development) Nanoident
Metalloporphyrins O2 Oxygen sensing Sentronic

Table 3.1: Commercially available and research-grade solid-state organic vapor-phase chemical sensors and e-nose instruments
as of May 2007. Abbreviations: Organic Conjugated Polymer (OCP), Phthalocyanine (Pc), Metallophthalocyanine (MPc),
Carbon Nanotube (CNT), Volatile Organic Compound (VOC), Chemical Warfare Agents (CWA), Toxic Industrial Compound
(TIC), Organic Thin Film Transistor (OTFT), Surface Plasmon Resonance (SPR), and Photoluminescent (PL).
98
Chapter 3 References

[1] B. C Sisk and N. S Lewis. Comparison of analytical methods and calibra-

tion methods for correction of detector response drift in arrays of carbon
black-polymer composite vapor detectors. Sensors and Actuators B-Chemical,
104(2):249–268, 2005.

[2] L. Turin. Structure-odor relations: A modern perspective. Handbook fo Olfaction

and Gustation. Marcel Dekker, New York, 2002.

[3] M. Thompson and D. C Stone. Surface-Launched Acoustic Wave Sensors, volume

144 of Chemical Analysis. John Wiley & Sons, Inc., New York, 1997.

[4] R. Lucklum and P. Hauptmann. Acoustic microsensors-the challenge behind

microgravimetry. Analytical and Bioanalytical Chemistry, 384(3):667–682, 2006.

[5] C. K O’Sullivan and G. G Guilbault. Commercial quartz crystal microbalances

- theory and applications. Biosensors & Bioelectronics, 14(8-9):663–670, 1999.

[6] Y. Jin, K. M Wang, and R. Jin. Development of novel and sensitive sensors
based on microcantilever of atomic force microscope. Progress in Natural Science,
16(5):445–451, 2006.

[7] A. Hierlemann, O. Brand, C. Hagleitner, and H. Baltes. Microfabrication tech-

niques for chemical/biosensors. Proceedings of the IEEE, 91(6):839–863, 2003.

[8] T. Gao, M.D. Woodka, B.S. Brunschwig, and N.S. Lewis. Chemiresistors for
Array-Based vapor sensing using composites of carbon black with low volatility
organic molecules. Chemistry of Materials, 18(22):5193–5202, October 2006.

99
 [9] A. J Heeger. Semiconducting and metallic polymers: The fourth generation of
polymeric materials (Nobel lecture). Angewandte Chemie-International Edition,
40(14):2591–2611, 2001.

[10] W. Zheng, Y. Min, A. G MacDiarmid, M. Angelopoulos, Y. H Liao, and A. J

Epstein. Effect of organic vapors on the molecular conformation of non-doped
polyaniline. Synthetic Metals, 84(1-3):63–64, 1997.

[11] J. X Huang, S. Virji, B. H Weiller, and R. B Kaner. Polyaniline nanofibers:

Facile synthesis and chemical sensors. Journal of the American Chemical Society,
125(2):314–315, 2003.

[12] J. Huang, S. Virji, B. H Weiller, and R. B Kaner. Nanostructured polyaniline

sensors. Chemistry-a European Journal, 10(6):1315–1319, 2004.

[13] B. Liu and G. C Bazan. Optimization of the molecular orbital energies of con-
jugated polymers for optical amplification of fluorescent sensors. Journal of the
American Chemical Society, 128(4):1188–1196, 2006.

[14] R. Kessick and G. Tepper. Electrospun polymer composite fiber arrays for the de-
tection and identification of volatile organic compounds. Sensors and Actuators
B-Chemical, 117(1):205–210, 2006.

[15] B. Li, G. Sauve, M. C Iovu, M. Jeffries-El, R. Zhang, J. Cooper, S. Santhanam,

L. Schultz, J. C Revelli, A. G Kusne, T. Kowalewski, J. L Snyder, L. E Weiss,
G. K Fedder, R. D McCullough, and D. N Lambeth. Volatile organic compound
detection using nanostructured copolymers. Nano Letters, 6(8):1598–1602, 2006.

[16] L. M Dai, P. Soundarrajan, and T. Kim. Sensors and sensor arrays based
on conjugated polymers and carbon nanotubes. Pure and Applied Chemistry,
74(9):1753–1772, 2002.

[17] M. K Ram, O. Yavuz, V. Lahsangah, and M. Aldissi. CO gas sensing from

ultrathin nano-composite conducting polymer film. Sensors and Actuators B-
Chemical, 106(2):750–757, 2005.

100
[18] S. L. Tan. Velocity-optimized diffusion for ultra-fast polymer based resistive gas
sensors. IEEE Proc.-Sci. Meas. Technol., 153(3):94–100, 2006.

[19] T. C Pearce, S. S Schiffman, H. T Nagle, and J. W Gardner, editors. Handbook

of Machine Olfaction. Wiley-VCH, Weinheim, 2003.

[20] B. C Sisk and N. S Lewis. Vapor sensing using polymer/carbon black composites
in the percolative conduction regime. Langmuir, 22(18):7928–7935, 2006.

[21] E. S Tillman, M. E Koscho, R. H Grubbs, and N. S Lewis. Enhanced sen-

sitivity to and classification of volatile carboxylic acids using arrays of linear
poly(ethylenimine)-carbon black composite vapor detectors. Analytical Chem-
istry, 75(7):1748–1753, 2003.

[22] N. Kobayashi. Phthalocyanines. Current Opinion in Solid State & Materials

Science, 4(4):345–353, 1999.

[23] B. Schollhorn, J. P Germain, A. Pauly, C. Maleysson, and J. P Blanc. Influence

of peripheral electron-withdrawing substituents on the conductivity of zinc ph-
thalocyanine in the presence of gases. part 1: reducing gases. Thin Solid Films,
326(1-2):245–250, 1998.

[24] J. P Germain, A. Pauly, C. Maleysson, J. P Blanc, and B. Schollhorn. Influence

of peripheral electron-withdrawing substituents on the conductivity of zinc ph-
thalocyanine in the presence of gases. part 2: oxidizing gases. Thin Solid Films,
333(1-2):235–239, 1998.

[25] W. F Qiu, W. P Hu, Y. Q Liu, S. Q Zhou, Y. Xu, and D. B Zhu. The gas sensitiv-
ity of a substituted metallophthalocyanine, tetra-iso-propoxyphthalocyaninato
copper (II). Sensors and Actuators B-Chemical, 75(1-2):62–66, 2001.

[26] Y. L. Lee. NO2 sensing characteristics of copper phthalocyanine films: Effects of

low temperature annealing and doping time. Sensors and Actuators B-Chemical,
119:174–179, 2006.

101
[27] S. Iijima. Helical Microtubules of Graphitic Carbon. Nature, 354(6348):56–58,
1991.

[28] J. Li, Y. Lu, and M. Meyyappan. Nano chemical sensors with Polymer-Coated
carbon nanotubes. IEEE Sensors Journal, 6(5):1047–1051, 2006.

[29] J. Li, Y. Lu, Q. Ye, M. Cinke, J. Han, and M. Meyyappan. Carbon nanotube
sensors for gas and organic vapor detection. Nano Letters, 3(7):923–933, 2003.

[30] I. Sayago, E. Terrado, M. Aleixandre, M. C Horrillo, M. J Fernandez, J. Lozano,

E. Lafuente, W. K Maser, A. M Benito, M. T Martinez, J. Gutierrez, and
E. Munoz. Novel selective sensors based on carbon nanotube films for hydrogen
detection. sensors and Actuators B-Chemical, 122:75–80, 2007.

[31] H. Shibata, M. Ito, M. Asakursa, and K. Watanabe. A digital hygrometer using a

polyimide film relative humidity sensor. IEEE Transactions on Instrumentation
and Measurement, 45(2):564–569, 1996.

[32] R. Igreja and C. J Dias. Capacitance response of polysiloxane films with interdigi-
tal electrodes to volatile organic compounds, volume 455-456 of Materials Science
Forum, pages 420–424. Trans Tech Publications Ltd, Zurich-Uetikon, 2004.

[33] S. V Patel, T. E Mlsna, B. Fruhberger, E. Klaassen, S. Cemalovic, and D. R

Baselt. Chemicapacitive microsensors for volatile organic compound detection.
Sensors and Actuators B-Chemical, 96(3):541–553, 2003.

[34] A. M Kummer, A. Hierlemann, and H. Baltes. Tuning sensitivity and selec-

tivity of complementary metal oxide semiconductor-based capacitive chemical
microsensors. Analytical Chemistry, 76(9):2470–2477, 2004.

[35] T. Boltshauser and H. Baltes. Capacitive Humidity Sensors in SACMOS Technol-

ogy with Moisture Absorbing Photosensitive Polyimide. Sensors and Actuators
A-Physical, 26(1-3):509–512, 1991.

102
[36] T. Boltshauser, L. Chandran, H. Baltes, F. Bose, and D. Steiner. Humidity
Sensing Properties and Electrical Permittivity of New Photosensitive Polyimides.
Sensors and Actuators B-Chemical, 5(1-4):161–164, 1991.

[37] F. Josse, R. Lukas, R. N Zhou, S. Schneider, and D. Everhart. AC-impedance-

based chemical sensors for organic solvent vapors. Sensors and Actuators B-
Chemical, 36(1-3):363–369, 1996.

[38] F. L Dickert, G. K Zwissler, and E. Obermeier. LIQUID-CRYSTALS ON IN-

TERDIGITAL STRUCTURES - APPLICATIONS AS CAPACITIVE CHEMI-
CAL SENSORS. Berichte Der Bunsen-Gesellschaft-Physical Chemistry Chemi-
cal Physics, 97(2):184–188, 1993.

[39] Van Gerwen P.[1], Laureyn W., Laureys W., Huyberechts G., Op De Beeck M.,
Baert K., Suls J., Sansen W., Jacobs P., Hermans L., and Mertens R. Nanoscaled
interdigitated electrode arrays for biochemical sensors. Sensors and Actuators
B: Chemical, 49:73–80, June 1998.

[40] A. M Kummer and A. Hierlemann. Configurable electrodes for capacitive-type

sensors and chemical sensors. IEEE Sensors Journal, 6(1):3–10, 2006.

[41] R. Igreja and C. J Dias. Organic vapour discrimination using sorption sensitive
chemocapacitor arrays, volume 514-516 of Materials Science Forum, pages 1064–
1067. Trans Tech Publications Ltd, Zurich-Uetikon, 2006.

[42] A. Tsumura, H. Koezuka, and T. Ando. Macromolecular Electronic Device -

Field-Effect Transistor with a Polythiophene Thin-film. Applied Physics Letters,
49(18):1210–1212, 1986.

[43] L. Torsi, A. Dodabalapur, L. Sabbatini, and P. G Zambonin. Multi-parameter

gas sensors based on organic thin-film-transistors. Sensors and Actuators B-
Chemical, 67(3):312–316, 2000.

[44] J. T Mabeck and G. G Malliaras. Chemical and biological sensors based on

103
 organic thin-film transistors. Analytical and Bioanalytical Chemistry, 384(2):343–
353, 2006.

[45] B. K Crone, A. Dodabalapur, R. Sarpeshkar, A. Gelperin, H. E Katz, and Z. Bao.

Organic oscillator and adaptive amplifier circuits for chemical vapor sensing.
Journal of Applied Physics, 91(12):10140–10146, 2002.

[46] J. Locklin and Z. N Bao. Effect of morphology on organic thin film transistor
sensors. Analytical and Bioanalytical Chemistry, 384(2):336–342, 2006.

[47] J. B Chang, V. Liu, V. Subramanian, K. Sivula, C. Luscombe, A. Murphy, J. S

Liu, and J. M. J Frechet. Printable polythiophene gas sensor array for low-cost
electronic noses. Journal of Applied Physics, 100(1):7, 2006.

[48] M. Bouvet. Phthalocyanine-based field-effect transistors as gas sensors. Analyt-

ical and Bioanalytical Chemistry, 384(2):366–373, 2006.

[49] C. S Huang, B. R Huang, Y. H Jang, M. S Tsai, and C. Y Yeh. Three-terminal

CNTs gas sensor for n-2 detection. Diamond and Related Materials, 14(11-
12):1872–1875, 2005.

[50] H. Chen, M. Josowicz, and J. Janatax. Chemical effects in organic electronics.

Chemistry of Materials, 16(23):4728–4735, 2004.

[51] L. Wang, D. Fine, and A. Dodabalapur. Nanoscale chemical sensor based on

organic thin-film transistors. Applied Physics Letters, 85(26):6386–6388, 2004.

[52] L. Wang, D. Fine, S. I Khondaker, T. Jung, and A. Dodabalapur. Sub 10 nm

conjugated polymer transistors for chemical sensing. Sensors and Actuators B-
Chemical, 113(1):539–544, 2006.

[53] N. A Rakow and K. S Suslick. A colorimetric sensor array for odour visualization.
Nature, 406(6797):710–713, 2000.

[54] K. S Suslick. An optoelectronic nose: ”Seeing” smells by means of colorimetric

sensor arrays. MRS Bulletin, 29(10):720–725, 2004.

104
[55] N. A Rakow, A. Sen, M. C Janzen, J. B Ponder, and K. S Suslick. Molecular
recognition and discrimination of amines with a colorimetric array. Angewandte
Chemie-International Edition, 44(29):4528–4532, 2005.

[56] K. S Suslick, N. A Rakow, and A. Sen. Colorimetric sensor arrays for molecular
recognition. Tetrahedron, 60(49):11133–11138, 2004.

[57] T. Nakamoto, M. Yosihioka, Y. Tanaka, K. Kobayashi, T. Moriizumi, S. Ueyama,

and W. S Yerazunis. Colorimetric method for odor discrimination using dye-
coated plate and multiLED sensor. Sensors and Actuators B-Chemical, 116(1-
2):202–206, 2006.

[58] J. Y Wang, Z. A Luthey-Schulten, and K. S Suslick. Is the olfactory receptor a

metalloprotein? Proceedings of the National Academy of Sciences of the United
States of America, 100(6):3035–3039, 2003.

[59] M. C Janzen, J. B Ponder, D. P Bailey, C. K Ingison, and K. S Suslick. Col-

orimetric sensor arrays for volatile organic compounds. Analytical Chemistry,
78(11):3591–3600, 2006.

[60] A. V Samoylov, V. M Mirsky, Q. Hao, C. Swart, Y. M Shirshov, and O. S

Wolfbeis. Nanometer-thick SPR sensor for gaseous HCl. Sensors and Actuators
B-Chemical, 106(1):369–372, 2005.

[61] T. Basova, E. Kol’stov, A. K Ray, A. K Hassan, A. G Gurek, and V. Ahsen.

Liquid crystalline phthalocyanine spun films for organic vapour sensing. Sensors
and Actuators B-Chemical, 113(1):127–134, 2006.

[62] X. Bevenot, A. Trouillet, C. Veillas, H. Gagnaire, and M. Clement. Surface

plasmon resonance hydrogen sensor using an optical fibre. Measurement Science
& Technology, 13(1):118–124, 2002.

[63] Y. C Kim, W. Peng, S. Banerji, and K. S Booksh. Tapered fiber optic surface
plasmon resonance sensor for analyses of vapor and liquid phases. Optics Letters,
30(17):2218–2220, 2005.

105
[64] J. Mavri, P. Raspor, and M. Franko. Application of chromogenic reagents in
surface plasmon resonance (SPR). Biosensors & Bioelectronics, 22(6):1163–1167,
2007.

[65] E. B Feresenbet, F. Taylor, T. M Chinowsky, S. S Yee, and D. K Shenoy. Optical

chemical sensing using nematic liquid crystal. Sensor Letters, 2(2):145–152, 2004.

[66] E. B Feresenbet, E. Dalcanale, C. Dulcey, and D. K Shenoy. Optical sensing of the

selective interaction of aromatic vapors with cavitands. Sensors and Actuators
B-Chemical, 97(2-3):211–220, 2004.

[67] B. Culshaw. Optical fiber sensor technologies: Opportunities and-perhaps-

pitfalls. Journal of Lightwave Technology, 22(1):39–50, 2004.

[68] S. Khalil, L. Bansal, and M. El-Sherif. Intrinsic fiber optic chemical sensor for
the detection of dimethyl methylphosphonate. Optical Engineering, 43(11):2683–
2688, 2004.

[69] E. Scorsone, S. Christie, K. C Persaud, and F. Kvasnik. Evanescent sensing of

alkaline and acidic vapours using a plastic clad silica fibre doped with poly(o-
methoxyaniline). Sensors and Actuators B-Chemical, 97(2-3):174–181, 2004.

[70] T. Shinquan, C. B Winstead, R. Jindal, and J. P Singh. Optical-Fiber sensor

using tailored porous Sol-Gel fiber core. IEEE Sensors Journal, 4:322–328, 2004.

[71] P. A. S Jorge, M. Mayeh, R. Benrashid, P. Caldas, J. L Santos, and F. Farahi.

Applications of quantum dots in optical fiber luminescent oxygen sensors. Applied
Optics, 45(16):3760–3767, 2006.

[72] S. Christie, E. Scorsone, K. Persaud, and F. Kvasnik. Remote detection of

gaseous ammonia using the near infrared transmission properties of polyaniline.
Sensors and Actuators B-Chemical, 90(1-3):163–169, 2003.

[73] C. Elosua, I. R Matias, C. Bariain, and F. J Arregui. Volatile organic compound

optical fiber sensors: A review. Sensors, 6(11):1440–1465, 2006.

106
[74] A. Messica, A. Greenstein, and A. Katzir. Theory of fiber-optic, evanescent-wave
spectroscopy and sensors. Applied Optics, 35(13):2274–2284, 1996.

[75] I. R Matias, J. Bravo, F. J Arregui, and J. M Corres. Nanofilms onto a hollow

core fiber. Optical Engineering Letters, 45:1–3, 2006.

[76] Y. L Hoo, W. Jin, H. L Ho, D. N Wang, and R. S Windeler. Evanescent-wave

gas sensing using microstructure fiber. Optical Engineering, 41(1):8–9, 2002.

[77] S. Luo, Y. Liu, A. Sucheta, M. Evans, and R. V Tassell. Applications of LPG fiber
optical sensors for relative humidity and chemical warfare agents monitoring.
Proceedings of SPIE, 4920:193–204, 2002.

[78] R. Falate, R. C Kamikawachi, M. Muller, H. J Kalinowski, and J. L Fabris. Fiber

optic sensors for hydrocarbon detection. Sensors and Actuators B-Chemical,
105(2):430–436, 2005.

[79] M. Giordano, M. Russo, A. Cusano, and G. Mensitieri. An high sensitivity

optical sensor for chloroform vapours detection based on nanometric film of d-
form syndiotactic polystyrene. Sensors and Actuators B-Chemical, 107:140–147,
2005.

[80] P. D O’Neal, A. Meledeo, J. R Davis, B. L Ibey, V. A Gant, M. V Pishko, and

G. L Cote. Oxygen sensing based on the flourescence quenching of a ruthenium
complex immobilized in a biocompatible Poly(Ethylene glycol) hydrogel. IEEE
Sensors Journal, 4:728–734, 2004.

[81] P. A. S Jorge, P. Caldas, C. C Rosa, A. G Oliva, and J. L Santos. Optical fiber

probes for flourescent based oxygen sensing. Sensors and Actuators B-Chemical,
130:290–299, 2004.

[82] W. Q Cao and Y. X Duan. Optical fiber-based evanescent ammonia sensor.

Sensors and Actuators B-Chemical, 110(2):252–259, 2005.

107
[83] K. D Weston, P. J Carson, J. A DeAro, and S. K Buratto. Single-molecule
detection fluorescence of surface-bound species in vacuum. Chemical Physics
Letters, 308(1-2):58–64, 1999.

[84] S. W Thomas, G. D Joly, and T. M Swager. Chemical sensors based on amplifying

fluorescent conjugated polymers. Chemical Reviews, 107:1339–1386, 2007.

[85] J. F Callan, A. Prasanna de Silva, and D. C Magri. Luminescent sensors and

switches in the early 21st century. Tetrahedron, 61:8551–8588, 2005.

[86] D. Filippini, A. Alimelli, C. Di Natale, R. Paolesse, A. D’Amico, and

I. Lundstrom. Chemical sensing with familiar devices. Angewandte Chemie-
International Edition, 45:3800–3803, 2006.

[87] C. Di Natale, R. Paolesse, and A. D’Amico. Metalloporphyrins based artificial

olfactory receptors. Sensors and Actuators B-Chemical, 121(1):238–246, 2007.

[88] D. T McQuade, A. E Pullen, and T. M Swager. Conjugated polymer-based

chemical sensors. Chemical Reviews, 100(7):2537–2574, 2000.

[89] S. J Toal, J. C Sanchez, R. E Dugan, and W. C Trogler. Visual detec-

tion of trace nitroaromatic explosive residue using photoluminescent metallole-
containing polymers. Journal of Forensic Sciences, 52(1):79–83, 2007.

[90] I. Sanchez-Barragan, J. M Costa-Fernandez, M. Valledor, J. C Campo, and

A. Sanz-Medel. Room-terperature phosphorescence (RTP) for optical sensing.
Trends in Analytical Chemistry, 25(10):958–967, 2006.

[91] S. J Zhang, F. T Lu, L. N Gao, L. P Ding, and Y. Fang. Fluorescent sensors for
nitroaromatic compounds based on monolayer assembly of polycyclic aromatics.
Langmuir, 23(3):1584–1590, 2007.

[92] S. Bencic-Nagale, T. Sternfeld, and D. R Walt. Microbead chemical switches:

An approach to detection of reactive organophosphate chemical warfare agent
vapors. Journal of the American Chemical Society, 128(15):5041–5048, 2006.

108
[93] S. Y Tao, Z. Y Shi, G. T Li, and P. Li. Hierarchically structured nanocom-
posite films as highly sensitive chemosensory materials for TNT detection.
Chemphyschem, 7(9):1902–1905, 2006.

109
110
Chapter 4

Bilayer Heterojunction
Photoconductors

Fundamental progress has to do with

the reinterpretation of basic ideas.

ALFRED NORTH WHITEHEAD

Philosopher

The bilayer heterojunction photoconductor combines the concept of a heterojunc-

tion with the device architecture of a photoconductor. Unlike traditional organic
light emitting diodes (OLEDs) or photovoltaics, the heterojunction is aligned along
the normal of the applied field. To better understand and model this novel device
architecture, we can build upon previously conducted research in organic heterojunc-
tions and photoconductivity of organic solids.

4.1 Organic Heterojunction Theory

Organic heterojunctions play a central role in practically all organic optoelectronic
devices. Ever since Ching Tang’s [1] seminal work in 1987 that incorporated a het-
erojunction into an efficient OLED, the organic electronics field has been utilizing
the heterostructure in a number of other device structures in order to manipulate
carrier concentrations and to boost efficiencies. The fundamental idea behind the

111
Figure 4-1: (a) Type-I (Straddling) heterojunction (b) Type-II (Staggered) hetero-
junction (c) Type-III (Broken-gap) heterojunction.

heterojunction is a union of two dissimilar materials, where the heterojunction is

categorized by the energy band alignment of the two materials. A Type-I (strad-
dling) [Fig. 4-1(a)] heterojunction consists of a smaller bandgap material that lies
entirely within the energy bands of the wider bandgap material. The discontinuities
of the bands are such that both types of carriers (electrons and holes) need energy
to hop into the larger bandgap material from the smaller bandgap material, while
carriers gain energy in the opposite situation. This heterojunction is quite common,
and can be found in a number of inorganic optoelectronic devices such as quantum
cascade lasers [2] due to the confinement of charge carriers to small regions which
enhance electro-optical interactions. Another type of heterojunction is the Type-II
(staggered) heterojunction [Fig. 4-1(b)]. In this case, the bandgaps of the two mate-
rials overlap and disrupt the symmetrical carrier transport across the interface. For
instance, electrons moving from material 1 will find it energetically favorable to hop
into material 2, whereas holes will find a large potential barrier to entering material
2. Type-II heterojunctions form the basis for the most efficient OLEDs and organic
photovoltaics and are what helped to fuel the developments in the nascent field of
organic optoelectronics back in the 1980’s. The third and final type of heterojunction
is the Type-III (broken-gap) heterojunction, where the bandgaps do not overlap at
all. The situation for carrier transfer from material 1 to material 2 is similar to the
Type-II case, but more pronounced.

112
4.1.1 Excitons at Heterojunction Interfaces

Aside from playing a direct role in carrier blocking and injection, heterojunctions also
affect the exciton dynamics in devices. As discussed in a previous chapter, the weak
intermolecular forces in organic thin films are responsible for the formation of excited
states consisting of bound electron-hole pairs, which can be treated as quasi-particles
known as excitons. The formation and dissociation of excitons is largely a function
of the potential energy landscape in the materials and is accentuated by the abrupt
changes in energy present at a heterointerface.

An example of the utility of a heterointerface in an organic optoelectronic device

is the role it plays in OLED technology. High efficiency OLEDs require the balanced
injection of electrons and holes within the active organic layer [Fig. 4-2] and confine-
ment of the charge to the proper luminescent molecule, thus restricting non-radiative
decay pathways and controlling the spectrum of luminescence. Both requirements can
be met through the use of a Type-II heterojunction between an electron transporting
layer (ETL) and a hole transporting layer (HTL). Ideally, the ETL has a LUMO level
near the work function of the cathode to facilitate injection and transport of electrons
while the HTL has a HOMO level near the work function of the anode to facilitate
injection and transport of holes. Under a forward bias, VF , electrons are injected into
the ETL from the cathode (metal) and get blocked at the interface by the HTL, while
holes are injected into the HTL from the anode (ITO) and get blocked at the inter-
face by the ETL. Thus, the energy barriers at the interface act to spatially confine
the injected carriers near the interface, which enhances the probability of radiative
recombination. In addition, the barriers also prevent recombination from occurring
too close to the device contacts, which act as exciton quenching centers.

The energy level alignment at a heterojunction also plays a critical role in the
operation of organic photodetectors and solar cells. Power conversion efficiency in
photovoltaics is a figure of merit that measures a device’s ability to convert inci-
dent light into electrical energy collected at the contacts [Fig. 4-3]. Incident light
is absorbed by the organic materials to create excitons, which can diffuse with some

113
Figure 4-2: (a) Cross section and (b) energy band diagram of a typical OLED device
consisting of a transparent indium tin oxide (ITO) anode, a hole transport layer
(HTL), an electron transport layer (ETL), and a metal cathode. q is the charge of
an electron and VF is the forward voltage bias.

probability to the interface. Once there, the band offsets make it energetically fa-
vorable for the excitons to dissociate, transferring a hole (in the case of a donor
exciton) from the donor material to the acceptor material. Excitons created in the
acceptor material will transfer an electron to the donor material. Exciton dissociation
approaches 100% efficiency in most Type-II heterojunctions if the exciton binding en-
ergy is less than the LUMO (in the case of an acceptor exciton) or the HOMO (in the
case of a donor exciton) offsets [3]. The band offsets present at the interface enhances
the electric-field dependent Onsager dissociation in these devices [4], boosting device
efficiency.

4.1.2 Energy Band Alignment

One area of active research is the energy band alignment of dissimilar materials at
an organic-organic heterointerface, where the factors determining a common vacuum
level and the amount of band bending in organic materials are not well understood.
Ultraviolet photoemission spectroscopy (UPS) is one experimental technique used to
probe the energy levels of organic materials. The UPS measurement takes advantage
of the photoelectric effect and is performed by irradiating a sample film, in vacuum,
with high energy monochromatic light while measuring the energy distribution of elec-

114
Figure 4-3: (a) Cross section and (b) energy band diagram of a typical photovoltaic
device consisting of a transparent indium tin oxide (ITO) anode, an acceptor material
(A), a donor material (D), and a metal cathode. q is the charge of an electron and
VR is the reverse voltage bias.

trons emitted from the sample. With this technique the ionization energy or HOMO
level can be determined. Ionization energy, I, and electron affinity, A, are defined
as the difference in energy between the vacuum level and the HOMO and LUMO,
respectively. The work function, ΦF , is defined as the energy separation between the
Fermi level and the vacuum level [Fig. 4-4]. Because the electrons of a molecular
solid fill its energy levels according to Fermi statistics, the concept of a Fermi level is
still valid [5]. In principle, the value of the LUMO level can be determined through
inverse photoemission spectroscopy (IPES), but this measurement often causes radia-
tion damage to the sample [6]. Thus, the LUMO level is typically estimated from the
value of the HOMO level minus the optical bandgap1 (Eg ) of the material, which can
be determined through absorption measurements. Because all of the relevant energy
levels are referenced to the vacuum level, it is important to consider how to define
this quantity.

1
The optical bandgap is in general different from the actual energy required to form an exciton,
due to the Coulombic stabilization energy between the electron-hole pair [5] and the polarization
energies of the electron and the hole. Only when the Coulombic stabilization energy is equal to the
sum of the polarization energies does the optical bandgap become equal to the true bandgap.

115
Figure 4-4: Schematic of relevant energy band parameters: ionization energy (I),
electron affinity (A), and work function (ΦF ). After [7].

Defining the Vacuum Level

Theoretically, the vacuum level for an electron usually refers to the reference energy
where an isolated electron is at rest an infinite distance from the system of interest.
For convenience, we can denote this energy as VL(∞). Experimentally, the vacuum
level of a solid corresponds to the energy of an electron at rest just outside of the
solid, VL(s). Therefore, the experimentally determined VL(s) cannot be used as the
invariant reference level VL(∞). More concretely, it has been demonstrated that the
work function on the surface of a single crystal of tungsten is dependent upon the
crystal orientation of the surface itself [8]. Because the Fermi level is a common level
inside the solid material, it stands to reason that the difference in measured work
function is due to the differences in the vacuum levels of the surfaces.

For a metal, the difference between the energies of the VL(∞) and VL(s) is mostly
due to the surface dipole layer formed by the tail of the electron density distribution
at the surfaces [Fig. 4-5(a)]. The extension of the electron cloud into vacuum makes
the vacuum side negative, while just inside the surface becomes positive. From the
perspective of an electron a small distance x from a dipole layer of finite length L
(x ≪ L), the dipole layer acts like a plane of charge approaching infinite length. In

116
Figure 4-5: (a) Electron density in metal with tailing at the surface to form a dipole
layer. (b) Potential energy [V(x)] plot vs. distance of an electron a distance x from
a dipole layer. Fermi level (EF ) is uniform throughout solid. Φm 1 and Φm 2 are work
functions of surfaces with different crystal orientation. After [7]

117
this case, the potential energy forms a step function across the dipole layer and V(x)
on each side becomes a constant value independent of x. When x ≫ L, the dipole
layer can be regarded as a point dipole where the potential V(x) decreases as x-2 [Fig.
4-5(b)]. The contribution of the surface dipole layer makes the potential energy for
an electron in and out of a metal take the shape shown in Figure 4-5(a). As the
distance from the surface becomes large (x ≫ L), the effect of the surface dipole layer
diminishes, and the energy of the free electron converges to a common value, VL(∞).
Note that the dependence of the work function on the surface crystal orientation can
be attributed to the differences in the electron cloud distributions at different surfaces
[7].

Extending this idea of vacuum level to the interface between a metal and an or-
ganic film can lead to insights for organic-organic interfaces. When a metal and an
organic solid are far away, their energy levels are aligned so that they share a common
VL(∞) [Fig. 4-6(a)]. When the solids come into contact, neglecting rearrangement
of charge, the energy levels merge to form a common vacuum level at the interface,
(VL), in an extremely narrow interfacial gap [Fig. 4-6(b)/(c)]. The vacuum level at
the interface is a hypothetical concept useful for discussion of interface dipoles and
may actually be lower than the VL(s) immediately outside of the interface, due to the
merging of potential wells. In the actual system, a dipole layer may be formed at the
interface due to a number of processes (i.e. charge transfer across the interface, redis-
tribution of the electron cloud, chemical reaction at the interface, etc.) and recognize
that this is an additional dipole to the dipole layer of the free metal surface. The
interfacial dipole formation causes an abrupt shift of the potential across the interface
[Fig. 4-6(d)], leading to a shift (∆) of the virtual vacuum level at the interface. The
value of ∆ is determined by the magnitude of the dipole. A positive charge in the
organic material causes a downward shift of VL in the organic layer relative to the
VL in the metal layer, making delta negative. This follows the convention in the field
of surface science, where delta is positive when VL is raised by deposition [9].

The origin of the interface dipole layer is not well understood, and has been
theorized to form as a result of a number of different mechanisms. Charge transfer

118
Figure 4-6: (a) Potential surface of a metal and organic which are not in contact. (b)
Contact of the metal and a thin film of organic causes the common potential barrier
to drop. (c) Schematic representing common VL at interface. ΦB n /ΦB p denote the
injection barriers for electrons/holes. (d) Energy diagram showing shift of VL (∆)
due to dipole formation at the interface [7].

119
is widely invoked to explain VL shift. However, there have been reported cases of
VL shifts even when charge transfer is not possible (for example, using ultra-wide
bandgap organic materials) [10, 11].
While there can be large shifts in the VL between a metal and organic (∆ ≈ -1
eV for Alq3 on Au), the largest reported shift between two organic materials reaches
∆ ≈ -0.5eV for PCTDA on Alq3 [12]. One explanation for this trend is that charge
transfer is the only available mechanism for formation of an interface dipole due
largely to the lack of free carriers, the nonpolar nature of organic molecules, and
the lack of chemical interactions between organics. Thus, the only combinations of
organic materials expected to have large deltas are strong electron donor and acceptor
pairs such as PTCDA/Alq3 .

Band Bending and Fermi Level Alignment in Organics

In general, the work functions for the two different organic layers are different, which
raises two important questions: 1) Does the Fermi level align in organic semiconduc-
tors? 2) Is there band bending in organic semiconductors? Fermi level alignment
will only occur if the total number of available mobile carriers in the organic layer is
sufficiently large so that charge redistribution around the interface can occur within
the experimental time window. This is the case when either the organic layer is thick
or the organic layer has good semiconductor properties. The number of intrinsic car-
riers is not expected to be large in purified organic materials under ultrahigh vacuum
(UHV), since the bandgaps are typically much larger than the thermal energy at room
temperature. Thus, for the very thin layers of materials for UPS measurements, it is
a good approximation to use a flat-band description of the organic films at the inter-
face, and unaligned Fermi levels. However, recent experimental work (UPS studies)
on a number of organic/organic heterojunction interfaces yields a different insight
[13]. Tang et. al. theorize and demonstrate that Fermi level alignment can occur in
thermal equilibrium conditions, and that the determining factor is the alignment of
the Fermi level in the underlying film with either the HOMO or LUMO of the top
film.

120
Figure 4-7: Schematic diagrams of organic-organic heterojunctions, where ΦF is work
function, EF is Fermi level, A is electron affinity, I is ionization potential, and Eg is
the bandgap. Subscripts denotes which material the properties belong to. (a) depicts
a situation where ΦFI <AII before and after contact. (b) depicts ΦFI >III before and
after contact. (c) depicts ΦFI >AII before and after contact. After [13].

121
 Kelvin probe microscopy work done by Ishii et. al. [14] demonstrated Fermi level
pinning in the bulk of organics deposited on different metal substrates. In other words,
the Fermi levels measured in the organics were mostly independent of the thickness
of the film (for thicknesses >10 nm) and the energy of the Fermi level depended on
the underlying metal substrate, suggesting that the Fermi level in the organic film
on a metal is not the same as it is in the bulk of an isolated organic film. Tang
noted that aside from the abrupt vacuum level shift at the metal-organic interface,
the Fermi level in the first organic film (Org I) is determined solely by the substrate
it is deposited on. Thus, if a second organic film (Org II) is deposited on top of the
first film and the pinned Fermi level of Org I is either above the LUMO or below the
HOMO level of Org II, charge redistribution will take place, causing significant band
bending to occur as well as an abrupt vacuum level shift at the interface. Charge
redistribution is expected to occur in this case because there are a sufficient number
of states (due to the tail ends of the density of states in the HOMO and LUMO) that
extend into the gap near the Fermi level. On the other hand, if the pinned Fermi
level of Org I is near the middle of the energy gap of Org II, then little to no charge
redistribution takes place because of the negligible density of states in the middle of
the Org II’s bandgap. In this case, the traditional picture of vacuum level alignment
correctly predicts the band offsets [Fig. 4-7].

4.2 Photoconductor Device Theory

A photoconductor is a two-terminal device that changes its conductivity upon illumi-

nation. Typically, interdigitated electrodes are used to maximize both the semicon-
ductor area exposed to the light source and the device bandwidth [15]. As photons are
absorbed by the semiconductor, excitons are generated in proportion to the intensity
of the incident photon flux. Those excitons can then diffuse with some probability to
the surfaces or bulk defects where they can dissociate into free charge carriers in the
film. When a bias is applied across the electrodes, an electric field develops across the
photoconductive film which sweeps the charges out as photocurrent. In crystalline

122
inorganic semiconductors, the Wannier-Mott exciton binding energy is small enough
(∼ 0.1 eV) that excitons readily separate into electrons and holes. However, in organic
semiconductors, the photons will generate strongly bound (1 eV) Frenkel-type singlet
excitons that are difficult to dissociate without the presence of carriers, dopants, or
charged species at interfaces [16]. Much of the photoconductive phenomena in a de-
vice can be described by a set of basic processes. In organic photoconductors these
processes include 1) optical absorption by which excitons are generated, 2) exciton
dissociation to generate free carriers, 3) electrical transport by which the free carriers
contribute to electrical current, and 4) the capture of free carriers leading either to
recombination or trapping.

4.2.1 Carrier Generation

For simplicity, we can model the photoconductor as a thin film with thickness L
[Fig. 4-8]. Illumination occurs from one side with intensity Io (photons/second).
Throughout the film the density of photogenerated excitons is ε(x), and the excitons
can diffuse in either direction (positive or negative x-direction) based on the gradient
of the exciton concentration. We can assume that the excitons will recombine in the
bulk of the film due to fluorescence or non-radiative decay at a rate represented by the
exciton lifetime, τex . Additionally, excitons are continuously being generated in the
film at a rate that is proportional to the intensity of illumination at that point. The
dε(x)
general rate equation describing the change in exciton density, dt , in an interval
dx will be equal to the number of photogenerated excitons (G) minus those lost by
diffusion (δ) and spontaneous decay (R)

dε(x)
= G − δ − R. (4.2.1)
dt

Optical absorption can be quantitatively described by an absorption coefficient, α,

which is wavelength dependent. Neglecting reflection and scattering effects, if light
of intensity Io is incident on a material of thickness L with absorption coefficient α,
then the intensity of the transmitted light I is given by the empirical relation known

123
as Beer’s Law,
I(x) = Io e−αL . (4.2.2)

In a region dx, that is a distance of x away from the illuminated surface, the rate of
exciton generation is of the form G = Ae−αx . To find the coefficient A we can imagine
creating the same number of excitons as incident photons for an infinitely thick film.
∞
This leads to the relation Io = ∫0 Ae−αx dx = A
α so that A = Io α [17].

The carrier pairs that leave the region by diffusion can be represented by the
relation δ = dξ/dx, where ξ is the exciton flux. Exciton flux can be described by the
following equation
dε
ξ = −D . (4.2.3)
dx
where D is the exciton diffusion coefficient. The negative sign here implies that
excitons will flow from areas of higher to lower concentration. Finally, the rate of
recombination (encompassing all exciton loss mechanisms) can be represented by
R = ε/τex , where τex is the exciton lifetime in the bulk of the film. This relation
holds assuming that we can neglect multiple-particle effects (i.e. exciton annihilation,
exciton-polaron quenching) which occur at very high light intensities. Thus,

dε(x) d2 ε(x) ε(x)

= αηP L Io e−αx + D − , (4.2.4)
dt dx2 τex

where ηP L is the photoluminescent efficiency (the efficiency of generating excitons

from each incident photon). Under steady-state conditions dε (x)/dt = 0, which gives
the general solution

x/ −x/ αηP L Io τex −αx

ε (x) = C1 e LD + C2 e LD + e , (4.2.5)
1 − α 2 LD 2
√
where LD = Dτex is the exciton diffusion length. Surface-enhanced exciton dissoci-
ation models are based on the assumption that the absorption of a photon creates an
exciton that diffuses to a surface where it dissociates into a free electron-hole pair, or
a free and deeply trapped carrier of opposite sign, through a donor-acceptor interac-
tion associated with the surface [18]. Assuming a dopant and trap-free film, exciton

124
dissociation should only occur at the surfaces. The exciton flux, ξ, at the surfaces
where dissociation occurs is important for determining the boundary conditions. As-
suming that the dissociation rate at each surface can be represented by an exciton
flux into the surface proportional to the exciton density at the surface

dε (0)
x = 0, −ε (0) S0 = −D
dx
dε (L)
x = L, ε (L) SL = −D
dx

where S0 and SL are the surface dissociation velocities for both surfaces and act as
proportionality constants with units of velocity. Substituting these boundary condi-
tions into Equation (4.2.5) above and solving for the constants C1 and C2 yields,

γ (SL − αD) (S0 + Dβ) e−αL + (S0 + αD) (Dβ − SL ) e−βL

C1 =
β 2 − α2 (SL − Dβ) (S0 − Dβ) e−βL − (SL + Dβ) (S0 + Dβ) eβL
[ ]

γ (SL − αD) (Dβ − S0 ) e−αL + (S0 + αD) (SL + Dβ) eβL

C2 =
β 2 − α2 (SL − Dβ) (S0 − Dβ) e−βL − (SL + Dβ) (S0 + Dβ) eβL
[ ]

1 αηP L Io
where β = LD and γ = D . Examples of exciton concentration profiles for a single
film using common organic material parameters and fixed illumination intensity are
plotted in [Fig. 4-8]. Note that at any given point x, the exciton concentration has a
linear dependence on the light intensity, Io .

We can use a similar rate equation to calculate the free carrier profiles in the film
assuming that carrier generation does not occur in the bulk of the film, but rather
at the surfaces where one type of carrier gets trapped (for example, holes) and the
other goes free (for example, electrons) [20]

dn (x) d2 n (x) n (x)

= Dn − . (4.2.6)
dt dx2 τn

Here n is the electron concentration, Dn is the electron diffusion coefficient, and τn is

the lifetime of an electron. Under steady-state conditions dn (x)/dt = 0, which gives
the general solution
x/ −x/
n (x) = C1 e Ln + C2 e Ln , (4.2.7)

125
Figure 4-8: Simulated exciton carrier concentration profile across a thin film of organic
semiconductor with L = 50 nm, α = 5x105 cm-1 , D = 1x10-3 cm2 s-1 [19], 100 mW/cm2
of 2.25 eV photons, and various boundary conditions. SL =S0 =0 (violet), SL =S0 =1020
(green), SL =0 and S0 =1020 (blue), SL =1020 and S0 =0 (red).

where Ln is the electron diffusion length. If we can assume that one of the surfaces
(at x = 0) is responsible for carrier generation (due to unsatisfied bonds) and the
other surface (at x = L) is passivated and has a surface recombination velocity of 0,
then we get the following boundary conditions

n (0) = n0
dn (L)
n (L) SL = 0 = −Dn
dx

Solving Equation (4.2.7) with these boundary conditions yields

−L/
n0 e Ln
C1 =
L/ −L/
e Ln + e Ln
L/
n0 e Ln
C2 =
L/ −L/
e Ln + e Ln

126
To find the value of n0 we can use the following electron continuity equation

dn (0) dε (0) d2 n (0)

=D − Dn , (4.2.8)
dt dx dx2

where the first term represents the generation rate of electrons at the surface (flux of
excitons into surface) assuming every exciton generates an electron and the second
term represents the electron diffusion current away from the surface.

4.2.2 Carrier Transport

Under steady-state illumination the rate of charge recombination (n/τn ) must equal
the rate of free carrier generation, Gc , giving the basic relationship that n = Gc τn . If
more than one charge recombination process is present, the individual rates add. As
carriers are generated in the photoconductive film, they drift towards the appropri-
ate electrode given the polarity of the applied field. Those carriers are transported
through the film by hopping from one molecule to the next and only stopping at im-
perfections in the film. A captured carrier at an imperfection can do one of two things;
1) recombine with a carrier of the opposite type or 2) be thermally re-excited to the
nearest energy band before recombination occurs (in which case the imperfection is
a trap). The rate of capture of a free carrier concentration n by imperfections with
density N is given by Rc = βnN , where β is the capture coefficient. The thermal de-
trapping rate is given by Rd = nt νe−∆E/kt , where nt is the density of trapped carriers,
ν is a characteristic attempt to escape frequency, and ∆E is the activation energy for
detrapping. If Rc > Rd then the imperfection acts like a recombination center or like
a trap if Rc < Rd . The process of electrical current flow under illumination involves
two processes: the effect of contacts and the nature of the transport of free carriers.
The ability of the contacts to replenish carriers to maintain charge neutrality in the
material, if carriers are drawn out of the opposite contact by an electric field, plays a
key role in determining the photocurrent characteristics. Electrical contacts can be
organized into three classes of behavior: 1) blocking contacts, 2) ohmic contacts, and
3) injecting contacts. The characteristic current-voltage curve of each type of contact

127
Figure 4-9: Representative current-voltage characteristics for a blocking contact
(green), an ohmic contact (red), and an injecting contact (blue).

is shown in [Fig. 4-9]. If a contact is unable to replenish photogenerated carriers when

they are leaving the device, then the current varies sublinearly with applied voltage,
and may become saturated. An ohmic contact is able to replenish photogenerated
carriers that exit the device, such that current varies linearly with applied voltage
(with a slope equal to the conductivity of the bulk material). An injecting contact is
an ohmic contact at low applied fields, but under sufficiently high fields allows carri-
ers to traverse the distance between electrodes before the material is able to achieve
charge neutrality by dielectric relaxation. The result is a regime in which the material
is not electrically neutral, and the current flow becomes limited by the build-up of
space charge at the injecting contact, hence the name “space-charge-limited” (SCL)
current.

The charge transport process is described in terms of drift-diffusion theory. The

total current density Jtot can be expressed as the sum of a drift current Jdr and
a diffusion current Jdif f . The drift current is given by Jdr = σE, where σ is the
electrical conductivity, and the diffusion current (for electrons) is expressed as Jdif f =
qDn dn/dx. Neglecting space charge effects in the lateral photoconductor geometry

128
implies that the carrier profiles are uniform throughout the channel length, indicating
that Jdif f = 0. One distinction to be made here is that Jlight = Jphoto + Jdark . In other
words, the current under illumination is the sum of the photogenerated current plus
the current that flows under dark conditions.
Consider the effects on the light current to dark current ratio when blocking
contacts are used. Jdark will decrease with the increased injection barriers from the
contacts to the semiconducting film. However, there may be negative effects on Jphoto
as well that may reduce the benefits of decreased carrier injection. For instance, if the
photoexcited carriers raise the effective conductivity of the film during their lifetime,
then there is the opportunity for injected charge to travel multiple times through the
film before the excited carriers relax. A prerequisite for this amplifying behavior is the
ability for charge to be injected to replace carriers that are exiting the device. Thus,
blocking contacts would prevent amplification by preventing the injection of charge
that effectively allows for multiple transits through the film. If Jphoto is strongly
dependent on the photoconductive gain in the device, then blocking contacts may
reduce the light to dark current ratio by preventing amplification. However, in cases
where gain is not a factor, blocking contacts should improve the light to dark current
ratio by suppressing the injection of carriers without illumination.

4.2.3 Figures of Merit

There are three main features of a photoconductor: photosensitivity, the spectral

response, and the speed of response.

Photosensitivity

In a material with intrinsic free carriers, the electrical conductivity in the dark can
be given by
σd = q (no µno + po µpo ) (4.2.9)

where q is the charge of the carrier, no and po are the intrinsic carrier concentrations,
and µno and µpo are the effective mobilities. Under illumination, this conductivity is

129
increased by the photoconductivity ∆σ,

σL = σo + ∆σ. (4.2.10)

The possibility that photoexcitation may change both the carrier density and carrier
mobility yields the following set of changes:

no → no + ∆n

po → po + ∆p

µno → µno + ∆µn

µpo → µpo + ∆µp

σo + ∆σ = q[(no + ∆n)(µno + ∆µn ) + (po + ∆p)(µpo + ∆µp )].

Therefore,

∆σ = q[∆nµno + (no + ∆n)∆µn + ∆pµpo + (po + ∆p)∆µp ]. (4.2.11)

Under steady state illumination, we know that the carrier generation rate equals the
recombination rate
∆n = Gc τn = ∆p, (4.2.12)

where Gc is the carrier generation rate [m-3 s-1 ] (which includes exciton formation and
dissociation) and τn is the electron lifetime. Thus,

∆σ = q [Gc τn µno + n∆µn + Gc τp µpo + p∆µp ] . (4.2.13)

To add to the complexity, the lifetime may be a function of the excitation rate. Thus,
there are three types of effects that can occur:

1. Increase in the carrier density with constant lifetime and no change in mobility,
such that
∆σ = qGc (τn µno + τp µpo ) . (4.2.14)

130
 Plotting photoconductivity on a log-log scale versus G would show a slope of 1.

2. Increase in carrier density with lifetime a function of intensity of illumination. If

we assume τn varies as Gγ−1 , then ∆σ varies as Gγ . Thus, if γ < 1, the lifetime
decreases with increasing excitation rate, which yields sublinear behavior. If
γ > 1, the lifetime increases with the excitation rate which is called supralinear
behavior.

3. Increase in carrier mobility, so that

∆σ = q (n∆µn + p∆µp ) . (4.2.15)

One likely mechanism can give rise to this phenomenon: photoexcitation that
excites carriers to fill in traps, raising the effective mobility of the material. In
addition, mobilities have been shown to be dependent on carrier concentrations
[21, 22].

To first order approximation, given sufficiently high photoexcitation rates, constant

mobility, and the fact that organic materials are insulators, ∆n ≫ no and ∆p ≫ po ,

∆σ ≈ Gc q (τn µn + τp µp ) . (4.2.16)

This equation can be rearranged to give the ratio of change in conductivity to excita-
tion rate, which is a figure of merit (lifetime-mobility product) for a photoconductor’s
photosensitivity,
∆σ
= τ n µn + τ p µp . (4.2.17)
Gc q

Photoconductive Gain

Under steady state conditions, the photoconductive gain represents the ratio of the
photogenerated charge carrier lifetime to the transit time across the device of charge
carriers injected at the contacts. In the steady state, the total carrier generation rate
must be equal to the total carrier recombination rate, which can be written as the

131
following relation:
n
L = ηP L Io (4.2.18)
τn
where n is the charge carrier number density, τn is the carrier lifetime, L is the device
thickness, ηP L is the photoluminescence efficiency of the active material, and Io is
the incident photon flux. Assuming uniform charge density in the lateral direction
along current flow, device photocurrent can be modeled by the one dimensional drift
current density equation:
Jphoto = qn (µn + µp ) E (4.2.19)

where electron and hole densities are equal (n = p), µn and µp are the electron and
hole mobilities, respectively, and E the applied electric field. Substituting the steady
state condition relation and solving for the total current density yields:

d 1 1
Jphoto = qnΦ × τn ( + ) (4.2.20)
L tn tp

where d is the device channel length and tn and tp are the electron and hole transit
times respectively. The photocurrent gain increases when either of the carrier transit
times decreases or if the carrier lifetime increases. When photogenerated carriers are
created through the dissociation of an exciton, the carriers will hop among molecular
states in response to the applied electric field. In amorphous organic thin films it has
been shown that energetic and spatial disorder [23] can lead to a significant population
of energetic traps, molecular states that limit carriers from hopping to another site
[24]. As the energetic traps are filled by the excess photogenerated carriers, the film
will have a higher effective mobility while the trapped charge persists in this non-
equilibrium state. The time it takes for trapped charge to relax is dependent on the
disorder in the film, so much so that the carrier transit time can be shorter than the
carrier lifetime in strongly disordered materials. Thus, for every photoexcited carrier,
many more carriers can be transported across the film before the photoexcited carriers
relax. This process leads to photoconductive gains above unity.

132
Speed of response

In a trap-free material, the free carrier lifetime determines the time constant asso-
ciated with the decay curve (decrease of photoconductivity to the dark value after
turning off illumination). The governing rate equation assuming n ≈ ∆n is

dn n
= Gc − . (4.2.21)
dt τn

Integrating this equation with Gc = 0 at time t = 0 yields,

−t/
n (t) = Gc τn e τn . (4.2.22)

Given these simplified conditions, we observe that the photocurrent decay follows a
single exponential. Thus, plotting current versus time on a semi-logarithmic scale
would yield a slope equal to the decay time constant. The capture process of a free
carrier at an imperfection can be described through a capture coefficient, C [m3 s-1 ],
which can be thought of as the product of a capture cross section U and the average
thermal velocity of the free carrier, v,

C = Uv (4.2.23)

where the magnitude of C depends on the details of the capture process. The rate of
capture Rc [m-3 s-1 ] of a species with density n by another species with density N is
given by
Rc = CnN. (4.2.24)

When the capture of free carriers leads to recombination with the opposite species, the
lifetime of a free carrier, τn , (the average time the carrier is free before recombining),
is given by the density of carriers divided by their rate of capture

τn = 1/CN. (4.2.25)

133
Comparison with Equation (4.2.24) shows that the rate of capture (in this case re-
combination) is equal to n/τn . In a generalized photoconductor model, the carrier
generation rate Gc and thermal excitation rate g are balanced by the carrier recom-
bination across the bandgap with a rate Rc , such that with illumination,

2
Gc + g = C (no + ∆n) (po + ∆p) = C(no + ∆n) = Rc (4.2.26)

since ∆n = ∆p. With no illumination

g = Cno po = Cn2o (4.2.27)

since no = po . For insulating organic materials, g ≪ Gc , no ≪ ∆n, so that Equation

(4.2.26) becomes
Gc ≈ C∆n2 . (4.2.28)

Therefore,

1/2
∆n = (Gc /C) (4.2.29)
1 1
τn = =√ . (4.2.30)
C∆n CGc

In this case ∆n ∝ Gc 1/2 , which indicates a case of sublinear photoconductivity. In

other words, higher intensity light, which yields higher exciton generation rates and
subsequently higher carrier generation rates, will have less and less of an effect on
the photocurrent. Correspondingly, the carrier lifetime decreases with increasing
intensity, τn ∝ Gc −1/2 . Cases like this are dominated by bimolecular recombination.

4.3 Bilayer Heterojunction Photoconductor

In a typical organic photodetector, organic thin films absorb the incident light to form
excitons (bound electron-hole pairs) that can dissociate into charge carriers that are
extracted by applying an external electric field. The vertical photodetector device ge-

134
ometries task the photoactive films with the dual role of exciton photogeneration and
charge transport. The coupling of these two dissimilar roles in a single organic film
constrains the design of photodetectors, requiring use of highly absorptive films that
can simultaneously efficiently transport the photogenerated charge to the electrodes.
Typically, amorphous organic thin films, while having desirable optical absorption
characteristics, have low charge carrier mobilities due to the weak electrostatic inter-
actions between van-der-Waals bonded molecules [16]. The key to overcoming this
fundamental limitation is to recognize that the coupled functionality is due largely to
the device architecture. Thus, by introducing a heterojunction into a lateral device
geometry we physically separate the exciton generation from the charge transport,
enabling independent optimization of the two processes in an amplifying photocon-
ductor structure.
The two-terminal, lateral photoconductor incorporates a heterojunction between
an optically absorptive thin film that is primarily responsible for exciton generation,
and another film that is (preferably) optically transparent at the operating wavelength
and is primarily responsible for charge transport. The dual-layer structure has its
roots in xerography, where photoreceptor drums contain a charge generation layer
for absorbing the light from the image exposure and forming free carriers, and a
charge transport layer responsible for transferring those carriers to the corona-charged
surface, where they recombine to form the electrostatic latent image [18]. By adding
the lateral, planar electrodes, we extend the dual-layer photoreceptor paradigm and
demonstrate photoactive device structures that also enable direct study of the charge
generation and recombination rates at the bilayer heterojunction interface.

4.3.1 Device Model

The lateral bilayer heterojunction photoconductors developed in this model consist

of two semiconducting thin films deposited onto interdigitated gold electrodes [Fig.
4-10(a)]. The rest of the figure illustrates two different views of the physical processes
that govern operation of the photoconductors. Figure 4-10(b) shows the proposed en-
ergy band diagram indicating the highest occupied molecular orbital (HOMO) and

135
the lowest unoccupied molecular orbital (LUMO) for the two semiconducting organic
layers, while Figure 4-10(c) depicts the cross section of the proposed device stack. As
shown, the exciton generation layer (EGL) is deposited on top of the charge transport
layer (CTL), which is in intimate contact with the gold electrodes. Under steady-state
bias conditions, conduction can occur through either the CTL or the EGL. With the
proper choice of materials, conduction can predominantly occur in the channel formed
between the CTL/EGL [20]. Under illumination, light with an appropriate emission
spectrum will be absorbed by the EGL, resulting in the formation of EGL excitons. A
fraction of the photogenerated excitons will diffuse to the EGL/CTL interface. If the
EGL and CTL are aligned such that their offsets in either the LUMOs or HOMOs are
greater than the binding energy of EGL excitons [3], as shown in Figure 4-10(b), then
it is energetically favorable for excitons to dissociate at the interface. Exciton disso-
ciation in disordered organic materials has been attributed to both surface and bulk
effects [16], with appropriately designed Type-II heterojunction interfaces achieving
near complete exciton dissociation. Once created, the photogenerated charge will con-
tinue to accumulate at the EGL/CTL interface until equilibrium is achieved between
carrier generation, charge recombination (both Langevin and Shockley-Read-Hall),
and charge diffusion away from the interface. In these structures, the accumulated,
interfacial holes in the CTL will balance the electron charge accumulated in the EGL.
The conductivity of the CTL is enhanced by the population of photogenerated charge
carriers, which we can measure as an increase in photocurrent from an illuminated,
biased device.

To model the carrier concentrations in a device, we can build upon the rate equa-
tions used for a single photoconductive film derived earlier. Given that the hetero-
junction is deposited on glass, the primary surface that is expected to dissociate
excitons is the EGL/CTL interface itself. Assuming the EGL preferentially absorbs
the illumination, the exciton concentration profile for a bilayer heterojunction would
have an exciton quenching surface (S → ∞) at the interface and an exciton reflecting
surface (S = 0) at the interface with air. This would yield an exciton profile that
resembles the blue curve in Figure 4-8 assuming the EGL/CTL interface is located

136
Figure 4-10: The interdigitated gold fingers form 60 photoconductive channels each
(1000 ± 1) µm long and (10 ± 1) µm wide yielding an effective device area of (0.6
± 0.06) mm2 . (a) An optical microscope image of a set of device electrodes. (b)
Energy band diagram of a lateral bi-layer heterojunction photoconductor consisting
of an exciton generation layer (EGL) and a charge transport layer (CTL). (c) A
cross-sectional view of the same bi-layer device is also shown. Both illustrations
depict the physical processes involved in steady-state device operation: (1) Light
absorption in EGL, (2) Exciton diffusion through EGL, (3) Exciton dissociation at
EGL/CTL interface and charge transfer to CTL, and (4) Charge transport in the
photoconductive channel.

137
at x = 0 and the EGL is to the right of the interface.
Again, the carrier profiles can be simulated using the previously derived equa-
tions for free carriers, however the boundary conditions need to be developed further.
Given the proposed band diagram, when excitons dissociate electrons will remain
in the EGL while holes are transferred to the CTL. Thus, only excess holes will be
considered for x < 0 and excess electrons for x > 0. When considering the boundary
conditions at the EGL/air and the CTL/glass interfaces, a surface recombination
velocity of 0 would imply reflecting surfaces. This assumption is plausible as long
as there are negligible deep trap states (i.e. unsatisfied bonds) and therefore few
recombination centers:

dp (−L) dn (L)
−p (−L) S−L = −Dp = 0 = −Dn = n (L) SL . (4.3.1)
dx dx

The boundary condition at the EGL/CTL interface can be determined by calculating

the carrier concentration at that surface. To do so requires further consideration of
the processes that generate free carriers.
Initially, exciton dissociation at the heterointerface produces a pair of charges
which are still coulombically bound (a charge-transfer state or polaron pair) [25].
Thus, the generation rate of polaron pairs (GP ) at the EGL/CTL interface is directly
proportional to the exciton flux there −D dε(0)
dx = GP . These CT states are intermediate

states between excitons and free carriers. The coulombic attraction between the
electron and hole in the CT state make it quasi-stable and capable of undergoing
multiple dissociations and subsequent formations over its lifetime [26]. The CT states
can either disappear by recombining monomolecularly (with a rate constant of krec )
[27] or can dissociate into free carriers with a field-dependent rate constant kdiss (E)
[26]. Additionally, CT states can be formed by Langevin-type (bimolecular capture)
recombination of free carriers (Λnp) [28]. In the case of a pristine, trap-free material,
the recombination constant is governed by the sum of the electron and hole mobilities
q
Λ =  (µn + µp ), where  is the dielectric constant of the material. Therefore, the

138
general rate equation for the polaron pairs (P (x)) may be expressed as

dP (0)
= GP − krec P (0) − kdiss P (0) + Λn (0) p (0) . (4.3.2)
dt

To derive a free carrier concentration profile, we can also express the free carrier
continuity equation (in this case electrons) as

dn (0) n (0) p (0)

= kdiss P (0) − Λn (0) p (0) − kSRH , (4.3.3)
dt n (0) + p (0)

where the last term refers to the Shockley-Read-Hall (SRH) form of recombination,
which is a trap-assisted recombination effect that is common in polymer interfaces
[29]. The SRH rate coefficient kSRH takes into account both the trap densities and
capture coefficients of the carriers, while n and p represent the free carrier densities
which are assumed to be much larger than the intrinsic carrier densities. However,
given that we are concerned primarily with purified organic materials we can neglect
this last term. Note that both Equation (4.3.2) and Equation (4.3.3) have units
associated with surface densities (particles per area). Combining Equations (4.3.1),
(4.3.2), (4.3.3) under steady-state conditions yields,
√
kdiss GP
n (0) = . (4.3.4)
krec Λ

Neglecting high intensity effects such as exciton-exciton annihilation and exciton-

polaron quenching, we found earlier that the generation rate of polaron pairs is linearly
dependent on intensity (GP ∝ αIo ). This implies that the carrier concentration at
the EGL/CTL interface will follow a sublinear dependence on intensity n (0) ∝ Io 1/2 .

139
4.4 Model System: TPD/PTCBI Bilayer Hetero-
junction Photoconductors

To test this device model we used two well characterized molecular organic mate-
rials: N,N’-bis(3-methylphenyl)-N,N’-diphenyl-1,1’-biphenyl-4,4’-diamine (TPD) as
CTL and 3,4,9,10-perylenetetracarboxylic bis-benzimidazole (PTCBI) as EGL, whose
absorption spectra and electron energy band levels are depicted in Figure 4-11.
TPD (obtained from H.W. Sands) is a common hole transport material (µp ∼ 10−3
cm2 V-1 s-1 ) used in organic LEDs and is optically transparent for wavelengths of
λ > 400 nm, identifying it as suitable for use as the CTL. PTCBI (obtained from
Sensient GmBh) is a perylene dye derivative that has found use as an optically ab-
sorptive material in organic solar cells [30] with strong absorption across the visible
spectrum, making it suitable for use as an EGL. In addition, the high ionization energy
of PTCBI makes it a suitable hole donor to TPD and creates large HOMO/LUMO
band offsets to aid in dissociation of excitons. From the absorption spectra in Figure
4-11, we note that PTCBI can be optically excited independent of TPD by using a
green LED centered at λ = 532 nm (Lamina Light Engine, BL-3000). The thermally
evaporated TPD/PTCBI films are deposited on a series of interdigitated gold elec-
trodes that are photolithographically defined on glass [Fig. 4-10(a)]. These electrodes
form a serpentine channel that is (60 ± 1) mm wide and (10 ± 1) µm across, resulting
in an effective device area of (0.6 ± 0.06) mm2 .
Thin film absorption measurements are captured using a UV-VIS-NIR spectropho-
tometer (NKD-8000, Aquila Instruments). The green LED light engine, powered
by a sourcemeter (Keithley 2400), provides narrow band excitation (FWHM ∼ 50
nm) with intensities up to 40 mW/cm2 as measured with a calibrated photodetector
(Newport, 818-UV). The emission spectrum of the green LED was measured using a
spectrometer (Ocean Optics, USB4000). Device current-voltage (I-V) characteristics
were measured in a dark box using a picoammeter (Keithley 6487). Photocurrent
spectra were measured using a digital lock-in amplifier (Stanford Research Systems,
SR830) with monochromatic illumination provided by a 1000 W Xenon arc lamp

140
Figure 4-11: Absorption spectra (50 nm thick films), chemical structure, and
proposed energy band levels of the materials used in the bi-layer heterojunction
photoconductor: N,N’-bis(3-methylphenyl)-N,N’-diphenyl-1,1’-biphenyl-4,4’-diamine
(TPD) as CTL and 3,4,9,10-perylenetetracarboxylic bis-benzimidazole (PTCBI) as
EGL. Also shown is the emission profile of the green LED (dotted line, λpeak = 532
nm) operated with a forward current of 2 A and an intensity of 40 mW/cm2 . Energy
band levels represent experimental values for the HOMO/LUMO based on photoemis-
sion spectroscopy data [6, 14]. As shown, the energy bands correspond to transport
levels that provide upper and lower limits for the optical bandgap and emphasize the
Type-II heterojunction between TPD/PTCBI.

141
(Oriel) passed through a monochromator (Princeton Instruments/Acton Spectrapro
300i). The optical chopping frequency was set at 40 Hz, giving the device enough time
to respond to the illumination. Photocurrent transient measurements were performed
with a standard voltage divider circuit, an oscilloscope (Tektronix 3500 Series), and
an arbitrary waveform generator (Agilent 3350A). All measurements were taken in
a nitrogen glove box (MBraun) environment to prevent device degradation during
measurements.

4.4.1 Current-Voltage Characteristics

Figure 4-12 plots the steady state I-V characteristics, both with and without LED
illumination, from a series of devices containing either single films or bilayer hetero-
junctions of PTCBI and TPD. The device that contains a single (50 ± 0.5) nm thick
film of TPD generates no measurable photoresponse, which reflects the inability of
TPD to absorb green (λ = 532 nm) light. The thin film of PTCBI does generate a
moderate photocurrent in response to the green light illumination, which can be seen
by comparing the solid blue line to the dashed blue line in Figure 4-12. Despite the
absence of a heterojunction, photogenerated excitons are still able to dissociate in
the single film (most likely at polycrystalline grain boundaries) of PTCBI contribut-
ing to the enhanced film conductance [31]. Comparing the dark currents of PTCBI
and TPD, it is evident that PTCBI has a lower overall dark conductivity. Given
that there should be little to no electron injection barrier from the gold contacts into
PTCBI [6], the decreased conductivity can largely be attributed to the low electron
mobility of PTCBI (∼ 10−6 cm2 V-1 s-1 ). The heterojunction devices show a relative
improvement in the photoresponse and a light to dark current ratio of over 100. The
lower conductivity exhibited by the PTCBI/TPD device (green curves) is most likely
due to the PTCBI film having a lower electron mobility than the hole mobility in
TPD [32]. This dataset suggests that the TPD/PTCBI heterojunction interface is
efficiently dissociating excitons and generating charge under illumination, which is
transported through the CTL. It is also worth mentioning that both heterojunction
devices demonstrated an increase (up to a factor of five) in dark currents (red and

142
green dashed lines) over the dark currents (black and blue dashed lines) of the under-
lying films alone. This may be evidence of the formation of an interface dipole due to
charge transfer at the interface, which forms a conductive channel in dark conditions.
The strong electron donor-acceptor nature of TPD/PTCBI makes this a plausible
scenario [7]. Fitting the curves on a log-log scale reveals a near-linear dependence
of current on electric field, which suggests that over the range of voltages tested,
the conduction remained ohmic and that gold does not form a blocking contact with
TPD.

By varying the thickness of the underlying CTL it is possible to show, with another
series of measurements, the formation of a more conductive channel that is primarily
concentrated at the heterointerface. Figure 4-13(a) shows a series of I-V curves that
were taken under green LED illumination. Despite illuminating the device through
the glass substrate and up through the TPD films, the negligible absorption coefficient
of TPD at λ = 532 nm [33] means that the intensity of light reaching the PTCBI
film will essentially remain unchanged from the 50 nm to 500 nm thick TPD/PTCBI
device. In addition, the interface exciton dissociation efficiency should not be affected
by the underlying CTL thickness, which suggests that the photogeneration rate of
carriers should remain the same for all of the devices in the series.

For clarity, the dark currents are not included as the noise floor of the measure-
ment obscures any observable trends. As the thickness of TPD is varied over the
range of 50 nm to 500 nm, a clear decrease in the light current occurs for thicker
and thicker CTLs. In order for the current levels of the devices to decrease in the
linear regime there must be either a decrease in the applied voltage or an increase in
the effective resistance of the device. Because the range of applied voltages is fixed
the effective resistances of the thicker devices must be increasing. The best way to
explain this phenomenon is to model the bilayer device as a simple resistor network
[Fig. 4-13(a)], where the bulk resistivities of TPD and PTCBI are much higher than
the resistivity of the interface due to charge transfer [20] and the large carrier concen-
tration. Knowing that the interface is the most conductive pathway implies that the
conduction in the bilayer devices will be limited by the transport of injected charge

143
Figure 4-12: Semi-logarithmic plot of the current-voltage characteristics from a series
of lateral photoconductor devices using TPD and PTCBI. Current-voltage sweeps
are taken in the dark (dashed) and under illumination (solid) through the glass sub-
strate from a 40 mW/cm2 green LED (λpeak ∼ 532 nm). The inset cross-sectional
diagrams describe the thicknesses and device structures. Both heterojunction devices
(Au/PTCBI/TPD [green] and Au/TPD/PTCBI [red]) exhibit steady-state photore-
sponse. The Au/TPD device (black) exhibits no photoresponse to the green LED,
consistent with its absorption spectrum. The Au/PTCBI device (blue) exhibits a
steady-state photoresponse with lower currents than heterojunction devices.

144
from the bottom contacts through the TPD into the conductive channel at the in-
terface. The bulk resistivity of TPD and injection cross section are independent of
thickness, thus we would expect a linear increase in resistance through the TPD film
with increasing thickness and a consequent inverse relationship with current. It is
not possible, however, to linearly extrapolate the resistance of the interface without
assuming linear control over thickness. The height of the gold contacts (70 ± 0.5 nm)
can disturb the linearity of resistance with film thickness because TPD films that are
below or near the thickness of the contacts may exhibit nonlinear growth character-
istics at the sidewalls of the contacts [Fig. 4-13(b)]. This has the potential to yield
much more conductive pathways through the sidewalls than would be expected from
the thickness of the TPD film. Electron microscopy of the cross section of the devices
could be used to ensure that the TPD films are thick enough to ensure conformal
growth characteristics.

Another effect to take note of is the transition of the 500 nm thick device (blue
curve) from ohmic conduction to space charge limited conduction (SCLC) (the dotted
lines in Figure 4-13 are guides for the eye). Space charge builds up in insulators
because of the typically slow dielectric relaxation times that cannot quickly screen
injected charge. When the transit time of injected carriers is greater than the dielectric
relaxation time, SCLC become apparent in the I-V curves and follows a V2 dependence

9 µV 2
J= . (4.4.1)
8 L3

The critical voltage onset of SCLC can be derived from the criterion discussed above,

L2 
ttr = ≥
µV σ
σL2
Vcrit = ,
µ

where L is the distance between electrodes,  is the dielectric constant of the insulator,
µ is the mobility of the insulator, and V is the voltage applied across the electrodes.
The assumptions here are that 1) there is only one type of charge carrier present

145
Figure 4-13: (a) Log-log plot of current-voltage characteristics for a series of
TPD/PTCBI devices under illumination. The dashed lines are provided to show the
transition between ohmic and SCLC in the blue curve (500 nm TPD/50 nm PTCBI).
Inset: a circuit model of conduction pathways in the bilayer devices, where the rela-
tive size of the resistor is indicated. (b) A schematic drawing depicting the non-linear
sidewall growth that can occur with thin films evaporated on bottom contacts.

in the insulator, 2)  and σ are constant throughout the film, and 3) the insulator
is trap-free. Thus, the transition from ohmic to SCLC in Figure 4-13(a) indicates
trap-free TPD/PTCBI bilayer devices.

4.4.2 Spectral Response

Wavelength-resolved photocurrent spectra from three lateral photoconductors, con-

taining films of PTCBI and TPD, are shown in Figure 4-14. The photoconductor
with a single (5 ± 0.5) nm thick film of TPD yields no measurable photocurrent and
is omitted from the Figure 4-14 plot. This is most likely due to the amorphous na-
ture of thermally evaporated TPD, which contains few centers or surfaces for exciton
dissociation. The spectral response of the photocurrent is measured with the lock-in
amplifier locked to the light chopping frequency, effectively filtering out the DC dark
current. All of the measured photocurrent curves spectrally match their respective de-

146
vice’s absorption spectra. Comparing the response of the photoconductor with a neat
film of PTCBI to the photoconductors containing the TPD/PTCBI heterojunction,
we note that the introduction of a heterojunction interface improves the photocurrent
over 1000-fold for all wavelengths, reaching an external quantum efficiency of (12 ±
1)% at λ = 552 nm. This suggests that excitons generated in either TPD or PTCBI
are efficiently dissociated by the Type-II heterojunction formed at the interface. Ac-
counting for the number of photons absorbed in our thin films at that wavelength, the
internal quantum efficiency for Au/TPD(5 ± 0.5 nm)/PTCBI(10 ± 0.5 nm) is (140 ±
2)%, when illuminated with λ = 552 nm wavelength light and operated at an applied
electric field of 1.25 x 105 V/cm. An internal quantum efficiency of greater than 100%
indicates an optical-to-electrical conversion gain in the device. Despite the order of
deposition of the constituent films, both heterojunction devices yield improvements
in photoresponse as compared with the single films. The PTCBI/TPD device’s lower
quantum efficiency (comparing the green curve to the red curve) can be attributed
to the low electron mobility in PTCBI.

4.4.3 Intensity Measurements

Figure 4-15 demonstrates the current as a function of voltage at different incident

light intensities for a 50 ± 0.5 nm TPD/50 ± 0.5 nm PTCBI device. At all intensities,
the device exhibits ohmic conduction over the range of voltages swept. At 1 V the
light current shows a sublinear dependence on intensity as evidenced by the inset of
Figure 4-15. The power law fit to the relative photocurrent vs. relative intensity
follows a near square-root dependence with an exponent of 0.56. As discussed earlier,
bimolecular recombination leads to a square-root dependence of the excited carrier
concentration on incident illumination, because of the dependence of carrier lifetime
on intensity. Monomolecular recombination leads to a linear dependence on intensity.
Models have been developed to describe carrier density versus light intensity curves
that fall somewhere between these two regimes with a power law exponent between
0.5 and 1. For monomolecular recombination, the carrier lifetime is independent of
the carrier concentration. In the case of bimolecular recombination, the lifetime varies

147
Figure 4-14: Semi-logarithmic plot of external quantum efficiency (EQE) versus wave-
length, for a series of device structures illustrated in the inset. Electric fields applied
during the measurement are indicated to the left of each curve. The Au/TPD device
yields no measurable photoresponse. Both heterojunction devices (solid and dashed
lines) exhibit over 1000-fold improvement in the EQE over the Au/PTCBI device
(dotted line), at 2-fold lower bias voltages. At λ = 552 nm, the Au/PTCBI/TPD
device and Au/TPD/PTCBI device yield EQEs of (8 ± 1)% and (12 ± 1)%, respec-
tively. Accounting for the absorption of λ = 552 nm light in the PTCBI/TPD device,
we find an internal quantum efficiency (IQE) of (140 ± 2)%. This result demonstrates
that lateral bi-layer heterojunctions are capable of quantum gain.

148
with the inverse of carrier concentration. It stands to reason that if these two recom-
bination processes are in parallel then there may be an intermediate exponent value
as the dependence transitions from linear to square-root. However, this transition
will be quite abrupt and should only be observed over a factor of 2 change in light
intensity.

Figure 4-15 shows a 0.56 exponent over 3 orders of magnitude suggesting that
these bilayer devices follow the model proposed by Tabak and Warter [34]. This
model assumes a single film and that there is bimolecular recombination at the illu-
minated surface and monomolecular recombination in the bulk. This assumption is
consistent with the idea that only photogenerated holes are in TPD, while photogen-
erated electrons are confined to PTCBI. Another consequence of the model is that
the bulk of the transporting film is trap-free. This is consistent with the lack of ob-
servable trap-charge limited currents in the I-V sweeps. Finally, back surface effects
are neglected, which is consistent with our shining light through the glass substrate
and having negligible reflection from the films.

Qualitatively, there are three separate regimes of operation for strongly absorbed
light. At low intensities, the monomolecular bulk recombination dominates and cur-
rent is a linear function of light intensity. At intermediate light levels, the bimolecular
surface recombination becomes sufficient to cause the current to follow a square-root
dependence of the photocurrent on light intensity. At high light levels, the surface
recombination becomes a sink for electron-hole pairs which pins the carrier concentra-
tion at the surface to a much lower level than the peak concentration levels in the bulk.
So, the photocurrent becomes linear with light intensity again, but the sensitivity is
much lower than at the low intensity regime. The shape of the photocurrent-intensity
curve is controlled by the transition points, which is dependent on one parameter - the
product of the carrier diffusion length to the absorption coefficient (Lα). Considering
the TPD layer does not absorb much of the green light, the effective penetration depth
of the light approaches 0 (α → ∞). Thus, the ratio of the carrier diffusion length to
the absorption depth in TPD becomes very large. This has the effect of pushing the
transition points further apart, ultimately extending the range of intensities for which

149
Figure 4-15: Log-log I-V characteristics of 50 nm TPD/50 nm PTCBI device under
varying levels of green LED illumination. Inset: Relative photocurrent as a function
of relative intensity reveals sublinear dependence.

150
Figure 4-16: Simulated variation of photocurrent with light intensity for the Tabak-
Warter model, where L is carrier diffusion length and α is absorption coefficient.
Breakpoints between different regimes of dependence are indicated by dashed red
lines. After [34].

the current will follow a square-root dependence [Fig. 4-16]. In the measurements
of the bilayer photoconductor, the range of intensities samples this large sublinear
region between the two linear regions.

4.4.4 Transient Measurements

When measuring the transient aspects of photoconductivity, we can either look at

the build-up or relaxation of steady-state photocurrents. While the turn-on transient
reflects both the generation and recombination of carriers, the turn-off transient only
reflects the recombination effects and is easier to analyze. Turn-off transients that
follow a simple exponential decay will only be observed when a unique recombination

151
path is followed and usually occurs in exceptional cases [35].

When the steady-state illumination is terminated, the generation rate drops to

zero while the carrier density and recombination rates are not affected. Thus, the
initial photocurrent decay will be governed by whatever recombination mode existed
under steady-state conditions. Relaxation of photogenerated carriers in amorphous
semiconductors can occur through either monomolecular or bimolecular recombina-
tion. Monomolecular recombination occurs if a free carrier encounters a recombina-
tion center. As a reminder, recombination centers are traps that have a lower rate
of detrapping relative to recombining. Due to the various recombination processes
in operating devices and the lack of information regarding the mechanisms involved,
the response times have traditionally been used as estimates to carrier lifetimes. An
exponential fit to the initial part of the decay is often used to make this approxima-
tion [35]. This initial decay does not have to represent the most significant part of
the decay process, such as in the case where persistent photoconductivity occurs (one
of the relaxation times observed is longer than the observation time). However, in
analyzing the transient photocurrent curves for the bilayer devices, we have observed
an initial fast decay followed by a much slower response. In this case we take the
initial decay to represent the best estimate of carrier lifetime.

Figure 4-17(a) displays the transient photocurrent decay with varying intensities of
light ranging from 0.25 to 74 mW/cm2 . The photocurrent is induced by illumination
with a green LED (λ = 532 nm) incident through the glass substrate. All curves
are normalized to their steady-state photocurrent levels and time-shifted so that the
illumination cuts off at time t = 0. A constant voltage of 150 V is applied between the
interdigitated electrodes throughout the measurement. It is evident that the decay
rate decreases significantly with increasing intensity. Fitting the different intensity
curves with a double exponential yields a fast and a slow component time constant.
Two-component transient responses are common in experimental work and are an
artifact of using photoconductors with a finite length [36]. However, there is ambiguity
regarding the proper model to use for extracting time constants from transient data
[37]. The decay curve can exhibit two distinct exponential regions if two different

152
Figure 4-17: (a) Semi-logarithmic plot of normalized photocurrent decay under var-
ious intensity illumination and an applied electric field of 15 V/µm. Light is turned
off at time t = 0. Dark current is subtracted from all curves before fitting double
exponentials over the range 2.8 µs to 12 µs. (b) Fitted time constants plotted versus
intensity.

153
recombination mechanisms dominate at different carrier concentration levels, which is
consistent with all recombination mechanisms being carrier concentration dependent.
In this work we associate the initial fast time constant to bimolecular recombination
across the interface, and those values are provided in the inset table of Figure 4-17(a).
The values are all in the microsecond range, which is much slower than the turn off
time of the LED light source (∼ 100 ns) used for excitation. The slower of the two
time constants is associated with bulk recombination in the TPD film. Because TPD
is considered an intrinsic semiconductor we would expect the bulk recombination to
have a slower rate than at the heterointerface where there are many more of the
oppositely charge carrier present one molecular hop away. The dependence of the
carrier lifetime on intensity is visualized in Figure 4-17(b) and fits a power law relation
with an exponent of γ = -0.56, which is consistent with the bimolecular recombination
model exhibited by the photocurrent dependence on intensity. This is a signature of
bimolecular recombination, where the carrier lifetime follows an inverse square-root
dependence on intensity in a trap-free, insulating film.
The lateral bi-layer photoconductor structure presented here exhibits gain and is
shown to generate photocurrent exclusively from absorption in the PTCBI film. In
contrast to a single layer photoconductor, the bi-layer device separates the charge
generation mechanism from the charge transport mechanism. Doing so allows the
engineering of each layer for optimized device performance. In addition, the physical
processes that take place in device operation are not specific to the material set.
The critical requirement is that the energy band alignment is favorable for exciton
dissociation at the interface between the materials.

4.5 Other Material Systems

The heterojunction photoconductor platform is a general device architecture that can

exhibit similar optical and electrical characteristics across a wide range of material
sets. The only major requirement is the ability to form a Type-II heterojunction
between the two materials of interest. Because of the existence of interface dipoles

154
and the inherent uncertainty in energy level measurements (i.e. UPS/IPES) it will
be necessary to try different materials with similar band offsets in order to account
for shifts in vacuum level and changes in energy levels due to deposition conditions.
Until the origin of the interface dipole is well understood, this process of trial and
error may be unavoidable. Two examples of Type-II heterojunction photoconductors
that utilize different materials are detailed below: small molecule organic/quantum
dots and J-aggregate dye/metal oxide.

4.5.1 CdSe Quantum Dot/Spiro-TPD Heterojunction Pho-

toconductor

Colloidal CdSe (Cadmium Selenide) quantum dots (QD) have been used in con-
junction with the hole-transporting small molecule N,N’-Bis(3-methylphenyl)-N,N’-
bis(phenyl)-9,9-spirobifluorene (Spiro-TPD) to fabricate bilayer heterojunction pho-
toconductors. QDs are chemically synthesized QD nanocrystals which can be pro-
cessed from solution and have tunable absorption/emission characteristics based on
the material selection and the physical size of the nanocrystals [38, 39].
The heterojunction photoconductor utilizes an evaporated film of the hole-transporting
material, Spiro-TPD (Luminescence Technologly Corp.), as the CTL, which is sen-
sitized across visible wavelengths by a thin film of TOPO-capped CdSe nanocrystal
QDs, which act as EGL. The device structure is shown in Figure 4-18(a). Electrically,
the hole mobility (10−5 cm2 V-1 s-1 ) of a molecule that is the same in nomenclature
as the Spiro-TPD used in this work, but with a slightly different chemical struc-
ture is demonstrated to be an order of magnitude higher than the electron mobility
in the film of tri-octylphosphine oxide (TOPO) capped CdSe QDs (10−6 cm2 V-1 s-1 )
[40, 41]. Thus, conduction in the device should predominantly occur in the bulk of
the Spiro-TPD film. Optically, the wide bandgap of Spiro-TPD provides an onset
in the absorption spectrum at shorter wavelengths (λ < 400 nm), allowing for the
onset of absorption at longer wavelengths to be determined by the QD layer. Under
broadband illumination, excitons are generated in the organic and QD layers which

155
Figure 4-18: (a) Section view of the device structure. The thickness of the Spiro-
TPD layer is 200 nm. For each device there are 60 sets of Au electrodes arranged
in an interdigitated array forming a serpentine channel (not shown). The channel
length is 10 µm and the electrode width is 1000 µm. (b) Chemical structure of the
hole-transporting material, Spiro-TPD. (c) Schematic of CdSe nanocrystal (∼ 5-7 nm)
passivated with TOPO ligands [42].

together form a Type-II heterojunction [Fig. 4-19(a)]. Excitons within a diffusion

length of this interface may be dissociated, resulting in the transfer of free electrons
to the QDs and holes to the organic charge transport layer. The increase in car-
rier concentration results in an increase in current when a bias is applied across the
electrodes.
The devices consist of interdigitated chrome/gold electrodes (20nm/50nm) on
glass that form a serpentine channel with a width of 60 mm and a length of 10
µm. A 200 nm thick Spiro-TPD film is then thermally evaporated over the elec-
trodes. Subsequently, the samples are transferred directly into a nitrogen-filled glove
box without exposure to air. CdSe QD cores (∼ 5 - 7 nm in diameter) with a lowest
energy absorption peak at wavelength λ = 590 nm are synthesized and passivated
with TOPO ligands to provide solubility in chloroform [46]. Micro-contact printing
from an untreated polydimethylsiloxane (PDMS) stamp is used to deposit the QDs
onto the substrate. Stamps are cleaned with chloroform prior to spin-coating with

156
Figure 4-19: (a) Current-voltage characteristics measured for Spiro-TPD (control)
and Spiro-TPD/QD (QD) devices. Dashed lines represent dark currents. Solid lines
show the device response under illumination from a green LED with an intensity of
74 mW/cm2 and with emission centered at λ = 521 nm. The inset shows the energy
band diagram for the QD device. The energy levels for the Spiro-TPD were taken
from [43]. Energy levels for the CdSe QDs were calculated following the approach
reviewed in [44] and using bulk CdSe parameters obtained from [45]. (b) Absorption
spectra of Spiro-TPD (dashed line) and CdSe QDs (solid line). Emission from the
LED (dotted line) is absorbed primarily by the QD film [42].

157
QDs at 4000 RPM for 60 seconds. The stamps coated with QDs are then placed
in a vacuum environment (∼ 10−3 Torr) for approximately 1 hour to remove excess
solvent. Finally, the PDMS stamp is applied to the device substrate with a small
amount of force to transfer the QD film. The thicknesses of transferred QD films
were measured to be approximately 50 nm, and completed devices were tested in a
nitrogen environment.

I-V measurements are shown in Figure 4-19(a) for a control device, consisting of
a 200 nm thick film of Spiro-TPD, and for a bilayer Spiro-TPD/QD heterojunction
device. Following the absorption spectra of the materials used, the green LED illu-
mination is primarily absorbed by the QDs [Fig. 4-19(b)]. The Spiro-TPD device
shows little photoresponse as evident by comparing the gray solid line to the gray
dashed line in Figure 4-19. This behavior is consistent with Spiro-TPD’s absorption
spectrum, and its inability to absorb the green illumination. The 50% decrease in the
dark current from the control device to the bilayer device, while rather large, occurs
when the QD film is sufficiently thick. In devices with a single layer of Spiro-TPD, as
the Spiro-TPD films grow thicker there is a consistent increase in dark current, sug-
gesting that conduction occurs in the bulk of the Spiro-TPD film. The QD stamping
process can negatively affect the overall dark conductivity by removing some of the
underlying CTL while transferring the QDs. There exists some evidence to support
this theory given that untreated PDMS stamps can be used to lift-off organic and
metal thin films [47]. However, this theory relies on the assumption that the organic
material that is being stamped is “soft” enough to squeeze through the interstitial
spaces or cracks in the QD film to interact with the PDMS stamp. Anecdotally, this
seems to be the case given that stamping QD films has been shown to embed the QDs
into underlying small-molecule organic materials such as Spiro-TPD. Complete QD
films remove little to no underlying organic layer, suggesting that the organic mate-
rial is being removed through cracks in incomplete films. More careful studies of the
microcontact printing process and the effects on the underlying organic material still
need to be undertaken to elucidate the physical mechanism for removal of underlying
material with QD stamping.

158
 Another insight can be inferred from the decrease in dark conductivity post-
stamping. Because we have seen up to this point that dark currents rise after the
introduction of the heterojunction due to charge transfer across the interface, a re-
duced dark current could indicate little to no charge transfer occurring between the
QD film and the underlying Spiro-TPD film, which would make the heterointerface
less conductive than the bulk Spiro-TPD. Data to support this assumption is shown
in Figure 4-20 where for varying thicknesses of Spiro-TPD and varying thicknesses
of QDs, the dark current decreases indicating no charge transfer in addition to re-
moval of underlying Spiro-TPD. In order to explain this phenomenon, another piece
of information is required concerning the QDs ability to transfer charge. To vary the
thickness of the stamped QD films, different concentrations of QD solutions are used,
the more dilute solutions yielding on average thinner spun films on the PDMS stamps.
When QDs are synthesized, they are naturally capped by TOPO ligands which are in
a dynamic equilibrium between being bound to the surface of the dot and being free
in solution [48]. This implies that the more dilute the solution of QDs, the lower the
density of ligands at the nanocrystal surface. Fewer ligands suggest that there may be
a higher probability of charge transfer to occur between the QD and its surrounding
medium. Therefore, the ability for charge to transfer between the Spiro-TPD and the
QDs may be a function of the concentration of the QD solution used.

Assuming the electric field distribution between the electrodes is uniform, a bias
of 100 V is equivalent to an electric field strength of 105 V/cm. At this bias, the
ratio of the light current to the dark current increases from 1.1, in the control device,
to 3.5 for the bilayer QD device. The energy band alignment and high CTL hole
mobility in the Spiro-TPD/QD devices would suggest similar device performance to
the TPD/PTCBI devices. However, there is a difference of roughly two orders of
magnitude in the light to dark current ratios. Comparing data from the I-V curves it
is evident that the increased dark current in the Spiro-TPD/QD device is responsible
for the reduction in the ratio of light to dark current. Given that Spiro-TPD has a
slightly lower field effect mobility than TPD, it cannot account for the higher relative
dark conductivity. One possible explanation for this behavior is the energy band

159
Figure 4-20: Log-log current-voltage characteristics of various Spiro-TPD/QD de-
vices. Dotted lines represent single films of Spiro-TPD, while solid lines represent
the QD-stamped bilayer devices. The magenta dotted line is a guide to the eye that
represents the functional form of SCLC.

160
Figure 4-21: EQE spectra for a variety of biases. Inset shows EQE at λ = 590 nm
as a function of bias. Photocurrent follows a nearly V 2 power law indicative of space
charge limited conduction in Spiro-TPD.

alignment of Spiro-TPD with respect to the gold contacts. Spiro-TPD has a HOMO
at 4.9 eV, effectively presenting a lower barrier to hole injection (as compared to
TPD) from gold, which has a work function of c.a. 5.1 eV and a large VL shift of ∼
1 eV [14]. Thus, the lower barrier to charge injection into Spiro-TPD would serve to
increase the overall dark conductivity of the device.

Figure 4-21 shows the EQE of the QD device for a variety of biases. A large
response peak that occurs between λ = 300 nm and λ = 400 nm is attributable to
light that is absorbed in the Spiro-TPD layer. There is also a strong photoresponse
for λ > 400 nm that closely follows the absorption spectrum of the QDs. A maximum
in the EQE occurs at λ = 590 nm, the lowest energy absorption peak of the QDs,
which reaches 13% at 300V. The corresponding internal quantum efficiency (IQE)
at λ = 590 nm is calculated to be 79%. For a sufficiently high bias applied across

161
the electrodes an EQE of greater than 100% may be realized, which would suggest
that the device is operating in a regime of photoconductive gain. The inset reveals
that the EQE measured at λ = 590 nm follows a V 2 power law dependence with
bias. This is consistent with space charge limited conduction in the ideal case of a
trap-free insulator [49, 50]. We note that there is a strong peak in photoresponse at λ
= 400 nm, precisely where Spiro-TPD absorption begins to fall off [Fig. 4-21(b)]. A
photoresponse in a layered heterojunction device that is strongest at wavelengths for
which absorption is weakest can be described as antibatic (vs. symbatic) and suggests
that the heterointerface is effectively dissociating excitons [51]. The presence of the
antibatic peak at λ = 400 nm therefore supports the assertion that the interface
between Spiro-TPD and the QD film is the site of exciton dissociation in the device.
Comparing the IQE value of the Spiro-TPD/QD device with that of the TPD/PTC-
BI device, it is evident that at the absorption peak of the EGL, the TPD/PTCBI
devices have higher efficiencies at lower applied fields. This is another signature of
the lower mobility of Spiro-TPD and of the diminished charge transfer dynamics at
the organic/QD interface due to the presence of the organic ligands. It is well known
that QD interspacing (which is set by the length of the ligand) affects not only the
conductivity of the QD films, but also the dissociation efficiency of excitons where one
of the charges has to tunnel through a potential barrier to an adjacent QD [52, 53].
Annealing the QD films is one way to reduce the negative electrical effects of the
ligands. However, the CTL would have to be replaced with a material set that can
withstand the high temperatures required for annealing (> 100○ C) [46].

4.5.2 TDBC J-aggregate dye/Zinc Indium Oxide Heterojunc-

tion Photoconductor

Cyanine dyes have traditionally been used to sensitize silver halides in photographic
processing. Unlike most organic dye materials, this special class of dyes can form
thin films that have very narrow emission and absorption spectra (∼ 10 nm FWHM)
and exhibit record high optical absorption coefficients (∼ 106 cm-1 ) in solid state [54].

162
Figure 4-22: Absorption spectra of TDBC dye in monomer phase (blue) and J-
aggregate phase (red). Inset: Chemical structure of TDBC dye.

The cyanine dyes aggregate in such a way that their transition dipoles align to form
J-aggregates, which are aggregate states that have enhanced resonance in the J-band
[Fig. 4-22]. We have used a well-characterized J-aggregating dye [5,6-dichloro-2-[3-
[5,6-dichloro-1-ethyl-3-(3-sulfopropyl)-2(3H)-benzimidazolidene]-1-propenyl]-1-ethyl
-3-(3-sulfopropyl) benzimidazolium hydroxide (inner salt, sodium salt from Nippon
Kankoh Shikiso Kenkyusho Co., Ltd.) (TDBC), in conjunction with Zinc Indium
Oxide (ZIO) as the CTL to fabricate bilayer photoconductors.
The TDBC film, acting as the EGL with peak absorption at λ = 595 nm [Fig.
4-22], was deposited using a layer-by-layer (LBL) process [55–57]. The sequential
immersion of the substrate into cationic and anionic solutions allows the J-aggregate

163
to form highly absorptive films that contain only a few molecular layers. The cationic
solution is a 30 mM solution of poly(diallyldimethylammoniumchloride) [(PDAC)
obtained from Sigma Aldrich] in deionized water. The anionic solution is a 50 µM
solution of the J-aggregating cyanine dye TDBC. Immersing the devices into aqueous
solutions during fabrication necessitates the use of a CTL that is impervious to the
solvent. A natural material for use as the CTL is a metal oxide, which typically has
a wide bandgap (visibly transparent) and is mechanically robust (insoluble and high
glass transition temperature).

The heterojunction photoconductor consists of interdigitated chrome/gold elec-

trodes (20nm/50nm) on glass that form a serpentine channel with a width of 60 mm
and a length of 10 µm. A 50 ± 0.5 nm RF-sputter-deposited film of the n-type metal
oxide, ZIO (target obtained from Plasmaterials, Inc.) is deposited onto the elec-
trodes before dip-coating a 5 ± 1 nm TDBC J-aggregate film. ZIO is a wide bandgap
semiconductor whose absorption falls off quickly at wavelengths longer than λ = 550
nm [Fig. 4-23]. In order to reduce the conductivity of the metal oxide film to the
range of the measurement equipment, the ZIO was grown in a 30% O2:70% Ar gas
environment (filling the oxygen vacancies resulted in more resistive thin films of ZIO).

I-V measurements are shown in Figure 4-24(a) for a device consisting of a 5 ± 1 nm

thick film of TDBC, a device consisting of a 50 ± 0.5 nm thick film of ZIO, and a bilayer
ZIO/J-aggregate heterojunction device. While the carrier mobility in TDBC has not
been measured, the photoconductor made with a single TDBC film (black) exhibits
current levels below the ZIO device (blue). This would suggest that the ZIO (with a
resistance of ∼ 700 MΩ/◻) would be primarily responsible for electrical conduction in
the bilayer device. According to the absorption spectra of the materials used, amber
LED illumination (λpeak = 590 nm) is primarily absorbed by the J-aggregate [Fig. 4-
23]. However, the low conductivity of the TDBC film yields low currents in both dark
and light conditions (black lines). The ZIO device shows a small photoresponse as
evident by comparing the blue solid line to the blue dashed line in Figure 4-24. This
behavior is consistent with ZIO’s absorption spectrum and its long decaying tail that
extends across the visible spectrum. The greater than order of magnitude increase

164
Figure 4-23: Absorption spectra of ZIO and TDBC films. Emission spectrum of
amber LED centered at λ = 595 nm is also shown.

165
Figure 4-24: Log-log current-voltage characteristics for a series of devices containing
ZIO and TDBC. I-V sweeps are collected under both light (solid) and dark (dashed)
conditions. Illumination is provided by an amber LED (λpeak = 595 nm). Inset
cross-sectional diagrams detail device structures and film thicknesses.

in the dark current from the ZIO device (black dashed line) to the ZIO/J-aggregate
bilayer device (blue dashed line) suggests charge transfer across the heterointerface
and the creation of a Type-II heterojunction. Examining the red set of lines in Figure
4-24, it is clear that the photoresponse of the ZIO/J-aggregate has benefited from the
presence of the interface (where efficient exciton dissociation can occur) as well as
the higher conductivity in the underlying ZIO film. OLEDs were demonstrated with
TDBC J-aggregate as the emitting layer [58], suggesting that TDBC has a lower
electron affinity than the ZIO. Thus, as excitons are dissociated at the interface,
electrons travel into the ZIO, while holes remain in the TDBC film.

These bilayer devices demonstrate a light-to-dark current ratio of over 1000, due
to a high photocurrent. There are two possible factors that can contribute to this
phenomenon, either carriers have a long lifetime or a short transit time. Measurements
of the carrier lifetime [Fig. 4-25] yield a 90 µs decay time constant which is on the same
order or slightly larger than that of the TPD/PTCBI devices for a given intensity.

166
Figure 4-25: Plot of photocurrent decay for a 50.0 ± 0.5 nm ZIO/5.0 ± 0.5 nm TDBC
bilayer heterojunction photoconductor. Illumination by an amber LED (λpeak = 595
nm) is turned off at time t = 0 s.

Therefore, it must be that the good conductivity of ZIO and the favorable injection
barrier (relative to TPD) keeps carrier transit times short and allows for multiple
transits through the channel for every photoexcited carrier.

Figure 4-26 shows the EQE of the ZIO/TDBC device (red curve) at an applied
field of 1x106 V/µm. The single film of TDBC yielded no measurable quantum ef-
ficiency due to the low intensity of the monochromatic illumination as well as the
low conductivity of the film. The single film of ZIO (blue curve) exhibits significant
photocurrent symbatic with its absorption spectrum. The bilayer device results in an
EQE of c.a. 300% at the TDBC peak absorption wavelength. In addition, the ZIO’s
photoresponse is improved up to a factor of 50 over its absorption spectrum. The
improvement in photoresponse of the bilayer device is a signature of efficient exciton
dissociation at the Type-II heterojunction. The corresponding internal quantum ef-
ficiency (IQE) at λ = 590 nm is calculated to be over 2000%, consistent with a large
photon-to-electron conversion gain and a much shorter transit time relative to carrier

167
Figure 4-26: Spectral response data for a 50.0 ± 0.5 nm ZIO device and a 50.0 ±
0.5 nm ZIO/5.0 ± 0.5 nm TDBC bilayer device. Inset cross-sectional diagrams detail
device structure and film thicknesses. An electric field of 1 x 106 V/µm is applied to
both devices.

lifetime.

4.6 Summary
In this chapter we have focused on understanding the device physics of a hetero-
junction and a traditional single layer photoconductor. Understanding the role of
both concepts in the bilayer heterojunction photoconductor is crucial to modeling
device operation. The device model of the bilayer heterojunction photoconductor
is based largely on existing models of photoconductor operation as well as hetero-
junction dynamics in solar cells. Experimental data collected from heterojunction
photoconductors that incorporate a wide variety of material sets, have corroborated
our models and yield further insight into pathways for optimization.

168
Chapter 4 References

[1] C. W Tang and S. A Vanslyke. Organic Electroluminescent Diodes. Applied

Physics Letters, 51(12):913–915, 1987.

[2] T. Chakraborty and V. M. Apalkov. Quantum cascade transitions in nanostruc-

tures. Advances in Physics, 52(5):455, 2003.

[3] B. P. Rand, D. P. Burk, and S. R. Forrest. Offset energies at organic semicon-

ductor heterojunctions and their influence on the open-circuit voltage of thin-
film solar cells. Physical Review B (Condensed Matter and Materials Physics),
75(11):115327–11, March 2007.

[4] P. Peumans and S. R Forrest. Separation of geminate charge-pairs at donor-

acceptor interfaces in disordered solids. Chemical Physics Letters, 398:27–31,
2004.

[5] J. D. Wright. Molecular Crystals. Cambridge University Press, Cambridge, 2nd

edition, 1987.

[6] W. R Salaneck, K. Seki, A. Kahn, and J. Pireaux. Conjugated Polymer and

Molecular Interfaces. Marcel Dekker, Inc., New York, 2002.

[7] H. Ishii, K. Sugiyama, E. Ito, and K. Seki. Energy level alignment and inter-
facial electronic structures at Organic/Metal and Organic/Organic interfaces.
Advanced Materials, 11(8):605–625, 1999.

[8] R. W. Strayer, W. Mackie, and L. W. Swanson. Work Function Measurements

169
 by Field-emission Retarding Potential Method. Surface Science, 34(2):225–248,
1973.

[9] J. Hoelzel, F. K. Schulte, and H. Wagner. Solid State Surface Physics. Springer
Verlag, Berlin, 1979.

[10] H. Ishii, K. Sugiyama, D. Yoshimura, E. Ito, Y. Ouchi, and K. Seki. Energy-level

alignment at model interfaces of organic electroluminescent devices studied by
UV photoemission: Trend in the deviation from the traditional way of estimating
the interfacial electronic structures. IEEE Journal of Selected Topics in Quantum
Electronics, 4(1):24–33, February 1998.

[11] E. Ito, H. Oji, H. Ishii, K. Oichi, Y. Ouchi, and K. Seki. Interfacial elec-
tronic structure of long-chain alkane/metal systems studied by UV-photoelectron
and metastable atom electron spectroscopies. Chemical Physics Letters, 287(1-
2):137–142, April 1998.

[12] A Rajagopal, CI Wu, and A Kahn. Energy level offset at organic semiconductor
heterojunctions. Journal of Applied Physics, 83(5):2649–2655, March 1998.

[13] J. X. Tang, C. S. Lee, and S. T. Lee. Electronic structures of organic/organic

heterojunctions: From vacuum level alignment to fermi level pinning. Journal
of Applied Physics, 101(6):064504–4, March 2007.

[14] H. Ishii, N. Hayashi, E. Ito, Y. Washizu, K. Sugi, Y. Kimura, M. Niwano,

Y. Ouchi, and K. Seki. Kelvin probe study of band bending at organic semicon-
ductor/metal interfaces: examination of fermi level alignment. Physica Status
Solidi (A), 201(6):1075–1094, 2004.

[15] S. M. Sze and K. K. Ng. Physics of Semiconductor Devices. John Wiley & Sons,
Hoboken, third edition, 2007.

[16] M. Pope and C. E Swenberg. Electronic processes in organic crystals and poly-
mers. Oxford University Press, New York, 2nd edition, 1999.

170
[17] H. B. DeVore. Spectral distribution of photoconductivity. Physical Review,
102(1):86, April 1956.

[18] P. M Borsenberger and D. S. Weiss. Organic photoreceptors for xerography.

Optical Engineering. Marcel Dekker, New York, 1998.

[19] A.J. Lewis, A. Ruseckas, O. P.M. Gaudin, G.R. Webster, P.L. Burn, and I.D.W.
Samuel. Singlet exciton diffusion in MEH-PPV films studied by exciton-exciton
annihilation. Organic Electronics, 7(6):452–456, December 2006.

[20] H. Alves, A. S. Molinari, H.. Xie, and A. F. Morpurgo. Metallic conduction at

organic charge-transfer interfaces. Nature Materials, 7(7):574–580, July 2008.

[21] W. F. Pasveer, J. Cottaar, C. Tanase, R. Coehoorn, P. A. Bobbert, P. W. M.

Blom, D. M. deLeeuw, and M. A. J. Michels. Unified description of Charge-
Carrier mobilities in disordered semiconducting polymers. Physical Review Let-
ters, 94(20):206601, May 2005.

[22] H. J. Snaith and M. Gratzel. Light-Enhanced charge mobility in a molecular

hole transporter. Physical Review Letters, 98(17):177402–4, April 2007.

[23] C. Madigan and V. Bulovic. Solid state solvation in amorphous organic thin
films. Physical Review Letters, 91(24):247403, 2003.

[24] C. Madigan and V. Bulovic. Modeling of exciton diffusion in amorphous organic

thin films. Physical Review Letters, 96:046404, 2006.

[25] D. Cheyns, J. Poortmans, P. Heremans, C. Deibel, S. Verlaak, B. P. Rand, and

J. Genoe. Analytical model for the open-circuit voltage and its associated resis-
tance in organic planar heterojunction solar cells. Physical Review B (Condensed
Matter and Materials Physics), 77(16):165332–10, April 2008.

[26] C. L. Braun. Electric field assisted dissociation of charge transfer states as

a mechanism of photocarrier production. The Journal of Chemical Physics,
80(9):4157–4161, May 1984.

171
[27] J. Nelson, J. Kirkpatrick, and P. Ravirajan. Factors limiting the efficiency of
molecular photovoltaic devices. Physical Review B, 69(3):035337, 2004.

[28] M. M. Mandoc, F. B. Kooistra, J. C. Hummelen, B. de Boer, and P. W. M.

Blom. Effect of traps on the performance of bulk heterojunction organic solar
cells. Applied Physics Letters, 91(26):263505–3, December 2007.

[29] M. M. Mandoc, W. Veurman, L. J.A. Koster, B. deBoer, and P. W. M. Blom.

Origin of the reduced fill factor and photocurrent in MDMO-PPV:PCNEPV
All-Polymer solar cells. Advanced Functional Materials, 17(13):2167–2173, 2007.

[30] P. Peumans, A. Yakimov, and S. R. Forrest. Small molecular weight organic thin-
film photodetectors and solar cells. Journal of Applied Physics, 93(7):3693–3723,
2003.

[31] S. B. Rim, R. F. Fink, J. C. Schoneboom, P. Erk, and P. Peumans. Effect of

molecular packing on the exciton diffusion length in organic solar cells. Applied
Physics Letters, 91(17):173504–3, October 2007.

[32] B. P. Rand, J. Xue, S. Uchida, and S. R. Forrest. Mixed donor-acceptor molecular

heterojunctions for photovoltaic applications. i. material properties. Journal of
Applied Physics, 98(12):124902–7, December 2005.

[33] F. Celii, T. Harton, and O. F. Phillips. Characterization of organic thin films

for OLEDs using spectroscopic ellipsometry. Journal of Electronic Materials,
26(4):366–371, April 1997.

[34] M. D. Tabak and P. J. Warter. Sublinear photocurrents and surface recombina-

tion in cadmium sulfide crystals. Physical Review, 148(2):982, 1966.

[35] M. Brinza, J. Willekens, M. Benkhedir, E. Emelianova, and G. Adriaenssens.

Photoconductivity methods in materials research. Journal of Materials Science:
Materials in Electronics, 16(11):703–713, November 2005.

172
[36] N. M. Haegel, D. R. Palmieri, and A. M. White. Current transients in extrinsic
photoconductors: comprehensive analytical description of initial response. Ap-
plied Physics A: Materials Science & Processing, 73(4):433–439, October 2001.

[37] D. A. Redfern, C. A. Musca, E. P. G. Smith, J. M. Dell, and L. Faraone. On the

transient photoconductive decay technique for lifetime extraction in HgCdTe. In
Optoelectronic and Microelectronic Materials Devices, 1998. Proceedings. 1998
Conference on, pages 275–278, 1998.

[38] J. S. Steckel, S. Coe-Sullivan, V. Bulović, and M. G. Bawendi. 1.3 um to 1.55

um tunable electroluminescence from PbSe quantum dots embedded within an
organic device. Advanced Materials, 15(21):1862–1866, 2003.

[39] E. H. Sargent. Infrared quantum dots. Advanced Materials, 17(5):515–522, 2005.

[40] T. P. I. Saragi and J. Salbeck. Organic heterostructure field-effect transis-

tors using C60 and amorphous spirolinked compound. Applied Physics Letters,
89(25):253516–3, December 2006.

[41] D. Yu, C. Wang, and P. Guyot-Sionnest. n-Type conducting CdSe nanocrystal

solids. Science, 300(5623):1277–1280, May 2003.

[42] T. P. Osedach, S. M. Geyer, J. C. Ho, A. C. Arango, M. G. Bawendi, and

V. Bulovic. Lateral heterojunction photodetector consisting of molecular organic
and colloidal quantum dot thin films. Applied Physics Letters, 94(4):043307–3,
2009.

[43] T. P. I. Saragi, T. Fuhrmann-Lieker, and J. Salbeck. Comparison of Charge-

Carrier transport in thin films of Spiro-Linked compounds and their correspond-
ing parent compounds. Advanced Functional Materials, 16(7):966–974, 2006.

[44] A. L. Efros and M. Rosen. The Electronic Structure of Semiconductor Nanocrys-

tals. Annual Review of Materials Science, 30, November 2003.

173
[45] T. C. Chiang and F. J. Himpsel. Electronic Structure of Solids: Photoemission
Spectra and Related Data, volume 23 of New Series. Springer-Verlag Berlin,
Berlin, 1989.

[46] M. V. Jarosz, V. J. Porter, B. R. Fisher, M. A. Kastner, and M. G. Bawendi. Pho-

toconductivity studies of treated CdSe quantum dot films exhibiting increased
exciton ionization efficiency. Physical Review B, 70(19):195327, November 2004.

[47] J. J. Yu. Improving OLED Technology for Displays. Ph.D. thesis, MIT, 2008.

[48] J. C Martins, I. Moreels, and Z. Hens. Solution NMR spectroscopy as a useful

tool to investigate colloidal nanocrystal dispersions from the capping ligand’s
point of view. In 2006 MRS Fall Proceedings, volume 984, 2006.

[49] M. A. Lampert. Simplified theory of Space-Charge-Limited currents in an insu-

lator with traps. Physical Review, 103(6):1648, 1956.

[50] A. Rose. Recombination processes in insulators and semiconductors. Physical

Review, 97(2):322, 1955.

[51] A. K Ghosh and T. Feng. Merocyanine Organic Solar-Cells. Journal of Applied

Physics, 49(12):5982–5989, 1978.

[52] C. A. Leatherdale, C. R. Kagan, N. Y. Morgan, S. A. Empedocles, M. A. Kastner,

and M. G. Bawendi. Photoconductivity in CdSe quantum dot solids. Physical
Review B, 62(4):2669, July 2000.

[53] A. Franceschetti and A. Zunger. Exciton dissociation and interdot transport in

CdSe quantum-dot molecules. Physical Review B, 63(15):153304, March 2001.

[54] M. S. Bradley, J. R. Tischler, and V. Bulović. Layer-by-Layer J-Aggregate

thin films with a peak absorption constant of 106 cm-1 . Advanced Materials,
17(15):1881–1886, 2005.

[55] H. Fukumoto and Y. Yonezawa. Layer-by-layer self-assembly of polyelectrolyte

and water soluble cyanine dye. Thin Solid Films, 327-329:748–751, August 1998.

174
[56] M. Kawasaki, T. Sato, and T. Yoshimoto. Controlled layering of Two-
Dimensional J-Aggregate of anionic cyanine dye on Self-Assembled cysteamine
monolayer on au(111). Langmuir, 16(12):5409–5417, June 2000.

[57] C. Peyratout and L. Daehne. Aggregation of thiacyanine derivatives on poly-

electrolytes. Physical Chemistry Chemical Physics, 4(13):3032–3039, 2002.

[58] J. R. Tischler, M. S. Bradley, V. Bulovic, J. H. Song, and A. Nurmikko. Strong

coupling in a microcavity LED. Physical Review Letters, 95(3):036401, July 2005.

175
176
Chapter 5

Multilayer Heterojunction
Photoconductors

It is through science that we prove,

but through intuition that we
discover.

JULES HENRI POINCARE

Mathematician/Physicist

Having detailed the carrier dynamics and processes that take place at the het-
erointerface of the bilayer photoconductors, we turn our focus now towards modifying
those same processes to engineer better device performance. Because gain is linearly
dependent on carrier lifetime, we must look for methods of limiting the possibility
of carrier recombination processes at the interface in order to fabricate devices with
higher photocurrents. Detailing a number of different device architectures used to
manipulate carrier lifetimes will help to motivate the use of multiple layers in the
heterojunction devices.

5.1 Quantum Well in an Electric Field

During the 1980’s inorganic multiple quantum well (MQW) structures were receiv-
ing considerable attention for enabling the ability to engineer their photoluminescent

177
Figure 5-1: Schematic of energy bands and wavefunctions of electrons and holes in a
quantum well under an applied electric field.

properties by varying barrier heights and widths, which would allow for the creation
of solid-state lasers and LEDs. Around this time it was demonstrated that an applied
electric field perpendicular to the layers defining the quantum well could decrease or
completely quench the luminescence associated with the MQW structure [1]. This
phenomenon was originally associated with field induced separation between the elec-
trons and holes in the quantum well [2]. Thus, increasing the well width significantly
decreases luminescence [Fig. 5-1]. However, this initial model neglected the effect of
free carrier density and the ability of free charge to screen the applied electric field.
A study which detailed the effect of injection current on the recombination rate in a
quantum well [3], found that at much higher carrier densities the free-carrier-induced
field acted to cancel the applied electric field, which causes the separation of electrons
and holes to have a much weaker dependence on the applied field.

5.2 n-i-p-i Structures

In photodetection, the creation of luminescence through recombination represents

a loss mechanism to the extracted photocurrent. Thus, one method of improving
photocurrent levels in a photodetector is to reduce the bimolecular recombination rate

178
by physically separating the electrons and holes. One particular class of engineered
material that accomplishes this separation is the n-i-p-i crystal. n-i-p-i crystals refer
to doped superlattices where periodic n- and p-doping can be used to tune various
material properties. The basic idea of doping superlattices is to create a periodic
energy band modulation in real space. This is accomplished by growing alternating
layers with opposite polarity of dopant. One commonly used scheme to create n-i-p-i
crystals is to fabricate a series of alternating homogeneous doped n and p layers. In
this case, the intrinsic regions consist of the space-charge regions between the layers
due to depletion region formation. Undoped layers can be grown between the two
doped layers in order to increase the intrinsic regions. In the most extreme case,
intrinsic layers are separated by monolayers of doped semiconductor (also known as
δ-doping), which creates a sawtooth potential profile [4, 5].

One benefit of the n-i-p-i architecture is the built-in potential modulation [Fig.
5-2], which leads to a spatial separation of photogenerated electron-hole pairs and
thus to a drastic enhancement of the excess carrier lifetime as compared to the bulk
semiconductor. This separation can be made nearly perfect by appropriate choice
of doping concentrations and layer thicknesses. According to Dohler [6], there are
two mechanisms that limit the lifetime in a p-i-n structure: tunneling through the
potential barrier due to the quantum mechanical overlap of the electron and hole
wave functions, and thermal excitation across the potential barrier so that electrons
and holes can meet. Due to the overlap of the electron and hole wavefunctions photon
absorption is also possible for photon energies less than the bandgap.

One particular device architecture that has emerged from this class of structures is
the p-i-n photodiode. The p-i-n photodiode consists of an insulating film sandwiched
between a p-type doped film and an n-type doped film. This architecture physically
extends the width of the depletion region (where excitons can be dissociated by the
electric field) thereby increasing the area available for capturing light [7]. The p-i-n
photodiode has a number of benefits over the p-n photodiode. Increasing the width
of the depletion layer reduces junction capacitance and therefore the associated time
response. However, the transit time of photogenerated carriers increases with the

179
width of the depletion region.
Recently, Sista et. al. reported using an interfacial layer between the donor
and acceptor layers of an organic solar cell to increase open circuit voltage [8]. In
that work, the interfacial layer is used to create a pair of new interfaces where the
HOMO-LUMO offsets are much greater between the interfacial layer and the donor
(or acceptor) material as compared with the donor-acceptor HOMO-LUMO offset
in the original bilayer cell. While it is widely accepted that open circuit voltage is
solely due to the difference between the HOMO of the donor and the LUMO of the
acceptor [9], this is only the case when recombination is high [10]. Work done by
Arango et. al. demonstrated record-high open circuit voltages in nanostructured
solar cells incorporating CdSe quantum dots [11]. In this case, the organic ligands
that passivate the nanocrystals are believed to prevent bimolecular recombination
across the interface between the CdSe QDs and the organic transport material. By
reducing the recombination rate at the interface, charge carriers must tunnel through
or hop over the barrier posed by the capping groups in order to recombine on the
quantum dots, yielding open circuit voltages that are close to the actual band gap
of the CdSe (∼ 1.5 eV). This work suggests the use of interfacial layers to reduce
interfacial recombination rates relative to charge transfer at the interface, providing
an experimental mechanism to enhance device efficiency.

5.3 Trilayer Photoconductor Model

Having delineated the benefits of physically separating free carriers to extend their
lifetimes, and the use of an interfacial layer to achieve that separation, a new class
of trilayer heterojunction photoconductors can be conceived of and fabricated. This
multilayer architecture extends the bilayer device by introducing an interfacial layer
to form a cascading energy band structure.
Neglecting the effects of interfacial dipoles, it becomes feasible to engineer novel
trilayer devices based on the band energies of individual organic materials. These
devices would ideally consist of an exciton generation layer (EGL), a charge spacer

180
Figure 5-2: Schematic of a n-i-p-i doping superlattice. Space charge regions between
the doped slices creates insulating zones. The conduction and valence band edges
modulate parabolically with spatially separated electron and hole wavefunctions in
their ground states.

181
layer (CSL), and a charge transport layer (CTL). In devising a working model for
device operation, it is useful to start with the energy bands of a bilayer and illustrate
all of the possible carrier interactions and decay pathways [Fig. 5-3(a)]. Figure
5-3(b) depicts the same structure with a CSL acting to separate the charge carrier
populations over its thickness. In order for the introduced interfacial layer to have the
intended effect of increasing carrier lifetimes by reducing the recombination rate across
the donor-acceptor interface, the change in that loss mechanism must be greater than
the sum of the new loss pathways. In other words, kCT L−CSL + kCSL−EGL < kCT L−EGL
so that the dominant recombination pathway is across the thickness of the CSL. This
rate should be reduced relative to the bilayer case due to the smaller overlap of the
carrier wavefunctions. Because the interfacial recombination rates are dependent on
the carrier concentrations that are built up on either side of the interface, ke and
kh should be greater than zero to ensure the previous assertion. These rates are
diffusion-limited and therefore will be highly dependent on the thickness of the CSL.
Thus, there will be an optimum thickness above which carriers can no longer diffuse
across the thickness of the CSL, resulting in bilayer operation between the CSL/EGL
and CTL/CSL interfaces. This crude model captures the larger, qualitative trends
we expect to observe. However, it also neglects the availability of other energetically
favorable states (i.e. triplet excitons) that may play a significant role in experimental
observation of the recombination rates.

5.4 Experiments

While simulations can provide useful guidelines to inform the choice of materials to use
in fabricating the trilayer heterojunction photoconductors, simplifying assumptions
(such as complete, homogeneous, trap-free films, no interface dipole formation, etc.)
must be made to yield reasonable computation times. As the experimental results
below will show, these ideal film properties are much more difficult to realize in
practice. Chemical structures of all of the materials used in the trilayer experiments
can be found in Figure 5-4.

182
Figure 5-3: Energy band diagrams of (a) a bilayer device consisting of CTL/EGL and
(b) a trilayer device consisting of CTL/CSL/EGL. Relevant interfacial recombination
rates and carrier diffusion rates are depicted.

5.4.1 Cascading Energy Bands

The first set of experiments involves the use of materials which are likely to form a
cascading set of energy bands. For consistency with the previous bilayer data, all of
the following trilayer structures make use of TPD as CTL and PTCBI as EGL. Due to
the inherent uncertainty in photoemission measurements of energy bands (± 0.1 eV)
and the dependence of the energy levels on morphology and deposition conditions
[12], the isolated HOMO/LUMO levels in vacuum are used as a starting point for
engineering trilayer devices.

TPD/Alq3 /PTCBI Trilayer

Tris(8-hydroxyquinolinato)aluminum(III) (Alq3 ) is commonly used as a green emit-

ting layer in OLEDs and was part of the first demonstration of an efficient OLED,
along with TPD [13]. The isolated energy bands of the individual materials are de-
picted in Figure 5-5, which shows a cascade in energy bands and justifies the use of
Alq3 as CSL. Optically, the introduction of Alq3 plays little role in the absorption
characteristics of the device. The CSL is typically much thinner than either the CTL
or the EGL making the absorption of the CSL negligible relative to the other films
in the device. The absorption spectra of the materials also indicate that the use of

183
Figure 5-4: Chemical structures of materials used in trilayer devices. Clockwise from
top-left corner: N,N’-bis(3-methylphenyl)-N,N’-diphenyl-1,1’-biphenyl-4,4’-diamine
(TPD), 3,4,9,10-perylenetetracarboxylic bis-benzimidazole (PTCBI), N,N’-diphenyl-
N,N’-bis(1-naphthyl)-1,1’-biphenyl-4,4”-diamine (α-NPD), oleic acid-capped lead se-
lenide quantum dot (4.7 nm PbSe QD), p-bis(triphenylsilyl)benzene (UGH2), tris
(8-hydroxyquinolinato) aluminum(III) (Alq3 ).

184
Figure 5-5: Absorption spectra of TPD, Alq3 , and PTCBI films (50 nm thick). The
green LED emission spectrum (λpeak = 532 nm), operated at an intensity of 40
mW/cm2 , is included for reference (dashed line). Inset: Suggested energy bands
of TPD [14], Alq3 [15], and PTCBI [16] based on photoemission spectroscopy data
and optical bandgaps and digital images of materials in purified, powder form.

a green LED will continue to predominantly excite the PTCBI film [Fig. 5-5]. AFM
images [Fig. 5-6] demonstrate good wetting characteristics of the materials following
deposition, which ensures complete thin film formation.

The simulations above suggested that there would be an improvement in the pho-
tocurrent characteristics of the trilayer devices (relative to the bilayer) with similar
energy band alignment as that of the TPD/Alq3 /PTCBI devices used in the exper-
iments. However, after conducting several independent device trials, no clear trend
emerged in the photocurrent. Figure 5-7 compiles the results of nine different device
runs, where the Alq3 thickness was varied over a wide range of thicknesses. All of the
values were obtained with the same applied electric field (10 V/µm) and have been
plotted relative to the quantum efficiency of the bilayer device (control) for that par-

185
Figure 5-6: Set of AFM images of various films consisting of (a) 50 ± 0.5 nm TPD (b)
50 ± 0.5 nm of TPD/6.0 ± 0.5 nm Alq3 (c) 50 ± 0.5 nm TPD/6.0 ± 0.5 nm Alq3 /50
± 0.5 nm PTCBI. All images have same height range of 5 nm. Low RMS roughness
in all cases suggests good wetting properties of deposited films.

ticular device trial, near the PTCBI absorption peak (λ = 532 nm). In addition, all
devices were tested in a nitrogen-ambient environment to prevent device degradation
during testing.
Individual current-voltage characteristics of the TPD/Alq3 /PTCBI devices varied
significantly in magnitude and curvature, making it difficult to assess the conduction
mechanisms. However, aggregating the current-voltage characteristics of the various
trilayer devices reveals an observable trend of increased dark current [Fig. 5-8]. This
evidence indicates charge transfer across either the TPD/Alq3 or Alq3 /PTCBI in-
terface. Comparison of the dark current levels of the trilayer device with those of a
bilayer consisting of TPD/Alq3 suggests that much of the charge transfer occurs at
that interface.
In order to explain the wide variation in the observed device characteristics, we
have to take a closer look at Alq3 . The electrical characteristics of organic devices
are strongly determined by the inter- and intramolecular charge transport mecha-
nisms, which are influenced strongly by the orientation and packing of the deposited
molecules. Thus, the variation in observed current values can mainly be explained by
the lack of control over a number of important parameters during deposition of Alq3 .
Although Alq3 is a well characterized material, it has a number of unique prop-
erties that can significantly affect morphological, optical, and electrical film charac-
teristics. Alq3 is calculated to have a strong, permanent dipole moment of 4.1 D

186
Figure 5-7: Scatter plot of nine separate trilayer device trials consisting of 50 ± 0.5
nm TPD/Alq3 /50 ± 0.5 nm PTCBI, with varying Alq3 thickness. Quantum efficiency
is plotted versus the thickness of Alq3 and suggests no definite trend. All quantum
efficiencies are obtained with the same applied field and wavelength and normalized
to the control (bilayer device) of that particular device trial.

187
Figure 5-8: Scatter plot of eight separate trilayer device trials consisting of 50 ± 0.5
nm TPD/Alq3 /50 ± 0.5 nm PTCBI, with varying Alq3 thickness. Dark current is
plotted versus Alq3 thickness and suggests an overall increase due to the presence of
Alq3 . All current levels are obtained at the same applied field and are plotted relative
to the current levels of the bilayer (control) device for that particular device trial.

188
and 7.1 D for its two geometric isomeric forms: meridional and facial, respectively
[17]. These properties can result in amorphous films with a net polarization and
significant concentrations of electron traps. Previous studies have demonstrated the
ability of light irradiation to have a significant effect on the orientation of deposited
Alq3 molecules, which can influence the energy level alignment of Alq3 relative to a
metal electrode [14]. It has been theorized that this effect is due to depolarization of
the molecules after photon absorption (i.e. rotation of the excited state or formation
of paired molecules whose dipoles cancel). In addition, substrate temperature has
been shown to have a substantial effect on the RMS roughness and conductivity of
amorphous, vacuum-sublimated Alq3 [18, 19]. There is no substrate temperature or
light control for the deposition chambers used in our experiments; the only variable
parameters are power delivered to the crucible (i.e. deposition rate) and chamber
pressure. Thus, it becomes highly likely that the deposited Alq3 films will present
varying electrical and optical properties and varying interfacial dipole formation be-
tween TPD and PTCBI from run to run. The inconsistency of measured current
levels in devices incorporating Alq3 necessitates the use of a less reactive, non-polar
material to form the cascading trilayer photoconductor.

TPD/NPD/PTCBI Trilayer

NPD is another well characterized material [20–23] that has a similar chemical struc-
ture to TPD, finding use as a hole transport and electron blocking layer in efficient
OLED structures [24, 25]. Given NPD’s wide bandgap, it is optically transparent to
illumination with a green LED [Fig. 5-9]. In addition, the cascading energy band
alignment formed from the introduction of NPD makes it a good candidate for the
trilayer devices. Although AFM data of NPD deposited on TPD was not obtained,
the similarities in chemical structure between the two materials suggests that a thin
film of TPD would present a similar free surface energy to deposited NPD molecules
as a thin film of NPD [26, 27]. Because evaporated molecules prefer to stick together
to form clusters, we can estimate that NPD will exhibit good wetting behavior when
deposited on TPD [28, 29].

189
Figure 5-9: Absorption spectra of TPD, NPD, and PTCBI (50 nm thick). The green
LED emission spectrum (λpeak = 532 nm), operated at an intensity of 40 mW/cm2 ,
is included for reference (dashed line). Inset: Suggested energy bands of TPD [14],
NPD [15], and PTCBI [16] based on photoemission spectroscopy data and optical
bandgaps.

190
Figure 5-10: Log-log current-voltage characteristics of 50 ± 0.5 nm TPD/NPD/50
± 0.5 nm PTCBI, with 0, 3.5, and 4 ± 0.5 nm of NPD. Arrows indicate increasing
NPD thickness and dashed red lines are guides that depict ohmic and trap-limited
conduction.

Figure 5-10 shows the current-voltage characteristics of a series of 50 ± 0.5 nm

TPD/NPD/50 ± 0.5 nm PTCBI trilayer devices with varying NPD film thickness. The
50 ± 0.5 nm TPD/50 ± 0.5 nm PTCBI bilayer device exhibits the lowest current levels
and demonstrates a dark current transition from ohmic conduction to trap-charge-
limited conduction at applied fields greater than 30 V/µm. Introduction of a 3.5 ±
0.5 nm NPD film as CSL, raises the dark current by up to two orders of magnitude
relative to the bilayer device, indicating strong charge transfer across the interfaces.
However, the light current exhibits a moderate factor of 2 increase. Increasing the
thickness of the NPD to 4.0 ± 0.5 nm yields another factor of 2 increase in both light
and dark current over the trilayer device with 3.5 ± 0.5 nm of NPD.

191
 Figure 5-11 compiles the dark and light current values for several trilayer device
trials of varying NPD thickness at an applied field of 10 V/µm. All of the values are
normalized to the bilayer device current values for that particular trial. As projected
by the simulations, there is an optimum NPD thickness as evidenced by the peak
in light current near 4 nm. The increase in light current due to increasing NPD
thickness is attributed to the separation of free carriers at the interface, which will
reduce recombination. However, when the NPD is too thick the free holes can no
longer diffuse to the TPD/NPD interface and the resulting device characteristics will
resemble that of a bilayer between NPD/PTCBI.
The relative dark current levels vary widely and do not show any observable trend.
This is again likely due to our inability to adequately control the film deposition
process. The distribution of trap states in the devices will largely affect the bilayer
dark current as there are not enough injected carriers to populate all of the traps over
the range of applied electric fields. Thus, the ratio of the trilayer dark currents to the
bilayer currents will be sensitive to changes in trap distribution, which are largely a
result of the deposition process.
More evidence supporting the positive influence of a CSL on photocurrent levels
can be found by measuring the spectral response of the trilayer devices. Figure 5-
12 displays the relative external quantum efficiency values at both the TPD and
PTCBI absorption peaks (λ = 368 nm and 532 nm, respectively) for various bias
voltages. The increase in photoresponse with increasing NPD thickness is consistent
with the data obtained in the steady-state current-voltage characteristics. It is also
important to note that increase in photoresponse due to increasing NPD thickness
is more pronounced with larger applied electric fields. The optimum NPD thickness
represented by the peak in EQE shifts from 4.5 nm at 10 V/µm to 4.0 nm at 20 and
30 V/µm.

5.4.2 Alternate Energy Band Alignments

In the previous section we demonstrated a trilayer heterojunction photoconductor

with cascading energy bands. However, there remain other energy band alignments

192
Figure 5-11: Scatter plot of three separate trilayer device trials consisting of 50 ± 0.5
nm TPD/NPD/50 ± 0.5 nm PTCBI, with varying NPD thickness. Light current (▽)
is plotted versus NPD thickness and suggests an overall increase due to the presence
of NPD up to 4 nm. All current levels are obtained at the same applied field and
are normalized to the current levels of the bilayer (control) device for that particular
device trial.

193
Figure 5-12: Scatter plot of three separate trilayer device trials consisting of 50 ±
0.5 nm TPD/NPD/50 ± 0.5 nm PTCBI, with varying NPD thickness. External
quantum efficiency is plotted versus NPD thickness at λ = 368 nm (∎) and at λ =
532 nm (◻), suggesting an increase in photoresponse due to the presence of NPD. All
measurements are obtained at the intensity and are normalized to the photocurrent
levels of the bilayer (control) device for that particular device trial. Larger electric
fields result in larger improvements.

194
that can be engineered in trilayer devices. The following trilayer experiments help to
expand our understanding of the trilayer device structure.

TPD/UGH2/PTCBI Trilayer

Ultrahigh energy gap hosts such as p-bis(triphenylsilyl)benzene (UGH2) are designed

to act as wide bandgap hosts for emissive dopant molecules in efficient OLEDs [30, 31].
In order to prevent charge and energy transfer between the dopant molecules and the
host, the optically transparent UGH materials are synthesized to have large singlet
and triplet energies (∼ 3 - 4 eV). Given the absorption spectrum and the wide bandgap
of UGH2 we can expect it to prevent excitons generated in PTCBI from dissociating,
resulting in reduced photocurrent levels [Fig. 5-13]. However, excitons generated
in the TPD film will continue to experience a Type-II heterojunction and efficiently
dissociate.
Figure 5-14 exhibits the current-voltage characteristics of 50 ± 0.5 nm TPD/UGH2/
50 ± 0.5 nm PTCBI trilayer devices, with varying UGH2 thickness. As is customary
with the bilayer device we see ohmic conduction in the light (illumination with green
LED) and trap-charge-limited conduction in the dark (black). A 2.0 ± 0.5 nm thick
CSL of UGH2 increases the dark current by many orders of magnitude relative to
the bilayer, indicating greater charge transfer between the Type-II heterojunction at
the UGH2/TPD interface. The light current of this trilayer device also increases by
a factor of 10, which is inconsistent with the band alignment at the UGH2/PTCBI
interface that should prevent exciton dissociation. One possible explanation for this
phenomenon is that 2 nm of UGH2 may be too thin to form a complete film, especially
if UGH2 has poor wetting behavior when deposited on TPD. A thicker, presumably
more complete film, of 7.0 ± 0.5 nm of UGH2 suppresses the light current by an
order of magnitude while marginally affecting the dark current relative to the bilayer
device.
AFM images of 10 ± 0.5 nm and 50 ± 0.5 nm thick UGH2 deposited on TPD
[Fig. 5-15] confirms the poor wetting characteristics of UGH2. This implies that
the trilayer devices with thin (2 nm) UGH2 films are primarily TPD/PTCBI bilayer

195
Figure 5-13: Absorption spectra of TPD, UGH2, and PTCBI (50 nm thick). The
green LED emission spectrum (λpeak = 532 nm), operated at an intensity of 40
mW/cm2 , is included for reference (dashed line). Inset: Suggested energy bands
of TPD [14], UGH2 [30], and PTCBI [16] based on photoemission spectroscopy data
and optical bandgaps.

196
Figure 5-14: Log-log current-voltage characteristics of a TPD/PTCBI bilayer device
(black), and two TPD/UGH2 (1.8 nm, 6.9 nm)/PTCBI trilayer devices (red and
green respectively).

devices with “islands” of UGH2 embedded at the interface.

The EQE plot vs. UGH2 thickness in Figure 5-16 corroborates the trends in the
current-voltage characteristics of the trilayer devices containing UGH2 - increased
photocurrent for thin films, reduced photocurrent for thick films. For UGH2 films
thinner than 4 nm there is an increase in photoresponse from both TPD and PTCBI.
This relative increase over the bilayer device is more prominent for higher applied
electric fields. However, beyond 4 nm the trilayer device shows a reduction in pho-
toresponse at all bias levels. Above 6 nm the EQE of the devices cannot be measured
with the available equipment as the light current levels are overtaken by the dark
current.

197
Figure 5-15: Set of AFM images of various films consisting of (a) 50 ± 0.5 nm TPD/4
± 0.5 nm UGH2 (b) 50 ± 0.5 nm of TPD/50 ± 0.5 nm UGH2. All images have same
height range of 100 nm. Island growth suggests poor wetting properties of UGH2
deposited on TPD.

In order to explain the increase in photoresponse due to exciton generation in

PTCBI, we can recall the current-voltage data from the previous chapter that demon-
strated a single film of PTCBI’s capability of dissociating excitons without the pres-
ence of an interface. While these free charges are traveling laterally along the interface
they will hop from PTCBI molecule to PTCBI molecule and either: 1) recombine
monomolecularly with the opposite charge on a PTCBI molecule 2) get blocked from
entering TPD by a UGH2 molecule or 3) hop (in the case of holes) onto a TPD
molecule. The first scenario is unlikely given the low intensity of illumination (<1
mW/cm2 ) and the low carrier mobility in the PTCBI film preventing free carriers from
interacting with each other. The more likely scenario is that carriers in PTCBI will re-
combine across the interface with an oppositely charged carrier in TPD, as evidenced
by bilayer intensity data from the previous chapter. On average, the UGH2 molecules
will reduce the likelihood of this bimolecular recombination process (by occasionally
blocking the hopping of a hole from PTCBI into TPD), effectively increasing carrier
lifetime and consequently carrier concentrations. This increased steady-state carrier

198
Figure 5-16: External quantum efficiency of TPD/UGH2/PTCBI trilayer device ver-
sus UGH2 thickness at λ = 368 nm (∎) and at λ = 532 nm (∎). Various applied
electric fields from 10 V/µm to 30 V/µm are indicated in the legend.

199
concentration results in an increased photoresponse as demonstrated in Figure 5-16
for thin films of UGH2 (below 4 nm). When UGH2 is made thicker and approaches a
complete film, monomolecular recombination in PTCBI will dominate and result in
a decreased photoresponse consistent with the observed data. Experimentally, this
would result in observing a linear dependence of photocurrent on intensity for thicker
UGH2 trilayer devices.

The increase in TPD photoresponse (at λ = 368 nm) is due to a similar phe-
nomenon. Unlike PTCBI and despite its higher mobilities, TPD’s amorphous nature
makes the probabilities of field-assisted exciton dissociation in a single TPD film
difficult to measure. However, TPD forms a Type-II heterojunction with UGH2 ac-
cording to the isolated energy band levels. Thus, the excitons generated in the TPD
layer will dissociate at the TPD/UGH2 interface to generate a hole in TPD and
an electron in UGH2. The electron in UGH2 can diffuse to the PTCBI (provided
UGH2 is thin enough) where it will have a longer effective lifetime due to the abil-
ity of UGH2 molecules to occasionally block the bimolecular recombination across
the TPD/PTCBI interface. For thicker UGH2 films, electrons in UGH2 will not
be able to diffuse to the PTCBI and the trilayer device effectively operates like a
TPD/UGH2 bilayer (when exciting TPD). Neglecting the effect of interfacial trap
states, bimolecular recombination across the interface will be limited by the smaller
band offset (either ∆LUMO or ∆HOMO) [32]. Thus, the thick UGH2 trilayer device
will have higher recombination rates than the TPD/PTCBI bilayer device due to the
∆LUMO of TPD/UGH2 being smaller than the ∆HOMO of TPD/PTCBI, resulting
in a reduced photocurrent.

After accounting for the EQE dependence on thickness we are left to explore the
dependence on bias voltage. If we consider the improvement in EQE relative to the
bilayer at both wavelengths as the bias voltage increases from 200 to 300 V, then
it is clear that this improvement is maximized for thinner UGH2 films below 4 nm
in thickness. This improvement is likely due to the onset of a different conduction
regime at the higher voltages in the thinner UGH2 trilayer devices. Comparing the
light current sweeps in Figure 5-14 of the 1.8 nm trilayer device with the bilayer device

200
gives experimental verification. At the higher biases the bilayer device maintains
ohmic conduction while the trilayer device begins to transition to another regime.
Because of the inability to operate these devices at higher applied fields, it is difficult
to infer whether SCLC or trap-limited conduction is dominant.

TPD/PTCBI/PbSe QDs Trilayer

Lead selenide (PbSe) nanocrystals (or QDs) have recently been investigated as active
materials in optoelectronic devices for their ability to emit and absorb infrared light
[33–35]. The PbSe QD cores can be coated with an insulating shell or capped with
organic ligands such as oleic acid to passivate surface states. The monodisperse
PbSe QDs used in the following experiments were synthesized following the procedure
outlined in [36, 37]. Oleic acid is used as the capping group to passivate surface states
and increase solubility in chloroform (of unknown concentration). The QD solution is
spun-cast onto a PDMS stamp and is deposited onto TPD/PTCBI bilayers following
a micro-contact printing process [38].
Given the energy band alignment of PbSe QDs relative to TPD and PTCBI,
we can expect excitons generated in PTCBI to dissociate at both of the Type-II
heterointerfaces [Fig. 5-17]. However, the PbSe QD’s broad absorption across the
visible spectrum makes it impossible to preferentially generate excitons in a single
layer. Thus, more careful analysis of the I-V curves will be necessary.
Figure 5-18 depicts the current versus voltage sweeps of a 50 ± 0.5 nm TPD/50
± 0.5 nm PTCBI device (blue) and a similar 50 ± 0.5 nm TPD/50 ± 0.5 nm PTCBI
device stamped with a 10 ± 0.5 nm thick PbSe QD film (red). Illumination is provided
by a green LED peaked at λ = 532 nm and is absorbed by both PTCBI and PbSe
QDs. The TPD/PTCBI bilayer shows typical photoresponse and ohmic conduction
in the light curve. The stamped bilayer exhibits a reduced dark current relative to the
bilayer. It is unclear at this point what phenomenon is causing this reduction. Likely,
the printing of QDs effectively embeds the dots into the PTCBI film in addition to
removing some of the layer. Potentially, the dots would then be close enough to
the TPD/PTCBI interface to negatively affect charge transport there, as untreated

201
Figure 5-17: Absorption spectra of TPD, PTCBI (each 50 nm thick) and PbSe (10
nm). The green LED emission spectrum (λpeak = 532 nm), operated at an intensity of
40 mW/cm2 , is included for reference (dashed line). Inset: Suggested energy bands of
TPD [14], PTCBI [16], and oleic acid-capped PbSe QDs [39] based on photoemission
spectroscopy data, charge transport data, and optical bandgaps.

202
Figure 5-18: Log-log current-voltage characteristics of a TPD/PTCBI bilayer device
(blue) and a TPD/PTCBI/PbSe QD trilayer device (red). Inset device cross sections
indicate relevant film thicknesses.

PbSe QDs are poor hole transporters [39]. Systematic surface analysis studies of
the stamped QD layer in conjunction with electrical characterization and varying
thicknesses of the PTCBI film would be required to explore the cause of reduced dark
current.

In addition to the reduced dark current there is an order of magnitude increase

in the light current that cannot be explained by the increased absorption from the
stamped PbSe QD film alone, which only increases the absorption by a factor of 2 at
λ = 532 nm [Fig. 5-19]. Also, the Type-I heterojunction between PTCBI/PbSe QDs
would prevent excitons generated in PbSe from dissociating into free carriers. Thus,
we must rely on spectrally resolved data to help explain this enhancement.

Observing the EQE data [Fig. 5-20] we see that the spectral response of the
PbSe QD trilayer device does not match its absorption profile. In fact, the response
appears to share the same spectral profile as that of the bilayer, largely enhancing

203
Figure 5-19: Absorption spectra of stamped film of PbSe QD on glass (black), on
TPD (red), on TPD/PTCBI (blue).

204
Figure 5-20: Spectral photoresponse of a TPD/PTCBI bilayer device (blue) and a
TPD/PTCBI/PbSe QD trilayer device (red) with an applied electric field of 10 V/µm.
Inset device cross sections indicate relevant film thicknesses.

the peaks of TPD and PTCBI absorption. This suggests that although PbSe may
be generating excitons by absorbing illumination, free carriers are not generated at
the PTCBI/PbSe QD interface, consistent with the proposed band alignments. The
increase in photocurrent from carriers generated in TPD and PTCBI is likely due to
longer lived free electrons generated at the TPD/PTCBI interface that diffuse into the
PbSe QDs. Thus, PTCBI effectively acts as a CSL in this case to prevent bimolecular
recombination between holes in TPD and electrons in the PbSe QDs.

205
5.5 Summary
The CSL concept has been demonstrated to enhance photocurrent levels in trilayer
devices. One possible explanation for this enhancement is the separation of free
carriers, which increases carrier lifetime. Careful time response measurements must
be carried out in order to support this hypothesis, and are left as areas for further
scientific exploration.
Device modeling is a useful tool to aid the device design process. However, exper-
iments using various materials have indicated the need to factor in other properties
in addition to the energy bands, such as polarizability and morphology. While it
is important to compare the results of the different trilayer device experiments, we
currently do not have enough data to support that discussion and must treat each
set of experiments on a case by case basis. Until better control over the energy bands
and morphology of deposited films is attained, engineering trilayer devices will re-
main a difficult trial-and-error process. Because as evidenced by the previous set
of experiments, differences in energy band alignment have a significant effect on the
recombination processes at the interfaces and consequently the electrical character-
istics. Systematic UPS studies and surface analysis techniques should be used in
conjunction with electrical characterization in order to elucidate the role of energy
bands and deposition conditions on overall photoconductivity in trilayer devices.

206
Chapter 5 References

[1] E. E. Mendez, G. Bastard, L. L. Chang, L. Esaki, H. Morkoc, and R. Fischer.

Effect of an electric field on the luminescence of GaAs quantum wells. Physical
Review B, 26(12):7101, December 1982.

[2] G. Bastard, E. E. Mendez, L. L. Chang, and L. Esaki. Variational calculations

on a quantum well in an electric field. Physical Review B, 28(6):3241, 1983.

[3] F. D. Sala, A. D. Carlo, P. Lugli, F. Bernardini, V. Fiorentini, R. Scholz, and J.

Jancu. Free-carrier screening of polarization fields in wurtzite GaN/InGaN laser
structures. Applied Physics Letters, 74(14):2002–2004, April 1999.

[4] E. F. Schubert. Delta doping of III–V compound semiconductors: Fundamentals

and device applications. Journal of Vacuum Science & Technology A: Vacuum,
Surfaces, and Films, 8(3):2980–2996, May 1990.

[5] E. F. Schubert. Optical properties of [delta]-doped doping superlattices. Surface

Science, 228(1-3):240–246, April 1990.

[6] G. Dohler. Doping superlattices (”n-i-p-i crystals”). Quantum Electronics, IEEE

Journal of, 22(9):1682–1695, 1986.

[7] S. M. Sze and K. K. Ng. Physics of Semiconductor Devices. John Wiley & Sons,
Hoboken, third edition, 2007.

[8] S. Sista, Y. Yao, Y. Yang, M. L. Tang, and Z. Bao. Enhancement in open circuit
voltage through a cascade-type energy band structure. Applied Physics Letters,
91(22):223508–3, November 2007.

207
 [9] B. P. Rand, D. P. Burk, and S. R. Forrest. Offset energies at organic semicon-
ductor heterojunctions and their influence on the open-circuit voltage of thin-
film solar cells. Physical Review B (Condensed Matter and Materials Physics),
75(11):115327–11, March 2007.

[10] J. Nelson, J. Kirkpatrick, and P. Ravirajan. Factors limiting the efficiency of

molecular photovoltaic devices. Physical Review B, 69(3):035337, 2004.

[11] A. C. Arango. A quantum dot heterojunction photodetector. M.S. thesis, MIT,

2005.

[12] H. Ishii, K. Sugiyama, D. Yoshimura, E. Ito, Y. Ouchi, and K. Seki. Energy-level

alignment at model interfaces of organic electroluminescent devices studied by
UV photoemission: Trend in the deviation from the traditional way of estimating
the interfacial electronic structures. IEEE Journal of Selected Topics in Quantum
Electronics, 4(1):24–33, February 1998.

[13] C. W Tang and S. A Vanslyke. Organic Electroluminescent Diodes. Applied

Physics Letters, 51(12):913–915, 1987.

[14] H. Ishii, N. Hayashi, E. Ito, Y. Washizu, K. Sugi, Y. Kimura, M. Niwano,

Y. Ouchi, and K. Seki. Kelvin probe study of band bending at organic semi-
conductor/metal interfaces: examination of fermi level alignment. physica status
solidi (a), 201(6):1075–1094, 2004.

[15] I. G. Hill and A. Kahn. Energy level alignment at interfaces of organic semicon-
ductor heterostructures. Journal of Applied Physics, 84(10):5583–5586, Novem-
ber 1998.

[16] W. R Salaneck, K. Seki, A. Kahn, and J. Pireaux. Conjugated Polymer and

Molecular Interfaces. Marcel Dekker, Inc., New York, 2002.

[17] A. Curioni, M. Boero, and W. Andreoni. Alq3 : ab initio calculations of its

structural and electronic properties in neutral and charged states. Chemical
Physics Letters, 294(4-5):263–271, September 1998.

208
[18] K. C. Chiu, Z. A. Jian, Y. Z. Luo, J. M. Chung, S. J. Tang, M. C. Kuo, J. L.
Shen, W. C. Chou, C. S. Yang, Zhenan Bao, and David J. Gundlach. Effects
of substrate temperature on properties of alq3 amorphous layers prepared by
vacuum deposition. In Organic Field-Effect Transistors V, volume 6336, pages
63361A–9, San Diego, CA, USA, 2006. SPIE.

[19] Z. Jian, Y. Luo, J. Chung, S. Tang, M. Kuo, J. Shen, K. Chiu, C. Yang, W.

Chou, C. Dai, and J. Yeh. Effects of isomeric transformation on characteristics
of alq3 amorphous layers prepared by vacuum deposition at various substrate
temperatures. Journal of Applied Physics, 101(12):123708–6, June 2007.

[20] M. Chun, T. Teraji, and T. Ito. Interfacial charge transfer and hole injection
in α-NPD organic overlayer-CVD diamond substrate system. Applied Surface
Science, 237(1-4):470–477, October 2004.

[21] M. Chun, T. Teraji, and T. Ito. Characterization of α-NPD thermally de-

posited on Hydrogen-Terminated Single-Crystal Chemical-Vapor-Deposited dia-
mond films. Japanese Journal of Applied Physics, 42:5233–5238, 2003.

[22] T. Chasse, C. I. Wu, I. G. Hill, and A. Kahn. Band alignment at organic-

inorganic semiconductor interfaces: alpha-NPD and CuPc on InP(110). Journal
of Applied Physics, 85(9):6589–6592, May 1999.

[23] T. P. I. Saragi and J. Salbeck. Organic heterostructure field-effect transistors

using c[sub 60] and amorphous spirolinked compound. Applied Physics Letters,
89(25):253516–3, December 2006.

[24] C. Yang, C. Lin, C. Wu, Y. Yeh, C. Cheng, Y. Kuo, and T. Chen. High-contrast
top-emitting organic light-emitting devices for active-matrix displays. Applied
Physics Letters, 87(14):143507–3, October 2005.

[25] J. S. Park, J. W. Lee, Y. M. Kim, S. J. Bae, J. Jang, J. K. Kim, and B. K. Ju.

Enhancement of efficiency for White-Organic Light-Emitting diode with a thin

209
 Electron-Blocking layer. Journal of The Electrochemical Society, 152(11):H196–
H199, November 2005.

[26] E. Bauer. Phaenomenologische theorie der kristallabscheidung an oberflaechen

i. Z. Kristallogr., 110:372–394, 1958.

[27] A. Gavezzotti. Molecular free surface: a novel method of calculation and its
uses in conformational studies and in organic crystal chemistry. Journal of the
American Chemical Society, 107(4):962–967, February 1985.

[28] P. Niyamakom. Influence of deposition parameters on morphology, growth and

structure of crystalline and amorphous organic thin films (the case of perylene
and NPD). PhD thesis, RWTH Aachen University, 2008.

[29] A. Farahzadi. Systematic Studies on Structural and Optical Properties of Organic

Thin Films on Inorganic Substrates. PhD thesis, RWTH Aachen University,
2007.

[30] X. Ren, J. Li, R. J. Holmes, P. I. Djurovich, S. R. Forrest, and M. E. Thompson.

Ultrahigh energy gap hosts in deep blue organic electrophosphorescent devices.
Chemistry of Materials, 16(23):4743–4747, November 2004.

[31] R. J. Holmes, B. W. D’Andrade, S. R. Forrest, X. Ren, J. Li, and M. E. Thomp-

son. Efficient, deep-blue organic electrophosphorescence by guest charge trap-
ping. Applied Physics Letters, 83(18):3818–3820, November 2003.

[32] B. Mazhari. Impact of interfacial barriers on recombination profile in bilayer or-

ganic light-emitting diode. Organic Electronics, 6(5-6):229–236, December 2005.

[33] J. S. Steckel, S. Coe-Sullivan, V. Bulović, and M. G. Bawendi. 1.3 um to 1.55

um tunable electroluminescence from PbSe quantum dots embedded within an
organic device. Advanced Materials, 15(21):1862–1866, 2003.

[34] V. Sholin, A.J. Breeze, I.E. Anderson, Y. Sahoo, D. Reddy, and S.A. Carter.
All-inorganic CdSe/PbSe nanoparticle solar cells. Solar Energy Materials and
Solar Cells, 92(12):1706–1711, December 2008.

210
[35] R. J. Ellingson, M. C. Beard, J. C. Johnson, P. Y., O. I. Micic, A. J. Nozik,
A. Shabaev, and A. L. Efros. Highly efficient multiple exciton generation in
colloidal PbSe and PbS quantum dots. Nano Letters, 5(5):865–871, May 2005.

[36] S. Coe-Sullivan. Hybrid Organic/Quantum Dot Thin Film Structures and De-
vices. Ph.d. thesis, MIT, 2005.

[37] D. C. Oertel. Photodetectors based on colloidal quantum dots. Ph.D. thesis, MIT,
2007.

[38] L. Kim. Deposition of Colloidal Quantum Dots by Microcontact Printing for

LED Display Technology. M. eng. thesis, MIT, 2006.

[39] T. S. Mentzel, V. J. Porter, S. Geyer, K. MacLean, M. G. Bawendi, and M. A.

Kastner. Charge transport in PbSe nanocrystal arrays. Physical Review B (Con-
densed Matter and Materials Physics), 77(7):075316–8, February 2008.

211
212
Chapter 6

Bilayer Heterojunction Sensors for

Vapor-phase Explosives Detection

There are very few personal

problems that cannot be solved
through a suitable application of
high explosives.

SCOTT ADAMS
Cartoonist

Having explored the multilayer heterojunction device structure and proven its
ability to be utilized as a photodetector, we now turn our attention towards adapting
the heterojunction photoconductor for chemical sensing applications. An increasingly
important, life-saving sensing application is explosives detection. On the battlefields,
improvised explosive devices (IEDs) have resulted in 54% of all coalition deaths in
Iraq and 59% in Afghanistan [1]. The U.S. military has responded with an intensive
program of technology development, spending $12.4 billion over the past three years
on counter-IED equipment, technology R&D, training, and other measures through
the Joint Improvised Explosive Device Defeat Organization (JIEDDO). “Sniffing”
technology, based on fluorescent polymers, has emerged as one of the most sensi-
tive tools in the military’s explosives detecting arsenal. Fluorescent polymer sensors
(such as the Fido XT—) have demonstrated effective ultra-trace, vapor-phase detec-

213
tion of TNT at security checkpoints and during vehicular sweeps. Standoff detection
capability can be achieved by attaching the Fido portable explosives detector onto
autonomous robots that can find hidden explosives at safe distances from the opera-
tors.
While the Fido system works well, it does so for only a limited number of explo-
sives. Synthesizing new polymers that are sensitive and specific to different analytes
can be challenging, but engineering those polymers to work within the Fido’s tight
operating tolerances can be even more difficult. Therefore, sensing platforms that uti-
lize fluorescent polymers would benefit from a more direct transduction mechanism
and simpler sensor architecture.

6.1 Device Model and Theory of Operation

We employ the bilayer heterojunction to develop an improved explosives detection

device with streamlined transduction and simplified architecture. The chemosensor
operates in a similar fashion to the bilayer heterojunction photoconductor, with the
added functionality of analyte detection in the EGL. This design can improve trans-
duction efficiency as compared to the current generation Fido XT—. Because the
heterointerface between the CTL/EGL converts the absorbed photons directly into
free carriers, a photodetector is no longer required. Therefore, we eliminate the opti-
cal coupling losses associated with collecting the transduction photons emitted by the
fluorescent polymer. By designing the energy band alignment between the CTL and
EGL to yield a Type-II heterointerface, any chemosensitive fluorescent polymer can
be used in this platform. Elimination of the photodetector results in a miniaturizable
device that enables wide spread distribution of these sensors for more comprehensive
protection.
Conceptually, the device consists of a Type-II bilayer heterojunction deposited
on planar electrodes that enables the application of an electric field in-plane with
the interface [Fig. 6-1]. The structure of the lateral heterojunction photoconductor
exposes the top surface of the EGL to the external environment. If the EGL is a

214
chemosensing film, the bi-layer photoconductor structure can be used as a chemire-
sistor. In a traditional chemiresistor, the introduction of a chemical analyte can either
cause changes in the chemosensitive film density [2] or can chemically adsorb to the
surface of the film while transferring charge [3]. Both processes result in changes in
device conductivity. State-of-the-art chemiresistors, attempt to enhance sensor speci-
ficity by functionalizing the active material, which directly affects film conductivity.
In contrast, the bilayer devices separate the chemosensing and conduction functions
across the EGL and CTL, respectively, enabling independent optimization of the two
device functions. This separation relaxes the constraint of using a single film to serve
both functions. Typically, functionalization of a polymer film (to make it specific to
a particular analyte) will negatively impact the film’s conductivity. In the bilayer
sensor, the resulting device inherits its electrical properties from the CTL and its
optical and chemosensing properties from the EGL. Therefore, chemosensitive fluo-
rescent polymers can be tailored to a particular analyte with a focus on specificity
and sensitivity without regard for conductivity, just as the CTL can be optimized for
carrier transport and collection without regard for specificity.

By incorporating a chemosensitive fluorescent conjugated polymer [4–6] as the

EGL, we envision a vapor-phase, solid-state chemical sensor architecture that utilizes
the sensing properties of the fluorescent polymer while maintaining the chemiresis-
tor’s simple transduction of a chemical to electrical signal. Substitution of the EGL
with a chemosensitive fluorescent polymer allows the exciton generation rate of the
film to not only depend on the intensity of incident light excitation, but also the pres-
ence of analytes in the local environment. The photoluminescence efficiency, η, of
chemosensitive fluorescent polymers can be strongly modulated by a chemical signal,
which is a signature of a change in the population of excitons that can radiatively
decay. Therefore, in the bilayer heterojunction chemosensor the change in exciton
population will correspond to a change in charge carrier density at the EGL/CTL
heterojunction interface, which will be measured as a change in photoconductivity.

Because the fluorescent polymer produces excited charge carriers from its photo-
generated excitons, there remains the possibility that the carrier lifetimes can exceed

215
the transit times, leading to amplification of the transduced chemical signal. The pho-
toconductive gain is related to the incident light intensity. This characteristic implies
that the sensor’s sensitivity can be tuned by changing the incident light signal, giving
the sensor a potentially large dynamic range. In addition, the chemosensitive lateral
heterojunction photoconductor can be driven with a modulated light signal, allow-
ing for locked-in measurements of the photogenerated, transduced chemical signal.
These AC measurements remove the contributions from the background DC current
flowing through the CTL, relaxing the requirement of high signal-to-noise ratios for
extracting useful measurements from chemiresistors.

6.2 Material Sets and Fabrication

In practice, the bilayer heterojunction is realized by spin-casting a chemosensitive,
fluorescent polymer on top of a metal oxide. The polymer (insulator) acts primarily
as an exciton generation layer (EGL) and the metal oxide (semiconductor) serves as
a charge transport layer (CTL). Before investigating the electrical characteristics of
the bilayer sensors, it is useful to consider the materials and processes involved in
making the devices.

6.2.1 Fluorescent Conjugated Polymers

The advantage of using chemosensitive conjugated polymers (CPs), instead of func-

tionalized small molecule or oligomeric materials, is the ability to collectively respond
to small perturbations through the polymer backbone [7]. In contrast to CPs, the
luminescent properties of an individual chemoresponsive molecule are only influenced
by the local environment. However, electrical transport properties of CPs are delo-
calized over each conjugated polymer strand, which can yield an amplified response
to analytes in a luminescence quenching system.
In order for signal amplification to occur, the properties of a large number of
lumophore units must simultaneously change with the presence of a single analyte
molecule. The π-electrons that are delocalized along the polymer chain create ex-

216
Figure 6-1: Illustration depicting cross-sectional views and energy band diagrams of a
chemosensitive bilayer heterojunction photoconductor. These illustrations highlight
the impact of the analyte on carrier concentrations at the EGL/CTL interface. The
sample graph of photocurrent vs. time demonstrates the effect of the analyte on
device photoconductivity.

217
tended states that form energy bands (conduction and valence) along the polymer
backbone. Excitons can travel through these energy bands and the polymer film by
either Forster (long-range) or Dexter (short-range) energy transfer. These efficient
transport mechanisms allow the excited states to migrate among many more attached
lumophore units. In the case of fluorescence quenching, any analyte molecule that is
bound to a receptor site within the exciton diffusion length will create an energy trap
to quench the excited state, preventing it from emitting light. In other words, the de-
localization of excitons along the CP backbone effectively amplifies the luminescence
quenching.

Initial luminescence quenching demonstrations in CPs used dilute solutions with

integrated receptor units [8], restricting energy transport along a one-dimensional
polymer chain. Thin films of conjugated polymers increase sensitivity by allowing
excitons to sample receptor sites in three dimensions (inter-chain transport) rather
than redundantly sampling sites along the same backbone (intra-chain). In the solid
state, however, conjugated polymers often form excimers or exhibit π-stacking, both
of which can decrease photoluminescent (PL) efficiencies [9]. In one effective solution
to this problem, rigid three-dimensional pentiptycene molecules are incorporated into
the polymer backbone. This rigid scaffolding serves three important roles: 1) to pre-
vent aggregation of the polymer, increasing PL efficiency, 2) to increase porosity of
thin films for faster analyte diffusion, and 3) to create cavities that can selectively bind
analytes based on size. This approach has been used with great success for the selec-
tive detection of electron-deficient nitroaromatic compounds such as trinitrotoluene
(TNT) [10]. By tuning the electrostatic interaction and film porosity/thickness, it
is possible to detect TNT at sub-equilibrium vapor pressures (10-100 femtograms),
which is comparable to canine olfactory performance [11].

To further enhance the three dimensional diffusion of excitons, energy migration in

three dimensions has been explored through the use of multi-layer films deposited by
the Langmuir-Blodgett technique. The construction of these multi-layer films creates
aligned polymer backbones that channel exciton migration towards lower bandgap
materials deposited on top [12]. This effectively creates a staircase of energy levels,

218
making it energetically favorable for the excitons to transfer their energy from layer
to layer towards smaller bandgap materials. Thus, if the energy migration efficiency
is high, then the probability that excitons will revisit molecules on the same chain
will be reduced. In addition, some research has centered on trying to extend excited
state lifetimes to increase the probability of encountering a bound receptor site. This
vein of research continues to yield promising materials for sensing applications [13].
Mathematical modeling of CP sensors has shown that the sensitivity of the CP to
changes in analyte concentration has a non-monotonic dependence on the number of
receptors on each polymer backbone, increasing with the number of receptors up to
a critical value at which point it decreases. In addition, the dynamic range of CP
sensors is strongly dependent on the number of polymer backbones and the number
of receptors on each backbone [14].

Lasing Action in Fluorescent Conjugated Polymers

As previously discussed, one approach to increasing the signal amplification in CP

chemosensors is to change the energy transport properties by increasing excited state
lifetimes or improving order in the polymer backbone to improve mobilities. Another
approach that can complement other sensitivity enhancement schemes is to use the
lasing action of CPs to generate greater differential emission signals.
The sensitivity enhancement due to lasing action is a general principle that is
independent of the optical cavity architecture, and it has demonstrated more than a
30-fold increase in detection sensitivity [4]. One of the challenges with this technique
is to use a polymer that exhibits high durability, enabling operation at high incident
pumping powers in ambient atmosphere. The polymer used in the study cited above
incorporated pendant aromatic rings that result in hydrocarbon side chains running
parallel to the polymer backbone. These side chains effectively encapsulate the back-
bone, preventing the aggregation that leads to self-quenching, and may be the source
of the polymer’s stability in thin films. Another feature of the polymer design is the
extensive π-orbital interactions that promote longer exciton diffusion lengths (∼ 7-15
nm).

219
 Another challenge with the lasing approach is the push to use thinner CP films
for optimal sensitivity. Unfortunately, thinner films cannot sustain the waveguided
modes needed for amplified stimulated emission (ASE) and lasing action. Thinner
films are desirable because TNT and DNT (electron-deficient molecules) bind strongly
to electron-rich CPs so that short exposures prevent analytes from diffusing into
the bulk of the film [9]. Therefore, thin CP films ensure that the effect of exciton
quenching at the surface is not overshadowed by emission from the unquenched bulk.

Many CPs, with high solid state PL efficiencies, exhibit low lasing thresholds
[15, 16]. This is due mainly to the large Stokes shift that minimizes light reabsorption
losses within the film. In addition, the molecular disorder inherent in organic materials
creates a dispersion of energies through which excitons can diffuse downhill in energy.
Exciton diffusion complements the energy relaxation due to molecular reconfiguration
and increases the observed Stokes shift [17]. In general, the Franck-Condon principle
is the approximation that electronic transitions occur much faster than nuclei can
react. Following absorption and the nearly instantaneous promotion of an electron to
an excited state, the surrounding nuclei must rearrange their energetic configurations
to reach an equilibrated excited state. After emission, the nuclei are again in a
nonequilibrium state and must change configuration to accommodate the relaxed
electron. In all organic materials, the nuclear relaxation process causes the observed
red shift in the emission profile relative to the absorption profile. These processes
can be modeled using a four level system [Fig. 6-2], where the different energy levels
represent the following molecular configuration energies: 1) the electronic and nuclear
ground state of the molecule, 2) the electronic ground state and the nuclear excited
state, 3) the electronic excited state and the nuclear ground state and 4) the electronic
and nuclear excited state.

The model idealizes a CP thin film by disregarding energetic disorder (energy

levels instead of bands) and places the film within an optical cavity that is excited by
an ideal laser source (single wavelength, constant power). Following the treatment in
[4], the following approximations can be made. First, the rates of transition from level
4 to level 3 and from level 2 to level 1 are much faster than the transition from level 3

220
Figure 6-2: Schematic of a 4-level system representative of amplified stimulated emis-
sion in a conjugated polymer. Ex are the energies, Nx are the electron populations,
and Rxy are the rates of transition between the different levels of the 4-level system.
After [4].

to level 2. This relative rate has two effects: 1) electrons are approximated as being
excited directly into level 3 and 2) the electron population in level 2 is assumed to be
negligible. The transition from level 1 to level 4 occurs when an exciton is generated
by the optical pumping rate G. The transition from level 3 to 2 occurs through one of
four processes: spontaneous emission (with rate Rsp ), stimulated emission (with rate
Rst ), spontaneous non-radiative decay (with rate Rnr ), and quenching (with rate Rq ).
Thus, the following rate equation governs the population in level 3 (N3 ),

dN3
= G − R32 , (6.2.1)
dt

where
⎧
⎪ Rst + Rsp + Rnr + Rq , above lasing threshold
⎪
⎪
R32 = ⎨
⎪
⎪ Rsp + Rnr + Rq , below lasing threshold
⎪
⎩
In the absence of lasing under steady-state conditions (all time derivatives equal zero),
the exciton generation rate is equal to the transition rate from level 3 to level 2, which

221
is the sum of all the relaxation rates:

G = R32 = Rsp + Rnr + Rq . (6.2.2)

Given that the population in level 2 is negligible (N2 ∼ 0), the constant population
difference between level 3 and level 2 can be represented by

1 R32 1 1 1
N = Gτ32 , = = + + ,
τ32 N τsp τnr τq

where the τ ’s are the lifetimes associated with each relaxation pathway.

To achieve lasing, N must surpass the required threshold population difference

(Nth ) by having the gain due to stimulated emission exceed the optical cavity losses.
The optical cavity losses can be categorized by losses due to the gain medium and
losses due to the cavity. We can evaluate the losses in the gain medium by first noting
that absorption of the lasing wavelength is negligible due to the large Stokes shift in
CPs as discussed earlier. Thus, only the mirror losses contribute. Assuming that
both mirrors are identical (symmetric cavity), the distributed cavity loss coefficient
(γ) can be written as,
1 1
γ= ln ( ) , (6.2.3)
d Γ
where d is the cavity length and Γ is the reflectivity of the mirrors [18]. Below
threshold, the effect of stimulated emission on the population is small because the
photon flux (Φ) present in the cavity at the lasing mode is relatively low. So, the
“small-signal” gain, g(λ), is given by,

g (λ) = N σ32 (λ) = Gτ32 σ32 (λ) , (6.2.4)

where σ32 (λ) is the cross section for stimulated emission at wavelength λ. It is
important to note that the product of the photon flux with the cross section for
stimulated emission between levels 3 and 2, yields the stimulated emission rate, Rst :

Rst = Φσ32 (λ) . (6.2.5)

222
This equation shows that the stimulated emission rate is not a function of the quench-
ing rate and will remain the same in the presence of an analyte. Thus, the lasing
threshold should occur when,

γ = g (λ)
1 1
⇒ ln ( ) = Gτ32 σ32 (λ) (6.2.6)
d Γ
1 1
⇒ Gth = ln ( ) .
dτ32 σ32 (λ) Γ

From this expression, it is clear that the threshold pumping rate will be modified by
the quenching rate through τ32 alone.

When lasing, nearly every exciton generated above threshold contributes to optical
power in the lasing mode, as can be seen by the following relation,

Plaser ∝ Ppump − Pth . (6.2.7)

Pth is the pumping optical power necessary to reach threshold and is linearly propor-
tional to the threshold generation rate of excitons, Gth . Thus, if the introduction of
an analyte causes the overall carrier lifetime (τ32 ) to change, then the quenched laser
power becomes,
′ ′ ′ τ32
Plaser ∝ Ppump − Pth , Pth = ′
τ32 Pth
′
where τ32 is the carrier lifetime associated with the transition from level 3 to 2 after
quenching. Since quenching will reduce this lifetime by introducing another relaxation
pathway, the pumping threshold will shift to higher energies. Now, it is possible to
specify the fractional change in intensity of the lasing CP film in the presence of
analyte using a fixed pump power,

′ −P Pth (1 − ττ32
′ )
∆I Plaserlaser 32 Pth (1 − β)
= = = , (6.2.8)
Io Plaser Ppump − Pth Ppump − Pth

223
where,
τ32
β= ′
τ32 .
This expression can be compared to the fractional change in PL intensity of the CP
film to specify the sensitivity enhancement due to lasing action. The PL efficiency
(η) of the CP film can be written as,

Rate of photon emission Re τ32

η= = = . (6.2.9)
Rate of exciton generation G τe

Considering that the PL intensity is directly proportional to the PL efficiency, the

fractional change in PL intensity becomes,

η′ − η
τ ′ 32 τ32
∆I τe − τ e 1−β
= = τ32 = . (6.2.10)
Io P L η τe β

From these expressions it becomes clear that operating the chemosensing CP film near
its lasing threshold post-exposure will yield enhancements in fluorescence quenching,
thereby enhancing sensitivity. Simple inspection of Figure 6-3 shows that the optimal
sensitivity is obtained when the pump power of the CP is at the onset of amplified
stimulated emission (ASE). The devices used for the initial lasing enhanced sensitivity
studies were easily fabricated and convenient for evaluating and comparing the lasing
characteristics of different CPs. However, these low quality laser cavities give high loss
coefficients and higher lasing thresholds. Higher Q-factor laser cavities would lower
device operating powers, lengthen CP film lifetimes, and reduce the lasing threshold,
allowing for thinner CP active layers. These thinner active layers are expected to
provide increased sensitivity and faster analyte response times.

6.2.2 Metal Oxides

In order to utilize chemosensitive fluorescent polymers in a bilayer structure, the poly-

mer must be deposited out of solution to form the topmost layer and EGL. This format
ensures that the fluorescent polymer’s exciton generating capabilities are exposed to
the changing environment. Therefore, the underlying CTL must consist of a material

224
Figure 6-3: An example sketch of the lasing threshold increase after exposure of the
CP film to a quenching analyte, such as TNT. The change in the output emission
intensity is largest at pump energies near the lasing threshold of analyte-exposed
films. After [19].

225
that is compatible with solvent processing and can withstand the spin-coating pro-
cess. Metal oxides make a good choice for the CTL for several reasons, including high
carrier mobilities, chemical/mechanical robustness and tunable conductivity. Metal
oxides typically have the chemical formulas Mx Oy or Mx Ny Oz , where M and N are
metallic elements and O is oxygen [20]. Metal oxides are used in a wide range of ap-
plications including jewelry, transparent conductors, gas sensors, and catalysts. All of
the optical, electrical, and mechanical properties of metal oxides can be traced back
to the chemical bonds formed between the atoms of metal and oxygen. There are
two types of bonding in these materials: ionic and covalent. Ionic bonds form as the
result of the Coulombic attraction between oppositely charged ions, occurring only if
the energy of the molecular unit is lower than the energy of the separated ions. These
bonds typically result in insulating materials with a strong, non-directional bonding
pattern. Covalent bonds form as the result of electron sharing between neighboring
atoms. This configuration results in directional bonds that depend on which electronic
orbitals are sharing the electrons. Therefore, covalent bonds can range in strength
and lead to materials that are metallic to semiconducting in nature.
Some of the methods used to deposit amorphous metal oxide thin films include
molecular beam epitaxy (MBE) [21], chemical vapor deposition (CVD) [22], atomic
layer deposition (ALD) [23], and sol-gel [24]. In this work we use radio-frequency (RF)
magnetron sputtering to deposit metal oxides due to the relatively fast deposition
rates and high quality films that are achievable.

6.3 Proof of Concept: ZZ-TOP and SnO2

In fabricating the first proof-of-principle device, we used a robust CP termed ZZ-

TOP as the chemosensitive EGL [Fig. 6-4]. This CP has demonstrated stable lasing
ability in air and good sensitivity to TNT vapor [4, 19], making it a robust, practical
material to use. Pairing this polymer with an appropriate CTL to make a Type-II
heterojunction requires first the determination of its HOMO and LUMO levels.
The HOMO level can be obtained experimentally through a technique known

226
Figure 6-4: Emission spectra of a 17 ± 1 nm thin film ZZ-TOP before (light green)
and after (dark green) exposure to saturated TNT vapor. Inset: Digital image of
ZZ-TOP thin film spin-coated on a glass substrate and excited by UV illumination.

227
Figure 6-5: (a)Helium I UPS data from ITO sample (100 Ω/◻) and ITO coated with
10.0 ± 0.5 nm film of ZZ-TOP polymer with no bias relative to reference electrode.
(b)Energy level schematic of UPS spectra superimposed over the energy bands of the
sample materials. The work function of ITO (Φm ) can be extracted from the onset of
photoemitted electron collection. Note the small vacuum level shift (∆) and the large
energy difference between the ZZ-TOP HOMO and the ITO Fermi level (εFHOM O ).

as ultraviolet photoemission spectroscopy (UPS). Data was gathered using a multi-

technique surface analysis system for UPS and ultra high vacuum scanning probe
microscopy (Omicron Nano Technology GmbH). A 0.1 mg/mL concentration solution
of ZZ-TOP is spin-coated onto an indium tin oxide (ITO) chip at 6000 RPM for 60
seconds, resulting in a 10 ± 0.5 nm thick polymer film as measured by AFM. To
reduce the effects of adsorbed surface contaminants on the polymer film such as
oxygen, water, or excess solvent, samples were left under high vacuum conditions
(< 10-10 Torr) in the UPS chamber over a 24 hour period. Ultraviolet light, from
a Helium lamp emitting at 21.2 eV (He I radiation), illuminated the sample. The
low photon energies in UPS means that deep core electron levels cannot be excited.
Therefore, only electrons in the valence band or shallow core levels are accessible with
this technique.
Figure 6-5(a) highlights the important areas of the UPS spectra at both energy

228
extremes, more specifically where the photoemitted electrons have the least kinetic
energy and the most kinetic energy. The low kinetic energy side is used to determine
the work function of a reference material and the vacuum level shift of another ma-
terial relative to the underlying reference material. The small vacuum level shift of
0.2 eV combined with the energy difference of 2 eV between ZZ-TOP’s HOMO and
ITO’s Fermi level yields a HOMO level of 6.5 ± 0.5 eV for ZZ-TOP [Fig. 6-5(b)]. The
large uncertainty in the data stems from variation in the UPS measurement itself and
from variations in the surface condition of the polymer film.

The LUMO level of ZZ-TOP can be determined empirically by adding the optical
band gap energy to the HOMO level. The optical bandgap energy can be inferred
from the onset of ZZ-TOP’s absorption spectrum, which starts at λ = 480 nm corre-
sponding to an energy of 2.5 eV [Fig. 6-6]. Therefore, the HOMO/LUMO levels of
ZZ-TOP are 6.5 eV and 4.0 eV, respectively.

Thin film absorption spectra of ZZ-TOP and SnO2 are captured using a UV-
VIS-NIR spectrophotometer (NKD-8000, Aquila Instruments). Pairing ZZ-TOP with
SnO2 as CTL yields a Type-II heterojunction with some overlap in the absorption
spectra [Fig. 6-6]. Because the ZZ-TOP absorption peaks at blue (shorter) wave-
lengths, it is difficult to find materials with wide enough bandgaps to prevent overlap
of absorption spectra. The UGH materials would certainly be better from an ab-
sorption standpoint, but the additional requirements of a LUMO greater than 4.0 eV
and solution processing of ZZ-TOP make those wide bandgap materials incompatible
with this particular polymeric system.

Once the optimal components were determined, the designed sensing devices were
fabricated. To deposit the SnO2 , an Ar/O2 gas mixture was used for the sputtering
process at a flow ratio of 7 sccm to 3 sccm. The residual gas pressure in the sputtering
chamber was roughly 4 mTorr. At a constant RF power of 160 Watts, the deposition
rate reached a maximum of 1.0 Å/s. To remove any target contamination and to
prepare the target, a 100 nm thick layer was presputtered at the desired process gas
flow and power. The film morphology of SnO2 films deposited by RF magnetron
sputtering depends strongly on the growth pressure. At higher process pressures,

229
Figure 6-6: Normalized absorption spectra of a 17.0 ± 0.5 nm thick film of ZZ-TOP
(green), a 100.0 ± 0.5 nm thick film of SnO2 (grey), and the emission spectra of a blue
(λpeak = 460 nm) and UV LED (λpeak = 385 nm). Inset: Suggested energy bands of
ZZ-TOP and SnO2 [25] based on photoemission spectroscopy data and optical band
gaps.

230
columnar structures are formed, whereas at lower pressures the film consists of closely
packed fibrous grains [26]. Figure 6-7 shows an AFM image of the deposited SnO2
films which suggests a polycrystalline film with an average crystallite size of 50 nm
in diameter.

In semiconducting metal oxides, oxygen vacancies that act as electron donors (n-
type semiconductors) or excess oxygen atoms that act as electron acceptors (p-type)
are responsible for the electronic conductance. Doping the SnO2 films with 30% oxy-
gen by volume fills oxygen vacancies and reduces conductivity. The practical benefit
is that the lower conductivity films are now within measurement range of available
testing equipment (Keithley 6487), whereas the undoped films were so conductive as
to overload the current measurement.

Figure 6-8 plots the steady state I-V characteristics, both with and without blue
LED illumination (Lamina, BL-3000), from a series of devices containing either single
films or a bilayer heterojunction of SnO2 and ZZ-TOP. The blue LED light engine,
powered by a sourcemeter (Keithley 2400), provides narrow band excitation (FWHM
50 nm) with intensities up to 30 mW/cm2 as measured with a calibrated photode-
tector (Newport, 818-UV). The emission spectrum of the blue LED [Fig. 6-6] was
measured using a spectrometer (Ocean Optics, USB4000). Device current-voltage (I-
V) characteristics were measured in a dark box using a picoammeter (Keithley 6487).
All measurements were conducted in a nitrogen ambient environment provided by a
glovebox (MBraun).

Notice that all of the curves are essentially ohmic in nature across the voltage
range. Electrons are the majority carrier in SnO2 making it n-type. That fact coupled
with the LUMO level of SnO2 suggests close Fermi level alignment with that of the
gold electrodes, which is consistent with the ohmic conduction observed. The device
containing a single (17.0 ± 0.5) nm thick film of ZZ-TOP generates little measurable
photoresponse, which reflects 1) the low absorption in the ZZ-TOP film and 2) the
inability of ZZ-TOP to dissociate excitons efficiently. The thin film of SnO2 does
generate a moderate photocurrent in response to the blue illumination, which can be
discerned by comparing the solid black line to the dashed black line. A thick 100

231
Figure 6-7: AFM images of various fluorescent polymers spin-coated onto glass sub-
strates and SnO2 films coated with those same polymers. All images have the same
height range of 20 nm.

232
Figure 6-8: Log-log plot of the current-voltage characteristics from a series of lateral
photoconductor devices using SnO2 and ZZ-TOP. Current-voltage sweeps are taken
in the dark (dashed) and under illumination (solid) through the glass substrate from
a 30 mW/cm2 Blue LED (λpeak = 532 nm). The dashed red line is a guide for the eye
and follows Ohm’s law. The inset cross-sectional diagrams describe the thicknesses
and device structures.

233
nm film significantly contributes to the absorption tail that extends into the LED’s
emission spectrum. The photogenerated excitons will diffuse to crystallite boundaries
[Fig. 6-7] where they can dissociate to contribute to the film’s photoconductivity.
Therefore, despite the absence of a heterojunction, the SnO2 film is able to generate
a photoresponse.

Comparing the dark currents, it is evident that ZZ-TOP has the lowest overall
conductivity. This presents one of the challenges of incorporating ZZ-TOP (and other
fluorescent polymers) into a traditional chemiresistor structure, because low conduc-
tivity makes the electrical signal more susceptible to environmental noise. On the
other hand, the SnO2 has a high carrier mobility (∼ 1 cm2 V-1 s-1 ) consistent with the
polycrystalline nature of metal oxides, resulting in a higher dark current curve. The
combined SnO2 /ZZ-TOP bilayer structure has nearly the same dark current as the
single film of metal oxide. This indicates that the level of charge transfer at the inter-
face is less than the intrinsic carrier concentration of the SnO2 film. Given the small
vacuum level shift between ITO and ZZ-TOP [Fig. 6-5], a negligible concentration of
transferred charge relative to the intrinsic metal oxide carrier concentration is to be
expected.

The increased photocurrent of the SnO2 /ZZ-TOP device relative to the SnO2
device can be explained by the presence of the Type-II heterojunction between ZZ-
TOP and SnO2 . Both ZZ-TOP and SnO2 absorb the blue light and generate excitons,
which produces enough carriers to make the interface the most conductive pathway
in the device. This dataset suggests that the SnO2 /ZZ-TOP heterojunction interface
is efficiently dissociating excitons and generating charge under illumination, which is
transported as electrons through SnO2 .

When introducing saturated TNT vapor onto the devices, there was no significant
change in any of the current-voltage characteristics. This observation leads to the
assertion that the differential signal is too small to be measured with steady-state I-V
sweeps. The main factor accounting for this result is the absorptive ability of the thick
SnO2 relative to the thin ZZ-TOP, which affects the ability to preferentially excite
ZZ-TOP by using a broad emission spectrum light source (i.e. LED). This absorption

234
makes it difficult to pick out the differential signal due to ZZ-TOP quenching given
the competition from the baseline photoresponse of the SnO2 . Therefore, a locked-in
photocurrent measurement with monochromatic excitation is needed to pick out the
differential current signal caused by quenching.

The testing setup consists of a 1000 W Xenon arc lamp (Oriel) incident on a
monochromator (Acton SpecraPro 300i) with a diffraction grating (600 L/mm) blazed
at λ = 500 nm providing monochromatic illumination across the visible spectrum. A
computer-controlled rotary filter wheel prevents secondary harmonics from exciting
the samples. In addition, the monochromatic light is fed into a glass fiber which is
coupled into a nitrogen-filled glovebox. The light is optically chopped at 40 Hz and
locked-in with a digital lock-in amplifier (Stanford Research Systems, SR830), which
selects only the photocurrent component of the input signal. Saturated TNT vapor is
produced by solid TNT particles placed into a bubbler (Chemglass). Air flow through
the headspace of the solid TNT is provided by a diaphragm pump and is controlled
by a flow meter, allowing for TNT vapor delivery rates ranging from 0 to 500 cc/min
[Fig. 6-9].

Wavelength resolved data enables a more detailed analysis of device operation.

Figure 6-10 shows the spectral response of a series of devices containing SnO2 and
ZZ-TOP. The SnO2 film shows a significant photoresponse out to 500 nm consistent
with its absorption profile, while the ZZ-TOP’s photoresponse could not be mea-
sured. Both results are consistent with the previously obtained current-voltage data.
Attempts to boost the relative absorption of the ZZ-TOP over SnO2 by depositing a
thinner SnO2 film yielded a more photoresponsive device (grey curve). This suggests
that thinner SnO2 films form smaller crystallites, which generates a larger concentra-
tion of defect sites for enhanced exciton dissociation in the SnO2 bulk [27]. The bilayer
device exhibits an improved photoresponse relative to SnO2 by more than an order
of magnitude, again consistent with the I-V data. From the spectrally resolved data
it is clear that the improvement in photoresponse occurs across both the SnO2 and
ZZ-TOP absorption spectra. However, the large photoresponse of the SnO2 clearly
dominates, with only a slight shoulder at λ = 456 nm to indicate the presence of the

235
Figure 6-9: Illustration of photocurrent measurement setup incorporating a TNT
vapor delivery mechanism.

236
ZZ-TOP.

After exposure to saturated TNT vapor, the bilayer sensor shows a clear reduction
in the shoulder centered at λ = 456 nm. This shoulder peaks at the point where the
ZZ-TOP absorption relative to SnO2 absorption is at a maximum. Practically, this
means that the shoulder peak is red-shifted from the polymer’s actual absorption
peak. The decrease in the quantum efficiency indicates quenching of the polymer’s
ability to generate excitons and consequently free carriers. Fixing the wavelength at
λ = 456 nm and measuring the photocurrent versus time provides further evidence for
a reduced photocurrent signal due to the introduction of TNT vapor. Aside from the
reduced photoresponse of ZZ-TOP, the rest of the bilayer profile remains the same,
indicating that TNT cannot diffuse to the SnO2 film or that SnO2 does not respond
to TNT. Further testing with a single film of SnO2 reveals that it has no response
to saturated TNT vapor. Thus, we have successfully demonstrated the transduction
of a chemical signal into a differential photocurrent, and provided more evidence to
support the bilayer sensor device model developed earlier.

Reusability of the fluorescent polymer for multiple detection events is directly re-
lated to the strength of the interaction between the analyte and the polymer. Weaker
interactions such as physisorption impart a greater ability to refresh the polymer (i.e.
with a clean gas stream), but it also results in a weaker quenching response. Strong
interactions such as chemisorption or charge transfer result in strong quenching but
make the removal of adsorbed analyte more difficult. In the SnO2 /ZZ-TOP sensors,
the TNT-polymer interaction is most likely a charge transfer event, based on the en-
ergy levels of the materials involved, and leads to significant difficulty in refreshing the
polymer. Some methods of removing the TNT from the polymer include heating to
desorb the analyte and rinsing with methanol to dissolve the TNT. Another effective
recovery method involves pulling vacuum (30 inHg) on the sample for 30 minutes.
While all of these methods represent effective ways of recovering the polymer film,
heating has been shown to be the most practical for real-world applications where
portability and speed of response are critical.

237
Figure 6-10: Semi-log spectral response characteristics for devices containing SnO2
and ZZ-TOP. All devices have same applied electric field (5 V/µm). 100.0 ± 0.5 nm
SnO2 film (black) is less quantum efficient than 20.0 ± 0.5 nm SnO2 film (grey). 100.0
± 0.5 nm SnO2 /17.0 ± 0.5 nm ZZ-TOP before TNT exposure (green) is plotted against
same device after TNT exposure (dark green). Inset: Real-time plot of photocurrent
decay at a fixed wavelength of λ = 456 nm upon TNT vapor exposure. TNT vapor
is introduced at time t = 0 s.

238
Figure 6-11: (a) Normalized emission spectra of various chemosensitive fluorescent
polymers. Inset: photograph of UV-excited polymers in solution and thin film. (b)
Chemical structures of various chemosensitive fluorescent polymers.

6.3.1 Developing a Sensor Platform Technology

One of the aforementioned benefits of the bilayer heterojunction photoconductor is

that it can act as a sensor platform to incorporate various chemosensitive fluorescent
polymers. Proving that this effect is not unique to the combination of ZZ-TOP and
SnO2 , we have experimented with a number of different TNT-sensitive fluorescent
polymers [Fig. 6-11].
Figure 6-12 compiles the results of photocurrent measurements performed on ZZ,
ED4, and HW polymers before and after exposure to saturated TNT vapor. All of
the polymers used were solvated in toluene at a concentration of 1 mg/mL and spun
at 3000 rpm for 60 seconds. The polymer/SnO2 sensors exhibited a differential signal
due to quenching of the polymer and an improved bilayer photocurrent relative to
the single SnO2 control device, suggesting the formation of a Type-II heterojunction
in all cases. However, both the ZZ and HW polymer were spun onto a thinner 20.0
± 0.5 nm SnO2 film. In both of these cases, the quenched signal is no longer limited

239
to the absorption profile of the polymer, affecting the entire absorption spectrum
including both the polymer and the metal oxide [Fig. 6-12(a)]. ZZ-TOP and ED4
were spin-coated onto thicker 100.0 ± 0.5 nm SnO2 films, which do not exhibit a
reduced photocurrent over SnO2 ’s absorption profile.

Without AFM data of the thinner SnO2 film it is difficult to conclusively explain
this phenomenon. However, many metal oxides will exhibit island growth when de-
posited at similar thicknesses on glass due to poor wetting. Thus, if the 20 nm SnO2
film is incomplete then the TNT quenching may be negatively affecting the charge
carrier collection efficiency of the device. Because as the dissociated carriers must
hop through some combination of metal oxide crystallites and polymer regions while
traversing the channel, the quenched polymer regions may exhibit lower conductivities
which would affect the carriers that were photogenerated by SnO2 .

While each bilayer sensor reacts to the TNT vapor, the response time and magni-
tude of response was different for each case. Plotting the aggregated real-time data
of all four polymer/metal oxide devices, it is evident that the HW/SnO2 device dis-
plays the best performance [Fig. 6-13]. The 70% quench response over 125 seconds is
comparable to the amount of fluorescence quenching that occurs in similar polymer
films [7, 11, 28, 29].

By varying the flow rate, it is possible to change the amount of TNT delivered to
the sensor. Using the HW polymer/metal oxide device as an example, we obtain a
plot of response time versus flow rate [Fig. 6-13]. There is a clear saturation point
where the time constant approaches 100 ms. By varying the time steps we were able
to determine that this saturation point was not an artifact of the measurement but
rather intrinsic to the device itself. Thus, it is possible to estimate the amount of
TNT necessary to elicit a response.

Without proper gas chromatography and mass spectrometry equipment, it is im-

possible to know the real dose of TNT being delivered to the samples. However, a
reasonable upper bound on sensitivity can be calculated by assuming we have satu-
rated TNT vapor and that there are no effects from convective flow. The following

240
241
Figure 6-12: (a) Normalized absorption spectra of a 100.0 ± 0.5 nm SnO2 film (grey), 5.0 ± 0.5 nm HW film (orange), 17.0 ±
0.5 nm ED4 film (magenta), 18.0 ± 0.5 nm ZZ film (blue), and 17.0 ± 0.5 nm ZZ-TOP film (green). Spectral response at 5
V/µm of (b) 20.0 ± 0.5 nm SnO2 /18.0 ± 0.5 nm ZZ (c) 100.0 ± 0.5 nm SnO2 /17.0 ± 0.5 nm ED4 (d) 20.0 ± 0.5 nm SnO2 /5.0 ±
0.5 nm HW before and after exposure to 300 cc/min saturated TNT vapor. Insets: Photocurrent vs. time plot at (b) λ = 456
nm and (c)/(d) at λ = 500 nm and TNT vapor on at t = 0 seconds.
Figure 6-13: Normalized photocurrent vs. time plot of various chemosensitive fluo-
rescent polymer/metal oxide devices. Inset: Time constants from exponential fits vs.
flow rate of 20.0 ± 0.5 nm SnO2 /5.0 ± 0.5 nm HW device.

formula, derived from the ideal gas law, can be used for this purpose:

PV
M WT N T × A × , (6.3.1)
RT

where M WT N T is the molecular weight of TNT (227 g/mol), A is the ratio of the
active device area to the size of the substrate (assuming even distribution of vapor),
P is the pressure (in atm), R is the universal gas constant, T is the temperature (in
K), and V is the volume (in mL) of TNT delivered to the sensor. The volume of TNT
delivered is a function of the flow rate and can be calculated with

V = ϕ × τ × ξ (T ) , (6.3.2)

where ϕ is the flow rate (in cc/s), τ is the response time (in s), and ξ (T ) is the
vapor pressure of TNT at a given temperature (10 ppb at room temperature) [30].
By using a signal to noise ratio of 3 to 1, where the noise is the variance of the signal

242
before TNT vapor is introduced, it is possible to extract a response time from the
real-time data. The response time is the difference between the time TNT vapor
is introduced and the time when the signal drops an amount equal to 3 times the
value of the noise (∼ 3 seconds). Thus, the calculation yields a TNT detection limit
of roughly 1 picogram. While this level of detection is still 3 orders of magnitude
inferior to the commercial Fido sensor, it is on par with other commercially available
explosives detection systems. A direct comparison of the proof-of-concept bilayer
sensor with the highly engineered Fido system is not entirely fair, given that the
bilayer sensor can be further optimized for improved performance. In addition, the
ability to transduce the chemical signal directly from any chemosensitive fluorescent
polymer into an electrical signal makes this an attractive sensor technology.

6.4 Summary

In this chapter we have successfully demonstrated a bilayer heterojunction explosives

sensor and a general sensor platform for chemosensitive fluorescent polymers. Further
characterizion of the bilayer heterojunction sensor will greatly improve performance.
Careful studies of response times versus polymer and metal oxide thickness are nec-
essary to optimize device sensitivity. Once the device structure is optimized, dosing
experiments can be applied to quantify the limits of detection. In addition, specificity
testing should be carried out to determine whether that particular polymer attribute
is inherited in the bilayer devices. Finally, measuring the change in sensitivity with
varying illumination intensity and bias are two more parameter spaces that could po-
tentially lead to the ability to tune the sensitivity, effectively enhancing the sensor’s
dynamic range.
In order to optimize the bilayer sensor, the devices should incorporate thinner
polymer films so that the TNT can quench as much of the film as possible without
having to diffuse into the bulk. This modification must be done in conjunction with
reducing the photoresponse of the underlying CTL so that the differential signal can
be extracted from the background response. One way to accomplish this would be to

243
synthesize polymers with absorption profiles that are red-shifted away from the metal
oxide absorption profile. This shift would prevent the metal oxide photoresponse from
swamping out the smaller signal from the thin polymer films.

244
Chapter 6 References

[1] G. Zorpette. Countering IEDs. IEEE Spectrum, September 2008.

[2] T. Gao, M.D. Woodka, B.S. Brunschwig, and N.S. Lewis. Chemiresistors for
Array-Based vapor sensing using composites of carbon black with low volatility
organic molecules. Chemistry of Materials, 18(22):5193–5202, October 2006.

[3] T. C Pearce, S. S Schiffman, H. T Nagle, and J. W Gardner, editors. Handbook

of Machine Olfaction. Wiley-VCH, Weinheim, 2003.

[4] A. Rose, Z. G Zhu, C. F Madigan, T. M Swager, and V. Bulovic. Sensitivity gains

in chemosensing by lasing action in organic polymers. Nature, 434(7035):876–
879, 2005.

[5] B. Lourdes, D. N Reinhoudt, and M. Crego-Calama. Design of fluorescent ma-

terials for chemical sensing. Chemical Society Reviews, 36:993–1017, 2007.

[6] S. W Thomas, G. D Joly, and T. M Swager. Chemical sensors based on amplifying

fluorescent conjugated polymers. Chemical Reviews, 107:1339–1386, 2007.

[7] T. M Swager. The molecular wire approach to sensory signal amplification.

Accounts of Chemical Research, 31(5):201–207, 1998.

[8] Q. Zhou and T. M Swager. Fluorescent chemosensors based on energy migration

in conjugated polymers: The molecular wire approach to increased sensitivity.
Journal of the American Chemical Society, 117(50):12593–12602, 1995.

245
 [9] J. S Yang and T. M Swager. Porous shape persistent fluorescent polymer films:
An approach to TNT sensory materials. Journal of the American Chemical
Society, 120(21):5321–5322, 1998.

[10] J. Yang and T. M. Swager. Fluorescent porous polymer films as TNT chemosen-
sors: Electronic and structural effects. Journal of the American Chemical Society,
120(46):11864–11873, November 1998.

[11] C. J Cumming, C. Aker, M. Fisher, M. Fox, M. J la Grone, D. Reust, M. G

Rockley, T. M Swager, E. Towers, and V. Williams. Using novel fluorescent
polymers as sensory materials for above-ground sensing of chemical signature
compounds emanating from buried landmines. IEEE Transactions on Geoscience
and Remote Sensing, 39(6):1119–1128, 2001.

[12] J. S Kim, D. T McQuade, A. Rose, Z. G Zhu, and T. M Swager. Directing

energy transfer within conjugated polymer thin films. Journal of the American
Chemical Society, 123(46):11488–11489, 2001.

[13] K. Kuroda and T. M Swager. Self-amplifying sensory materials: Energy migra-

tion in polymer semiconductors. Macromolecular Symposia, 201:127–134, 2003.

[14] J. Sung and R. J Silbey. Quantitative relationship between analyte concentration

and amplified signal intensity of a molecular wire sensor. Analytical Chemistry,
77(19):6169–6173, 2005.

[15] N. D Kumar, J. D Bhawalkar, P. N Prasad, F. E Karasz, and B. Hu. Solid-state

tunable cavity lasing in a poly(para-phenylene vinylene) derivative alternating
block co-polymer. Applied Physics Letters, 71(8):999–1001, 1997.

[16] F. Hide, B. J Schwartz, M. A Diaz-Garcia, and A. J Heeger. Conjugated polymers

as solid-state laser materials. Synthetic Metals, 91(1-3):35–40, 1997.

[17] C. Madigan and V. Bulovic. Modeling of exciton diffusion in amorphous organic

thin films. Physical Review Letters, 96:046404, 2006.

246
[18] M. Csele. Fundamentals of light sources and lasers. Wiley-IEEE, Hoboken, NJ,
2004.

[19] A. Rose, Z. Zhu, C. F Madigan, T. M Swager, and V. Bulovic. Chemosensory

lasing action for detection of TNT and other analytes. Proceedings of SPIE,
6333:6330Y1–6330Y7, 2006.

[20] J. L. G. Fierro and J. L. G. Fierro. Metal oxides. CRC Press, Boca Raton, 2006.

[21] W. Alan Doolittle, Alexander G. Carver, and Walter Henderson. Molecular beam
epitaxy of complex metal-oxides: Where have we come, where are we going, and
how are we going to get there? In J. Vac. Sci. Technol. B, volume 23, pages
1272–1276. AVS, May 2005.

[22] P. J. Wright, C. J. Anthony, M. J. Crosbie, P. P. Donohue, P. A. Lane, and

M. A. Todd. CVD of oxide materials for thermal imaging - the role of precursor
chemistry. Journal of Materials Chemistry, 14(21):3251–3258, 2004.

[23] M. Ritala, K. Kukli, A. Rahtu, P.. I. Risnen, M. Leskel, T. Sajavaara, and

J. Keinonen. Atomic layer deposition of oxide thin films with metal alkoxides as
oxygen sources. Science, 288(5464):319–321, April 2000.

[24] B. B. Lakshmi, P. K. Dorhout, and C. R. Martin. SolGel template synthesis

of semiconductor nanostructures. Chemistry of Materials, 9(3):857–862, March
1997.

[25] Y. Xu and M. A. A. Schoonen. The absolute energy positions of conduction

and valence bands of selected semiconducting minerals. American Mineralogist,
85(3-4):543–556, March 2000.

[26] M. Ruske, G. Bruer, J. Pistner, U. Pffflin, and J. Szczyrbowski. Properties

of SnO2 films prepared by DC and MF reactive sputtering. Thin Solid Films,
351(1-2):146–150, August 1999.

247
[27] J. L. Brousseau, H. Bourque, A. Tessier, and R. M. Leblanc. Electrical properties
and topography of SnO2 thin films prepared by reactive sputtering. Applied
Surface Science, 108(3):351–358, March 1997.

[28] D. Zhao and T. M Swager. Sensory responses in solution vs solid state: A

fluorescence quenching study of poly(iptycenebutadiynylene)s. Macromolecules,
38(22):9377–9384, 2005.

[29] J. S Yang and T. M Swager. Fluorescent porous polymer films as TNT chemosen-
sors: Electronic and structural effects. Journal of the American Chemical Society,
120:11864–11873, 1998.

[30] J. C. Oxley, J. L. Smith, K. Shinde, and J. Moran. Determination of the vapor

density of triacetone triperoxide (TATP) using a gas chromatography headspace
technique. Propellants, Explosives, Pyrotechnics, 30(2):127–130, 2005.

248
Chapter 7

Concluding Remarks and Future

Areas of Research

Great is the art of beginning, but

greater is the art of ending.

ROBERT A. HEINLEIN
Author

The interdisciplinary research undertaken in this thesis resulted in the develop-

ment of a novel solid-state vapor-phase explosives sensor. As an electrical engineer,
there was a motivation to develop a device platform that could be customized and
optimized while ultimately providing an output that is electrical in nature for down-
stream readout circuitry. This led to a using a lateral photoconductor architecture
that improved upon the simple transduction mechanism of the chemiresistor. Inspira-
tion for separating the chemical sensing function from electrical characteristics came
from understanding the benefits of dual-layer photoreceptor drums invented during
the heyday of xerographic development and the present challenges of developing flu-
orescent polymer sensors. Through the design, fabrication, and characterization of
lateral heterojunction photoconductors, we have not only developed a viable sensor
platform technology, but also a platform for investigating device physics at a het-
erointerface.

249
7.1 Research Highlights
Some of the useful findings in this thesis include:

ˆ Extending the dual-layer photoreceptor concept (from xerography) of a CTL

and EGL by incorporating the dual-layer structure into a lateral device

ˆ Demonstrating photoconductive gain in a bilayer device

ˆ Demonstrating that recombination rate is dominated by bimolecular recombi-

nation in bilayer devices

ˆ Integration of various material sets including QDs, J-aggregates, polymers, and

metal oxides into viable photoconductors

ˆ Design of trilayer devices that modulate interfacial carrier dynamics

ˆ Demonstrating solid-state vapor-phase detection of as little as 1 pg of TNT

ˆ Confirming the ability of the bilayer sensor to be a polymer-independent plat-

form

7.2 Research Opportunities

The work performed in this thesis has proved the utility of the bilayer heterojunction
photoconductor as a device physics toolkit as well as a platform for chemical sensing.
A number of interesting experimental results warrant further scientific attention and
investigation, representing opportunities to not only delve deeper into the physics of
organic materials at an interface but also the opportunity to extend and improve the
solid-state chemical sensing platform.
One pressing need in the field of chemical and biological sensors is the ability to
tune the sensitivity of a given sensor. For example, there may be situations that
require high sensitivity at the cost of a large number of false positives, while at other
times, lower sensitivity with fewer false positives will suffice. If this tuning ability

250
could be achieved with a simple adjustment, the sensor could be more useful in a
wider range of environments and applications. The bilayer sensor has the potential
for a wide dynamic range by adjusting either operating temperature or intensity of
illumination. In the FIDO—detector, the polymer is heated to promote faster TNT-
polymer absorption/desorption kinetics at the surface. Therefore, by heating the
bilayer sensor, the TNT will bind less strongly to the polymer and result in lower
sensitivities. Heating will also result in larger background dark currents as the metal
oxide’s free carriers become more mobile. This requires paying close attention to the
resulting signal to noise ratio (SNR). Changing the illumination intensity will modify
the SNR by providing a “knob” to tune exciton population and consequently the
free carrier concentration at the interface. Thus, altering the excitation intensity will
ultimately make the sensor more or less sensitive. These two methods of changing
sensitivity are simple enough to be incorporated as an electrically tunable parameter
made available to the end user.

A simple extension of the bilayer sensor work would be to extend the platform
to include solution-phase detection. The sensors in this work incorporate polymeric
and metal oxide materials, which will not solvate in aqueous solutions. This would
potentially enable the sensors to operate in aqueous environments, which are common
to many biomedical applications such as blood diagnostics.

We have demonstrated that extending the lateral heterojunction platform to in-

clude multilayers has potential benefits to overall device efficiency. Further explo-
ration of different groups of materials will be necessary to elucidate the physical
mechanisms involved. Key to this effort will be utilizing film deposition methods
such as atomic layer deposition (ALD) that can exert fine control over film thickness
and morphology, both of which are critical in the interpretation of experimental data.
Once multilayer devices are well understood, translating those insights into designing
multilayer chemical sensors will enable the engineering devices for better sensitivity.

In order to expand the utility of the bilayer chemical sensor platform, adding de-
tection of different high-value analytes will be an important piece of ongoing research.
One way to quickly build this capability is to develop a modular library of chemi-

251
cal building blocks that can be used to synthesize new fluorescent polymers that are
specific to any analyte of interest. Another interesting vein of research would be to
leverage the growing availability of functionalized nanoparticles such as quantum dots
or carbon nanotubes within the bilayer heterojunction device framework.
Finally, the ultimate goal of this work is to enable a new class of solid-state
chemical sensors that are amenable to miniaturization and engineering. Towards
this goal, this thesis strove to define the available parameter space for optimizing
the bilayer heterojunction photoconductor. The results of this work will certainly
facilitate the incorporation of many more chemosensitive fluorescent polymers into
viable sensors. The potential to fabricate arrays of highly specific and highly sensitive
sensors with wide dynamic range holds promise for the bilayer sensor platform to
find use in many future applications within the defense, medical, environmental, and
industrial processing industries. While this work opens many new avenues for future
endeavor, the focus must always remain on the end application, as the sensor will
only be as useful as the analyte it is searching for.

252