Excitonics for Organic Electronics by Grayson Ingram a Thesis Submitted

Home , Electronic component

Excitonics for Organic Electronics

Grayson Ingram

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy

Department of Materials Science and Engineering University of Toronto

Excitonics for Organic Electronics

Grayson Ingram

Doctor of Philosophy

Department of Materials Science and Engineering University of Toronto

2018

Abstract

Organic semiconductors have immense potential as replacements for traditional inorganic materials in optoelectronics applications, in particular for organic light-emitting diodes (OLEDs). At the core of the device physics governing the stability and efficiency of OLEDs are tightly bound electron-hole pairs known as excitons. Here, the link between OLED excitonics and operational stability is studied in active OLEDs.

First, the exciton distribution is investigated in active OLEDs with doped and undoped emissive layers. In both types of OLEDs, a surprisingly narrow exciton formation zone was measured given the bipolar nature of the materials used.

Next, the influence of defects on exciton diffusion is investigated. The effective singlet exciton diffusion length is measured as a function of defect concentration and operational conditions and described in a unified model. Exciton-defect interactions are central to the efficiency and stability of OLEDs; consequently, quantifying these interactions under realistic operating conditions is a major step towards a comprehensive understanding of OLED excitonics.

Finally, building on these results, a simple model for singlet exciton driven degradation in OLEDs in presented. This model accounts for the time and current density dependence of the host and defect emission in degrading OLEDs and is thoroughly validated using new experimental data, as well as literature data from leading academic and industrial research

groups. Results of a first look at applying these conclusions to photovoltaic devices are also included. There exists an abundance of evidence in the literature suggesting singlet exciton driven degradation is a major degradation mechanism. Undoubtedly, providing mathematical tools to quantify this process will be invaluable.

iii

Table of Contents

Table of Contents ...... iv

List of Tables ...... vi

List of Figures ...... vii

Abbreviations ...... xiii

Chapter 1 Introduction ...... 1

Chapter 2 Experimental Methods ...... 31

Chapter 3 Exciton Diffusion in CBP ...... 38

Chapter 4 Exciton Distribution in a Doped Emissive Layer OLED ...... 59

Chapter 5 Impact of Defects on Exciton Diffusion ...... 66

Chapter 6 Electroluminescence of Degrading CBP Films ...... 82

Chapter 7 Excitonic Degradation of OLEDs ...... 96

Chapter 8 OPV Degradation ...... 102

Chapter 9 Conclusions and Future Work ...... 114

References ...... 121

Appendix A: Boundary Condition Validation ...... 132

Appendix B: Supplementary Information for Chapter 5 ...... 134

Appendix C: Equation Derivations ...... 137

Appendix D: OLED External Degradation ...... 143

Appendix E: Supplementary Information for Chapter 6 ...... 144

Appendix F: Environmental Contaminant Diffusion Simulation ...... 146

List of Tables

Table 1-1: Selected physical properties of germanium and anthracene ...... 11

Table 3-1: Parameter values used in the calculation of OLED emission ...... 48

Table 5-1: Extracted relative defect concentrations for three sets of OLEDs ...... 77

Table 5-2: Extracted effective diffusion lengths and saturation intensities for three sets of OLEDs ...... 79

Table 8-1: Bright EL areas as a fraction of total pixel area for OPV’s driven at 25mA/cm2 forward bias, and Jsc as a fraction of its value at the start of outdoor testing, measured 4, 43, and 163 days after the outdoor testing period ended ...... 110

Table E-1: Details on extracted data from Figure 6-3 ...... 147

Table E-2: Fit metrics for fits to Equation 7-3 in Figure 7-1 ...... 148

Table E-3: Fit metrics for fits to Equation 7-3 in Figure 7-2 ...... 148

List of Figures

Figure 1-1: Single layer OLED schematic depicting the processes of 1) charge injection 2) charge transport 3) exciton formation and 4) exciton decay accompanied by light emission ...... 6

Figure 1-2: The bilayer OLED structure reported by Tang and Van Slyke. Reproduced with permission...... 7

Figure 1-3: Energy level diagram of a generic multilayer OLED including hole injection layer, electron blocking layer, emissive layer, hole blocking layer and electron injection layer sandwiched between two electrodes...... 8

Figure 1-5: Simplified potential energy curves and vibrational wavefunctions and demonstration of Franck-Condon transitions with strength determined by wavefunction overlap. This leads to

Figure 1-6: Illustration of Wannier-Mott, Frenkel, and charge transfer excitons. Filled circles represent molecules...... 18

Figure 1-7: Molecular state viewed a) through electron configuration and b) through exciton energy levels. The view taken in b) greatly simplified the intuitive picture of transitions between molecular states. Reproduced with permission, copyright 2015 John Wiley & Sons, Inc...... 19

Figure 1-8: Jablonski-diagram, indicating the singlet and triplet manifold with vibrational levels. Also indicated are radiative and nonradiative transitions as arrows between different states. On the left side, a schematic absorption spectrum is indicated. Reproduced with permission, copyright 2015 John Wiley & Sons, Inc...... 20

Figure 1-9: Illustration of the electron transitions involved in Förster and Dexter energy transfer...... 23

vii

Figure 1-10: Lifetime evolution of fluorescent (a) and phosphorescent (b) blue, green, and red OLEDs. The data points for the white OLEDs were not distinguished between stacked, hybrid, or separate phosphorescent/fluorescent devices. Reprinted with permission from Scholz, S., Kondakov, D., Lüssem, B. & Leo, K. Degradation mechanisms and reactions in organic light- emitting devices. Chem. Rev. 115, 8449–8503 (2015). Copyright 2015 American Chemical Society...... 29

Figure 1-11: Dark spots growing on an OLED. Reprinted with permission from Smith, P. F., Gerroir, P., Xie, S., Hor, A. M. & Popovic, Z. Degradation of Organic Electroluminescent Devices . Evidence for the Occurrence of Spherulitic Crystallization in the Hole Transport Layer. 80, 5946–5950 (1998). Copyright 1998 American Chemical Society...... 30

Figure 2-1: Kurt J. Lesker Luminos Cluster Tool ...... 35

Figure 2-2: Completed devices on a glass substrate with pre-patterned ITO anode. Organic materials are layered on top of the ITO in eight regions of the substrate and two Al cathode bars are deposited on top, perpendicular to the ITO. Each 2mm2 pixel is defined by the region of overlap between an Al and an ITO bar for 32 pixels total (every 5’th ITO bar is shorted to the Al)...... 35

Figure 2-3: a) Example organic shadow mask b) Al shadow mask and c) substrate holder on top of shadow mask holder demonstrating alignment of mask and substrate...... 36

Figure 2-4: a) Minolta LS-110 luminance meter and b) integrating sphere connected by optical fiber to an Ocean Optics USB4000 spectrometer ...... 37

Figure 2-6: Example of a) device architecture and b) emission intensity of different layers of an OLED incorporating exciton capturing layers...... 40

Figure 3-1: a) The LUMO and HOMO levels (upper and lower lines) and b) the singlet exciton energies of layers for devices used to measure the singlet exciton diffusion length in CBP...... 44

viii

Figure 3-2: Sensor emission intensity calculated using numeric solutions to Equation 3-11 for

R0=0nm (black circles), R0=3nm (blue triangles) and R0=5nm (red squares). The data is fitted with Equation 3-7 (dashed lines) which does not take into account long range energy transfer

from the host to dopants. For the R0=3 and R0=5, erroneous diffusion length values of 5.2nm and 6.5nm are extracted using Equation 3-7...... 49

Figure 3-3: Effective diffusion length (shown in color scale) for an intrinsic diffusion length of

a) 2.5nm and b) 5nm as a function of CA and R0. Data are calculated using Equation 3-11 and effective diffusion lengths are extracted from fits to Equation 3-7. Note the difference in scale between the two plots...... 50

Figure 3-4: CBP thin film photoluminescence (blue solid line) and Rubrene absorption (black dashed line) in toluene solution...... 52

Figure 3-5: a) Current density-voltage characteristics and b) OLED emission spectra at 12mA/cm2 for devices with sensing layers placed at 5nm, 10nm, 15nm, 20nm and 25nm from the ETL/HTL interface...... 53

Figure 3-6: Sensing layer emission intensity as a function of the sensing layer position and current density. The solid circles are experimental measurement and the dashed lines are fits to Equation 3-11...... 54

Figure 3-7: Extracted singlet exciton diffusion length as a function of current density...... 55

Figure 3-8: a) Fitting parameters, A’ (black squares) and B’ (blue circles) as a function of current density extracted from fitting sensing layer emission to Equation 3-11 and b) the same parameters divided by the current...... 56

Figure 3-9: a) Experimental sensor emission at 10mA/cm2 (blue data points) as well as fits to Equation 3-11 setting g=1nm and g=4nm (red solid and black dashed lines respectively) and b) Extracted singlet exciton diffusion lengths by fitting the data in a) with Equation 3-11 and g set to different values...... 57

Figure 3-10: Current voltage characteristics for devices with sensing layers at 5 nm from the interface and a thin HTL (dashed line) and 200 nm HTL (solid line) ...... 61

Figure 4-1: Schematic energy level diagram of organic layers in OLEDs. The lines at top and bottom of each box represent the LUMO and HOMO energies (in eV) respectively...... 63

Figure 4-2: a) Current vs voltage characteristics and b) normalized electroluminescence spectra

at 7.4 V of devices with x=1 nm, x=5 nm and a device with no C60 doped layer ...... 64

Figure 4-3: Normalized emission spectra of Ir(ppy)2(acac) (dashed green line) and normalized

absorption spectrum of C60 (solid black line) ...... 66

Figure 4-4: Measured quenching fraction (black dots) and fit (line) at 40mA/cm2 ...... 68

Figure 5-1: Schematic device structure of the OLEDs used in this study...... 71

Figure 5-2: OLED emission spectra for devices from three different sets of OLEDs run for nine seconds at 25 mA/cm2 with an exciton capturing layer at 10 nm. The data is averaged over two nominally identical devices and smoothed for clarity...... 75

Figure 5-3: Ratio of defect to host emission intensities as a function of the exciton capturing layer position for three sets of OLEDs at constant current density of a) 25 mA/cm2 and b) 100 mA/cm2 ...... 76

Figure 5-4: CBP emission intensities for three sets of OLEDs operated at 25 mA/cm2 and 100 mA/cm2...... 76

Figure 5-5: Extracted effective exciton diffusion length as a function of defect concentration. ..81

Figure 5-6: a) Driving voltages at a constant current density of 25 mA/cm2 for devices with different defect concentrations and spacer thicknesses as well as linear fits to the data. b) The y- intercept of the fits in a) vs. defect concentration as well as linear fits to the data...... 83

Figure 6-1: OLED emission at 9 s, 90 s and 900 s for a device aged at 25 mA/cm2, smoothed for clarity...... 90

Figure 6-2: a) CBP emission as a function of time and fits to Equation 6-3 and b) defect emission and fits to Equation 6-7. Devices are aged at constant current densities of 12.5 mA/cm2 to 50 mA/cm2...... 91

Figure 6-3: Values of the luminance half-life (t50) extracted from fits in Figure 6-2a (black circles) and fits in Figure 6-2b, (red squares). A power law fit to the black circles is also shown (solid line)...... 92

Figure 6-4: Emission of an OLED with exciton capturing layer at d=10nm from the CBP/TBPi interface at 9 s, 90 s and 900 s. The device is aged at a constant current density of 25 mA/cm2. Data is smoothed for clarity...... 94

Figure 6-5: Time dependent CBP emission for devices with exciton capturing layers at d=2.5 nm, 5 nm, 7.5 nm, 10 nm and 12.5 nm from the CBP/TPBi interface, and fits to Equation 6-5. .96

Figure 6-6: CBP luminance half-life, t50, with a fit to Equation 6-11 as well as CBP initial intensity, I0, with a fit to Equation 6-12. Both sets of data are extracted from the fits in Figure 6- 5...... 97

Figure 7-1: a) Degradation of HTM/TPBi bilayer OLEDs from ref. 33 driven at 20 mA/cm2 (solid dots) and fits to Equation 7-3 (lines). b) Luminance half-life as a function of the HTM singlet energy ...... 101

Figure 7-2: Degradation of state-of-the-art OLEDs (solid dots) and fits to Equation 7-3 (solid lines) ...... 103

Figure 7-3: Degradation data (solid dots) for the OLEDs in Figures 2a and 3 with normalized time (t50=1) and Equation 3 with L0=1 and t50=1 (dashed line)...... 104

Figure 8-1: EL spectra representative OPVs with Pc, Nc and Nc/Pc structures. Data are smoothed for clarity...... 107

Figure 8-2: Photographs of the EL of an OPV pixel 4, 43 and 163 days after outdoor testing. A photograph of the PL is also shown 163 days after outdoor testing. White outlines indicate the OPV pixel area...... 108

Figure 8-3: Current-voltage characteristic curves of an OPV device under simulated AM1.5 illumination and the same device with only the region dark under EL testing exposed to light...... 111

Figure 8-4: Normalized electroluminescence of a Pc OPV at multiple current densities, smoothed for clarity. The inset shows the M:A emission intensity as a function of current density and a linear fit to the data...... 114

Figure 8-5: a) Intercept (red squares) and inverse of the slope (black circles) extracted from fits to the emission ratio vs current density, and b) the driving voltage at 7.5 mA/cm2 plotted as a function of the number of days after device fabrication...... 115

Figure 8-6: Intercept (red squares) and inverse of the slope (black circles) extracted from fits to the emission ratio vs current density, plotted against the driving voltage at 7.5 mA. Solid lines are linear fits to each set of parameters...... 116

Figure 9-1: Illustration of the exciton diffusion process under different temperatures. (a) The downhill migration fully determines the exciton diffusion process at low temperatures. (b) At high temperatures, the thermally activated hopping also contributes to the exciton diffusion length. Reprinted with permission from Mikhnenko, O. V et al. Temperature dependence of exciton diffusion in conjugated polymers. J. Phys. Chem. B 112, 11601–4 (2008). Copyright 2008 American Chemical Society...... 120

xii

Abbreviations

a6T (alpha-sexithiophene)

BCP (2,9-Dimethyl-4,7-diphenyl-1,10-phenanthroline)

Born Oppenheimer (BO)

CBP (4’-bis(carbazol-9-yl)biphenyl)

Charge transfer (CT)

Cl-BsubNc (Boron subnaphthalocyanine chloride)

Cl-BsubPc (Boron subphthalocyanine chloride)

Electroluminescnce (EL)

Electron blocking layer (EBL)

Electron injection layer (EIL)

Electron transport layer (ETL)

Emissive layer (EML)

External quantum efficiency (EQE)

Förster resonant energy transfer (FRET)

Highest Occupied molecular orbital (HOMO)

Hole injection layer (HIL)

xiii

Hole transport layer (HTL)

Hole transport material (HTM)

ITO (Indium tin oxide)

Light emitting diode (LED)

Liquid crystal display (LCD)

Lowest unoccupied molecular orbital (LUMO) mcp (1,3-Bis(carbazol-9-yl)benzene)

MDMOPPV (poly[2-methoxy-5-(30,70-dimethyloctyloxy)-1, 4-phenylenevinylene])

Miller-Abrahams (MA)

NPB (N,N' -Bis(naphthalen-1-yl)-N,N' -bis(phenyl)-benzidine)

Organic light emitting diode (OLED)

Organic photovoltaic (OPV)

PEDOT:PSS (Poly(3,4-ethylenedioxythiophene)-poly(styrenesulfonate))

Photoluminescence (PL)

Power conversion efficiency (PCE)

Root mean squared error (RMSE)

Spectrally resolved photoluminsescent quenching (SRPLQ)

Spiro-cbp (2,2',7,7'-Tetrakis(carbazol-9-yl)-9,9-spirobifluorene)

Spiro-npb (N,N' -Bis(naphthalen-1-yl)-N,N' -bis(phenyl)-2,7-diamino-9,9-spirobifluorene)

Stretched exponential (SE)

xiv

TAPC ((1,1-bis-(4-bis(4-methyl-phenyl)-amino-phenyl)-cyclohexane))

TBADN (2-tert -Butyl-9,10-di(naphth-2-yl)anthracene

TCTA (4,4',4"-Tris(carbazol-9-yl)triphenylamine

TPBi (1,3,5-Tris(1-phenyl-1Hbenzimidazol- 2-yl)benzene)

Ultraviolet (UV)

Chapter 1 Introduction

1.1 Motivation

The field of organic electronics aims to replace inorganic semiconductors with organic semiconductors in a variety of applications, examples of which include light-emitting diodes, photovoltaic devices, and thin film transistors.

To date, OLED is the most successful organic electronic technology. OLEDs have several advantages over LED/LCD technology in display applications such as better black level, higher contrast, and wider viewing angle. These advantages have fueled the adoption of OLEDs in displays for mobile electronic devices and televisions. OLEDs also have potential for use in solid state lighting. For this application advantages of OLEDs include their lightweight and flexible nature, large emission area, and tunable emission spectra. Despite these advantages, efficiency and lifetime of bright, white OLEDs are lower than for display applications and they have yet to achieve widespread adoption.

Organic electronic technologies benefit from the mechanical properties of organic thin films, allowing for the fabrication of thin, light, flexible devices; however it is the electrical, optical, and excitonic properties of these materials which fundamentally enable all organic electronic devices. These properties in turn, are dependent on the structure of the organic molecules, as well as their solid state morphology. Morphology can be tuned to some extent by varying processing conditions, but is to a large extent a product of molecular structure. The connection between molecular structure and materials properties is exploited by synthetic chemists who are able to tune the electrical and optical properties of organic molecules through the addition or removal of functional groups.

The near infinite set of possible structures for organic molecules provides immense potential for synthesizing molecules with desirable properties but also gives researchers the difficult task of attempting to choose materials to focus on from a nearly unlimited supply of options. This task is further complicated by researchers’ very limited ability to assess the relevant materials

properties of newly synthesized organic materials. For example, reports of electronic energy levels, exciton diffusion lengths, and charge mobilities for a given material often vary by orders of magnitude depending on the method used and the researcher making the measurement. Assessing new materials is complicated further still by an inability to predict how a given material will perform, even given knowledge of materials properties, as performance depends on subtle aspects of how a material interacts with others in a device. These factors make identifying “good” organic molecules and device architectures similar to trying to find a needle in a haystack, without a good means of distinguishing between needles and hay.

Due to the difficulties outlined above, progress the field of organic electronics has been driven more significantly by large-scale, empirical, trial-and-error than by rational design of molecules and devices. In order to continue to push the boundaries of this field, a focus on improving our understanding of the underlying physical processes occurring in organic devices is necessary. This will allow for rational design through understanding of the influence of molecular structure, materials properties, and device architecture on device performance.

The focus of this work is on the physics of bound electron hole pairs, or excitons, in organic devices. As will be demonstrated in the subsequent sections, the properties of excitons play a central role in the determination of the efficiency and operational stability of OLED devices.

1.2 Background 1.2.1 OLED Operation and Selective History

This section will provide a high level overview of OLED operation.

The simplest OLED architecture includes a single organic layer between two electrodes, at least one of which is partially transparent to allow light to escape the device. This is the first structure which was used to exhibit organic electroluminescence, when Pope et al. observed emission from anthracene crystals driven at several hundred volts in 19631. When a voltage is applied across the electrodes, electrons and holes are injected into the device from opposite electrodes, and are transported across the device under influence of the applied electric field. When

electrons and holes come into close proximity as they are being transported through the organic material, their mutual coulomb attraction can lead to the formation of a tightly bound electron/hole pair known as an exciton. In organic materials excitons are typically localized onto a single molecule and often the terms “exciton” and “molecular excitation” are used interchangeably. The exciton may transfer between molecules and diffuse for a short period of time before the electron and hole recombine to release their potential energy as either heat or light. A portion of the light emitted during exciton recombination can be coupled out of the device through the transparent electrode, and thus light is produced through electrical excitation. These processes are illustrated in Figure 1-1.

Figure 1-1: Single layer OLED schematic depicting the processes of 1) charge injection 2) charge transport 3) exciton formation and 4) exciton decay accompanied by light emission

Although the processes described above are common to all OLEDs, the single organic layer structure usually has extremely poor performance. The main reason for the low performance is that most organic semiconductors have multiple orders of magnitude difference between electron and hole mobilities meaning that most charges traverse the entire device without forming

excitons and emitting light, and those that do form excitons will do so primarily near the electrodes where exciton quenching on electrodes is efficient.

The solution to this problem was first demonstrated by Tang and Van Slyke in 1987 when they demonstrated moderate efficiency (~1%) from an organic bilayer device2. In this device, the functions of electron and hole transport were separated onto two different materials in a bilayer configuration as shown in Figure 1-2. At the junction between these two organic materials, the offsets in the electronic energy levels of the two materials lead to a large steady state concentration of charge carriers. These large charge carrier concentrations lead to efficient exciton formation and prevented current leakage as in the single layer device.

Figure 1-2: The bilayer OLED structure reported by Tang and Van Slyke. Reproduced with permission2.

A second development by the same authors came two years later when they further separated the functions of emission and charge transport by introducing a low concentration of an efficient

emitter into the electron transport layer3. This technique allowed the authors to achieve a ~2.5% EQE.

The next major step forward in OLED technology was in the development of phosphorescent emitters4 in 1998. The emissive materials used up until this point were fluorescent, meaning that they were only capable of emitting efficiently from the singlet state. Triplet excitons which comprise ¾ of excitons produced under electrical excitation recombine non-radiatively. The development of phosphorescent emitters meant that triplet excitons could also contribute to the radiative output of OLEDs and increased the theoretical limit on OLED efficiency by a factor of four. Although this first phosphorescent device had a relatively low efficiency of ~4% EQE, the ability to harvest triplet excitons quickly enabled a jump in efficiency leading to ~19% EQE only a few years later5.

Figure 1-3: Energy level diagram of a generic multilayer OLED including hole injection layer, electron blocking layer, emissive layer, hole blocking layer and electron injection layer

sandwiched between two electrodes.

Further advances after this point were more incremental, with new materials and device architectures leading to improvements in the stability and efficiency of OLEDs. In response to increasing performance demands, device architecture of many devices has grown in complexity as can be seen in the schematic in Figure 1-3. This device includes a hole injection layer (HIL), electron injection layer (EIL), hole blocking layer (HBL), and electron blocking layer (EBL) in addition to the emissive layer (EML).

1.2.2 Overview of Organic Materials Properties

1.2.2.1 Bonding

The physical properties of organic materials are largely a function of the intra- and intermolecular bonding.

Intramolecular bonding in semiconducting organic molecules is characterized by a high degree of conjugation (alternating single and double bonds). In these molecules, the “backbone” of the molecule is formed by strong bonds, with a delocalized -electron system also contributing to bonding. The delocalized -electron휎 cloud typical of orga휋nic semiconductors is shown in Figure 1-4 for an anthracene molecul휋 e. The frontier orbitals, the highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO), in conjugated molecules are typically orbitals. Frontier orbitals are involved in absorption and emission of light as well as charge and휋 energy transfer, so it is the delocalized electron systems of conjugated molecules from which semiconducting organics derive may of휋 their useful properties. Most organic molecules used in optoelectronic devices have HOMO/LUMO gaps around 1.5eV-3.5eV.

Intermolecular bonding in molecular crystals and amorphous molecular solids is primarily van der Waals bonding. These weak intermolecular bonds lead to much stronger charge and exciton localization than in inorganic semiconducting crystals in which strong covalent bonds allow for delocalization. The weak intermolecular bonding also leads to reduced hardness and low melting points relative to inorganics. In many cases, organics in the solid state will be amorphous, or “nanocrystalline” i.e. characterized by only very short range order.

Selected physical properties of Germanium and Anthracene are provided in Table 1-1. The differences between these two archetype semiconductors highlight some of the major differences in physical properties between organics and inorganics 7.

Table 1-1: Selected physical properties of germanium and anthracene7

Property Germanium Anthracene

Melting Point (C) 937 217

Dielectric Constant 16 3.2

Electron mobility at 300K (cm2V-1s-1) 3800 1.06

Hole mobility at 300K (cm2V-1s-1) 1800 1.31

Intrinsic carrier concentration (cm-3) 1014 10-4

≈

1.2.2.2 Charge Transport

Strong electronic coupling between lattice sites in most inorganic crystalline semiconductors results in band-like charge transport. Due to the large amount of energetic and spatial disorder present in the amorphous films studied here, charge transport occurs through hopping between localized states. This hopping is most commonly described by the Miller-Abrahams (MA) hopping rate8:

exp = exp 퐸푗−퐸푖 , > (1-1) 1, �− 푘퐵푇 � 휈푖푖 휈0 �훾푟푖푖� � 퐸� 퐸푖 퐸� ≤ 퐸푖 vij is the hopping rate from site i to site j, v0 is a pre-factor is the inverse of the effective wavefunction radius, rij is the spatial separation between sites,훾 Ei and Ej are the initial and final site energies, kB is the Boltzmann constant and T is the temperature. The MA expression makes hops between sites close together more likely, while hops downwards in energy occur freely and hops upwards in energy are thermally activated.

Conduction through an organic film requires both the injection and transport of charge. If the charge transport is limited by one of these two processes, relatively simple analytical expressions for the resulting current can be applied.

Unipolar, injection limited current in a single organic layer (JILC) has been described by Scott and Malliaras 9:

= 4 exp( ) exp 1 (1-2) 퐵 2 푒휙 2 퐽퐼퐼퐼 푁0휓 푒푒푒 − 푘퐵푇 �푓 � where N0 is the density of chargeable sites, φB is injection barrier height, F is the electric field at the injection interface, μ is the charge carrier mobility and is a function of the reduced field

( = 4 ). 휓 3 2 2 푓 푒 �⁄ 휋휋푘퐵 푇 = + 1 + 2 1 (1-3) 1 1 2 −1 −2 −1 2 휓 푓 푓 − 푓 � 푓 � In the other limiting case, current is limited by space charge and can be described using the Mott- Gurney law:

= 2 (1-4) 9 푉 3 퐽 8 �0𝜖 푑 where V is the applied voltage and d is the thickness of the film.

It is important to note that the thermally activated hopping on which charge transport is based leads to a field dependence of the mobility. A Poole-Frenkel type filed dependence has been observed for many materials,

= exp (1-5)

� �0 �훽√�� where β is the electric field activation parameter and μ0 is the zero-field mobility.

1.2.2.3 Optical Properties

1.2.2.4 BO Approximation

An important approximation made in the calculation of electronic energy levels of organic molecules is the Born-Oppenheimer (BO) approximation. It states that due to the large difference in mass of nuclei and electrons, electrons will respond nearly instantaneously to

changes in the nuclear configuration. Although this work will not deal explicitly with the calculation of molecular energy levels, one of the major results of the BO approximation will be useful in understanding electronic energy levels in molecules. This result is that under the BO appromation the molecular wave function can be described as:

= (1-6)

Ψ 휓휓휓 where , and are the electronic, vibrational and rotational components of the overall molecular휓 � wavefunction.� The electronic component can be further broken down:

= (1-7)

휓 휓푒푒 휓푠푠�푠 where and are the components of the electronic wavefunction dependent on the spatial and spin휓 푒푒coordin휓ates푠푠�푠 of the electrons.

The electron energy can be described by:

= + + (1-8)

퐸 퐸푒푒푒푒 퐸푣�푣 퐸푟푟푟 where , and are the electronic vibrational and rotational contributions to the total energy.퐸 푒푒푒푒The 퐸rotational푣�푣 퐸 term푟푟푟 is typically neglected due to small contribution relative to electronic and vibrational components.

Given the ability to express the electron energy as the sum of its electrical and vibrational components, we can then graphically express the electron energy level diagram by the electronic energy levels and superimposed vibrational sublevels.

1.2.2.5 Electronic Transitions

Transitions between electronic states may occur through the emission or absorption of photons or phonons. In the context of molecular solids the terms “phonon” and “vibration” will be used interchangeably as will “excited state” and “exciton”. We will see shortly that visualizing an excited state as an “exciton” quasiparticle will be advantageous when energy transfer between

molecules is an important consideration. The most common means of inducing an electronic transition from the ground to an excited state is by absorption of a photon.

The rate of electronic transitions is often approximated using Fermi’s Golden Rule:

= (1-9) 2 2휋휋 ′ 푘 ℏ ��Ψ푓�퐻��Ψ푖�� where k is the transition rate between initial and final states with wavefunctions and ; is

the density of final states, and is the perturbation coupling the two states. Ψ푖 Ψ푓 휌 ′ 퐻� The relevant perturbation for absorption of light is the dipole operator ( = ). This operator ′ only interacts with the electronic component of the wavefunction, so we퐻� can re푒풓�-write the rate of absorption as:

= , , | , | , (1-10) 2휋휋 2 2 2 푘 ℏ ��ψ푒푒 푓�푒풓��ψ푒푒 푖�� 〈�푓 �푖〉� �〈ψ푠푠�푠 푓 ψ푠푠�푠 푖〉� If any of these three terms is equal to zero, the transition rate will be zero and the transition is termed “forbidden”. The terms corresponding to the electrical and vibrational wave function are the transition dipole moment and Franck-Condon factors.

1.2.2.6 Electronic Factor

The transition dipole moment, = , , , will be the main factor in determining the

strength of absorption between푇 two states.�ψ푒푒 푓 � 푒Since풓��ψ푒푒 the푖� dipole operator is odd, the initial and final states must have opposite symmetry or the transition dipole moment will be zero. A transition for which T=0 is “dipole-forbidden”. T is large for transitions with large spatial overlap between the initial and final states and when these states are delocalized over large regions of the molecule.

1.2.2.7 Vibrational Factor

Given the time-scale for photon absorption (~10-15s) is much faster than for nuclear re- arrangement, absorption occurs within a fixed nuclear framework (10-12s-10-13s). Such a transition is called vertical because it represents a transition parallel to the y-axis when viewing molecular transitions as a function of configuration coordinates (see Figure 1-5). From a quantum mechanical standpoint, this corresponds to transitions between initial and final states with a large overlap of vibrational wavefunctions. This overlap can be expressed through the Franck-Condon factors, which contribute to the relative absorption efficiency of different vibrational levels of electronic excited states:

= | (1-11) 2 �퐹 �〈�푓 �푖〉� Vibrational relaxation of the excited state occurs rapidly (10-12s-10-13s) following absorption leading to internal conversion to the lowest vibrational state of the new electronic excited state. Eventually, the molecule will return to its ground state configuration, releasing energy through light or heat. The time scale for internal conversion is much faster than for any of the processes returning the molecule to its ground state and thus these transitions will only occur with significant rates from the ground vibrational state of a given electronic level. This is referred to as Kasha’s rule. The same Franck-Condon Factors apply to the emissive transitions, which leads to mirror symmetry in the absorption and emission spectra of many organic molecules, as depicted in Figure 1-5.

Figure 1-5: Simplified potential energy curves and vibrational wavefunctions and demonstration of Franck-Condon transitions with strength determined by wavefunction overlap. This leads to

1.2.2.8 Spin considerations

Considering electron spin is important in understanding the spectral properties of organic molecules. The Pauli exclusion principle states that no two electrons may have the same quantum numbers, which for our purposes means that electrons in a filled molecular orbital must have antiparallel spins. If one of these two paired electrons is raised into a higher energy level through absorption of a photon the electron spins will remain paired due to the spin conservation rule, however in principle the excited and ground state electrons may have either parallel or antiparallel spin. Excitons composed of excited and ground state electrons with antiparallel and

parallel spin are referred to as singlet and triplet excitons due to the experimental observation that triplet states may be resolved into three separate components under influence of a magnetic field while singlets yield only one state.

The possible arrangements of the spin of the ground and excited state electrons forming an exciton are (1) (2), (1) (2), (1) (2) and (1) (2) where the up and down arrows indicate spin↑-up and↑ spin↓-down↓ electrons,↑ ↓and the numbers↓ ↑ in parentheses are the labels for the first and second electron. Since the wavefunction must be either symmetric or antisymmetric under particle exchange, the four states can be rewritten as:

( ) ( ) : ( ) ( ) ↑ 1 ↑ 2 ( ) ( ) ( ) ( ) ↓ 1 ↓ 2 (1-12) 푇𝑇𝑇𝑇 푆�푆𝑆� � 1 ( ) ( ) ( ) ( ) : √2�↑ 1 ↓ 2 +↓ 1 ↑ 2 � 1 푆𝑆푆𝑆� 푆�푆𝑆 √2�↑ 1 ↓ 2 −↓ 1 ↑ 2 � The first three are the triplet states which have a total spin of one and are symmetric under particle exchange while the last is the singlet state which has a total spin of zero and is antisymmetric under particle exchange.

1.2.3 Excitons

Excitons are quasiparticles composed of electrons and holes bound by the Coulomb attraction. Depending on the strength of this attraction, excitons may be loosely (Wannier-Mott exciton) or tightly (Frenkel exciton) bound. The binding energy of an exciton is defined as the difference in energy between two free charges and the exciton they form and is typically around 0.1eV to 1eV for organic materials and around 1 meV to 25 meV for inorganic materials.

Organic materials have a small dielectric constant (~ 3.5) resulting in tightly bound Frenkel excitons with the electron and hole localized, quite often on a single molecule, while inorganic materials have larger dielectric constants (~12) and weakly bound Wannier-Mott excitons which are easily dissociated at room temperature.

- - + - + +

Wannier-Mott Frenke l Charge Transf er

Figure 1-6: Illustration of Wannier-Mott, Frenkel, and charge transfer excitons. Filled circles represent molecules.

The binding energy of a Wannier-Mott exciton can be approximated by calculating the Coulomb energy between two point charges. The Wannier-Mott exciton binding energy is quantized, in analogy with the hydrogen atom energy levels:

= 2 (1-13) 푒 휇푒푒푒 1 2 2 2 2 퐸퐵 8휖 휖0ℎ 푚 EB is the exciton binding energy, e is the elementary charge, ueff is the effective reduced mass of the electron/hole pair, and m is the energy level of the exciton.

Two additional categories of excitons are useful when describing excited states in organic materials in which distributing an excitation is distributed among multiple molecules: charge transfer excitons and exciplexes.

A “charge transfer state” or “charge transfer exciton” is an exciton in which the electron and hole are localized on different molecules but are still weakly bound. CT excitons occur most commonly at the interface between different materials, at least one of which has strong electron donating or accepting properties. Exciplexes occur when an excitation is shared between multiple molecules, rather than localized on a single molecule. In principle an excited state

distributed over two molecules will generally have some exciplex and some CT character as described by:

= ( ) + ( )+ ( ) + ( ) (1-14) ∗ ∗ − + + − 휓 푐1Ψ 퐴 퐵 푐2Ψ 퐴퐵 푐3Ψ 퐴 퐵 푐4Ψ 퐴 퐵 In this case, when the coefficients c1, c2 >> c3, c4 we refer to the excitation as an exciplex, while if the relative magnitudes are reversed we call the excitation a CT complex. CT and exciplex excitons typically have binding energies much lower than Frenkel excitons (<100meV)

The Wannier-Mott, Frenkel, and CT excitons are illustrated in Figure 1-6. References to excitons in the following text refer to Frenkel excitons unless otherwise specified.

1.2.3.1 Exciton State (Jablonski) Diagrams

It is often convenient to represent the different excitonic states available to a molecule on an exciton state diagram rather than considering the equivalent two electron system and associated electronic states. This equivalency is illustrated in Figure 1-7. The exciton state diagram is referred to as a Jablonski diagram and is shown in Figure 1-8, including vibrational sublevels of each electronic level.

The gaps between singlet states are typically on the order of 1 eV, while the vibrational sublevels are typically spaced by ~150 meV.

The difference between the energies of singlet and triplet excitons depends on the electron-hole separation. To first order, the singlet-triplet splitting scales exponentially with the overlap of the electron and hole wave functions. In inorganic materials where there is minimal overlap between electron and hole wavefunctions, the singlet-triplet splitting is around 1 meV. In organic materials this splitting is several orders of magnitude higher, typically around 0.3 eV to 1 eV for Frenkel excitons and 25 meV to 0.1 eV for CT excitons or Frenkel excitons with the electron and hole wavefunctions localized to separate parts of the molecule.

The emission from singlet and triplet states is referred to as fluorescence and phosphorescence respectively. The timescale of fluorescence (10-9 s to 10-6 s) is usually orders of magnitude shorter than for phosphorescence (10-6 s to 101 s) since the transition from a triplet excited state to the ground state involves a forbidden spin flip. The spin flip converting a singlet into a triplet exciton is referred to as intersystem crossing. For most organic molecules, intersystem crossing occurs over a much longer time scale than the non-radiative triplet to ground state transition and so most triplet excitons do not contribute to the radiative output of the device.

The radiative transitions discussed above compete with nonradiative transitions between the same states. These transitions release heat instead of light, an undesirable process in OLEDs where a central goal is to produce optical energy as efficiently as possible.

Intersystem crossing is enabled by the mixing of singlet and triplet states. The mixing coefficient ( ) is determined by both the magnitude of the spin-orbit interaction ( ) and the

energetic distance휆 between the singlet and triplet states ( ): 퐻푆푆 Δ퐸푆� (1-15) 퐻푆푆 휆 ∝ Δ퐸푆푆 There are two main strategies to enable triplet emission. The first is to decrease the rate of nonradiative transitions competing with phosphorescence. This is usually done by engineering highly deuterated molecules11 however this is not a practical approach on an industrial scale. The second is to increase the rate of intersystem crossing by enhancing spin orbit coupling. Usually these phosphorescent molecules will incorporate a heavy metal atom such as Ir or Pt12,13.

1.2.3.2 Exciton Transfer

Excitons may transfer their energy from one molecule (donor) to another molecule (acceptor). There are three mechanisms through which an exciton can transfer its energy: trivial, Förster, and Dexter energy transfer.

Trivial energy transfer involves the emission and re-absorption of a photon. The exciton decays radiatively on a donor molecule and the emitted a photon is re-absorbed by another molecule 7.

This type of transfer occurs over a length scale determined by the absorption-length of the acceptor material at the emissive wavelength of the donor molecule.

Förster resonant energy transfer (FRET) occurs through dipole-dipole interactions and occurs in situations where there is significant overlap between the donor emission and acceptor absorption spectra. The rate of energy transfer between an isolated donor/acceptor pair is given by:

= (1-16) 1 푅0 6 푘퐸� 휏 � 푅 � where kET is the FRET rate, τ is the exciton lifetime in the absence of any acceptors, and R is the

donor/acceptor separation. R0, the Förster radius is the characteristic length scale for the energy transfer and is determined by the photophysical properties of the donor and acceptor. Typical 14 values for R0 are between 1nm and 10nm. The Förster radius is given by :

( ) = (1-17) 2 6 9 ln 10 푁퐴 휅 Φ퐷 2 4 푅0 128휋 �푅 퐽 where NA is Avogadro number, κ is an orientation factor, ΦD is the donor fluorescent quantum

yield, nR is the index of refraction of the medium containing the donor and acceptor and J is an the overlap integral of a normalized donor emission spectrum and an acceptor absorption spectrum. Efficient Förster energy transfer requires an overlap between the absorption of the acceptor and the emission of the donor. A high overlap gives a large value for J and thus leads

to a large value of R0, which in turn leads to more efficient long range energy transfer. A high absorption coefficient for the acceptor also contributes to a large value for J.

The Förster radius depends on the photoluminescence quantum efficiency of the donor and thus only occurs from donor states where radiative transitions are allowed. The Förster radius also requires overlap of the donor emission and acceptor absorption spectra and thus acceptor states with negligible absorption will also not participate in Förster transfer. The implications of these requirements are that for fluorescent materials in which emission and absorption of triplet states are disallowed, Förster energy transfer is not a significant mechanism of energy transfer. Förster transfer will play a significant role in phosphorescent materials where intersystem crossing allows for radiative transitions between the triplet and ground state.

Dexter energy transfer occurs through the simultaneous exchange of electrons in the HOMO and LUMO levels of the donor and acceptor molecules. Dexter energy transfer relies on wavefunction overlap, thus, due to the exponentially decaying wavefunctions in the radial direction, only has appreciable rates between adjacent molecules, similar to charge transfer. This sets the length scale of Dexter energy transfer at approximately 0.1nm to 1nm. The rate of Dexter energy transfer is given by:

(1-18) 2푅 푘퐸� ∝ 퐽�퐽퐽 �− � � where a is a characteristic length scale set by the decay of the electron wavefunctions away from the molecule, typically ~0.5 nm. J is the spectral overlap integral normalized to the extinction coefficient of the acceptor.

Figure 1-9: Illustration of the electron transitions involved in Förster and Dexter energy transfer.

Although Dexter energy transfer of singlet excitons is allowed, Förster energy transfer will usually be the dominant mechanism of singlet energy transfer

The energy transfer mechanisms described above may each occur between molecules of the same or different species. In an organic solid state device such as an OLED where multiple materials are used, consideration of energy transfer between dissimilar molecules is often necessary.

1.2.3.3 Exciton Formation

In an OLED excitons are typically formed at an organic heterojunction by uncorrelated charge carriers (i.e., electrons and holes), injected from opposite electrodes. In the vast majority of practical devices, the ratio of singlet and triplet excitons is 1:3 due to the degeneracy of the two states and assuming similar formation cross sections, although there have been reports of deviations from this ratio15. Exciton formation will occur primarily at the heterojunction between the electron transporting layer (ETL) and hole transporting layer (HTL) where positive and negative charges tend to accumulate. This region, referred to as the exciton formation zone is often localized very close to the ETL/HTL interface.

Exciton formation is usually quantified using a Langevin recombination model which expresses the rate of exciton formation as:

= (1-19)

퐺 훾�푛푒푛ℎ where G is the exciton generation rate, ne and nh are the electron and hole concentrations respectively, and is the Langevin recombination rate constant given by:

훾� = ( + ) (1-20) 푒 훾� 4휋�휖0 �푒 �ℎ where and are the electron and hole mobilities. Note that in an OLED ne and nh are

functions�푒 of the�ℎ position in the device.

There are two primary pathways to exciton formation in an OLED: 1) exciton formation on the host material(s) and 2) exciton formation on the dopants. In either case excitons tend to form where there are high concentrations of electrons and holes in a specific location.

Due to the energetic and mobility offsets at the ETL/HTL heterojunctions, these will usually be the primary location excitons are formed. Determining the precise shape of the exciton generation zone is very difficult, however the assumption of an exciton generation zone which decays exponentially away from the ETL/HTL interface was sufficient for several previous studies16,17:

= exp (1-21) 푥 퐺 퐺0 �− �� where G0 is the interface generation rate and g is the width of the exciton generation zone. Previously values of g are 3 nm16 and 0.5 nm17.

1.2.3.4 Exciton Migration

Exciton migration in an organic film occurs through multiple short range hops between similar or dissimilar molecules. In a uniform film, this motion can be described by a random walk which is often approximated using a diffusion equation:

= + 2 (1-22) 휕� 휕 � � 2 휕� 퐷 휕푥 − 휏 퐺 where n is the exciton density, t is time, x is position, τ is the exciton lifetime, and G is an exciton generation term, for example the exponential term given by equation 1-21.

In an organic device with multiple layers, possibly containing mixtures of different materials, the diffusion equation must be modified to consider the effects of these new materials. For example, at interfaces between layers exciton reflection by a higher energy material can be expressed as:

= 0 (1-23) 휕� −퐷 휕��푥=0 In addition to the effects of interfaces, bulk effects such as energy transfer in a mixed film can be accounted for by introducing additional terms into equation 1-22.

The spatial and energetic disorder present in an amorphous organic film lead to a Gaussian distribution of exciton and electronic states. Depending on the degree of the disorder it may or may not have a significant impact on exciton transport. Studies of exciton diffusion, quenching, etc. largely ignore this energetic disorder as the experimental and computational techniques necessary to accurately assess energetic disorder are significantly more complicated than for the disorder-free case. In this work we will neglect energetic disorder, however assessing its impact is suggested as a focus of future work.

1.2.3.5 Bimolecular Exciton Interactions

When exciton and charge carrier concentrations become large enough, exciton-exciton and exciton-charge interactions need to be taken into account. The interactions often contribute to “exciton quenching”, the set of processes leading to increased nonradiative decay, and thus lower efficiency. Since the concentration of charges and excitons increases with the applied current density, every OLED experiences exciton quenching given high enough current densities. The drop in efficiency at high excitation levels due to increased rates of exciton quenching is termed “OLED efficiency roll-off”.

Triplet triplet anihilation is often the most detrimental quenching process in an OLED and follows one of the two pathways shown below:

+ + + 2 3 (1-24) 1 → 푇� + 푆0 → 푇1 + 푆0 푇1 푇1 → 푋 � 4 5 → 푆� 푆0 → 푆1 푆0 where Tn is the n’th triplet state Sn is the n’th singlet state and X is an intermediate state, S0 is the gorund state. The arrows indicate the processes of: 1) formation of intermediate state 2) transfer to a high level (“hot”) triplet state plus a ground state molecule 3) thermalization of the excited triplet 4) transfer to a high level (“hot”) singlet state plus a ground state molecule 5) thermalization of the excited singlet state.

Singlet-triplet annihilation is usually the most significant quenching process affecting the singlet exciton population. This process involves the singlet exciton donor transferring its energy to a

triplet exciton acceptor through a Förster energy transfer after which the excited triplet rapidly thermalizes:

+ + + (1-25)

푆1 푇1 → 푆0 푇� → 푆0 푇1 Singlet-singlet annihilation follows an analogous pathway to singlet-triplet annihilation with a second singlet exciton playing the role of the triplet:

+ + + (1-26)

푆1 푆1 → 푆0 푆� → 푆0 푆1 This process uncommon during normal OLED operation as the concentration of singlet excitons is very low due to their short lifetimes.

Singlet-polaron and triplet-polaron annihilation are also possible. Each results in an excited charged state as an intermediate which thermalizes to release its excess energy:

+ + + (1-27) ∗ 푇1 � → 푆0 � → 푆0 � + + + (1-28) ∗ 푇1 � → 푆0 � → 푆0 � where p and p* indicate relaxed and excited charges respectively. The effect of these processes on the exciton population is usually accounted for using rate equations. For example the time dependence of a uniform triplet population in an organic material can be described by:

= (1-29) 휕�� 1 2 휕� − 휏� − 2 푘푇푇푛푇 where nT is the triplet exciton concentration is the monomolecular triplet exciton lifetime and

kTT is the bimolecular triplet-triplet annihilation휏푇 rate. In an OLED with a uniform exciton recombination zone with width w, where the efficiency is limited by triplet-triplet annihilation the external quantum efficiency (EQE) is given by18:

= 1 + 1 (1-30) 퐸퐸퐸 퐽0 8퐽 퐸퐸퐸0 4퐽 �� 퐽0 − � = (1-31) 4푒푒 2 퐽0 푘푡푡휏� 24

1.2.4 OLED Degradation

1.2.4.1 Literature and Commercial Lifetime

State of the art OLEDs have been reported with half-lives over 1 million hours at 1000 Cd/m219; however record OLED lifetimes vary depending on the type and color of emitter used. Wider band-gap emitters (blue emission) tend to have shorter lifetime than narrow band-gap materials (red or near IR emission), and phosphorescent emitters are typically less stable than fluorescent materials with similar emission wavelengths. This is evident in Figure 1-1020, which documents progress in OLED lifetime over the previous few decades. The relative instability of blue emitters is often attributed to the small difference between exciton energies and bond dissociation energies21.

1.2.4.2 Degradation Mechanisms

Some degree of degradation of OLED performance is unavoidable in all devices. This degradation may manifest itself in several ways, including catastrophic failure, dark spot growth, luminance loss and voltage rise.

Catastrophic failure is the sudden, irreversible failure of an OLED. This type of degradation is usually due to either the formation of a short in the device which provides an alternative pathway for current, or a runaway or due to a positive feedback loop, for example between device temperature and conductivity. For precisely engineered and carefully stored and tested devices, catastrophic failure is not typically an issue.

Dark spot growth was a dominant degradation mechanism in early OLEDs. It is characterized by circular nonemissive regions of the device which grow in diameter over time due to degradation of the cathode and/or cathode/organic interface. There are several mechanisms which may lead to dark spot growth, by far the most common being reactions involving moisture and oxygen which have penetrated into the device. In most cases, proper encapsulation is able to effectively mitigate this mode of degradation.

Luminance loss and voltage rise describe the decrease in emitted light and rise in driving voltage at a given current density as an OLED ages. Each may be a result of degradation at the electrodes, however in state-of-the-art devices which are well encapsulated, luminance loss and voltage rise are most commonly attributed to the formation of charge traps and nonradiative recombination centers in the organic layers of the device. Unlike dark spot growth, which is most often arises through extrinsic degradation, these modes of degradation are usually intrinsic, which is to say they arise during operation even in well encapsulated devices.

Luminance loss and voltage rise often occur simultaneously in degrading OLEDs, and on similar time scales. The connection between the electrical and luminance aging has been made explicit for several materials sets in which the concentration of trapped charge (presumably trapped on degradation induced defects) as a function of time correlates with the luminance loss23, suggesting that these two processes are fundamentally connected, although it is unclear whether they are two manifestations of the same underlying physical process, or reflect two independent but connected physical processes.

The work presented here focuses on intrinsic degradation leading to OLED luminance loss. The formation of charge traps and exciton quenching centers in OLEDs can be driven by charges24; singlet or triplet excitons25–30; or bimolecular interactions of charges and excitons 31–33 or multiple excitons 34.

1.2.4.3 Luminance/lifetime trade-off

The concentrations of singlet and triplet excitons, and charges increase with current density, leading to faster rates of degradation at higher current density. Since higher currents are necessary to increase the brightness of a given OLED, there is a fundamental trade-off between device brightness and lifetime. This creates a barrier to the adoption of OLEDs for solid state lighting, as the brightness needed for this application (~5000 Cd/m2) is much higher than for display applications (~500 Cd/m2), leading to lower than acceptable lifetimes.

The following empirical relationship has been widely observed between the brightness and lifetime of OLEDs:

(1-32) � 1 0 퐿 ∝ �50 For many devices, especially OLEDs using fluorescent emitters, brightness is linearly proportional to current density up to high brightness so that the following relationship also applies:

(1-33) � 1 퐽 ∝ �50 The value of n will differ slightly between equations 1-32 and 1-33 if the relationship between the current density and brightness is not linear. The value of n is referred to as the acceleration factor as it characterized the accelerating rates of degradation at elevated current density or brightness. Values of n typically range from 1 to 2. A value of 1 is characteristic of monomolecular exciton driven degradation while larger values may possibly indicate bimolecular processes, or the effects of increased temperature at elevated current densities.

1.2.4.4 Modelling Degradation

The time dependence of OLED degradation is most commonly described using empirical equations. These equations have no physical origin or interpretation and are judged on their ability to minimize fit residuals and extrapolate luminance loss forward in time. The most commonly used empirical formula is the stretched exponential (SE) function:

= exp (1-34) � � 훽 �0 �− �휏� � Physical models for OLED degradation have either been based on coupled rate equations. A model by Giebink et al. considered OLED degradation due to the formation of charge traps and exciton quenching centers 35. The authors fit experimental data under several scenarios including degradation driven by charges; triplet excitons; exciton-charge interactions; and exciton-exciton interactions. Based on the quality of the fits to the time dependent luminance and voltage of their devices under these different scenarios, the authors concluded that exciton-charge

interactions were the source of degradation. A later study using a similar model using a different device architecture concluded that triplet-triplet interactions led to degradation34.

Chapter 2 Experimental Methods

2.1 Materials

The materials used in this work were mainly commercially available mainly organic small molecules purchased from Lumtec with purity >99% and used as purchased.

The inorganic materials MoO3 (99.9% pure) and LiF (99.9% pure) were used as injection layers and purchased from Sigma Aldrich. Al was used for the cathode and purchased in pellet form with 99.999% purity from Kurt J. Lesker co. Substrates were pre-patterned ITO coated glass with a sheet resistance of 15Ω/square.

2.2 Fabrication

OLED samples were fabricated on pre-patterned ITO coated glass in a Kurt J. Lesker Luminos Cluster Tool with a base pressure of ~10-8 torr shown in Figure 1. Prior to deposition, the substrates were cleaned with a standard regiment of Alconox®, acetone, and methanol, followed by UV ozone treatment for 15 minutes.

Each substrate is divided into eight sections as shown in Figure 2-2. Using a set of interchangeable shadow masks different organic materials can be deposited on subsets of these regions of the substrate. Each of the eight sections contains four pixels defined by the overlap of the pre-patterned ITO anode and the vacuum deposited Al cathode.

Figure 2-1: Kurt J. Lesker Luminos Cluster Tool

Examples of an organic and Al shadow mask are shown in Figures 3a and 3b. The shadow masks are positioned below the substrate using a set of interlocking mask and substrate holders as shown in Figure 2-3c. Often a set of OLEDs fabricated on one substrate will share some layers in common while systematically altering others in order to study the effect of one aspect of OLED design on performance. This is achieved by switching between shadow masks blocking different portions of the substrate between subsequent depositions.

Figure 2-3: a) Example organic shadow mask b) Al shadow mask and c) substrate holder on top of shadow mask holder demonstrating alignment of mask and substrate.

Following the substrate cleaning procedure, samples are loaded into the cluster tool load lock chamber after which time device fabrication takes place entirely under vacuum.

Layer thicknesses were measured in-situ using a quartz crystal monitor, and calibrated by profilometry using a KLA-Tencor P16+.

Thin films are deposited by thermal evaporation at pressures of ~1E-6 Torr for metals and ~1E-7 Torr for all other materials. Deposition temperatures for organic molecules range from ~150C to 300C.

2.3 Characterization 2.3.1 OLED characterization

Current-voltage data was measured using a Keithley 6430 source meter or a HP4140B picoammeter. OLED electroluminescence spectra were collected using an integrating sphere connected by an optical fiber to an Ocean Optics USB4000 spectrometer. Luminance was measured using a Minolta LS-110 Luminance meter.

a b

Figure 2-4: a) Minolta LS-110 luminance meter and b) integrating sphere connected by optical fiber to an Ocean Optics USB4000 spectrometer

2.4 Device Architecture 2.4.1 Simplified OLED

The device structures used in much of this work are closely related to the “simplified OLED” structure recently reported by previous members of this research group and shown in Figure 2-5. This OLED structure drew considerable attention due to its very simple device architecture and record efficiency36–39. This OLED structure is ideal as a framework for the study of excitonics for three reasons: 1) The simple OLED structure makes interpretation of experimental results more straightforward 2) The high efficiency of monochromatic and white OLEDs based on this structure mean that an understanding of excitons in this structure will be relevant for high efficiency OLEDs and of interest to the OLED community and 3) The structure uses only common, well characterized organic materials allowing for comparison of results to the existing literature.

The simplified OLED structure uses ITO as the anode, MoO3 as a hole injection layer, 4,4'- N,N'-dicarbazole-biphenyl (CBP) as the hole transport layer (HTL), 1,3,5-tris(N- phenylbenzimidazole-2-yl)benzene (TPBi) as the electron transport layer (ETL), and bis(2-

phenylpyridine)(acetylacetonate)iridium(III) (Ir(ppy)2(acac)) doped CBP as the emissive layer (EML), LiF as the electron injection layer, and an aluminium cathode.

2.4.2 Exciton Capturing Layers

In many of the studies presented here, exciton capturing layers will be used to probe the exciton distribution. Exciton capturing layers are thin doped layers introduced at various positions in the organic stack with the intention of capturing a portion of the local exciton population. By varying the position of these layers and comparing the intensity of the emission from the capturing layer and/or the material from which the excitons are being captured, information about the exciton distribution can be determined.

This concept was first used by Tang et al. in their report of the first doped OLED3, with following reports refining the technique and analysis, mainly for the purpose of measuring exciton diffusion lengths 15–17,40,41.

An example of a device architecture including exciton capturing layers is shown in Figure 2-6a. Assuming the exciton formation region is localized close to the interface between the ETL and host, the capturing layer and host emission will appear similar to Figure 2-6b. As the capturing layer is moved further from the exciton generation zone, fewer excitons are transferred from the host to the capturing layer, resulting in a gradual drop in capturing layer emission and increase in host emission. The characteristic shape of these curves can be used to determine the exciton diffusion length.

Figure 2-6: Example of a) device architecture and b) emission intensity of different layers of an OLED incorporating exciton capturing layers.

If a fluorescent probe is used in the exciton capturing layer, only captured singlet excitons will be emitted. Alternatively, triplet excitons can be probed by using a phosphorescent emitter in the exciton capturing layer. This work will focus on singlet exciton dynamics using fluorescent emitters in the exciton capturing layer.

Chapter 3 Exciton Diffusion in CBP

This chapter originally published as: Ingram, G. L., Nguyen, C. & Lu, Z. H. Long-Range Energy Transfer and Singlet-Exciton Migration in Working Organic Light-Emitting Diodes. Phys. Rev. Appl. 5, 64002 (2016). Copyright 2016 American Physical Society.47

3.1 Introduction

Unprecedented performance of recent state of the art organic light emitting devices (OLEDs) 48,49 and the adoption of OLEDs for consumer electronics demonstrate the incredible progress made in the field since the first OLEDs were fabricated several decades ago 2. Researchers have now turned to the new challenge of fabricating white OLEDs with similarly high efficiency and long lifetime. The need for multiple emitters and high current densities in order to meet the strict technical requirements of high color quality and high luminance white light makes the design of white OLEDs more challenging than for their monochrome counterparts. In order to establish rational device design principles, a better understanding of certain fundamental processes, such as the diffusion, and energy transfer of excitons in a working device is becoming critical. Here, we investigate these fundamental processes in the archetype host and hole transport material CBP.

Many methods for measuring exciton diffusion length have been reported. These methods can be broadly grouped into those using optical or electrical excitation to produce an exciton population. For optical methods, the organic samples are excited by laser light or monochromated light from a broadband source. The sample’s photoluminescence 50–52 or photocurrent 53,54 is then monitored as a function of sample thickness, time, or excitation wavelength. The exciton diffusion length is then extracted by modelling the optical absorption profile and exciton diffusion. These techniques tend to be limited to materials with a significant photoluminescence

quantum yield or strong absorption, thus limiting their usefulness. Moreover, the broad region over which the light is typically absorbed leads to a broad exciton generation zone that is often much larger than the exciton diffusion length, which further complicates the analysis. The alternative to these optical methods is measurement of the diffusion length by producing excitons electrically in an OLED. In this case, the excitons are generated in a thin region localized at the interface between the electron transport layer (ETL) and the hole transport layer (HTL). Several variations in the measurement technique have been proposed, including incorporating a thin doped sensing layer at different positions in the OLED41,55 or varying the thickness of a spacer layer3,15–17,40 between the exciton formation and emissive regions. Here, we use the latter OLED-based method to study singlet exciton diffusion in CBP. The OLED-based measurements allow us to establish applicable physical parameters of exciton diffusion in working devices. In addition to carefully considering subtleties of this technique emphasized in previous studies, such as of the device boundary conditions, the finite width of the exciton generation zone, and the possibility of direct exciton formation, we also provide an in-depth investigation of long- range energy transfer known as FRET from the host to the sensing dopants. We show that this consideration is important as the “effective” diffusion length can be significantly enhanced relative to the “intrinsic” diffusion length under certain experimental conditions if FRET is not properly accounted for.

We note that studies of singlet exciton diffusion in host materials may also provide insight into triplet diffusion in phosphorescent materials in which triplets diffuse and quench through similar Förster type transitions as singlets51,56,57.

The choice of CBP and 1,3,5-tris(Nphenylbenzimidazole-2-yl)benzene (TPBi) as the HTL and ETL respectively, was motivated by the high performance of OLEDs based on these materials36,46. Understanding how excitons form and diffuse near the CBP/TPBi heterojunction could contribute to the understanding of the success of this materials set. Experimental details are provided in Section 2. Section 3 presents and compares models of exciton diffusion with and without including FRET from the host to the sensor. Experiments measuring the singlet exciton diffusion length in CBP are outlined in Section 4 followed by conclusions in Section 5.

3.2 Experimental Methods

To study singlet exciton diffusion we fabricated sets of OLEDs with thin sensing layers doped with a fluorescent emitter placed at different distances from the ETL/HTL interface. By using a fluorescent sensor any triplet excitons harvested by the sensing dopants will not contribute to the emission of the device. Each set of devices is fabricated on a single substrate and share common transport and emissive layers. This eliminates potential run-to-run variations in device fabrication, particularly at the anode and cathode interfaces, and thus makes it possible for direct comparison between devices within a set.

The HOMO/LUMO and singlet exciton energies of the organic layers used here are shown in

Figure 3- 1. ITO was used as the anode, MoO3 as a hole injection layer, CBP as the hole transport and host material under investigation, 2wt% rubrene doped CBP as the sensing layer, TPBi as the ETL, LiF as the electron injection layer, and an aluminium cathode. The devices were encapsulated in-situ by thermally evaporating a 200 nm thick layer of silicon monoxide source material. The thickness of the transport layers are adjusted such that the sensing layer remains in the same position in the cavity as its position relative to the ETL/HTL interface is varied. This ensures no variations in optical outcoupling between devices.

Excitons may either form on the transport materials or directly on the rubrene molecules. Those formed on the transport materials are assumed to primarily form near the ETL/HTL interface where mobility and charge transport energy offsets lead to the accumulation of charge. It is also possible that electrons penetrating into the CBP layer will recombine with holes trapped on the rubrene molecules forming excitons directly on the dopants. These two processes will both be accounted for in the following sections.

TPBi has a singlet energy that is 0.2 eV larger than that of CBP. This will inhibit singlet transfer from CBP to TPBi, leading to exciton blocking at this interface. CBP singlet excitons that diffuse to the rubrene sensing layer are harvested by the lower energy rubrene dopants and a fraction of those harvested excitons emit photons detected outside of the device. The following section will demonstrate how modelling the emission from these sensing layers allows for the measurement of the singlet exciton diffusion length.

Figure 3-1: a) The LUMO and HOMO levels (upper and lower lines) and b) the singlet exciton energies of layers for devices used to measure the singlet exciton diffusion length in CBP.

3.3 Theory

Singlet exciton diffusion in a thin film will occur through a series of Förster and/or Dexter type energy transfers between like molecules. In our measurements, the path of a singlet exciton will include such energy transfers between host molecules but may also include long range FRET to the sensing dopants. Although strictly speaking hops between host molecules may occur through a Förster mechanism, we will reserve the term FRET for energy transfer between two different molecular species (e.g. the host and the dopant) while referring to energy transfer between similar molecules (e.g. two CBP molecules) simply as diffusion.

Section 3.1 will present an analytical model for the exciton population in the OLEDs described in the previous section when FRET is ignored. Section 3.2 will outline how to account for the

effects of FRET. Section 3.3 will quantify the systematic errors introduced into singlet exciton diffusion length measurements when FRET is not appropriately compensated for by comparing the models in Sections 3.1 and 3.2.

3.3.1 Modelling Singlet Diffusion without FRET

In the absence of FRET, we model the exciton motion inside an OLED using the steady state diffusion equation:

+ exp = 0 2 (3-1) 휕 �푆 �푆 푥 2 퐷 휕푥 − 휏푆 퐺0 �− �� The terms in equation 3-1, from left to right, account for the diffusion, natural decay, and generation of excitons. The variable x measures the position relative to the ETL/HTL interface;

nS, D and are the density, diffusivity and natural lifetime of the singlet excitons, respectively;

G0 is the rate휏 of exciton generation at x=0, and g is the width of the exciton generation zone (the region in which excitons initially form). Following the example of Wünsche et al., the exciton generation zone is assumed to have an exponentially decaying profile away from the ETL/HTL interface, with a characteristic decay length g. Exciton quenching is assumed to be negligible.

As TPBi has a singlet exciton energy that is 0.2 eV larger than that of CBP, singlet excitons will be blocked at the ETL/HTL interface. Assuming perfect blocking

= 0. (3-2) 휕�푆 휕� �0 At the sensing layer (x=d), CBP singlet excitons will be efficiently transferred to the lower energy rubrene molecules, and given perfect quenching we can use the boundary condition

( ) = 0. (3-3)

푛푆 푑 This boundary condition is validated in Appendix A. If we define the exciton diffusion length as

= . (3-4)

퐿 �퐷휏푆 The exciton density given by the solution of equation 3-1 and using boundary conditions in equations 3-3 and 3-4 is given by

( ) = 푥 푥 푥 exp exp . (3-5) �퐺0휏푆 �� exp�−푔�+� exp�퐿�� cosh�퐿� 푥 푥 푆 2 2 푑 푛 � � −� � cosh�퐿� − 퐿 �� − 푔 �− �� The intensity of emission from the sensing layer ( ) will be proportional to the rate of exciton

capture by the sensing dopants. For a sensing layer퐼 of thickness 푙 ( ) + exp + . (3-6) 휕� 푑+� 푥 푃� 0 푒 퐼 푑 ∝ � � �−퐷 휕��푑 ∫푑 퐺 �− �� 푑� 푖 � The exciton capture has three sources corresponding to the three terms in equation 3-6. From left to right, these are: 1) exciton diffusion into the sensing layer followed by harvesting on the sensing dopants 2) exciton generation on the host inside the sensing layer due to the distributed nature of the exciton generation zone followed by harvesting on the sensing dopants 3) exciton formation directly on the dopants due to recombination of trapped charges with a non-negligible electron current ( ) through the CBP. Following the examples set in the literature, the electron

current will be assumed푖푒 to be a constant through the CBP layer. To relate the measured intensity to the exciton capture rate, the photoluminescence quantum efficiency ( ) and light out- coupling ( ) must also be considered. The photoluminescence and optical�푃� out-coupling efficiencies� are the same for each device within each set of OLEDs fabricated. Substituting equation 3-5 into Equation 3-6 and grouping constants we arrive at:

( ) = exp + exp tanh exp + exp (1 (1 2 � 푑 � 푑 푑 푑 푑 2 2 퐼 푑 −퐴 � −� �� − �� − �� − �� − − )(1 exp )) + . (3-7) � 2 � �� − �− �� 퐵 A and B in the above expression are fitting parameters that are proportional to the rates of exciton formation on the host ( ) and on the dopants, respectively. The generation zone width

and exciton diffusion length are푔 also퐺0 free parameters.

3.3.2 Modelling Singlet Diffusion Including FRET

It has been previously shown that FRET can lead to overestimates of singlet exciton diffusion lengths measured by photoluminescence quenching in a bilayer configuration 50. In addition, it

was pointed out by Hoffman et al. that a similar overestimation would occur in OLED-based measurements of singlet exciton diffusion if FRET was not taken into account.

To account for FRET, we model exciton diffusion as follows

+ exp = 0 2 (3-8) 휕 �푆 �푆 푥 2 퐷 휕푥 − 휏푆 − 푘퐹푛푆 퐺0 �− �� where is the FRET rate from the host to a thin doped sensing layer given by

푘퐹 ( ) = . (3-9) ( ) ( ) �퐴휋 6 1 1 3 3 푘퐹 � 6휏 푅0 � 푑−푥 − 푑−푥+� � is the concentration of the sensing dopants and is the Förster Radius, a characteristic

�length퐴 scale for the energy transfer between the donor푅0 and acceptor which is given by:

( ) = 1 2 (3-10) 9 𝑙 10 푁퐴 휅 훷퐷 6 2 4 푅0 � 128휋 �푅 퐽� where NA is Avogadro’s number, κ is an orientation factor, is the donor fluorescent quantum

yield, n is the index of refraction of the medium containing Φthe퐷 donor and acceptor and J is the overlap integral of a normalized donor emission spectrum and an acceptor absorption spectrum.

The OLED emission intensity in this case is given by:

( ) = + 푥 + ( ) ( ) + (3-11) 휕� 푑+� −푔 푑 0 퐹 퐼 푑 퐴′ �−퐷 휕��푑 ∫푑 퐺 푒 푑� ∫0 푘 � 푛 � 푑�� 퐵′ where A’ and B’ are fitting parameters analogous to A and B in the previous section and n(x) is the solution to Equation 3-8 using boundary conditions from Equations 3-2 and 3-3. Analytical solutions to Equation 3-8 are not available and so n(x) is determined by solving Equation 3-8 numerically using as defined in Equation 3-9. To avoid diverging at x=d, its value is capped at 1000/τS. 푘퐹 푘퐹

3.3.3 Comparison of Models with and without FRET

It may be tempting to assume that the diffusion length enhancement from FRET will be approximately equal to the Förster radius. This is not necessarily true, and the extent to which

energy transfer will enhance singlet exciton diffusion is dependent on the particular experimental conditions. We will now compare the models presented in the previous two subsections, in order to quantify the effect of FRET on singlet exciton diffusion length measurements.

To understand how FRET influences sensor emission intensity, Equation 3-11 was plotted in

Figure 3- 2 using R0=0nm, 3nm, and 5nm. The values of each parameter used in the calculation are shown in Table 3-1. As the rate of FRET increases, the emission intensity increases.

Note that τS was chosen to avoid rounding errors and its value does not affect the results.

To determine how closely each of the sets of data in Figure 3-2 correspond to the analytical model in Section 3.1, they were each fit with Equation 3-7 (setting g=1nm to match the value from Table 3-1), and using A, B and L as fitting parameters. The fits are plotted as dashed black lines in Figure 3- 2.

Table 3-1: Parameter values used in the calculation of OLED emission

Parameter Value(s)

A’ 1

B’ 0

L [nm] 5

g [nm] 1

-3 CA [nm ] 0.05

L [nm] 3

-3 -1 G0 [cm s ] 1

τS 1s

R0 [nm] 0, 3, 5

d [nm] 5, 10, 15, 20, 25

Figure 3-2: Sensor emission intensity calculated using numeric solutions to Equation 3-11 for

R0=0nm (black circles), R0=3nm (blue triangles) and R0=5nm (red squares). The data is fitted with Equation 3-7 (dashed lines) which does not take into account long range energy transfer

from the host to dopants. For the R0=3 and R0=5, erroneous diffusion length values of 5.2nm and 6.5nm are extracted using Equation 3-7.

The fits (not accounting for energy transfer) match closely with the data points (accounting for 2 energy transfer), with R >0.99 for all three fits. For the R0=0nm case, the “effective” diffusion length extracted from the fit agrees exactly with the 5nm value used in the numeric model, as

expected. The diffusion length extracted from fits to the data using R0=3nm and 5nm are 5.2nm and 6.5nm, respectively. In other words, for a material with a 5nm diffusion length and experimental conditions reflecting the parameter choices given above, the effective diffusion length measured would overestimate the true, or “intrinsic” diffusion length of the sample by

0.2nm (4%) and 1.5nm (30%) using materials with Förster radii of 3nm and 5nm, resectively, if FRET was not considered.

The Förster radius between the host and sensor (R0), and the sensor concentration (CA) both contribute to rate of FRET. Figures 3-3a and 3-3b plot the effective diffusion length for different values of R0 and CA, for materials with an intrinsic diffusion length of 2.5nm and 5nm, respectively. The effective diffusion lengths are extracted from fits to Equation 3-7 of data calculated using Equation 3-11. The calculated data use the same parameters as those listed in Table 3-1, unless otherwise indicated.

Figure 3-3: Effective diffusion length (shown in color scale) for an intrinsic diffusion length of

We can see clearly in Figure 3-3 that the effective and intrinsic diffusion lengths can differ greatly when R0 and/or CA become large, especially if the intrinsic diffusion length is small.

When R0 and CA are small, the intrinsic and effective diffusion lengths are similar as the small

rate of FRET does not produce an exciton distribution significantly different from what would be observed in the absence of FRET.

In the following section we will take several steps to minimize systematic errors due to FRET based on our findings here. First, we will fit experimental data using Equation 3-11, rather than

Equation 3-7. If precise values of R0 were available then this would be sufficient to eliminate the systematic error FRET introduces into measurements of singlet exciton diffusion. However, we note that calculated and experimental values of R0 reported in previous studies differ greatly

suggesting that calculating R0 is not always straightforward. To avoid uncertainty in our

calculated value of R0 leading to uncertainty in our measured value of the diffusion length, we will take additional steps to reduce the rate of FRET in our devices. To this end, we will

minimize R0 by choosing a suitable sensing dopant and minimize CA by using a low dopant concentration in the sensing layer (while being careful to maintain the boundary condition from

Equation 3-4). By minimizing R0 and CA, we reduce the rate of FRET such that the distribution is primarily determined by diffusion rather than energy transfer.

3.4 Results

This section will provide details on the measurement of the singlet exciton diffusion length in CBP.

3.4.1 Calculation of the Förster Radius

Figure 3-4: CBP thin film photoluminescence (blue solid line) and Rubrene absorption (black dashed line) in toluene solution.

Equations 3-9 and 3-10 indicate that the rate of FRET depends on the overlap between the donor emission and acceptor absorption spectra. The CBP thin film photoluminescence and rubrene solution absorption, shown in Figure 3-4, were used along with Equation 3-10 to calculate a Förster radius of 2.5nm (the values in Figure 3-4 have been normalized for clarity however absolute values are used in the calculation) . Rubrene absorption was taken in solution rather than thin film to simplify calculation of the molar extinction coefficient.

Figures 3a and 3b indicate that for an intrinsic exciton diffusion length between 2.5nm and 5nm, the enhancement due to FRET for a Förster radius of 2.5nm will be small as long as the sensor concentration is kept low.

3.4.2 Singlet Exciton Diffusion Length Measurement

The current density-voltage (J-V) characteristics of our devices are shown in Figure 3-5a. Differences in the J-V characteristics are not significant between devices, suggesting similar internal electric field distributions. We will demonstrate in section 4.3 that changes in charge balance do not affect the measured exciton diffusion length and thus we expect that slight shifts in charge balance due to charge trapping on the rubrene dopants will not affect our results.

Emission spectra for devices with a fluorescent sensing layer at different positions at a current density of 12mA/cm2 are shown in Figure 3-5b. As the sensing layer is moved away from the ETL/HTL interface, the sensor emission intensity (centered at 560nm) decreases while the host emission (centered at 390nm) increases as expected.

Figure 3-5: a) Current density-voltage characteristics and b) OLED emission spectra at 12mA/cm2 for devices with sensing layers placed at 5nm, 10nm, 15nm, 20nm and 25nm from the ETL/HTL interface.

The peak intensity of the sensor emission is plotted against the distance from the ETL/HTL interface in Figure 3-6 over a range of current densities. Fits to Equation 3-11 using A’, B’ and L as free parameters and setting g=1nm are plotted in Figure 3-6. The choice of g will be justified in section 4.3. Equation 3-11 provides good fits to the data over a wide range of current densities.

The singlet diffusion lengths extracted from the fits are plotted in Figure 3-7. The diffusion length remains roughly constant as the current density is varied from 5 mA/cm2 to 450 mA/cm2. Given the small variations in L as a function of current density and also taking into account the combined effects of random variations in spacer thicknesses, we report a value for the singlet exciton diffusion length in CBP of 4.3 ± 0.3 nm. Using the previously reported value for the CBP singlet exciton lifetime in CBP of 0.7 ns 51, this corresponds to a diffusivity of (2.6 ± 0.3)x10-4 cm2/s. For comparison, using Equation 3-7 produced an extracted diffusion length of

4.6 ± 0.2 nm. This shows that even with our low R0 and low dopant concentration including FRET in our model still provides a slight refinement on models that do not account for FRET.

Our value for the singlet diffusion length in CBP differs significantly from the previously reported value of 16.8nm measured by spectrally resolved photoluminescence quenching51. Further work would be necessary to determine if the differences were due to intrinic differences in the films studied, or in the interpretation and analysis of experimental data.

The constant diffusion length as a function of current density indicates that no bulk exciton quenching is occuring in the region being probed. This validates the omission of quenching terms in Equations 3-1 and 3-8. This differs from a previous study of singlet exciton diffusion in OLEDs which showed a decrease in the extracted diffusion length of an N,N’-di-1-naphthalenyl- N,N’-diphenyl-(1,1’:4’,1’’:4’’,1’’ ’-quaterphenyl)-4,4’’ ’-diamine (4P-NPD) at high current density which was attributed to exciton quenching processes17.

Figure 3-7: Extracted singlet exciton diffusion length as a function of current density.

The extracted values for A’, and B’ from the fits of the data in Figure 3-6 with Equation 3-11 are shown in Figure 3-8a. The same parameters are divided by the current density in Figure 3-8b. Clearly the fitting parameters have very different current dependencies, which is expected as they represent excitons formed through two different mechanisms. As the current density rises, the sensor emission becomes less dominated by excitons formed on the host.

We can see from Figure 3-8b that direct dopant exciton formation increases proportionally to the current density, suggesting that the emission is limited by the CBP electron current and not by the concentration of dopant molecules.

3.4.3 Generation Zone Width

In this section we will justify our choice of the generation zone width used in the previous subsection.

Excitons are typically generated in a narrow region of finite width close to the ETL/HTL interface, referred to as the exciton generation zone. Wünsche et al. have previously pointed out that neglecting the finite width of the generation zone may lead to overestimation of exciton diffusion lengths, especially for singlets which tend to have relatively short diffusion lengths16. The generation zone width is particularly important to consider for CBP which has a large minority carrier mobility, a quality that is expected to lead to wider generation zones.

The shape of the generation zone is difficult to measure directly. Previous studies have inferred the generation zone width from fits of experimental data16,17. We will now demonstrate that this is impractical when using our model, which differs from previous models in that it accounts for FRET as well as the harvesting of excitons generated inside the sensing layer.

Sensing layer emission data from the previous section at 10 mA/cm2 is plotted in Figure 3-9. Fits to Equation 3-11 with g fixed at 1nm and 4nm are also plotted in Figure 3-9a. Although the fits are nearly indistinguishable using g=1nm and g=4nm, the extracted diffusion lengths are 4.5nm and 2.3nm, respectively. The extracted diffusion lengths for multiple fits with different values for g are shown in Figure 3-9b. Figure 3-9b demonstrates that to accurately determine the diffusion length it is necessary to know the generation zone width, g. However, the nearly identical fits in Figure 3-9a indicate that different values of g do not produce easily distinguishable fits to experimental data.

Given that there is not a simple method of measuring the generation zone directly and that one cannot distinguish between different generation zone widths by fitting experimental data, we will mitigate this problem by investigating factors that theoretically have an influence on the generation zone shape. We will start by looking at the expression for the exciton generation rate.

Assuming Langevin recombination, the exciton generation rate (kL) is given by

( ) = ( ) ( ) (3-12)

푘� � 훾�푛 � � � is the Langevin recombination rate and n and p are the concentrations of electrons and holes,

훾respectively.� Charge transport and mobility offsets lead to accumulation of charges at the ETL/HTL interface, which according to Equation 3-12 leads to high exciton generation rates, explaining why the exciton formation zone is typically localized near the ETL/HTL interface. We can see from Equation 3-12 that the spatial dependence of the exciton generation rate (and thus the exciton generation zone shape) is dependent on the charge balance in the device (relative concentrations of electrons and holes). The charge balance in an OLED is dependent on the applied field as well as the thickness and mobility of the charge transport layers, among other factors.

The dependence of charge balance, and thus the shape of the exciton formation zone on the applied field, suggests that as the applied voltage is increased we may observe a broadening or narrowing of the generation zone. Given the short singlet exciton diffusion lengths for CBP measured in Section 4.2, changes in the generation zone width as small as several nanometers should be detectable as relative changes in the extracted exciton diffusion length. The lack of variation in the singlet exciton diffusion length with current density, observed in Figure 3-7, indicates that either: 1) the generation zone width is very narrow relative the diffusion length such that fluctuations in its width do not affect the measured exciton diffusion length or 2) that the changes in charge balance with current density are not large enough to cause a significant

shift in the exciton generation zone width. Note the extracted diffusion length does not show significant current density dependence for any choice of the generation zone width.

To rule out the second possibility we fabricated another set of devices in which the HTL thickness is increased to 200nm in order to induce an extreme shift in the charge balance inside the device. For these devices constant optical out-coupling was not maintained for each sensing layer in the set in favor of maintaining constant ETL and HTL thicknesses. Constant transport layer thicknesses ensure that the charge balance is the same for each device in the set. The optical out-coupling variations will introduce a systematic error into our measurements however these errors are expected to be minimal, as the singlet exciton diffusion length measured in Section 4.2 is much shorter than the characteristic length scale of optical outcoupling variations.

The current-voltage characteristics of OLEDs with a sensing layer 5nm from the ETL/HTL interface and different HTL thicknesses are shown in Figure 3-10. The dramatic change in the current-voltage characteristics of the devices is indicative of substantial changes in charge balance.

Sensing layer emission for this new set of devices was fit to Equation 3-11 using a generation zone width of 1nm. The CBP singlet exciton diffusion length extracted from these fits was 4.8±0.3nm, in agreement with the value measured in the previous section. The agreement of the two values using the same generation zone width in each case indicates that even extreme changes in charge balance do not affect the exciton generation zone width for this CBP/TPBi OLED. It is very unlikely that a broad generation zone would remain unperturbed by such large changes in charge balance, thus we conclude that the exciton generation zone width for the OLEDs studied here is narrow relative to the singlet exciton diffusion length.

We use a generation zone width of 1nm here, although according to Figure 3-9b values of g<1nm result in similar values for the extracted diffusion length. This suggests that exciton generation is limited to the first few monolayers of CBP from the CBP/TPBi interface.

The narrow generation zone can be explained by the mobility offset of the electron and hole transport materials. CBP has a hole mobility of 10-4 cm2V-1s-1- 10-2 cm2V-1s-1, while TPBi has an electron mobility of 10-6 cm2V-1s-1- 10-4 cm2V-1s-1 58–61. The higher hole mobility in our devices

will lead to hole accumulation at the ETL/HTL interface, electrons arriving at the interface will rapidly recombine with this high interfacial hole population limiting the width of the generation zone. The high hole mobility of CBP contributes to a high Langevin recombination rate which will also limit the thickness of the generation zone.

The observation of a narrow exciton generation zone width is surprising in this case as CBP is of ambipolar nature. There are numerous claims in the literature that the use of ambipolar hosts creates broad exciton generation zones. Our results suggest instead that another phenomenon is responsible for the high performance of OLEDs fabricated using of ambipolar hosts. In particular, our results suggest that the high efficiencies achieved using devices with a CBP/TPBi heterojunction36,46 are not due to a broadened exciton generation zone on the host, although charge transport on the dopant may contribute to broadening of the generation zone in these devices.

An alternative explanation for the widespread success of ambipolar host is that these materials can sustain high minority carrier currents which are necessary for direct exciton formation on dopant, as evidenced by highly efficient OLED made using a single layer ambipolar host62.

Figure 3-10: Current voltage characteristics for devices with sensing layers at 5nm from the interface and a thin HTL (dashed line) and 200nm HTL (solid line)

3.5 Conclusions

The impact of various parameters on measurement of singlet exciton diffusion in OLED has been studied. The effects of long range energy transfer from hosts to sensing dopants on exciton diffusion in OLEDs are identified as a major factor affecting the diffusion length measurements. After inclusion of all contributing factors, we measure a singlet exciton diffusion length of 4.3 ± 0.3 nm, corresponding to a diffusivity of (2.6±0.3)x10-4cm2/s. Contrary to expectations for an ambipolar-based CBP host, we demonstrate that a narrow exciton generation zone exists for the CBP/TPBi based OLEDs studied here. We expect that this work will provide a valuable guide in future studies of excitons in OLEDs.

Chapter 4 Exciton Distribution in a Doped Emissive Layer OLED

This chapter was adapted from: Ingram, G. L., Chang, Y.-L. & Lu, Z. H. Probing the Exciton distribution in Organic Light Emitting Diodes Using Long Range Energy Transfer. Can. J. Phys. 92, 845–848 (2014). Copyright 2014 Canadian Science Publishing.42

4.1 Introduction

In most studies of the exciton distribution of OLEDs, exciton diffusion through nominally identical organic films of different thickness is compared 3,15–17,40,41 . For materials with high run-to-run reproducibility, the comparison of nominally identical films is appropriate, however in some cases, either due to sensitivity to deposition conditions or limitations in control over the precise mixture in doped films, run-to-run reproducibility may be low. In these cases, we propose an alternative approach which enables study of an organic film deposited in a single run through analysis of the effect of sensing layers incorporated into the device outside the emissive region. In this way, the region of interest is identical between a set of devices, while thickness variations occur in the neat spacer layers adjacent to the emissive region. This technique enables studies of exciton diffusion in materials with large run-to-run variations in crystallinity, purity, or morphology.

4.2 Experimental Methods

Materials used in the device are: 4,4'-N,N'-dicarbazole-biphenyl (CBP) as the hole transport layer (HTL), 1,3,5-tris(N-phenylbenzimidazole-2-yl)benzene (TPBi) as the electron transport layer (ETL), and bis(2-phenylpyridine)(acetylacetonate)iridium(III) (Ir(ppy)2(acac)) doped CBP

as the emissive layer (EML). C60 fullerene was chosen as an exciton acceptor since its

absorption overlaps strongly with the emission wavelengths of the emitter and thus is expected to

promote Förster type energy transfer. The OLED structure is shown in Figure 4- 1: ITO/MoO3

(1nm)/CBP (35-d nm)/CBP:C60(20wt%) (3nm)/CBP (d nm)/CBP:Ir(ppy)2(acac) (6wt%)

(10nm)/TPBi(65nm)/LiF(1nm)/Al(100nm). Control devices with no C60 doped layer were also fabricated.

Figure 4-1: Schematic energy level diagram of organic layers in OLEDs. The lines at top and bottom of each box represent the LUMO and HOMO energies (in eV) respectively.

It is important that the C60 molecules only affect the exciton distribution in the devices and the

OLED optical and electrical properties are not be significantly altered by introducing the C60 doped layers into various locations in the OLEDs. The current vs. voltage curves and normalized electroluminescence (EL) spectra are shown in Figure 4- 2. The electrical and optical properties show almost no dependence on the position of the C60 doped layer. The invariance of the electrical characteristics is consistent with expectations based on knowledge of

charge trapping. The HOMO level of C60 is slightly deeper than that of CBP so the C60

molecules do not act as efficient hole traps. We note that similar devices using TPBi:C60 layers

exhibited large variations in the current-voltage characteristics with position of the C60 layer.

Figure 4-2: a) Current vs voltage characteristics and b) normalized electroluminescence spectra

at 7.4V of devices with x=1nm, x=5nm and a device with no C60 doped layer

4.3 Theory

The lack of CBP emission from these devices indicates that excitons are formed directly on the

Ir(ppy)2(acac) dopants following injection of electrons from TPBi, or are rapidly transferred

from the CBP to Ir(ppy)2(acac) following exciton formation on the CBP. In addition,

Ir(ppy)2(acac) exhibits fast intersystem crossing from the singlet to triplet state such that close to

100% of the excitons in the emissive layer are Ir(ppy)2(acac) are triplets. The steady state rate

equation for Ir(ppy)2(acac) triplet excitons is:

= 0 2 (4-1) 휕 �� 2 퐷 휕푥 − 휏� − 푘퐹푛 where D is the triplet exciton diffusivity, x is the position in the emissive layer measured from

the ETL/HTL interface towards the anode, nT is the triplet exciton concentration, τT is the triplet

exciton lifetime in the absence of the C60 layer, kF is spatially dependent the rate of Förster energy transfer from the Ir(ppy)2(acac) dopants to the C60 quenching layer. The rate of energy transfer from an excited donor to a collection of acceptors depends on the number and distribution of the acceptors. The energy transfer rate of a donor at position x to a film of acceptors is given by the following equation 43:

( ) = (4-2) ( ) ( ) ( ) �퐴 휋 6 1 1 3 3 푘퐹 �ℎ푖� 푓𝑓� � 휏 6 푅0 � 10−푥+푑 − 10−푥+푑+� � where CA is the molecular concentration of the acceptors, is the thickness of the acceptor film, and R0 is a characteristic length referred to as the Förster radius푙 defined as the distance at which the rate of energy transfer between an isolated donor-acceptor pair is equal to the energy transfer rate to all other sources. The Förster radius is given by 14:

( ) = 2 (4-3) 6 9 ln 10 푁퐴 휅 Φ퐷 2 4 푅0 128휋 �푅 퐽 where NA is Avogadro’s number, κ is an orientation factor, ΦD is the donor fluorescent quantum yield, nR is the index of refraction of the medium containing the donor and acceptor and J is an the overlap integral of a normalized donor emission spectrum and an acceptor absorption spectrum. Efficient Förster type energy transfer requires an overlap between the absorption of the acceptor and the emission of the donor. A high overlap gives a large value for J in Equation

4-3 and thus leads to a large value of R0, which in turn leads to more efficient long range energy transfer. A high absorption coefficient for the acceptor also contributes to a large value for J.

The triplet exciton energies of CBP and TPBi are significantly higher than that of Ir(ppy)2(acac) confining triplet excitons to the dopants. The exciton confinement and exciton generation at the ETL/HTL interface are expressed through the following boundary conditions:

= (4-4) 휕�� 퐽 −퐷 휕� �푥=0 푒 = 0 (4-5) 휕�� −퐷 휕� �푥=10 The triplet exciton concentration is calculated using Equation 4-1 and the boundary conditions in Equations in Equations 4-4 and 4-5. The triplet exciton distribution will depend on both the triplet exciton diffusion length, = , and the separation between the emissive layer and the exciton quenching layer, d. The퐿 fraction�퐷휏 푇of excitons which are transferred to the quenching layer can be calculated by integrating the triplet exciton distributions over the emissive layer with and without the energy transfer term in Equation 4-1.

Experimentally, the quenching fraction is determined by comparing the device external quantum efficiency with and without using an exciton quenching layer:

( ) ( ) = 1 (4-6) 휂푄 푑 푓 푑 − 휂 where and are the external quantum efficiencies with and without the exciton quenching

layer. 휂 � 휂

4.4 Results and Discussion

To account for FRET from the Ir(ppy)2(acac) to the C60, we will calculate the Förster radius for this donor acceptor pair.

Figure 4-3: Normalized emission spectra of Ir(ppy)2(acac) (dashed green line) and normalized

absorption spectrum of C60 (solid black line)

The normalized emission spectra of emitter (Ir(ppy)2(acac)) triplet excitons as well as the

normalized absorption spectrum of the acceptor (C60) are shown in Figure 4- 3. The emission

spectrum of Ir(ppy)2(acac) was taken from the electroluminescent spectra of an OLED using this

44 molecule as an emitter. The absorption spectrum of C60 is taken from the literature . It is clear

that there is a large overlap between C60’s absorption spectrum and the emission of

Ir(ppy)2(acac) triplets.

The Förster radius for Ir(ppy)2(acac) triplets with a C60 acceptor is calculated to be 3.8 nm using Equation 4-3. Assuming amorphous films with randomly oriented rigid dipoles κ=0.845 2/3 45. 46 The refractive index of the CBP host at the peak emission of the Ir(ppy)2(acac) is 1.8 . �The PL 5 efficiency of Ir(ppy)2(acac) is ~1 . The overlap integral was calculated using the data presented in Figure 4- 3. The molar absorption coefficient in the overlap integral is estimated by using the 44 extinction coefficient reported in the literature and dividing by the molecular density of C60 which is assumed to be 2.4mol/L 7.

The quenching fraction as a function of the proximity of the quenching layer to the emissive region is shown in Figure 4-4 for devices driven at 40 mA/cm2.

The data is fit with numeric solutions to Equation 4-1 and the boundary conditions in equations in Equations 4-4 and 4-5 using the exciton diffusion length as the fitting parameter, and calculated values for the Förster transfer rate. The extracted value for the exciton diffusion length is 2.0 ± 0.6 nm. The uncertainty takes into account variations in the quenching fraction as a function of current density as well as uncertainty in the calculated Förster radius. The extracted value for the diffusion length is consistent with relatively inefficient exciton transfer between spatially separated dopant molecules, resulting in exciton localization close to the ETL/HTL interface.

Figure 4-4: Measured quenching fraction (black dots) and fit (line) at 40mA/cm2

The excellent agreement between theory and experiment in Figure 4- 4 suggests that this method of probing exciton diffusion could be very useful in situations where run to run variations in materials are significant.

4.5 Conclusion

We have demonstrated that C60 doped CBP, which does not alter the device electrical characteristics, can be used as an acceptor layer to investigate the exciton distribution in functional OLEDs. C60’s absorption overlaps strongly with the wavelengths of common host and emitter materials and thus facilitates long range energy transfer, which allows us to probe the excitons in the emissive layer without placing probe layers inside the emissive zone. We have shown that, our results are consistent with experimental data indicating a long range energy transfer from the Ir(ppy)2(acac) triplets in the EML to C60 acceptors. We expect that this work will provide an alternative way to probe the exciton dynamics in OLEDs through the use of acceptor layers and Förster type energy transfer theory.

Chapter 5 Impact of Defects on Exciton Diffusion

This chapter originally published as: Ingram, G. L. & Lu, Z. H. Impact of defects on exciton diffusion in organic light-emitting diodes. Org. Electron. 50, 48–54 (2017). Copyright 2017 American Chemical Society.63 https://doi.org/10.1016/j.orgel.2017.07.017

5.1 Introduction

The diffusion of bound electron-hole pairs, or excitons, plays an important role in organic optoelectronic devices such as organic photovoltaics and organic light-emitting diodes (OLEDs). In the former, the diffusion of excitons to a dissociating interface is a necessary step in the generation of free charges from absorbed light. In the latter exciton transport determines both efficiency and operational stability of devices. In both applications, the diffusion of excitons in an organic thin film depends strongly on the photophysical properties of the individual molecules, which can be controlled to some extent through molecular design 64–66, as well as the properties of the solid film they form. Systematic studies of how the properties of organic films influence exciton diffusion have analyzed the importance of purity67–69 and film crystallinity70–76.

In this work we explore the relationship between the defect concentration and singlet exciton diffusion length in 4’-bis(carbazol-9-yl)biphenyl (CBP). CBP defect formation and its effect on OLED performance have been studied as a result of both operational degradation and post fabrication annealing in OLEDs33,77,78, but the impact of CBP defects on exciton diffusion has not yet been considered.

Here, we study the dependence of singlet exciton diffusion on defect concentration in CBP by monitoring the emission of a series of OLEDs with thin exciton capturing layers. CBP defects are emissive, allowing us to quantify the defect concentration through analysis of OLED emission spectra. The origin of varying defect concentrations is not known. However Wang et al. have demonstrated that CBP defects can be produced through thermal annealing33 and we

expect that excess heat produced during either the CBP deposition or during the high temperature cathode deposition may lead to defect formation during device fabrication. Modelling the changes in exciton diffusion as a function of the defect concentration allows us to determine the “defect-free” singlet exciton diffusion length in CBP, analogous to the impurity free exciton diffusion length reported by Curtin et al. in a photoluminescence based study of exciton diffusion in N,N’-bis(naphthalen-1-yl)-N,N’-bis(phenyl)-benzidine (α-NPD)67.

The method of using exciton capturing layers in an OLED to study exciton diffusion has been demonstrated previously to measure exciton diffusion lengths 3,15–17,40,41. It has the advantage of probing organic materials under the same fabrication and operational conditions as other organic layers used in real devices.

5.2 Experimental Methods

The OLED device architecutre used in this study is shown in Figure 5-1. All OLEDs have the same general structure ITO/MoO3 (1nm)/CBP (53-d nm)/CBP:rubrene (3nm, 2wt%)/CBP (d nm)/TPBi (70nm)/LiF (1nm)/Al (100nm), where CBP is used as the hole transport layer (HTL), and 1,3,5-tris(Nphenylbenzimidazole-2-yl)benzene (TPBi) is used as the electron transport layer (ETL).

Three sets of OLEDs are fabricated, with the distance between the ETL/HTL interface and CBP:Rubrene layer, d, varying across the devices within each set. Each set of OLEDs is fabricated on a single substrate with shared electrodes, injection layers, rubrene doped layer, and electron transport layer, while CBP layers of varying thickness are deposited before and after the CBP:rubrene layer to vary the position of the rubrene doped layer in the device, while keeping the total device thickness constant. Each set of devices is fabricated entirely under vacuum (except for the pre-patterned ITO anode) with varying thickness of the CBP layers enabled using a set of interchangeable shadow masks. Although the three sets of devices investigated are nominally identical, we find evidence of varying defect concentration between the three sets (see Section 4). To ensure that the variation in defect concentration only occurs between the sets of devices and not between the layers of a given set, we fabricate two control devices with the structure: ITO/MoO3 (1nm)/ CBP (45nm)/ CBP (10nm)/ TPBi (70nm)/ LiF (1nm)/ Al (100nm). All layers of the two devices are shared except for the 10nm CBP layer which was deposited

individually for each of the two devices. The emission spectrum of the two devices are identical (see Appendix B), confirming that layer-to-layer variations in defect concentration are not significant within a set of devices fabricated on a single substrate.

Although we cannot verify the mechanisms leading to the defect formation in these nominally identical sets of devices, it may be due to the samples’ inadvertent exposure to heat or light during the fabrication process and while being stored prior to testing. Alternatively, if defects arise due to CBP aggregation into a more thermodynamically favorable state as proposed by Wang et al.33, defect formation may occur over time even in the absence of stressors such as heat or light. We note that the diversity in photoluminescence spectra79–81 reported for CBP suggests that variations in CBP film composition are common.

The rubrene doped CBP is included to capture singlet excitons which diffuse from the exciton formation region at the ETL/HTL interface. The low singlet energy of the rubrene dopants ensures that any excitons diffusion to the layer are quickly captured, while the higher singlet energy of TPBi relative to CBP confines excitons to the CBP layer. Varying the position of this exciton capturing layer allows for the systematic variation of the singlet exciton concentration.

Figure 5-1: Schematic device structure of the OLEDs used in this study.

5.3 Theory

The singlet exciton population in the CBP layer of the OLED structure described above, under constant current operation can be modelled using the steady state diffusion equation.

( ) ( ) 0 = ( ) 2 (5-1) 휕 �푆 푥 �푆 푥 2 퐷 휕푥 − 휏푆 − 푘퐸�푛푆 � 푁 nS is the CBP singlet exciton concentration, x is the position in the device measured from the ETL/HTL interface towards the anode, D is the CBP singlet diffusivity, is the singlet exciton natural lifetime, is the average energy transfer rate of singlets from the휏푆 CBP to the defects, and N is the defect푘퐸 �concentration, which is assumed to be uniform. The terms in Equation 5-1 from left to right correspond to exciton diffusion, exciton decay and energy transfer to defects.

We express Equation 5-1 in a simpler form in Equation 5-2.

( ) ( ) 0 = 2 (5-2) 휕 �푆 푥 �푆 푥 2 2 휕푥 − �푒푒푒 The true diffusion length, = , is defined as the diffusion length in the absence of any defects and is related to the퐿 effective√퐷� diffusion length, , through Equation 5-3. 퐿푒𝑒 = (5-3) � 퐿푒𝑒 �1+휏푆푘퐸�푁 If the defect concentration increases, the effective diffusion length will decrease due to increased competition between singlet exciton diffusion and energy transfer to defects.

Equation 5-2 can be solved using the boundary conditions in Equations 5-4 and 5-5. Equation 5- 4 represents singlet exciton formation localized at the CBP/TPBi interface by equating the exciton flux at this interface to the rate of charge injection:

= (5-4) 휕�푆 퐽 −퐷 휕� �0 4푒 J is the current density and e is the elementary charge. The factor of ¼ on the right side of Equation 5-4 is equal to the fraction of excitons which are produced in the singlet state under

electrical excitation. A previous study shows a narrow exciton generation zone for this device structure47 which is why we localize exciton formation to the ETL/HTL interface here.

Equation 5-5 represents complete energy transfer of singlet excitons to the rubrene at the boundary of the exciton capturing layer positioned at a distance d from the ETL/HTL interface, as discussed in Appendix A:

( ) = 0 (5-5)

푛푆 푑 d is the position of the exciton capturing layer as measured from the ETL/HTL interface towards the anode. Direct exciton formation on the defects is also possible through the recombination of trapped holes with minority electrons. Given the modest defect concentrations present in our devices, significant direct exciton formation on defects is unlikely. A broad exciton generation zone or strong long-range energy transfer to the capturing dopants can be taken into account following the example set by previous studies16,17,47. The solution to Equation 5-2 using boundary conditions given by Equations 5-4 and 5-5 is given by Equation5- 6.

( ) = tanh + 1 cosh exp (5-6) 퐽�푒𝑓 푑 푥 푥 푛푆 � 4푒� �� 푒푒푒� � ��푒푒푒� − ��푒푒푒�� The CBP emission intensity, I, is proportional to the integrated singlet exciton density from the ETL/HTL interface to the boundary of the exciton capturing layer.

( ) = 1 sech (5-7)

푑 �푆 푥 푑 퐼 ∝ ∫0 휏푆 푑� 퐼푠𝑠 � − ��푒푒푒�� In Equation 5-7, as d becomes large relative to Leff, the emission intensity asymptotically approaches a saturation intensity, .

퐼푠𝑠 The singlet population trapped by defects is governed by a rate equation given in Equation (8).

0 = (5-8) �퐷 2 푘퐸�푛� − 휏퐷 − 훼푑 푛퐷 is the lifetime of singlet excitons trapped by defects, and is an unknown constant and nd is the휏퐷 concentration of excitons trapped by defects. Equations 훼5-1 and 5-8 are coupled through the

energy transfer term (first term in Equation 5-8). The second term in Equation 5-8 describes the defect exciton decay and the third is an empirical exciton quenching term dependent on the exciton capturing layer position. The empirical term is needed to explain the defect emission dependence on d observed in the subsequent sections. Although the physical origin is unknown, it is likely due to variations in the triplet or charge concentrations as a function of the capturing layer position. The d2 dependence was chosen as it shows a close match to experimental data.

CBP defects are luminescent and the relative intensity of the defect emission carries information about the defect concentration. The defect emission intensity is proporitional to the integrated defect exciton concnetration.

( ) = (5-9) 푑 �퐷 푥 푘퐸�푁휏푆 2 퐼퐷 ∝ ∫0 휏퐷 푑� 1+휏퐷훼푑 퐼 Rearranging Equation 9 allows us to express the ratio of the defect and CBP emission using Equation 5-10:

= (5-10) 퐼퐷 푁 2 퐼 훽 1+휏퐷훼푑 where we have introduced β as a proportionality constant.

The equations outlined in this setion will be used to analyze the CBP and defect emission in three sets of OLEDs with the structure described in the previous section.

5.4 Results and Discussion

The emission of an OLED from each of the three sets fabricated, with an exciton capturing layer 10nm from the ETL/HTL interface are shown in Figure 3-2. The emission bands centered at 390nm, 500nm and 560nm correspond to the CBP, defect and exciton capturing layer (CBP:rubrene) emission respectively. We attribute the 500nm emission band to defect emission based in its similarity in position and width to previously values for CBP defect emission 33. The OLEDs have each been run at 25mA/cm2 for nine seconds. The different intensities of the defect emission bands suggests different defect concentrations, which in turn will lead to different effective singlet exciton diffusion lengths according to Equation 5-3.

Set A

] 150 Set B s t

n Set C u o [

l a

n 100 g i S

r o t c

e 50 t e D

0 350 400 450 500 550 600 650 700

Wavelength [nm]

Figure 5-2: OLED emission spectra for devices from three different sets of OLEDs run for nine seconds at 25mA/cm2 with an exciton capturing layer at 10nm. The data is averaged over two nominally identical devices and smoothed for clarity.

The defect emission is convoluted by the tail of the CBP emission band. Therefore in order to independently analyze the CBP and defect emission, we deconvolute the OLED emission spectra into the CBP emission and the defect and rubrene emission. Defect emission intensity is then measured at wavelengths less than 505nm to avoid overlap with the rubrene.

Rapid luminance degradation of the OLEDs also needs to be taken into account. We are interested in the effects of defects which exist in the OLEDs prior to any operational degradation and so we extrapolate the emission intensities back to their initial values, with details provided in Appendix B. Subsequent discussion of emission intensity for CBP and defects will refer to these extrapolated initial values.

5.4.1 Extraction of Relative Defect Concentration

According to Equation 5-10, the ratio of the defect to the host emission can be used in determining the defect concentration. This ratio is plotted in Figure 5-4 as a function of the exciton capturing layer position. Clearly, the ratio of defect to host emission varies significantly across the three sets of OLEDs. The defect to host emission intensity ratio varies with the position of the exciton capturing layer, as anticipated by Equation 5-10.

According to Equation 5-10, as the exciton capturing layer is moved towards d=0, the ratio of defect to CBP emission approaches a value proportional to the defect concentration. The data from each set of OLEDs is fit with Equation 5-10 using and as free parameters and

plotted as solid lines for each set of OLEDs in Figure 5-3.훽� This provides휏푁훼 a close match to the data across the three sets of OLEDs and two current densities studied.

Figure 5-3: Ratio of defect to host emission intensities as a function of the exciton capturing layer position for three sets of OLEDs at constant current density of a) 25mA/cm2 and b) 100mA/cm2

The values of N extracted from the fits are supplied in Table 5-1. It is noted that βN is slightly 2 2 lower from 25mA/cm훽 to 100mA/cm , suggesting that β may have a current density dependence due to changes in exciton quenching on triplet excitons or charges, as a function of current density.

Table 5-1: Extracted relative defect concentrations for three sets of OLEDs

βN Current Density 2 [mA/cm ] Set A Set B Set C

25 2.12±0.09 1.08±0.08 0.18±0.02

100 1.72±0.03 0.96±0.05 0.154±0.007

We can now use βN as a relative measure of the defect concentration.

5.4.2 Extraction of Effective Diffusion Length

We will now analyze the CBP emission within the theoretical framework discussed in Section 3 to determine the effective diffusion length for CBP singlet excitons in each of our three sets of OLEDs and at two different current densities.

Figure 5-4 shows the CBP emission intensity for the three sets of OLEDs as a function of the position of the exciton capturing layer. For capturing layers close to the ETL/HTL interface, most of the CBP excitons are captured and the CPB emission intensity is low. As the capturing layer is moved away from the generation zone, the CBP emission intensity increases and then assymptotically approaches a saturation intensity as fewer excitons can diffuse far enough to be affected by the capturing layer.

Although each set of OLEDs shares these common features, they differ in how quickly the emission intensities approach saturation as well as the magnitude of this saturation value. If the

emission saturates quickly it tends to do so at a smaller saturation intensity (for a given current density). The differences in the sets of OLEDs are due to differing effective diffusion length. If the effective diffusion length is short, excitons do not move as far, on average, from their generation zone and so the effectiveness of the exciton capturing layer diminishes rapidly as it is moved further away. In addition, the short effective diffusion length means that energy transfer to defects is competitive with radiative decay, leading to reduced emission and lower saturation emisison intensity.

For both current densitities studied, the data points in each set of OLED in Figure 5-4 are fit with Equation 5-7 using Leff and Isat as fitting paramters. The fits are plotted in Figure 5-4. The extracted values for the fit parameters are provided in Table 5-2. From left to right in Figure 5-4, both the effective diffusion length and the saturation level increase monotonically for a given current density. It is worth noting that Isat, which corresponds to the CBP emission intensity in the absence of the exciton capturing layer, increases by a factor of 1.4 from set A to set C (highest to lowest defect concentrations as measured in the previous section).

Figure 5-4: CBP emission intensities for three sets of OLEDs operated at 25mA/cm2 and 100mA/cm2.

Table 5-2: Extracted effective diffusion lengths and saturation intensities for three sets of OLEDs

Current Density Effective Diffusion Length (Leff) Saturation Intensity (Isat)

[mA/cm2] [nm] [102 counts]

Set A Set B Set C Set A Set B Set C

25 3.3 ± 0.3 3.7 ± 0.3 4.4 ± 0.4 1.06 ± 0.06 1.28 ± 0.06 1.48 ± 0.08

100 3.1 ± 0.4 3.3 ± 0.4 3.8 ± 0.2 4.2 ± 0.3 4.6 ± 0.3 4.9 ± 0.1

We also note that the diffusion length within each set decreases as the current density is increased from 25mA/cm2 to 100mA/cm2 due to exciton quenching at high current densities. The difference in the diffusion lengths at the two current densities studied here is likely due to exciton quenching on charges or triplets, which increases in magnitude with the current density. With only two current densities examined here, we are not able to make any judgements about the precise nature of the quenching mechanism, or quantify the magnitude of the effect outside of noting its influence on the effective diffusion length. Future studies wishing to investigate the effect of current density on exciton diffusion length would be able to do so by including an additional term in Equation 1 corresponding to current density dependent exciton quenching, and systematically investigate this term as a function of current density.

We have now determined that there are variations in the effective exciton diffusion length between our three sets of OLEDs. We noted in Section 4.1 that the defect concentration varied between the three sets of OLEDs. In Section 4.3 we will demonstrate the link between the relative defect concentration and effective diffusion length.

5.4.3 Relationship between Leff and defect concentration

The effective diffusion length was determined by analyzing the CBP emission, while was

determined through analysis of the defect emission. Both provide information about the훽� defects;

Leff tells us how exciton diffusion is affected by the defect population, and gives us a means of comparing the relative magnitude of the defect concentration between sets훽� of OLEDs. We will now link these two quantities.

The effective diffusion length is plotted against in Figure 5-5. As expected the effective diffusion length decreases at higher defect concentration훽� s as energy transfer to the defects becomes more competitive with diffusion. The data in Figure 5-5 is fit using Equation 5- 3 at each current density, allowing us to extract true (defect free) singlet exciton diffusion lengths of 4.5 nm ± 0.3nm and 3.9nm ± 0.3nm at 25mA/cm2 and 100mA/cm2 respectively. Equation 5- 3 shows good agreement with the data, suggesting that our theoretical framework is suitable to describe the exciton dynamics in this system. The extracted values for the defect free diffusion lengths are both in agreement with the value of 4.3 ± 0.3 nm previously determined using a similar approach47, and much smaller than the value of 16.8nm reported using spectrally resolved photoluminescnce quenching (SRPLQ) 51. The difference between the numbers reported here and using the SRPLQ method may be due to systematic errors in one or both measurement techniques, or due to differences in film morphology between the two studies.

Figure 5-5: Extracted effective exciton diffusion length as a function of defect concentration.

5.4.4 Calculation of the defect concentration

We will now determine the approximate defect concentrations in the three sets of OLEDs. First

we need to define the form of kET, the rate constant for energy transfer from the CBP to defects. We will assume that this energy transfer is diffusion limited and has the form 7,82,83:

= 4 (5-11)

푘퐸� 휋�푟퐼 rI is a characteristic interaction distance for the energy transfer which we will assume to be a typical value of 1nm 7. If Förster energy transfer from the CBP to defects is significant, the

above expression may underestimate kET leading to an overestimate of the defect concentration. Isolating and optically characterizing the defects would allow for calculation of Förster energy transfer rate; however this task is beyond the scope of this work. Substituting Equation 5- 11 into Equation 5- 3, and using the definition of the diffusion length ( = ) we can rearrange

for the defect concentration which gives us an expression which can퐿 be used√퐷� to evaluate the defect concentration given our extracted values of Leff and L.

= (5-12) 1 1 1 2 2 푁 4휋푟퐼 ��푒푒푒 − � � 2 Using Equation 5- 12 and our extracted values for Leff and L at 25mA/cm we determine that the defect concentration in the three sets of OLEDs are approximately (3 ± 1) x 1018cm-3, (2 ± 1) x 1018cm-3 and (0.3 ± 0.7) x 1018cm-3.

Defects will exist to some extent in nearly all OLEDs, either created during device fabrication, or due to operational or intrinsic degradation. Calculating the defect concentration is desirable, but the extent to which defects influence the singlet exicton population is ultimately affected by the competition between natural exciton decay in host materials and energy transfer to defects at

rates of 1/τ and kETN respectively. In the absence of exciton capturing layers, the emission intensity in the presence of defects is decreased by a factor of relative to the defect free 1 1+휏푆푘2퐸�푁 value. If kET has the form of Equation 5- 11 and we substitute D=L /τ, this becomes , 1 2 from which we can see that, all else being equal, materials with longer diffusion lengths1+4휋 will� 푟퐼푁 be more susceptible to energy transfer to defects. This is supported by the intuitive argument that in order to diffuse further, an exciton must migrate through a larger number of molecules, increasing its chance of encountering a defect.

5.4.5 Relationship Between Defect Concentration and OLED Driving Voltage

Our analysis up until now has demonstrated the power of exciton capturing layers on investigating the link between exciton migration and the defect population in OLEDs. however fabricating sets of OLEDs incorporating capturing layers may not be readily applicable to all types of devices. Studies of OLED degradation often note an increase in driving voltage associated with defect formation due to operational degradation in cases where defects act as charge traps. As driving voltage can be easily measured for any OLED, we will now discuss the connection between the driving voltage and defect concentration in our devices.

The driving voltage extrapolated to its initial value, V, varies as a function of both the defect concentration and the position of the exciton capturing layer. The driving voltage at current densities of 25 mA/cm2 and 100 mA/cm2 for each device is ploted against the capturing layer position in Figure 5-6a . The voltages within each set, are fitted with a linear expression

= + for each current density. The y-intercepts of the fits, Vo, are plotted in Figure 5-6b

푉against푉0 the푚 values� of extracted from the fits to the ratio of defect and CBP emission in Section 4.1. V0 increases훽� monotonically with the defect concentration for a given current density, suggesting that driving voltage may be a useful indicator in assessing defect concentrations in situations where defects are non-radiative or the use of exciton capturing layers is impractical. The linear fits are plotted in Figure 5-6b suggest a simple linear relationship between the defect concentration and driving voltage.

Figure 5-6: a) Driving voltages at a constant current density of 25mA/cm2 for devices with different defect concentrations and spacer thicknesses as well as linear fits to the data. b) The y- intercept of the fits in a) vs. defect concentration as well as linear fits to the data.

5.5 Conclusions

We have demonstrated that the presence of defects in vacuum deposited CBP leads to variations in the effective singlet exciton diffusion length. We have presented a set of simple analytical equations to quantify the concentration of these defects and their impact on exciton diffusion by

analyzing OLED emission spectra. For three sets of OLEDs, we determined defect concentrations of approximately (3 ± 1) x 1018cm-3, (2 ± 1) x 1018cm-3 and (0.3 ± 0.7) x 1018cm-3. We observed a decrease in the effective diffusion length as the defect concentration increases, as well as with increasing current density. Extrapolating the experimental data, we determine “defect free” diffusion lengths of 4.5 ± 0.3nm and 3.9 ± 0.3nm at current densities of 25 mA/cm2 and 100 mA/cm2 respectively. The decrease in diffusion length with current density is attributed to increased exciton quenching at elevated current densities. We also analyze the differences in driving voltage in these devices and show that the driving voltage scales approximately linearly with the defect concentration. This work provides an analytical methodology to gain insights into the process of exciton transport and provides a template for analysis of defects and exciton diffusion in OLEDs.

Chapter 6 Electroluminescence of Degrading CBP Films

6.1 Introduction

Despite its success in commercial active matrix organic light-emitting diode (AMOLED) display applications, OLED’s poor stability at high brightness is currently holding back its broad application as a general solid-state lighting source. Establishing the fundamental device physics of exciton dynamics is a key to the design and fabrication of future generations of robust OLEDs for applications that require stability at high luminance. In many material systems the energy of tightly bound Frenkel excitons fuels molecular reactions leading to device degradation. These reactions may be driven directly by excitons 25–30 or through exciton-charge interactions32,33,31 or exciton-exciton interactions34. Once degradation products, or defects, are formed they will capture excitons, causing OLED luminance loss. As excitons are involved in the physical processes governing luminance loss in degrading OLEDs, the fundamental device physics of exciton dynamics is critical;

Exciton driven degradation is particularly detrimental for wide band-gap materials where singlet exciton energies are comparable to bond dissociation energies. This makes bond cleavage driven by singlet excitons the dominant degradation pathway for many of these materials.

As 4,4'-Bis(carbazol-9-yl)biphenyl (CBP) is an archetype wide band-gap hole-transport and host material, it will be used as a testbed for establishing exciton device physics. Degradation will be studied in neat layers of CBP in a simple bilayer configuration. While these neat CBP layers have much higher exciton concentration and accompanying shorter lifetime than the doped layers used in high efficiency devices, using pure CBP layers allows us to quantify degradation without needing to deconvolute the effects of dopant and host degradation. This will allow for a more straightforward assessment of the degradation mechanism of this wide band-gap material.

CBP degradation is known to create new emission bands from excitons captured on degradation induced defects, a phenomenon which has also been observed for a variety of other materials33,84.

Defect luminescence provides a useful measurable parameter for quantifying the dynamics of excitons captured on defects.

In this paper, we show that a set of analytical equations can be effectively applied to model how and where defects form in working OLEDs under exciton driven degradation. We show that this set of equations can quantify the influence of exciton capture by defects on the CBP emission intensity and what factors influence the subsequent radiative emission of these captured excitons.

6.2 Experimental Methods

Molybdenum Oxide (MoO3) and lithium fluoride (LiF) are used as the hole and electron injection layers respectively. CBP and 2,2’,2”-(1,3,5-Benzinetriyl)-tris(1-phenyl-1-H- benzimidazole) (TPBi) are used as the hole transport layer (HTL) and electron transport layer (ETL) respectively, CBP:Rubrene (2wt. %) is used as an exciton capturing layer where indicated.

To ensure environmental degradation is not a factor in our experiments, a set of nominally identical devices were fabricated and half were encapsulated in a nitrogen environment prior to any atmospheric exposure. The degradation profiles of encapsulated and unencapsulated devices shown in Appendix D are close to identical and thus we conclude that environmental degradation is not a factor over the short time scales in which we are interested.

6.3 Theory

Previous reports indicate that CBP degradation involves some combination of singlet excitons, charges and ETL materials. Based on the strong evidence put forward by Kondakov et al. 25,27, we will assume that the rate limiting step in the degradation process is driven solely by singlet excitons. This assumption will be tested through investigation of the degradation current density dependence.

Yu et al. studied CBP degradation at CBP/ETL heterojunctions using several different ETL materials including TPBi85. The results of these experiments indicate that the luminescent

defects are formed through reactions between CBP and the ETL materials. These reactions must require adjacent CBP and TPBi molecules and thus we expect that the degradation reactions leading to formation of luminescent defects will only occur directly at the ETL/HTL interface, driven by the interface singlet exciton population.

Using these assumptions, we can express the formation rate of defects with concentration, N, as:

( ) = ( = 0, ) (6-1) 푑� � 푑� 푘퐹푛 � 푡 kF is the rate constant for defect formation, n is the singlet exciton concentration and x is the position in the device measured from the CBP/TPBi interface into the CBP.

At the CBP/TPBi interface, excitons are generated due to the accumulation of positive and negative charges at the organic heterojunction. TPBi has a singlet exciton energy ~0.25eV greater than CBP (based on low temperature solid state emission spectra86,87), thus we expect good exciton confinement on CBP. Defects are also produced at this interface and will capture a portion of the interface exciton population. These two processes produce a net singlet exciton flux ( ) at this interface given by: 휕� −퐷 휕� ( , ) = ( = 0, ) ( ) (6-2) 휕� 푥 � 퐽 퐸� − 퐷 휕� �푥=0 4푒 − 푛 � 푡 푘 푁 푡 D is the CBP singlet exciton diffusivity, J is the current density, e is the elementary charge, and

kET is the energy transfer rate constant from CBP singlet excitons to defects. We assume that the energy transfer occurs over sufficiently short rate that only the interface CBP singlet population is captured efficiently. The first and second terms on the right side of Equation 6-2 correspond to exciton formation on CBP and capture of on defects, respectively. The factor of 1/4 in the first term on the right side of Equation 6-3 accounts for the fraction of total excitons which for in the singlet state. We assume that charge trapping on defects and exciton formation through recombination of these trapped charges will be minimal. This assumption is supported by the lack of field dependence in the shape of the luminance degradation curves observed in the next section (beyond the expected scaling with current density). If charge trapping on defects was

significant, the field dependence of the charge trapping process would contribute to the field dependence of the degradation rate

CBP singlet excitons with lifetime, τ, diffuse through the CBP layer as described by the steady- state diffusion equation:

( , ) ( , ) 0 = 2 (6-3) 휕 � 푥 � � 푥 � 2 퐷 휕푥 − 휏 The above equation is in steady state over the time scale on which the singlet exciton population rearranges itself (on the order of ns), although the slow increase in the defect population (on the order of s) means that the expression is not in a true steady state condition.

We also require that the exciton population decay to zero far from the CBP/TPBi interface:

( = , ) = 0 (6-4)

푛 � ∞ 푡 Equations 6-1 to 6-4 can be solved for the time and position dependent singlet exciton population (a detailed solution is available in the Appendix C). The time-dependent CBP emission intensity is found by integrating the singlet population throughout the CBP layer. This leads to an expression for the time dependence of the observed CBP emission intensity:

= (6-5) 퐼 1 퐼0 푡 �1+3푡50

I and I0 are the instantaneous and initial CBP emission intensity, respectively and t50 is the time at which the CBP intensity reaches half of its initial value. The half-life (t50) is given by:

= 2 (6-6) 3푒� 2 푡50 2퐽푘퐸�휏 푘퐹 Solving Equations 6-1 to 6-4 also provides an expression for the time dependence of the defect concentration:

( ) = 1 + 1 (6-7) � 3� 푁 푡 푘퐸�휏 �� 50 − �

As the defect concentration increases, a growing portion of the CBP singlet excitons will be captured by the defects. This is accounted for by the second term on the right side of Equation 6- 2. These captured singlet excitons can be described by a steady state rate equation: ( ) 0 = + ( = 0, ) ( ) ( ) ( ) (6-8) �푁 � − 휏푁 푘퐸�푛 � 푡 푁 푡 − 푘�푛푁 푡 푁 푡 where is the concentration of singlet excitons captured by defects, is the defect singlet

exciton푛 lifetime푁 and kQ is a concentration quenching rate constant. Concentration휏푁 quenching is a commonly observed phenomena in doped OLEDs in which the luminescence efficiency of a molecule decreases with concentration88,89. In our case, the luminescent defects act as dopants in the CBP host and as their concentration becomes large, defect-defect interactions reduce the defects photoluminescence efficiency.

Changes in defect emission as a function of time are proportional to the changes in the concentration of excitons captured by defects:

1 − = 푡 2 (6-9) 퐴�1−�1+3푡50� � 퐼푁 ∝ 푛푁 푡 1+퐵��1+3푡50−1�

where IN is the defect emission intensity and and will be used as fitting parameters.

퐴 퐵 6.4 Results and Discussion

OLEDs are fabricated based on a simple organic bilayer, with the structure ITO/ MoO3(1nm)/ CBP (56nm)/ TPBi(70nm)/ LiF(1nm)/ Al(100nm). The OLEDs are aged at constant current densities between 12.5mA/cm2 and 50mA/cm2 while emission spectra are periodically recorded. Sample emission spectra of an OLED aged for 9s, 90s, and 900s at 25mA/cm2 are show in Figure 6-1. The emission peaks centered at ~410nm and ~490nm correspond to CBP and defect emission respectively. As the OLEDs are aged, the CBP emission peak decreases, while the defect peak initially increases and then decreases slightly with aging time. These effects can be seen more clearly by resolving the emission spectra into two peaks and plotting each as a function of time, as shown in Figure 6-2. The CBP emission intensity is determined by fitting the

OLED emission spectra with a pure CBP emission spectrum. The defect peak emission intensity is determined by taking the maximum of the residual of the CBP fit.

We note that the lifetime of these OLEDs is much shorted than state-of-the-art devices due to the use of neat CBP layers. In long lived OLEDs energy transfer from host to dopant is used to reduce the exciton concentration on unstable wide band-gap materials. Our focus is on understanding the device physics rather than on extending the lifetime of these devices.

Figure 6-1: OLED emission at 9s, 90s and 900s for a device aged at 25mA/cm2, smoothed for clarity.

If the data in Figure 6-2 is consistent with the theory presented in the previous section, we expect that: 1) the time dependence of the CBP luminance degradation will follow Equation 6-5; 2) the

half-life (t50) extracted from fits of the time dependent CBP emission using Equation 6-5 will be inversely proportional to the current density as predicted by Equation 6-6, 3) the time dependence of the defect emission will follow Equation 6-9; 4) the half-life (t50) extracted from

fits of the time dependent defect emission using Equation 6-9 will agree with those extracted from the CBP degradation using Equation 6-5.

The time dependent CBP and defect emission are presented in Figures 2a and 2b and fit at each current density with Equations 6-5 and 6-9, respectively. The fit in Figure 6-2a uses L0 and t50 as free parameters, while the fit in Figure 6-2b uses t50, A and B as free parameters.

Figure 6-2: a) CBP emission as a function of time and fits to Equation 6-3 and b) defect emission and fits to Equation 6-7. Devices are aged at constant current densities of 12.5mA/cm2 to 50mA/cm2.

The values of t50 extracted from fits to the CBP emission is plotted as a function of current

density in Figure 6-3. The t50 values are fit with a power law and the fit is also plotted in Figure 6-3. The exponent extracted from the fit is -1.00 ± 0.01, consistent with the value of -1 predicted by Equation 6-4, while the extracted coefficient is (2.61 ± 0.09) x 103 (mA/cm2)(1/n). The reciprocal relationship between the current density and luminance half-life is commonly observed in OLEDs and is sometimes referred to as “coulombic degradation” 90. We note that the

inverse relationship between t50 and J is not consistent with the exciton-polaron interaction as a degradation mechanism as proposed by Wang et al.33 as such an interaction would lead to a

higher order dependence of t50 on the current density. It is possible that polarons play a role in a reaction step that is not rate limiting.

So far, we have concluded that the time and current density dependence of our OLED degradation matches closely to predictions based on singlet exciton driven degradation, we will now focus on the defect emission. The time dependence of the defect emission is very well described by Equation 6-9, as demonstrated by the close fits in Figure 6-2b. The quality of these fits may either reflect the validity of the underlying model, or the flexibility of Equation 6-9. If Equation 6-9 is to be useful, it should provide physically relevant results. A good test to

determine if the values for t50 extracted from fits to the defect emission agree with the more

physically meaningful values determined from analyzing the CBP emission. To be clear, the t50 value extracted from the fit to the defect emission using Equation 6-9 corresponds physically to the CBP luminance half-life, but is extracted from the defect emission, independently of any input about the CBP degradation. We also note that for the data Figure 6-2b, there is no obvious

way to determine at what time the device has reached t=t50 by inspection. Noting these constraints, we plot the values for t50 extracted from the fits to the defect emission in Figure 6-3 on the same axes as those extracted from fitting the CBP emission data. A power law fit to this second set of data yields a coefficient and exponent of (3 ± 2) x 103 (mA/cm2)(1/n) and -1.0 ± 0.2.

These values agree with those extracted for the t50 values from the CBP fits.

The ability to extract the half-life from the defect emission time dependence independently of any knowledge of the host emission may prove useful in several cases. For example: 1) analysis of defect emission could provide information about degradation kinetics driven by non-emissive (“dark”) excitons as long as the defects produced are emissive; 2) comparing defect emission to luminance degradation could be used to determine if unaccounted for degradation mechanisms are present in the device; 3) defect emission provides a window into defect-defect interaction through the last term in Equation 6-8; 4) using the combination of defect emission and luminance degradation is likely to provide more accurate extrapolation of luminance forward in time than using luminance alone.

The results outlined above suggest that the CBP luminance degradation and corresponding defect emission exhibited by our devices is consistent with the mechanism of exciton driven degradation.

Using exciton capturing layers we will now demonstrate the ability to probe the defect spatial distribution.

Exciton capturing layers can be used to systematically vary the singlet exciton concentration in a set of OLEDs. This is often done with the goal of measuring exciton diffusion lengths 3,15–17,40,41. Here we will use the variations in the singlet exciton distribution to test the hypothesis that defects in our OLEDs form at the ETL/HTL interface.

A new set of OLEDs was fabricated with the structure ITO/ MoO3 (1nm)/ CBP (56-d nm)/ CBP:Rubrene (2wt. %) (3nm)/ CBP (d nm)/ TPBi (70nm)/ LiF (1nm)/ Al (100nm). The purpose of the CBP:Rubrene layer is to capture singlet excitons which diffuse to the rubrene doped layer. The distance between the exciton capturing layer and the CBP/TPBi interface, d, is varied from 2.5nm to 12.5nm. These OLEDs have three emission bands corresponding to the CBP, defects, and rubrene, as shown in Figure 6-4 for a device with d=10nm.

Figure 6-4: Emission of an OLEDs with exciton capturing layer at d=10nm from the CBP/TBPi interface at 9s, 90s and 900s. The device is aged at a constant current density of 25mA/cm2. Data is smoothed for clarity.

With this new device architecture, the boundary condition given by Equation 6-4 must be revised. The new boundary condition requires that the exciton population decays to zero at the exciton capturing layer:

( = , ) = 0 (6-10)

푛 � 푑 푡 Complete quenching of the CBP singlet exciton population at the capturing layer is supported by the absence of the CBP electroluminescence when 3nm of CBP:Rubrene (2wt. %) is located at the CBP/TPBi interfaceas well as the arguments presented in Appendix A. Exciton formation on the rubrene molecules through recombination of trapped charges is assumed to occur at equal rates for all values of d.

The CBP luminance degradation is still described by Equation 6-5, however the luminance half- life now has a dependence on the position of the exciton capturing layer (d), due to the variations in CBP singlet exciton concentration (a detailed derivation is provided in the Appendix C):

= 2 (6-11) 6푒� 50 2 2푑 푡 퐽푘퐸�휏 푘퐹 tanh 퐿 The initial CBP luminance also has a dependence on the position of the capturing layer:

1 sech (6-12) 푑 퐿0 ∝ � − �� The CBP peak emission intensity is determined by fitting the OLED emission spectrum with a pure CBP emission spectrum as in the previous section. The time dependence of the CBP emission is plotted in Figure 6-5, along with fits to Equation 6-5 using L0 and t50 as free parameters. For instances where two devices were available with the same d, only one is shown in Figure 6-5 and the extracted fit parameters are included in the subsequent analysis.

Figure 6-5: Time dependent CBP emission for devices with exciton capturing layers at d=2.5nm, 5nm, 7.5nm, 10nm and 12.5nm from the CBP/TPBi interface, and fits to Equation 6-5.

The extracted values for CBP initial luminance, I0, are plotted in Figure 6-6, along with a fit to Equation 6-12 using the singlet exciton diffusion length, L, and a proportionality constant as free parameters. A weighted fit was used to account for the variations in uncertainty in I0 for different devices. The data matches closely to the fit with an extracted singlet exciton diffusion length of L = (4.5 ± 0.3) nm. This value agrees with previously a reported value of (4.3 ± 0.3) nm47, but is significantly smaller than the value of 16.8nm determined by a photoluminescence quenching study51.

Figure 6-6: CBP luminance half-life, t50, with a fit to Equation 6-11 as well as CBP initial intensity, I0, with a fit to Equation 6-12. Both sets of data are extracted from the fits in Figure 6- 5.

The t50 values extracted from the data in Figure 6-5 are plotted in Figure 6-6. Equation 6-11 predicts tanh ; we test this prediction by fitting the data using L, and a proportionality −2 푑 constant 푡as50 free∝ parameters,� and comparing the extracted value of L to the one determined from

the initial intensities. A weighted fit was used to account for the variations in uncertainty in t50 for different devices. The fit is plotted in Figure 6-6 and matches closely with the

experimentally determined values of t50. A value of L= (5.1 ± 0.5) nm was extracted from the fit, which is within error of the value determined above through analysis of the initial CBP intensities.

The close match to the predicted tanh behavior, and consistency with the diffusion −2 푑 length extracted using initial CBP푡 50intensities∝ strongly� supports our assumption that the

degradation process is localized to the ETL/HTL interface due to the involvement of TPBi in the rate limiting degradation step.

6.5 Conclusion

In summary, we have derived a set of analytical equations to quantify exciton dynamics of luminescent defects in organic light-emitting diodes. To test the usefulness of this set of equations, we studied the degradation of neat CBP films in simple bilayer OLEDs. The time dependence of the CBP emission intensity matched well to our predictions based on singlet exciton driven degradation. We also observed a linear dependence of the degradation rate on current density, consistent with defect formation driven by singlet excitons. These conclusions match with those of Kondakov et al. based on comparison of degradation products for devices aged by UV exposure and electrical stressing25,27.

The emission of the luminescent defects was also analyzed. We found that the time dependence of the defect emission was described very well by our proposed equation based on singlet exciton driven degradation. We found that the CBP luminance loss can be completely attributed to energy transfer to the luminescent defects, and the CBP luminance half-life can be accurately predicted using only the defect emission. The drop in defect emission in heavily degraded devices was accounted for using a concentration quenching term, suggesting that defect-defect interactions significantly affect the observed defect emission intensity.

The singlet exciton distribution was systematically varied using a set of devices incorporating exciton capturing layers. The degradation of these devices indicated that degradation is localized to the ETL/HTL interface, supporting observations made by Yu et al85.

Chapter 7 Excitonic Degradation of OLEDs

7.1 Introduction

In the previous chapter, the degradation of CBP was successfully described using a rate equation model based on singlet exciton driven degradation. In this chapter, we attempt to extend our results to a broader range of materials, including bilayer OLEDs using different hole transport materials and state of the art OLEDs using doped emissive layers. It should be emphasized that there are many possible OLED degradation mechanisms and we do not expect that every OLED will degrade due to the exciton driven degradation mechanism described here. The theory section outlines a simple model for how the luminance is expected to evolve in time for OLEDs experiencing exciton driven degradation, and the results section demonstrates several OLEDs which follow this time dependence.

7.2 Theory

The time dependence of the defect concentration will determine the aging behavior of the OLED. If the rate limiting step in the formation of degradation products is assumed to be limited by the number of available excitons, then the defect formation rate can be expressed as

= (7-1) 푑� 푑� 푘퐹푛 where N is the degradation product concentration, t is time, kF is the reaction rate constant, and n is the exciton concentration. In general, the reaction rate is thermally activated, i.e., =

퐹 exp , where A and EA are the coefficient and activation energy. The exciton푘 퐸퐴 퐴concentration�− 푘퐵푇� n may be either host or dopant excitons in either singlet or triplet state. The host exciton population is typically low due to rapid energy transfer to dopants in the case of a mixed dopant-host system. Due to their possessing higher exciton energies, however, hosts may be very unstable as compared to dopants. The relative degradation by a small population of unstable host excitons versus a large population of stable emitter excitons will depend on the host-dopant combinations used. In most fluorescent materials, singlet excitons are significantly more

energetic than triplets and thus we expect that singlets will be the main contributors to degradation. In phosphorescent materials, rapid intersystem crossing will lead to a negligible singlet population and triplet excitons may contribute significantly.

The exciton population in the presence of degradation products in a uniform exciton recombination zone can be described by the steady-state rate equation

0 = (7-2) 퐽� � 푒푒 − 휏 − 푘퐸�푛� where: J is the current density; e is the elementary charge; w is the exciton recombination zone width; φ is a spin factor to account for whether singlet or triplet excitons cause degradation; τ is

the exciton lifetime; and kET is the exciton energy transfer rate constant to the degradation products.

Equations 7-1 and 7-2 form a set of coupled differential equations for the exciton and defect concentrations, and can be solved analytically for n and N. Assuming the luminance is proportional to the exciton concentration, the solution can be expressed as follows:

= (7-3) � 1 �0 푡 �1+3푡50

where L0 and L are the initial and instantaneous device luminance and = is the 3푒푒 2 time at which the luminance drops to half of its initial value. The inverse푡50 relationship퐽�2푘퐸�휏 푘 between퐹

t50 and the current density, J, has been observed for many OLEDs and termed “coulombic degradation” by early studies.

Equation 7-3 predicts the drop in normalized OLED luminance as a function of time is

completely characterized by t50 when the defect formation rate follows Equation 7-1.

7.3 Results

Given the agreement between theory and experiment for the time and current density dependence of the degradation we observe in CBP bilayer devices in the previous chapter, we analyze the

results of a previous study by Wang et al. of degradation in neat hole transport materials (HTM) in a simple organic bilayer structure consisting of a single HTM/TPBi heterojunction 33. The OLEDs are each driven at a constant current density of 20 mA/cm2. The molecular structures of the HTMs analyzed in this study are provided in Figure 7-1b.

Of the materials considered in this previous study, CBP25,27, TCTA28, NPB26 and TAPC26 have been identified as undergoing exciton driven degradation reactions. Given that spiro-CBP, and spiro-NPB have chemical structures closely related to those of CBP and NPB, it is likely that these molecules will also undergo exciton driven degradation.

Degradation of the HTM emission intensity is shown in Figure 7-1a along with fits to Equation

7-3 using t50 as the only free parameter. With the exception of the device with an NPB HTM, each of the OLEDs in Figure 7-1a show a close match to the model; fit quality metrics are provided in the suppporting information. It is noted that the time scale for the degradation of these neat layers is much shorter than for host:dopant OLEDs. This is in line with expectations based on exiton driven degradation since the exciton concentration on doped hosts is much lower due to rapid energy transfer from the host to dopants.

The extracted values for t50 from the devices in Figure 7-1a are plotted against the HTM singlet

exciton energy in Figure 7-1b. The general trend towards smaller t50 with larger singlet energy supports the hypothesis that the energy for degradation events is supplied by excitons.

Figure 7-1: a) Degradation of HTM/TPBi bilayer OLEDs from ref. 33 driven at 20mA/cm2 (solid dots) and fits to Equation 7-3 (lines). b) Luminance half-life as a function of the HTM singlet energy

Given the wide range of possible degradation mechanisms and the large variation in the molecular structure of the HTMs in this study, the observation of a common degradation profile in Figure 7-1a is a surprising result. The close match between the data and the expression for exciton driven degradation given by Equation 7-3 suggests that each of these materials undergoes degradation reactions driven by excitons as described above.

We have shown that Equation 7-3 works well in describing the degradation of common HTMs in OLEDs with a simple structure. We now extend our analysis to literature degradation profiles for state-of-the-art OLEDs reported by various industrial and academic laboratories. The

degradation data of these OLEDs are shown in Figure 7-2 along fits to Equation 7-3 using t50 as the only free parameter. The data in Figure 7-2 include fluorescent blue91,92, phosphorescent blue31, phosphorescent green 93, thermally activated delayed fluorescence green94 and white95 OLEDs. The excellent fits to the reported data (see Appendix E for fit quality metrics) over this broad range of OLEDs suggests that exciton driven reactions described above plays a major role

in dictating the aging in state-of-the-art OLEDs. We note that Ref. 94 concluded in their work that the excited state instability of the host was the primary degradation mechanism.

Figure 7-2: Degradation of state-of-the-art OLEDs31,91–95 (solid dots) and fits to Equation 7-3 (solid lines)

We now discuss the uniqueness of Equation 7-3 in modeling OLED aging. We plot in Figure 7-3 the luminance decay as a function of time from each OLED shown in Figures 1a and 2 on a normalized time scale (t=t/t50 using extracted t50). The NPB device from Figure 7-1a is excluded due to its poor fit to Equation 7-3. It is clear from this figure that despite the broad range of materials used and wide span of lifetime, the aging characteristics of these OLEDs follow the

same time profiles. Equation 7-3 is also plotted on the same axes with L0=1 and t50=1. Equation 3 matches closely to the data, and is able to capture the time dependence of the degradation even down to very low luminance levels. Figure 7-3 highlights that, although Equation 3 is able to closely fit the data from a wide range of OLEDs, it is a fairly rigid model. This suggests that Equation 3 can be useful as a tool for diagnosing degradation mechanisms in various OLEDs.

100

Figure 7-3: Degradation data (solid dots) for the OLEDs in Figures 2a and 3 with normalized time (t50=1) and Equation 3 with L0=1 and t50=1 (dashed line).

7.4 Conclusions

In summary, we have derived a simple analytical formula that can be used to describe the time dependence of luminance degradation in a broad range of OLEDs. The wide range of materials and device architectures which match closely with the analytical model strongly suggests that exciton fueled degradation is a highly prevalent in OLEDs. While other degradation pathways may present in different devices, the unique signature provided by Equation 3 is useful in diagnosing exciton driven degradation. Thus, Equation 3 will find broad use as a diagnostic and predictive tool in analyzing OLED degradation.

101

Chapter 8 OPV Degradation

This work was conducted in collaboration with Prof. Bender in the University of Toronto’s Department of Chemical Engineering. David Josey fabricated the devices and carried out the rooftop aging experiments. David Josey and Prof. Bender participated in the planning and ongoing discussion of results of work presented here; however any experiments and analysis presented here were conducted independently, unless otherwise specified.

Although OPVs are not explicitly optimized for efficient electroluminescence (EL) in forward bias, many of the design requirements are similar for OLEDs and OPVs and most OPVs will produce EL when operated in forward bias. Here, we examine several OPV architectures which emit light from the singlet state of the electron acceptor when operated in forward bias.

In the previous chapters, we have demonstrated that electroluminescence can be a powerful tool to probe degradation in OLEDs. This section will attempt to extend the work from the previous chapters to the study of degradation in OPVs through analysis of their EL. In other words, by operating OPVs as OLEDs we can gain valuable information about the degradation mechanisms of these devices which are not available through more traditional means. While the previous chapters have focused only on intrinsic or operational degradation, this chapter will also deal with environmental, or extrinsic, degradation due to environmental stressors outside of the devices.

Part I analyzes the variations in EL intensity across OPV pixels to determine the extent to which external degradation has affected the performance of a set of OPVs following a period of outdoor photovoltaic testing. Part II focuses the intrinsic degradation of an OPV using a Cl-BsubPc electron acceptor.

8.1 Extrinsic Degradation

There are several imaging techniques available to assess environmental (extrinsic) degradation in OPVs. These include photoluminescence imaging, EL imaging and light beam induced current mapping, often in combination96–100. Here, EL and photoluminescence imaging are used to evaluate the environmental degradation of OPVs following outdoor aging, and assess the mechanism of extrinsic degradation.

Current-voltage data was measured using an HP4140B picoammeter. Photoluminescence imaging used a hand-held UV lamp for excitation. Photoluminescence and EL imaging photographs were taken using a Canon Rebel T3i DSLR camera. Device fabrication and testing was similar to a previous report101

Three OPV structures were studied:

1)”Pc” structure PEDOT:PSS / a6T (50 nm) / Cl-BsubPc (20 nm) / BCP (7 nm) / Ag (80 nm) / encapsulation

2) “Nc” structure PEDOT:PSS / a6T (50 nm) / Cl-BsubNc (25 nm) / BCP (10 nm) / Ag (80 nm) / encapsulation and

3) “Nc/Pc” structure PEDOT:PSS / a6T (55 nm) / Cl-BsubNc (12 nm) / Cl-BsubPc (15 nm) / BCP (7 nm) / Ag (80 nm) / encapsulation.

Where a6T is alpha-sexithiophene, Cl-BsubPc is chloro-boron subphthalocyanine, BCP is bathocuprine, and Cl-BsubNc is boron subnaphthalocyanine chloride. The encapsulation

includes a MoO3 barrier layer followed by UV curable epoxy and a glass slide. The primary difference between these three structures is the choice of electron acceptor(s).

EL will not always be observable in OPVs, however, due to the favorable energetic alignment of the OPVs studied, electrically generated excitons are passed to the Cl-BsubPc or Cl-BsubNc layers which have moderate photoluminescence efficiency, allowing us to observe EL in forward bias. Sample emission spectra of each of the three device architectures studied are shown in Figure 8-1.

103

Figure 8-1: EL spectra representative OPVs with Pc, Nc and Nc/Pc structures. Data are smoothed for clarity.

The OPVs were fabricated on August 8’th 2016 and tested outdoors from September 9’th to October 31’st 2016. Following this outdoor testing period, EL and PL imaging were used to determine the extent to which external degradation contributed to the loss in PCE.

The first three images in Figure 8-2 show photographs a Pc OPV driven at forward current densities of 25mA/cm2. The cell was photographed on three dates; 4, 43, and 163 days after outdoor testing was completed.

104

The boundary between bright and dark regions of the OPVs run parallel to the encapsulation edge (not shown) suggesting environmental degradation through the ingress of water, oxygen or other environmental contaminants from the encapsulation edge. In several devices, we observed a curved boundary between the bright and dark EL regions, an explanation for this phenomena is included in Appendix F.

The average area of bright EL as a fraction of the total pixel area for each device structure is given in Table 8-1. The devices are further broken down into those pixels at the edge and center of the substrate, since the pixels at the edge of the substrate were more susceptible to extrinsic degradation. Table 8-1 also provides the short circuit current density, Jsc, as a fraction of the

initial Jsc measured the morning before outdoor testing began.

105

Four days after the outdoor testing period ended, the magnitude of the losses in bright EL area and Jsc are similar. This may suggest that pixel shrinkage contributed significantly to the observed Jsc loss. To further probe the connection between the bright EL area loss and Jsc loss, the devices were allowed to continue degrading in dark conditions. This decouples the extrinsic degradation due to contaminant diffusion under the encapsulant edges from the operational degradation experienced during outdoor testing. Over the next two testing periods, 43 and 163 days after outdoor testing the Jsc remained constant (within experimental error) for all of the sets of devices in Table 8-1, except for the edge Nc devices, which experienced more significant pixel shrinkage than the other devices. Over the same period, the bright EL areas of each device continued to decrease. These observations suggest pixel shrinkage was not a major source of Jsc loss in these devices. To better understand the mechanism leading to pixel shrinkage, we employed photoluminescence (PL) imaging.

106

Jsc/Jsc,initial Fraction of Bright EL Area

Device Pixel name Location 43 163 163 No. of 4 Days Days Days 4 Days 43 Days Days Devices

0.89 ± 0.91 ± 0.895± 0.95 ± 0.91 ± 0.79 ± Pc Center 0.02 0.02 0.009 0.02 0.03 0.02 2

0.89 ± 0.88 ± 0.88 ± 0.878 ± 0.846 ± 0.78 ± Nc Center 0.04 0.04 0.03 0.007 0.003 0.01 2

0.67 ± 0.569 ± 0.35 ± Edge 0.65 0.74 0.74 0.02 0.005 0.02 2*

0.90 ± 0.90 ± 0.91 ± 0.91 ± 0.90 ± 0.86 ± Nc/Pc Center 0.04 0.04 0.03 0.05 0.05 0.05 4

0.87 ± 0.90 ± 0.90 ± 0.945 ± 0.7 ± Edge 0.02 0.03 0.04 0.009 0.8 ± 0.3 0.1 3

* Initial Jsc not available for one of two Nc edge pixels.

PL of a Pc OPV under UV illumination 163 days after rooftop testing is shown in Figure 8-2. Uniform PL was observed over the entire pixel, with similar observations for all other devices in

107

this study. In contrast the EL images shown in Figure 8-2, there is no evidence of environmental degradation, for example reduced PL close to the encapsulant edges.

Uniform PL in regions that are dark in the EL imaging is characteristic of degradation of one or both electrode/organic interfaces, without degradation of the bulk materials. This degradation may be due to local delamination of the silver electrode resulting in a loss of charge injection at the Ag/organic interface, or due to increased contact resistance at either of the electrode/organic interfaces, resulting in reduced charge injection efficiency. The independence of the Jsc and bright EL area observed in Table 8-1 suggests that the latter is more likely.

Probing the spatial dependence of photovoltaic performance allows us to distinguish between electrode delamination and increased contact resistance at electrodes. Cathode delamination will result in the complete loss of both OLED and OPV activity, while increased contact resistance will only reduce OLED and OPV efficiency but not eliminate it entirely.

The Pc OPV was masked in such a way that only regions dark under the EL imaging were exposed to light. The resulting current-voltage characteristics under AM1.5G irradiation are shown in Figure 8-3 along with an unmasked device for comparison. Note that the current densities are calculated using the area of the device exposed to light, not the entire pixel area.

Figure 8-3: Current-voltage characteristic curves of an OPV device under simulated AM1.5 illumination and the same device with only the region dark under EL testing exposed to light.

108

The open-circuit voltage, Jsc, and fill factor are all reduced when exposing only the degraded portion of the device to AM1.5G radiation. This observation not consistent with electrode delamination, where we would expect negligible photocurrent, thus we conclude that the shrinkage in EL area is due to increased contact resistance at one or both electrodes as a result of ingress of environmental contaminants.

In summary, for 163 days following outdoor testing short circuit current densities remained roughly constant for OPVs with three different structures, while the emissive EL area shrunk. The independence of these quantities suggested that dark areas during EL imaging were not photovoltaically inactive. This conclusion was supported by comparisons between EL and PL imaging as well as testing the photovoltaic performance of the dark EL regions. While improving the encapsulation quality is desirable, it is unlikely that pixel shrinkage contributed significantly to the device degradation over the outdoor testing period.

8.2 Intrinsic Degradation

Cl-BsubPc and its derivatives have demonstrated a tendency to aggregate in solution and solid state54,102,103. Aggregation is characterized by the appearance of an emission band at lower energies than the monomer emission in the photoluminescence and electroluminescence spectra.

Aggregate emission is observed in many newly fabricated devices and films. One study also observed an increase in intensity of aggregate photoluminescence over the period of several days following deposition for a thin film while being stored in an inert environment54. It is unclear how the presence of aggregates affects exciton diffusion, exciton dissociation, and charge transport.

This section will analyze how the relative intensity of the aggregate and monomer emission change over the lifetime of one of the Pc OPVs described in the previous section to determine if aggregation is a source of degradation in this device.

109

EL spectra of one Pc OPV were recorded throughout the testing period to probe the connection between aggregate formation and OPV degradation. OPV electroluminescence spectra were collected by coupling the emission into an optical fiber connected to an Ocean Optics USB4000 spectrometer. Current-voltage data was measured using an HP4140B picoammeter.

Electroluminescence from the Pc solar cell exhibits both monomer and aggregate emission centered at 625nm and 716nm respectively. The normalized emission spectra 4 days after device fabrication at several current densities are shown in Figure 8-4a. The relative intensity of the monomer emission increases monotonically with the current density.

The ratio of monomer:aggregate emission (M:A) is plotted as a function of current density in Figure 8-4b. The relationship between the two quantities is approximately linear and was fit using a weighted linear regression. This linear relationship was also observed when similar measurements were repeated on the OPV throughout the study period, although the slope and intercept of the fits varied in each measurement.

The dependence of the M:A emission ratio on current density is similar to the color shift commonly observed in OLEDs with multiple emitters. The field dependence of charge trapping, exciton formation, and exciton quenching processes may contribute to this shift in relative emission intensity.

110

The slope and intercept of the fits described above characterize the empirical relationship between the applied current density and the M:A emission ratio. The dependence of these parameters on the testing date is shown in Figure 8-5a. The first measurement was taken several days after fabrication, while subsequent measurements take place before, during and after the outdoor testing period. The relationship between these empirical parameters and time is highly nonlinear.

The OPV driving voltage at the current density of 7.5mA/cm2 is shown in Figure 8-5b. The driving voltage provides a measurable indicator of the degradation induced changes in the electrical characteristics of the device. The driving voltage also changes nonlinearly with time.

111

Figure 8-5: a) Intercept (red squares) and inverse of the slope (black circles) extracted from fits to the emission ratio vs current density, and b) the driving voltage at 7.5mA/cm2 plotted as a function of the number of days after device fabrication.

The changes in the slope and intercept over time means that for a given current density, different M:A ratios are observed on different testing dates. This could indicate shifts in the composition of the Cl-BsubPc layer of this device. Alternatively, since the M:A ratio depends on the electrical properties of the device (as shown in Figure 8-4b) and the electrical characteristics change over time (as shown in Figure 8-5b) the composition could be unchanged and the differences in emission spectra due purely to the changes in electrical characteristics.

Plotting the intercept and the inverse of the slope against the driving voltage at 7.5mA applied current in Figure 8-6, we find that these quantities depend linearly on the driving voltage. The observation of a simple linear relationship between these measureable quantities suggests a physical connection, especially when considering the significant nonlinearity observed for each as a function of time (as demonstrated in Figure 8-5). This indicates that the variations in the observed M:A ratio are a result of changes in the electrical characteristics, for example through variations in emission efficiency linked to charge balance, rather than morphological changes.

112

Figure 8-6: Intercept (red squares) and inverse of the slope (black circles) extracted from fits to the emission ratio vs current density, plotted against the driving voltage at 7.5mA. Solid lines are linear fits to each set of parameters.

In summary, the intrinsic degradation of a Pc OPV was analyzed. Variations in the current density dependence of the ratio of monomer to aggregate emission were attributed to degradation of the electrical characteristics of the OPV rather than changes in film morphology. EL analysis may be a useful tool for understanding of exciton dynamics in mixed monomer/aggregate films of Cl-BsubPc and its derivatives.

113

Chapter 9 Conclusions and Future Work

9.1 Conclusions

As outlined in the Introduction section, a major limiting factor in the progress of the field of organic electronics is the immense number of possible organic compounds which are potential candidates for use in organic electronic devices. Given the wide range of physical properties exhibited by these materials, conclusions which can be generalized to multiple materials are extremely valuable. This is important to consider when evaluating the merit of degradation models proposed in this work. A model which can describe the degradation of a single material is a valuable contribution to the field, but the true value of the equations described in Chapters 6 and 7 is that they appear to be applicable to a broad range of materials and device architectures. This is not to say that the singlet exciton driven degradation mechanism which we describe is the only mechanism at play in state of the art OLEDs, but it does appear to be important in dictating the degradation of a wide range of host materials. The approach taken here to studying degradation here is to look at systems that are as simple as possible and look at more complicated systems after firmly establishing the device physics of simple cases. This is in contrast to most studies of OLED degradation which focus only on state-of-the-art devices.

In Chapters 5, 6 and 8 defect emission was used to quantify the defect concentration in OLEDs. Although this may seem like an obvious choice, defect emission has not been quantified in a similar way in the past due to two factors: 1) in many materials defect emission is extremely weak making accurate measurement difficult and 2) defect-defect, defect-host and defect-charge interactions mean that defect emission and defect concentration do not have a simple linear relationship. Materials systems such as the one described here, which allow for the quantification of defect concertation via emission spectra may be useful to study other aspects of defects which are more difficult in most common materials. This again demonstrates the benefits of looking at simple systems first and applying the results to a broader range of materials and devices.

A recurring conclusion throughout this work is the importance of interfaces in governing the efficiency and stability of OLEDs. Given the high electron and hole mobilities reported for

CBP, conventional wisdom would predict a broad exciton generation zone. In contrast to this expectation, exciton generation is localized to the ETL/HTL interface for both doped and undoped emissive layer OLEDs. In addition to localized exciton generation, we also observe an exciton recombination zone localized to the interface in an OLED with a doped emissive layer. This is a surprising result given the high efficiency of this device; the narrow recombination zone implies high exciton concentrations which typically lead to high rates of exciton quenching. Interfaces also play a key role in degradation of the OLEDs studied. The importance of interfaces in OLED degradation has been previously noted; however it has not been clarified if the importance of interfaces is purely due to the higher exciton and charge concentrations near interfaces, or if the interfaces themselves contribute to instability. Here, it has been demonstrated that excitons close to interfaces indeed contribute disproportionately to the overall device degradation. A natural question to ask is whether the spatially coincident exciton generation zone and primary degradation zone are related. If charge carriers participate in a non- rate-limiting step in the degradation process this would explain the connection between these two phenomena since exciton generation occurs in regions of high charge concentration.

CBP is the focus of several of the chapters of this thesis. Given its limited stability, use in commercial application is unlikely, however the wide range of studies reporting on CBP make it one of the best characterized materials available. Given the still limited understanding of fundamental degradation processes, CBP remains a good candidate for fundamentals studies such as the ones presented here.

Exciton device physics is at the heart of the efficiency and stability of OLEDs. Applying the tools and knowledge of studies of excitons in organic materials to the problem of OLED and OPV degradation has provided useful results in this work, and has significant potential for moving the field forward in future studies, which are described in the next section.

9.2 Future Work

This section outlines experiments which may extend, support, or supplement the work presented in this thesis. Experiments are grouped into those involving exciton diffusion and those focused

115

on operational degradation, although in practice there may (and hopefully will) be some overlap between these topics.

9.2.1 Exciton Diffusion

9.2.1.1 Energetic Disorder

In this work and elsewhere, energetic disorder is assumed to be minimal such that exciton diffusivity can be assumed to be time, position, and temperature independent. In reality, due to the energetic disorder present in molecular solids, exciton diffusivity will decrease in time as the exciton relaxes into the lower energy states through energy transfer between molecules. The two methods that I suggest to analyze the impact of energetic disorder are time and temperature dependent studies of exciton diffusion.

Time dependent studies of exciton diffusion are more suitable for photoluminescence based studies rather than electroluminescence. Time resolved studies of singlet excitons produced through electrical excitation are hindered by the slow electrical response of OLEDs relative to the singlet lifetime. Phosphorescence lifetimes are sufficiently long to probe triplet excitons produced through electrical excitation, however time resolved analysis of electroluminescence is complicated by the slow recombination of charges on the same time scale as the phosphorescence. Time resolved photoluminescence can be used to probe exciton relaxation into lower energy states by either monitoring the spatial distribution of the photoluminescence as a function of time, or by monitoring the spectral shifts as a function of time.

Temperature dependence can also be used to probe energetic disorder and its effect on exciton diffusion. Several studies have approached this problem by analyzing temperature dependence of photoluminescence quenching in organic bilayers 69,104,105. Figure 9-1 demonstrates how Mikhnenko et al. visualize the temperature dependence of exciton diffusion in the energetically disordered MDMO-PPV film.

116

Extending the work in these studies to measurements of exciton diffusion in active OLEDs allows for the possibility of studying triplet excitons in fluorescent materials as well as fluorescent materials. Identifying how molecular structure influences energetic disorder and how energetic disorder in turn effects exciton diffusion in OLEDs and OPVs would be an important contribution to the field and address one limitation of this work.

117

9.2.1.2 Heavy Atom Effect in OPV

Extending the exciton diffusion length in organic materials is a topic of major interest in the OPV community as the power conversion efficiency of OPVs is often limited by the diffusion of excitons to dissociating heterojunctions. Enhancing singlet exciton diffusion in OPVs to increase exciton harvesting has been the topic of many previous studies which have attempted to use solid state dilution 106, molecular engineering 64 and triplet sensitization107–109 to increase the effective diffusion length of excitons. The subset of these studies that attempt to increase effective exciton diffusion length by promoting production of triplet states rely on the assumption that triplet excitons have significantly longer exciton diffusion lengths than singlets.

Although the internal heavy atom effect has been exploited through the use of triplet sensitizers, the external heavy atom effect has yet to be exploited for diffusion length increase in OPVs. The external heavy atom effect describes an increase in spin-orbit coupling in molecules adjacent to heavy atom containing molecules or polymers. Incorporation of a relatively small amount of heavy atom containing material which does not absorb significantly in the visible region or contribute to charge trapping in the device may allow for exciton diffusion enhancement in a wide range of materials.

9.2.2 Operational Degradation

9.2.2.1 Temperature Dependence of OLED Degradation

The temperature dependence of OLED degradation can provide useful information in two respects. First, from a practical perspective understanding the temperature dependence of OLED degradation is useful for accelerated aging studies used to determine the potential lifetime of OLEDs at standard operating conditions without requiring tests to run for thousands or tens of thousands of hours. These tests are instead usually run at elevated temperatures and the degradation is assumed to be described by single activation energy, allowing for predictable acceleration of degradation without changing the degradation mechanism. Second, the

118

temperature dependence may provide useful information about chemical processes and kinetics of the degradation mechanism.

In the context of exciton driven degradation as describe in this work, the activation energy may correspond to a physical value such as the difference between the singlet excited state energy and a higher energy dissociative state which leads to photochemical degradation. By studying the thermal activation of exciton driven degradation across multiple materials, we may be able to better understand the degradation mechanism and possible set guidelines for reducing the rate of degradation.

A study integrating temperature and current density dependence of both stability and efficiency roll-off is a very interesting path forward.

9.2.2.2 Understanding Multi-Step Reactions

A major discrepancy exists between the observed previous studies of OLED degradation in CBP. Kondakov et al. have suggested that degradation rates increase proportionally with the applied current density 27, suggesting degradation by monomolecular reactions of singlet excitons. This is the same conclusion drawn in this work. These observations contrast the work Wang et al. who observed elevated degradation rates in situations where excitons and charges were coincident 33. This observation was taken to indicate that bimolecular interactions of singlets and charges were leading to degradation events. This mechanism was also proposed for a different materials system by Geibink et al. 31. A resolution of these two apparently conflicting results is to suggest that the reaction occurs in multiple steps, with the rate limiting step controlled by monomolecular singlet exciton reactions and another (either prior or subsequent) step involving charges. The possibility of multi-step reactions is accepted in the community and reaction pathways involving multi-step reactions are often proposed based on observed degradation products.

Alternate light and current stressing may provide evidence for a multi-step reaction as well as indicate the order of the steps if this is the case. For example, if a primary step involves only

119

charge carriers and a second step involves only light, photodegradation will be enhanced by prior electrical stressing due to the build-up of primary degradation products.

Involving synthetic chemists the future studies such as those above would be beneficial, since ultimately the processes we are interested in are chemical reactions. For materials which have emissive defects, the possibility of synthesizing possible degradation products and comparing their emission to defect emission from devices presents an exciting possibility.

Experiments such as those described above could provide very useful information about reaction kinetics which is currently a topic of speculation.

9.2.3 Exciton Driven OPV Degradation

Exciton driven degradation is not limited to OLED materials; studies which suggest that similar degradation processes may be important in determining the lifetime of OPVs 110. Resistance to operational degradation is important if OPVs are to meet the demanding lifetime requirements for most photovoltaic applications.

The time dependence of the photocurrent degradation due to exciton driven degradation of OPVs would likely follow similar time dependence as that observed for OLEDs in this work. For the case of OPVs, changes in absorption, the spatial distribution of excitons, and the effects of charge trapping on PCE may need to be considered. The availability of physically based equations for degradation of OPVs would undoubtedly be highly valued.

120

References

1. Pope, M., Kallmann, H. P. & Magnante, P. Electroluminescence in Organic Crystals. J. Chem. Phys. 38, 2042–2043 (1963).

2. Tang, C. W. & VanSlyke, S. A. Organic electroluminescent diodes. Appl. Phys. Lett. 51, 913 (1987).

3. Tang, C. W., VanSlyke, S. A. & Chen, C. H. Electroluminescence of doped organic thin films. J. Appl. Phys. 65, 3610–3616 (1989).

4. Baldo, M. A., Thompson, M. E. & Forrest, S. R. High-efficiency fluorescent organic light- emitting devices using a phosphorescent sensitizer. Nature 403, 1998–2001 (2000).

5. Adachi, C., Baldo, M. a., Thompson, M. E. & Forrest, S. R. Nearly 100% internal phosphorescence efficiency in an organic light-emitting device. J. Appl. Phys. 90, 5048 (2001).

6. Pope, M. Electric Currents in Organic Crystals. Sci. Am. 216, 86–97 (1967).

7. Pope, M. & Swenberg, C. E. Electronic Processes in Organic Crystals and Polymers. (Oxford University Press, 1999).

8. Miller, A. & Abrahams, E. Impurity conduction at low concentrations. Phys. Rev. 120, 745–755 (1960).

9. Scott, J. C. & Malliaras, G. G. Charge injection and recombination at the metal – organic interface. 115–119 (1999).

10. Köhler, A. & Bässler, H. Electronic Processes in Organic Semiconductors: An Introduction. (2015).

11. Hirata, S. et al. Efficient Persistent Room Temperature Phosphorescence in Organic Amorphous Materials under Ambient Conditions. Adv. Funct. Mater. 23, 3386–3397 (2013).

121

12. Baldo, M. A. et al. Highly efficient phosphorescent emission from organic electroluminescent devices. Nature 395, 151–154 (1998).

13. Lamansky, S. et al. Highly phosphorescent bis-cyclometalated iridium complexes: synthesis, photophysical characterization, and use in organic light emitting diodes. J. Am. Chem. Soc. 123, 4304–12 (2001).

14. Förster, T. Transfer mechanisms of electronic excitation energy. Discuss. Faraday Soc. 27, 7–17 (1959).

15. Baldo, M. A., O’Brien, D. F., Thompson, M. E. & Forrest, S. R. Excitonic singlet-triplet ratio in a semiconducting organic thin film. Phys. Rev. B 60, 14422–14428 (1999).

16. Wünsche, J., Reineke, S., Lüssem, B. & Leo, K. Measurement of triplet exciton diffusion in organic light-emitting diodes. Phys. Rev. B 81, 245201 (2010).

17. Hofmann, S., Rosenow, T. C., Gather, M. C., Lüssem, B. & Leo, K. Singlet exciton diffusion length in organic light-emitting diodes. Phys. Rev. B 85, 245209 (2012).

18. Baldo, M. a. & Forrest, S. R. Transient analysis of organic electrophosphorescence: I. Transient analysis of triplet energy transfer. Phys. Rev. B 62, 10958–10966 (2000).

19. Jarikov, V. V., Kondakov, D. Y. & Brown, C. T. Efficient and extremely long-lived organic light-emitting diodes based on dinaphthylperylene. J. Appl. Phys. 102, (2007).

20. Scholz, S., Kondakov, D., Lüssem, B. & Leo, K. Degradation mechanisms and reactions in organic light-emitting devices. Chem. Rev. 115, 8449–8503 (2015).

21. Schmidbauer, S., Hohenleutner, A. & K??nig, B. Chemical degradation in organic light- emitting devices: Mechanisms and implications for the design of new materials. Adv. Mater. 25, 2114–2129 (2013).

22. Smith, P. F., Gerroir, P., Xie, S., Hor, A. M. & Popovic, Z. Degradation of Organic Electroluminescent Devices . Evidence for the Occurrence of Spherulitic Crystallization in the Hole Transport Layer. 80, 5946–5950 (1998).

122

23. Kondakov, D. Y., Sandifer, J. R., Tang, C. W. & Young, R. H. Nonradiative recombination centers and electrical aging of organic light-emitting diodes: Direct connection between accumulation of trapped charge and luminance loss. J. Appl. Phys. 93, 1108 (2003).

24. Aziz, H. Z. D. P. N.-X. H. A.-M. H. G. X. Degradation Mechanism of Small Molecule – Based Organic Light-Emitting Devices. Science (80-. ). 283, 1900–1902 (1999).

25. Kondakov, D. Y., Pawlik, T. D., Nichols, W. F. & Lenhart, W. C. Free-radical pathways in operational degradation of OLEDs. J. Soc. Inf. Disp. 16, 37–46 (2008).

26. Kondakov, D. Y. Role of chemical reactions of arylamine hole transport materials in operational degradation of organic light-emitting diodes. J. Appl. Phys. 104, (2008).

27. Kondakov, D. Y., Lenhart, W. C. & Nichols, W. F. Operational degradation of organic light-emitting diodes: Mechanism and identification of chemical products. J. Appl. Phys. 101, (2007).

28. Scholz, S., Walzer, K. & Leo, K. Analysis of complete organic semiconductor devices by laser desorption/ionization time-of-flight mass spectrometry. Adv. Funct. Mater. 18, 2541–2547 (2008).

29. Seifert, R., Scholz, S., Lüssem, B. & Leo, K. Comparison of ultraviolet- and charge- induced degradation phenomena in blue fluorescent organic light emitting diodes. Appl. Phys. Lett. 97, 1–4 (2010).

30. Bell, B. M. et al. Degradation of Hole Transport Materials via Exciton-Driven Cyclization. ACS Appl. Mater. Interfaces 9, 13369–13379 (2017).

31. Giebink, N. C. et al. Intrinsic luminance loss in phosphorescent small-molecule organic light emitting devices due to bimolecular annihilation reactions. J. Appl. Phys. 103, (2008).

32. Giebink, N. C., D’Andrade, B. W., Weaver, M. S., Brown, J. J. & Forrest, S. R. Direct evidence for degradation of polaron excited states in organic light emitting diodes. J. Appl.

123

Phys. 105, (2009).

33. Wang, Q., Sun, B. & Aziz, H. Exciton-polaron-induced aggregation of wide-bandgap materials and its implication on the electroluminescence stability of phosphorescent organic light-emitting devices. Adv. Funct. Mater. 24, 2975–2985 (2014).

34. Lee, J. et al. Hot excited state management for long-lived blue phosphorescent organic light-emitting diodes. Nat. Commun. 8, 15566 (2017).

35. Giebink, N. C. et al. Intrinsic luminance loss in phosphorescent small-molecule organic light emitting devices due to bimolecular annihilation reactions. J. Appl. Phys. 103, 44509 (2008).

36. Helander, M. G. et al. Chlorinated indium tin oxide electrodes with high work function for organic device compatibility. Science 332, 944–7 (2011).

37. Wang, Z. B. et al. Highly simplified phosphorescent organic light emitting diode with >20% external quantum efficiency at >10,000 cd/m2. Appl. Phys. Lett. 98, 73310 (2011).

38. Chang, Y.-L. et al. Enhancing the efficiency of simplified red phosphorescent organic light emitting diodes by exciton harvesting. Org. Electron. 13, 925–931 (2012).

39. Chang, Y.-L. et al. Highly Efficient Warm White Organic Light-Emitting Diodes by Triplet Exciton Conversion. Adv. Funct. Mater. 23, 705–712 (2013).

40. D’Andrade, B. W., Thompson, M. E. & Forrest, S. R. Controlling Exciton Diffusion in Multilayer White Phosphorescent Organic Light Emitting Devices. Adv. Mater. 14, 147– 151 (2002).

41. Lebental, M. et al. Diffusion of triplet excitons in an operational organic light-emitting diode. Phys. Rev. B 79, 165318 (2009).

42. Ingram, G. L., Chang, Y.-L. & Lu, Z. H. Probing the Exciton distribution in Organic Light Emitting Diodes Using Long Range Energy Transfer. Can. J. Phys. 92, 845–848 (2014).

43. Haynes, D. R., Tokmakoff, A. & George, S. M. Distance dependence of electronic energy

124

transfer between donor and acceptor adlayers: p-terphenyl and 9,10-diphenylanthracene. J. Chem. Phys. 100, 1968–1980 (1994).

44. Burroughs, J., Bradley, D. & Brown, A. Solid C60: a new form of carbon. Nature 347, 354–358 (1990).

45. Maksimov, M. & Rozman, I. On the energy transfer in rigid solutions. Opt. Spectrosc. 12, 337–338 (1962).

46. Wang, Z. B. et al. Unlocking the full potential of organic light-emitting diodes on flexible plastic. Nat. Photonics 5, 753–757 (2011).

47. Ingram, G. L., Nguyen, C. & Lu, Z. H. Long-Range Energy Transfer and Singlet-Exciton Migration in Working Organic Light-Emitting Diodes. Phys. Rev. Appl. 5, 64002 (2016).

48. Kim, S. Y. et al. Organic light-emitting diodes with 30% external quantum efficiency based on a horizontally oriented emitter. Adv. Funct. Mater. 23, 3896–3900 (2013).

49. Kim, K. et al. Phosphorescent dye-based supramolecules for high-efficiency organic light- emitting diodes. Nat. Commun. 5, 1–8 (2014).

50. Scully, S. R. & McGehee, M. D. Effects of optical interference and energy transfer on exciton diffusion length measurements in organic semiconductors. J. Appl. Phys. 100, 34907 (2006).

51. Lunt, R. R., Giebink, N. C., Belak, A. a., Benziger, J. B. & Forrest, S. R. Exciton diffusion lengths of organic semiconductor thin films measured by spectrally resolved photoluminescence quenching. J. Appl. Phys. 105, 53711 (2009).

52. Bergemann, K. J. & Forrest, S. R. Measurement of exciton diffusion lengths in optically thin organic films. Appl. Phys. Lett. 99, 243303 (2011).

53. Banerjee, S., Parhi, A. P., Iyer, S. S. K. & Kumar, S. Method of determining the exciton diffusion length using optical interference effect in Schottky diode. Appl. Phys. Lett. 94, 223303 (2009).

125

54. Gommans, H., Schols, S., Kadashchuk, A. & Heremans, P. Exciton Diffusion Length and Lifetime in Subphthalocyanine Films. J. Phys. Chem. C 2974–2979 (2009).

55. Sun, Y. et al. Management of singlet and triplet excitons for efficient white organic light- emitting devices. Nature 440, 908–912 (2006).

56. Divayana, Y. & Sun, X. Observation of Excitonic Quenching by Long-Range Dipole- Dipole Interaction in Sequentially Doped Organic Phosphorescent Host-Guest System. Phys. Rev. Lett. 99, 143003 (2007).

57. Staroske, W., Pfeiffer, M., Leo, K. & Hoffmann, M. Single-Step Triplet-Triplet Annihilation: An Intrinsic Limit for the High Brightness Efficiency of Phosphorescent Organic Light Emitting Diodes. Phys. Rev. Lett. 98, 8–11 (2007).

58. Baek, H.-I., Lee, C. & Chin, B. D. Comparison of the carrier mobility, unipolar conduction, and light emitting characteristics of phosphorescent host–dopant system. Synth. Met. 162, 2355–2360 (2012).

59. Lee, S., Yasuda, T., Yang, M., Fujita, K. & Tsutsui, T. Charge carrier mobility in vacuum- sublimed dye films for light emitting diodes studied by the time-of-flight technique. Mol. Cryst. Liq. Cryst. 405, 67–73 (2003).

60. Hung, W.-Y. et al. Employing ambipolar oligofluorene as the charge-generation layer in time-of-flight mobility measurements of organic thin films. Appl. Phys. Lett. 88, 64102 (2006).

61. Wong, T. C., Kovac, J., Lee, C. S., Hung, L. S. & Lee, S. T. Transient electroluminescence measurements on electron-mobility of N-arylbenzimidazoles. Chem. Phys. Lett. 334, 61–64 (2001).

62. Hudson, Z. M., Wang, Z., Helander, M. G., Lu, Z.-H. & Wang, S. N-heterocyclic carbazole-based hosts for simplified single-layer phosphorescent OLEDs with high efficiencies. Adv. Mater. 24, 2922–8 (2012).

63. Ingram, G. L. & Lu, Z. H. Impact of defects on exciton diffusion in organic light-emitting

126

diodes. Org. Electron. 50, 48–54 (2017).

64. Mullenbach, T. K., McGarry, K. a, Luhman, W. a, Douglas, C. J. & Holmes, R. J. Connecting molecular structure and exciton diffusion length in rubrene derivatives. Adv. Mater. 25, 3689–93 (2013).

65. Masri, Z. et al. Molecular weight dependence of exciton diffusion in poly(3- hexylthiophene). Adv. Energy Mater. 3, 1445–1453 (2013).

66. Raisys, S. et al. Exciton diffusion enhancement in triphenylamines via incorporation of phenylethenyl sidearms. J. Mater. Chem. C 2, 4792 (2014).

67. Curtin, I. J., Blaylock, D. W. & Holmes, R. J. Role of impurities in determining the exciton diffusion length in organic semiconductors. Appl. Phys. Lett. 108, 163301 (2016).

68. Mikhnenko, O. V. et al. Trap-limited exciton diffusion in organic semiconductors. Adv. Mater. 26, 1912–1917 (2014).

69. Lin, J. D. A., Mikhnenko, O. V., Van Der Poll, T. S., Bazan, G. C. & Nguyen, T. Q. Temperature dependence of exciton diffusion in a small-molecule organic semiconductor processed with and without additive. Adv. Mater. 27, 2528–2532 (2015).

70. Lunt, R. R., Benziger, J. B. & Forrest, S. R. Relationship between crystalline order and exciton diffusion length in molecular organic semiconductors. Adv. Mater. 22, 1233–6 (2010).

71. Mikhnenko, O. V et al. Effect of thermal annealing on exciton diffusion in a diketopyrrolopyrrole derivative. Phys. Chem. Chem. Phys. 14, 14196–14201 (2012).

72. Yang, J., Zhu, F., Yu, B., Wang, H. & Yan, D. Simultaneous enhancement of charge transport and exciton diffusion in single-crystal-like organic semiconductors. Appl. Phys. Lett. 100, 103305 (2012).

73. Sim, M. et al. Dependence of exciton diffusion length on crystalline order in conjugated polymers. J. Phys. Chem. C 118, 760–766 (2014).

127

74. Siebbeles, L. D. a., Huijser, A. & Savenije, T. J. Effects of molecular organization on exciton diffusion in thin films of bioinspired light-harvesting molecules. J. Mater. Chem. 19, 6067–6072 (2009).

75. Pinto, R. M. et al. Effect of Molecular Stacking on Exciton Diffusion in Crystalline Organic Semiconductors. J. Am. Chem. Soc. 137, 7104–7110 (2015).

76. Ayzner, A. L. et al. Impact of the Crystallite Orientation Distribution on Exciton Transport in Donor-Acceptor Conjugated Polymers. ACS Appl. Mater. Interfaces 7, 28035–28041 (2015).

77. Wang, Q. & Aziz, H. The different influence of singlet and triplet excitons in the degradation of phosphorescent organic light-emitting devices due to exciton–polaron- induced aggregation of host materials. Org. Electron. 26, 464–470 (2015).

78. Zhang, Y. & Aziz, H. Influence of the Guest on Aggregation of the Host by Exciton- Polaron Interactions and Its Effects on the Stability of Phosphorescent Organic Light- Emitting Devices. ACS Appl. Mater. Interfaces 8, 14088–14095 (2016).

79. Adachi, C., Baldo, M. A. & Forrest, S. R. Electroluminescence mechanisms in organic light emitting devices employing a europium chelate doped in a wide energy gap bipolar conducting host. J. Appl. Phys. 87, 8049–8055 (2000).

80. Li, F. et al. Highly efficient electrophosphorescence devices based on rhenium complexes. Appl. Phys. Lett. 84, 148–150 (2004).

81. Schneider, D. et al. Ultrawide tuning range in doped organic solid-state lasers. Appl. Phys. Lett. 85, 1886–1888 (2004).

82. Powell, R. C. & Soos, Z. G. Singlet exciton energy transfer in organic solids. J. Lumin. 11, 1–45 (1975).

83. S. Chandrasekhar. Stochastic problems in physics and astronomy. Rev. Mod. Phys. 15, 89 (1943).

128

84. Seifert, R. et al. Chemical degradation mechanisms of highly efficient blue phosphorescent emitters used for organic light emitting diodes. Org. Electron. physics, Mater. Appl. 14, 115–123 (2013).

85. Yu, H., Zhang, Y., Cho, Y. J. & Aziz, H. Exciton-Induced Degradation of Carbazole- Based Host Materials and Its Role in the Electroluminescence Spectral Changes in Phosphorescent Organic Light Emitting Devices with Electrical Aging. ACS Appl. Mater. Interfaces 9, 14145–14152 (2017).

86. Highly Efficient OLEDs with Phosphorescent Materials. (2008).

87. Wang, F. et al. A low-cost phenylbenzoimidazole containing electron transport material for efficient green phosphorescent and thermally activated delayed fluorescent OLEDs. J. Mater. Chem. C 3, 5533–5540 (2015).

88. Kawamura, Y. et al. 100% phosphorescence quantum efficiency of Ir(III) complexes in organic semiconductor films. Appl. Phys. Lett. 86, 71104 (2005).

89. Kawamura, Y., Brooks, J., Brown, J. J., Sasabe, H. & Adachi, C. Intermolecular Interaction and a Concentration-Quenching Mechanism of Phosphorescent Ir(III) Complexes in a Solid Film. Phys. Rev. Lett. 96, 17404 (2006).

90. Van Slyke, S., Chen, C. H. & Tang, C. W. Organic electroluminescent devices with improved stability. Appl. Phys. Lett. 69, 2160–2162 (1996).

91. Heil, H. et al. High-performance OLED materials. Dig. Tech. Pap. - SID Int. Symp. 45, 495–497 (2014).

92. Kawamura, M. et al. 39.4: Highly Efficient Fluorescent Blue OLEDs with Efficiency- enhancement Layer. SID Symp. Dig. Tech. Pap. 41, 560 (2010).

93. Fukagawa, H. et al. Highly efficient and stable organic light-emitting diodes with a greatly reduced amount of phosphorescent emitter. Sci. Rep. 5, 9855 (2015).

94. Sandanayaka, A. S. D., Matsushima, T. & Adachi, C. Degradation Mechanisms of

129

Organic Light-Emitting Diodes Based on Thermally Activated Delayed Fluorescence Molecules. J. Phys. Chem. C 119, 23845–23851 (2015).

95. Pang, H., Michalski, L., Weaver, M. S., Ma, R. & Brown, J. J. Thermal behavior and indirect life test of large-area OLED lighting panels. J. Solid State Light. 1, 7 (2014).

96. Seeland, M., Rsch, R. & Hoppe, H. Luminescence imaging of polymer solar cells: Visualization of progressing degradation. J. Appl. Phys. 109, (2011).

97. Hoyer, U. et al. Comparison of electroluminescence intensity and photocurrent of polymer based solar cells. Adv. Energy Mater. 1, 1097–1100 (2011).

98. Rösch, R. et al. Investigation of the degradation mechanisms of a variety of organic photovoltaic devices by combination of imaging techniques—the ISOS-3 inter-laboratory collaboration. Energy Environ. Sci. 5, 6521 (2012).

99. Feron, K., Nagle, T. J., Rozanski, L. J., Gong, B. B. & Fell, C. J. Spatially resolved photocurrent measurements of organic solar cells: Tracking water ingress at edges and pinholes. Sol. Energy Mater. Sol. Cells 109, 169–177 (2013).

100. Klumbies, H. et al. Water ingress into and climate dependent lifetime of organic photovoltaic cells investigated by calcium corrosion tests. Sol. Energy Mater. Sol. Cells 120, 685–690 (2014).

101. Josey, D. S. et al. Outdoor Performance and Stability of Boron Subphthalocyanines Applied as Electron Acceptors in Fullerene-Free Organic Photovoltaics. ACS Energy Lett. 2, 726–732 (2017).

102. Morse, G. E., Castrucci, J. S., Helander, M. G., Lu, Z.-H. & Bender, T. P. Phthalimido- boronsubphthalocyanines: new derivatives of boronsubphthalocyanine with bipolar electrochemistry and functionality in OLEDs. ACS Appl. Mater. Interfaces 3, 3538–44 (2011).

103. Helander, M. G. et al. Pentafluorophenoxy boron subphthalocyanine as a fluorescent dopant emitter in organic light emitting diodes. ACS Appl. Mater. Interfaces 2, 3147–52

130

(2010).

104. Matthew Menke, S. & Holmes, R. J. Evaluating the role of energetic disorder and thermal activation in exciton transport. J. Mater. Chem. C 4, 3437–3442 (2016).

105. Mikhnenko, O. V et al. Temperature dependence of exciton diffusion in conjugated polymers. J. Phys. Chem. B 112, 11601–4 (2008).

106. Menke, S. M., Luhman, W. A. & Holmes, R. J. Tailored exciton diffusion in organic photovoltaic cells for enhanced power conversion efficiency. Nat. Mater. 12, 152–157 (2013).

107. Luhman, W. a. & Holmes, R. J. Enhanced exciton diffusion in an organic photovoltaic cell by energy transfer using a phosphorescent sensitizer. Appl. Phys. Lett. 94, 153304 (2009).

108. Roberts, S. T. et al. Observation of triplet exciton formation in a platinum-sensitized organic photovoltaic device. J. Phys. Chem. Lett. 2, 48–54 (2011).

109. Rand, B. P. et al. Photocurrent enhancement in polymer:fullerene bulk heterojunction solar cells doped with a phosphorescent molecule. Appl. Phys. Lett. 95, 173304 (2009).

110. Forrest, S. R. Excitons and the lifetime of organic semiconductor devices. Philos. Trans. R. Soc. A Math. Phys. Eng. Sci. 373, 20140320–20140320 (2015).

111. Adams, J. et al. Water ingress in encapsulated inverted organic solar cells: Correlating infrared imaging and photovoltaic performance. Adv. Energy Mater. 5, 1–11 (2015).

131

Appendix A: Boundary Condition Validation

Figure A-1: a) The LUMO and HOMO levels (top and bottom of boxes) of the organic layers for devices used to determine efficiency of exciton quenching of the rubrene sensing layer.

To determine if the condition of complete exciton quenching at the exciton sensing layer is satisfied, we fabricated devices with the structure shown in Figure A-1 using a CBP: Rubrene (2wt%) or a CBP: Rubrene (8wt%) sensing layers at d=10nm and d=20nm. The emission from these OLEDs at 10mA/cm2 and 80mA/cm2 are shown in Figure 9-S2. The ratio of sensing layer emission between the sensors at 10nm and 20nm does not vary significantly between sensors with 2wt% and 8wt% dopant concentrations at either 10mA/cm2 or 80mA/cm2. This indicates that there is no increase in the quenching ability for dopant concentration greater than 2wt%, suggesting that this concentration is sufficient to maintain complete quenching.

Figure A-2: OLED emission intensity at 10mA/cm2 and 80mA/cm2 and sensing layers with either 2wt% or 8wt% CBP:rubrene.

133

Appendix B: Supplementary Information for Chapter 5

1 Controlling for layer-to-layer variations

To ensure that the variation in defect concentration only occurs between the sets of devices and not between the layers of a given set, we fabricate two control devices with the structure: ITO/MoO3 (1nm)/ CBP (45nm)/ CBP (10nm)/ TPBi (70nm)/ LiF (1nm)/ Al (100nm). All layers of the two devices are shared except for the 10nm CBP layer which was deposited individually for each of the two devices. The emission spectrum of the two devices is shown in Figure B-1. The devices do not show significant variations, confirming that layer-to-layer variations in defect concentration are not significant within a set of devices fabricated on a single substrate.

0.8

0.6

0.4

Signal Intensity [au] Intensity Signal 0.2

0 350 400 450 500 550 600 650 Wavelength [nm]

Figure B-1: Normalized emission spectra of two nominally identical devices aged for 9s at 25mA/cm2. Devices have the structure ITO/MoO3 (1nm)/ CBP (45nm)/ CBP (10nm)/ TPBi (70nm)/ LiF (1nm)/ Al (100nm) where 10nm CBP layer has been deposited in two separate depositions. Data has been smoothed for clarity.

2 Correcting for operational degradation

Rapid luminance degradation of the OLEDs needs to be taken into account. We are interested in the effect of defects which exist in the OLEDs prior to any operational degradation and so we extrapolate the emission intensities back to t=0s to remove the effects of degradation. Discussion of emission intensities in the main text refer to these extrapolated values.

Figure B-2 shows the operational degradation for 5 devices from Set A, with d=2nm, 4nm, 6nm, 8nm, and 10nm (circles) and fits (solid lines) to the data using the empirical formula:

= (B-1) �1 푐 퐼 1+�2� 3 where I is the CBP emission intensity, t is the time and c1, c2, and c3 are fit parameters.

Figure B-2: CBP degradation for OLEDs from Set A with d=2nm, 4nm, 6nm, 8nm, and 10nm (circles) and fits using Equation B-1 (solid line). Emission intensity increases monotonically with d.

135

The defect to CBP emission ratio (ID/I) is also extrapolated to t=0s using the first 6 observed

values. The values of ID/I for 5 devices from Set A, with d=2nm, 4nm, 6nm, 8nm, and 10nm (circles) and linear fits (solid lines) are provided in Figure B-3.

Figure B-3: Ratio of defect to CBP emission intensity as a function of time for OLEDs from Set A with d=2nm, 4nm, 6nm, 8nm, and 10nm (circles) and linear fits (solid line). The ratio of intensities decreases monotonically with d.

136

Appendix C: Equation Derivations

1 Derivation of Equations 6-5 to 6-7 and 6-9

The equation for the CBP exciton population:

0 = 2 (C-1) 휕 � � 2 퐷 휕푥 − 휏 The solution to C-1 is:

( ) = + 푥 푥 (C-2) 퐿 −퐿 푛 � 푐1푒 푐2푒 The first boundary condition on C-1 is:

( ) = 0 (C-3)

푛 ∞ Applying this condition to C-2 yields:

( ) = 푥 (C-4) −퐿 푛 � 푐2푒 The second boundary condition is:

= (0) (C-5) 휕� 퐽 퐸� − 퐷 휕��푥=0 4푒 − 푛 푘 푁 Applying this condition to C-4:

( ) = 퐽� 푥 (C-6) 퐷� −퐿 푘퐸�퐿 푛 � 1+ 퐷 푁 푒 The defect formation rate is given by:

= ( = 0) (C-7) 푑� 푑� 푘퐹푛 � Substituting C-6 into C-7:

= 퐽� (C-8) 푑� 퐷� 푓 푘퐸�퐿 푑� 푘 1+ 퐷 푁 C-8 is a separable ordinary differential equation which can be solved analytically for N:

= 1 + 1 2 (C-9) � 2퐽��푘퐸�휏 2 푁 푘퐸�휏 �� 푒� 푡 − �

The definition of the diffusion length ( = ) has been used to simplify the above expression. 퐷 퐿 �휏 The CBP emission intensity is proportional to the integrated exciton concentration:

( ) = 퐽퐽 (C-10) ∞ 푒 푘 𝜏 0 퐸� 퐼 ∝ ∫ 푛 � 푑� 1+ 퐿 Substituting C-9 into C-10, introducing a proportionality constant, and grouping constants:

= (C-11) 퐼 1 퐼0 3푡 �1+푡50 where:

= 2 (C-12) 3푒� 2 푡50 2퐽푘퐸�휏 푘퐹 The concentration of excitons captured by defects is given by:

0 = + (0) (C-13) �푁 − 휏푁 푘퐸�푛 푁 − 푘�푛푁푁 Substituting C-6 and C-9 and C-12 into C-13:

138

1 − = 퐽�푁 푡 2 (C-14) 푒 �1−�1+3푡50� � 푘 � 퐿 푛푁 푄 푁 푡 1+ 푘퐸�� 1+3푡50−1� The defect emission intensity is proportional to the defect exciton concentration. Grouping constants in C-14:

1 − 푡 2 (C-15) 퐴�1−�1+3푡50� � 퐼푁 ∝ 푡 1+퐵��1+3푡50−1�

2 Derivation of Equations 6-11 and 6-12

With exciton capturing layers, C-1 must be solved with the boundary condition given by C-5 and the following new boundary condition, which replaces C-3:

( ) = 0 (C-16)

푛 푑 Substituting C-2 into C-5 and C-16, and solving the resulting system of equations:

= 퐽� (C-17) 푒� 1 푘 푁� 2푑 푘 푁� 퐸� 퐿 퐸� 푐 퐷 −1−푒 �1+ 퐷 �

= 퐽� (C-18) 푒� 2 푘 푁� −2푑 푘 푁� 퐸� 퐿 퐸� 푐 퐷 +1+푒 �1− 퐷 �

푥 푥 ( , ) = + − (C-19) 퐽� 푒퐿 푒 퐿 푘 푁� 2푑 푘 푁� 푘 푁� −2푑 푘 푁� 푒� 퐸� 퐿 퐸� 퐸� 퐿 퐸� 푛 � 푑 � 퐷 −1−푒 �1+ 퐷 � 퐷 +1+푒 �1− 퐷 �� The CBP emission intensity is given by:

( ) ( , ) (C-20) 푑 퐼 푑 ∝ ∫0 푛 � 푑 푑� Substituting C-19 into C-20:

139

푑 푑 ( ) + − (C-21) 푒퐿−1 1−푒 퐿 푘 푁� 2푑 푘 푁� 푘 푁� −2푑 푘 푁� 퐸� 퐿 퐸� 퐸� 퐿 퐸� 퐼 푑 ∝ � 퐷 −1−푒 �1+ 퐷 � 퐷 +1+푒 �1− 퐷 �� Rearranging C-21:

( ) 푑 (C-22) 1−sech�퐿� 푘퐸�푁� 푑 퐼 푑 ∝ 1+ 퐷 tanh�퐿� The interface exciton density is:

( = 0, ) = + (C-23) 퐽� 1 1 푘 푁� 2푑 푘 푁� 푘 푁� −2푑 푘 푁� 푒� 퐸� 퐿 퐸� 퐸� 퐿 퐸� 푛 � 푑 � 퐷 −1−푒 �1+ 퐷 � 퐷 +1+푒 �1− 퐷 �� Rearranging C-23:

( = 0, ) = 푑 (C-24) 퐽� tanh�퐿� 푘퐸�푁� 푑 푛 � 푑 푒� �1+ 퐷 tanh�퐿�� Substituting C-24 into C-7:

= 푑 (C-25) 푑� 퐽� tanh�퐿� 퐹 푘퐸�푁� 푑 푑� 푘 푒� �1+ 퐷 tanh�퐿�� C-25 is a separable ordinary differential equation which can be solved analytically for N:

= 1 + 1 (C-26) � 3� 푑 50 푁 푘퐸�휏 tanh�퐿� �� − � where:

= 2 (C-27) 3푒� 50 2 2 푑 푡 2퐽푘퐸�휏 푘퐹 tanh �퐿� Substituting C-26 into C-22:

( , ) = (C-28) ( ) 퐼 푑 � 1 퐼0 푑 3푡 �1+푡50 140

C-28 is identical to C-11, but uses a new definition of t50 given by C-27.

3 Derivation of Equation 7-3

Beginning from Equations C-29 and C-30 which correspond to Equations 7-1 and 7-2 from the main text and using the same variable definitions as the main text:

= (C-29) 푑� 푑� 푘퐹푛 0 = (C-30) 퐽� � 푒푒 − 휏 − 푘퐸�푛� C-29 can be re-written as:

= (C-31) 1 푑� 푛 푘퐹 푑� Substitute C-29 into C-30:

0 = (C-32) 퐽� 1 푑� 푘퐸�푁 푑� 푒푒 − 푘퐹휏 푑� − 푘퐹 푑� Rearrange C-32:

= 퐽퐽푘퐹� (C-33) 푑� 푒푒 푑� 1+휏�퐸�푁 This first order ordinary differential Equation for N has the solution:

= 1 + 2 1 (C-34) 1 퐽𝐽 푁 푘퐸�휏 �� 푘퐸�휏�퐹 푒푒 푡 − � Substitute C-34 into C-33:

= 퐽퐽퐽 (C-35) 푒푒 퐽퐽퐽 푛 �1+2푘퐸�휏�퐹 푒푒 � Assuming no spectral drift, the luminance is proportional to exciton density for a given device:

141

( ) ( ) (C-36)

퐿 푡 ∝ 푛 푡 Using C-36, SC-35 can be re-written as:

= (C-37) � 1 �0 푡 �1+3푡50

where t50 is defined by:

= (C-38) 3푒푒 2 푡50 퐽�2푘퐸�휏 푘퐹 C-37 is Equation 7-3.

142

Appendix D: OLED External Degradation

To ensure environmental degradation is not a factor in our experiments, bilayer OLEDs with the same structure as those in the main text were fabricated using a custom thermal deposition system. Half of the devices were encapsulated in a nitrogen environment prior to atmospheric exposure. The encapsulated and unencapsulated devices show identical degradation behavior and thus we conclude that environmental degradation is not a factor over the short time scales in which we are interested.

Figure D-1: Degradation of encapsulated and unencapsulated OLEDs aged at constant current density (25 mA/cm2).

Appendix E: Supplementary Information for Chapter 6

1 Details on Data Extracted from Literature

Data from Fig. 2 are the monomer emission intensities at the t=0 peak monomer emission wavelength. The Figure number and label of extracted data for the OLEDs in Fig. 3 are summarized in Table E-1 below

Table E-1: Details on extracted data from Figure 6-3

Reference Figure Data Label

Giebink et al. 3 1000cd/m2

Heil et al. 6b None

Kawamura et al. 5 w/o EEL

Fukagawa et al. 2c, inset CBP

Sandanayaka et al. 2 L/L0 of OLED

Pang et al. 2 J=10mA/cm2

2 Fit Metrics for Literature Data

The root-mean-squred-error (RMSE) and adjusted R2 values for the fits in Fig 2 and Fig 3 are provided in Tables E-2 and Table E-3.

Table E-2: Fit metrics for fits to Equation 7-3 in Figure 7-1

Sample RMSE Adjusted R2

TBADN 0.029 0.9891

Spiro-NPB 0.0356 0.8593

CBP 0.0219 0.9946

Spiro-CBP 0.025 0.9919 mcp 0.0217 0.9951

TAPC 0.0345 0.9839

TCTA 0.0427 0.9618

NPB 0.0911 0.7851

Table E-3: Fit metrics for fits to Equation 7-3 in Figure 7-2

Reference RMSE Adjusted R2

Giebink et al. 0.0192 0.993

Heil et al. 0.0079 0.987

Kawamura et al. 0.0067 0.992

Fukagawa et al. 0.0114 0.995

Sandanayaka et al. 0.0124 0.992

Pang et al. 0.0083 0.977

145

Appendix F: Environmental Contaminant Diffusion Simulation

In several of the devices investigated, the boundary between bright and dark regions showed significant curvature. We attribute this curvature to a lower diffusivity of environmental contaminants through the bulk organic layers than at the organic/Ag interface. This leads to a concentration gradient of contaminants at the right and left electrode edges which drives contaminants horizontally. This process results in the curved profiles observed in some devices.

This phenomenon can be reproduced in simulations as shown in Figure F-1. The top of the figure is held at a constant concentration of contaminants, representing the edge of the encapsulation. The diffusivity, D, is set to D=1 in the left and right regions, while the center region has D=10. The solid lines in Figure F-1 represent contours of constant concentration after running the simulation for an arbitrary period of time, and demonstrate a curved degradation profile, as we observed.

146

Figure F-1: a) simulation of contaminant diffusion with varying diffusivity on and off of the Ag cathode. b) photograph of EL from a Nc/Pc OPV 4 days after outdoor testing ended, demonstrating curved boundary between bright and dark EL.

The consistency of the curved diffusion front in Figure F-1a and the observations in Figure F-1b suggest that fast diffusion occurs in regions where an Ag cathode has been deposited. This could be due to enhanced diffusion rates along the organic/Ag interface, or could be due to elevated temperatures in cathode regions during outdoor testing.

We note that previous studies have also reported devices with similarly curved degradation profiles, although no explanation was provided at the time111.

147