116
Novel Polymer Thin Films for Application in Flexible Organic Light-emitting Diodes Kyung Min Lee A Dissertation Presented to the Faculty of Princeton University in Candidacy for the Degree of Doctor of Philosophy Recommended for Acceptance by the Department of Electrical Engineering Adviser: Professor Barry Rand September 2019

Novel Polymer Thin Films for Application in Flexible Organic ......Novel Polymer Thin Films for Application in Flexible Organic Light-emitting Diodes Kyung Min Lee A Dissertation Presented

  • Upload
    others

  • View
    4

  • Download
    0

Embed Size (px)

Citation preview

  • Novel Polymer Thin Films for

    Application in Flexible Organic

    Light-emitting Diodes

    Kyung Min Lee

    A Dissertation

    Presented to the Faculty

    of Princeton University

    in Candidacy for the Degree

    of Doctor of Philosophy

    Recommended for Acceptance

    by the Department of

    Electrical Engineering

    Adviser: Professor Barry Rand

    September 2019

  • © Copyright by Kyung Min Lee, 2019.

    All rights reserved.

  • Abstract

    Organic light-emitting diodes (OLEDs) offer potentially cost effective and energy effi-

    cient alternatives to their inorganic counterparts. Superior mechanical flexibility and

    relaxed processing conditions have allowed OLEDs to become widespread in displays

    and mobile devices. Next generation OLEDs are emerging, which include flexible

    white OLED lighting and biodegradable electronics. These applications have practi-

    cal implications for reducing energy consumption and electronic waste as electronics

    have become ubiquitous.

    In this thesis, we explore two emerging areas of research in which OLEDs can

    be used to meet the increasing energy demands. First, we demonstrate a practical

    method to increase luminous efficacy of white OLEDs. This involves developing a

    process that spontaneously forms micro- and nano- pores in an optically transparent

    and high-index plastic. The porous polymer films are optimized to enable broadband

    outcoupling from white OLEDs.

    Second, we introduce the properties of low ceiling temperature polymers and their

    potential as plastics with high recyclability. By incorporating photocatalysts in the

    polymer and integrating an OLED directly atop, we demonstrate a proof-of-concept

    device that can realize room-temperature depolymerization reactions.

    iii

  • Acknowledgements

    Wow! I cannot believe this day has finally come. Last five years were truly interesting

    and time well spent. I got to make things glow, developed really good night vision

    running experiments in the dark, and learned some valuable life lessons. And this is

    also where I get to thank everyone who made it possible. First and foremost, this

    work would not be possible without unconditional love and support from my family. I

    would also like to thank my advisor Prof. Barry Rand and my mentors Profs. Claire

    Gmachl, and Craig Arnold, and Antoine Kahn for your support.

    I had the greatest privilege to work closely with many talented scientists. I want

    to say special thanks to Dr. Tae–wook Koh for being an amazing OLED guru and an

    amazing mentor. Dr. Romain Fardel taught me how to zap things with lasers and

    good lab practices. Many thanks to Dr. Nakita Noel, Dr. Lisa Lin, Dr. Kwangdong

    Roh, Zhuozhi Yao, and Amanda Du; your scientific expertise and unwavering enthu-

    siasm are truly inspiring. Andrew Ma and MC Otani, you two were the best mentees

    I could ask for, and you brought so much fresh perspective.

    I would also like to give special thanks to the members of Cool Kidz and AT&T

    Family, Taylor, Gabi, and Piper for their companionship. Thank you Dr. Nan Yao,

    John Schreiber, Colleen Conrad, Roelie Abdi, Jess Johnson, Lidia Stokeman, and

    Linda Dreher for your technical and administrative support. Finally, this work

    received funding from the DOE EERE SSL Program and DARPA Young Faculty

    Award.

    iv

  • To my family.

    v

  • Contents

    Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii

    Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv

    List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii

    List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix

    1 Introduction to organic light-emitting diodes and current challenges 1

    1.1 Background on organic light-emitting diodes . . . . . . . . . . . . . . 1

    1.2 Light outcoupling in OLEDs . . . . . . . . . . . . . . . . . . . . . . . 7

    1.2.1 Analysis of OLED efficiency . . . . . . . . . . . . . . . . . . . 13

    1.2.2 Dipole radiation near interfaces . . . . . . . . . . . . . . . . . 14

    1.2.3 Previous work on OLED outcoupling . . . . . . . . . . . . . . 18

    1.3 Opportunities for lighting and transient electronics . . . . . . . . . . 20

    2 Experimental determination of OLED efficiency and comparison

    with an electromagnetic model 25

    2.1 Measurement and characterization of OLEDs . . . . . . . . . . . . . . 25

    2.1.1 Comparison of model with experiment . . . . . . . . . . . . . 31

    2.2 Increasing outcoupling efficiency via substrate modification . . . . . . 37

    2.2.1 Preparation and characterization of porous scattering films . . 38

    3 Substrate light outcoupling in flexible white OLEDs 47

    vi

  • 3.1 Colorless polyimide-silver nanowire composite as a flexible substrate

    for white OLEDs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

    3.2 Experimental details . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

    3.3 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 57

    3.3.1 Characterization of scattering polyimide film . . . . . . . . . . 57

    3.3.2 Outcoupling enhancement in green and white PHOLEDs on

    scattering substrates . . . . . . . . . . . . . . . . . . . . . . . 58

    4 Application of OLEDs in transient electronics 67

    4.1 Background on transient electronics . . . . . . . . . . . . . . . . . . . 67

    4.1.1 Low ceiling temperature transient polymers . . . . . . . . . . 69

    4.2 Transient substrate for OLED . . . . . . . . . . . . . . . . . . . . . . 75

    4.2.1 Experimental details . . . . . . . . . . . . . . . . . . . . . . . 90

    5 Conclusion and outlook 93

    A Publications and presentations 98

    A.1 Publications contributing to this thesis . . . . . . . . . . . . . . . . . 98

    A.2 Presentations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

    Bibliography 100

    vii

  • List of Tables

    1.1 Performance of various TCEs in literature. . . . . . . . . . . . . . . . 6

    1.2 Performance of white OLEDs published in a DOE report. . . . . . . . 22

    2.1 Relative powers radiated into each mode. . . . . . . . . . . . . . . . . 34

    2.2 Comparison of simulated outcoupling efficiency and measured EQEs. 40

    4.1 Transient materials and devices reported in literature. . . . . . . . . . 68

    4.2 Ceiling temperatures of various polymers. . . . . . . . . . . . . . . . . 69

    viii

  • List of Figures

    1.1 Molecules and structure used in the first OLED. . . . . . . . . . . . . 2

    1.2 Cross-section and energy diagram of a bottom-emitting OLED. . . . . 3

    1.3 Energy diagram of a guest-host system. . . . . . . . . . . . . . . . . . 4

    1.4 Highest EQEs reported for fluorescent, phosphorescent, and TADF

    emitters by year. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

    1.5 Ray optics representation of losses in a bottom-emitting OLED. . . . 8

    1.6 Dispersion relation for a stratified OLED. . . . . . . . . . . . . . . . 9

    1.7 Propagation of surface plasmon polariton along a metal-dielectric in-

    terface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

    1.8 Patterns of power emitted by a dipole with various orientations. . . . 15

    1.9 Dipole radiation near interfaces. . . . . . . . . . . . . . . . . . . . . . 16

    1.10 Thin-film interference effects in OLEDs. . . . . . . . . . . . . . . . . 17

    1.11 Luminous efficacy of various lighting elements. . . . . . . . . . . . . . 21

    1.12 Photograph of electronic waste. . . . . . . . . . . . . . . . . . . . . . 23

    2.1 Photopic sensitivity function. . . . . . . . . . . . . . . . . . . . . . . 26

    2.2 Measurement geometry for OLEDs in this work. . . . . . . . . . . . . 27

    2.3 Chemical structures of iridium phosphorescent emitters. . . . . . . . . 28

    2.4 The CIE 1931 color map with x, y coordinates of the iridium emitters

    used in this work. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

    2.5 Structure and optical properties of the OLED for simulation. . . . . . 32

    ix

  • 2.6 Simulated power radiated into each wavevector at emission peak. . . . 33

    2.7 Simulated power spectral density plotted over wavelength and in-plane

    wavevector. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

    2.8 Comparison of measured and simulated spectra at two viewing angles. 35

    2.9 Comparison of the measured EQEs of thick-ETL OLEDs and simulated

    outcoupling efficiencies. . . . . . . . . . . . . . . . . . . . . . . . . . . 36

    2.10 Performance of thick-ETL OLEDs. . . . . . . . . . . . . . . . . . . . 36

    2.11 Fabrication of a porous polyimide film by precipitation immersion . . 38

    2.12 Characterization of a pristine Kapton polyimide and a porous Kapton

    polyimide film. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

    2.13 Performance of green OLEDs on glass/ITO with and without a scat-

    tering film. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

    2.14 Angle-dependent performance of green OLEDs on glass/ITO with and

    without a scattering film. . . . . . . . . . . . . . . . . . . . . . . . . . 42

    2.15 Performance of white OLEDs with and without a scattering film. . . 44

    2.16 Angle-dependent performance of white OLEDs on glass/ITO with and

    without a scattering film. . . . . . . . . . . . . . . . . . . . . . . . . . 45

    3.1 Proposed mechanism for extracting substrate loss from flexible OLEDs. 48

    3.2 Outcoupling efficiencies simulated for OLEDs on two different substrates. 49

    3.3 Synthesis of Kapton polyimide. . . . . . . . . . . . . . . . . . . . . . 50

    3.4 Synthesis of colorless polyimide. . . . . . . . . . . . . . . . . . . . . . 51

    3.5 Comparison of Kapton polyimide and colorless polyimide. . . . . . . 52

    3.6 Thermogravimetric analysis of colorless polyimide. . . . . . . . . . . . 53

    3.7 Fabrication scheme for flexible scattering substrates. . . . . . . . . . . 54

    3.8 Device structures of the green and white PHOLEDs. . . . . . . . . . 55

    3.9 Electron microscope images and optical characterizationof a scattering

    substrate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

    x

  • 3.10 Performance of green OLEDs on various substrates. . . . . . . . . . . 60

    3.11 Performance of white OLEDs on various substrates. . . . . . . . . . . 62

    3.12 Angle-dependent spectra of various WOLEDs. . . . . . . . . . . . . . 63

    3.13 Angle-dependent CIE 1931 coordinates of green and white PHOLEDs

    on various substrates. . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

    3.14 Angle-dependent CIE 1931 coordinates and CCT for WOLEDs on var-

    ious substrates. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

    3.15 Normalized outcoupling efficiency over many bending cycles. . . . . . 65

    3.16 Reproducibility of flexible scattering substrates. . . . . . . . . . . . . 66

    4.1 Chemical structures of PHA monomer and cyclic PPHA. . . . . . . . 70

    4.2 Schematic of a low ceiling temperature polymer undergoing cyclic poly-

    merization and depolymerization. . . . . . . . . . . . . . . . . . . . . 71

    4.3 Acid generating mechanism of a Rhodorsil-Faba PAG under UV exci-

    tation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

    4.4 Energy diagram of photo-induced electron transfer between a PAG and

    a sensitizer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

    4.5 Chemical structure and optical properties of BPET sensitizer. . . . . 74

    4.6 Storage modulus of sensitized PPHA before and after photo-exposure. 75

    4.7 Acid generation mechanism via photo-induced electron transfer and

    proposed mechanism for OLED-induced transience. . . . . . . . . . . 76

    4.8 Optical characterization of Yb, Al, and PPHA, and normalized resis-

    tance of AgNW embedded in UV-exposed s-PPHA. . . . . . . . . . . 78

    4.9 Characterization of a green OLED on a pristine PPHA substrate. . . 79

    4.10 Characterization of OLEDs on s-PPHA/glass/ITO and s-PPHA/AgNW

    substrates. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

    4.11 Spectroscopic evidence of sensitizer photobleaching. . . . . . . . . . . 81

    xi

  • 4.12 Photographic evidence of OLED-induced transience and mechanical

    degradation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

    4.13 Evolution of a phototriggered s-PPHA substrate over 72 h. . . . . . . 84

    4.14 Quantification of depolymerization via FTIR. . . . . . . . . . . . . . 84

    4.15 Photographs of OLED-induced depolymerization and OLED lifetimes

    on depolymerizing s-PPHA substrates. . . . . . . . . . . . . . . . . . 86

    4.16 Projected transmission for different sensitizer loadings. . . . . . . . . 87

    4.17 Performance of a polymer LED fabricated on a pristine PPHA substrate. 88

    4.18 Photographs of polymer LEDs on a pristine PPHA substrate and a

    s-PPHA substrate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

    5.1 Photographs of porous polyimide directly formed on polymer sub-

    strates by spin coating. . . . . . . . . . . . . . . . . . . . . . . . . . . 94

    xii

  • Chapter 1

    Introduction to organic

    light-emitting diodes and current

    challenges

    1.1 Background on organic light-emitting diodes

    The 1950 discovery of electroluminescence (EL) in organic materials led to the devel-

    opment of the first organic light-emitting diode (OLED) 27 years later[1]. This first

    OLED consisted of a 75 nm aromatic diamene hole transport layer and a 60 nm elec-

    tron transporting fluorescent emitter, tris-(8-hydroxyquinoline)aluminum (Alq3). The

    two active organic layers were vapor deposited between an indium-tin-oxide (ITO)

    transparent electrode and a Mg:Ag reflective top electrode. This green OLED had

    about 1% external quantum efficiency (EQE), which is still roughly the maximum

    EQE achievable with similarly structured Alq3 OLEDs today. Since the demonstra-

    tion of the first fluorescent OLED, a variety of OLEDs with multilayer structures and

    other fluorescent emitters were demonstrated[2].

    1

  • Figure 1.1: Molecular structures of a) Alq3, and b) diamine hole transport layer,and c) structure of the first OLED; reprinted from [1], with the permission of AIPPublishing.

    Many high-efficiency OLEDs today are fabricated in a multilayer structure for

    better charge balance and exciton confinement, but the basic structure of the OLEDs

    remains unchanged from that of the first OLED (Figure 1.1). The basic structure

    of a bottom-emitting OLED includes a substrate followed by a transparent conduct-

    ing electrode (TCE). Most commonly, an ITO-coated glass serves as a hole-injecting

    transparent contact due to its deep work function as well as its excellent electrical and

    optical properties. A wide bandgap hole transport layer (HTL) selectively transports

    holes into the emissive layer (EML). Electrons are transported by an electron trans-

    port layer (ETL) deposited between the EML and the reflective top contact, cathode.

    We note that the lowest unoccupied molecular orbital (LUMO) levels of most wide

    bandgap ETLs are high (2-3 eV) and therefore, a cathode with a low workfunction

    is needed to achieve a low energetic barrier to inject electrons[3]. Other OLED struc-

    tures include a top-emitting OLED and a transparent OLED, which differ from the

    bottom-emitting OLED by the location of the transparent contact and the number

    of transparent contacts, respectively. The physical structure and the light generation

    mechanism in a bottom-emitting OLED is shown schematically in Figure 1.2.

    In 1993, Kido et al. demonstrated the first white OLED with three active organic

    layers containing red, green, and blue emitters[4]. The fluorescent OLEDs had lim-

    ited EQEs however, due to the spin statistics in which electrically excited organic

    2

  • Figure 1.2: a) Cross-section of a bottom-emitting OLED, and b) energy diagram ofcharge injection and radiative recombination in the OLED. The top and the bottomof the HTL, EML, and ETL represent LUMO and the highest occupied molecularorbital (HOMO) levels.

    molecules form singlet excitons and triplet excitons in a 1:3 ratio. This is because

    excited state organic molecules form bound electron-hole pairs (excitons) localized to

    each molecule, as opposed to free electron-hole pairs that form in inorganic materials.

    Fluorescent emitters with non-emissive triplets therefore can at most achieve 25% in-

    ternal quantum efficiency (IQE) in an electrically driven device, and their EQE is well

    below 5% due to the low outcoupling efficiency (∼20%) of planar OLEDs. The lim-

    ited efficiency of fluorescent OLEDs was first lifted by Baldo et al. in 1998 by use of a

    phosphorescent red emitter[5]. The phosphorescent OLED (PHOLED) consisted of a

    platinum organometallic compound, 2,3,7,8,12,13,17,18-octaethyl-21H,23H-porphyrin

    platinum(II) (PtOEP) in which the heavy platinum atom facilitates intersystem cross-

    ing, allowing the emitter to phosphoresce from its triplet state. By doping a small

    concentration of PtOEP (10%) into a host material with larger singlet and triplet en-

    ergies, the energy transfer (resonant Förster and Dexter) from the excited-state host

    to the dopant becomes energetically favorable and can result in up to 100% capture

    rate of the triplet excitons by the dopant. Most PHOLEDs today utilize this guest-

    3

  • host system. The mechanism governing the radiative recombination in PHOLEDs is

    illustrated in Figure 1.3.

    Figure 1.3: The Jablonski diagram of a guest-host system. Triplet excitons form onthe guest dopant via energy transfers from the host where both singlets and tripletsare first excited.

    Various phosphorescent emitters derived from iridium and platinum organometal-

    lic compounds are among the highest-performing OLED emitters today, with EQEs

    exceeding 38%[6]. These emitters contain rare-earth metals in which the heavy metal

    atom increases spin–orbital interactions and facilitates intersystem crossing[5]. This

    process is also known as the heavy atom effect and allows for the typically none-

    missive triplet excitons–which account for 75% of the initially excited excitons– to

    contribute to electroluminescence. As such, phosphorescent emitters can achieve up

    to 100% internal quantum efficiency (IQE). Despite the high efficiencies achievable

    with the organometallic emitters, there is a large demand to replace them with a more

    abundant material as iridium is one of the rarest metals. Only 2 parts per billion of

    iridium is known to be located in the earth’s crust[7]. A promising alternative to the

    rare-earth compounds includes all-organic compounds engineered to exhibit a very

    small triplet-to-singlet energy gap[8,9]. This allows for efficient reverse intersystem

    crossing at room temperature, and these compounds are able to convert the nonemis-

    4

  • sive triplets to singlets. The class of such emitters are known as thermally activated

    delayed fluorescence (TADF) emitters, and their efficiencies are fast approaching that

    of the phosphorescent emitters. As TADF emitters are synthesized with all-organic

    constituent molecules, they may provide lower-cost alternatives to the phosphorescent

    emitters containing precious metals. The evolution of the reported EQEs in literature

    by year is shown in Figure 1.4.

    Figure 1.4: Highest EQEs reported in literature by year. Sourced from [10].

    Similarly, there exists a large demand to replace ITO with another TCE due to

    its mechanical brittleness and high concentration of a rare metal indium. In today’s

    literature, the performance of ITO serves as a benchmark; a commercially available

    ITO-coated glass substrate exhibits high visible transmission (>87% at 550 nm) and

    5

  • Substrate/TCE Rs (Ω/2) Transmission at550 nm

    Depositionmethod

    Glass/ITO 10 87% SputteringPET/ITO1 60 84% Sputtering

    PET/AgNW[13] 38 95% Bar coatingPolyimide/AgNW(this work)

    25 85% Spin or spray coat-ing

    PET/CuNW[14] 99.1 86.6% Bar coating

    Table 1.1: Performance of various TCEs.

    low sheet resistance (Rs ∼ 10 Ω/2). However, high temperature (400 °C) anneal-

    ing is necessary to produce highly crystalline ITO films with optimum conductivity

    and visible transmission, making ITO incompatible with plastic substrates with much

    lower thermal tolerance. As a result, ITO-coated plastic substrates such as polyethy-

    lene terephthalate/ITO (PET/ITO) are produced without the aggressive thermal

    annealing step and thus exhibit poorer performance with Rs ∼60 Ω/2 and ∼85%

    transmission.

    Metal nanowires and metal nanotubes have been presented as promising alterna-

    tives to ITO[11,12]. This class of TCEs exhibits superior mechanical flexibility and

    can outperform glass/ITO when optimized. In addition, certain metal nanotubes

    and nanowires are derived from more abundant materials such as copper (Cu) or

    graphene, offering advantages over ITO which relies on precious indium. The com-

    parison of ITO and select TCEs is summarized in Table 1.1. In particular, silver

    nanowires (AgNWs) gained much attention for favorable processing conditions (solu-

    tion deposition in an ambient environment) and superior performance. The AgNW

    electrodes have been combined with a variety of plastic substrates to serve as flexible

    transparent conducting substrates for OLEDs. While silver is more expensive than

    indium, AgNWs are highly compatible with large-area, roll-to-roll processes and are

    generally regarded as a more cost effective flexible TCE compared to ITO.

    1Available from Sigma Aldrich.

    6

  • Highly optimized OLED materials such as TADF emitters and emitters with spe-

    cific dipole orientations are commercially available. Thanks to the high IQE of these

    emitters, the efficiency of optimized OLEDs is now mostly limited by outcoupling, a

    process in which emitted photons can escape the physical structure of the OLED. To

    illustrate this point, a standard bottom-emitting OLED on a glass/ITO substrate is

    shown in Figure 1.2. The outcoupled photons must escape the device past the organic

    core and the substrate, into air. Outcoupling in this structure is impeded by three

    main loss channels. First, the proximity between the emitter molecules and the metal

    cathode gives rise to surface plasmon polaritons (SPPs). Once coupled to SPPs, light

    propagates parallel to the metal-organic interface before being absorbed by the metal.

    Second, the high-index organic core between the reflective metal contact and the low-

    index substrate supports waveguiding, trapping light in the organic core. Finally, the

    light incident on the organic-substrate interface below the critical angle may enter

    the substrate but can still undergo total internal reflection (TIR) at the substrate-air

    interface. This loss due to the substrate-air index mismatch is known as substrate

    loss. Without modifying the planar OLED structure, for every exciton formed on

    the emitter molecule, only about 20-30% result in outcoupled photons[15]. A variety

    of structures have been incorporated in OLEDs to outcouple one or more of the loss

    channels in planar OLEDs. In section 1.2.3, we will discuss some of the relevant work

    in literature.

    1.2 Light outcoupling in OLEDs

    We consider the optical environment of a simplified planar bottom-emitting OLED in

    Figure 1.5 to study outcoupling in similar OLEDs that will be discussed later in this

    thesis. We note that the in-plane wavevector kx is a conserved quantity throughout

    the structure as the structure is invariant in x̂. For the light generated in the emissive

    7

  • Figure 1.5: A ray optics representation of a bottom-emitting OLED and light prop-agating in air (1), substrate (2), organic core (3), and metal-organic interface (4).

    medium with refractive index nem propagating at an angle θem relative to the surface

    normal, the in-plane wavevector kx can be written in terms of the wavevector (k) and

    the emission wavelength (λ).

    kx = |k|sinθ =2π

    λnemsinθem (1.1)

    Only a select set of θem is able to escape the OLED structure, as large θem or light

    incident on an interface at an oblique angle can undergo total internal reflection. Four

    different modes are identified in Figure 1.5 in order of increasing θem: air mode (1),

    substrate mode (2), organic/waveguided mode (3), and SPP mode (4). The critical

    angle at each interface places the upper limit on the value of kx that allows light to

    escape each medium (processes 1-3).

    We plot the dispersion relation for the same structure, ω(kx) in which the propa-

    gating mode in each medium with relative permittivity � satisfies the following rela-

    tionship where c is the speed of light:

    ω =c|k|√�

    =c√�

    √k2x + k

    2z (1.2)

    The linear lines plotted in Figure 1.6 represent the condition, kz = 0, and are

    known as light lines. Above the light line, the wave propagates in a given medium

    8

  • with the form, eikxxeikzze−iωt where both kx and kz are real. Below the light line, kz is

    imaginary, and it indicates that the wave has an exponentially decaying y-component

    (i.e. evanescent). The kx values at which the light lines intersect the light with a

    given frequency, ω (3.1 eV or 400 nm in this example, noted by a horizontal line in

    Figure 1.6) is directly related to the maximum angle θem below which a propagating

    solution is supported in the medium. This corresponds to the critical angle above

    which TIR traps light in the organic stack or the substrate. The regions labeled (1),

    (2), and (3) are air, substrate-trapped, and organic/waveguided mode, respectively.

    Figure 1.6: Dispersion relation for the OLED in Figure 1.2. The horizontal linecorresponds to 400 nm. The lines intersecting the x-axis represent the maximum in-plane wavevector allowed in each numbered region. Note that nsub = 1.5, norg = 1.9.

    The SPP excited at a metal-dielectric interface propagates parallel to the interface;

    this wave involves oscillation of free electrons on the metal surface. As we have alluded

    to earlier, this wave exponentially decays away from the interface. The conditions for

    such surface wave to exist can be found by solving Maxwell’s equations and applying

    9

  • boundary conditions for the given interface. We look for the solution in which a

    transverse magnetic (TM) wave propagates along the interface formed by a metal (�m)

    and a dielectric (�d). The wave propagating in x but confined in z has wavevectors,

    ~kd = kxx̂+ iγdẑ in the dielectric (z > 0), and ~km = kxx̂− iγmẑ in the metal (z < 0).

    Note that γd, γm > 0. We express ~E and ~H for this TM wave in each medium.

    Figure 1.7: Surface plasmon polariton along a metal-dielectric interface

    First, in the dielectric:

    ~Ed =< Ex,d, 0, Ez,d > eikxx+γdz−iωt

    ~Hd =< 0, Hy,d, 0 > eikxx+γdz−iωt

    (1.3)

    And in the metal,

    ~Em =< Ex,m, 0, Ez,m > eikxx−γmz−iωt

    ~Hm =< 0, Hy,m, 0 > eikxx−γmz−iωt

    (1.4)

    10

  • We solve the last curl equation in the source-free Maxwell’s equations (Equation 1.5)

    where i is for the metal or the dielectric.

    ∇ · �i ~Ei = 0

    ∇ · ~Hi = 0

    ∇× ~Ei = µ0∂ ~Hi∂t

    ∇× ~Hi = �i∂ ~Ei∂t

    (1.5)

    At the boundary (z = 0), the tangential components of ~E and ~H, and the normal

    component of ~D = � ~E are continuous.

    Ex,m = Ex,d

    Hy,m = Hy,d

    �mEz,m = �dEz,d

    (1.6)

    Solving the last curl equation in Equation 1.5 and applying Equation 1.6, we arrive

    at the condition for SPP to exist at the interface:

    γm�m

    = −γd�d

    (1.7)

    We have assumed that γd,m are positive real numbers. For a positive real �d, �m

    must be a negative real number, which is possible for an ideal Drude metal with the

    dielectric function below its plasma frequency ωp[16]:

    �m(ω) = 1−ω2pω2

    (1.8)

    11

  • Noting that k2d,m = k2x−γ2d,m, Equations 1.7 and 1.8 can be rewritten for kx as follow:

    kx =ω

    c

    √�d�m�d + �m

    c

    √√√√ �d(1− ω2pω2 )�d + 1−

    ω2pω2

    (1.9)

    Without explicitly solving the above expression for ω(kx), we consider the following

    limits on ω where ωspp =ωp√�d+1

    is the surface plasmon frequency.

    (a) ω → 0 =⇒ kx → ωc√�d

    (b) ω → ωspp =⇒ kx →∞

    The observation indicates that at small ω, the SPP dispersion curve approaches that

    of the light line for the dielectric medium (norg). For a large wavevector, the frequency

    approaches the surface plasmon frequency, ωspp. This observation is consistent with

    the actual dispersion curve shown in Figure 1.6 solved for the interface between alu-

    minum and a dielectric with �d = n2org. The plasma frequency of aluminum was taken

    from literature to be h̄ωp = 14 eV[16]. The part of the solution that lies below the

    light lines indicates that the metal-dielectric interface supports evanescent waves (i.e.

    SPP). We see that for 400 nm light, a coupling to SPP takes place for large kx. It also

    shows that above the bulk plasmon frequency ωp, the metal becomes transparent, and

    hence the solution lies above the light lines. We have assumed an ideal lossless metal

    and a dielectric but real materials can have complex dielectric functions. Accounting

    for the loss leads to a qualitatively similar result but includes attenuation in x.

    It should be noted that SPP propagates with a large in-plane wavevector due to

    its evanescent nature. That is, propagating (or far-field) waves in any of the media in

    the given OLED structure, cannot excite SPP since their kx is always smaller than the

    kx required for SPP. This means that SPP cannot be excited with external sources

    such as lasers or lamps without a coupling mechanism such as a prism. However,

    SPP can be excited in OLEDs because the OLED emitters are located within tens of

    12

  • nanometers away from the metal surface. Due to the proximity between the emitters

    and the metal surface, SPP can be excited via near-field coupling. In this near-field

    regime, the OLED emitters may radiate with arbitrarily large kx and hence can couple

    to SPP. Once excited, the SPP mode is unable to couple to far-field radiation due to

    wavevector mismatch and results in reduced outcoupled light.

    1.2.1 Analysis of OLED efficiency

    We have shown that OLED emitters in a stratified structure radiate power into various

    modes that do not result in far-field radiation in air. The structure inherently traps

    light in the high-index media and prevents outcoupling. This limited outcoupling

    efficiency is the bottleneck for achieving high EQE in today’s OLEDs. The EQE is a

    composite effect of various electrical and optical phenomena in an OLED[5].

    EQE = γ η STηeff ηout (1.10)

    The first term γ represents electrical efficiency, or the efficiency at which injected

    charges form excitons in a single layer, single emitter EML. Typically, a near unity

    electrical efficiency can be achieved by doping transport layers and/or by using charge

    blocking layers as recombination is a bimolecular process requiring balanced injection

    and transport of electrons and holes. In an OLED with multiple EMLs (or emitters)

    to achieve broadband emission spectrum, excitons are distributed among the EMLs.

    In such multi-EML OLEDs, each EML may have a distinct exciton capture rate

    depending on its location in the device stack. Modeling multi-layer, multi-emitter

    OLEDs requires extracting the exciton capture rate of each emitter[17]. The discussion

    in this work assumes a single layer, single emitter OLED with γ = 1. The second

    term, ηST , describes the spin selection rule for singlet and triplet excitons that can

    radiatively recombine. We assume that this value is 25% for fluorescent emitters, and

    13

  • 100% for phosphorescent emitters. The third term, ηeff , is the effective IQE of an

    emitter placed in a stratified medium. This differs from the IQE of the same emitter

    radiating in free space by the Purcell factor, F whose discussion will follow later.

    This is because in a stratified environment such as in an OLED, there is a secondary

    field doing work on the source emitter via reflection from interfaces. Generally, if

    this reflected field in a particular cavity structure has an interference maximum at

    the emitter location, it will increase the emitter decay rate (i.e. resonance). We note

    that the intrinsic radiative decay rate, Γr, is modified by F . The non-radiative decay

    rate is given by Γnr and is assumed to be small for the emitters of our interest. The

    effective IQE is then given by the ratio of the radiative decay rate and the total decay

    rate including non-radiative losses.

    ηeff =F · Γr

    F · Γr + Γnr(1.11)

    For very efficient emitters with vanishingly small non-radiative losses (i.e. Γnr → 0,

    Γr → 1), the effective IQE approaches 1. This makes intuitive sense as an emitter

    with zero non-radiative losses will radiate with 100% efficiency in free space or in a

    cavity, hence the effective IQE is unaffected by F . We assume that the phosphorescent

    emitters that we will use in this work have 100% intrinsic IQE. Then the expression

    for the EQE for a charge balanced phosphorescent OLED with 100% IQE depends

    solely on the outcoupling efficiency of the structure.

    1.2.2 Dipole radiation near interfaces

    In this section, we characterize the OLED emission pattern by studying a dipole emit-

    ter near interfaces. An OLED emitter is an oscillating point charge, ~p = pe−iωtδ(~r) p̂,

    and generates current density, ~J = d~pdt

    = −iω~p. The electric field generated by such

    a point source in free space has been established in many early works[18]. The cross-

    14

  • sections of the power emitted by the dipoles with three possible orientations are shown

    in Figure 1.8. Each dipole radiates most strongly in the direction perpendicular to its

    axis. It is clear why vertically oriented emitters suffer from low outcoupling efficiency,

    as it preferentially emits in the plane.

    Figure 1.8: Power emitted by a a) vertical dipole (TM), b) horizontal dipole (TM),and c) horizontal dipole (TE).

    The total power radiated by a resistive or dissipative dipole is given by the Poynt-

    ing flux or its own electric field doing work on the dipole.

    Prad =1

    2

    ∫S

    Re[ ~E × ~H∗] dS

    = −12

    ∫V

    Re[ ~E · ~J∗] dV(1.12)

    Since our current source is a point or delta source, the second integral in Equation

    1.12 simply reduces to Equation 1.13. The power radiated by a dipole is proportional

    to the electric field at the source location (r0) in the direction of the dipole moment,

    thus doing work on it.

    Prad = −1

    2Re[ | ~J | ~Ep̂(r0)] (1.13)

    For a source near interfaces such as our OLED, secondary fields ( ~Es) upon reflec-

    tion can be added to the total field at the dipole location, ~E(r0) = ~E0(r0) + ~Es(r0)

    where ~E0 is the electric field generated by a dipole in free space. Therefore, calcula-

    15

  • tion of the total field at the emitter location is our main task for solving for the power

    radiated by a dipole in a stratified medium such as the one in Figure 1.9. In short,

    Fresnel reflection and transmission coefficients are computed at each interface, and

    by using the transfer matrix method[19], ~Es can be computed at the dipole location.

    Figure 1.9: Dipole emitter in medium ne with its primary field and secondary fieldsresulting from reflections from neighboring media ne−1 and ne+1.

    The electric field of an oscillating dipole source, ~E, can be written as a Fourier

    integral over the normalized in-plane wavevector, u = kxk

    = sinθem[20]. For interested

    readers, rigorous formulations can be found in literature[20–22]. In this formalism, the

    dipole emission can be represented as a set of plane waves propagating in ẑ, and this

    allows us to characterize it in the manner similar to the approach we took in section

    1.2. Equation 1.13 can be rewritten as an integral of power spectral density per u per

    wavelength, K(λ, u). In this representation, we can rewrite the power emitted into

    each in-plane wavevector u.

    Prad(λ) =

    ∫ ∞0

    uK(λ, u)du (1.14)

    Note that K(λ, u) depends on dipole orientation and polarization since the radiated

    power (Equation 1.13) depends on the direction of the electric field parallel to the

    dipole moment. Following the notations in the work of Furno et al., we decompose K

    into horizontal (“h”) and vertical (“v”) dipoles[23]. Note that horizontal dipoles can

    16

  • assume both TE and TM polarizations.

    KTM,v =3

    4Re[

    u2√1− u2

    (1 + a+TM)(1 + a−TM)

    1− aTM]

    KTM,h =3

    8Re[√

    1− u2 (1− a+TM)(1− a

    −TM)

    1− aTM]

    KTE,h =3

    8Re[

    1√1− u2

    (1 + a+TE)(1 + a−TE)

    1− aTE]

    (1.15)

    Where,

    a+TM,TE = r+TM,TE e

    2ikzd+

    a−TM,TE = r−TM,TE e

    2ikzd−

    aTM,TE = a+TM,TEa

    −TM,TE

    (1.16)

    A close observation at Equation 1.15 reveals that the numerators of the K ′s refer

    to wide-angle interference in which the secondary field interferes with the primary

    field. The denominators refer to Fabry-Perot interference that depends on the phase

    accumulated over multiple reflections by the same beam[24]. This is illustrated in Fig-

    ure 1.10. For a bottom-emitting OLED where only one of the electrodes is strongly

    reflective and the other is transparent, the effect of wide-angle interference is dom-

    inant. This means that the distance between the emitter and the strong reflective

    cathode has a large influence on the microcavity environment in bottom-emitting

    OLEDs.

    Figure 1.10: a) Wide-angle interference, and b) Fabry-Perot interference.

    17

  • For a dipole with an arbitrary orientation in both horizontal and vertical direc-

    tions, the power spectral density is then the sum of the two scaled by the anisotropy

    factor, a. An isotropic emitter has a = 13, and a perfectly horizontal emitter has

    a = 0.

    K = aKv + (1− a)Kh = aKTM,v + (1− a)(KTM,h +KTE,h) (1.17)

    Finally, the outcoupling efficiency can be written as the ratio of the power radiated

    into air and the total power radiated into all in-plane wavevectors. Recall that only

    a small set of kx is within the light line for propgation in air (Figure.1.6). This

    condition can be determined by calculating the critical angle between air and the

    emitting medium (i.e. ucrit = sinθcrit,air =nairnem

    ).

    ηout(λ) =Pair(λ)

    Prad(λ)=

    ∫ ucrit0

    uK(λ, u)du∫∞0uK(λ, u)du

    (1.18)

    The power radiated into each mode (air, substrate, waveguide) can be obtained by

    setting the upper limit on the numerator integral in Equation 1.14 as the maximum

    allowed u, or ninem

    where ni = nair, nsub, norg. The power radiated into SPP corresponds

    to large in-plane wavevectors, u > 1 (evanescent in ẑ). This will be revisited in

    Chapter 2 with an example. We also note that the Purcell factor, F , is defined as

    Prad normalized by the power that would be radiated by the dipole in free space.

    1.2.3 Previous work on OLED outcoupling

    We have identified that the prescence of a microcavity surrounding OLED emitters

    limits efficient outcoupling of photons to air. We also noted that outcoupling ef-

    ficiency is the determining factor of EQE as long as the OLED emitter has very

    small non-radiative losses. As a result, various strategies have been shown to out-

    18

  • couple more light from OLEDs. Some improvement is possible for planar OLEDs by

    modifying the optical environment; this includes the use of low-index ETLs which

    can reduce waveguiding and SPP[25]. Engineering emitter orientations such that the

    emitters grow with a preferentially horizontal orientation has also shown a lot of

    promise[26,27]. Without any modification to the planar structure, these preferentially

    horizontal emitters have higher outcoupling efficiency (∼30%) compared to isotropic

    emitters (∼20%).

    Outcoupling efficiency can be further enhanced by adopting non-planar structures.

    For example, substrate trapped light can be recovered by attaching a hemispherical

    lens index matched to the substrate[28]. In this way, the rounded lens edge can be used

    to reduce the incident angle, thus preventing TIR. However, this is only effective when

    the hemispherical lens is much larger than the OLED making it bulky and impractical.

    Mladenovski et al. demonstrated that even higher outcoupling efficiency is possible if

    the hemispherical lens is combined with a high-index substrate (n ∼1.8, sapphire)[29].

    Compared to the glass+lens OLED, the sapphire+lens OLED is expected to have

    reduced waveguided loss as the index contrast between the organic/ITO stack and

    the substrate is reduced and the substrate is optically incoherent.

    A more elegant solution to extracting substrate loss includes the use of microlens

    arrays attached to the substrate backside. Various arrays of shapes (pyramidal, hemi-

    spherical, hexagonal, etc) and sizes can be created with lithography[30,31]. Microlenses

    are also compatible with flexible substrates unlike macrolenses. Other outcoupling

    mechanisms include modification of substrates by incorporating scattering structures,

    such as nanoparticles[32] or roughness-induced scattering surfaces[33].

    The aforementioned nanostructures extract substrate losses by external modifica-

    tion of the substrates, and therefore these outcoupling nanostructures do not affect

    the electrical performance of the OLEDs. On the other hand, extraction of waveguid-

    ing and SPP tends to be more challenging because texturing thin film organic/TCE

    19

  • stack can modify the optical as well as electrical properties of the stack. The inter-

    nal modes (waveguiding and SPP) are usually extracted by introducing periodic or

    irregular corrugations to the thin-film stack[34,35]. Propagated corrugation is believed

    to increase chances of electrical shorts however, as many OLEDs have active layers

    with thicknesses (typically

  • Commercially available white LED bulbs and WOLED panels currently have luminous

    efficacies of ∼170 lm/W2 and ∼61 lm/W3, respectively. The discussion of luminous

    efficacy will be presented in Chapter 2. If WOLEDs can be produced at lower costs,

    Figure 1.11: Luminous efficacy of lighting elements and projections for LEDs andOLEDs. Source: DOE (2012b, p. 38).

    they have a real potential for offering higher-efficiency and light-weight alternatives

    to current lighting elements. Large area OLED lighting panels published in the 2016

    DOE report[40] have already exceeded 50 lm/W, as shown in Table 1.2. The first

    flexible white OLED panel is now commerically available from OLEDWorks4. These

    flexible WOLEDs are reported to be fabricated on Corning’s 100 µm-thick flexible

    2MAS LEDtube 1500 mm UE 21.5 W 840 T8. Available from Philips.3Lumiblade Bright 3. Available from OLEDWorks.4LumiCurve. Available from OLEDWorks.

    21

  • LG Display N6S OLEDWorksFL300

    Kaneka Corp.

    Panel size 32×32 cm2 5×20 cm2 10×10 cm2Correlated colortemperature

    3000 K, 4000 K 2500 K, 2900 K 3000 K 4000 K

    Luminous effi-cacy

    55-60 lm/W 42-50 lm/W 29-40 lm/W

    Table 1.2: Performance of white OLEDs published in the DOE SSL report in 2016[40].

    glass for substrates and encapsulation. The company claims that their flexible white

    OLEDs offer power efficiencies upwards of 60 lm/W. Despite the continued growth,

    WOLEDs have yet to become mainstream, and more work remains to be done to

    produce higher-efficiency devices via outcoupling while keeping the processing costs

    low. In Chapters 2 and 3, we will discuss our approach to enable a scalable outcoupling

    mechanism in flexible WOLEDs for lighting applications.

    Other emerging applications using OLEDs include wearable devices such as wear-

    able displays[41], medical sensors[42], textiles, and garments. These mechanically flexi-

    ble and thin film devices can conform to surfaces such as skin and be worn comfortably

    without generating appreciable heat.

    The electronics market is changing fast, presenting new and improved technology

    annually, and this turns many functioning devices obsolete. This consumption cy-

    cle now has turned electronic waste (E-waste) into the fastest-growing waste stream;

    most E-waste ends up in landfills[43]. The emergence of inexpensive plastic electronics

    is worrisome as plastic materials have extremely poor recycling efficiency. This is be-

    cause most plastics today cannot be broken down to pristine monomers that can be

    repolymerized into a new product. Typically, a lower-quality recycled material is pro-

    duced by mechanically and thermally treating used plastics as these processes result

    in lower molecular weight products[44]. Thermoplastics such as polyethylene (PE),

    polypropylene (PP), and polyvinyl chloride (PVC) can be melted by heating, and

    22

  • they can be remoulded into another shape. Thermoset plastics such as polyimides,

    polyurethanes, and polyester resin undergo irreversible crosslinking processes, and

    they typically decompose before melting. Thermoset plastics can therefore be reused

    as filler materials after being broken down by grinding or milling. The presence of

    impurities such as additives and contaminants also lower the recycling efficiency. In

    this current recycling landscape, some estimate that less than 10% of plastics get

    recycled[45]. Alternatively, plastics can be incinerated to avoid being landfilled and

    recover industrial gases and oil mixtures that can be used to produce fuels[44]. The

    heat generated as a by-product of incineration may also be used to drive turbines.

    Figure 1.12: Photograph of electronic waste in Fresno in May 2019. Source: photoby Christie Hemm Klok for TIME[43].

    Chemical recycling is believed to be a more effective recycling mechanism as it

    chemically breaks down a polymer into its monomers by performing chemical reac-

    tions at ambient or moderate temperatures. This allows for pristine monomers to

    be recovered without much loss in material properties. As such, novel polymers are

    being synthesized to exhibit efficient depolymerization chemistry that can turn them

    into pure monomers[46–48]. Particularly, low ceiling temperature (Tc) polymers are

    gaining attention because they can readily depolymerize at ambient temperatures

    23

  • with specific triggers such as acid. Therefore, this class of polymers are emerging

    plastic materials with superior recyclability as the recovery of pristine monomers is

    facilitated. In addition to monomer recovery, these polymers are being used as “tran-

    sient” materials that can be reintegrated back to the environment without producing

    solid waste. For example, transience may be achieved by first triggering a low Tc

    polymer to depolymerize into monomers, which can then vaporize. Other modes of

    transience include liquification and dissolution in solvents.

    In Chapter 4, we will discuss how OLEDs can be fabricated on a transient polymer

    substrate. Then, the OLEDs will serve as a triggering mechanism to depolymerize the

    supporting substrate. Our work was the first in literature that realized integration of

    the triggering mechanism (OLED, in this case) and the transient polymer. Such inte-

    grated devices may pave a way to developing compact systems that can be deployed

    to remote environments or biological media where transience of obsolete devices is

    desired.

    24

  • Chapter 2

    Experimental determination of

    OLED efficiency and comparison

    with an electromagnetic model

    2.1 Measurement and characterization of OLEDs

    Here, we introduce several terms and definitions that are used to evaluate perfor-

    mance of OLEDs, LEDs, and other devices that output light to interact with human

    eyes. Instead of using standard radiometric quantities to measure absolute radiant

    energy (in units of Watt, Joule, etc), the human perception of the radiant energy is

    taken into account as a weighting factor. This practice is known as photometry, and

    the conversion between radiometric and photometric quantities is achieved by using

    the photopic sensitivity function as a weighting function. Shown in Figure 2.1, the

    photopic sensitivity function is an approximation of the average human vision that

    describes the relative perceived brightness of visible wavelengths of the same radio-

    metric power (Φrad). This means that 1 W of monochromatic light emitting at 555

    nm will appear twice as bright as 1 W of 510 nm monochromatic light.

    25

  • Figure 2.1: Photopic sensitivity function.

    The measurement geometry for our OLEDs involves an area source OLED and a

    photodiode rotating about the OLED to measure the radiant intensity. The photodi-

    ode rotates about the OLED from polar angles θ =0° to 90°, and we assume that the

    OLED emission is independent of the azimuthal angle, φ. In this setup, the radiant

    flux from a Lambertian source received by the photodiode will decrease as a cosine

    of the polar angle because the projected area of the OLED decreases by the same

    amount. The Lambertian source has uniform radiance however, since radiance (in

    unit of W sr−1 m−2) is defined as the radiant flux emitted by a surface per unit pro-

    jected area per unit solid angle. The expression for radiance in our setup is defined

    in Equation 2.1 where ΦPD(θ) is flux received by the photodiode as a function of

    polar angle, and AOLED is the OLED area. Note that the human perception of the

    Lambertian source is also uniform as a function of angle as our eyes adjust for the

    decreasing projected area with angle.

    Lrad(θ) =ΦPD(θ)

    AOLEDcos(θ)dΩ(2.1)

    26

  • Figure 2.2: a) Measurement geometry for OLEDs in this work, and b) simplifiedgeometry independent of azimuthal angle.

    The radiance of an OLED is often converted to and reported as luminance, a

    photometric quantity in units of lumen (lm) obtained by the following formula where

    ρ(λ) is the normalized spectrum of the OLED. In this way, the wavelengths that have

    greater perceived brightness will result in larger luminance as the OLED spectrum is

    normalized by the photopic responsivity function. In other words, 1 W of monochro-

    matic light with an emission wavelength of 555 nm has 683 lm of power in photopic

    unit.

    Lphotopic(θ) = 683lm

    WLrad(θ)

    ∫ρ(λ, θ)f(λ)dλ (2.2)

    Another quantity that becomes especially relevant for evaluating the performance of

    white OLEDs is correlated color temperature (CCT). Understanding of CCT requires

    discussion of color spaces; in particular, we will discuss the first CIE color space

    (CIE 1931). The CIE 1931 color space was created by the international commission

    of illumination (CIE) in a series of experiments performed on human subjects in

    the 1920s. In short, the human subjects were asked to recreate the colors of single

    wavelengths in the visible spectrum by tuning the intensities of three primary-colored

    laser sources (436 nm, 546 nm, and 700 nm in Smith’s work[49]). The study found

    that most single wavelengths were perceptually equivalent to mixtures of the primary

    colors except for the wavelengths in the 450-530 nm range (bright greens and cyan).

    More details can be found in the original studies[49,50], and here we state some of its

    27

  • findings that will become useful in later chapters. The outer line of the CIE 1931

    color space represents a series of pure (or saturated, or single wavelength) colors,

    and the inside of the shape is populated with spectrally broad colors that can be

    created by mixing more than one colors. The x, y coordinates on the CIE 1931 of

    any spectrum can be determined by using empirically obtained weighting functions,

    known as color matching functions. The use of the weighting functions represents the

    limits of human color perception in which two different spectra can appear the same.

    Figure 2.3: The chemical structures of the iridium phosphorescent emitters used inthis work. The peak PL emission wavelengths are λPL.

    Three iridium emitters (Ir(MDQ)2acac, Ir(ppy)2acac, and FIrpic) are commercially

    obtained and used in the OLEDs in this work. Their molecular structures are shown

    in Figure 2.3. The CIE 1931 x and y coordinates of these iridium phosphorescent

    emitters are plotted in Figure 2.4. The triangle formed by the three coordinates

    represent all possible colors that can be created by mixing the three emitters at the

    vertices. Also plotted in the figure is the Planckian locus, whose coordinates have

    been obtained by taking the spectra of an ideal blackbody radiator at temperature,

    Tc(K). For a near-Planckian source, a correlated color temperature (CCT) can be

    assigned in which the non-Planckian source is perceptually equivalent to the ideal

    28

  • Planckian source[51]. In colloquial terms, the CCT tells a consumer whether a light

    source is “warm” or“cool” white.

    Figure 2.4: The CIE 1931 color map with the x, y coordinates of three iridium emit-ters.

    Returning to our earlier discussion, there are several characterization methods to

    evaluate the performance of an OLED. The most widely used performance metric for

    an OLED include external quantum efficiency and power efficiency (PE). The EQE

    of an OLED is defined as the ratio of outcoupled photons and injected electrons.

    The calculation of this quantity requires electrical characterization (current-voltage

    relationship) as well as measurement of the generated light that can be collected ex-

    ternally from the device (i.e. outcoupled light). The PE is equivalent to the luminous

    efficacy discussed in Chapter 1.

    For a bottom-emitting OLED, the outcoupled light is measured along the hemi-

    sphere surrounding the OLED through the backside of the transparent substrate.

    In our experimental setup, a photodiode and a fiber optic cable are mounted on a

    29

  • goniometer to rotate around an OLED to capture the outcoupled power and the out-

    coupled spectrum, respectively. We assume that the OLED is homogeneous across its

    active area and that its emission characteristic is independent of the azimuthal angle.

    Therefore, only several measurements are taken for angles 0 to 90°.

    A calibrated photodiode exhibits a near-linear relationship between optical power

    impinging on the photodiode and the output current. This relationship is given by the

    photodiode responsivity curve, R(λ). The measured photodiode current represents

    the cumulative response of the photodiode due to a spectrally broad source; this

    quantity alone is independent of emission wavelength. To calculate the contribution

    to the photodiode current from each wavelength present in the emitter spectrum, the

    spectrum of the source ρ(λ) and the photodiode responsivity curve must be taken into

    account. The total power received by the photodiode, accounting for the wavelength-

    dependent photodiode responsivity, is given by:

    ΦPD(θ) =iPD(θ)∫

    ρ(λ, θ)R(λ)dλ(2.3)

    To calculate the total flux of photons impinging on a hemispherical surface en-

    closing our OLED, we need a differential photodiode power per solid angle. The

    relationship can be derived by considering the solid angle that the OLED makes with

    the photodiode. Since the measurement is taken by the photodiode with a finite area,

    the solid angle that the photodiode makes with the OLED is approximatelyApdr2

    . For

    a small photodiode area or a large separation, this is a good approximation for the

    true solid angle, which requires the area on a curved spherical surface. Then the

    differential photodiode power per unit solid angle is given as follows.

    dΦPD(θ, φ) =iPD(θ)∫

    ρ(λ, θ)R(λ)dλ

    r2

    APD(2.4)

    30

  • Then the total power received by a hemisphere can be given by integrating the above

    expression over the hemisphere. The number of photons per solid angle is then given

    by considering the energy of a photon (E = hcλ

    ) and the emission spectrum.

    dn(θ, φ) = dΦPD(θ, φ)

    ∫dλρ(λ, θ)λ

    hc(2.5)

    The expression for EQE in terms of empirical quantities is then given by integrat-

    ing the above expression over all solid angles and dividing by the number of injected

    charges, Ie.

    EQE(%) =

    ∫ 2π0dφ

    ∫ π2

    0sin(θ)dθdn(θ, φ)

    Ie

    × 100 (2.6)

    The expression for EQE can be obtained entirely using radiometric measurements,

    but power efficiency is defined in photometric terms which includes the spectral sen-

    sitivity of the average human vision. The power efficiency of an OLED is defined

    as the total luminance integrated over a hemisphere divided by the electrical power

    consumed by the OLED. It has the following form in the unit of lm/W.

    PE =683 lm

    WAOLED

    ∫ΩLrad(θ)

    ∫f(λ)ρ(λ, θ)dλ

    IV(2.7)

    An ideal monochromatic green OLED emitting at 555 nm can have a luminous efficacy

    or PE of at most 683 lm/W, assuming absolutely zero loss. Similarly, theoretical limits

    for white OLEDs are 260-300 lm/W, depending on their CCT[52].

    2.1.1 Comparison of model with experiment

    The electromagnetic model of the OLED emission in stratified media was discussed

    in Chapter 1. In this section, we apply the model to the device structure in Figure

    2.5. The refractive index measured by ellipsometry is plotted for each thin film

    31

  • in the stack. The small, but nonzero extinction coefficients of the layers were also

    measured by ellipsometry and were accounted for in the simulation although they

    are not presented here. The photoluminescence spectrum of Ir(ppy)3 is used as an

    input to the simulation. The emitter Ir(ppy)3 has been documented to be an isotropic

    Figure 2.5: The structure of the OLED (left), and the refractive index and PL mea-surements (right).

    emitter when thermally evaporated, with ∼100% internal quantum efficiency. The

    emitter orientation is thus accounted for by assigning an anisotropy factor a of 13.

    This allows us to understand selective polarization-dependent coupling to SPP and

    waveguide modes; it should be noted that SPP coupling (u > 1) only exists for TM-

    polarized emission and the two waveguide peaks (0.87< u

  • Figure 2.6: Power radiated into each wavevector at emission peak (510 nm) for threepossible dipole orientations and corresponding polarizations.

    Chapter 1. Here, the light lines are obtained with respect to the emission wavelength.

    λ =2π

    kx

    ninem

    where i = air, sub, org (2.8)

    Two waveguiding modes and strong SPP coupling can be identified for all wavelengths

    of interest. In Table 2.1, the relative powers radiated into the four different modes are

    summarized. The transmission of the substrate trapped light back into the organic

    stack is considered absorbed.

    Some of the simulated results can be compared to the experimental result. The

    OLED with the device structure in Figure 2.5 was fabricated and tested for perfor-

    mance in an experiment. The portion air mode simulated (8.98%) is equivalent to

    the measured EQE for the emitter with near-unity internal quantum efficiency, and

    33

  • Figure 2.7: Natural log of power spectral density, uK, plotted over wavelength andin-plane wavevector, kx.

    air substrate waveguide SPP absorption

    Portionpower (%)

    8.98 47.02 15.01 20.19 8.79

    u 0-0.573 0.573-0.87 0.87-1 1-1.2 N/A

    Table 2.1: Relative powers radiated into each mode.

    we observe that the measured EQE peaks around 8.9% showing good agreement with

    the simulation. In addition, the measured angle-dependent spectrum can be com-

    pared with the simulation. Typically, planar OLEDs exhibit outcoupled spectrum

    that blueshifts with viewing angle; this is because the path length depends on both

    wavelength and viewing angle. In this case, shorter wavelengths are being preferen-

    tially outcoupled at large viewing angles. The measured spectrum at two extreme

    34

  • viewing angles (0° and 80°) are plotted against the simulated spectra, and it can be

    seen that the blueshift of the spectrum with increasing viewing angle can be effectively

    predicted.

    Figure 2.8: Comparison of measured and simulated spectra at two viewing angles.

    The validity of the simulation can be further verified by fabricating a set of OLEDs

    with increasing ETL thickness, x. The structure of the OLEDs fabricated in this

    experiment is shown in Figure.2.9a.

    The simulation (Figure 2.9b) predicts an oscillatory outcoupling efficiency as a func-

    tion of ETL thickness. This trend is confirmed by the measured EQEs at various

    ETL thicknesses.

    It should be noted that the devices with large ETL thicknesses above 100 nm

    are nontrivially realized by doping the ETL with a n-type dopant[3]. The doping

    efficiency is evident in the current density (J) vs. voltage (V ) data in Figure 2.10 in

    which the current density is virtually independent of the ETL thickness. Without the

    doped ETL, thick ETLs will cause the current density to decrease more significantly

    due to increased series resistance associated with the large ETL thickness. This will

    35

  • Figure 2.9: a) Device structure, and b) portion powers simulated to reside in eachmode. The white dots are experimentally measured EQEs at the corresponding ETLthickness.

    Figure 2.10: a) The J − V − L curves, and b) EQE vs J curves of the OLEDs inFigure 2.9a.

    not only prevent efficient and balanced charge injection into the emissive layer but

    also prevent the OLED from achieving high luminance due to limited current through

    the device. However, in our device results, the nearly identical J − V characteristic

    confirms that the increase in the device thickness does not increase series resistance.

    It also allows us to determine that the varying luminance at each ETL thickness is

    the result of an oscillating outcoupling efficiency. This determination is not possible

    36

  • for the devices with thick, undoped ETLs as thicker devices will suffer from hindered

    electron injection and transport, resulting in resistive and poorly charge-balanced

    devices.

    2.2 Increasing outcoupling efficiency via substrate

    modification

    As seen in Chapter 1, substrate loss due to the index mismatch at substrate-air inter-

    face presents a significant challenge in developing high-efficiency OLEDs. Thankfully,

    there are many outcoupling strategies for extracting substrate loss. This typically in-

    volves incorporating scattering particles in the substrate to modify the incident angle

    arriving at the interface with air. In 2015, we presented a scalable approach to ex-

    tract susbtrate loss by introducing a scattering film on the backside of a glass/ITO

    substrate[53]. The scattering polymer comprised of a high-index Kapton® polyimide

    (Dupont, referred to as Kapton hereafter) film with embedded air voids that presents

    a large index contrast, leading to an optically hazy film. The aim of the scattering

    film was twofold. First, it should be fabricated with a low-cost and scalable process

    and set of materials for large area lighting applications. Second, the outcoupling

    enhancement should be broadband and demonstrated for both monochromatic and

    white OLEDs. In this Chapter, we show outcoupling enhancement for rigid glass/ITO

    OLEDs assisted with the porous scattering films. In Chapter 3, we will apply this

    technique to flexible OLEDs.

    37

  • 2.2.1 Preparation and characterization of porous scattering

    films

    Here, we introduce the fabrication method for producing the aforementioned porous

    polymer scattering films. We also carried out optical characterization to assess their

    applicability to broadband WOLEDs. Finally we will present the outcoupling en-

    hancement obtained by applying the porous scattering films to a glass/ITO OLED.

    Figure 2.11: a) Fabrication of a porous polyimide film by phase inversion. b) Dynamicformation of pores in the polyimide film. Reprinted (adapted) with permission from[53]. Copyright (2015) American Chemical Society.

    The porous polyimide films are produced via immersion precipitation in which a

    film cast from the precursor polyamic acid (polymer and solvent) is introduced to a

    coagulation bath containing a nonsolvent. Nonsolvent-induced phase inversion and

    solvent-nonsolvent exchange take place, and the polymer solidifies to a membrane-

    like porous film[54]. The air voids formed spontaneously in this way are randomly

    distributed within the polyimide host, giving rise to broadband haze. In addition,

    Go et al. has shown that the choice of nonsolvent has a significant impact on the

    morphology and thickness of the resulting porous polymer film, and hence its outcou-

    pling efficiency[55]. In this work, we use deionized water as our nonsolvent which is

    both scalable and highly miscible with our choice of solvent (N-methyl-2-pyrrolidone,

    38

  • NMP), thus accelerating the immersion precipitation (or phase inversion) process.

    The schematic of the phase inversion process is shown in Figure 2.11.

    Figure 2.12: (a) Refractive index (n) and extinction coefficient (k) of the Kaptonpolyimide. (b) Transmission (T ) and absorption (A) spectra of an 870 nm thickplain polyimide film and the intrinsic photoluminescence spectra of blue (FIrpic),green (Ir(ppy)3), and red (Ir(dmpq)2(acac)) phosphorescent dopants for the OLEDsemployed in this work. (c) Measured total and diffuse transmission and calculatedhaze spectra for the p-PI scattering layer. (d) Top-down confocal microscopy image(the inset is a photograph of the p-PI layer on a glass substrate), (e) top-down SEMimage, and (f) cross-sectional SEM image of the p-PI scattering layer. Reprinted(adapted) with permission from [53]. Copyright (2015) American Chemical Society.

    The porous polymer (p-PI or pPI) film obtained via immersion precipitation takes

    advantage of the large index contrast between itself (in this case, Kapton with n >1.7)

    39

  • and nonabsorbing air pockets. The refractive index and extinction coefficient of a con-

    trol, pristine Kapton film without pores is measured by ellipsometry and is presented

    in Figure 2.12a. As a result, the porous Kapton film exhibits high haze across the

    full visible spectrum (Figure 2.12c) and appears visibly hazy (Figure 2.12d). The

    microstructures of the non-conductive porous Kapton film can be studied via envi-

    ronmental scanning electron microscopy (ESEM). The top-down microscopy images

    (Figure 2.12e) show circular pores ranging from nanometers to several microns, which

    together with the large index contrast between the Kapton film and voids, accounts

    for the large broadband haze. The cross-sectional ESEM image (Figure 2.12f) of

    the porous Kapton film shows that the circular pores from the top-down view are

    columnar.

    Air/EQE Substrate Waveguide SPP Absorption

    Simulatedpower (%)

    21.1 31.5 15.6 27.2 4.6

    ControlOLED (%)

    18.2 N/A N/A N/A N/A

    OLED+pPI(%)

    30 N/A N/A N/A N/A

    Table 2.2: Portion powers simulated for a control green OLED and measured peakEQEs of the OLEDs with and without a porous Kapton film. Internal modes (waveg-uiding, SPP, and absorption) could not be measured experimentally, thus some valuesare not available (N/A). Reprinted (adapted) with permission from [53]. Copyright(2015) American Chemical Society.

    In addition, we characterized the optical properties of the control Kapton film

    prepared without the immersion precipitation step. The total transmission through

    the control Kapton film as shown in Figure 2.12b is about 90% in the wavelength

    range of the OLED emitter spectra and shows minimal absorption in the visible.

    The absorption is especially negligible for thin films (< 1µm); in this particular

    case, the control Kapton film was 870 nm, which increases to 2 µm when porosity

    is introduced. In Chapter 3, we will address the small but nonzero absorption in

    40

  • the short wavelengths; this will allow us to increase the polyimide thickness without

    drastically increasing the absorption in the shorter wavelengths.

    We applied the OLED model to the control structure without the porous Kapton

    film (Figure 2.13a) to estimate the portion power residing in each mode. Table 2.2

    summarizes the simulation results.

    Figure 2.13: a) Structure of green OLEDs with and without a porous Kapton film,b) current density-luminance-voltage (J − L − V ) characteristics of the OLEDs, c)EQE vs J curves, and d) power efficiency of the OLEDs. Reprinted (adapted) withpermission from [53]. Copyright (2015) American Chemical Society.

    Finally, the outcoupling efficiency of the porous Kapton film was experimentally

    measured by fabricating green OLEDs on glass/ITO substrates with and without the

    scattering film. The application of the porous Kapton film is external to the OLEDs,

    so the J − V curves of the control device and the device with the porous Kapton

    film are identical. The luminance increases significantly for the green OLED with

    41

  • the porous Kapton film, resulting in a 65% enhancement in EQE. The peak EQE of

    the control green OLED is 18.2% which is slightly lower than the simulated 21.1%

    air mode (Table 2.2); this could be due to imperfect electrical efficiency and/or less-

    than-unity internal quantum efficiency of the emitter. The increase of 11.8% in EQE

    upon introducing the porous Kapton scattering film to the control OLED amounts

    for approximatley 44% of the simulated substrate loss (31.5%), indicating that the

    porous Kapton film can extract nearly half of the substrate loss for this particular

    device structure.

    Figure 2.14: a) Angle-dependent emission intensity, and b) angle-dependent spectrafor the green OLEDs with and without a porous scattering layer. Reprinted (adapted)with permission from [53]. Copyright (2015) American Chemical Society.

    Angle-dependent measurements provide additional insights about the OLED cav-

    ity environment. In Figure 2.14a, we show the emission intensity profiles as a function

    of viewing angle for the control green OLED, pPI-assisted green OLED, and a Lam-

    bertian surface. The emission pattern of the control OLED is significantly wider than

    that of a Lambertian surface, which indicates that this particular OLED microcavity

    preferentially outcouples in off-normal directions. With the application of the porous

    polyimide film on the glass substrate backside, the emission profile becomes closer to

    the Lambertian profile. This means that the strongly angle-dependent OLED emis-

    42

  • sion pattern incident on the glass-air interface is scattered by the porous polyimide

    film and redirected to resemble that of an ideal diffuse surface, a Lambertian surface.

    Similarly, the angle-dependent spectra of the control OLED and the pPI-assisted

    OLED are shown in Figure 2.14b. The spectra of the control OLED exhibit flattened

    peaks, as opposed to the spectra of the pPI-assisted OLED which show a prominent

    peak around 510 nm and a shoulder at 550 nm. The pPI-assisted spectra resemble

    that of the photoluminescence spectrum of Ir(ppy)3 in solution (i.e. free space), hence

    free of microcavity effects. In a strong microcavity such as the control OLED, the first

    peak at 510 nm is suppressed, compared to the shoulder at 550 nm. The first peak

    at 510 nm is recovered upon application of the pPI film, as shown in Figure 2.14b

    (right). This means that the control OLED microcavity preferentially traps shorter

    wavelengths in the substrate, and the pPI film is able to outcouple such trapped light

    by scattering.

    To demonstrate the efficacy of the pPI films for broadband applications, we re-

    peated the experiment for white OLEDs. The structure of the WOLEDs is shown in

    Figure 2.15 in which three distinct emissive layers were used to produce broadband

    spectra. Note that the presence of multiple emitters distributed along the device

    stack prevents us from using the model we applied to the green OLED. Without

    the knowledge of the relative exciton capture efficiency by each emitter, the results

    presented here are limited to empirical observations.

    The J−V characteristics of the white OLEDs remain unchanged with and without

    the porous polyimide film; this is consistent with the observation we made with the

    green OLEDs. The J −V −L plot in Figure 2.15b shows increased luminance for the

    WOLED with the scattering film, resulting in 60% enhancement in EQE. The EQE

    of the control WOLED (11.9%) is increased to 19% when the porous polyimide film

    is applied to the glass substrate backside. The PE at 100 cd m−2 was 18 lm W−1 for

    the control WOLED, and 32 lm W−1 for the WOLED with the scattering film (Figure

    43

  • 2.15d). This corresponds to 78% enhancement in PE. We note that, in electrically

    equivalent devices such as the OLEDs with and without the externally applied porous

    polyimide films, the enhancement in PE is expected to be larger than that in EQE.

    This is because the definition of PE takes into account the power (i.e. voltage and

    current) while EQE only accounts for current. As a result, the outcoupling-enhanced

    OLED is able to reach the same luminance at a lower voltage as well as at a lower

    current density, resulting in a larger enhancement in PE.

    Figure 2.15: a) Structure of white OLEDs with and without a porous Kapton film,b) current density-luminance-voltage (J −L− V ) characteristics of the WOLEDs, c)EQE vs J curves, and d) power efficiency of the OLEDs. Reprinted (adapted) withpermission from [53]. Copyright (2015) American Chemical Society.

    Finally, the angle-dependent behavior of the WOLEDs is shown in Figure 2.16.

    The initially broad emission intensity profile as a function of viewing angle becomes

    narrower and closer to the Lambertian pattern with the application of the porous

    44

  • Figure 2.16: a) Angle-dependent emission intensity, and b) angle-dependent spectrafor the white OLEDs with and without a porous scattering layer. Reprinted (adapted)with permission from [53]. Copyright (2015) American Chemical Society.

    polyimide film. As we discussed earlier for the green OLEDs, this result indicates

    that the control WOLED initially has a strongly directional emission pattern due to

    the microcavity effects. The porous polyimide film is able to modify this pattern to a

    Lambertian pattern which is characteristic of a diffuse scattering surface. The angle-

    dependent spectra shown in Figure 2.16b evidence that larger spectral variations are

    exhibited by the control WOLED. This is indicated by the arrows in Figure 2.16b

    (left) that represent spectral narrowing and the enhanced blue and green peaks with

    increasing angle. Such effects are negligible in the angle-dependent spectra of the

    WOLED with the scattering film, which show virtually identical spectra irrespective

    of viewing angle. This result is especially significant for WOLEDs intended for lighting

    applications as ideal lighting elements should appear the same color when viewed from

    various angles.

    We thus far demonstrated an effective scattering mechanism to extract substrate

    loss from rigid glass/ITO substrates. The pPI scattering films were externally applied

    to OLEDs, and therefore it allowed us to study outcoupling enhancement in isolation

    (i.e. without affecting J − V relationship). In addition, both monochromatic and

    45

  • white OLEDs were realized to highlight the broadband scattering efficiency; the EQEs

    of the green and white OLEDs were enhanced by 73% and 60%, respectively. In the

    following chapter, we will discuss how the porous polyimide scattering film can be

    modified for application in flexible OLEDs.

    46

  • Chapter 3

    Substrate light outcoupling in

    flexible white OLEDs

    3.1 Colorless polyimide-silver nanowire composite

    as a flexible substrate for white OLEDs

    In Chapter 2, we introduced an inexpensive and scalable process for extracting sub-

    strate loss by using porous Kapton films. The amount of available substrate loss and

    the portion of the substrate loss extracted were quantified by comparing the simu-

    lated substrate loss and the measured EQEs. Despite the efficacy of the outcoupling

    efficiency of the porous Kapton film, the application was limited to rigid glass/ITO

    substrates, and the weak yet nonzero absorption of the Kapton films in the short

    wavelengths resulted in a prominent yellow hue in the films. In this chapter, we will

    improve on the previous work by integrating the porous scattering films to flexible

    OLEDs and presenting an alternative polyimide with improved visible transmission.

    The resulting substrates should be flexible and more transparent than Kapton, mak-

    ing them particularly useful for flexible white OLEDs. To achieve this goal, the new

    substrate material must be a robust host for a TCE with resistance to scratching,

    47

  • rubbing, or mechanical bending that could take place during fabrication, handling,

    and testing. Second, the new material should have negligible absorption across the

    entire visible spectrum so that it can be cast as a film of substantial thickness for

    handling around a laboratory. Usually this requires a thickness above 20 µm. Be-

    low this thickness, the films are difficult to handle and are not mechanically robust.

    Third, the material should become porous in a scalable manner to serve as a scat-

    tering layer that is index-matched to a nonporous substrate. Figure 3.1 compares

    the substrate light extraction mechanism presented in the previous chapter and the

    proposed mechanism adopted on a flexible plastic platform.

    Figure 3.1: a) A rigid glass/ITO substrate enhanced with a porous Kapton scatteringfilm, and b) proposed flexible scattering substrate with nanowire electrode, substrate,and scattering film. Reprinted from [56].

    The elimination of the ITO electrode allows for improved mechanical flexibility

    and improved processibility. We proposed the structure in Figure 3.1b in which the

    high-index ITO thin film is replaced with metal nanowires, and the low-index glass

    substrate is replaced with a high-index polymer. With this structure, a larger portion

    of the generated light is trapped in the high-index substrate while waveguiding in

    the high-index core is reduced. This is an intuitive result since more reflection at

    the substrate-air interface takes place as the index of the substrate increases. The

    simulation was performed for a green Ir(ppy)3 OLED on a glass/ITO substrate and

    a high index (1.7 at 400 nm) substrate without an electrode as it is difficult to model

    metal nanowires. For larger ETL thicknesses above 50 nm, the portion of substrate

    48

  • loss in the high-index case becomes significantly larger compared to the glass/ITO

    case. This means that we are starting with larger substrate trapped light available

    for extraction. An ideal substrate loss extraction mechanism with 100% outcoupling

    efficiency is expected to result in a higher efficiency device on a high-index substrate

    compared to a low-index glass substrate.

    Figure 3.2: Portion power simulated for a green Ir(ppy)3 OLED as a function ofETL thickness for a) glass/ITO substrate, and b) high-n (1.7) substrate without anelectrode.

    The high-index transparent material we chose is colorless polyimide (CPI) and

    adopted the same immersion precipitation principle to obtain porous scattering films,

    similar to the case for Kapton. Colorless polyimide is an optically transparent mate-

    rial that inherits the excellent thermal, mechanical, and chemical stability of aromatic

    polyimides. Unlike conventional polyimides such as Kapton that absorb strongly in

    the short visible wavelengths, CPIs have improved visible transmission by introducing

    alicyclic[57] or semialicyclic[58] substituents or highly electronegative substituents[59] in

    the synthesis. These substituents shift the absorption band edge of polyimides by in-

    hibiting the formation of charge transfer complexes between the diamine (donor) and

    the dianhydride (acceptor) moieties[60].

    49

  • Figure 3.3: Synthesis and imidization process for Kapton polyimide, starting fromprecursors dianhydride and dianiline moieties (top), followed by polymerization intopolyamic acid (middle), and imidization into polyimide (bottom).

    In Figure 3.3a, the general synthesis for Kapton polyimide is shown. The polymer-

    ization of dianhydride and dianiline moieties form polyamic acid by oxidization of the

    dianiline, and the polyamic acid imidizes to polyimide at high temperatures. Follow-

    ing the procedures in the work of Spechler et al.[61], our homemade colorless polyimide

    is synthesized from polymerizing two precursor moieties, a typical dianhydride accep-

    tor and a fluorinated dianiline donor. The synthesis for colorless polyimide is shown

    in Figure 3.4. The homemade CPI showed visible improvement in the transmission

    and lacks the yellow hue that was especially severe for the films thicker than several

    microns.

    50

  • Figure 3.4: Imidization process of colorless polyimide starting from precursors, a typ-ical dianhydride and a fluorinated donor dianiline (top), followed by polymerizationinto polyamic acid (middle), and imidization into colorless polyimide (bottom).

    Several thin films with thicknesses 10-20 µm were prepared by spin coating and

    imidizing precursors of Kapton and colorless polyimide. The total transmission spec-

    tra through these polyimide films with supporting glass substrates were obtained

    with an integrating sphere coupled to a monochromator. The transmission spectra

    are presented in Figure 3.5a. As the Kapton film thickness is increased from 11.7

    µm to 14.5 µm, the transmission across 400-500 nm decreases significantly. However,

    the CPI film with a larger thickness than the two Kapton films does not suffer from

    severe absorption in the 400-500 nm range, and this can be visually confirmed by ob-

    serving the photographs of the freestanding films of Kapton and colorless polyimide

    (note that the dark rectangular patterns are densely deposited silver nanowires). As

    such, we are able to fabricate thicker CPI films while Kapton films limit the thickness

    before significant absorption takes place. This allows us to use the CPI films as a

    51

  • standalone, flexible substrate with thicknesses upwards of 50 µm for handling in a

    laboratory setting. A typical scattering CPI substrate is comprised of two parts: a

    50 µm thick solid CPI substrate and a ∼8 µm-thick porous CPI film (pCPI). The

    ability to achieve substantial thickness without significant absorption combined with

    a porous scattering film made from the same CPI precursor creates a smooth and

    index-matched interface with minimal reflection loss.

    Figure 3.5: a) Transmission spectra of various polyimide films (glass/Kapton andglass/CPI), b) photograph of a patterned AgNW/Kapton substrate, and c) photo-graph of a patterned AgNW/CPI substrate.

    One of the advantages of polyimides over other plastic materials such as PET and

    polyethylene naphthalate (PEN) is its thermal and mechanical robustness. Our CPI

    substrates are thermally imidized at 360 °C and exhibit excellent resistance against

    solvents and high annealing temperatures. The thermogravimetric analysis (TGA)

    thermal curve of the homemade colorless polyimide shows an onset of weight loss at

    560 °C, showing an excellent potential for high-temperature application. The control

    Kapton polyimide under the same TGA analysis shows the onset temperature of 557

    °C.52

  • Figure 3.6: Thermogravimetric analysis (TGA) thermal curve of colorless polyimide.

    3.2 Experimental details

    Colorless polyimide has proven to be an excellent host material for AgNWs, and

    composite CPI/AgNW substra