Optical and Electronic Processes in Organic Photovoltaic Devices


Material Information

Optical and Electronic Processes in Organic Photovoltaic Devices
Physical Description:
1 online resource (207 p.)
Myers,Jason David
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Materials Science and Engineering
Committee Chair:
Xue, Jiangeng
Committee Members:
So, Franky
Sinnott, Susan B
Douglas, Elliot P
Guo, Jing


Subjects / Keywords:
bifunctionality -- microlens -- opvs -- organics -- photocarriers -- photovoltaics -- simulation
Materials Science and Engineering -- Dissertations, Academic -- UF
Materials Science and Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation


Organic photovoltaic devices (OPVs) have become a promising research field. OPVs have intrinsic advantages over conventional inorganic technologies: they can be produced from inexpensive source materials using high-throughput techniques on a variety of substrates, including glass and flexible plastics. However, organic semiconductors have radically different operation characteristics which present challenges to achieving high performance OPVs. To increase the efficiency of OPVs, knowledge of fundamental operation principles is crucial. Here, the photocurrent behavior of OPVs with different heterojunction architectures was studied using synchronous photocurrent detection. It was revealed that photocurrent is always negative in planar and planar-mixed heterojunction devices as it is dominated by photocarrier diffusion. In mixed layer devices, however, the drift current dominates except at biases where the internal electric field is negligible. At these biases, the diffusion current dominates, exhibiting behavior that is correlated to the optical interference patterns within the device active layer. Further, in an effort to increase OPV performance without redesigning the active layer, soft-lithographically stamped microlens arrays (MLAs) were developed and applied to a variety of devices. MLAs refract and reflect incident light, giving light a longer path length through the active layer compared to a device without a MLA; this increases absorption and photocurrent. The experimentally measured efficiency enhancements range from 10 to 60%, with the bulk of this value coming from increased photocurrent. Additionally, because the enhancement is dependent on the substrate/air interface and not the active layer, MLAs are applicable to all organic material systems. Finally, novel architectures for bifunctional organic optoelectronic devices (BFDs), which can function as either an OPV or an organic light emitting device (OLED), were investigated. Because OPVs and OLEDs have inherently opposing operation principles, BFDs suffer from poor performance. A new architecture was developed to incorporate the phosphorescent emitter platinum octaethylporphine (PtOEP) into a rubrene/C60 bilayer BFD to make more efficient use of injected carriers. While the emission was localized to a PtOEP emitter layer by an electron permeable exciton blocking layer of N, N'-bis(naphthalen-1-yl)-N,N'-bis(phenyl)-benzidine (NPB), total performance was not improved. From these experiments, a new understanding of the material requirements for BFDs was obtained.
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility:
by Jason David Myers.
Thesis (Ph.D.)--University of Florida, 2011.
Adviser: Xue, Jiangeng.
Electronic Access:

Record Information

Source Institution:
Rights Management:
Applicable rights reserved.
lcc - LD1780 2011
System ID:

This item is only available as the following downloads:

Full Text




2 2011 Jason D avid Myers


3 To Anne


4 ACKNOWLEDGMENTS This research and my graduate studies we re only possible because of contributions from many incredible people in my life. First, I extend my gratitude to my PhD advisor, Dr. Jiangeng Xue. Dr. Xue taught the best course I had while I was completing my undergraduate degree at the University of Florida and I was happy to j oin his group. My years working with him have been challenging and fruitful and I feel thank Dr. Susan B. Sinnott for accepting me as an undergraduate research assistant and especially for sending me to the 2005 AVS International Symposium to present my research. Exposure to the wider scientific research community was a major influence on my decision to continue my schooling and obtain a doctorate. Your patience and co ncern for your students are an inspiration. I also acknowledge my other advisory committee members, Drs. Franky So, Elliot Douglas, and Jing Guo f or their time and interest. This research could not have been completed without the help of my current and fo rmer labmates. To the more senior members of the Xue research group: your experience and instruction were invaluable to my research. I owe my initial training to Teng Kuan Tseng, Ying Zheng was a great sounding board for ideas and guidance, and it was a pleasure collaborating with Sang Hyun Eom on my microlens research. I especially thank Bill Hammond, who was a great partner in the trenches as the years went on. To my other group members, Yixing Yang, Weiran Cao, Renjia Zhou, Ed Wrzesniewski John Mudr ick Nate Shewmon, and Matt Ripp e, you have all contributed to my success as a graduate student I specifically acknowledge Weiran for his


5 collaboration in our optical management projects and John for being an apt pupil and shouldering responsibility in m aintaining the lab I also thank the undergraduate students who worked with me, Vincent Cassidy and Erik Klump. My research has been made possible by partial financial support from the National Science Foundation CAREER Fellowship, the Department o f Energ y Solar Energy Technologies Program, the Florida Energy Systems Consortium, and the University of Florida Alumni Fellowship. I also thank Karl Zawoy and the University of Florida Office of Technology Licensing for their financial support and assistance in moving forward with patenting this work My most profound gratitude is extended to my friends and family. This long journey would not have been possible without you all I was grateful to be able to share my experiences as a graduate student with my bro ther, Daniel. I especially thank my parents for their constant support encouragement, and interest in my results and progress. Finally, I thank my wife, Anne. Anne, y ou been there from the time I was a nervous undergrad looking school with my doctorate. This work is dedicated to you.


6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 9 LIST OF FIGURES ................................ ................................ ................................ ........ 10 LIST OF ABBREVIATIONS ................................ ................................ ........................... 15 ABSTRACT ................................ ................................ ................................ ................... 18 CHAPTER 1 INT RODUCTION TO ORGANIC SEMICONDUCTORS ................................ ......... 20 1.1 Overview ................................ ................................ ................................ ........... 20 1.2 Classification of Organic Materials ................................ ................................ .... 21 1.3 Electrical and Excitonic Properties of Organic Semiconductors ........................ 23 1.3.1 Origin of Electronic Structure ................................ ................................ ... 24 1.3.2 Transport Behavior ................................ ................................ .................. 26 1.3.3 Excitons ................................ ................................ ................................ ... 28 1.4 Processing Techniques of Organic Semiconductors ................................ ......... 31 1.4.1 Small Molecule Purification ................................ ................................ ..... 32 1.4.2 Vacuum Thermal Evaporation ................................ ................................ 34 1.4.3 Spin Coating ................................ ................................ ............................ 37 1.4.4 Emerging Techniques ................................ ................................ .............. 39 1.5 Common Organic Based Devices ................................ ................................ ..... 40 1.5.1 Organic Photovoltaics ................................ ................................ .............. 40 1.5.2 Organic Light Emitting Devices ................................ ............................... 42 1.6 Research Scop e ................................ ................................ ............................... 43 2 INTRODUCTION TO ORGANIC PHOTOVOLTAIC DEVICES ............................... 46 2.1 Basic Concepts ................................ ................................ ................................ 46 2.2 Overview and History ................................ ................................ ........................ 50 2.3 Operation Principles ................................ ................................ .......................... 52 2.3.1 Basic Processes ................................ ................................ ...................... 52 2.3.2 Fundamental Limitations ................................ ................................ ......... 56 2.4 Progress in Organic Photovoltaic Device Performance ................................ .... 57 2.4.1 Small Molecule Organic Photovoltaic Devices ................................ ........ 58 2.4.2 Polymer Based Organic Photovoltaic Devices ................................ ........ 61 2.4.3 Tandem Organic Photovoltaic Devices ................................ .................... 64 2.4.4 Optical Management ................................ ................................ ............... 66


7 3 ORGANIC OPTOELECTRONIC DEVICE CHARACTERIZATION ......................... 69 3.1 Chapter Overview ................................ ................................ ............................. 69 3.2. Organic Photovoltaics ................................ ................................ ...................... 69 3.2.1 Calibration, Spectral Mismatch, and Current Voltage Measurement ....... 6 9 3.2.2 Quantum Efficiency ................................ ................................ ................. 75 3.2.3 Synchronous Photocurrent ................................ ................................ ...... 78 3.3 Organic Light Emitting Diodes ................................ ................................ .......... 81 4 OPTICAL SIMULATION OF ORGANIC PHOTOVOLTAIC DEVICES .................... 85 4.1 Monte Carlo Based Ray Optics Modeling ................................ ......................... 85 4.1.1 Basic Implementation Scheme ................................ ................................ 88 4.1.2 Simulation of Optical Structures ................................ .............................. 92 4.1.3 Quantitative Significance and Verification ................................ ............... 96 4.2 T ransfer Matrix Wave Optics ................................ ................................ ............ 99 4.2.1 Concept ................................ ................................ ................................ ... 99 4.2.2 Optical Field Calculation ................................ ................................ ........ 100 4.2.3 Photocurrent Calculation ................................ ................................ ....... 108 4.3 Review of Optical S imulation Techniques ................................ ....................... 110 5 PHOTOCURRENT GENERATION AND TRANSPORT BEHAVIOR IN ORGANIC PHOTOVOLTAIC DEVICES ................................ ............................... 112 5.1 Overview ................................ ................................ ................................ ......... 112 5.2 Effect of Heterojunction Architecture ................................ ............................... 112 5.3 Wavelen gth Dependent Photocurrent Behavior in Mixed Heterojunction Devices ................................ ................................ ................................ .............. 119 5.4 Exciton Dissociation Behavior in Bilayer Organic Photov oltaics ..................... 125 5.5 Review ................................ ................................ ................................ ............ 130 6 OPTICAL MANAGEMENT IN ORGANIC PHOTOVOLTAIC DEVICES ................ 133 6.1 Introduction and Background ................................ ................................ .......... 133 6.2 Microlens Array Fabrication ................................ ................................ ............ 137 6.3 Enhancement Characteristics ................................ ................................ ......... 139 6.4 Optical Field Optimization ................................ ................................ ............... 144 6.4.1 Bilayer Heterojun ction Devices ................................ .............................. 144 6.4.2 Bulk Heterojunction Devices ................................ ................................ .. 150 6.5 Geometric Effects ................................ ................................ ........................... 153 6.6 Ideal Architectures for Enhancement ................................ .............................. 163 6.7 Review ................................ ................................ ................................ ............ 164 7 BIFUNCTIONAL ORGANIC OPTOELECTRONIC DEVICES ............................... 166 7.1 Fundamentals of Organic Bifunctional Devices ................................ .............. 166 7.2 Novel Device Architectures for Phosphorescent Bifunctional Devices ............ 169 7.3 Requirements for Efficient Bifunctional Device Design ................................ ... 178


8 8 CONCLUSIONS AND FUTURE WORK ................................ ............................... 181 8.1 Photocurrent Generation and Transport ................................ ......................... 181 8.2 Optical Management in Organic Photovoltaic Devices ................................ ... 183 8.3 Bifunctional Organic Optoelectronic Devices ................................ .................. 187 8.4 Afterword ................................ ................................ ................................ ........ 189 LIST OF REFERENCES ................................ ................................ ............................. 191 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 207


9 LIST OF TABLES Table page 1 1 Structures and optical properties of the polyacene family. ................................ 26 3 1 Spectral mismatch factor M for various devices ................................ ................. 72 6 1 Performance characteristics for several different device architectures and active layer materials with and without microlens arrays ................................ 144


10 LIST OF FIGURE S Figure page 1 1 Examples of different categories in the spectrum of organic materials. .............. 22 1 2 Molecular structures of several organic small molecules and polymers used in this work. ................................ ................................ ................................ ......... 23 1 3 Diagram of and bonding within an ethane molecule. ................................ .. 24 1 4 Schematic energy level diagram of a discrete organic molecule. ....................... 25 1 5 Schematic representation of diff erent classes of excitons ................................ 29 1 6 Gradient zone sublimation, with colors approximating that of CuPc during high vacuum purification. ................................ ................................ .................... 33 1 7 A representative vacuum thermal evaporation (VTE) system. ............................ 34 1 8 Diagram of shadow mask geometry. ................................ ................................ .. 36 1 9 Stages of spin coating. ................................ ................................ ....................... 38 1 10 Examples of organic photovoltaic devices. ................................ ......................... 41 1 11 Example s of commercial OLED products. ................................ .......................... 43 2 1 Current voltage characteristics of a representative photovoltaic device. ............ 47 2 2 Equivalent photovoltaic device circuit and typical schematic of an organic photovoltaic device. ................................ ................................ ............................ 49 2 3 Basic processes in power generation in a bilayer organic photovolt aic device. .. 53 2 4 Optical absorption spectra for several organic photovoltaic materials. ............... 54 2 5 HOMO and LUMO energy levels for several common OPV materials. ............... 56 2 6 Representations of unoptimized and nanoscale phase segregated bulk heterojunction OPV microstructures with two constituent materials. .................. 59 2 7 Ideal interdigitated heterojunction for organic photovoltaics. .............................. 59 2 8 Typical device structure and optical field plot for a tandem organic photovoltaic device. ................................ ................................ ............................ 66 2 9 Three previous examples of optical enhancement techniques. .......................... 67


11 3 1 Reference solar spectra. ................................ ................................ .................... 70 3 2 Reference AM1.5 Global and simulated Xe arc lamp spectra. ........................... 71 3 3 Arrangement of the solar simulator and associated optical components. ........... 73 3 4 Optical intensity distribution over a calibrated 100 mW/cm 2 white light beam ... 74 3 5 External quantum efficiency characterizati on system based on the ASTM E1021 standard. ................................ ................................ ................................ 76 3 6 External quantum efficiency spectrum and current voltage curve for a rubrene/C 60 (35/25 nm) device. ................................ ................................ .......... 78 3 7 Example photocurrent data. ................................ ................................ ................ 79 3 8 Irradiance spectra of two different white light bias lamps compared to the absorption spectra of three representative OPV materials within the visible region. ................................ ................................ ................................ ................ 80 3 9 OLED current voltage and luminance characterization system. ......................... 82 4 1 Schematic diagram of light propagation via ray optics. ................................ ....... 86 4 2 Simplified flow chart of a Monte Carlo ray optics simulator. ............................... 89 4 3 Typical simulated stack. ................................ ................................ ..................... 90 4 4 Typi cal simulated device structure with a convex microlens array. ..................... 92 4 5 Simulated microlens arrays. ................................ ................................ ............... 93 4 6 Concept and mathematical details to simulate lenses of different contact angles. ................................ ................................ ................................ ................ 94 4 7 Simulated lens arrays with contact angles of 90 and 30 ................................ 95 4 8 Verification of basic simulator functions. ................................ ............................. 97 4 9 Simulated air mass 1.5G solar spectrum and active layers absorption coefficients f or various material systems. ................................ ........................... 98 4 10 Calculated transmission, absorption, and reflection of a multilayer CuPc/C 60 structure. ................................ ................................ ................................ .......... 103 4 11 Transfer m atrix calculated optical fields. ................................ .......................... 106 4 12 Exciton generation plots in two different bilayer CuPc/C 60 devices. .................. 107


12 4 13 Transfer matrix calculated short circuit current in a 20 nm CuPc/ x nm C 60 device. ................................ ................................ ................................ .............. 109 5 1 Three different small molecule device architectures. ................................ ........ 113 5 2 Example photocurrent characteristics for three different device architectures. 116 5 3 Relative contr ibutions of drift and diffusion currents to the net photocurrent at a small forward b ias for two different device architectures. .............................. 117 5 4 Relative contributions of drift and diffusion currents to the net pho tocurrent at a large forward bias for two different device architectures. ............................... 118 5 5 Photocurrent measurements of a 90 nm 1:1 CuPc:C 60 device at various wavelengths. ................................ ................................ ................................ ..... 120 5 6 Measured photocurrent values for two different mixed layer thicknesses at different wavelengths. ................................ ................................ ....................... 121 5 7 V 0 vs. wavelength for three different CuPc:C 60 (1:1) layer thicknesses. ........... 122 5 8 Experimental V 0 vs. wavelength data and transfer matrix calculated charge generation fields for three different CuPc:C 60 (1:1) active layer thicknesses. ... 123 5 9 V 0 vs. wavelength for three different P3HT:PCBM devices. ............................. 125 5 10 Measured photocurrent characteristics for planar heterojunction devices at different wavelengths. ................................ ................................ ....................... 126 5 11 Transfer matrix optical simulations for three different bilayer devices. ............. 127 5 12 Inversion voltage and exciton generation profiles for CuPc/C 60 planar devices with different thicknesses. ................................ ................................ ................ 128 5 13 The drift, diffusion, and net photocurrent for excitons dissociated by either the interface or the electric field. ................................ ................................ ............. 129 6 1 Schematic diagram of light interaction and path length through the active layer in a device with and without a microlens array ................................ ........ 135 6 2 Processing steps in microlens array fabrication. ................................ ............... 138 6 3 Current voltage and external quantum efficiency characteristics of a SubPc/C 60 (12/40 nm) device with and without a microlens array ................... 140 6 4 Current voltage and quantum efficiency characteristics for a high e fficiency PBnDT DTffBT:PCBM OPV ................................ ................................ ............. 14 2


13 6 5 Current voltage characteristics of hybrid PCPDTBT:CdSe polymer:inorganic nanoparticle devices with and without a ZnO optical spacing layer. ................. 143 6 6 Calculated optical fields for SubPc/C 60 (12/60 nm) OPVs at normal (0 ) and 30 incidence. ................................ ................................ ................................ ... 145 6 7 Effect of varied C 60 thickness on J SC and P enhancement for a SubPc/C 60 (12/ y nm) device with and without a microlens array. ................................ ....... 147 6 8 Distributions of light incident angle upon the active layer in SubPc:C 60 (1:4 by weight) films calculated by ray optics simulations. ................................ ............ 148 6 9 Power conversion efficiency P for SubPc/C 60 (12/ y nm) with and without a microlens array ................................ ................................ ................................ 149 6 10 Short circuit current enhancements for P3HT:PCBM bulk heterojunction devices with and without a microlens array. ................................ ..................... 150 6 11 Effect of ZnO optical spacer thickne ss on mixed P3HT:PCBM devices. .......... 152 6 12 Effect of device active area on relative enhancement with SubPc/C 60 (12 nm/60 nm) devices. ................................ ................................ .......................... 154 6 13 Simulation results of different geometric arrangements. ................................ ... 156 6 14 Effect of device spacing on simulated absorption enhancement in a 70 nm thick SubPc:C 60 (1:4) device. ................................ ................................ ............ 158 6 15 Effect of contact angle variations on simulated enhancements in 70 nm thick SubPc:C 60 (1:4) films. ................................ ................................ ....................... 159 6 16 Performance of 1 cm 2 SubPc/C 60 (12/40 nm) devices under 5 mW/cm 2 white light illumination with a va riable incident angle. ................................ ................ 161 7 1 Basic OPVs and OLEDs device architectures, with charge carrier behavior diagrammed. ................................ ................................ ................................ ..... 166 7 2 Auger up conversion process for half gap electroluminescence in rubrene/C 60 BFDs ................................ ................................ ............................ 168 7 3 Effect of an NPB electron blocking layer on a rubrene/C 60 BFD. ...................... 170 7 4 BF Ds using a doped phosphorescent emissive layer. ................................ ...... 172 7 5 Jablonski diagram of exciton energies for a system containing NPB, PtOEP, and rubrene. ................................ ................................ ................................ ..... 173 7 6 Ir(ppy) 3 phosphorescent BFD architecture and emission spectra for devices with and without doping into NPB. ................................ ................................ .... 174


14 7 7 Adjusted phosphorescent BFD architecture. ................................ .................... 175 7 8 Photovoltaic and LJV characteristics of phosphorescent BFDs. ....................... 176 7 9 Light emitting characteristics of phosphorescent BFDs. ................................ ... 177 8 1 Device schematics for concave and convex microlens array rear reflectors. ... 186 8 2 Schematic diagram of localized effects of ferroelectric nanoparticles polarization on the potential barrier for electron injection at a rubrene/C 60 interface. ................................ ................................ ................................ ........... 188


15 LIST OF ABBREVIATIONS Alq3 aluminum tris(8 hydroxyquinoline) AM 1.5G Air M ass 1.5 Global AMU Atomic mass unit ASTM American Society for Testing and Materials BCP Bathocuproine or 2,9 d imethyl 4,7 diphenyl 1,10 phenanthroline BFD Bifunctional device CBP 4,4' N,N' dicarbazole biphenyl CIGS CuIn x Ga (1 x) Se 2 copper indium gallium diselinide ClAlPc Chloroaluminum phthalocyanine ClInPc Chloroindium phthalocyanine CRZ Carrier recombination zone CT Charge transfer CuPc Copper phthalocyanine EQE External quantum efficiency F8TBT poly((9,9 dioctylfluorene) 2,7 diyl alt [4,7 bis(3 hexylthien 5 yl) 2,1,3 benzothiadiazole] diyl) FDTD Finite difference time domain FF Fill factor HJ Heterojunction HOMO Highest occupied molecular orbital IQE Internal quantum efficiency Ir(ppy) 3 fac tris (phenylpyridine) iridium ITO Indium tin oxide


16 LED Light emitting diode LUMO Lowest unoccupied molecular orbital MDMO PPV poly( 2 methoxy 5 (3,7 dimethyloctyloxy) 1,4 phenylene vinylene) MEH PPV poly(2 methoxy 5 ethyl hexyloxy) 1,4 phenylene vinylene) MFP Mean free path MLA Microlens array MO Molecular orbital NPB N, N' bis(naphthalen 1 yl) N,N' bis(phenyl) benzidine OLE D Organ ic light emitting diode OPV Organic photovoltaic OVJP Organic vapor jet printing OVPD Organic vapor phase deposition P3HT poly(3 hexylthiophene) PBC Periodic boundary conditions PBnDT DTffBT poly(benzo[1,2 b:4,5 (5,6 difluoro 4,7 dithi en 2 yl 2,1,3 benzothiadiazole) PbPc Lead phthalocyanine PCBM [6,6] phenyl C 61 butyric acid methyl ester PCDTBT poly[ N 9'' hepta decanyl 2,7 carbazole alt 5,5 (4',7' di 2 thienyl 2',1',3' benzothiadiazole) PCPDTBT p oly[2,6 (4,4 bis [2 ethylhexyl] 4H cyclop enta[2,1 b;3,4 b] dithiophene) alt 4,7 (2,1,3 benzothiadiazole)] PDMS poly(dimethylsiloxane) PEDOT:PSS poly(3, 4 ethylenedioxythiophene):poly(styrenesulfonate) PPV poly(1,4 phenylene vinylene) PTCBI 3,4,9,10 perylene tetracarboxylic bis benzimidazole


17 PtOEP Platinum octaethylporphine QTH Quartz tungsten halogen SnPc Tin phthalocyanine SubPc Boron subphthalocyanine chloride TP yP 5,10,15, 20 tetra(3 pyridyl)porphyrin VMD Visual Molecular Dynamics VTE Vacuum thermal evaporation ZnPc Zinc phthalocyanine


18 Abstrac t of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy OPTICAL AND ELECTRONIC PROCESSES IN ORGANIC PHOTOVOLTAIC DEVICES By Jason Dav id Myers August 2011 Chair: Jiangeng Xue Major: Materials Science and Engineering Organic photovoltaic devices (OPVs) have become a promising research field. OPVs have intrinsic advantages over conventional inorganic technologies : they can be produced from inexpensive source materials using high th roughput techniques on a variety of substrates, including gla ss and f lexible plastics However, o rganic semiconductors have radically diffe rent operation characteristics which present challenges to achieving high performance OPVs To increase the efficiency of OPVs, knowledge of fundamental operation principles is crucial. Here, the photocurrent behavior of OP Vs with different heterojunction architectures was studied using synchronous photocurrent detection It was revealed that photocurrent is always negative in planar and planar m ixed heterojunction devices as it is dominated by photocarrier diffusion. In mixed layer devices, however, the drift current dominates e xcept at biases where the internal electri c fiel d is negligible At these biases, the diffusion current dominates, exhibiting behavior that is correlated to the optic al interference patterns within the device active layer. Further, in an effort to increase OPV performance without redesigning th e active layer, soft lithographically stamped microlens arrays (MLAs) were developed and


19 applied to a variety of devices. MLAs refract and reflect incident light, giving light a longer path length through the active layer compared to a device without a ML A; this increases absorption and photocurrent. The experimentally measured efficiency enhancement s range from 10 to 60%, with the bulk of this value coming from increased photocurrent. Additionally, because the enhancement is dependent on the substrate/a ir interface and not the active layer, MLAs are applicable to all organic material systems Finally, novel architectures for bifunctional organic optoelectronic device s (BFDs), which can function as either an OPV or an organic light emitting device (OLED) were investigated Because OPVs and OLEDs have inherently opposing operation principles, BFDs suffer from poor performance. A new architecture was developed to incorporate the phosphorescent emitter platinum octaethylporphine (PtOEP) in to a rubrene/C 60 bilayer BFD to make more efficient use of injected carriers While the emission was localized to a PtOEP emitter layer by an electron permeable exciton blocking layer of N, N' bis(naphthalen 1 yl) N,N' bis(phenyl) benzidine (NPB), total performance was no t improved. From these experiments, a new understanding of the material requirements for BFDs was obtained.


20 CHAPTER 1 INTRODUCTION TO ORGANIC SEMICONDUCTORS 1.1 Overview This C hapter introduces the reader to the electrical and physical properties of organic semiconductors to better understand their application to optoelectronic devices. As a class, organic semiconductors have fueled much interest in the scientific community for over three decades. They have the potential to revolutionize several a spects of society: ubiqu itous photovoltaic power, transparent displays, efficient and inexpensive solid state white lighting and truly flexi ble and rugged electronics are merely a few examples 1 7 Interest in these materials is driven by their intrinsic advantages over inorganic semiconductors such as Si or GaAs. Organic materials are generally inexpensive and compatible with lar ge area, low cost and low temperature manufa cturing techniques Many of these manufacturing techniques are compatible with high throughput roll to roll processing. Further, while most inorganic devices require high purity crystalline substrates, organic devices can be produced on glass, plastic fi lms, or metal foils without concern of lattice matching or strain induced defect states. However, organic materials suffer from low charge carrier mobilities due to weak intermolecular interactions, which lower their performance relative to inorganic devic es. Additionally, most organic materials are damaged by exposure to oxygen and water vapor, requiring extensive environmental e ncapsulation to achieve acceptable device lifetimes. The purity of organic materials is also much less than inorganics introdu cing electronic defect states that further reduce performance.


21 Section 1.2 covers the basic definition and classification of organic materials. Their electronic nature is discussed in Section 1.3, along with an introduction to excitons. Processing techn iques are covered in Section 1.4, and a brief introduction to different organic electronic devices is in Section 1.5. 1.2 Classification of Organic Materials carbon atoms For the purposes of this research organic materials are restricted to those that have conjugated molecular structures and exhibit semiconducting properties. In strong contrast to conventional inorganic semiconductors based on co v alently bonded silicon or III V, organic materials are loosely bound molecular solids held together by weak van der Waals interactions; this has a profound effect on their electrical properties. Organic materials can be further subdivided into three categories based on comple xity: discrete small molecules, polymers, and biological molecules (Fig ure 1 1 ). The most complex organic molecules known are biological, such as the absorbing chromophore antenna of Rhodopseudomonas acidophila a purple colored photosynthetic bacterium 8 Biological molecules have not been incorporated into organic electronic devices. The current field of organic photovoltaics is ins tead centered on the first two classes of materials, small molecules and polymers. This work has made extensive use of both types of materials.


22 Figure 1 1 Examples of different categories in the spectrum of organic materials. Small molecules are the simplest type of organic solid and, des pite their moniker, can be relatively massive with ty pical masses of several hundred atomic mass units ( AMU ) Regardless of their size, all small molecules are distinct units. The bulk of the organic materials considered in this work fall into this category. Moving up the complexity scale, one arrives at polymers, long chains of repeating units based on a backbone o f carbon carbon bonds. Polymer masses can vary greatly, ranging from tens to thousands of repeat ing units, with masses up to a million AMU. The complexity of semiconducting polymers also varies greatly from relatively simple polythiophenes to intricate donor acceptor complexes. Several examples of organic small molecules and polymers used in this work are shown in Figure 1 2


23 Figure 1 2 Molecular structures of several organic small molecules (top row) and polymers (bottom row) used in this work 1.3 Electrical and Excitonic Properties of Organic Semiconductors As a class, organic semiconductors have very different electrical properties when compared to traditional, inorganic semiconductors. In this section, a brief overvi ew of the electronic structure and charge carrier behavior of an organic semiconductor will be discussed. Additionally, excitons, which couple optical and electronic processes in organics are introduced. Consequently o ptoelectronic devices based on excitonic semiconductors have different operation principles and design requirements compared to those based on traditional inorganic materials An understanding of organic charge transport and exciton formation is therefore crucial in effective device design and optimization.


24 1.3.1 Origin of Electronic Structure All organic semiconducting materials, whether they are small molecules, polymers, or more complex structures, rely on conjugated electron systems for conduction. Systems are considered con jugated when alternating carbon containing single and double bonds are present in their molecular structure. A straightforward example of this system is an ethene molecule C 2 H 4 (Figure 1 3 ) Figure 1 3 Diagram of and bonding within an ethane molecule. Each carbon atom in ethane is sp 2 hy b r idized, with three sp 2 orbitals created per atom and o ne leftover unhybridized p z orbital The six sp 2 b onds within the system ( four C H bonds and one C C ), with t he leftover dumbbell shaped p z orbitals around each carbon atom f o r m ing a C C bond. Due to the shape of the p z orbitals, the C bond has weak interaction due to small electron cloud overlap above and below the molecular plane The strength of the bonds leads to orbitals (MOs). The weaker interactions of the parallel p z


25 molecule. This is schemati cally represented in Figure 1 4 Because of the importance lowest energy option antibonding MO is named t (LUMO) The HOMO and LUMO, respectively, are analogous to the valence and conduction bands in inorganic semiconducting materials. Fig ure 1 4 Schematic energy level diagram of a discrete organic molecule. The e lectronic band gap (HOMO LUMO) is taken as the gap The degree of conjugation within an organic solid has a large impact on its electrical properties. Increased conjugation length causes a greater degree of electron delocalization, increasing the bonding system. Similarly, short conjugation length localizes electrons, reducing their ability to freely move about a system. This is reflected archetypically in the polyacenes, conjugated systems of conjoined benzene r ings (Table 1 1 ) Increased conjugation (more


26 conjoined benzene rings) corresponds with red shifted absorption spectra caused by decreasing HOMO LUMO separation 9 This illustrates a prime strength of organic semiconductors: simple changes to a base molecule can alter its electronic transport and optical properties. Table 1 1. Structures and optical properties of the polyacene family. Molecule Structure Absorption Maximum Benzene 255 nm Napthalene 315 nm Anthracene 380 nm Tetracene 480 nm Pentacene 580 nm 1.3.2 Transport Behavior Charge transport within organic based materials is a combination of two processes: intramolecular carrier movement and intermolecular charge transfer. Within a molecule, conjugation enables charge carriers to move freely. In organic materials, transport is limited by the weak van der Waals intermolecular coupling, drastically


27 lowerin g charge carrier mobility to typical values of 10 5 to 10 2 cm 2 /Vs within the photovoltaic materials of interest 10 12 Because of the weak coupling, charge carriers are strongly localized on individual molecules, preventing continuous band transport. Intermolecular t ransport typically occurs through a hopping process as a charge carrier ov ercomes an energy barrier to move from one molecule to the next. The mobility in this situation is dependent on the energy barrier height, electric field, and temperature according to where k B is the Boltzmann constant, F is the el ectric field, T is the temperature, E A is the energy barrier height, and is a material dependent constant. The situation can change substantially based on the degree of interaction between adjacent molecules. The van der Waals interacti on force can be approximated by where A and B are empirically derived constants, r i s the distance between molecules, and U i s the interaction potential energy. This relationship is known as the Lennard Jones potential and is used widely in molecular dynamics simulations 13 In this relationship, small deviations in r can have large effects on the degree of interaction within the solid, increasing coupling between molecules, decreasing charge carrier localization, and lowering the energy barrier for hopping transport. In a highly ordered molecular crystal charge carriers are sufficiently delocalized that band transport is realized, much like in inorganic s emiconductors. Charge mobility in highly pure crystals of 5 ,6,11,12 tetraphenylnaphthacene (rubrene) has reached values of 1 4 0 cm 2 /Vs 14 17


28 1.3.3 Excitons Bound electron hole pairs, or excitons, are crucial to the operation of optoelectronic organic devices, including organic photovoltaics and organic light emitting devices 9,18 In OPVs, excitons are the byproduct of photon absorption, where an electron is excited to the LUMO level of the molecule and coulombically binds with the hole left behind in the HOMO to slightly lower the total system energy. T he exciton must then be broken b ack into free charge carriers (dissociated) to extrac t power from the device. In organic light emitting devices injected charge carriers form excitons in an emissive layer, which then recombine to emit a photon. In either case, an electron a nd hole are separated by a distance r c based on the coulombic att raction force and dielectric constant of the material The material will form a tightly bound exciton if r c is larger than the Bohr radius r B of the material as in where q is the elementary charge, k B is the Boltzmann constant, m e and m eff are the standard and effective electron masses, and r 0 is the Bohr radius of hydrogen, 0.53 If the ratio > 1, the semiconductor is excitonic; < 1 is a traditional inorganic semiconductor. A semiconductor is also excitonic if the critical radius is larger than the particle itself, as in inorganic quantum dots 19 Based on the spin of the electron and hole, the exciton can eit her be classified as a triplet (total spin = 1) or a singlet (total spin = 0), so named because triplets are created at a 3:1 ratio relative to singlets 20 Direct recombination of a trip let exciton is forbidden by spin conservation giving it a much longer lifetime than a singlet, on th e order of 10 6 s vs. 10 9 s 21,22 E xciton type is seldom considered in organic photovoltaic


29 devices but huge advances in light emission efficiency have resulted by forcing recombination to occur in the lower energy triplet state using phosphorescent emitters giving i nternal quantum efficiencies approaching 1 00% 23 26 Figure 1 5. Schematic representation of different classes of excitons: (a) Frenkel (b) charge transfe r and (c) Wannier Mott, with varying degrees of delocalization indicated. There are three types of excitons that have been observed : Frenkel, charge transfer (CT), and Wannier Mott (Figure 1 5) Frenkel excitons are formed with the electron hole binding distance smaller than a single molecule or (in the case of inorganics) the lattice constant of the crystallographic unit cell. CT excitons occur when the bound carriers are delocalized over adjace nt molecules. The third class, Wannier Mott excitons, are found in inorganic semiconductors, where the large dielectric constant screens the coulombic attraction of the electron and hole and allows them to


30 delocalize over a long distance. Binding energie s of Frenkel and CT excitons are greater than 0.1 eV; Wannier Mott binding energies are only a few meV. At room temperatures, Wannier Mott excitons are dissociated by thermal energy consequently, any excitons formed upon photon absorption are immediately dissociated into free charge carriers. While bound, excitons can move throughout a solid much like other fundamental particles. Because excitons are charge neutral, applied electric fields do not control their motion. Excitons instead diffuse through a material either in a band to band direct energy transfer method (F rster transfer), or in a molecule to molecule hopping process (Dexter transfer). In the former, temporary electric dipoles are fo electrons upon initial excitation which induce sympathetic dipoles in an adjacent the second molecule. This can be thought of a s a photon emi ssion and absorb tion process where the first molecule relaxes to the ground state and simultaneously excites the second molecule, though no photon is actually emitted. This process is highly dependent on both the spectral absorption overlap of the two mo lecules and the distance between them. In Dexter trans fer, an excited electron moves directly to the LUMO of the acceptor molecule and an electron of the opposite spin is transferred from the acceptor to the donor HOMO. F rster transfer occurs at distanc es up to 10 nm 27 while Dexter transfer happens at shorter distances typically 5 10 Much as is the case with charge carrier tran sport, the weak intermolecular interactions in organic solids limit exciton mobilities and diffusion lengths; most excitons in organic solids can diffuse


31 on the order of 10 nm prior to recombination 27 29 though micrometer diffu sion lengths have been observed in highly pure rubrene crystals 30 The large binding energy of Frenkel and CT excitons does not lend itself to easy dissociation. Thermal dissociation is not practical due to the low decomposition temperatures of organic materials. Excitons can be dissociated by application of an electric field, but the field strength must be in excess of 10 6 V/ m, a prohibitively large value. The preferred route to induce exciton dissociation is the introduction of a heterojunction between organic materials with differing electron affinities and ionization potentials. When an exciton encounters such an interface, provided that the offset between the HOMO and LUMO levels is greater than the exciton binding energy, it is energetically favorable for that exciton to dissociate back into free charge carriers This process will be discussed in detail in Section 2.3.1 1.4 Processing Techniques of Organic Semiconductors One of the most distinct differences between inorganic and organic semiconductors is the processing techniques required with each. Inorganic semiconductors are produced on expensive, highly pure crystall ine substrates with high temperature low throughput techniques. In contrast, organic semiconductors can be produced on inexpensive substrates such as glass, plastic, and metal foil, and their molecular nature allows for solution based techniques. Prior to device fabrication, most small molecule materials are purified using multiple rounds of gradient zone sublimation 31,32 The concept and practical application of gradient zone sublimation are discussed in Section 1.4.1.


32 Two film deposition techni ques are used extensively in this research : vacuum thermal evaporation and spin c oating respectively covered in Section s 1.4.2 and 1.4.3 Thin, high quality films can be produced with either of these techniques. There are a variety of other techniques for both polymers and small molecules that are promising for industrial production of organic electronic s. T hese will be briefly highlighted in Section 1.4.4. 1.4.1 Small Molecule Purification As purchased, most small molecule source materials are of insufficient purity for use in electronic devices. Listed material purities are typically in excess of 99% based on residual metal content. H owever, as electronic transport and quenching can be do minated by a small minority of defects and recombination centers, additional purification is required to obtain electronically acceptable source material Materials will typically be purified from one to four times, depending on the quality of the source material. High purity crystalline yields for each purification are 25 75% of the source material mass vary ing greatly based on the source material quality. Materials can either be purified under high vacuum or under an inert gas flow, such as nitrogen or argon. The sublimation procedure occurs within a quartz tube heated by a multi zone furnace; prior to purification, the tube is cleaned under high vacuum at high temperatures (600 700 C) to eliminate all residual organic material. Several grams of source material are then added within a smaller inner quartz sleeve and placed in the hottest zone of the furnace. Two additional quartz collection sleeves are also added to the main tube to serve as high and low purity collection tubes. Finall y, a plug of quartz wool is used to prevent excessive material contamination of the


33 vacuum pumps and a mesh screen prevents stray quartz wool fibers from collecting in the pumping chamber and damaging the high vacuum turbo pump. The system is then sealed and pumped to the desired pressure (< 10 5 Torr for high vacuum, ~100 Torr with 0.1 0.2 scfh gas flow for inert gas flowthrough, moderated by a vacuum switch controlled solenoid valve and rotary vane roughing pump). The furnace temperature is then slow ly increased to the sublimation temperature of the source material (generally 300 400 C) and a temperature gradient of up to 50 C is applied relative to the middle zone. The temperatures are selected such that light impurities and the high purity crystallin e material sublimate in the hottest zone. The middle zone temperature allows the high purity crystalline material to deposit; light impurities deposit at the coolest end of the furnace. A schematic of this process is shown in Figure 1 6 The entire puri fication process can take s everal days, with high pressure flowthrough purifications generally taking less time than high vacuum. After the process is complete, the high purity crystals are collected and added to clean tubes as the source material for an additional purification cycle, if desired. Figure 1 6 Gradient zone sublimation, with colors approximating that of CuPc during high vacuum purification


34 1.4.2 Vacuum Thermal E vaporation Vacuum thermal evaporation ( VTE ) is the preferred method for growing thin films of organic small molecules. Polymers are ill suited to VTE, as they decompo se at temperatures lower than their evaporation temperature. The size and complexity of VTE systems can vary greatly. A simple VTE system is shown in Figure 1 7 consisting urately control film thickness, and a shadow mask to pattern the substrate. Figure 1 7. A representative vacuum thermal evaporation (VTE) system. The entire chamber is held at 10 6 to 10 7 Torr during evaporation. The source to mask distance in the primary chamber used for this work is app roximately 20 cm. A large source to mask distance results in less efficient source material usage due to a smaller fraction of the molecular beam impacting the substrate


35 but will improve the uniformity of the laye r thicknesses across multiple substrates. Additionally, the substrate platter can be rotated to increase film thickness uniformity. The source materials are loaded into boats made from refractory metals (i.e. tungsten, molybdenum, tantalum) or insulating crucibles (i.e. boron nitride, aluminum oxide, quartz). The system is evacuated to high vacuum (< 10 6 Torr) and the boats are resistively heated to the evaporation (or sublimation) temperature of the source material. To increase uniformity in t he molec ular beam, the boats can be designed to act as a point source. The evaporated molecules exhibit ballistic transport behavior after exiting the boat, with the mean free path of each molecule determined according to 33 where k B T is the temperature of the molecular flux, P dep is the pressure in the de position chamber, and d 2 is the collision area between molecules, assuming that all molecules are spherical. In a typical high vacuum VTE system MFP is much greater than h the source to mask distance. This means that molecules can be assumed to follow a straight line between the source and the substrate without any collisions in between to alter their path, making geometric analysis of the mask and resulting feature size simple to determine. The feature resolution limit p of the system is where s is the substrate to mask separation, h is the source to mask distance, t is the thickness of the mask and l is the width of the source. As shown in Figure 1 8 divergent beams from the source, mask thickness, and substrate s eparation create a


36 shadowing effect that increases the feature size beyond the mask aperture. Slightly increased feature size must be accounted for when calculating efficiencies to report accurate values, but is generally not a concern in large area organ ic photovoltaic devices. During deposition, the film thickness and deposition rate are monitored with a calibrated quartz crystal microbalance. A properly calibrated system can achieve average film thicknesses less than one nanometer (i.e. a partial mono layer), allowing for detailed investigation of organic film growth behavior 34 Figure 1 8 Diagram of shadow mask geometry. VTE has several advantage s, offering high quality film deposition without the expense of a molecular beam epitaxy system, the ability to deposit metals, organics, and some inorganic dielectrics in the same system, very fine thickness control and the cap ability to deposit complex, multilayer structures Additionally, VTE is a preferred route for doping, as multiple crystal monitors can be used to precisely measure the rates of coevaporated materials. Doping has not found widespread use in organic photov oltaic devices, but organic light emitting d evices rely extensively on it. VTE is not suitable for all applications, however it cannot process polymers and materials with


37 low decomposition temperatures, and a large percentage of the source material is w asted. Still, it is a m ainstay technique for the fabrication of small molecule organic electronics devices. 1.4.3 Spin Coating As discussed in the previous section, polymers and other materials with decomposition temperatures less than their evaporation temperatures cannot be processed with vacuum thermal evaporation. Instead, these materials are processed using solution based methods, the most common laboratory technique being spin coating. Spin coating is an inexpensive method to achieve uniform thin films with a defined thickness from a wide variety of starting solutions. It is a well established technique, with the most familiar industrial application being photoresist application for photolithography in inorganic m icroelectronic device fabrication and patterning There are four stages to the process: 1. S olution deposition, 2. S pin up/acceleration, 3. S pin off, and 4. S olvent evaporation. In step (1), the source solution is added to the substrate. There are few constraints on the solution, save that a n excessive amount of solution is deposited onto the substrate and it must be able to flow The solution should also be free of dust and other particulates, as these can lead to defects in the final film. In the second step, the substrate is accelerated (either gradually or rapidly) to a final desired spin speed (typically several thousand revolutions per minute). Centrifugal forces create a wave front in the solution and it flows to the edge of the substrate. In the third stage, excess


38 solution is cast off the edge of the film and a uniform thickness is achieved based on a balance of the centrifugal force from acceleration and the solution viscosity. The fi lm then dries in the final stage as excess solvent evaporates. The spin coating proc ess is diagr ammed in Figure 1 9 Figure 1 9 Stages of spin coating. The final film thickness is dependent on the initial solution concentration, acceleration rate, solvent evaporation rate, and final spin speed making fine thickness control and repeatability possible. However, because film s are initial ly deposited in the liquid state, multilayer structures are difficult to process u nless underlying layers are not compatible with upper layer solvents O therwise, the underlying layers will redissolve and mix with the new layer. Also, metal electrodes must still be processed in vacuum after the polymer layer is deposited though solution processed electrodes are currently under investigation 35 37 Additionally, a large proportion of material is lost during the spin off stage making this a relatively expensive technique from a material usage standpoint. While spin coating is ultimately limited as a production scale method due to


39 wasted source material and limited substrate diameter, its low equipment cost and simplicity make it the standard laboratory technique for solution processed devices. 1.4.4 Emerging Techniques While vacuum evaporation and spin coating ar e the primary techniques used for device fabrication in this work there are a myriad of other techniques that are gaining increasing interest for industrial scale production and laboratory scale investigation. Foremost amongst these techniques is inkjet printing, a solution processed technique that offers high throughput fabr ication and non lithographic patterning as small as 5 m, depending on substrate preparation 38,39 Inkjet printing has very little overspray and is therefore much less wasteful than most other solution based methods. However, there are challenges to overcome: ink formulation is difficult and nozz le alignment is extremely important to pattern small features. Spray deposition is anot her alternative solution based fabrication method that is highly compatible with high throughput processing and large deposition areas 40 42 T he spray is either formed within an atomizer or pumped in liquid form to an ultrasonic nozzle. After the spray is formed it is guided by an inert gas flow to the substrate. Because the solution ar rives at the substrate as micrometer diameter droplets, a prime challenge in spray deposition is achieving uniform, high quality films. The main competitor to VTE for small molecule deposition is organi c vapor phase deposition (OVPD) 33,43 In OVPD, materials are evaporated in boat s or effusion cells under an inert gas atmosphere. The resulting molecular beam is direc ted towards the substrate by an inert carrier gas flow This method uses source materials more efficiently than VTE, but deposition rates are sensitive to pressu re, temperature, and


40 flow rate making OVPD more difficult to control. A related technology is organic vapor jet printing (OVJP), a hybrid between OVPD and inkjet printing, where organic vapor is fed through a printing nozzle via an inert carrier gas, mak ing patterning possible without shadow masking 44 46 1.5 Common Organic Based Devices Organic semiconductors have been successfully applied to a wide variety of electronic devices. The following sec tion is a brief overview of the two main organic optoelectronic device types: photov oltaics and light emitting devices 1.5.1 Organic Photovoltaics Efficient OPV devices were first demonstrated in 1986 by Tang with a ~1% efficient bilayer heterojunction consist ing of copper phthalo cyanine and a perylene derivative 47 Since the introduction of this architecture, rapid advancements have been made in both basic scientific understanding and device performance. Through the understanding and development of new active layer materials and optimization of device architecture s, state of the art OPVs now have efficiencies of over 8% 48 Processing technology has also evolved to the point that commercial pro duction and market viability are increasing. Some examples of OPV products are shown in Figure 1 1 0 highlighting their main advantage over inorganic PV modules: they are lightweight, flexible, and produced using roll to roll processing. These advantages make OPVs an excellent source for integration into building materials and everyday objects for portable power generation, such as clothing. OPVs can also be made to mimic natural shapes, such as leaves, for aesthetically pleasing or concealed installatio n.


41 Figure 1 10. Examples of organic photovoltaic devices. Clockwise from the far left: roll to roll production line of polymer OPVs, leaf shaped flexible OPVs, and a commercially available OPV module, Power Plastic by Konarka, Inc. This work has focus ed on two areas of organic ph otovoltaics research: Chapter 5 concerns how free charge carriers move within the device after exciton dissociation; Chapter 6 discusses the impact that controlled light propagation has on device performance, with the goal of d emonstrating practical enhancement techniques to push efficiencies towards commercially desirable values. For a detailed history of organic photovoltaics and a discussion of device operation principles, the reader is referred to Chapter 2. Proper measurem ent system calibration and device characterization techniques are discussed in Chapter 3.


42 1.5.2 Organic Light Emitting Devices To date, organic light emitting devices (OLEDs) have found the widest commercial acceptance within the field of organic electroni cs. In many ways, OLEDs operate as the reverse of OPVs charges are funneled into a light emitting organic material to form an exciton, which then recombines to emit light. The color of light is dependent on the optical gap of the emitting molecule. Whil e organic electroluminescence has been known since the 1960s 49 practical OLEDs were not realized until 1987 with the demonstration of a bilayer heterojunction of a hole transporting layer of N, N' bis(naphthalen 1 yl) N,N' bis(phenyl) benzidine (NPB ) and an electron transporting layer of aluminum tris(8 hydroxyquinoline) (Alq3) 50 In this device, holes and electrons are efficiently transported to the heterojunction interface by their respective layers and exciton formation and recombination occurs within the Alq3 layer to emit green light. The electron to photon conversion efficiency was approximately 1% at a driving voltage of less than 10 V, substantial improvements over previous technology. The largest jump in OLED performance came with the introduction of heavy metal complex phosphorescent emitters, which enable radiative recombination of triplets via spin orbit coupling 23 As discussed in Section 1.3.3, the triplet to singlet ratio in organic materials is 3:1, and radiative recombination of the triplet state is forbidden. The heavy metal atoms permit the singlet and triplet exci ton states to mix, allowing all excitons to contribute to light emission. Phosphorescent OLEDs have demonstrated ~ 100% internal quantum efficiency 24 26 OLEDs have rapidly matured into a commercially viable technology for activ e matrix displays, promising greater efficiency, flexibility, and truer color reproduction than


43 liquid crystal technology. A second application area is in solid state white lighting, with many laboratories demonstrating highly efficient, true white device s. Several commercially available and prototype applications are shown in Figure 1 1 1. Figure 1 1 1 television (LG Display),large area white light OLED panel (Fraunhofer IPMS), semitransparent automobile heads up display (Neoview KOLON/Hyundai), 1.6 Research Scope There are two main topics to the results presented here : first, understanding and controlling the behavior of charge carriers within organic optoelectronic devices by modifying the heterojunction architecture and second, increasing performance by


44 controlling the interaction of incident light with the device. Chapters 1 and 2 provide background organic electronics, particularly organic photovoltaics. Chapter 3 describes the experimental measurement methods and charact erization of device p erformance, including the introduction to a novel characterization technique developed for use in Chapter 5. The subsequently presented research is heavily related to understanding optical behavior within the device active area. Therefore, Chapter 4 descr ibes two different optical simulation techniques, Monte Carlo ray optics and transfer matrix wave optics. In addition to the mathematical underpinnings of these methods, example results are presented to demonstrate the potential of each technique and desc ribe some specific aspects of their implementation for this work. Chapter 5 explores the correlations between optical field, charge carrier motion, and heterojunction architecture in organic photovoltaic devices. Synch ronous photocurrent detection was use d to directly measure the photocurrent contribution from a variety of devices and a detailed qualitative model is presented to explain the observed behavior. The model is then extended further to explain optical field dependence and provide evidence for f ield assisted dissociation of excitons in bilayer OPVs. Optical management is the topic of Chapter 6. Transparent, stamped microlens arrays were used to increase performance in a wide range of organic photovoltaic devices. Their macro scale optical behav ior and geometric effects on enhancement were probed us ing Monte Carlo ray optics simulations and modified transfer matrix


45 wave optics simulations were applied to understand the effect of microlens arrays on the internal optical field. In both cases, qual itative and quantitative agreement s with experimental results w ere obtained. Chapter 7 discusses organic bifunctional optoelectronic devices (BFDs). BFDs can operate as either an OPV or an OLED, but the optimal designs of these two device types are in opp osition. Namely, OLEDs funnel charge carriers into the center of the device for recombination and light emission, whereas OPVs are designed to quench excitons and efficiently remove photogenerated charge carriers from the device interior. This places uni que design constraints on BFDs, which were explored. Ultimately, significant redesign of BFDs is required to maintain reasonable performance in both operation modes. Finally, Chapter 8 summarizes the results of this research and offers possible routes of further investigation.


46 CHAPTER 2 INTRODUCTION TO ORGANIC PHOTOVOLTAIC DEVICES 2.1 Basic Concepts The conventional silicon photovoltaic device is over half a century old and has established itself as a promising clean, alternative energy source. While th ese devices offer high power conversion efficiencies, their total cost per watt is still too high to be competitive with non renewable resources due to expensive source materials, processing costs, and installation restrictions. A push has been made recen tly to develop so than $1/watt 51 Extremely high pe rformance compound semiconductors and inorganic thin film devices are two examples of Class III technology; a third, and the focus of this work is organic based photovoltaics (OPVs). Organic materials hold several intrinsic advantages over inorganics, am ong them processability on inexpensive substrates using high throughput methods, intrinsic flexibility and mechanical robustness, and low source material cost. These beneficial characteristics have generated a huge amount of scientific interest, and performance has steadily increased from a reported value of 1% in 1986 to ov er 8% today 47,48 This chapter will describe the basic operation of organic photovoltaic devices and the history of their development, along with the current major avenues of investigation for increased performance. Before this, it will be beneficial to the reader to understand the parameters used t o quanti fy OPV per formance. Figure 2 1 shows the current voltage (J V) characteristics of a representative device. Three curves are depicted: the dark current, characteristic


47 of a typical diode, the photocurrent, the current generated by the device under illumination, and the total current, the summation of the dark and photocurrents. Because the dark current is always present, it is common to not consider the photocurrent as a separate component as it is difficult to directly measure. Figure 2 1 Cur rent voltage characteristics of a representative photovoltaic device. Two points indicated on the current and voltage axes are the short circuit current density, J SC the photocurrent generated at zero applied bias, and the open circuit voltage, V OC the point at which the summation of the photo and dark currents equals zero. Also indicated are the current and voltage at the maximum power point, within the power generating fourth quadrant (where V < V OC )


48 The fill factor ( FF ) is the ratio of the measured maximum power to the ideal such that An ideal fill factor will therefore be unity. Modern optimized OPV devices have FF 0.7 48 The metric of greatest interest is the total power conversion efficiency, defined as where P 0 is the incident power intensity. A final parameter of interest is the external quantum efficiency (EQE) a ratio of the number electron generation to incident photons at a certain wavelength, w ith q as the elementary charge, h c as the speed of light. The details of me asuring quantum efficiency are presented in Chapter 3. Photovoltaic devices can be represented by an equivale nt circuit, shown in Figure 2 2 The equivalent circuit contains an ideal d iode in parallel with a current source, representing the photocurrent generation within the device. There are two resistors that are additionally present the series resistance, R s which represents the intrinsic electrical resistance of the organic semi conducting layers and contact resistances, and


49 the shunt resistance, R sh a parallel resistance that represents leakage between the electrical contacts. From this circuit, an expression for the current voltage relationship can be developed, known as the S hockley equation: where J s is the saturation current density of the diode, n is the diode ideality factor, k is T is the temperature. From this relationship, it is clear that J is maximized when R s is minim ized and R sh is maximized. Increases in R s are associated with decreased FF and J SC ; decreased values of R sh reduce FF and V OC The principle components of an OPV are also shown in Figure 2 2. The specific mechanisms of operation and materials selection criteria for these layers will be discussed in Section 2.3. 1. Figure 2 2 Equivalent photovoltaic device circuit and typical schematic of an org anic photovoltaic device For an ideal photovoltaic device ( R s = 0, R sh = ), the photocurrent is directly proportional to the incident power and remains a constant value regardless of the


50 magnitude of P 0 The open circuit voltage is defined as the voltage at which the total current J = 0, or giving V OC a logarithmic dependence on J ph and, therefore, on P 0 Assuming no loss mechanisms, p should increase logarithmically with incident power due to the increase in V OC and the constant value of J SC / P 0 In an actual device, there are loss mechanisms. The series and shunt resistances are finite values, reducing FF J SC and V OC from their ideal values. Additionally, bimolecular recombination increases with free carrier concentration (and, therefore, illumination intensity), causing the J SC /P 0 ratio to reduce with increased illumination intensity according to 18 Therefore, t he highest value of P for a given OPV device is obtained at an incident power intensity where the increase in V OC is greater than the decrease in J SC / P 0 due to bimolecular recombination 2.2 Overview and History While the photovoltaic effect has been observed in organic materials dating back to 1959 52 modern effi cient organic photovoltaic d evices emerged in the 1980s with the introduction of the bilayer heterojunction architecture. Prior to this advancement, OPVs were inefficient Schottky diodes that relied on the strong electric field near a metal


51 electrode organic interface to split the p hotogenerated exciton and create free charge carriers 53,54 T his is intrinsically inefficient, as excitons can be quenched at the metal interface and exciton dissociation only occurs in a narrow band near the electrode, wasting a large percentage of absorbed photons. In an organic heterojunction, e xcitons are instead split apart by the energy level offsets at the heterojunction interface. A notable early success in bilayer OPV devices was reported by Harima, et al in 1984 55 By combining the electron transporting material 5,10,15, 20 tetra(3 pyridyl)porphyrin (TP yP) and the hole transporting material zinc phthalocyanine (ZnPc) they observed a roughly thirty times increase in p hotocurrent compared to a reference single layer ZnPc Schottky device. Power conversion efficiency under weak, 430 nm monochromatic illumination was ~2%, but poor exciton transport behavior limited performance. Additionally, photocurrent contribution was limited to the TPyP layer. The true breakthrough in OPV device performance, with a power conversion efficiency under simulated AM2 solar illumination of nearly 1%, was reported by Ching Tang in 1986 47 a nearly ten fold increase over earl ier efforts. Tang used a hole transporting layer of copper phthalocyanine (CuPc) and an electron transporting layer of 3,4,9,10 perylene tetracarboxy lic bis benzimidazole (PTCBI). This architecture shows photocurrent contribution from both layers, improving photocurrent generation and spectral response. The Tang cell became the archetype for future OPV devices. It had four components: an electron t ransporting/electron accepting layer (PTCBI), a hole transporting/electron donating layer (CuPc), a transparent high work function electrode (indium tin oxide, or ITO) and a low work function electrode (silver). C onvention has


52 respectively named these layers the acceptor, donor, anode, and cathode. The selection of these materials must follow certain guidelines for an efficient cell: the anode and cathode should correspondingly have Ohmic contact with the donor and acceptor, and the heterojuncti on formed between the donor and acceptor must have energy level offsets great enough to efficiently dissociate excitons from both materials. This is the interface can only di ssociate TPyP excitons, but the CuPc/PTCBI interface can dissociate both CuPc and PTCBI excitons. Additionally, the donor and acceptor can either be deposited as neat layers, to form a planar (or bilayer) heterojunction or mixed together to form a mixed ( or bulk) heterojunction. 2.3 Operation Principles 2.3.1 Basic Processes current world record efficiency of 8.3% 5,48 it is beneficial to describe the basic operation of an OPV device. There are four primary, s equential processes (Figure 2 3 ) that must occur in order to extract power from an OPV 56 : Photon absorption (exciton generatio n) Exciton diffusion Charge transfer (exciton dissociation) Charge collection As there are efficiencies associated with each of these processes, the overall quantum efficiency of a photovoltaic device can be summarized as the product of its constituents,


53 Or, the latter three processes can be considered as their own product, the internal quantum efficiency, reducing the quantum efficiency to: Thinking of device efficiency in these terms can be advantageous; absorption and internal quantum efficiency are inherently opposed in most organic systems. The first process, light absorption, is characterized by an optical absorption length of where is the wavelength dependent absorption coefficient of the material. Typical values of for organic materials are 10 4 to 10 5 cm 1 leading to absorption lengths of at least 100 nm. The absorption spectra for several photovoltaic materials are shown in F igure 2 4. Figure 2 3 Basic processes in power generation in a bilayer organic photovoltaic device.


54 The donor and acceptor materials should be chosen to maximize absorption ngly between 500 700 nm. Recent advances have been realized by incorporating small bandgap materials to increase near infrared absorption, such as lead phthalocyanine (PbPc ) and new conjugated polymers as discussed in Section 2.4.1. Figure 2 4 Optica l absorption spectra for several organic photovoltaic materials calculated from extinction coefficient data measured with spectrographic ellipsometry Upon absorption, an exciton, or bound electron hole pair is formed within the material, with a typical binding energy of 0.1 1 eV 18 Excitons are mobile particles, and will diffuse within the organic material, characterized by the exciton diffusion length with D ex as the diffusivity and as the exciton lifetime. Most organic


55 material exciton diffusion lengths are on the order of 10 nm 27 29 The smaller exciton diffusion length relative to the optical absorption length in most organic materials results in ED << 1 for bilayer heterojunctions that are thick enough to absorb a substanti al proportion of the incident light. Bulk heterojunction devices do not have this limitation. While the exciton is present and mobile within the active layer, two basic processes can occur. First, the exciton can recombine, either within the organic laye r after a certain time has elapsed or at a metal organic interface; this will reduce ED Otherwise the exciton will be dissociated by either the electric field or at the heterojunction interface. Schottky type OPVs rely on field assisted dissociation, but as the field required to efficiently dissociate an exciton is quite large, ~ 10 5 to 10 6 V/m 1,57 this process is typically ignored in organic heterojunction devices. Charge generation is assumed to occur solely through dissociation at the organic heterojunction interface, where it is energetically favorable for the bound exciton to split and have the free electron and hole reside on different molecules (electron on the acceptor, hole on the donor). In other words, the separation betwee n the donor HOMO and acceptor LUMO is greater than the binding energy of the exciton 1 The exciton dissociation process is exceedingly fast, on the order of a few hundred femtoseconds 58 In a properly designed hete rojunction, the charge transfer efficiency CT is taken as unity. The HOMO and LUMO levels for several common OPV donor and acceptor ma terials are shown in Figure 2 5 After dissociation the photogenerated charges are able to move to their respective elec trodes (holes to the anode, electrons to the cathode) for collection and power generation in the external circuit. This process can be highly efficient in bilayer devices,


56 where the heterojunction interface separates free electrons and holes. Since bimol ecular recombination is dependent on the product of the number of free electrons and holes np this is not a large loss mechanism in a bilayer device away from the heterojunction interface. In a bulk heterojunction device, free electrons and holes have a significant spatial overlap while moving towards their respective collection electrodes. Thus, the np product is large, recombination is significant, and CC << 1 is typical. Figure 2 5 HOMO and LUMO energy levels for several common OPV materials. Donors: CuPc, PbPc, SubPc, P3HT, and PCPDTBT. Acceptors: C 60 C 70 PTCBI, and PCBM. 2.3 .2 Fundamental Limitations As discussed in the previous section, o ne of the greatest obstacles to efficient OPVs is the fundamental tradeoff between photon absorption a nd internal quantum


57 efficiency In almost all relevant organic materials, the optical absorption length 1/ (>100 nm) is greater than the effe ctive exciton diffusion length (~10 nm) and charge collection length (< 100 nm). Thus, the optical absorption ef ficiency A is inherently opposed with either the exciton diffusion or charge collection efficiencies CC and ED depending on the device architecture and active layer materials. In general, bil ayer heterojunction devices are limited by ED due to their single plane of exciton dissociation; bulk heterojunction devices are limited by poor charge collection because of strong bimolecular recombination and poor charge transport morphologies related to segregation of the constituent donor and acceptor materials. Bec ause of these limitations, many avenues have been investigated to ameliorate the tradeoff between A and IQE These are reviewed in the next section. 2.4 Progress in Organic Photovoltaic Device Performance T here have been two primary avenues to improve the performance of OPVs: developing and incorporat ing new active layer materials and designing and optimizing device architectures and morphologies. A third route, optical management, has been sparsely reported in the literature but has shown promise. Wh ile there are many processes in the operation of an OPV that can be improved, the predominant trend has been towards alleviating the fundamental tradeoff between A and IQE The following s ections are an over view of notable examples of improved device pe rformance for small molecule and polymer OPVs. In Section 2.4.4 enhancements from improved optical management are highlighted.


58 2.4 .1 Small Molecule Organic Photovoltaic Devices The earliest successful OPV devices almost exclusively consisted of small mo lecule s it was realized that a primary limitation was the efficiency of exciton dissociation and charge generation. Several architectural changes were intro duced to attempt to remedy this. First, the planar mixed heterojunction, a sandwich of a mixed donor acceptor layer between two neat donor and acceptor layers, was developed to attempt to balance exciton dissociation efficiency (in the mixed layer) with c harge transport and collection efficiency (in the neat, planar layers). While early efforts showed marked improvement in photocurrent generation and J SC there were large decreases in fill factor that were attributed to poor morphology within the mixed l ayer 59,60 The poor mixed layer morphology can be overcome by controlling the phase segregation of the coevaporated materials 34 typically by thermal annealing or deposition on hot substrates. Increased crystallinity and phase segregation result in large surface roughness and poor device performance due to pinhole leakage pathways 61,62 but annealing after deposition of the metal cathode constrains film reorganization during annealing, preventing pinholes and increased roughness while still allowing for internal phase segregation 63 A cartoon of an ideal nano phase segregated morphology is shown in Figure 2 6. In the unoptimized case, there is a large interface area for exciton dissociation, but poor conducting pathways lead to recombination and reduced charge collection efficiency. In the ideal phase segregated case, free carriers have easy tr ansport routes to their collection electrodes, increasing CC From these findings, the planar mixed architecture has become a highly efficient choice for small molecule OPVs 64 67


59 Figure 2 6 Representations of unoptimized and nanoscale phase segregated bulk heterojunction OPV microstructures with two constituent materials. An example of charge carrier transport is shown for each. A second ideal architecture for small molecule OPVs is the so called interdigitated heteroju nction architecture (Figure 2 7 ). Here, neat pillars of donor and acceptor form a regularly spaced heterojunction with spacing on the order of the exciton diffusion length Figure 2 7. Ideal interdigitated heterojunction for organic photovoltaics.


60 Th us, very high exciton diffusion and charge collection efficiencies are possible. However, forming this stru cture on the nanoscale is quite challenging. The typical route taken is to form pillars of one material by either glancing angle deposition 68 71 of small molecules or synthesis of inorganic nanorods 72 and infilling with a solution processed molecule or polymer to complete the heterojunction. Another prominent architectural feature in small molecul e OPVs is the addition of a thin electron transporting layer between the acceptor and cathode th at prevents exciton quenching at the electrode /organic interface 29 2,9 d imethyl 4,7 diphenyl 1,10 phenanthroline, i.e. b athocuproine (BCP) is used as the exciton blocking layer for the small molecule devices in this work The greatest increases in small molecule device performance have come from in incorporation of advanced active layer materials, to increase exciton diffusion length, open circuit voltage, or incident photon absorption. Early advances were made by changin g the acceptor from PTCBI, with a short exc iton diffusion length of ~3 nm to C 60 with a diffusion length on the order of 40 nm 29 Additionally, C 60 has strong absorption at short wavelengths, offering better coverage of the solar spectrum than PTCBI when paired with metal phthalocyanines. Even better spectral coverage and device performance is obtained with C 70 which has strong ab sorption through the blue and green wavelength regions 73 C 60 and C 70 are the current standard smal l molecule acceptor materials. M ost material advances have resulted in new electron donors. Compared to the archetypal CuPc used by Tang, various metal and metal chloride phthalocyanines (Pcs) (tin Pc (SnPc) lead Pc (PbPc) chloroaluminum Pc (ClAlPc) and chloroindium Pc


61 (ClInPc) ) with red shifted absorption into the near infrared have been incorporated 74 80 To extend absorption further into the infrared and improve charge transport properties, the metal Pc layer can be depos ited on a thin, highly ordered templating material to encourage preferential molecular alignment 81 This is especially beneficial for PbPc, to increase the ratio of triclinic phase to monoclinic phase 79,82 The predominate crystal phase is also of importance in ClInPc which requires solvent annealing to inc rease near infrared absorption 83 Alternately, the donor layer can be chosen to maximize the open circuit voltage by minimizing the reverse saturation dark current J S 84 The most prominent of these materials is boron subphthalocyanine chloride, SubPc. With a V OC > 1 V and P > 3%, SubPc has become a standard donor layer in high efficiency bilayer OPVs 85 88 Enhancement in V OC has also been observed in zinc pthahlocyanine (ZnPc) synthesized such that the deposited films have reduced concentrations of electronic defect states, decreasing J S 89 Currently, the highest efficiency small molecule OPV is a tandem cell ( Section 2.4.3 ) produced by Heliatek GmbH using proprietary red, green, and blue absorbe rs for a power conversion efficiency of 8.3% 5,48 2.4 .2 Polymer Based Organic Photovoltaic Devices As was the case for small molecule OPVs, early conjugated polymer devices were single layer Schottky diodes with universally terrible performance made from polythiophene 90 polyacetylene 91 or polyviny lenes 92,93 The first appreciable performan ce was observed in bilayer devices made with a polymer donor of poly(2 methoxy 5 ethyl hexyloxy) 1,4 phenylene vinylene) (MEH PPV) and a thermally


62 evaporated C 60 acceptor 94,95 Eventually, a functionalized derivative of C 60 [6,6] phenyl C 61 butyric acid methyl ester ( PC 61 BM or more commonly, PCBM) was incorporated and has rema ined the standard acceptor though some modern high efficiency devices use PC 71 BM a functionalized derivative of C 70 96 Virtually all other material work on polymer OPVs has focused on optimizing the donor. The first class of donor polymers to sh ow strong performance was the poly(1,4 phenylene vinylene)s (PPVs), specific ally (MEH PPV) 97 and poly( 2 methoxy 5 (3,7 dimethylocty loxy) 1,4 phenylene vinylene) (MDMO PPV) 98,9 9 These early devices had high quantum efficiency under low intensity light that rolled off drastically as the illumination intensity increased, evidence of strong bimolecular recombination and poor film morphology. Sariciftici demonstrated the importan ce of morphological control and interface modification in polymer OPVs. Films cast from solutions of either chlorobenzene or toluene were compared and a strong correlation between reduced aggregation and improved device performance was observed 98 Further, modifying the cathode interface with LiF induces a strong interface dipole and increases charge extraction. This, combined with a hole extraction and film planarization layer of poly(3, 4 ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS) resulted in a device efficiency of 3.3%, comparable to the best small molecule devices of the day 100 The most widespread donor polymer to date has been poly(3 hexylthiophene) (P3HT), replacing the PPVs as the archetypal choice. The absorption spectrum of P3HT is sl ightly red shifted compared to MEH PPV or MDMO PPV, allowing it to absorb a greater portion of the solar spectrum and generate a larger photocurrent. Extensive studies on the morphology of P3HT:PCBM films to generate an ideal nanoscale phase


63 segregated st ructure has increased the ir efficiency to 5 6% 101 104 This underscores the value of morphological control in OPVs Additionally, it is now understood that the structural regularity and molecular weight of the donor polymer are important 105 107 Improved donor materials have followed two routes, both relying heavily on synthesis. The first is to develop low bandgap materials with absorption edges shifting to the near infrared. Attempts to increase the conjugation length of P3HT 108 o r change the thiophene subgroup to the more electronegative selenophene 109,110 were successful in reducing the band gap, but ultimately did not improve performance over that of P3HT. A more successful approach is to design new polymers wi th electron rich and electron poor (i.e. donor acceptor) const ituents 111 113 From the wide variety of synthesized polymers, a few will b e highlighted here. The first, p oly[2,6 (4,4 bis [2 ethylhexyl] 4H cyclopenta[2,1 b;3,4 b] dithiophene) alt 4,7 (2,1,3 benzothiadiazole)] (PCPDTBT), ha s a greatly red shifted absorption spectrum (with peak absorption observed under 825 nm illumination) and has found widespread adoption in the research community 114,115 A larger band gap polymer, poly[ N 9'' hepta decanyl 2,7 carbazole alt 5,5 (4',7' di 2 thienyl 2',1' ,3' benzothiadiazole) (PCDTBT) has a higher V OC than P3HT based cells due to a lower HOMO level, resulting in less energy loss upon charge separation 116 Heeger has reported a 6% efficient device using PC 71 BM that has almost 100% internal quantum efficiency under monochromatic gree n illumination 117 The final materials to be discussed are fluorinated donor acceptor polymers, one of which, poly(benzo[1,2 b:4,5 (5,6 difluoro 4,7 dithi en 2 yl 2,1,3 benzothiadiazole) (PBnDt DTffBT), was used in Chapter 6. By incorporating the


64 strongly electronegative fluorine atoms into a common 2,1,3 benzothiadiazole acceptor subgroup, the HOMO level is reduced an d interchain interaction is increased due to a more rigid aromatic structure 118 Devices made in conjunction with PBCM are greater than 7% efficient, among the few reports to reach this value 119,120 As of writing, the highest efficiency polymer OPV is a proprietary donor acceptor blend reported by Konarka, Inc. with an efficiency of 8.3% 48 2.4.3 Tandem Organic Photovoltaic Devices Given the inherent tradeoff between A and IQE in most organic systems, one possible mediation route is to stack multiple devices with high IQE in series. This efficiency ino rganic photovoltaics 48,121 by allowing greater coverage of the solar spectrum with different materials in the constituent subcells. This same concept can also be applied to organic mat erials, incorporating, for example, blue and green onal spectral coverage, simple tandem cell s can increase performance by doubling the output voltage of the structure F or photovoltaic cells in tandem, the current can be taken as the lowest of the subcell currents and the voltage as the sum of the subcell voltages, i.e. a two cell tandem would h ave J = min( J 1 J 2 ) and V = V 1 + V 2 where 1 and 2 denote the individual subcells A tandem organic photovoltaic has a notable architectural requirement that single junction cells lack the carrier recombination zone (CRZ) that connects the subcells. In the CRZ, photogenerated electrons and holes from the front and back cells


65 respectively, and recombine. Most of the development in tandem OPVs has focused on this region. The first tandem devices utilized a 15 nm thick semitransparent layer of gold whic h has high conductivity but blocks a large proportion of incident light 122 Yakimov and Forrest presented a better CRZ made from high conductivity small molecule layers embedded with a vacuum deposited 0.5 nm thick (nominal) layer of silver nanoparticles 123 A t these dimensions the nanoparticles do not absorb a significant amount of light and can even enhance the optical field at certain wavelengths due to plasmonic effects 124 This interlayer was used in ~2.5% efficient CuPc/PTCBI based devices and a 5.7% efficient tandem based on planar mixed CuPc/C 60 devices, both records for those materials 66,123 Tandem devices based on polymer subcells are much more difficult to fabricate tha n those built from small molecules. Specifically, the front cell must be either not be soluble in or completely protected from (via the CRZ) the solvent used in processing the back cell. Therefore, polymer cells can either be used as the front cell with a vacuum deposited small molecule back cell 125,126 or the CRZ can be engineered to isolate the front cell using nanoparticle or solvent resistant organic layers 127 130 Regardless of the material c hoices, the architecture of a tandem OPV is highly complex compared to a si ngle junction cell. Figure 2 8 shows a typical architecture for a two cell tandem OPV and the optical field optimized for CuPc/C 60 planar mixed subcells 66 Optimizing both the charge transport properties and optical fields in the front and back cells is a non trivial task. Typically, the back cell is situate d in the first blue green interference peaks and the CRZ thickness is such that the front cell is in the in the


66 first green red interference peak, or into the second blue region, depending on the absorption spectra of the organic layers. Figure 2 8 Typical device structure and optical field plot for a tandem organic photovoltaic device. The optical field is calculated for front and back cells of planar mixed CuPc/C 60 heterojunction, with optimized thicknesses 66 2.4 .4 Optical Management The previous sections discussed changes in active layer materials and device architectures to increase efficiency. History has certainly borne out that these are valid improvement routes, but a third approach, optical management, has been relatively ignor ed. Instead of altering the internal components of a device, changes can be made to alter how the entire device (substrate and active layers) interacts with incoming light. This is accomplished by altering the substrate geometry Non geometric enhanceme nts, including luminescent concentrating layers 131 and plasmon enhanced


67 absorption 124,132 134 can also be considered optical management but will not be discussed further. Figure 2 9 shows three examples of changes to substrate geometry to increase absorption by forcing light to pass th rough the active layer multiple times. Figure 2 9 Three previous examples of op tical enhancement techniques. A ) V aligned solar cells for enhanced light trapping 135 137 B ) OPVs produced on microprism substrates, to induc e total internal reflection 138 C ) A mirror and microlens light trap: a metalized layer is deposited with gaps at the microlens foc al point to admit light 139,140


68 Figure 2 9 A shows two substrates in a V pattern to form a light trap 135 137 Light that enters is funneled deeper into the trap, passing through the active layers multiple times. Further, this trap will cause multiple reflections at almost all incident angles, though a large portion of the total active area will be shaded as the illumination angle changes. Figure 2 9 B details a device fabricated on a prism shaped substrate such that light reflecting off of the cathode will undergo total internal reflection at the substrate/air interface, returning for two more potential passes t hrough the active layer 138 The prismatic features in this case are on the order of 100 m; other textured substrates have been use d with much smaller features intended to serve as diffraction gratings 141,142 Finally, Figure 2 9 C shows a complex structure consisting of a microlens array that focuses incoming light through gaps in a photolithographically patterned metal layer that forms a reflecting cavity with the cathode; su bsequent reflection s have a very small probability of escaping, in theory greatly increasing absorption 139,140 These examples are ultimately ill suited for use in commercial OPVs, where low processing cost, large device areas, and high throughput are required. A geometric substrate modification tha t meets these criteria, soft lithographically stamped microlens arrays, is presented in Chapter 6.


69 CHAPTER 3 ORGANIC OPTOELECTRONIC DEVICE CHARACTERIZATION 3.1 Chapter Overview The proper measurement of optoelectronic device performance is key to ensure valid results that are directly comparable to those within the scientific community at large 143 This includes correct design and i nstallation of characterization systems and equipment, control of systematic measurement errors, and redundant calibrations. As this work is largely focused on OPVs, specifics of their characterization. The background and calibration of a simulated solar light source are discussed in Section 3.2.1 along with the most prevalent OPV characterization technique, cu rrent voltage (J V) measurement Section 3.2.2 covers another common characterization method, quantum efficiency both external (EQE) and internal (IQE) Finally, Section 3.2.3 introduces synchronous photocurrent detection, a novel technique used extensively in Chapter 5. The bifunctional photovoltaic and light emitting devices investigated in Chapter 7 require characterization of their emission efficiencies; therefore, t he principles of OLED characterization are briefly presented in Section 3.3. 3. 2. Organic Photovoltaics 3.2 .1 Calibration, Spectral Mismatch and Current Voltage Measurement A supreme concern in proper OPV characterization is the quality of the simulated solar light source in comparison to the reference solar spectrum. The typical refe rence solar spectrum is currently defined by the ASTM International G173 03 standard 144


70 which specifies the optical power (Wm 2 nm 1 ) at regular intervals from the ultraviolet (280 nm) to infrared (4000 nm) wavelengths at precise irradiance angles over the contiguous United States (37 tilt relative to the equator). The actual spectra are calcula ted 145 147 using several assumptions, the most significant of which is the absolute air mass of 1.5, for a solar zenith angle of 48 There are three reference spectra included in the G173 3 standard: extraterrestrial, global, an d direct (Figure 3 1A ) Figure 3 1 Reference solar spectra. A ) Reference Air Mass 1.5 spectra f rom the ASTM G173 3 standard. B ) AM1.5 Global spectrum with the normalized integrated optical intensity within visible wavelengths. The extraterrestrial spectrum does not include absorption by the atmosphere and other effects and is not used for this work The direct spectrum includes only the light that is directly incident on the 37 surface, while the global spectrum additionally includes light diffus ed by the atmosphere; the latter is used for the calibrations Since the total integrated power in the global spectrum is ~100 mW/cm 2 o ne sun is


71 defined as this val ue. Figure 3 1B shows the AM1.5G (global) spectrum within the visible region, from 350 to 1000 nm, and integrated irradiance valu es across this wavelength range The devices characterized in this work were measured under illumination from a 150W Xe arc bulb with a parabolic rear reflector to increase irradiance. The Xe arc bulb has strong UV emission, so an AM1.5G filter is used to remove some of this light from the spectrum. The incident power intensity was measured using a certified Si reference cell with an active area of 0.9 cm 2 Fig ure 3 2 Reference AM1.5 Global and simulated Xe arc lamp spectra. Data is normalized to 495 nm for comparison. Because the simulated solar light spectrum even with the AM1.5G filter, does not exactly match the r eference spectrum (Figure 3 2 ) and organic materials have different absorption spectra compared to the Si detector, a spectral mismatch factor M is used to adjust the measured intensity, according to


72 w here E R is the reference spectr al intensity, E S is the simulated source intensity, S R is the responsivity of the Si reference cell, and S T is the responsivity of the OPV under consideration. Each function should be integrated over the spectral response range of the test cell. Note tha t a KG1 standardized colored glass filter has been added to the Si reference cell to reduce its long wavelength absorption giving it better spectral responsivity agreement with organic materials. Despite the difference in spectral irradiance between the s olar simulator and reference spectrum, most OPVs have a mismatch close to unity in this system. The final, corrected optical intensity value is taken as arc reference cell and the test OPV device The spectral mismatch factor M for several different organic materi al systems in this characterization setup is shown in Table 3 1. Table 3 1. Spectral mismatch factor M for various devices P3HT:PCBM CuPc/C 60 ZnPc / C 60 SubPc/C 60 PBnDt DTffBT:PCBM M 0.98 0.99 1.01 0.98 0.98


73 The typical arrangement of the solar simulator and associated optical components for a J V measurement are shown in Figure 3 3. Neutral density filters are used to adjust the incident power P 0 from 0.1 sun to greater than 1 sun. The testing pocket is placed at a position calibrated to be 1 sun intensity under a certain filter combination using the Si referen ce cell with an assumed mismatch factor of unity. The intensity at this position can be adjusted slightly to account for variability in the spectral mismatch factors and lamp age by varying the lamp driving power (to 150 5 W). Figure 3 3 Arrangement of the solar simulator and associated optical components. In this arrangement, multiple irises are used to isolate a specific area of the larger generated beam to increase beam uniformity. Small area devices are measured with the final, small iris in pla ce; this ir is can be flipped out of the optical path to illuminate larger areas. Unfortunately, the Xe arc bulb and parabolic reflecto r combination still


74 results in significan t variations across the final ~4 mm diameter spot. Figure 3 4 details the varia bility in power intensity across the final spot, represented as the short circuit current generated from the Si reference cell through a pinhole opening. Figure 3 4 Optical intensity distribution over a calibrated 100 mW/cm 2 white light beam, measured as a function of current from a Si reference cell using a pinhole aperture.


75 There is an obvious region of high intensity near (2 mm, 2 mm), with intensity decreasing drastically toward the spot edge. While the total intensity o ver the spot size in this instance is ~100 mW/cm 2 extra care must be taken to account for this when measuring devices with an active area different from the standard 0.04 cm 2 The current voltage characteristics are measured using an Agilent 4155C semicon ductor parameter analyzer, which both biases the test OPV and measures the current output. For dark current measurements, a shutter is used to block the white light beam and the test pocket is covered with a blackout cloth to e nsure a light free environme nt. Typically, double scans are used to detect any hysteresis in device performance, which can be indicative of charge trapping or other electronic defects within the device active layer. For illuminated scans, the relevant photovoltaic performance param eters (short circuit current ( J SC ), open circuit voltage ( V OC ), and the maximum power point ( P max )) are averaged from the two individual scans. 3.2 .2 Quantum Efficiency A second vital technique to assess and compare OPV performance is quantum efficiency measurement. T he quantum efficiency is calculated in one of two ways: the external quantum efficiency (EQE), a representation of t he number of electrons collected per number of photons incident on a device, and the internal quantum efficiency (IQE), the r atio of electrons collected per absorbed photon. EQE is also known as the internal photon to electron conversion efficiency, IPCE. The quantum efficiency measurem ent system is shown in Figure 3 5 conforming to the ASTM E1021 testing standard 148


76 EQE is typically measured across the visible spectrum to show the full spectral response of an OPV. A quartz tungsten halogen ( QTH) lamp feeds white light into a monochromator, where diffraction gratings are used to split the white light into a low intensity (1 10 W/cm 2 ) monochromatic beam of the desired wavelength. From there, the diverging monochromatic beam is collimated and chopped by a mechanical blade to create an alternating signal. The beam is then condensed by another lens onto either the test device or, for incident power measurement, a calibrated Si photodetector. Figure 3 5 External quantum efficiency characte rization system based on the ASTM E1021 standard.


77 The test device is also illuminated by non chopped white light of typically 0.5 1 sun intensity to mimic the optical and electrical field within a device measured under one sun conditions in the J V charac terization system. The photocurrent is routed through a current a mplifier to increase the signal/ noise ratio, and then input into a lock in amplifier. The lock in amplifier uses the reference frequency from the chopper controller to isolate the current c omponent that comes solely from the chopped monochromatic light. The EQE is then calculated as where c is the speed of light, h q is the elementary charge, I T ( ) is the lock in measured photocurrent from the test OPV, I D ( ) is the Si photodetector current, and R D ( ) is the responsivity of the reference detector. To calculate IQE, the absorption efficiency A ( ) is taken into account. Ideally, the monochromatic b eam should be slightly smaller than the device itself. If the beam is larger than the device area, the incident power I D ( )/ R D ( ) must be adjusted to account only for the power incident on the device area (this adjustment is used in Chapter 6 to measure E QE under large area monochromatic illumination). The EQE spectrum for a rubrene/C 60 (35/25 nm) device is shown in Figure 3 6. The short circuit current under reference 100 mW/cm 2 AM1.5G illumination can also be calculated from EQE data, via w here S ( ) is the reference AM1.5G power intensity. If both the EQE and J V measurement systems are calibrated correctly, the difference between the integrated


78 and experimental J SC values should be minor. For the example rubrene/C 60 device, the mismatch is approximately 3%, indicating acceptable agreement between the two systems and proper calibration Figure 3 6 External quantum efficiency spectrum and current voltage curve for a rubrene/C 60 (3 5/25 nm) device. The EQE calculate d and experimentally measured short circuit currents are indicated. 3.2 .3 Synchronous Photocurrent Synchronous photocurrent detection is a novel technique developed and used extensively for this work The basic characterization setup is the same as for an EQE measurem ent (Figure 3 5) with the light source either being monochromatic or simulated solar illumination However, a bias is applied across the device by the current amplifier. The bias is swept from negative to positiv e (typically to +1 V) to reconstruct


79 the photocurrent component of the total current curve. The measurement can be performed under chopped simulated solar illumination to obtain a photocurrent curve of a directly comparable magnitude to the total an d dark currents, as in Figure 3 7A for a 20 nm zinc phthalocyanine (ZnPc)/40 nm C 60 bilayer device Note that this demonstrates how the photocurrent curve dominates total current from V = 1 V to V 0.25 V. The dark current dominates the total current at larg er forward biases, as expected. Figure 3 7 Example photocurrent data. A ) Dark, total, and photocurrents of a 20 nm ZnPc/40 nm C 60 bilayer device under simulated solar illumination. B ) Photocurrent measured under monochromatic illumination of various wavelengths for a 90 nm CuPc:C 60 (1:1 by weight) mixed heterojunction device. Inset: phase information recorded from the lock in amplifier. The more common usage is under monochromatic illumination, to study the photocurrent behavior as a function of wa velength (Figure 3 7B ) In this configuration, a strong white light bias (ideally 100 mW/cm 2 but device instability at large voltages can necessitate lower intensities to reduce measurement error) is applied to create a representative electrical field ac ross the active layers of the test OPV. Additionally, the


80 white light bias spectrum should be matched with the absorption spectrum of the device. For example, the spectra of three different b ias lamps are shown in Figure 3 8, along with the absorption sp ectra of CuPc, C 60 and PbPc Figure 3 8 Irradiance spectra of two different white light bias lamps compared to the absorption spectra of three representative OPV materials within the visible region. The LED lamp irradiance spectrum is closely matched to the absorption peaks of CuPc ( = 600 700 nm) and C 60 ( = 450 nm), making it a suitable choice for that material system. PbPc, however, absorbs at longer wavelengths ( > 700 nm) and is not matched well with the LED lamp spectrum. The halogen lamp is a better choice for PbPc to better mimic the AM1.5G conditions. The halogen lamp also demonstrates spectral match with CuPc and C 60 Precaution must be taken to cool the OPV when


81 using a halogen bulb due to its large heat output; extra heat will degrade the device over the course of the measurement. As the bias is increa sed across the test OPV during testing the magnitude of the photocurrent can decrease by several orders of magnitude. This rapidly decreases the signal to noise ratio of the meas urement To compensate, a longer integration constant (300 ms or 1 s), greater noise filtering within the lock in amplifier, and much longer hold times (~7 time constants) compared to an EQE measurement are used. S can ti mes can range from one to three ho urs depending on the number of wavelengths considered, necessitating device encapsulation to minimize degradation during this period. Measurement times can be reduced by varying the integration constant and hold time based on the magnitude of the photocur rent signal Refer to Chapter 5 to see synchronous photocurrent measurements applied to study photocarrier generation and transport behavior based on OPV heterojunction design and for other purposes. 3.3 Organic Light Emitting Diodes Complete characteriza tion of an OLED encompasses the measurement of current voltage performance, luminance and emission spectra. From these characteristics the luminous, power, and quantum efficiencies are determined. The emission spectrum is measured using a spectrometer (O cean Optics JAZ), which accepts light through a 400 m optical fiber. The incoming light is split with a diffraction grating and reflected onto a charge coupled device (CCD) array which is configured to correlate response from each pixel with a specific w avelength. In addition


82 to photon count measurements, the system is calibrated with a reference tungsten halogen light source for absolute power measurement. Current voltage and luminance characteristics are measured using th e system depicted in Figure 3 9 Figure 3 9 OLED current voltage and luminance characterization system. The most important assumption made in this characterization system is that the OLED emission pattern is Lambertian, or equal in intensity in the forward direction regardless of the solid angle 149 In reality, there are slight deviations from Lambertian emission, and any type of optical structure designed t o increase light outcoupling wi ll alter the emission pattern 150,151 The system is initially calibrated using a commercial luminance meter to determine a conversion factor between the actual luminance L and the measured


83 detector photocurrent I det The value of is obtained by curve fitting of L vs. I d et over a representative luminance range. Additionally, depends on the spectral response of the photodetector and a geometric factor f the fraction of the total emitted light that the detector can observe. By assuming that the emission spectrum of the device is constant regardless of driving voltage, f is calculated as where g ( ) is the normalized photopic respon se (responsivity of the human eye in bright light) curve with a peak value 0 = 683 lm/W 152 S ( ) is the device emission spectrum, R ( ) is the detector responsivity, and A is the device area. Typically, the emission spectra of different devices are ignored in the calculation of f making it dependent only on the geometry of the measurement system, such as detector area and distance from the device. With the geometric factor, emission spectrum, I det and the device driving current I D known, the luminous efficiency L power efficiency P and external quantum efficiency EQE can be readily calculated, ,


84 where h c is the speed of light, and q is the elementary charge. Ideally, the emission spectrum S ( ) should be measured at a variety of voltages to account for any changes as the driving conditions change. Note also that the method of calculating f only accounts for forward emission, ignoring waveguided and back reflected emission; an integrating spher e can be used to account for these components 152


85 CHAPTER 4 OPTICAL SIMULATION OF ORGANIC PHOTOVOLTAIC DEVICES 4.1 Monte Carlo Ba sed Ray Optics Modeling The propagation of light can be represented in two ways: as a ray with an origin point and a direction or as a wave traveling with a certain frequency wavelength, and direction. Because the wavelength of a visible light photon is less than 1 m, optical simulations at larger dimensions than this can be suitably approximated by treating the light wave a s a ray. At dimensions near or less than the wavelength of light the governing equations of ray optics are not an accurate approximation and the wave nature of light must be taken into account. For the purposes of this work Monte Carlo ray optics simulations were used to primarily examine the geometric effects of optical structures on light propagation behav ior, not model the behavior of light within the active layer itself. In the realm of organic devices, ray optics have primarily been used for studying light extraction in light emitting diodes 150,153 but there have been applications in photovoltaics research 154,155 At macroscopic scales, the equations that govern the behavior of light are well known. Transmission and reflection probabilities are cal cu lated via the Fresnel e quations, which are dependent on the polarization of the incident light ray (either parallel or perpendicular to the incident plane, respectively notated as p polarization and s polarization), the indic es of refraction of the media under consideration and the incident angle of the light ray relative to the interface surface normal. A simple diagram denoting the relationship between the incident angle I the reflected angle r and the


86 transmitted angle t as light moves from o ne medium into a second (with a higher refractive index) is shown in Fig. 4. 1 Fig ure 4 1 Schematic diagram of light propagation via ray optics. Note that n 2 > n 1 in this situation. The relative proportions of R and T are calculated using the Fresnel equations, Where n is the refractive index and the angles i and t are referenced to the surface normal as shown in Fig ure 4 1 law, This relationship can then be used to reduce the Fresnel equations to functions of n and i


87 For nonpolarized light, the total reflection and transmission coefficient s are then taken as: With this, the proportions of an incoming light ray with intensity I that have been reflected ( R ) and transmitted ( T ) at the interface between two media with differing refractive indices can be readily calculated The final component from Fig ure 4 1, r is simply i only the direction of the ray changes, not its angle relative to the interface normal. Practical implementation of these equations in a computer simulation can take two forms: first, a recursive algorithm that starts with a single ray with a set incident intensity, which is then split into two components at the first interface reflected and transmitted, with the relative intensities of those two rays calculated using the above formulations. Those rays are then furthe r split into reflected and transmitted components at the next interface they encounter, and so on. The second approach relies on simulating a lar ge number of rays to generate a statistical ly valid distribution of reflected and transmitted rays at each i nterface, where


88 the number of rays reflected is proportional to the Fresnel calculated R value For example, the calculated R value for normally incident light on an air glass interface (with refractive indices of 1 and 1.5, respectively) is approximately 0.04. The first method would take a single ray with intensity 1.00 and split it into a transmitted ray with an intensity of 0.96 and a reflected ray with intensity of 0.04. The second method would generate 100 rays, 4 of which would be reflected and 96 of which would be transmitted without any change in intensity. This is considered a Monte Carlo type simulation method. The latter approach has been implemented. The following discussion contains details of the specific implementations of the basic ray optics method, and how that technique has been adapte d to work with microlens arrays with the necessary refinements to give quantitatively significant results. The results of this simulator are used extensively in Chapter 6. 4.1.1 Basic Implementation S cheme A basic ray optics simulator of planar layers must perform several functions: 1. Create a simulated stack of several layers and remember their locations in space (layer start/end coordinates) and refractive indices 2. Generate rays with a randomized origin and desired vector 3. position, vector, and the coordinates of the next layer interface 4. Perform Fresnel calculations to determine the reflection and transmission probabilities 5. Determine whether the ray will reflect or transmit at the interface 6. Calculate new vectors after either reflection or transmission


89 7. Repeat steps 2 through 6 unti l the ray will no longer interact with the simulated stack due to reflection away from the layer s, transmission through all layers, or absorption within a layer 8. Accurately track the behavior of the ray, and whether it was absorbed, transmitted, or reflected 9. Repeat steps 2 through 8 for a large number of rays. A flow chart describing th is process is s hown in Figure 4 2. Fig ure 4 2 Simplified flow chart of a Monte Carlo ray optics simulator


90 For a system composed of purely planar layers, the implementation and mathematics of the simulator are straightforward A basic simulated stack of pla nar layers is shown in Figure 4 3, with the different device, substrate, and illumination areas back side of the active layer; the rest of the active layer is taken as a djacent to air, so rays can be lost by transmitting through the entire stack. Fig ure 4 3 Typical simulated stack. Note that layer thicknesses are not to scale. The intersection point between the ray and the next planar layer is a simple projection onto the layer. After the Fresnel reflection and transmission coefficients, R and T are determined, a random number within the set [0,1] is obtained; if the random number is less than or equal to R the ray is reflec ted and a new vector is calculated acc ording to: w here d is the direction vector of the ray, and n is the surface normal Otherwise, the ray is transmitted through the interface and the new refracted vector is:


91 w here n 1 and n 2 are the refractive indices of the layers involved. Eventually, the ray will transmit to the organic ac tive layer, where it can be absorbed according to the Beer Lambert law where is the wavelength dependent absorption coefficien t and d is the path length of light within the layer Similarly to the determination mechanism for reflection and transmission, if a randomly generated number is less than or equal to A the ray is absorbed. This calculation is performed whenever the ray is within the active layer. Should the ray not be absorbed on a first pass, there are subsequent opportunities for absorption on additional passes. If the ray is outside of the device area, Fresnel calculations are performed to determine if the ray is reflected or transmitt ed at the organic/air interface; within the device area, the ray is reflected off of the cathode back into the active layer, absorption calculations are repeated, and the simulation cycle continues. Several parameters are tracked dur ing the simulation for later analysis, among them the number of rays absorbed in the device area, absorbed outside of the device area, reflected upon initial interaction with the stack, reflected at any other point in the simulation, and transmitted throug h the stack. Additional tracked quantities include the ray pathlength through the active layer, initial generation point (inside or outside of the device area) and wavelength of each absorbed ray.


92 4.1.2 Simulation of Optical Structures The basic simulato r framework is suitable for a stack that contains non planar layers, such as arrays of pyramids or hemispherical microle nses, but significant care must be taken to properly describe the interaction around the non planar optical structure. This takes sever al forms: 1. Correct initialization of the simulated optical structure layer, with the flexibility to accommodate a variety of structures, such as microlens arrays with different contact angles, packing factors, and lens diameters. 2. Determination of whether th e next intersection will be with a planar layer or the optical structure 3. Accurate calculation of ray intersection points with the optical structure 4. Error checking to ensure interaction with only one optical structure, in the case of an array of features F or the purposes of this section, the optical structure is a convex hemispherical microlens array unless otherwise noted The typical simulated device s tructure is shown in Figure 4 4 with typical layer thicknesses and indices of refraction indicated. No te that the layer thicknesses are not to scale. Figure 4 4 Typical simulated device structure with a convex microlens array.


93 Proper simulation of the lens geometry is crucial. For a convex hemispherical microlens array with a contact angle of 90 a s phere is first created at a desired height within the structure the midpoint of the sphere is placed at the interface between air and the buffer layer, leaving a perfect hemisphere above and below the interface. The sphere is then repeated in space to s imulate the array. Two different packing factors are considered (Figure 4 5 ), a square array and a close packed hexagonal array. Image files are generated by recording the coordinates at every point the ray interacts with the structure; the resulting coo rdinate file is output and viewed using Visual Molec ular Dynamics (VMD) 156 Figur e 4 5 Simulated microlens arrays. A ) square and B ) hexagonal close packed (right) arrays of 1 mm diameter, 90 contact angle convex lenses ( visualized using VMD 156 ) With the full array simulated, the concavity of the array is defined by disallowing interactions between the ray and sphere below (for convex arrays) or above (for concave arrays) the air/buffer interface. To simplify later calculations, the center point


94 of each lens is saved during array generation. The lens array can either be created to fill the entire defined array area or as a small array that is spatially shifted to catch rays as they approach an intersection with the lens layer. The former approach is computat ionally expensive but simple to program; the latter requires additional care to accurately shift the array but vastly decreases expense and simulation time The latter approach was used in this simulator. Simulation of different contact angles was also required. Figure 4 6 illustrates the shifting process to create a lens with a contact angle less than 90 First, a 90 lens is made by the previously described process. Then, the lens is shifted a distanc e of into the buffer layer, where r is the radius of the sphere and is the desired contact angle. The array is then generated by either using the original sphere radius to leave gaps between the individual lenses, or by using the new lens radius to create a close packed array. Figure 4 6 C oncept and mathematical details to simulate lenses of different contact an gles.


95 Close packed arrays were used to isolate the effect of contact angle. Images of 90 and 30 hexagonally close packed arrays generated from 1 mm spheres are shown in Figure 4 7 visualized with VMD. The images appear semi transparent b ecause the coo rdinate files are generated from a finite number of interactions. Figure 4 7 Simulated lens arrays with contact angles of 90 and 30 (visualized with VMD 156 ). With the array properly created, the next concern is accurately defining interactions between the ray and the lens layer. Whenever there is a chance that the ray will intersect with the lens array an intersection point is sought; if no valid intersection is found, the next interaction surface is planar and the ray is projected to the next planar interface and Fresnel reflection and transmission calculations are performed. The translatio sphere is a quadratic relation, calculated as


9 6 , Where d is the ray direction, s is the ray o rigin, c is the sphere center point and r is the sphere radius If the quan tity is negative, no valid intersection exists between the ray and the sphere. In a convex array, the only allowed solutions are positive values of t such that the intersection point is above the basal plane of the lens array. This calculation is repeated for all lenses in the array and the smallest positive value of t is taken as the correct intersection point. If no suitable p ositive value of t exists within the array the intersection is treated as a planar. After an appropriate intersection point is calculated, the local surface normal of the lens at the intersection point is calculated for use in the Fresnel transmission and reflection calculations, and the ray is either reflected or refracted according to the previous equations. 4.1.3 Quantitative Significance and Verification Since this simulator is intended to show the effect that optical structures have on a real photovol taic device additional steps must be taken to ensure that the data is quantitatively significant. First, elementary hand calc ulations can be compared with simulated results to check that rays are being generated in an appropriately random manner across th e coefficient behaves appropriately for an air glass ( n = 1.5) interface (Figure 4 8 A ) and that ray generation occurs in an appropriately random manner (Figure 4 8B)


97 Figure 4 8 Verification of basic simulator functions. A ) Calculated Fresnel reflection coefficients as a function of incident angle. B ) Verification of randomized ray location generation. To better approximate and analyze the absorption characteristics induced in real organic photovoltaic devices, the organic active layer is assigned wavelength dependent absorption coefficients and rays are assigned wavelength values to mimic the wavelen gth distribution of light in the reference AM 1.5G solar spectrum 144 Absorption coefficients were derived from extinction coefficient values measured us ing spectroscopic ellipsometry according to T he absorption coefficients and simulated AM 1.5 G spectra are shown in Figure 4.9. The simulated AM 1.5G spectrum is created by first normalizing individual wavelength photon counts to the total photon count within the reference spectrum, then generating a random number. The ray is defined as having wavelength if the random number R s atisfies


98 where G is the list of ordered, normalized photo n counts. As shown in figure 4 9, close reproduction of the reference spectrum is obtained with a simulation of only one hundred thousand rays. Figure 4 9 Simulated air mass 1.5G solar spectrum and active layers absorption coefficients for various material systems Increasing the ray count to one million gives an almost exact reproduction. C ombining an accurate simulated spectrum with realistic absorption values gives a fairly accurate reproduction of solar cell absorption. Finally, periodic boundary conditions (PBCs) are enforced using the array area as a boundary, allowing the device to be effectively infinite (device area = array area) or isolated (device area << arr ay area). Simulated absorption spectra and the results of various geometric arrangements are discussed extensively in Chapter 6.


99 4.2 Transfer Matrix Wave Optics 4.2.1 Concept For dimensions near and less than the wavelength of visible light, ray optics methods are not suitable to describe the behavior of light. In OPVs, active layer thicknesses are on the order of 100 nm, and are typically adjacent to a highly reflective surface, the thermally evaporated metal cathode. In the area close to a reflective surface, incident and reflected light waves will constructively and destructively interact, creating areas of high and low optical intensity. The emergent optical interference patterns can have a profound effect on OPV performance. In bilayer devices, th icknesses should be optimized to center regions of high optical intensity near the heterojunction interface, ensuring that a large proportion of generated excitons are dissociated at the interface. Mixed heterojunction devices do not have this requirement but regions of strong optical absorption should be centered within the active region to maximize photocurrent generation. This can be accomplished by either increasing the layer thickness or, if charge carrier transport lengths or film morphologies limi t how thick the active layer can be made, adding an optical spacer between the cathode and active layer. Additionally, tandem OPVs require extensive optical design to balance photocurrent between the front and back cells to maximize efficiency 66 In all of these cases, transfer matrix optical simulations are a computationally efficient, robust method to examine optical interference effects in planar stacks of optical media. Here the transfer matrix method of Pettersson 157 and Peumans 1 is implemented. This approach calculates the optical field within a series of planar layers and then determines the resulting exciton generation and photocurrent output from the

PAGE 100

100 device, along with calculating properties such as reflection, transmission, and device absorption. The following section is a summary of the derivation of this method, which has been imple mented in numerous studies 27,66,79,158 163 4.2.2 Opt ical Field Calculation Given the assumptions that all l ayers are uniform and isotropic with parallel, flat interfaces (for convention, the interface normals are taken to be parallel to the x direction within the simulation), and that the incident light can be de s cribed by planar waves, a basic formulation emerges. At any point within the system, the electric field can be described with two components, one of which is propagating in the posit ive x direction and one in the negative x direction. Each interface within the stack can therefore be described by a 2 x2 matrix, where r ij and t ij are the complex Fresnel reflection and transmission coefficients at each interface. This work only considers s polarized light, with the electric field perpendicular to the plane of incidence, giving Fresnel coefficients of: and with

PAGE 101

101 where i is the complex refractive index of layer i 0 is the refractive index of air (1.0), and 0 is the angle of incidence on the stack. Within each layer, the propagation is described by a layer matrix where d i is the thickness of layer i and The quantity is also known as the phase thickness, a representation of how much the phase of an incident light wave is changed by passing through that layer. This has importan t ramifications for interference, as shifts in the phase will change the spatial locations of constructive and destructive interference. Ultimately, the electric fields at each end of a system of m planar layers must be related by a transfer matrix S as in S is itself calculated according to The transmission and reflection coefficients for the entire structure can be describe with elements of the total transfer matrix S ,

PAGE 102

102 This is allowed by realizin g that light incident at upon the first layer propagating in the positive x direction will have no component propagating in the negative x direction at layer m+1 making Thus, the r eflection and transmission coefficients follow a logical formulation; the proportion of reflected light is the ratio of negatively propagating light to positively propagating light at the first layer, and the transmitted light is the ratio of positively pr opagating light at the final layer to positively propagating light at the first layer. The absorption of the multilayer stack is calculated according to where the transmittance and reflectance are respectively calculated as and This can be further modified to remove the reflections at the air/substrate and substrate/multilayer interfaces, leaving only the reflection and transmission (and, therefore, absorption) through the multila yer stack itself 1,164 This form is , where

PAGE 103

103 The transmission, reflection, and absorption ( ) for a multilayer stack of ITO/CuPc/C 60 /BCP/Ag (150/20/40/8/100 nm) calculated using this method is shown in Figure 4 10 As expected, there is very little tra nsmission through the structure ( given the metal cathode ) giving absorption and reflection an inverse relationship. Figure 4 10 Calculated transmission, absorp tion, and reflection of a multilayer CuPc/C 60 structure This is a fine situation to look at the behavior of the system as a whole, but the layers, necessitating a mod ification of the total system matrix to Now, each layer can be dealt with by its own transfer matrix, according to with

PAGE 104

104 Here, and refer to the left hand boundary of layer j A similar rewriting based on gives terms for the right hand boundary of layer j , with Within this framework, reflection and transmission coefficients for layer j can be defined from both the right and left hand directions in terms of elements of their respective partial system matrices: , Next a transmission coefficient that relates the incident wave to the wave propagating within layer j in the positive x direction at the interface of layers j and ( j 1) can be derived using the above equations as

PAGE 105

105 where The negative x propagating component can also be determined, These last two equations can be used to describe the electric fiel d at any distance x away from the interface of layers j and ( j 1), provided that 0 x d j In terms of the incident forward propagating wave, the final solution is The optical field at position x is proportional to The total energy absorbed at a position x per second is given by where c is the speed of light, 0 is the permittivity of free space, and j is the absorption coefficient of light within layer j This is a q uantity of great interest in OPV device simulation, as it represents the time averaged absorption of light and (assuming a 100% photon to exciton conversion) exciton generation Substituting the expression for into this yields where is the incident light intensity and and are, respectively, the absolute value and the argument of

PAGE 106

106 It is now clear that the energy absor bed at any position x within a layer j in a structure of planar layers is proportional to the summation of three components (from left to right inside of the bracketed portion of the proceeding equation): incident light propagating in the positive x direction, reflected light propagating in the negative x direction, and interference between these two waves. While there are numerous reflecting interfaces inside of a typical device structure, the primary reflecting surface is the metal cathode. Thus, the regions of high intensity are closely related to the distance from this surface with peaks coming at distances of where m e int eger order of interference. The proportional optical fields of two different structures, ITO (150 nm)/CuPc (20 nm)/C 60 (40 nm)/BCP (8 nm)/Ag (100 nm) and ITO (150 nm)/CuPc:C 60 (1:1) (60 nm)/BCP (8 nm)/Ag (100 nm) are shown in Figure 4 11 Figure 4 11 Transfer matrix calculated optical fields Both A) 1:1 (by we ight) mixed and B ) planar heterojunction CuPc/C 60 architectures are shown

PAGE 107

107 While there are differences between the optical fields within the bilayer and mixed devices, the general trend of remains true in either case; shorter wavelengths have intensity peaks closer to the cathode and the regions of high intensity for longer wavelengths are further away. In the case of the bilayer device, it is highly desirable to maximize the absorption and photocurrent in each layer by ensuring that regions of high intensity fall within the absorption band of the proper material and close to the heterojunction interface to maximize exciton diffusion efficiency. This can either be solved mathematically by c omputing the photocurrent generation from the optical field, or can be qualitatively shown by where G is the photon absorption/ exciton generation rate, E is the electric field intensity, and is the wavelength dependent absorptio n coefficient. The resulting exciton generation pl ot s for two different bilayer structures, CuPc/C 60 20/40 nm and 40/80 nm, are shown in Figure 4 12 Figure 4 12 Exciton generation plots in two different bilayer CuPc/C 60 devices.

PAGE 108

108 The discontinuities in the absorption field come from the diff erent absorption spectra in each layer. Absorption in this model is only assumed to occur in the active layers. Examination of Figure 4 12 gives a good qualitative feel for the resulting device performance. In the case of a 20/40 nm bilayer, there is strong optical intensity in both CuPc and C 60 near the heterojunction interface and the layers are reasonably thin, which will maximize charge collection. In a 40/80 nm thick device, however, the heterojunction int erface has been moved away from the bulk of exciton generation in C 60 the maximum intensity of absorption in CuPc has been decreased, and the layers are now much thicker, limiting charge collection efficiency. It is therefore reasonable to conclude that 20/40 nm is a preferred architecture over 40/80 nm. While this approach is fine for qualitative analysis, quantitative evaluations are needed for more accurate optimization. 4.2.3 Photocurrent Calculation With the calculated photon absorption/exciton g eneration field Q ( x ) already known, a steady state diffusion equation can be constructed and solved for each layer that is defined as a contributor to the photocurrent (i.e. donor and acceptor), Where and are respectively the e xciton diffusivity and lifetime within layer j and is the exciton density at position x Since a properly designed Type II D A heterojunction and metal electrodes serve as perfect quenching/dissociation site f or excitons, boundary conditions of can be set at these locations. Solving for the exciton diffusion current at the location of the D A interface with this condition gives

PAGE 109

109 The quantum efficiency is found by normalizing the current to the incident power, The total current for the cell is then computed by summing the individual quantum efficiencies of each layer and integrating with respect to the AM1.5G spectrum, outputting the one sun photocurrent regardless of the incident power intensity. Revisiting the previous CuPc/C 60 structure, the optimum value for C 60 thickness can be calculated as 40 nm, in agreement with qu alitative predictions (Figure 4 13). Figure 4 13 Transfer matrix c alculated short circuit current in a 20 nm CuPc/ x nm C 60 device.

PAGE 110

110 Note that this presumes no loss from both the exciton dissociation efficiency CT and the charge collection efficiency CC which results in an overestimation of short circuit current. Corrections can be made if values for charge collection length and dissociation efficiency are known by modifying the current generation equation to Where is the distance from the dissociation interface to a charge collecting interface (i.e. anode or cathode) and can vary based on charge carrier type, and is the charge carrier collection length. This is especially suitable for bulk heterojunction devices, where charge collection losses can be substantial. The transfer matrix method has been used extensively in Chapters 5 and 6 to explore the relationship between optical field and photocurrent be havior. 4.3 Review of Optical Simulation Techniques In this C hapter, the reader was presented with the theoretical background, practical implementation, and capabilities of two different optical simulation techniques, Monte Carlo ray optics and transfer m atrix wave optics. Both Chapters 5 and 6 use these techniques to analyze OPVs. Ray optics simulations are used extensively in Chapter 6 to investigate the geometric dependence of microlens arrays on absorption enhancement. This technique is readily imp lemented and is suitable for a variety of optical structures. The basic

PAGE 111

111 governing equations and iteration scheme were described. The Monte Carlo framework makes the program easily customizable and suitable for recording many different aspects of ray beha vior. To obtain quantitatively significant data, rays are generated to fit the simulated AM1.5G solar spectrum and the active layer mimics the absorption characteristics of real materials. In Section 4.2, the transfer matrix method for wave optics simulat ion was explained. This method is appropriate for simulating the optical field, including interference patterns that emerge within a stack of planar layers. The optical field can then be used to calculate absorption and, with certain assumptions, the sho rt circuit current in an OPV; this is useful to quickly optimize layer thicknesses in complex structures without the expense of device fabrication. Transfer matrix calculations are used in Chapter 5 to correlate photocurrent behavior with the internal opt ical field in a variety of devices. In Chapter 6, a slightly modified calculation is used to approximate the behavior of devices with and without microlens arrays

PAGE 112

112 CHAPTER 5 PHOTOCURRENT GENERATION AND TRANSPORT BEHAVIOR IN ORGANIC PHOTOVOLTAIC DEVICES 5.1 Overview Despite significant advances in device performance, there are still fundamental questions regarding the nature of photocurrent generation and transport in organic photovoltaic devices. In this chapter, the novel technique of synchronous pho tocurrent detection is used to study the impact of device architecture and thickness on charge transport behavior ( Section 5.2). Transfer matrix optical simulations are then used to correl ate measured photocurrent characteristics and the optical field wit hin the device active layer s ( Section 5.3). Finally, exciton dissociation behavior in bilayer OPVs is qualitatively described using a combination of optical simulation and synchronous photocurrent detection. 5.2 Effect of Heterojunction Architecture OPV device architectures can take several different forms. The three most common small molecule based architectures are (1) bilayer or planar, with neat layers of donor and acceptor and a single heterojunction interface, (2) bulk or mixed, where the donor and acceptor are uniformly mixed, creating dissociation interfaces throughout the active layer, and (3) planar mixed, where neat layers of donor and acceptor sandwich a mixed layer in an attempt to maximize photocurrent generation without compromising charge extraction. A schematic of these three arc hitectures is shown in Figure 5 1.

PAGE 113

113 Figure 5 1 Three different small molecule device architectures. The different placement of heterojunction interfaces in each of these architectures will have a significant impact on the locations of photocarrier generation. In a planar device, exciton dissociation will only efficiently occur at one plane within the active layers (discounting any effect of field assisted exciton dissociation). Free charge carriers in this d evice therefore come only from the heterojunction interface. In a mixed device, charge generation will be more uniform throughout the active layer due to the abundance of readily available exciton dissociation interfaces (neglecting the effect of optical interference for the time being). A planar mixed device is a hybrid between these two extremes: strong locations of carrier generation at the donor/mixed and mixed/acceptor interfaces with uniform generation inside the mixed portion of the active layer. Based on this surface analysis, the movement of photogenerated charge carriers will differ due to the large variability in generation location and the charge transport characteristics of each architecture A planar heterojunction will act as charge block ing

PAGE 114

114 interface, causing carrier pileup. For example, electrons in a bilayer CuPc / C 60 device can move freely from CuPc to C 60 but have a large energy barrier (~1 eV) preventing movement from C 60 to CuPc 29,165 This arrangement should cause a concentration gradient to emerge point ing away from the in terface, strongly influencing the diffusion motion of charge carriers. To probe this behavior, devices consisting of an electron donor of CuPc an d an electron acceptor of C 60 archetypical small molecule OPV materials were studied All devices were fabri cated on glass substrates prepatterned with a tin doped indium oxide (ITO) anode (15 / sheet resistance) using vacuum thermal evaporation ( Section 1.4.2). Before deposition, the substrates were sonicated in successive solutions of surfactant, deionized water, acetone, and isoproponal to remove any residual debris from the ITO surface. Finally, substrates were exposed to a UV ozone treatment for 15 minutes to destroy any remaining organic particles and increase the work function of ITO for better hole extraction 166,167 The planar heterojunction devices have a 20 nm thick CuPc donor layer and a 40 nm thick C 60 acceptor layer. Planar mixed devices were a sandwich structure of CuPc/CuPc:C 60 (1:2 by weight)/C 60 (10 nm/20 nm/30 nm). Mixed heterojunction devices had varying active layer thicknesses (60 to 120 nm) and mixing ratios. All devices were finished with an 8 nm thick exciton blocking layer of BCP and an aluminum cathode. CuPc was purified using high vacuum gradient zone sublimation ( Section 1.4.1) prior to use; high purity C 60 and BCP were used as purchased. The devices were characterized according to the techniques described in Chapter 3. All photocurrent measurements were performed under a white light bias of

PAGE 115

115 approximately one sun (100 mW/cm 2 ) with a monochromatic beam intensity on the o rder of 1 10 W/cm 2 conforming to ASTM standards 148 In synchronous photocurrent detection, three components are fed to the lock i n amplifier: dark current, photocurrent from the white light bias, and periodically varying photocurrent from the mechanically chopped monochromatic light. The last is the parameter of interest. Experimental results will be presented primarily as externa l quantum efficien c y (EQE), calculated as where V is the applied voltage bias, is the incident monochromatic wavelength, and h c and q respectively. The only difference between the photocurrent quantum efficiency and a standard EQE measurement is the voltage applied across the device. Under forward bias, the ITO anode is connected to the positive terminal, and positive current is referenced as positive when it flows from the anode to the cathode. Photocurrent characteristics for three different architectures (planar, planar mixed, and mixed 1:1 (by weight) CuPc:C 60 ) at = 650 nm are shown in Figure 5 2. In Figure 5 2A, the planar and planar mixed devices show monotonic decreases in photocurrent as bias is increased, from EQ E > 10% at V = 0 V to < 0.05% at V = 1 V. Beyond V = 1 V data was unreliable due to overloading and noise in the measurement system. The mixed HJ device shows markedly different behavior. There is a steady decrease in EQ E until it reaches a minimum value at V 0.55 V. After this point EQ E begins to increa se. Shown in the inset of Figure 5 2A is the phase recorded by the lock in amplifier. While there is no change for the planar or p lanar mixed devices (excepting noise at high biases), there is a 180 shift in the mixed device phase at the same voltage where EQ E

PAGE 116

116 is at a minimum. This indicates that the direction of photocurrent changes at this voltage, i.e. it changes from negative to positive at a specific zero photocurrent voltage V 0 In the planar and planar mixed devices the photocurrent stays negative regardless of the voltage. Figure 5 2 Example photocurrent characteristics for three different device architectures A) R aw dat a ( inset: lock in amplifier phase). B ) P hase adju sted photocurrent characteristics This behavior can be explained by examining the relative behavior of the drift and diffusion components of the photocurrent. The photocurrent at any point in the act ive layer can be simply defined as a summation of the drift and diffusion subcomponents, w here e is the electron mobility, q is the elementary charge, E is the local electric field, D is the electron diffusivity and n is the number of electrons A similar equati on can be written for holes, making the total photocurrent Drift occurs as carriers are swept by the electric field; diffusion acts to reduce concentration gradients.

PAGE 117

117 Figure 5 3 diagrams the relative contributions of drift and diffusion at small forward biases (or at the short circuit or negative biases), where the internal electric field points from the cathode to the anode, in planar and mixed architectures. The behavior of pl anar mixed devices is qualitatively identical to planar and will therefore not be included in further discussions. Figure 5 3 Relative contributions of drift and diffusion currents to the net photocurrent at small forward biases for two different device architectures. The direction of the internal electric field is indicated. This is the standard operating condition of an OPV device. In the planar device, all charges are generated at the heterojunction interface. After dissociation, the interna l electric field is able to sweep charges towards the electrodes for collection, giving a negative drift current. The diffusion current is also negative due to the concentration of cha rges at the dissociation interface but it will be a minor contributor compared to drift. The internal electric field is continually sweeping charges away from the interface, reducing the concentration gradient and driving force for diffusion. The net photocurrent

PAGE 118

118 for a planar device is therefore negative. In a mixed devic e, the lack of a singular heterojunction interface to block charge movement simplifies things. Charges will freely move with the electric field, giving a negative drift current. There is additionally no concentration gradient across the width of the acti ve layer, reducing the diffusion photocurrent This also leads to a negative net photocurrent. The situation changes greatly when the forward bias is increased enough to overcome the built in field across the activ e layer, depicted in Figure 5 4 Figure 5 4 Relative contributions of drift and diffusion currents to the net photocurrent at large forward biases for two different device architectures. The direction of the internal electric field is indicated. The planar device has very different behavior at large forward biases. Now, the internal electric field will sweep charge carriers towards the heterojunction interface, greatly increasing the concentration gradient. This will in turn increase the diffusion current to a sufficiently large value to m aintain a negative net photocurrent regardless of the applied bias. The experimentally measured photocurrent curves show a steady decrease as the forward bias is increased. Two mechanisms are identified to explain

PAGE 119

119 this. First, defect states exist within the band gap in these materials that can act as charge recombination centers. As the bias is increased and the concentration of electrons and holes increases the recombination current will increase accordingly, decreasing the driving force for diffusion. Second, an increased forward bias will cause sharper band bending within the active layer, increasing the efficiency of thermionic assisted tunneling acro ss the heterojunction interface and allowing the positive drift current to correspondingly increase. There is no great change in the relative contributions of drift and diffusion in the mixed device; diffusion is still a minor constituent of the total photocurrent and drift dominates due to the lack of charge blocking interfaces. However, the drift cu rrent has changed direction with the electric field, making the total photocurrent positive. This explains the 180 phase shift and subsequent increase in photocurrent amplitude experimentally measured. This explanation infers that the experimental zero photocurrent voltage V 0 must occur at a forward bias with a negligible internal electric field. A detailed study of the origin of V 0 and its wavelength dep endence is covered in the Section 5.3 5.3 Wavelength Dependent Photocurrent Behavior in Mixed Heterojunction Devices S ection 5.2 explained the general relationship between the drift and diffusion components of the total photocurrent with respect to device architecture. One of the important conclusions to arise from this concerns mixed heterojuncti on devices; namely, the internal electric field must be negligible at the experimentally measured V 0 voltage because drift dominates the total photocurren t. Note that Figure 5 2 only presents one

PAGE 120

120 wavelength of incident monochromatic light (650 nm). If V 0 is solely dependent on the internal electric field direction and drift current, there should be no difference between inversion voltages at differ ent wavelengths. As Figure 5 5 shows, this is not true. Figure 5 5 Photocurrent measurements of a 90 nm 1:1 CuPc:C 60 device at various wavelengths. Lock in amplifier phase is inset. In fact, V 0 falls at a wide range of voltages. There is also no obvious dependence of V 0 with respect to wavelength, i.e. the shortest and longest wavelengths (400 and 700 nm) do not correspond to the smallest and largest inversion voltages. Further, the order and magnitude of inversion voltages change with the active layer thickness. In a 90 nm thick 1:1 CuPc:C 60 device (Figure 5 6A ) the minimum inversion voltage is ~0.52 V with 450 nm illumination. Increasing the active layer thickness to 120 nm (Figure 5

PAGE 121

121 6B ) decreases the minimum inversion voltage to ~0.45 V, now with 500 nm light; V 0 at 450 nm illumination is ~0.48 V. Figure 5 6 Measured photocurrent values for two different mixed layer thicknesses at different wavelengths. A) 90 nm CuPc:C 60 (1:1 by weight). B) 120 nm CuPc:C 60 (1:1 by weight). To gain a clearer picture, linear interpolation of the measured photocurrent spectra near the inversion voltage can give a more precise value for V 0 at each wavelength. Figure 5 7 shows the resulting curves for three different layer thicknesses (60, 90, a nd 120 nm) of 1:1 CuPc:C 60 devices measured at 10 nm wavelength increments from 350 to 750 nm. After processing, it is clear that V 0 shows a definite relationship to wavelength. To explain the wavelength dependence, the details of carrier generation within the active layer must be studied. This is readily accomplished with transfer matrix optical simulations to replicate the optical field within the device, taking into account interference between incident and reflected light from the cathode (refer to Sec tion 4.2.2 for details on transfer matrix optical field calculations). At a 1:1 weight

PAGE 122

122 ratio CuPc and C 60 mix homogenously 10 making it suitable to approximate charge carrier generation with optical intensity. Figure 5 7 V 0 vs. wavelength for three different CuPc:C 60 (1:1) layer thicknesses. Figure 5 8 shows the calculated carrier generation fields for three different active layer thicknesses, 60, 90, and 120 nm. Also shown is a r eproduction of Figure 5 7 for comparative purposes. The weighted average centers of charge generatio n x c for each wavelength are shown in white on the transfer matrix plots. The values of x c at each wavelength show a clear relationship to the optical interference patterns formed within the active layer. Interference patterns in an OPV

PAGE 123

123 result in peak in tensity values at an approximate distance of away from the reflecting metal cathode, since it is the primary reflecting surface in the system. For example, in the 90 nm thick device, the maximum optical intensity/carrier generation at 450 nm shifts away from the cathode with the as wavelength in creases; x c undergoes a similar shift. However, at 400 nm, a second order interference peak emerges proximate to the anode, shifting x c closer to the ITO/organic interface. The average location of charge carrier generation is closest to the cathode a t 440 nm. Figure 5 8 Experimental V 0 vs. wavelength data and transfer matrix calculated charge generation fields for three different CuPc:C 60 (1:1) active layer thicknesses. The weighted center of charge generation is shown in white on the transfe r matrix plots. The ITO/organic interface is at the top of the plots; organic/cathode is at the bottom.

PAGE 124

124 The observed correlation between V 0 and x c is attributed to the diffusion current in these devices. In a CuPc:C 60 (1:1) mixed film, the hole mobility (~10 5 cm 2 /Vs) is approximately three orders of magnitude lower than the electron mobility (~10 2 cm 2 /Vs) 10 Therefore, charge extraction from the active layer will be limited by hole transport. When holes are generated close to the cathode, the corresponding hole diffusion current is low; holes must travel a long distance to reach the anode. Since the total diffusion current is reduced, a smaller positive drift current (and smaller forward bias) is required to reach a zero photocurrent value. In contrast, if x c is close to the anode, the diffusion current will be large, resulting in a larger V 0 A comprehen sive device model is needed to fully understand this relationship. One simple verification of this model is to examin e the effect of device thickness on V 0 A thicker active layer will require charge carriers to diffuse over a longer distance for collecti on, reducing the diffusion component of the photocurrent. In effect, V 0 will be reduced across the entire visible spectrum Figure 5 9 shows a plot of V 0 vs. wavelength for three different poly(3 hexylthiophene) (P3HT) : [6,6] phenyl C 61 butyric acid meth yl ester ( PCBM ) (1:0.8 weight ratio) bulk heterojunction OPVs. The thickness is varied from ~80 nm for a 1000 rpm, 18 mg/mL device to >150 nm for a 1000 rpm, 36 mg/mL active layer. The thicker devices show universally reduced V 0 voltages, reducing from ~0.82 V in a 1000 rpm, 18 mg/mL device to ~0.70 V in the thickest device, 1000 rpm, 36 mg/mL. These results add more support to the earlier conclusions regarding diffusion current and inversion voltage.

PAGE 125

125 Figure 5 9 V 0 vs. wave length for three different P3HT:PCBM devices. 5.4 Exciton Dissociation Behavior in Bilayer Organic Photovoltaics The photocurrent behavior presented in Section 5.2 outlined the general relationship between drift and diffusion currents in typical bilayer O PV devices. In a 20 nm CuPc/40 nm C 60 device the photocurrent remains negative up to the limits of the measurement setup (Figure 5 10A) However, increasing the CuPc donor thickness from 20 nm to 40 nm introduces inversion at some wave lengths, as shown i n Figure 5 10 B Specifically, the 40 nm/40 nm device has inversion under 350, 400, 550, 600, 650, and 700 nm monochromatic light and no inversion under 450 and 500 nm illumination.

PAGE 126

126 Figure 5 10 Measured photocurrent characteristics for planar heterojunction devices at different wavelengths. A) 20 nm/40 nm CuPc/C 60 B) 40 nm/40 nm CuPc/C 60 There is no obvious correlation between the absorption spectrum of CuPc and the inversion voltages, as inversion occurs at short wavelengths in addition to 600 700 nm. Other variations in active layer thickness show different patterns of inversion. For f urther analysis, transfer matrix simulations were used to examine a relationship between optical absorption and inversion (and the lack thereof). The results for three different thickness combi nations are shown in Figure 5 11 T he transfer matrix plots in Figure 5 11 are representative of exciton generation, not carrier generation. The weighted average centers of exciton generation are shown in white on each transfer matrix plot. Note that the experimental V 0 plots have a much wider range of inversion vo ltages in a single devices than the bulk heterojunction devices presented in Section 5.3. Additionally, the plateau from 450 500 nm presented in the 40 nm/40 nm device does not represent inversion. Instead, it is the voltage at

PAGE 127

127 which the minimum value is observed for each wavelength; since inversion is not observed, this is the maximum voltage scanned in the system. F igure 5 11 Transfer matrix optical simulations for three different bilayer devices. Also shown are experimentally calculated V 0 values. The weighted average centers of exciton generation are shown in white. There is no clear correlation between the direct exciton generation fields and the measured inversion voltage plots, as there was for bulk heterojunction devices. Since exciton gene ration and diffusion occurs in distinct layers in a bilayer device as opposed to a mixed medium in a bulk heterojunction device, it is important to take the exciton diffusion length into account. The exciton diffusion length in C 60 is much longer than in

PAGE 128

128 CuPc (40 nm vs. 10 nm) 1,66 ; this is why C 60 layers can be thicker than phthalocyanine layers and still show decent int ernal quantum efficiencies. A more useful anal ysis is presented in Figure 5 12 B where the center of exciton generation has been adjusted to show the distance from the heterojunction interface in terms of exciton diffusion length. Figure 5 12 Inversion voltage and exciton generation profiles for CuPc/C 60 planar devices with different thicknesses. A) Experimentally measured V 0 v alues. B) T ransfer matrix calculated average exciton generation distance from the heterojunction interface A relationship emerges upon analysis of this figure in comparison to the experimental V 0 values. If excitons are on average generated close to th e heterojunction interface, the inversion voltage is large, or there is no inversion. A s exciton generation moves further from the interface, the inversion voltage decreases. Since the percentage of interface dissociated excitons will increase as the ave rage generation location moves closer to the heterojunction, there is a correlation between increased interface dissociated excitons and increased inversion voltage.

PAGE 129

129 All excitons have a chance of being dissociated by the electric field within the device, instead of at the heterojunction interface. This is generally ignored as a contributor to photocurrent, since it is an extremely inefficient and minor contributor compared to interface dissociation. However, at large forward biases the photocurrent resu lting from interface dissociation will be limited to the diffusion component. This can be overcome by drift of carriers generated by field assisted dissociation, where charges will not be blocked by the interface. The drift, diffusion, and net photocurre nt for two different excitons created in the donor layer one dissociated by the interface and one by the electric field, are diagrammed in Figure 5 13 for a device at a large forward bias. Figure 5 13 The drift, diffusion, and net photocurrent for exc itons dissociated by either the interface or the electric field. The internal electric field is representative of a device at a large forward bias. The first exciton is created close to the interface and dissociated by the heterojunction This results in carrier pileup at the heterojunction interface and an increased diffusion current, as explained in Section 5.2. The net photocurrent from

PAGE 130

130 interface dissociated excitons will be negative, but with a small magnitude due to the thicker act ive layers, which reduce the long range concentration gradient. The exciton created far from the interface has very different characteristics. When it dissociates, it will introduce a hole into the donor HOMO and an electron into the LUMO. The lack of a strong concentration gradient away from the heterojunction int erface will reduce to diffusion co mponent of photocurrent. Because the electron is no longer blocked by the heterojunction interface and has a clear path to the anode, the drift current will instead be primary component of net photocurrent from the field dissociated exciton Thus, the net photocurrent from the field dissociated excitons will be positive. In this model, inversion will appear under illumination from a certain wavelength if the positive field assisted drift current is larger than the negati ve interface dissociated diffusion current. While this model is internally consistent and agrees well with experimental results, more study is needed to quantify the significance of field assisted exciton dissociation and its contributions of drift and dif fusion photocur rent. Potential future research opportunities are discussed in Chapter 8. 5.5 Review In this Chapter, the charge generation and transport processes in OPVs were probed to gain a better understanding of fundamental device processes. A sync hronous photocurrent detection method was developed to isolate the contribution of photogenerated current at differe nt wavelengths and applied device biases. First, the photocurrent characteristics of thin bilayer and planar mixed heterojunction devices we re explored and the photocurrent was found to always remain

PAGE 131

131 negative. As the photocurrent results from both the drift and diffusion motion of photocarriers, a model using the relative magnitudes of these two components was developed. At small forward bi ases, the drift current dominates in these architectures as the built in field sweeps charges away from the interface. When the bias is increased sufficiently and the direction of the int ernal electric field reverses, carriers are instead swept towards th e heterojunction interface, where the large energy band offsets result in carrier pileup. The increased concentration gradient leads to an increase in the diffusion current and the photocurrent therefore remains negative, but at a much smaller magnitude as leakage pathways create a small positive current that increases with the applied bias The lack of charge transport barriers in a mixed heterojunction device causes the drift current to dominate at almost all applied biases. At large forward bi ases, th e internal electric field and drift current direction will reverse and become positive when the applied bias overcomes the built in field However, at a certain narrow applied bias range the internal electric field will be negligible, enabling the diffusi on current to dominate. The specific voltages where photocurrent inversion occurs were found to be highly correlated to the average charge carrier generation location within the active layer as determined using transfer matrix wave optics calculations. A dditional studies provided evidence that field assisted exciton dissociation can be an important contributor to photocurrent under some conditions. In contrast to most bilayer heterojunction devices, where photocurrent is persistently negative, devices wi th thicker donor or acceptor layers display photocurrent inversion at certain wavelengths. Comparisons between the calculated optical field and measured inversion voltage s

PAGE 132

132 reveal a relationship between the proximity of exciton location relative t o the dis sociation interface and increased inversion voltage. Because there is no change in the heterojunction interface, t his suggests that field assisted exciton dissociation is introducing free charge carriers past the heterojunction interface, where they creat e a positive drift current. Ultimately, the characteristics observed here require more qualitative evaluation and simulation. The insights into photocarrier behavior from this study can be used to develop and verify the results of an advanced device simula tor. In Chapter 8, the basic framework of the simulator is introduced and discussed along with a proposed implementation scheme.

PAGE 133

133 CHAPTER 6 OPTICAL MANAGEMENT IN ORGANIC PHOTOVOLTAIC DEVICES 6 .1 Introduction and Background In the search for clean, renewable energy sources, photovoltaics (PVs) have emerged as a strong contender. Most of the existing PV technologies rely on inorganic active layer materials, such as crystalline silicon or III V compound semiconductors like GaA s. While these materials have excellent power conversion efficiencies, they are prohibitively expensive for widespread adoption. Thin film inorganic technologies like CdTe and CuIn x Ga (1 x) Se 2 (CIGS) are supplanting legacy technologies, but expense and me chanical robustness remain intrinsic issues with inorganic materials. OPVs offer a promising alternative. They offer mechanical durability, inexpensive, light weight modules and are compatible with economical, high throughput production techniques. With recent power conversion efficiencies exceeding 8%, performance is close to the values for market acceptance. Most of the efforts in increasing performance have focused on synthesizing new active layer materials, optimizing processing techniques and condi tions, and developing new device architectures, little work has been expended investigating potential enhancements from manipulating the coupling of incident light into the active layer. Regardless of the method taken, the primary goal is to ameliorate the fundamental tradeoff between light absorption and internal quantum efficiency in OPVs The external quantum efficiency is defined by these two terms, The internal quantum efficiency (IQE) itself has three subcomponents, the charg e collection efficiency, charge transfer efficiency, and exciton diffusion efficiency, so

PAGE 134

134 The charge transfer efficiency CT is approximately unity in a properly designed OPV heterojunction and will not be considered further The other two components vary with architecture. In a bilayer heterojunction, exciton diffusion efficiency ED falls off quickly with increased layer thickness. As a benefit, however, the thinner, neat layers that results have favorable charge transpo rt properties and C C is high A bulk heterojunction has the opposite issue ED is considered to be unity due to the intermixing of the two materials and the resulting ubiquitous heterojunction interface; however, poor internal morphologies and phase se paration disrupt charge transport and collection, decreasing C C In either case, the simplest method to increase IQE without altering the active layer material is to decrease layer thickness. However, this has a deleterious effect on light absorption, which can be described by the Beer Lambert Law for a ray optics based system where is the wavelength dependent absorption coefficient and d is the path length of light through the active layer (typically taken as the layer thickness). Given this, there are two ways to increase light absorption without increasing the active layer thickness: increase the absorption coefficient, which would req uire a different active layer material, or increase the path length in the layer. The first option is ultimately desirable but requires considerable time and expense for synthesis and reoptimization of processing parameters; the second can be accomplished by manipulating how light interacts with and propagates within the active layer.

PAGE 135

135 There have been few reports in the literature of optical designs that accomplish this in OPVs. Among them are using prism shaped substrates to induce total internal reflect ion 138 V aligned solar cells 135,136 and a patterned mirror and lens light trap 139,140 While these are certainly effective methods to increase light absorption, they are incompatible with high throughput, inexpensive processing techniques, or cannot be implemented effectively with large area, production scale devices. The solution presented here is a transparent microlens array (MLA) applied to the light incident surface of the device using a soft lithographic stampin g technique. When light strikes a MLA, two advantageous processes w ill occur (Figure 6 1 ). Figure 6 1 Schematic diagram of light interaction and path length through the active layer in a device with and without a microlens array (MLA). 1) Refraction o f light due to the curved lens surface. 2) Reflection of incident light into a neighboring feature. Without a MLA, the path length of light is just the layer thickness t a for normal incident light. Should the light not be absorbed on the first pass, it will reflect off the cathode and pass through the layer again with the same path length, and then be lost to

PAGE 136

136 the device. With a MLA, the situation is greatly changed. Figu re 6 1 p rocess (1) illustrates an incident light ray refracted by the non normal surface of the lens. The refraction will add an additional angular component to the light ray, increasing the path length to where is the angle rela tive to the surface normal; the degree of refraction is dependent on the specific angle of impact at the lens surface. Since the reflection coefficient will increase with the impact angle, at some angle a substantial percentage of the incident light will be reflected at the lens/air interface The periodic structure of the microlens array becomes advantageous in this situation, as the incident light can be reflected into a neighboring feature at an angle that is now favorable for transm ission (process (2 ), Figure 6 1 ). In either case, the path length has been increased and t he absorption probability will accordingly rise without any changes to the active layer. Further, because this is an optical effect based on interaction of incident light with a textu red substrate, path length enhancements are present regardless of the active layer material. In the subsequent sections, the performance of small molecule, polymer, and inorganic quantum dot/polymer hybrid devices will be detailed for various architecture s ( Section 6.3). There are also numerous geometric effects on enhancement performance which are probed using a ray optics simulation package ( Chapter 4) in Section 6.4, along with experimental results T he effects of optical field shifts in these devices upon application of a MLA are described in Section 6.5. Finally, Section 6.6 describes device architectures and characteristics that are favorable for MLA enhancement.

PAGE 137

137 6 .2 Microlens Array Fabrication Si nce the primary advantages of OPVs over conventional PV technology are their flexibil ity and inexpensive processing the MLA must be compatible with these characteristics. The MLAs fabricated for this work were produced from a poly(dimethylsiloxane) (PDMS) stamp patterned using convective capillary self assembly of 100 m polystyren e microspheres. The convective capillary self assembly method is a low c ost, repeatable laboratory scale technique to achieve high quality, large area (3 in 2 ), close packed micr osphere monolayers 168 To form the close packed monolayers, polystyrene microspheres in aqueous solution are dropped onto a silico n wafer. The wafer is then tilted to introduce convective flow of microsphe res to the liquid air interface. A s the water evaporates, capillary forces pull the microspheres together into a he xagona l close packed array. Other techniques to form arrays inc luding inkjet printed lenses 169 liquid crystal droplets 170 the melting of self assembled polymer microspheres 171 and photolithographic techniques 172,173 either cannot make arra y s of sufficient quality, are too expensive, or are incompatible with high throughput processing. The subsequent steps in mold fabr ication are shown in Figure 6 2 After a monolayer of desired size is assembled (6 2 A ) PDMS precursors are added in a 10: 1 weight ratio of polymer to curing agent and poured onto the monolayer. The PDMS is then thermally cured in a vacuum oven for 2 hours at ~60 C and the cured polymer is removed from the silicon substrate, leaving a polymer mold with the polysty rene micros pheres embedded (Figure 6 2 B ). The spheres are removed from the PDMS surface with a scotch tape liftoff technique, lea ving behind a concave mold ( Figure 6 2 C ). The mold fabrication process is identical to the method used in previous reports of

PAGE 138

138 light outc oupling enhancements in organic light emitting diodes using microlens arrays 174,175 Figure 6 2 Processing steps i n microlens array fabrication. A ) Convective capillary self assembly of polystyrene microspheres and addition of poly(dimethylsiloxane) (PDMS) precursors. B ) PDMS after curing, with polystyrene microspheres embedded. C ) Removal of microspheres with scotch tape liftoff, leaving behind a concave mold. D ) Addition of optical adhesive to concave mold and application to substrate, and the resulting array after mol d removal. Image courtesy of S. H. Eom parts reprinted with permission 174

PAGE 139

139 To form a MLA on the substrate, optical adhesive (Norl and Optical Adhesive #63) is added to the mold and brought into contact with the substrate. After allowing time for the optical adhesive to spread and infill the microlens pattern (approximately three minutes), 365 nm UV light is shone through the mold for three minutes to cure the adhesive. The mold is then gently removed, leaving behind a convex array (Figure 6 2D ) This process can be repeated several dozen times with a single mold before it fails due to UV induced embrittlement and degradation. The re sulting lenses have contact angles of (85 5) and a high packing factor. The cured adhesive has a refractive index of ~1.56, closely matched to the glass substrate, and is highly transparent past 360 nm. The lens application procedure typically requires exposing the organic layers to UV light while the optical adhesive cures, which can induce slight degradation in photovoltaic performance This can be circumvented for thermally evaporated devices by prefabricating arrays on the substrate prior to active layer deposition, but no satisfactory laboratory scale workaround was found for solution processed devices. 6 .3 Enhancement Characteristics When a MLA is added to a device there are noticeable increases in both short circuit current ( J SC ) and power conversion efficiency ( P ), with the majority of the P increase attributed to enhancement in J SC T here are only minor enhancements in the open circuit voltage ( V OC ) and fill factor ( FF ) corresponding to a higher photocurrent relative to dark current. T he current voltage (J V) characteristics for a thermally evaporated bilayer boron subphthalocyanine chloride (SubPc)/C 60 (12/40 nm) device with an 8 nm thick BCP exciton blocking layer are shown in Figure 6 3 along with

PAGE 140

140 external quantum efficiency respons e across the visible spectrum. Note that all devices in this study were fabricated on indium tin oxide (ITO) coated glass substrates that were successively cleaned in a solution of Tergitol surfactant, deionized water, acetone, and isopropanol, then expos ed to a UV ozone environment for 15 minutes prior to organic layer deposition. With the MLA, J SC is significantly increased from (5.4 0.2) mA/cm 2 from (4.6 0.1) mA/cm 2 approximately 17% enhancement. Combined with minimal increases in V OC and FF P also increases from (3.1 0.1)% to (3.7 0.1)%, a 20% increase. The device active area is approximately 2 mm x 2 mm, and a large area 2.25 cm 2 rear reflector is used to simulate the geometric characteristics of a large device. Exact device areas were measured using optical microscopy. The methodology and justification for using a rear reflector is explained fully in Section 6.5 Figure 6 3 Current voltage and external quantum efficiency characteristics of a SubPc/C 60 (12/40 nm) device with and wit hout a microlens array (MLA). Relative enhancement is indicated across the visible spectrum.

PAGE 141

141 Comparing the external quantum efficiency spectra of the device with and without a MLA shows that the enhancement is present across all wavelengths, though it is not constant. In general, the observed enhancement is greater in regions where absorption is relatively weak (e.g. > 620 nm) and smaller at wavelengths where the absorption is strong (e.g. at ~ 575 nm, the peak of SubPc absorption). This is in agree ment with predictions from the Beer Lambert law. For instance, a hypothetical 50% increase in the optical path length will result in only a 15% increase in A if 60% of the light is initially absorbed; if only 10% is absorbed initially, the increase in A is 46%. The reduced EQE of the device with a lens array near 350 nm is caused by absorption of the microlens material. We also observe a slight shift in the EQE spectrum with and without a lens array (most observable in the slight blueshift of the peak near 425 nm), attributed to an optical field shift within the active layer. The origin and ramifications of the optical field shift are discussed in Section 6.4 One of the most attractive features of MLAs is that their enhancement mechanisms are universal and work with a wide variety of active layer mate rials. Significant enhancements have been realized for very high efficiency poly mer:fullerene cells (Figure 6 4 ) and polymer nanoc rystal hybrid cells (Figure 6 5 ). High efficiency polymer device s with an active layer of 1:1 by weight poly(benzo[1,2 b:4,5 (5,6 difluoro 4,7 dithi en 2 yl 2,1,3 benzothiadiazole) (PBnDT DTffBT): [6,6] phenyl C 61 butyric acid methyl ester (PCBM) 137 show enhancements of P from (6.2 0.3)% to (7.0 0.4) %, a relative increase of approximately 13%.

PAGE 142

142 Figure 6 4 Current voltage and quantum efficiency characteristics for a high e ff iciency PBnDT DTffBT:PCBM OPV Inset: PBnDT DTffBT molecular structure. As with the SubPc/C 60 bilayer devices, a correlation between strong EQE/absorbance and reduced enhancement is observed in the EQE spectra. PBnDT DTffBT and PCBM were dissolved in dichlor o benzene and spin coated at ~120 C on top of a 40 nm thick PEDOT:PSS layer, and then solve nt annealed for 12 hours prior to thermal deposition of a 1 nm thick LiF interfacial electron extraction layer and an aluminum cathode. H ybrid PV cells wi th active layers consisting of the low gap polymer poly[2,6 (4,4 bis (2 ethylhexyl) 4H cyclopenta[2, 1 b;3,4 alt 4,7 (2,1,3 benzothiadiazole)] (PCPDTBT) and CdSe nanoparticles 176,177 (~7 nm diameter) in a 9:1 (by weight) ratio were spin coated from a 9:1 chloroform :pyridine solution, then annealed at 150 C for 30 minutes prior to aluminum cathode deposition. Devices with a solution processed ZnO nanoparticle based electron transport/optical spacing layer

PAGE 143

143 show a 18% increase in P ; those without ZnO have a 32 % increase (Figure 6 5) The ZnO was spin coated from an ethanol solution and annealed at 85 C for 15 minutes. Figure 6 5 Current voltage characteristics of hybrid PCPDTBT:CdSe polymer:inorganic nanoparticle devices with and without a ZnO optical spacing layer. H ybrid devices are noteworthy in that their increases in P are substantially greater than the enhancements in J SC Devices without ZnO have a much greater enhancement in P due to significant increases in fill factor of (6 1)% and (3 1)% in V OC For comparison, the device without ZnO only has increases of (3 3)% in fill factor and negligible increases in V OC This is attributed to the increased dark current in devices without ZnO. The higher dark current limits FF and V OC by beginning to dominate the photocurrent at lower voltages Therefore, the increased photocurrent induced by the microlens array therefor e has a greater effect on these devices than on a device that is not limited in this manner. The performance metrics for several different device architectures and active layer mate rials are summarized in Table 6 1.

PAGE 144

144 Table 6 1. Performance characteristic s for several different device architectures and active layer materials with and without microlens arrays (MLAs). Architecture J SC (mA/cm 2 ) Enhancement (%) P (%) Enhancement (%) w/o MLA w/ MLA w/o MLA w/ MLA SubPc/C 60 (1) 4.6 0.2 5.4 0.3 17 3.1 0.2 3.7 0.2 19 SubPc /C 60 (2) 1.5 0.1 2.3 0.1 53 0.9 0.1 1.4 0.1 56 P3HT:PCBM (1) 9.2 0.5 10.5 0.5 14 3.4 0.2 3.9 0.2 15 P3HT:PCBM (2) 5.9 0.3 7.4 0.4 25 1.9 0.1 2.4 0.1 26 PCPDTBT:CdSe (1) 9.1 0.5 10.3 0.5 13 2.8 0.1 3.3 0.2 18 PCPDTBT:CdSe (2) 7.5 0.4 9.0 0.5 20 2.2 0.1 2.9 0.2 32 PBnDT DTffBT:PCBM 11.8 0.6 13.1 0.7 11 6.2 0.3 7.0 0.4 13 SubPc/C 60 (1): 12 nm SubPc/40 nm C 60 SubPc/C 60 (2): 12 nm SubPc/80 nm C 60 P3HT:PCBM (1): 100 nm mixed layer thickness, no ZnO optical spacing P3HT:PCBM (2): 100 nm mixed layer thickness, 45 nm ZnO PCPDTBT:CdSe (1): 85 nm mixed layer thickness, 20 nm ZnO PCPDTBT:CdSe (2): 85 nm mixed layer thickness, no ZnO PBnDT DTffBT:PCBM: 140 nm mixed layer thickness. 6.4 Optical Field Optimization 6.4.1 Bilayer Heterojunction Devices It was noted in the previous section that the EQE spectrum of a SubPc/C 60 bilayer OPV exhibits a slight blueshift at shorter wavelengths in addition to the expected increases from MLA enhanced absorption. This is consistent with changing optical interference patterns within the device active layer. Under normal illumination, regions of high optical intensity will emerge at distances of away from the cathode/organic interface, where m integer order of interference between the incoming wave and the reflected wave off of the cathode. The location of the intensity peak is related to the phase shift that the

PAGE 145

145 reflected wave undergoes as it travels through the organic layers Each layer therefore has a certain phase thickness, defined as where i is the complex refractive index of the layer i is the angle of wave propagation through the layer, d i is the layer thickness, and is the wavelength of light. Therefore, if the ang le of incidence is increased, the phase shift is correspondingly altered and the locations of constructive and destructive interference will move further away from the cathode. This effect can be modeled using tran sfer matrix optical simulations 157 by making the key assu mption that the MLA refracted light can be modeled as non normal incidence light on a series of planar layers. Figure 6 6. Calculated optical fields for SubPc/C 60 (12/60 nm) OPVs at nor mal (0 ) and 30 incidence. Figure 6 6 shows the calculated optical field i ntensities within a SubPc/C 60 (12/60 nm) bilayer OPV for normal (0 ) and 30 incidence. The effects of the phase thickness

PAGE 146

146 shift are most observable in the altered intensity profile at > 700 nm and the intensity tail bleeding into the SubPc layer at 575 nm with 30 incidence. Since 60 nm is larger than the optimized C 60 thickness (40 nm) for SubPc devices, shifting the 5 75 nm high intensity region into the SubPc laye r has significant ramifications, as the maximum SubPc absorption occurs in this wavelength region. This suggests that the optimum device thickness for a device under normal illumination is not necessarily the optimum thickness for a device with a MLA. The optical field shift has a significant effect on bilayer devices, where carrier generation can be highly dependent on the location of optical intensity peaks relative to the het erojunction interface, especiall y as strongly absorbed wavelengths are moved in and out of a layer. Figure 6 7 shows this situation for a SubPc/C 60 device (12 nm/ y nm), where the C 60 thickness is varied to serve as an optical spacing layer When the device has a thinner C 60 layer, rela tive enhancements in J SC and P are less than a device with a thicker layer. For example, a 40 nm C 60 layer has an enhancement of 19% in P ; this rises to 56% when the C 60 thickness is increased to 80 nm. With an 80nm thick active layer, the optical field within the SubPc layer under normal incidence is not optimized specifically, the SubPc absorption maximum at 575 nm is still predominantly within the C 60 layer. By adding the MLA and inducing the opti cal field shift, this intensity region is pushed into the SubPc region, greatly increasing the photocurrent generation from that layer. The inset of Figure 6 7 shows the transfer matrix calculated EQE response within the SubPc layer at normal (0) and 30 incidence. The optical field shift in this case is extremely favorable, significantly increasing response near the SubPc absorption peak.

PAGE 147

147 Fig ure 6 7 Effect of varied C 60 thickness on J SC and P enhancement for a SubPc/C 60 (12/ y nm) device with and without a microlens array. The results of transfer matrix simulations are also shown. Inset: calculated quantum efficiency response within the SubPc layer for y = 80 nm at normal (0 ) and 30 incidence, showing the effect of the optic al field shift. Figure 6 7 also shows calculated enhancements in short circuit current using transfer matrix simulations; these are in general agreement with the experimental data. To obtain qualitatively significant data the calculated currents for seve ral different incident angles must be averaged together, as a MLA do es not equally refract all light to a single inc ident angle. Monte Carlo r ay optics simulations were used to obtain a proper angular distribution. The details of the ray optics simulator are explained in Chapter 4. To determine the angular distribution, SubPc:C 60 (1:4 by weight) active layers of various thicknesses

PAGE 148

148 were simulated with 90 contact angle, 100 m lens arrays. The device, lens array, and illumination areas were treated as in finite by applying periodic boundary conditions. Mixed films were used in place of bilayer SubPc/C 60 films for computational simplicity T o correlate results, the total active layer thicknesses were compared (i.e. 20 nm SubPc : C 60 is comparable to 12 nm S ubPc/10 nm C 60 ) Whenever a ray is absorbed its path length through the active layer is recorded; the incident angle was back calculated from the ratio of path length to film thickness. Fi gure 6 8 shows the incident angle distributions for absorbed rays in thin (20 nm) and thick (120 nm) SubPc:C 60 devices. The results follow expectations from the Beer Lambert law. For a very thin device, the proportion of light absorbed with an increased path length will be more significant. In a thicker device, a subst antial percentage of light with no angular component will already be absorbed, making the relative contributions at with longer path lengths less significant. Figure 6 8 Distributions of light incident angle upon the active layer in SubPc:C 60 (1:4 by w eight) films calculated by ray optics simulations.

PAGE 149

149 From these results, the relative contributions for rays sorted into five different angle bins ( with bin centers of 9.6 22.4 35.2 48 and 60.8 ) are calculated and used to weight the short circuit current values calculated using tr ansfer matrix simulations. While this is an adequate method to approximate the relative enhancements with and without a MLA using the transfer matrix method a more rigorous implementation is needed for full quantitative simulations. While the relative enhancements increase with C 60 thickness due to the favorable optical shift, the total efficiency do es not. Because a thicker C 60 layer will also reduce ED and CC t he optimum C 60 thickness with a MLA remains at the same value as that of a bare device, 40 nm (Figure 6 9) A dedicated electron transport/optical spacing layer could circumvent this issue. Figure 6 9 Power conversion efficiency P for SubPc/C 60 (12/ y nm) with and without a microlens array (MLA).

PAGE 150

150 6.4.2 B ulk Heterojunction Devices The optical field shift has a different impact on bulk heterojunction devices, where exciton dissociation is not limited to a single plane within the active la yer. Instead, the shift can move optical intensity peaks out of the a ctive layer entirely, negating any benefit from enhanced path length. For example, the enhancements in a poly(3 hexylthiophene ( P3HT ) :PCBM (1:0.8 by weight) device with varying active layer thicknesses show two distinct trends, depending on whether the device is semi transparent (and therefore removed from the optical interference effect) or conventional, with a refle cting metal cathode (Figure 6 10 ). Figure 6 10 Short circuit current enhancements for P3HT:PCBM bulk heterojunction devices with and wi thout a microlens array. Conventional devices have a reflecting metal cathode; semi transparent devices have a cathode of ITO/ZnO and a vacuum deposited trilayer anode of MoO 3 /Au/MoO 3

PAGE 151

151 The conventional device has an anode of tin doped indium oxide (ITO) and a vacuum deposited aluminum cathode. The semi transparent device, however, has an electron selective cathode of ITO/ZnO nanoparticles and a vacuum deposited anodic trilayer electrode 178,179 of MoO 3 /Au/MoO 3 ( 3/10/40 nm) An aluminum coated piece of glass is attached to the top of the semi trans parent device with optical adhesive to still the active layer, strong interference patterns will not emerge and govern the optical field within the device. Instead, r ay optics approximations are appropriate. In both cases, the P3HT:PCBM active layer was spun coat from a 1:0.8 (by weight) chlorobenzene solution and annealed inside a nitrogen filled glovebox. The measured enhancement for the semi transparent device ther efore closely follows predictions from the Beer Lambert law, that the increase in path length is exponentially more significant as the active layer thickness decreases. A 140 nm thick device has current enhancements of ~10%, while an active layer thicknes s less than 35 nm yields enhancements above 30%. In a conventional device, the opposite is true enhancement monotonically decreases with the active layer thickness. In a thin device, the optical shift introduced by the MLA begins to move strongly absorb ed wavelengths out of the active layer, negating any additional benefit from increased path length. Enhancements vary from ~12% for a 170 nm thick device to a negligible 3% with a 60 nm thick active layer. As with the bilayer device, however, the optical field shift can be understood and compensat ed for to increase enhancements. In this case, a solut ion processed ZnO nanoparticle 177,180 182 layer can be added as an electron transport/optical spacing layer

PAGE 152

152 between the cathode and the active layer, decoupling layer thickness and position with in the optical field. Now, the optical field shift can have one of two effects, based on ZnO thickness. First, (considering only the first order interference peaks) it can shift short wavelength intensity peaks out of the ZnO layer and into the active layer. Second, longer wavelength peaks can be moved out of the active laye r. Figure 6 11 Effect of ZnO optical spacer thickness on mixed P3HT:PCBM devices. A ) Enhancements in short circuit current and power conversion efficiency for P3HT:PCBM/ZnO (100/ y nm) devices. The transfer matrix calculated enhancements are also sho wn. B ) Compariso n of short circuit current with and without a microlens array and the relative enhancement. When the ZnO thickness in a P3HT:PCBM is varied, a periodic oscillation of J SC emerges, roughly following the spacing of optical intensity peaks (Figure 6 11B). S tarting with an optimized P3HT:PCBM layer thickness of 100 nm, the highest observed values of J SC are present for ZnO thicknesses y of 0 and ~130 nm, which corresponds well to the expected spacing between the first and second optical intensity peaks for = 500 nm, in the middle of the P3HT:PCBM absorption region according to the relationship (assuming n = 1.8, the first intensity peak occurs ~ 70 nm away

PAGE 153

153 from the cathode). The minimum in optical intensity will occur between the inten sity peaks, i.e. at ~ 105 nm away from the cathode, explaining the observed minimum in J SC with a 45 nm thick ZnO optical spacer (the experimentally measured maxima and minima will vary from the calculation due to the wide band of P3HT:PCBM absorption). The relative enhancements in J SC and P (Figure 6 11 A ) also vary according to the optical intensity profile. The greatest enhancement is observed when the optical intensity within the active area is weak ( y = 45 nm) and the optical field shift works positively, moving high intensity regions into the active layer; J SC is increased by ~24%. Conversely, a negligible enhancement of ~1% is measured when the optical field is already favorable for device performan ce ( y = 130 nm) and the optical field shift moves strongly absorbed intensity regions out of the active area. When there is no ZnO ( y = 0) enhancements are still modest at ~13% in J SC as the high intensity regions are not completely shifted out of the ac tive layer. The transfer matrix cal culated enhancement values in Figure 6 11 A are in very rough qualitative agreement with experiment, first showing increased enhancement and then redu cing as ZnO thickness increases, but the experimentally observed charac teristics are not effectively reproduced. This underscores the need for more rigorous modeling to fully understand and exploit the optical shift. 6 .5 Geometric Effects Microlens arrays introduce several dependencies on the geometric relationship between the illumination, device, and lens array areas. These effects arise because the arrays diverge light in a periodic pattern over the illumination area. The even dispersion creates the favorable characteristic that enhancements increase with device area as

PAGE 154

154 loss mechanisms reduce. The effect of device area on J SC enhancement for SubPc/C 60 (12/60 nm) bilayer devices is shown in Figure 6 12 for both large area illumination and device area illumination. Considering a device where the illumination area is equal to the device area, a portion of light near the edge of the device is refracted and diverted outside of the active are a. W ith small, laboratory scale devices, the perimeter length is relatively long compared to the total device area and a large proportio n of incident light will be lost. As the device active area becomes larger, the proportion of light lost around the edges decreases accordingly. Figure 6 12 Effect of device active area on relative enhancement with SubPc/C 60 (12 nm/60 nm) devices. diameter (large area illumination) or masked off so only the active area was exposed to light (device area illumination). Ray optics simulated absorption enhancement in a 70 nm SubPc:C 60 (1:4) device is i ncluded.

PAGE 155

155 Extending the concept, an infinitely large device would not exhibit this effect. Lost light can be partially compensated for by making the illumination area larger than the device area, where light can be incoupled from outside of the device area This effect has important ramifications for production scale devices. In the laboratory, device areas are typically kept small for ease of fabrication and characterization. Commercial devices, however, need to be as large as possible to maximize power generation and minimize module cost. Any optical enhancement technique should accordingly be compatible with large areas, as MLAs are. Also shown in Figure 6 12 are calculated absorption enhancements using Monte Carlo ray optics simulations (simulation m ethodology is described in Chapter 4) for large (20 mm x 20 mm) and device area illumination with a lens array area of 100 mm x 100 mm. The simulated enhancements agree very well with experimental results, both qualitatively and quantitatively. For this reason, studies of lens geometry and relative array, device, and illumination areas were simulated rather than performed experimentally The first point of investigation is the relationship between device area and illumination area. Figure 6 13 A shows the relative absorption enhancements for two simulated 70 nm SubPc:C 60 (1:4) devices with either 1 mm 2 or 1 c m 2 active areas. The illumination area was varied in each case, and the enhancement is plotted as a function of the device area to illumina tion area ratio. The array width was set at 100 times the device width in each case to mimic completely isolated devices. When the device area is small, there is a significant increase in enhancement when the illumination area is changed from one to four times the device area, rising fr om a decrease in absorption of

PAGE 156

156 80% to no enhancement. Increasing the illumination area to 100 times the device area gives an absorption enhancement of 25%. For the large 1 cm 2 device, there is a much more gradual increase in enhancement as the illumination area changes, rising from 11% for equal device and illumination areas to 48% when the illumination to device area ratio is 100. This underscores the importance of perimeter loss for different device sizes. Since the smaller device has much more significant leakage around the device edge, compensation by increasing the illumination area has a much more significant effect. Figure 6 13 Simulation results of different geomet ric arrangements. A ) Absorption enhancement vs. device area/illumination area for SubPc:C 60 (1:4 by weight, 70 nm active layer thickness) devices of either 1 mm 2 or 1 cm 2 device area. B ) Percentage of absorbed rays sorted by generation loca tion for diffe rent device sizes (i llumination area = 20 mm x 20 mm, array area = 100 mm x 100 mm ) The perimeter effect can also be examined by comparing the generation locations of absorbed rays as a func tion of device area (Figure 6 13 B ). When the device area is small (1 mm 2 ), less than 40% of the absorbed rays are initially generated outside of the

PAGE 157

157 device area, demonstrating how significant light leakage around the device perimeter is with a small area device. As the device area increases, the percentage of intr a device generated absorbed rays increases monotonically to ~85% in a 1 cm 2 device. The behavior as area increases closely follows the experimental data in Figure 6 12 indicating that light leakage is the primary loss mechanism. Because the enhancements of small are a devices are compromised by light leakage, the experimental results presented in Section 6. 3 and Section 6.4 are small area devices with a large area aluminum rear reflector to mimic the geometric arrangement of a large area device. The refle ctor is insulated from the cathode by a ~100 nm thick spin coated layer of Cytop fluoropolymer ; the Cytop layer is cured in a high vacuum environment prior to aluminum deposition This situation is analogous to sampling a 2 mm x 2 mm area in a 2.25 cm 2 de vice. The amount of reflecting area around a device has a marked impact on enhancement. This can be readily simulated by enforcing and varying periodic boundary conditions (PBCs) in the ray optics simulations. When PBCs are enforced, a box defined as the main simulation area is effectively surrounded by identical imaginary boxes on all borders. Should a ray exit from one side of box, and identical ray is created to enter from the opposite side, as if it came from a neighboring box. For these simulations the periodic boundary area is taken as the array (or substrate) area. In Fi gure 6 14 the device and illumination areas are held constant at 1 cm 2 and the array area (periodic boundary area) is varied to simulate an array of devices with various packing densities. When the array area and device area are the same size, the device area is effectively infinite since no light can be lost due to transmission through

PAGE 158

158 the substrate (the cathode covers the entire simulated substrate area); calculated absorption enhancement is ~50% for a SubPc:C 60 (1:4, 70 nm) device. Enhancement falls exponentially with increasing device spac ing, event ually decreasing to 15% with a d evice to device spacing of 180 m m. This has important implications for commercial photovoltaic m odules, which are comprised of large arrays of small er devices. With a sufficiently close packed array, enhancements can approach the theoretical limit for an infinitely large device. Figure 6 14 Effect of device spacing on simulated absorption enhancement in a 70 nm thick SubPc:C 60 (1:4) device. Periodic boundary conditions are enforced with a variable periodic area to mimic device arrays with different device packing densities. Independent of the relative sizes of the device, illumination, an d substrate areas, the microlens geometry itself can affect the degree of enhancement. The two identified enhancement mechanisms, refraction and reflection, will be dominant at different

PAGE 159

159 locations on the lens surface, based on the probability of transmiss ion (considering only the initial ray lens interaction for normally incident light). Near the crown of the lens, the incident angle will be low enough to still have a high probability of transmission; refraction will dominate in this region. As the incid ent angle increases (i.e. as the intersection point moves from the crown to the base of the lens), reflection becomes a more significant component. By varying the contact angle of the simulated lens array (Figure 6 15 ) the relationship between incident angle and surface reflection can be probed. Figure 6 15 Effect of contact angle variations on simulated enhancements in 70 nm thick SubPc:C 60 (1:4) films. Devices are treated as infinite using periodic boundary conditions. The device, illumination, and substrate areas are treated as infinite in these simulations to isolate the effect of lens geometry on enhancement. Further, the MLA is

PAGE 160

160 close packed regardless of the contact angle, so the same non planar surface is available in all cases. As shown i n Figure 6 15 the simulated absorption enhancement in 70 nm thick SubPc:C 60 (1:4) films decreases with the contact angle. The decrease is not strictly monotonic, however. The highest calculated enhancement of ~52% is obtained for = 80 . In a microlen s with = 90, rays incident near the lens base have an extremely high probability of reflection, but as the incident and reflection angles are equal, the path length increase will be negligible. When the contact angle is reduced slightly to 80 these r eflected rays will have a more significant angular deflection, increasing absorption enhancement. As the contact angle reduces further, the reflection mechanism first reduces due to a reduction in surface area that will allow a reflected ray to strike a n eighboring feature; below = 45 this mechanism does not occur. At smaller contact angles the de gree of refraction of incident r ays is decreased. Eventually, a close packed array with = 1 demonstr ates an enhancement of only 12%. Finally, an important performance metric for any optical enhancement technology is its performance versus the incident angle of illumination, as t he solar illumination angle varies through the day for a photovoltaic module without a solar tracking system. Since OPVs are inten ded as a low cost PV solution, expensive solar tracking systems are not desirable. To test the behavior of MLAs under variable illumination angle, 1 cm 2 SubPc/C 60 (12/40 nm) were fabricated and illuminated under uniform, ~5 mW/cm 2 white light. Two differ ent illumination areas were considered: illumination area > device area (Figure 6 16 A ) and illum ination area = device area (6 16 B ).

PAGE 161

161 Figure 6 16 Performance of 1 cm 2 SubPc/C 60 (12/4 0 nm) devices under 5 mW/cm 2 white light illumination with a variable incident a ngle. A ) Illuminatio n area greater than device area. B) I llumination area equal to device area. Relative enhancements with and without a microlens array (MLA) are shown. Dashed lines are cos predictions.

PAGE 162

162 When the illumination area is much greater than the device area, there is a monotonic decrease in J SC as the tilt angle is increased, closely following a predicted relationship. This suggests compensation between inc reased surface reflectivity and increased path length through the active layer due to angled illumination, allowing the change in illumination intensity to dominate. The relationship deviates at high angles for a device without a MLA as surface reflection dominates. With a MLA, the measured J SC outperforms the cos dependence. Enhancements are mostly constant for < 50. However, it sharply increases from ~15% (when < 60) to ~90% at = 80. The drastic increase in enhancement for large incident an gles is attributed to the curved microlens surface. The refl ectivity of light in this region is highly sensitive to the exact incident angle. Consequently, the small changes in the surface normal due to lens curvature have a significant impact on the ref lection probability and the MLAs reduce the reflection of very high angle incident light. When the illumination and device areas are equal, the predicted cos relationship is adjusted to account for the increase in shadowed device area and the same trends in J SC reduction are observed. The relative enhancement behavior differs, however. It steadily decreases until = 60, after which it increases sharply to ~35% at = 80. The monotonic enhancement decrease when < 60 is a byproduct of light leakage around the device perimeter, which will increase r elative to a bare device with th increasing illumination angle. When the illumination area is much larger t han the d evice area, the extra light lost is compensated by incoupling more light from outside of the active area.

PAGE 163

163 6.6 Ideal Architectures for Enhancement After testing several different architectu res and active layer materials, general guidelines can be s et forth for suitable candidate systems for microlens enhancement. Microlens arrays enhance the photocurrent generated by a device, allowing for either a larger J SC or an equivalent J SC with thinner device layers, which can have a beneficial effect on fil l factor and reduced material usage when compared to a bare device of a set thickness. Additionally, the device must b e limited by light absorption, as a device that already absorbs a substantial portion of the incident light will not benefit from additio nal path length. An ideal material system will therefore have a relatively low J SC and large V OC For example, SubPc/C 60 with V OC in excess of 1.1 V and small baseline J SC of 4.6 mA/cm 2 for an optimized device, fits this pattern and shows significant enh ancement of ~ 20% in P for a 12 nm / 40 nm system. Material systems that are constrained with low fill factors and open circuit voltages due to high dark currents can also have significant enhancements in these regions from the improved photocurrent component. T he hybr id PCPDTBT:CdSe system (without ZnO) exemplifies this, with enhancement in J SC of 20% and P enhancement of 32 %. Given the se enhancement characteristics, tandem organic photovoltaic devices could be ideally su ited for microlens arrays. These devices have high open circuit voltages and can be designed to capture different portions of the solar spectrum in the front and back cells, creating wide spectral regions suitable for absorption enhancement. Further, the carrier recombination zone can serve as a built in optical spacing layer that can be tuned to exploit the optical field shift. There are significant

PAGE 164

164 challenges in optimizing the optical field in a tandem device, however. Additional experiments must be p erformed to determine the suitability of MLAs for tandem devices. 6.7 Review In this Chapter, soft lithographically stamped microlens arrays were demonstrated to enhance power conversion efficiency enhancements of10 60%. Enhancement is due to an increase in the average path length that light travels through the active layer, increasing the absorption probability without having to increase the active layer thickness, change the materials, or implement a more complicated heterojunction architecture. The pa th length is increased by refraction and reflection processes at the array air interface due to the curved, periodic nature of the arrays. Because the enhancement is due to light interaction with the air/substrate interface, it is applicable to all organ ic material systems. Small molecule, high efficiency polymer:fullerene, and hybrid inorganic nanoparticle/organic polymer systems all show enhancements. Dependencies of enhancement on the internal optical field intensity distribution were revealed and ex ploited to further increase enhancement by altering layer thicknesses and using optical spacers. Further, Monte Carlo based ray optics simulations were used to understand t he geometric dependence of microlens array enhancement s on device packing density, a ctive area, and illumination area ; enhancement increases with all of these. Finally, general guidelines were described for devi ces that are well suited to enhancement from MLAs : high V OC low J SC and low dark current. Because MLAs were demonstrated to ha ve several properties that align with commercial device requirements: enhancement increases with area, it is present at all

PAGE 165

165 incident angles, and soft lithography is compatible with roll to roll processing. For this reason, steps should be taken to complet e development and commercialize this technology. In Chapter 8, several necessary steps and suggestions for further development are presented.

PAGE 166

166 CHAPTER 7 BIFUNCTIONAL ORGANIC OPTOELECTRONIC DEVICES 7.1 Fundamentals of Organic Bifunctional Device s One of the great advantages of organic semiconducting materials is their versatility. One niche application of this is a device that can function as either an OPV or an OLED based on operat ing conditions, dubbed here as a bifunctional organic optoelectron ic device, or BFD. A BFD with respectable performance in both operating modes would have many different applications, but this has yet to be achieved. Examples of BFDs in the literature have been sparse 183 190 but two different trends have emerged, with device design either based off of an OPV or OLED architecture. General ly, an OLED based BFD has respectable light emission performance but poor photovoltaic efficiency (typically only under UV illumination) 186,187 OPV based BFDs are the opposite 185,191 This is due to the opposing operation principles o f these two processes (Figure 7 1). Figure 7 1 Basic OPVs and OLEDs device architectures, with charge carrier behavior diagrammed

PAGE 167

167 OLEDs are designed to efficiently move charge carriers from the electrodes into the interior of the device, where they recombine in an emitting layer. Conversely, OPVs are intended to efficiently quench photogenerated exciton s in the interior of the device and move charge carriers to the electrodes without transport barriers. This is the primary tradeoff that must be overcome in BFDs to have appreciably efficient operation in both modes A promising architecture based on a 5, 6,11,12 tetraphenylnaphthacene ( rubrene ) /C 60 heterojunction has been previously presented by Pandey and Nunzi 185 They report a large open circuit voltage of ~0.9 V (consistent with other bilayer rubrene/C 60 OPVs 84 ) and respectable photovoltai c performance. OLED efficiency was poor, but the turn on voltage was ~1 V, half of the optical gap of the fluorescent rubrene emitting layer. This architecture functions as a BFD for three reasons. First, there is no barrier to charge carrier extraction from the rubrene/C 60 interface giving high OPV power conversion efficiency. Second, the rubrene/C 60 heterojunction can efficiently dissociate C 60 excitons, but cannot dissociate rubrene excitons. Thus, luminescence excitons are not immediately quenched after formation near the interface. Third there is an A uger assisted energy up conversion process (Figure 7 2) that occurs at the interface, enabling electrons to overcome the large energy offset between the C 60 and rubrene LUMOs This is responsible fo r the half gap turn on voltage. The Auger up conversion process has also been o bserved in other organic and hybrid inorganic nanoparticle/organic polymer systems 182,192 There are three processes that can describe the Auger assisted injection process for light emissio n :

PAGE 168

168 1. Non radiative recombination of an electron (in C 60 ) and a hole (in rubrene) across the heterojunction interface 2. Transfer of the recombination energy to an electron in C 60 and excitation of that electron over the rubrene/C 60 energy barrier. 3. Recombination and light emission in rubrene. Figure 7 2 Auger up conversion process for half gap electroluminescence in rubrene/C 60 BFDs, with HOMO and LUMO energy levels indicated. 1) Non radiative recombination across the interface. 2) Energy transf er to and excitation of an electron in C 60 3) Recombination and light emission. Without the up conversion process, prohibitively large voltages would be required to excite electrons into the rubrene LUMO Unfortunately, Auger recombination is an extreme ly inefficient excitation method. At best, half of the injected charge carriers are wasted in non radiative recombination; in actuality, a larger percentage will be lost.

PAGE 169

169 The rubrene/C 60 system has many opportunities for optimization to improve emissive e fficiency while still maintaining respectable photovoltaic performance. This architecture was chosen as the starting point to increase total BFD efficiency. 7.2 Novel Device Architectures for Phosphorescent Bifunctional Devices The large energy barrier t o electron injection in a rubrene/C 60 BFD means that this will be limiting factor in OLED performance. Therefore, either t he injection barrier must be lowered or more efficient use of whatever electrons are injected through the Auger process must be made The former approach requires a change in the emissive/donor material, acceptor, or both. Because the rubrene/C 60 architecture already has several beneficial exciton energy properties and respectable performance as an OPV, this route was not taken. Inst ead, several different architectures and material systems were investigated to improve control of the BFD emissive proper ties and increase performance. I mproved OLED performance was ultimately not realized, but greater control and understanding of excito n behavior in BFD devices was obtained. There are several possible avenues to more efficiently use excitons in the donor/light emitting layer. First, rubrene is a fluorescent material and has inherently low emission effic iency ; replacing it with a phosphorescent emitter should increase performance Second, there is a possibility that electrons are traveling through the rubrene layer and escaping into the ITO ano de before they form excitons The latter assumption can be quickly tested. Patterned indium tin oxide substrates were cleaned in successive baths of surfactant, deionized water, acetone and isopropanol, and then dried under a nitrogen flow. The substrates were then treated by UV generated ozone, then a 40 nm thick layer of pol y(3, 4

PAGE 170

170 ethylenedioxythiophene):poly( styrenesulfonate) (PEDOT:PSS) was spin coated in air and annealed for 10 minutes at 140 C. Then, the substrate was placed in a high vacuum thermal evaporator and a 10 nm thick electron blocking layer of N, N' bis(naphth alen 1 yl) N,N' bis(phenyl) benzidine (NPB) a 35 nm thick rubrene emissive layer a 40 nm thick C 60 acceptor, a 8 nm thick bathocuproine (BCP) exciton blocking layer, and an aluminum cathode were deposited (Figure 7 3 A ). Note that all other devices in th is section were also thermally evaporated. Figure 7 3 Effect of an NPB electron blocking layer on a rubrene/C 60 BFD. A ) device architecture, with HOMO LUMO energy levels. B ) Current voltage beha v ior under 1 sun illumination. C ) Current Luminance Vo ltage characteristics. D ) OLED external quantum efficiency.

PAGE 171

171 The high LUMO level of NPB confines electrons within the rubrene layer, while its HOMO level is approximately the same as that of rubrene, so there should be a limited impact on hole transport a nd extraction and OPV efficiency. NPB will additionally serve as an exciton blocking layer to prevent quenching at the anode/organic interface. The NPB electron confinement layer has almost no impact on photovoltaic performance, as anticipated. In both cases, the power conversion efficiency P is virtually the same, ~0.9% with only minor diffe rences in fill factor (Figure 7 3B ) When driven as an OLED, there is a large increase in quantum efficiency, from EQE 0.12% without NPB to EQE 18% with NPB at a brightness level of 50 nits (cd/m 2 ) (Figure 7 3D ) While this is not the desired order of magnitude improvement it indicates that there is some benefit to an electron blocking layer. The remainder of this S ection focuses on integratin g phosphorescent emitters into a rubrene/C 60 BFD for more efficient use of injected electrons. Fluorescent emitters, such as rubrene, can only radiatively recombine singlet excitons. Phosphorescent emitters, however, can radiatively recombine triplet exc itons, which are produced at a ratio of 3:1 versus singlets 20 Additionally, singlet excitons produced in a phosphorescent emitter are converted to triplets via spin orbit coupling, allowing approximately 100% of generated excitons to contribute to light emission 23 26 Attempts to c ompletely remove rubrene and replace it with the phosphorescent emitter platinum octaethylporphine (PtOEP) were not successful. PtOEP has similar HOMO LUMO alig nment to rubrene and PtOEP/C 6 0 heterojunctions result in respectable OPV performance 84 However, b ecause PtOEP triplet excitons can be dissociated at the hetero junction interface, there is extremely inefficient light emission

PAGE 172

172 from such a structure Any excitons that for m proximate to the interface will be dissociated back into free charge carriers. Further, neat PtOEP films exhibit strong triplet triplet annihi lation behavior, where two triplet excitons c ombine and create a singlet 193 Doping PtOEP into the wide bandgap host material 4,4' N,N' dicarbazole biphenyl (CBP) to reduce triplet triplet quenching does not improve BFD performance The rubrene/C 60 hetero junction is therefore crucial: rubrene excitons are not efficiently dissociated at the interface, allowing for emissive recombinati on. Figure 7 4 BFDs using a doped phosphorescent emissive layer. A ) Basic phosp horescent BFD architecture and B ) photovoltaic performance under 1 sun illumination. C ) Emission spectra of BFDs with and without PtOEP doped into the NPB electron blocking layer. D ) External quantum efficiency behavior for two representative devices.

PAGE 173

173 A more successful approach is to place PtOEP immediately next to a thin electr on injecti on layer of rubrene, which will isolate PtOEP triplets from immediate dissociation at the interface (Figure 7 4A). The resulting structure is ITO/ x /rubrene/C 60 /BCP/Al (20/10/40/8 nm, excluding the electrodes), where x is either neat NPB or NPB:PtOEP (10% by weight). In this architecture, there is no change in photovoltaic performance between devices with and without P tOEP (Figure 7 4 B ). There is a slight alteration in the emission spectrum with the addition o f PtOEP, as shown in Figure 7 4C but the emiss ion pattern is still predominately that of rubrene. Despite the addition of phosphorescent emission, total luminance and quantum efficiency actually decrease with the add ition of PtOEP (Figure 7 4D ). The reduced luminance is caused by energy transfer to the lowest energy state in the system, the rubrene triplet state, T 1, rub enabled by the strong spin orbit coupling behavior of PtOEP. Figure 7 5 shows a basic Jablonski diagram with the energy transfe r pathways indicated. Figure 7 5 Jablonski diagram of exciton energies for a system containing NPB, PtOEP, and rubrene.

PAGE 174

174 Because the triplet energy of PtOEP is in between the singlet and triplet energies of rubrene, phosphorescent emission from PtOEP will be quenched as excitons instead move to the non emissive rubrene T 1 One option to prevent this is to replace PtOEP with a different phosphorescent material with T 1 > 2.1 eV, such that the lowest energy state in the system would be the fluorescent rubrene singlet (once excitons are transferred to t he rubrene S 1 state they are forbidden to move to the rubrene T 1 level) One material that fits this requirement is the common green phosphorescent dopant fac tris (phenylpyridine) iridium (Ir(ppy) 3 ), which exhibits ~2.4 eV T 1 emission. However, devices wi th and without 10% Ir(ppy) 3 doping into NPB have no appreciable differences in either photovoltaic or light emitting performance (Figure 7 6 ) Either the LUMO level of Ir(ppy) 3 is too high relative to rubrene (2.8 vs. 3.2 eV), limiting electron injection, or the Ir(ppy) 3 T 1 to rubrene S 1 transition is quenching phosphorescent emission. In either case, this is not a viable option. Figure 7 6 Ir(ppy) 3 phosphorescent BFD architecture and emission spectra for devices with and without doping into NPB. No changes in perf ormance we re observed.

PAGE 175

175 Instead of changing materials, a new architecture was developed to isolate PtOEP from rubrene to prevent exciton energy transfer to the non emissive rubrene T 1 state. This is accomplished by inserting a thin neat layer of NPB between the doped PtOEP region and rubrene such that electrons can tunnel through to reach PtOEP but exciton energy transfer back to rubrene cannot occur. This archit ecture is depicted in Figure 7 7 In addition to the thin NPB layer between rubrene and PtOEP, an addition al neat NPB layer is added to prevent quenching at the ITO /organic interface Figure 7 7 Adjusted phosphorescent BFD architecture. Thin NPB layer s isolate excitons on PtOEP while still allowing electrons to tunnel into PtOEP for excit on formation.

PAGE 176

176 The NPB blocking layer functions as an effective exciton barrier due to the short triplet triplet exchange distance via the Dexter process, which has a typical range of a few nanometers ( Section 1.3.3 ). Tunneling can occur at longe r distances than this, making the layer both electron permeable and exciton blocking. Figure 7 8 A shows the photovoltaic perfo rmance under 1 sun illumination. T here is no significant change as the NPB buffer thickness is altered with power conversion efficiencies of approximately 1% for all structures Figure 7 8 Photovoltaic and LJV characteristics of phosphorescent BFDs. A ) Current voltage characteristics of ITO/NPB/NPB:PtOEP (20%)/NPB/r ubrene/C 60 /BCP/Al ( 5/15/ y /10/40/8 nm) devices under 1 sun illumination B ) Luminance current voltage characteristics. While the injection behavior of each device is similar regardless of the NPB buffer layer thickness, d evices with no buffer or a thin 2 nm thick buffer have reduced luminance, indicating that quenching is an issu e in these structures (Figure 7 8B ). However, emission is predominately from the phosphorescent PtOEP T 1 state in these tw o devices, detailed in Figure 7 9A Note that the PtO EP doping ratio is 20% in this structure compared to 10% in the previous example (Figure 7 4C ), giving increased

PAGE 177

177 650 nm phosphorescent emission without a buffer layer. When the buffer layer thickness is increased to 5 nm and beyond electron tunneling in to the PtOEP layer is reduced and rubrene emission dominates. However, the emissive efficiencies of these devices are all extremely low. A 2 nm thick NPB buffer has predominately phosphorescent emission, but peak efficiencies are only EQE = 0.024%, lum = 0.04 cd/A and power efficiency P = 0.10 lum/W at ~1 nit. For comparison, a fluorescent device with a 10 nm thick buffer has EQE = 0.022%, lum = 0.06 cd/A and power efficiency P = 0.13 lum/W at the same brightness. Figure 7 9 Light emitting characteristics of phosphorescent BFDs. A ) Emission spectra for ITO/NPB/NPB:PtOEP (20%)/NPB/rubrene/C 60 /BCP/Al (5/15/ y /10/40/8 nm) devices, with B) luminous efficiency, C) power efficiency, and D ) external quantum efficiency characteristics for the same devices.

PAGE 178

178 Efficiencies roll off drastically for all structures at higher luminance levels. The comparable values of EQE indicate that phosphorescent emission is inherently more efficient, but the reduced luminance/increased quenching reduce efficiency. While the developed architecture is successful at isolating emission on PtOEP, the compromise between increased energy back transfer to non emissive states and reduced electron tunneling prevent an increase in OLED performance. Additional opti mization attempts with this structure were not successful in appreciably increasing performance. It is promising that all of the experimental architectures maintained approximately equivalent photovoltaic performance, with power conversion efficiencies of approximately 1%, demonstrating good understanding and control of the requirements for efficient BFD photovoltaic characteristics. U ltimately, either a fundamental redesign of BFD architecture or the incorporation of new active layer materials is requ ire d to realize a device that has respectable efficiencies as both an OPV and an OLED. 7.3 Requirements for Efficient Bifunctional Device Design Efficient BFDs are intrinsically difficult to achieve. As descr ibed in Figure 7 1, the operati n g mechanisms for power generation and light emission are inherently opposed. In Section 7.2 new device architectures were developed in an effort to shift emission to phosphorescent materials, but no real improvements in performance were seen. From this failure, however, several materials selection guidelines are now understood and can be described to support future advances in BFD performan ce and architecture development.

PAGE 179

179 For respectable photovoltaic performance: No barriers to hole or electron transport away from the hetereojunction/dissociation interface (high fill factor) Absorption over a large portion of the solar spectrum (high short circuit current) R educed dark curr ent (high open circuit voltage) For efficient light emission, the device requires: Reduced hole or electron injection barriers (depending on which side of the heterojunction has the light emitting component) Isolation of emissive excitons from the heterojunction interface, or Inability of the heterojunction interface to dissociate emissive excitons A phosphorescent or high efficiency fluorescent emitter preferably doped into a wide bandgap material to improve emissive efficiency. These necessities make materials selection extremely complex Fullerenes, such as C 60 and C 70 have high electron mobilit y and perform well as an acceptor for photovoltaic applications, but their deep LUMO level creates a large barrier for charge injection into most emissive materials. Changing to other accepting materials, such as 3,4,9,10 perylene tetracarboxylic bis benz imidazole ( PTCBI ) or poly((9,9 dioctylfluorene) 2,7 diyl alt [4,7 bis(3 hexylthien 5 yl) 2,1,3 benzothiadiazole] diyl) ( F8TB T) will reduce the injection barrier, but at the cost of reduced photovoltaic performance. F8TBT can actually function as ei ther a donor or acceptor, depending on which materials it is paired with, making it a very versatile option 194 This research as a lso made clear that fluorescent dopants are preferred to phosphorescent emitters. Phosphorescent materials allow energy transfer to non emissive states, necessitating extra steps to isolate the emitter from materials with lower energy triplets

PAGE 180

180 In Chapte r 8, a redesigned architecture is proposed that could reduce the necessary tradeoff between OPV and OLED performance using inorganic nanopa rticles to selectively tune the electric field and potential energy barrier to electron injection at the heterojuncti on interface

PAGE 181

181 CHAPTER 8 CONCLUSIONS AND FUTURE WORK 8.1 Photocurrent Generation and Transport In Chapter 5, the charge generation and transport processes in OPVs were explored using a novel characterization technique, synchronous photocurrent detection to isolate the contribution of photogenerated current at different wavelengths and device biases It was found that the device architecture and heterojunction structure have a significant impact on the resulting photocurrent behavior In thin bilayer and planar mixed heterojunction devices, the photocurrent always remains negative. At small forward biases, the drift current dominates in these architectures as the built in field sweeps charges away from the interface to be collected at their respective ele ctrodes When the bias is increased sufficiently and the direction of the internal electric field is reversed, carrier pileup at the heterojunction interface due to the large charge transport barriers results in an in crease in the diffusion current large enough to compensate for positive photocurrents created via leakage pathways at the interface T he photocurrent therefore remains negative, but at a much smaller magnitude. The lack of charge transport barriers in a mixed heterojunction device causes the drift current to dominate at almost all applied biases. At large forward biases, the photocurrent direction reverses and becomes positive as the internal electric field switches directions relative to the built in field. However, at a certain narrow appl ied bias range the internal electric field will be negligible, enabling the diffusion current to dominate. The distribution of charge carriers within the active layer then becomes

PAGE 182

182 important, and correlations between the wavelength dependent optical field and the inversion voltage ( at which the photocurrent changes from negative to positive ) were observed. Additional studies were performed on bilayer devices with layer thicknesses larger than typical optimal values. While an optimized bilayer device has a persistent negative photocurrent, the thicker devices display photocurrent inversion at certain wavelengths. Correlation between the optical field and inversion voltage reveal a relationship between the proximity of exciton location relative to the dissoc iation interface and inversion voltage. Namely, exciton generation closer to the heterojunction interface is observed with increased inv ersion voltage. This indicates that field assisted exciton dissociation, which can introduce free charge carriers past the heterojunction interface, becomes a significant contributor to photocurrent at large forward biases. While these explanations agree qualitatively with experimental photocurrent behavior, there is a need for simulation of photocurrent behavior to quant itatively determine the relative contributions of the drift and diffusion current. This can be accompli shed by combining the Gummel iteration method 195 with transfer matrix optical simulations to calculate electrical and optical field profiles, respectively. This approach has been applied to OPVs before 158,196 as a basic device simulator for the optimization of device thickness and charge carrier mobility, not with the goal of examining photocurrent transport and behavior. A basic implementation scheme would use the Gummel method to iteratively calculate the electric field, charge carrier distribution, and dark current profile within the device through manipulations of Poisso

PAGE 183

183 are applied to determine the optical field profile within the device. From this, the total charge carrier profile can be determined by calculating absorption and exciton dissociation probabilities. T he exp erimental technique of synchronous photocurrent detection can be approximated by moving a small amount of charge carriers for each wavelength (in proportion with the AM1.5G spectrum) and recording the amount of charge carriers that arrive at each electrode via either drift or diffusion. The total current curve can then be taken as the summation of the dark, drift, and diffusion currents. Finally the experimentally determined photocurrent behavior provides an extra level of verification that previous simu lation techniques have lacked. This approach to simulation should be highly accurate, and will enable holistic simulation of device behavior. 8 .2 Optical Management in Organic Photovoltaic Devices Most of the advancements in OPV performance have come through the synthesis of new active layer materials and development of new architectures. While important, these are expensive, time consuming routes. In Chapter 6, a third improvement method was expl ored, optical management. Enhancements of 10 60% in power conversion efficiency were demonstrated using soft lithographically stamped microlens arrays on the light incident surface of the device. This serves to increase the average path length that light travels through the active layer, increasing the absorption probability without having to increase the active layer thickness, change the materials, or implement a more complicated heterojunction architecture. The path length is increased by refraction a nd reflection processes at the array air interface.

PAGE 184

184 This enhancement was demonstrated to be present regardless of the active layer materials. Small molecule, high efficiency polymer:fullerene, and hybrid inorganic nanoparticle/organic polymer systems al l show enhancements. The degree of enhancement is strongly dependent on the optical intensity profile within the device. Microlens arrays change the direction of the incident light which will alter the interference patterns in the active layers. This w as exploited to increase the level of enhancement. T he geometric dependence of microlens array enhancement was also explored. Ray optics simulations revealed relationships between the device packing density, active area, and illumination area, with enhanc ement increasing as all of these factors increase. MLAs were demonstrated to have several properties that are matched with commercial device requirements: enhancement increases with area, it is present at all incident angles, and soft lithography is compa tible with roll to roll processing. Finally, general guidelines were described for devices that are well suited to MLA enhancement: high V OC low J SC and low dark current. There are numerous steps that can be taken to further study optical management in OPVs. First, a robust simulation method that can accurately couple millimeter scale substrate and lens dimensions with nanometer scale device active layer thicknesses is needed. The separate transfer matrix and ray optics simulations presented in Chapter s 4 and 6 are not suitable for complete simulation of MLAs, especially in understanding and predicting the change in optical interference patterns. One such method is a finite difference time domain (FDTD) calculation 197 FDTD is based on numerical iteration of s in

PAGE 185

185 space. A wide variety of mes h sizes can be accommodated within a single simulation, allowing simultaneous calculation of optical field propagation within the lens array and the device layers. Numerous commercial software packages are available. Additional work should also be done to scale up MLA production techniques to speeds and yields suitable for a manufacturing environment. This will require installation of a small scale roll to roll production line with a stamping and curing assembly or a system designed to imprint t he MLA pattern directly into a plastic substrate using a heated mold. Though stamped and cured lens arrays were used in this work a directly textured substrate should show the same enhancement characteristics provided that the microlens shape and packing factor are consistent between the two methods. The proposed suitability for commercial development makes demonstration of functioning high throu gh put roll to roll produced MLAs key. MLAs can theoretically be adapted for use in inorganic thin film PVs, such as CdTe or CuIn x Ga (1 x) Se 2 Because the refractive indices of these materials are much larger than in organic materials (i.e. n > 2.5 vs. n 1.7 1.8) the lens array itself must be conducted out of a higher index material. As it is not practical to have a transparent polymer with n > 2, high index inorganic nanoparticles (i.e. BaTiO 3 ) could be mixed with a conventional transparent polymer, such as optical adhesive, to increase the refractive index, since particles with diameters much less than the wa velength of light will only increase the effective refractive index, not scatter light This method is limited by the processability and refractive index of the nanoparticles. Alternately, transparent inorganic materials c ould be used to form the MLAs, b ut this requires drastically different fabrication techniques compared to soft lithography stamping. Additionally,

PAGE 186

186 b ecause the MLA refractive index is now much larger, a low index polymer could be pla ced on top of the MLA to serve as a grade d index antire flection coating and an encapsulation layer. A final proposed route is the development of textured rear reflectors for OPVs. In this case, the texture will be a microlens array in either a concave or convex orientation relative to t he device, as shown in Figure 8 1. Figure 8 1 Device schematics for concave and convex microlens array rear reflectors. Light is first incident on the substrate in this geometry. In this situation, the reflector will scatter light to both increase the average path lengt h during subsequent passes through the active layer and induce total internal reflection for a portion of the reflected light. The effect could be further enhanced when combined with a MLA on the light incident surface. Preliminary investigations have sh own that reflectors increase performance relative to a planar reflector by 5 10% in

PAGE 187

187 polymer based devices, and no statistically significant difference between concave and convex reflector geometries has been observed. 8 .3 Bifunctional Organic Optoelectronic Devices Chapter 7 concerns BFDs. These devices can function as either light emitting or power generating, depending on operating conditions. However, this prese nts a fundamental challenge. The operating mechanisms of OLEDs favor retainin g charges in the interior of the device to maximize emissive recombination, while OPVs require efficient movement of charges from the interior dissociation interface to the electrodes. To ameliorate this tradeoff, a promising bilayer rubrene/C 60 architect ure was modified to shift emission from rubrene (a fluorescent emitter) to PtOEP ( a phosphorescent emitter ) to make more efficient use of the electrons that are able to overcome the large injection barrier between rubrene and C 60 This proved to be a comp licated task and revealed many design requirements for future bifunctional devices. Initial attempts to simply replace rubrene with PtOEP were unsuccessful. While photovoltaic performance was respectable, PtOEP emission was quenched at the interface wit h C 60 Isolating PtOEP from C 60 with a thin layer of rubrene enables phosphorescent emission, but energy transfer through PtOEP to the previously forbidden rubrene triplet state reduces total luminance and efficiency. Finally, a novel architecture that is olates PtOEP from rubrene using a thin electron permeable exciton blocking layer of NPB was developed. This was successful in inducing predominately phosphorescent emission but the tradeoff between electron injection and exciton blocking limited total ef ficiency. Ultimately, substantial redesign is required to create an efficient BFD. This could either encompass better material

PAGE 188

188 selection to lower injection barriers while still allowing for exciton dissociation, or radically different architectures such as using inorganic nanoparticles to create favorable band bending for either OLED or OPV operation Figure 8 2. Schematic diagram of localized effects of ferroelectric nanoparticles polarization on the potential barrier for electron injection at a ru brene/C 60 interfa ce. Ferroelectric nanoparticle polarization is shown ab ove each set of band diagrams. A) I ncreased potential for thermionically assisted tunneling electron injection due to the thinned barrier between C 60 to rubrene, for OLED operation. B ) Interfacial band diagram under opposite ferroelectric nanoparticle poling for OPV operation Instead of relying solely on Auger assisted injection, ferroelectric nanoparticles can be used to selectiv ely induce band bending at the heterojunction interface and promote thermionic assisted tunneling at a lower applied bias than is otherwise possible (Figure 8 2A ). This arrangement should result in increased power efficiency and luminance from the device as a greater amount of carriers is injected than would be otherwise If the nanoparticles are polarized in the opposite direction (Fig ure 8 2 B ), the

PAGE 189

189 altered interfacial electric field should help sweep charges away from the interface and towards the electrodes, a potential benefit for photovoltaic performance. There are challenges in ferroelectric nanoparticle synthesis and incorporation at the interface. The nanoparticles must be small, to localize the induced electric field to the interface layers, but still large enough to allow significant pola rization. There could also be different effects on band bending based on whether the nanoparticles are placed exactly at the interface, or slightly embedded into either of the layers. An additional possibility is to isolate emission on inorganic quantum d ots embedded in a solution processed layer and electrically insulated by an encapsulating organic ligand. Previous studies have found that tunneling barriers can be tuned by the application of a large electric field, allowing charge carriers to move from an organic molecule to the inorganic nanoparticle 198 If the nanoparticles are dispersed in a bulk heterojunction OPV, the trouble some charge blocking heterojunction interface would be removed. This could be an effective way to separate the OPV and OLED operation modes but tuning the hole and electron tunneling barriers will be demanding. BFDs are a revolutionary technology, but ex tensive investigation is required to realize their potential. The work presented here was a first step in understanding the material requirements for future development. 8.4 Afterword I n the past three decades, organic photovoltaic devices have emerged f ro m a niche research topic to a prospective competitive technology in the photovoltaic marketplace. However, much work remains to make OPVs a viable alternative energy source. This dissertation presented two detailed studies to advance OPV technology:

PAGE 190

190 fi rst, investigation of photocurrent generation and transport behavior, including the development analytical techniques to enable future studies, and second, a method to universally improve performance in OPVs by manipulating the behavior of incident light that is both inexpensive and compatible with a commercial production environment. This study has also given insight into the necessary device characteristics and design requirements to fully exploit the optical enhancement effect a promising avenue to push device efficiencies past 10%. Inexpensive, rugged, and high efficiency organic photovoltaics have the potential to revolutionize solar energy around the world and diligent research effort will realize this goal.

PAGE 191

191 LIST OF REFERENCES 1. Peumans, P., Yakimov, A. & Forrest, S.R. Small molecular weight organic thin film photodetectors and solar cells. J. Appl. Phys. 93 3693 3723 (2003). 2. Forrest, S.R. The path to ubiquitous and low cost organic electronic appliances on plastic. Nature 428 911 918 (2004). 3. Shirota, Y. Organic materials for electronic and optoelectronic devices. J. Mater. Chem. 10 1 25 (2000). 4. Shaheen, S.E., Ginley, D.S. & Jabbour, G.E. Organic Based Photovoltaics: Toward Low Cost Power Generation. MRS Bull. 30 10 19 (2005). 5. Xue, J. Perspectives on Organic Photovoltaics. Polym. Rev. 50 411 (2010). 6. Sun, Y. et al. Management of singlet and triplet excitons for efficient white organic light emitting devices. Nature 440 908 912 (2006). 7. Scott, J.C. & Bozano, L.D. Nonvolatile Memory Elements Based on Organic Materials. Adv. Mater. 19 1452 1463 (2007). 8. Sundstrm, V. Femtobiology. Annu. Rev. Phys. Chem. 59 53 77 (2008). 9. Brtting, W. Physics of organic semiconductors (Wiley VCH: 2005). 10. Rand, B.P., Xue, J., Uchida, S. & Forrest, S.R. Mixed donor acceptor molecular heterojunctions for photovoltaic applications. I. Material properties. J. Appl. Phys. 98 124902 7 (2005). 11. Bozano, L., Carter, S.A., Scott, J.C., Malliaras, G.G. & Brock, P .J. Temperature and field dependent electron and hole mobilities in polymer light emitting diodes. Appl. Phys. Lett. 74 1132 1134 (1999). 12. Choulis, S.A. et al. High ambipolar and balanced carrier mobility in regioregular poly(3 hexylthiophene). Appl. Phys. Lett. 85 3890 3892 (2004). 13. Brenner, D.W. et al. A second generation reactive empirical bond order (REBO) potential energy expression for hydrocarbons. J. Phys.: Condens. Matter 14 783 802 (2002).

PAGE 192

192 14. Bisri, S.Z., Takenobu, T., Takahashi, T. & I wasa, Y. Electron transport in rubrene single crystal transistors. Appl. Phys. Lett. 96 183304 (2010). 15. de Boer, R.W.I., Gershenson, M.E., Morpurgo, A.F. & Podzorov, V. Organic single crystal field effect transistors. Phys. Status Solidi A 201 1302 1331 (2004). 16. Takeya, J. et al. Very high mobility organic single crystal transistors with in crystal conduction channels. Appl. Phys. Lett. 90 102120 (2007). 17. Yamagishi, M. et al. High mobility double gate organic single crystal transistors wi th organic crystal gate insulators. Appl. Phys. Lett. 90 182117 (2007). 18. Pope, M. & Swenberg, C.E. Electronic Processes in Organic Crystals and Polymers (Oxford U: New York, 1999). 19. Sun, S. S. & Sariciftci, N.S. Organic Photovoltaics: Mechanisms, Materials, and Devices (CRC Press: Boca Raton, FL, 2005). 20. singlet triplet ratio in a semiconducting organic thin film. Phys. Rev. B 60 14422 (1999). 21. Baldo, M.A. & Forrest, S.R. Transient analysis of organic electrophosphorescence: I. Transient analysis of triplet energy transfer. Phys. Rev. B 62 10958 (2000). 22. Sokolik, I., Priestley, R., Walser, A.D., Dorsinville, R. & Tang, C.W. Bimolecular reactions of singlet excitons in t ris(8 hydroxyquinoline) aluminum. Appl. Phys. Lett. 69 4168 (1996). 23. Baldo, M.A. et al. Highly efficient phosphorescent emission from organic electroluminescent devices. Nature 395 151 154 (1998). 24. 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). 25. Kawamura, Y. et al. 100% phosphorescence quantum efficiency of Ir(III) complexes in organic semiconductor films. Appl. Phys. Lett 86 071104 (2005). 26. Williams, E.L., Haavisto, K., Li, J. & Jabbour, G.E. Excimer Based White Phosphorescent Organic Light Quantum Efficiency. Adv. Mater. 19 197 202 (2007).

PAGE 193

193 27. Luhman, W.A. & Holmes, R.J. In vestigation of Energy Transfer in Organic Photovoltaic Cells and Impact on Exciton Diffusion Length Measurements. Adv. Funct. Mater. 21 764 771 (2011). 28. Shaw, P.E., Ruseckas, A. & Samuel, I.D.W. Exciton Diffusion Measurements in Poly(3 hexylthiophene). Adv. Mater. 20 3516 3520 (2008). 29. Peumans, P. & Forrest, S.R. Very high efficiency double heterostructure copper phthalocyanine/C[sub 60] photovoltaic cells. Appl. Phys. Lett. 79 126 128 (2001). 30. Najafov, H., Lee, B., Zhou, Q., Feldman, L.C. & Pod zorov, V. Observation of long range exciton diffusion in highly ordered organic semiconductors. Nat. Mater. 9 938 943 (2010). 31. Forrest, S.R. Ultrathin Organic Films Grown by Organic Molecular Beam Deposition and Related Techniques. Chem. Rev. 97 1793 1896 (1997). 32. Karl, N. Organic Semiconductors: Purification and Crystal Growth. Mol. Cryst. Liq. Cryst. 171 157 (1989). 33. Shtein, M., Peumans, P., Benziger, J.B. & Forrest, S.R. Micropatterning of small molecular weight organic semiconductor thin fil ms using organic vapor phase deposition. J. Appl. Phys. 93 4005 (2003). 34. Zheng, Y. et al. Computational and experimental studies of phase separation in pentacene:C[sub 60] mixtures. J. Vac. Sci. Technol. B 27 169 (2009). 35. Krebs, F.C. All solution r oll to roll processed polymer solar cells free from indium tin oxide and vacuum coating steps. Org. Elec. 10 761 768 (2009). 36. Wu, J. et al. Organic solar cells with solution processed graphene transparent electrodes. Appl. Phys. Lett. 92 263302 (2008) 37. Lee, J. Y., Connor, S.T., Cui, Y. & Peumans, P. Solution Processed Metal Nanowire Mesh Transparent Electrodes. Nano Lett. 8 689 692 (2008). 38. Derby, B. Inkjet Printing of Functional and Structural Materials: Fluid Property Requirements, Feature St ability, and Resolution. Annu. Rev. Mater. Res. 40 395 414 (2010). 39. de Gans, B. J, Duineveld, P.C. & Schubert, U.S. Inkjet Printing of Polymers: State of the Art and Future Developments. Adv. Mater. 16 203 213 (2004).

PAGE 194

194 40. Green, R. et al. Performance of bulk heterojunction photovoltaic devices prepared by airbrush spray deposition. Appl. Phys. Lett. 92 033301 033301 3 (2008). 41. Vak, D. et al. Fabrication of organic bulk heterojunction solar cells by a spray deposition me thod for low cost power generation. Appl. Phys. Lett. 91 081102 (2007). 42. Ishikawa, T., Nakamura, M., Fujita, K. & Tsutsui, T. Preparation of organic bulk heterojunction photovoltaic cells by evaporative spray deposition from ultradilute solution. Appl. Phys. Lett. 84 2424 (2004). 43. Baldo, M. et al. Organic Vapor Phase Deposition. Adv. Mater. 10 1505 1514 (1998). 44. Shtein, M., Peumans, P., Benziger, J.B. & Forrest, S.R. Direct, Mask and Solvent Free Printing of Molecular Organic Semiconductors. Adv. Mater. 16 1615 1620 (2004). 45. Shtein, M., Peumans, P., Benziger, J.B. & Forrest, S.R. Direct mask free patterning of molecular organic semiconductors using organic vapor jet printing. J. Appl. Phys. 96 4500 (2004). 46. Sun, Y., Shtein, M. & Forre st, S.R. Direct patterning of organic light emitting devices by organic vapor jet printing. Appl. Phys. Lett. 86 113504 (2005). 47. Tang, C.W. Two layer organic photovoltaic cell. Appl. Phys. Lett. 48 183 185 (1986). 48. Green, M.A., Emery, K., Hishikawa Y. & Warta, W. Solar cell efficiency tables (version 37). Prog. Photovolt: Res. Appl. 19 84 92 (2011). 49. Dresner, J. Double injection electroluminescence in anthracene. RCA Rev. 30 322 334 (1969). 50. Tang, C.W. & VanSlyke, S.A. Organic electrolumine scent diodes. Appl. Phys. Lett. 51 913 (1987). 51. Green, M.A. Third Generation Photovoltaics (Springer Verlag: Berlin, 2003). 52. Kallmann, H. & Pope, M. Photovoltaic Effect in Organic Crystals. J. Chem. Phys. 30 585 (1959). 53. Chamberlain, G.A., Coon ey, P.J. & Dennison, S. Photovoltaic properties of merocyanine solid state photocells. Nature 289 45 47 (1981).

PAGE 195

195 54. Ghosh, A.K. & Feng, T. Merocynanine organic solar cells. J. Appl. Phys. 49 5982 (1978). 55. Harima, Y., Yamashita, K. & Suzuki, H. Spectra l sensitization in an organic p n junction photovoltaic cell. Appl. Phys. Lett. 45 1144 1145 (1984). 56. Blom, P.W.M., Mihailetchi, V.D., Koster, L.J.A. & Markov, D.E. Device Physics of Polymer:Fullerene Bulk Heterojunction Solar Cells. Adv. Mater. 19 15 51 1566 (2007). 57. Popovic, Z.D., Khan, M.I., Atherton, S.J., Hor, A. M. & Goodman, J.L. Study of Carrier Generation in Titanyl Phthalocyanine (TiOPc) by Electric Field Induced Quenching of Integrated and Time Resolved Fluorescence. J. Phys. Chem. B 102 657 663 (1998). 58. Kraabel, B., McBranch, D., Sariciftci, N.S., Moses, D. & Heeger, A.J. Ultrafast spectroscopic studies of photoinduced electron transfer from semiconducting polymers to C_{60}. Phys. Rev. B 50 18543 (1994). 59. Hiramoto, M., Fujiwara, H & Yokoyama, M. p i n like behavior in three layered organic solar cells having a co deposited interlayer of pigments. J. Appl. Phys. 72 3781 (1992). 60. Rostalski, J. & Meissner, D. Monochromatic versus solar efficiencies of organic solar cells. Sol. Energy Mater. Sol. Cells 61 87 95 (2000). 61. Geens, W., Aernouts, T., Poortmans, J. & Hadziioannou, G. Organic co evaporated films of a PPV pentamer and C60: model systems for donor/acceptor polymer blends. Thin Solid Films 403 404 438 443 (2002). 62. W ang, X.H., Grell, M., Lane, P.A. & Bradley, D.D.C. Determination of the linear optical constants of poly(9,9 dioctylfluorene). Synth. Met. 119 535 536 (2001). 63. Peumans, P., Uchida, S. & Forrest, S.R. Efficient bulk heterojunction photovoltaic cells usi ng small molecular weight organic thin films. Nature 425 158 162 (2003). 64. Heremans, P., Cheyns, D. & Rand, B.P. Strategies for Increasing the Efficiency of Heterojunction Organic Solar Cells: Material Selection and Device Architecture. Acc. Chem. Res. 42 1740 1747 (2009). 65. Xue, J., Rand, B.P., Uchida, S. & Forrest, S.R. A Hybrid Planar Mixed Molecular Heterojunction Photovoltaic Cell. Adv. Mater. 17 66 71 (2005).

PAGE 196

196 66. Xue, J., Uchida, S., Rand, B.P. & Forrest, S.R. Asymmetric tandem organic photovol taic cells with hybrid planar mixed molecular heterojunctions. Appl. Phys. Lett. 85 5757 5759 (2004). 67. Yang, F. & Forrest, S.R. Photocurrent Generation in Nanostructured Organic Solar Cells. ACS Nano 2 1022 1032 (2008). 68. Zheng, Y., Bekele, R., Ouya ng, J. & Xue, J. Interdigitated Bulk Heterojunction Organic Photovoltaic Cells With Aligned Copper Phthalocyanine Nanorods. IEEE J. Sel. Top. Quantum Electron. 16 1544 1551 (2010). 69. Zheng, Y., Bekele, R., Ouyang, J. & Xue, J. Organic photovoltaic cells with vertically aligned crystalline molecular nanorods. Org. Elec. 10 1621 1625 (2009). 70. Zheng, Y. & Xue, J. Organic Photovoltaic Cells Based on Molecular Donor Acceptor Heterojunctions. Polym. Rev. 50 420 (2010). 71. Robbie, K. Advanced techniques f or glancing angle deposition. J. Vac. Sci. Technol. B 16 1115 (1998). 72. Nanorod/P3HT Solar Cells. J. Phys. Chem. C 111 18451 18456 (2007). 73. Pfuetzner, S., Meiss, J., Petrich, A., Riede, M. & Leo, K. Improved bulk heterojunction organic solar cells employing C[sub 70] fullerenes. Appl. Phys. Lett. 94 223307 (2009). 74. Bailey Salzman, R.F., Rand, B.P. & Forrest, S.R. Near infrared sensitive sm all molecule organic photovoltaic cells based on chloroaluminum phthalocyanine. Appl. Phys. Lett. 91 013508 (2007). 75. Dai, J., Jiang, X., Wang, H. & Yan, D. Organic photovoltaic cells with near infrared absorption spectrum. Appl. Phys. Lett. 91 253503 (2007). 76. Kim, D.Y., Sarasqueta, G. & So, F. SnPc:C60 bulk heterojunction organic photovoltaic cells with MoO3 interlayer. Sol. Energy Mater. Sol. Cells 93 1452 1456 (2009). 77. Lunt, R.R. & Bulovic, V. Transparent, near infrared organic photovoltaic so lar cells for window and energy scavenging applications. Appl. Phys. Lett. 98 113305 (2011). 78. Rand, B.P., Xue, J., Yang, F. & Forrest, S.R. Organic solar cells with sensitivity extending into the near infrared. Appl. Phys. Lett. 87 233508 233508 3 (20 05).

PAGE 197

197 79. Vasseur, K., Rand, B.P., Cheyns, D., Froyen, L. & Heremans, P. Structural Evolution of Evaporated Lead Phthalocyanine Thin Films for Near Chem. Mater. 23 886 895 (2011). 80. Xia, A.D. et al. Characteristics of het erojunction diode of C60/chloroindium phthalocyanine. Solid State Commun. 95 713 716 (1995). 81. Sullivan, P., Jones, T.S., Ferguson, A.J. & Heutz, S. Structural templating as a route to improved photovoltaic performance in copper phthalocyanine/fullerene (C60) heterojunctions. Appl. Phys. Lett. 91 233114 233114 3 (2007). 82. Iyechika, Y., Yakushi, K., Ikemoto, I. & Kuroda, H. Structure of lead phthalocyanine (triclinic form). Acta Crystallogr., Sect. B: Struct. Sci 38 766 770 (1982). 83. Wang, W., Place ncia, D. & Armstrong, N.R. Planar and textured heterojunction organic photovoltaics based on chloroindium phthalocyanine (ClInPc) versus titanyl phthalocyanine (TiOPc) donor layers. Org. Elec. 12 383 393 (2011). 84. Perez, M.D., Borek, C., Forrest, S.R. & Thompson, M.E. Molecular and Morphological Influences on the Open Circuit Voltages of Organic Photovoltaic Devices. J. Am. Chem. Soc. 131 9281 9286 (2009). 85. Li, N., Lassiter, B.E., Lunt, R.R., Wei, G. & Forrest, S.R. Open circuit voltage enhancement d ue to reduced dark current in small molecule photovoltaic cells. Appl. Phys. Lett. 94 023307 (2009). 86. Mutolo, K.L., Mayo, E.I., Rand, B.P., Forrest, S.R. & Thompson, M.E. Enhanced Open Circuit Voltage in Subphthalocyanine/C60 Organic Photovoltaic Cells J. Am. Chem. Soc. 128 8108 8109 (2006). 87. Tong, X., Lassiter, B.E. & Forrest, S.R. Inverted organic photovoltaic cells with high open circuit voltage. Org. Elec. 11 705 709 (2010). 88. Pandey, R. & Holmes, R.J. Graded Donor Acceptor Heterojunctions f or Efficient Organic Photovoltaic Cells. Adv. Mater. 22 5301 5305 (2010). 89. Fleetham, T.B. et al. Material dependence of the open circuit voltage in zinc phthalocyanine C60 organic photovoltaic cells. (in submission) 90. Glenis, S., Horowitz, G., Touri llon, G. & Garnier, F. Electrochemically grown polythiophene and poly(3 methylthiophene) organic photovoltaic cells. Thin Solid Films 111 93 103 (1984). 91. Kanicki, J. & Fedorko, P. Electrical and photovoltaic properties of trans polyacetylene. J. Phys. D: Appl. Phys. 17 805 817 (1984).

PAGE 198

198 92. Yu, G., Pakbaz, K. & Heeger, A.J. Semiconducting polymer diodes: Large size, low cost photodetectors with excellent visible ultraviolet sensitivity. Appl. Phys. Lett. 64 3422 3424 (1994). 93. Marks, R.N., Halls, J.J. M., Bradley, D.D.C., Friend, R.H. & Holmes, A.B. The photovoltaic response in poly(p phenylene vinylene) thin film devices. J. Phys.: Condens. Matter 6 1379 1394 (1994). 94. Sariciftci, N.S., Smilowitz, L., Heeger, A.J. & Wudl, F. Photoinduced Electron Tr ansfer from a Conducting Polymer to Buckminsterfullerene. Science 258 1474 1476 (1992). 95. Sariciftci, N.S. et al. Semiconducting polymer buckminsterfullerene heterojunctions: Diodes, photodiodes, and photovoltaic cells. Appl. Phys. Lett. 62 585 587 (1 993). 96. Hummelen, J.C. et al. Preparation and Characterization of Fulleroid and Methanofullerene Derivatives. J. Org. Chem. 60 532 538 (1995). 97. Yu, G., Gao, J., Hummelen, J.C., Wudl, F. & Heeger, A.J. Polymer Photovoltaic Cells: Enhanced Efficiencies via a Network of Internal Donor Acceptor Heterojunctions. Science 270 1789 1791 (1995). 98. Shaheen, S.E. et al. 2.5% efficient organic plastic solar cells. Appl. Phys. Lett. 78 841 (2001). 99. Wienk, M.M. et al. Efficient Methano[70]fullerene/MDMO PPV Bulk Heterojunction Photovoltaic Cells. Angew. Chem. Int. Ed. 42 3371 3375 (2003). 100. Brabec, C.J., Shaheen, S.E., Winder, C., Sariciftci, N.S. & Denk, P. Effect of LiF/metal electrodes on the performance of plastic solar cells. Appl. Phys. Lett. 80 12 88 (2002). 101. Ma, W., Yang, C., Gong, X., Lee, K. & Heeger, A.J. Thermally Stable, Efficient Polymer Solar Cells with Nanoscale Control of the Interpenetrating Network Morphology. Adv. Funct. Mater. 15 1617 1622 (2005). 102. Ma, W., Gopinathan, A. & Heeger, A.J. Nanostructure of the Interpenetrating Networks in Poly(3 hexylthiophene)/fullerene Bulk Heterojunction Materials: Implications for Charge Transport. Adv. Mater. 19 3656 3659 (2007). 103. Kim, Y. et al. A strong regioregularity effect in self organizing conjugated polymer films and high efficiency polythiophene:fullerene solar cells. Nat Mater 5 197 203 (2006).

PAGE 199

199 104. Li, G. et al. High efficiency solution processable polymer photovoltaic cells by self organization of polymer blends. Nat. Mater. 4 864 868 (2005). 105. Leclerc, M., Diaz, F.M. & Wegner, G. Structural analysis of poly(3 alkylthiophene)s. Die Makromolekulare Chemie 190 3105 3116 (1989). 106. McCullough, R.D., Tristram Nagle, S., Williams, S.P., Lowe, R.D. & Jayaraman, M. Self orien ting head to tail poly(3 alkylthiophenes): new insights on structure property relationships in conducting polymers. J. Am. Chem. Soc. 115 4910 4911 (1993). 107. Koppe, M. et al. Influence of Molecular Weight Distribution on the Gelation of P3HT and Its Im pact on the Photovoltaic Performance. Macromolecules 42 4661 4666 (2009). 108. Osaka, I. & McCullough, R.D. Chapter 3. Advanced Functional Regioregular Polythiophenes. Design and Synthesis of Conjugated Polymers (eds. M. Leclerc and J. F. Morin) (2010).at 109. Ballantyne, A.M. et al. Understanding the Influence of Morphology on Poly(3 hexylselenothiophene):PCBM Solar Cells. Macromolecules 43 1169 1174 (2010). 110. Ballantyne, A.M. et a l. Studies of Highly Regioregular Poly(3 hexylselenophene) for Photovoltaic Applications. Adv. Mater. 19 4544 4547 (2007). 111. Chen, J. & Cao, Y. Development of Novel Conjugated Donor Polymers for High Efficiency Bulk Heterojunction Photovoltaic Devices. Acc. Chem. Res. 42 1709 1718 (2009). 112. Boudreault, P. L.T., Najari, A. & Leclerc, M. Processable Low Bandgap Chem. Mater. 23 456 469 (2011). 113. Kroon, R., Lenes, M., Hummelen, J.C., Blom, P.W.M. & de Boer, B Small Bandgap Polymers for Organic Solar Cells (Polymer Material Development in the Last 5 Years). Polym. Rev. 48 531 (2008). 114. Peet, J. et al. Efficiency enhancement in low bandgap polymer solar cells by processing with alkane dithiols. Nat. Mater. 6 497 500 (2007). 115. Peet, J., Senatore, M.L., Heeger, A.J. & Bazan, G.C. The Role of Processing in the Fabrication and Optimization of Plastic Solar Cells. Adv. Mater. 21 1521 1527 (2009).

PAGE 200

200 116. Blouin, N., Michaud, A. & Leclerc, M. A Low Bandgap Poly( 2,7 Carbazole) Derivative for Use in High Performance Solar Cells. Adv. Mater. 19 2295 2300 (2007). 117. Park, S.H. et al. Bulk heterojunction solar cells with internal quantum efficiency approaching 100%. Nat Photon 3 297 302 (2009). 118. Zhou, H. et al Development of Fluorinated Benzothiadiazole as a Angew. Chem. Int. Ed. 50 2995 2998 (2011). 119. Chen, H. Y. et al. Polymer solar cells with enhanced open circuit voltage and efficiency. Nat Ph oton 3 649 653 (2009). 120. Liang, Y. et al. For the Bright Future Bulk Heterojunction Polymer Solar Cells with Power Conversion Efficiency of 7.4%. Adv. Mater. 22 E135 E138 (2010). 121. Zhao, J., Wang, A. & Green, M.A. 245% Efficiency silicon PERT cell s on MCZ substrates and 247% efficiency PERL cells on FZ substrates. Prog. Photovolt: Res. Appl. 7 471 474 (1999). 122. Hiramoto, M., Suezaki, M. & Yokoyama, M. Effect of Thin Gold Interstitial layer on the Photovoltaic Properties of Tandem Organic Solar Cell. Chem. Lett. 19 327 330 (1990). 123. Yakimov, A. & Forrest, S.R. High photovoltage multiple heterojunction organic solar cells incorporating interfacial metallic nanoclusters. Appl. Phys. Lett. 80 1667 (2002). 124. Rand, B.P., Peumans, P. & Forrest S.R. Long range absorption enhancement in organic tandem thin film solar cells containing silver nanoclusters. J. Appl. Phys. 96 7519 (2004). 125. Janssen, A.G.F., Riedl, T., Hamwi, S., Johannes, H. H. & Kowalsky, W. Highly efficient organic tandem solar cells using an improved connecting architecture. Appl. Phys. Lett. 91 073519 (2007). 126. Colsmann, A., Junge, J., Kayser, C. & Lemmer, U. Organic tandem solar cells comprising polymer and small molecule subcells. Appl. Phys. Lett. 89 203506 (2006). 127. Kim, J.Y. et al. Efficient Tandem Polymer Solar Cells Fabricated by All Solution Processing. Science 317 222 225 (2007). 128. Zhao, D.W. et al. Efficient tandem organic solar cells with an Al/MoO[sub 3 ] intermediate layer. Appl. Phys. Lett. 93 083305 (2008).

PAGE 201

201 129. Shrotriya, V., Wu, E.H. E., Li, G., Yao, Y. & Yang, Y. Efficient light harvesting in multiple device stacked structure for polymer solar cells. Appl. Phys. Lett. 88 064104 (2006). 130. Sista, S. et al. Highly Efficient Tandem Polymer Photovoltaic Cells. Adv. Mater. 22 380 383 (2010). 131. Currie, M.J., Mapel, J.K., Heidel, T.D., Goffri, S. & Baldo, M.A. High Efficiency Organic Solar Concentrators for Photovoltaics. Science 321 226 228 (2008). 132. Kim, S. S., Na, S. I., Jo, J., Kim, D. Y. & Nah, Y. C. Plasmon enhanced performance of organic solar cells using electrodeposited Ag nanoparticles. Appl. Phys. Lett. 93 073307 (2008). 133. Kulkarni, A.P., Noone, K.M., Munechika, K., Guyer, S. R. & Ginger, D.S. Plasmon Enhanced Charge Carrier Generation in Organic Photovoltaic Films Using Silver Nanoprisms. Nano Lett. 10 1501 1505 (2010). 134. O., Rahachou, A. & Zozoulenko, I.V. Surface plasmon increase absorption in polymer photovoltaic cells. Appl. Phys. Lett. 91 113514 (2007). 135. Rim, S. B., Zhao, S., Scully, S.R., McGehee, M.D. & Peumans, P. An effective light trapping configuration for thin film solar cells. Appl. Phys. Lett. 91 243501 (2007). 1 36. O. Folded reflective tandem polymer solar cell doubles efficiency. Appl. Phys. Lett. 91 123514 (2007). 137. O. Multifolded polymer solar cel ls on flexible substrates. Appl. Phys. Lett. 93 033302 (2008). 138. Niggemann, M. et al. Functional microprism substrate for organic solar cells. Thin Solid Films 511 512 628 633 (2006). 139. Tvingstedt, K., Dal Zilio, S., Ingans, O. & Tormen, M. Trappi ng light with micro lenses in thin film organic photovoltaic cells. Opt. Express 16 21608 21615 (2008). 140. Zilio, S.D., Tvingstedt, K., Ingans, O. & Tormen, M. Fabrication of a light trapping system for organic solar cells. Microelectron. Eng. 86 1150 1154 (2009).

PAGE 202

202 141. Cocoyer, C. et al. Implementation of submicrometric periodic surface structures toward improvement of organic solar cell performances. Appl. Phys. Lett. 88 133108 (2006). 142. Nalwa, K.S., Park, J. M., Ho, K. M. & Chaudhary, S. On Reali zing Higher Efficiency Polymer Solar Cells Using a Textured Substrate Platform. Adv. Mater. 23 112 116 (2011). 143. Shrotriya, V. et al. Accurate Measurement and Characterization of Organic Solar Cells. Adv. Funct. Mater. 16 2016 2023 (2006). 144. ASTM Standard G173 03, 2008, Standard Tables for Reference Solar Spectral Irradiances: Direct Normal and Hemispherical on 37 Tilted Surface. 145. Gueymard, C.A. Parameterized transmittance model for direct beam and circumsolar spectral irradiance. Sol. Energy 71 325 346 (2001). 146. applications and solar radiation models. Sol. Energy 76 423 453 (2004). 147. Gueymard, C.A. SMARTS2, A Simple Model of the Atmospheric Radiative Transfer of Sunshine: Algorithms and performance assessment FSEC PF 270 95 (Florida Solar Energy Center/University of Central Florida: Cocoa, FL, 1995). 148. ASTM Standard E1021, Standard Test Method for Measuring Spectral Response of Photovoltaic Cells. 149. Kim, J. S., Ho, P.K.H., Greenham, N.C. & Friend, R.H. Electroluminescence emission pattern of organic light emitting diodes: Implications for device efficiency calculations. J. Appl. Phys. 88 1073 (2000). 150. Sun, Y. & Forrest, S.R. Organic light emitting devices with enhanced outcoupling via microlenses fabricated by imprint lithography. J. Appl. Phys. 100 073106 (2006). 151. Lee, J., Chopra, N. & So, F. Cavity effects on light extraction in organic light emitting dev ices. Appl. Phys. Lett. 92 033303 (2008). 152. Forrest, S.R., Bradley, D.D.C. & Thompson, M.E. Measuring the Efficiency of Organic Light Emitting Devices. Adv. Mater. 15 1043 1048 (2003). 153. Chen, K. Y., Hsiao, Y. T., Lin, H.Y., Wei, M. K. & Lee, J. H. Partitioning pixel of organic light emitting devices with center hollowed microlens array films for efficiency enhancement. Opt. Express 18 18685 18690 (2010).

PAGE 203

203 154. Li, Y. et al. Origins of performance in fiber based organic photovoltaics. Appl. Phys. Le tt. 95 203503 (2009). 155. V. & Forrest, S.R. Efficient photon harvesting at high optical intensities in ultrathin organic double heterostructure photovoltaic diodes. Appl. Phys. Lett. 76 2650 (2000). 156. Humphrey, W., Dalke, A. & Schulten, K. VMD Visual Molecular Dynamics. J. Mol. Graphics 14 33 38 (1996). 157. O. Modeling photocurrent action spectra of photovoltaic devices based on organic thin films. J. Appl. Phys. 86 487 (1999). 1 58. Sievers, D.W., Shrotriya, V. & Yang, Y. Modeling optical effects and thickness dependent current in polymer bulk heterojunction solar cells. J. Appl. Phys. 100 114509 114509 7 (2006). 159. W. Exciton diffusion and optical i nterference in organic donor acceptor photovoltaic cells. J. Appl. Phys. 90 3632 (2001). 160. Hoppe, H., Arnold, N., Sariciftci, N.S. & Meissner, D. Modeling the optical absorption within conjugated polymer/fullerene based bulk heterojunction organic solar cells. Sol. Energy Mater. Sol. Cells 80 105 113 (2003). 161. 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 034907 (2006). 162. Hoppe, H., Arnold, N., Meissner, D. & Sariciftci, N.S. Modeling of optical absorption in conjugated polymer/fullerene bulk heterojunction p lastic solar cells. Thin Solid Films 451 452 589 592 (2004). 163. Gilot, J., Barbu, I., Wienk, M.M. & Janssen, R.A.J. The use of ZnO as optical spacer in polymer solar cells: Theoretical and experimental study. Appl. Phys. Lett. 91 113520 (2007). 164. Pe film J. Appl. Phys. 95 2938 (2004). 165. Rand, B.P., Burk, D.P. & Forrest, S.R. Offset energies at organic semiconductor heterojuncti ons and their influence on the open circuit voltage of thin film solar cells. Phys. Rev. B 75 115327 11 (2007). 166. Kim, S.Y. Effect of ultraviolet ozone treatment of indium tin oxide on electrical properties of organic light emitting diodes. J. Appl. Ph ys. 95 2560 (2004).

PAGE 204

204 167. Sugiyama, K., Ishii, H., Ouchi, Y. & Seki, K. Dependence of indium tin oxide work function on surface cleaning method as studied by ultraviolet and x ray photoemission spectroscopies. J. Appl. Phys. 87 295 (2000). 168. Malaquin, L., Kraus, T., Schmid, H., Delamarche, E. & Wolf, H. Controlled Particle Placement through Convective and Capillary Assembly. Langmuir 23 11513 11521 (2007). 169. Danzebrink, R. & Aegerter, M.A. Deposition of optical microlens arrays by ink jet processes. Thin Solid Films 392 223 225 (2001). 170. Cheng, C. C., Chang, C.A. & Yeh, J.A. Variable focus dielectric liquid droplet lens. Opt. Express 14 4101 4106 (2006). 171. Lu, Y., Yin, Y. & Xia, Y. A Self Assembly Approach to the Fabrication of Patterned, Two Dimensional Arrays of Microlenses of Organic Polymers. Adv. Mater. 13 34 37 (2001). 172. S. & Forrest, S.R. Improved light out coupling in organic light emitting diodes employing ordered microlens arrays. J. Appl. Phys. 91 3324 (2002). 173. Chang, S. I. & Yoon, J. B. Shape controlled, high fill factor microlens arrays fabricated by a 3D diffuser lithography and plastic replication method. Opt. Express 12 6366 6371 (2004). 174. Eom, S. H., Wrzesniewski, E. & Xue, J. Close packed hemispherical microlens arrays for light extraction enhancement in organic light emitting devices. Org. Elec. 12 472 476 (2011). 175. Eom, S. H., Wrzesniewski, E. & Xue, J. Enhancing light extraction in organic light emitting devices via hemispherical microlens arrays fabricated by soft lithography. J. Photon. Energy 1 011002 (2011). 176. Yang, J., Tang, A., Zhou, R. & Xue, J. Effects of nanocrystal size and device aging on performance of hybrid poly(3 hexylthiophene):CdSe nanocrystal solar cells. Sol. Energy Mater. S ol. Cells 95 476 482 (2011). 177. Qian, L. et al. Hybrid polymer CdSe solar cells with a ZnO nanoparticle buffer layer for improved efficiency and lifetime. J. Mater. Chem. 21 3814 (2011). 178. Ji, W., Zhao, J., Sun, Z. & Xie, W. High color rendering fle xible top emitting warm white organic light emitting diode with a transparent multilayer cathode. Org. Elec. 12 1137 1141 (2011). 179. Minami, T. Transparent conducting oxide semiconductors for transparent electrodes. Semicond. Sci. Technol. 20 S35 S44 ( 2005).

PAGE 205

205 180. Bera, D., Qian, L., Sabui, S., Santra, S. & Holloway, P.H. Photoluminescence of ZnO quantum dots produced by a sol gel process. Opt. Mater. 30 1233 1239 (2008). 181. Zhou, H., Yang, L., Price, S.C., Knight, K.J. & You, W. Enhanced Photovoltaic Performance of Low Bandgap Polymers with Deep LUMO Levels. Angew. Chem. Int. Ed. 49 7992 7995 (2010). 182. Qian, L. et al. Electroluminescence from light emitting polymer/ZnO nanoparticle heterojunctions at sub bandgap voltages. Nano Today 5 384 389 (20 10). 183. Chen, L.L., Li, W.L., Li, M.T. & Chu, B. Bifunctional electroluminescent and photovoltaic devices using bathocuproine as electron transporting material and an electron acceptor. J. Lumin. 122 123 667 670 (January). 184. Okada, H. et al. Organic Bifunctional Devices with Emission and Sensing Abilities. Jpn. J. Appl. Phys. 46 1328 1332 (2007). 185. Pandey, A.K. & Nunzi, J. M. Rubrene/Fullerene Heterostructures with a Half Gap Electroluminescence Threshold and Large Photovoltage. Adv. Mater. 19 36 13 3617 (2007). 186. Wei, H. et al. White organic electroluminescent device with photovoltaic performances. Appl. Surf. Sci. 252 2204 2208 (2006). 187. Hong, Z.R., Lee, C.S., Lee, S.T., Li, W.L. & Shirota, Y. Bifunctional photovoltaic and electroluminesce nt devices using a starburst amine as an electron donor and hole transporting material. Appl. Phys. Lett. 81 2878 (2002). 188. Narayan, K.S., Manoj, A.G., Nanda, J. & Sarma, D.D. Dual function hybrid polymer nanoparticle devices. Appl. Phys. Lett. 74 871 (1999). 189. Sahu, S. & Pal, A.J. Multifunctionality of Organic Devices: Light Emission, Photovoltage Generation, and Photodetection. J. Phys. Chem. C 112 8446 8451 (2008). 190. Bernards, D.A., Flores Torres, S., Abrua, H.D. & Malliaras, G.G. Observation of Electroluminescence and Photovoltaic Response in Ionic Junctions. Science 313 1416 1419 (2006). 191. Ng, A.M.C., Djurisic, A.B., Chan, W. K. & Nunzi, J. M. Near infrared emission in rubrene:fullerene heterojunction devices Chem. Phys. Lett. 474 141 145 (2009).

PAGE 206

206 192. Pandey, A.K. & Nunzi, J. M. Upconversion injection in rubrene/perylene diimide heterostructure electroluminescent diodes. Appl. Phys. Lett. 90 263508 (2007). 193. Baldo, M.A., Adachi, C. & Forrest, S.R. Transi ent analysis of organic electrophosphorescence. II. Transient analysis of triplet triplet annihilation. Phys. Rev. B 62 10967 (2000). 194. McNeill, C.R. et al. Dual electron donor/electron acceptor character of a conjugated polymer in efficient photovolta ic diodes. Appl. Phys. Lett. 90 193506 (2007). 195. Gummel, H.K. A self consistent iterative scheme for one dimensional steady state transistor calculations. IEEE Trans. Electron Devices 11 455 465 (1964). 196. Mandoc, M.M., Koster, L.J.A. & Blom, P.W.M. Optimum charge carrier mobility in organic solar cells. Appl. Phys. Lett. 90 133504 (2007). 197. Yee, K. Numerical solution of initial boundary value problems involving IEEE Trans. Antennas Propag. 14 302 307 (1966). 198. Ouyang, J., Chu, C. W., Szmanda, C.R., Ma, L. & Yang, Y. Programmable polymer thin film and non volatile memory device. Nat. Mater. 3 918 922 (2004).

PAGE 207

207 BIOGRAPHICAL SKETCH Jason Myers was born in 1985 in Kingsport, TN. During midd le school, he moved to Miami, FL and finished his preparatory education there graduating from Miami Palmetto Senior High School in 2003, after which he was awarded a National Merit s cholarship and enrolled at the University of Florida in Gainesville, FL He completed his Bachelor of Science degree in Materials Science and Engineering Summa cum L aude in 2007 and remained for his doctoral studies in the same field, obtaining his M.S. in 2008. ergraduate research assistant investigating gaseous diffusion through branched carbon nanotube structures using molecular dynamics simulations He was named the American Vacuum Society (AVS) Undergraduate Scholar of the Year in 2005 and presented his rese arch at numerous conferences. At the completion of his undergraduate degree, Jason shifted his efforts to Alumni Fellow. His research focused on organic based photovoltaic devices, with an emphasis on photocarrier motion and optical enhancements. His research has led to several refereed publications, numerous presentations at international conferences, and multiple United States patents. In 2010, he traveled to Kuala Lumpu r, Malaysia as the United States representative in the Institute of Materials, Mining, and Minerals World Lecture Competition, winning the grand prize. After receiving his doctorate in August 2011 Jason accepted a position with the Naval Research Labora tory in Washington, DC H e married his wife, Anne in 2009