This item is only available as the following downloads:
1 ORGANIC ELECTRONIC DEVICES USING GRAPHENE AND HIGHLY PURIFIED THIN FILMS OF CARBON NANOTUBE S AS TRANSPARENT CONDUCTIVE ELECTRODES By EVAN PETER DONOGHUE A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FL ORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2012
2 2012 Evan Peter Donoghue
3 To my many teachers
4 ACKNOWLEDGMENTS Writing a dissertation can be a humbling experience and as I wrote the pages that you are about to read, I was constantly reminded of how much of what I have accomplished is owed to the many friends, colleagues and teachers who have contributed both to my research efforts as well as to my own personal development. In light of the wide ranging contributions from many different people, it seems unfair that my name stands alone on the cover of this document. I hope that they all know that I am truly appreciative of what they have given me. First and foremost, I would like to thank my family who have helped to make me who I am and who have supported me in times of success and times of difficulty. They have always given me the love, care encouragement and support that I have needed while allowing me to cho o se my own p ath in life. On days that I was struggling in my research or classwork, there was always a degree of comfort that came from knowing I also need to thank my advisor Andre graduate career F rom the first day s, back in June 2006 I always felt that he listened to and valued my opinions and offered me his honest opinions in return. He has always been there by my side to teach me the fundamental laboratory skills that I have needed and make sure that, whatever I was doing, I was always thinking about the best way to do it from something as simple as washing a pair of tweezers to as important as working around hazardous materials It is hard to imagine getting a better training in hands on experimental science. Professor Rinzler and his research were a major factor in my decision to come to the University of Florida and one of the best things I can say
5 about that decision is tha t after six years of research, my curiosity and excitement about the future of this work has only grown with time. M y labmates have been essential to my work in this dissertation as well as my growth as a scientist. I am extremely fortunate to be surro unded by so many smart, driven and fun people who have helped me in my research and kept me sane in the process and for that I thank Bo Liu, Mitchell McCarthy, Pooja Wadhwa, Rajib Das, Maureen Petterson, Max Lemaitre, Svetlana Vasilyeva, Xiao Chen, Yu Shen Ramesh Jayaraman, Matt Gilbert, Tom Hayman, Stephen French, Kyle Dorsey, and Zhuangchun Wu. I c annot count how many problems I ha ve overcome simply by talking to my colleagues or how often they have been willing to contribute their time and effort to ad vancing my work. I want to particularly thank Bo Liu for helping me when I needed him most, offering me ideas for projects to undertake and keeping me calm with his even effo rt (in conjuction with Bill Malphurs of the Machine Shop) in designing the glovebox and evaporators that have enabled so much of the work in our lab. Ramesh Jayaraman (in addition to Maureen Petterson and others) has played an often underappreciated role in carbon nanotube synthesis to provide the material that drives all of the research in our lab and I owe him my thanks Similarly, Max Lemaitre and I w orked s i d e b y side o n the ef f o r t s in G VFE T s f o r Chapter 4 and t he w o r k t h e re was s h a r e d though h e also g rew al l th e g raphene h i m self and brought t h e idea of t h e i m prove d trans f e r tec h n i q u e I also need to thank all of our collaborators, starting with Prof John Reynolds and his group in the Department of Chemistry (now at Georgia Tech). Professor Reynolds
6 was always a calming presence who kept an eye on the big pict ure and on many occasions was there with an encouraging word and a suggestion for future work. He always seemed to be looking out for me and I am truly grateful for his role in my research I was also fortunate to meet his student, Ken Graham, early in m y career so that I could develop and learn alongside him on projects such as the organic light and I hope that can continue into the future. There are many others who I worked alongside in MCCL and the Chemistry Department that have helped me along this journey, namely Nathan Heston, Aubrey Dyer, Ryan Walczak, Egle Puodziukynaite Richard Farley, Caroline Grand, Patrick Wieruszewski, Justin Oberst and Danielle Salazar a nd I would like to thank all of them our laboratory and we have relied on each other for both equipment and insight. I want to particularly thank Patrick Mickel, Sefaattin Tongay and Kara B erke. Additionally, Dr. Franky So and his group ha ve assisted ours in many ways and I am appreciative of their contributions. During my first visit to the University of Florida, as Professor Hebard showed me around the building he made a point of raving about the excel l e nce of the support staff in the Physics Department. At the time, this statement did not mean much to me but after six years, I cannot fathom what the department would do without them. So much of the equipment in our lab h as been built by Ed Storch, Bill Malphurs and Marc Link that to remove their contributions would be to eliminate 75% of the results found in this document Similarly, I could always count on Jay Horton to help in any way he could
7 whether that was maintain ing pumps or installing fume hoods. Greg Labbe and John Graham in Cryogenics, everyone in the Electronics Shop the custodial staff and Tim Noland have always been there to keep our experiments and our lab running smoothly. Along the same lines, Darlene Lattimer is the unsung hero of our laboratory and has kept things going behind the scenes and never once complained (to me at least) about having to save me from my own mistakes in purchasing. I also need to thank Nathan Williams and Pam Marlin for keeping me on track and never getting upset over my often late paperwork. It was important to me that I assemble an advisory committee of people that I respected, admired and enjoyed being around. When I was deciding between graduate schools, one of the factors that drew me to Florida was the passion for science that I could immediately sense in meetings with Prof Tanner and Prof Hebard and it was this same passion for science that led to me asking them to serve on my committee. It is clear that they love their work and love to learn and I hope that I can always maintain the same excitement in my work. as his student in Optical Effects in Solids as well as a collaborator in multiple research project s in insight and brief chats in the hall. I first got to know Professor Monkhorst in taking Soli d State Physics where his excitement for the material and physics in general was always conveyed in discussions. When I see him, h down and I have always appreciated his encouragem have him on my committee Prof Kirk Ziegler has a shared interest in carbon nanotubes
8 and has supported my research by loaning our laboratory both equipment and materials so that we could test new ideas and I am thankful for his contributions. I am truly app reciative of all the time and energy that my committee has put into helping me become a better scientist and in helping me through this process I am indebted to my many friends throughout my life who have pushed me, supported me and driven me to be better I have had many friends to lean on here in Gartner, and Ronny Remmington for helping make my time in Gainesville more enjoyable. And I want to single out, above all, Kyl e Thompson who has been my roommate for the past 6 years. More than anyone else, he has always been there for me to understand what I was going through and listen patiently when needed There is no way that I can fully express my gratitude. if I could have made it through graduate school without him. not just here at the University of Florida but through my life. Thinking back to all the little school from first grip ping a pencil to learning my multiplication tables, there have been so many educators who have each played a key role in my journey towards a PhD. At the time they knew me my future in life was (and still is unknown) but they each imparted so many little pieces of knowledge that have summed together to help me achieve this PhD. I will never be able to thank them all but they have each made a contribution, big or small, to my progression to this point and the continuation of the journey onwards. As a toke n of my appreciation, I dedicate this dissertation to them.
9 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ .......... 11 LIST OF FIGURES ................................ ................................ ................................ ........ 12 LIST OF ABBREVIATIONS ................................ ................................ ........................... 15 ABSTRACT ................................ ................................ ................................ ................... 17 C H A P T E R 1 Introduction and Background ................................ ................................ .................. 19 Physical and Electronic Structure of Graphene ................................ ...................... 19 Physical and Electronic Structure of Single Walled Carbon Nanotubes ................. 20 Transparent Conductive Thin Films of Single Walled Carbon Nanotubes .............. 22 Organic Semiconductors ................................ ................................ ......................... 26 2 Purification of Single Walled Carbon Nanotubes ................................ .................... 28 Overview ................................ ................................ ................................ ................. 28 Single Walled Car bon Nanotube Synthesis by Dual Pulsed Laser Vaporization .... 29 Conventional Purification of Carbon Nanotubes in the Rinzler Laborato ry ............. 31 Nitric Acid Reflux ................................ ................................ .............................. 31 Cross Flow Filtration ................................ ................................ ......................... 32 Shorting Pathways from SWNT Thin Films through Organic Layers ...................... 35 Conformal Layers to Achieve Flat Films ................................ ................................ 40 Magnetic Purification ................................ ................................ ............................... 46 High Speed Centrifugation ................................ ................................ ...................... 51 Continuous Flow Centrifugation ................................ ................................ .............. 54 3 S c hottky Barrier Modulat ion in Vertical F ield Effect Transistors E nabled by Low Density of States Electrodes ................................ ................................ ................... 67 Overview ................................ ................................ ................................ ................. 67 Energy Band Alignment at a Metal Semiconductor Interface ................................ .. 70 Schottky Barrier Height Modulation ................................ ................................ .. 70 Schottky Barrier Wi dth Modulation ................................ ................................ ... 71 Schottky Barrier Modulation in the CN VFET ................................ .......................... 72 Fabrication of Vertical Field Effect Transistors ................................ ........................ 73 Effect of SWNT Film Porosity on VFET Performance ................................ ............. 76 Graphene Based Vertical Field Effect Transistors ................................ .................. 80
10 4 D evice P hsics of V ertical F ield E ffect Transistors with V arying C hannel P roperties ................................ ................................ ................................ ............... 91 Overview ................................ ................................ ................................ ................. 91 Role of Channel Layer Thickness in Vertical Field Effect Transistors ..................... 91 CN VFETs with a Solution Processable Channel Layer ................................ ......... 96 n Type Vertical Field Effect Transistors ................................ ................................ 103 5 Organ ic Light Emitting Diodes Using T hin F ilms of Single Walled Carbon N anotubes as Anodes ................................ ................................ ........................... 109 Introduction ................................ ................................ ................................ ........... 109 Theoretical Background ................................ ................................ ........................ 110 Technical Approach ................................ ................................ .............................. 113 Results and Discussion ................................ ................................ ......................... 116 Conclusions ................................ ................................ ................................ .......... 120 6 Light Emitti ng Electrochemical Cells Using T hin films of Single Walled Carbon N anotubes as Electrodes ................................ ................................ ...................... 122 Overview ................................ ................................ ................................ ............... 122 Operating Principles and Scientific Background ................................ ................... 122 Single Emissive Devices ................................ ................................ ....................... 125 Experimental Methods ................................ ................................ .................... 125 Results ................................ ................................ ................................ ........... 127 Dual Emissive Light Emitting Electrochemical Cells ................................ .............. 129 Experimental Methods ................................ ................................ .................... 129 Results ................................ ................................ ................................ ........... 130 7 Conclusions and Paths Forward ................................ ................................ ........... 132 LIST OF REFERENCES ................................ ................................ ............................. 134 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 147
11 LIST OF TABLES Table page 2 1 Characterization of SWNT material following continuous flow centrifugation with predeposited electrodes ................................ ................................ .............. 62 2 2 Characterization of SWNT material following continuous flow centrifugation with postdeposited electrodes ................................ ................................ ............ 63 2 3 Elect rical c haracterization of SWNT material following continuous flow centrifugation ................................ ................................ ................................ ...... 64 4 1 Comparison of P3HT CN VFET to published, low voltage P3HT TFTs ............ 100
12 LIST OF FIGURES Figure page 1 1 Schematic indicating graphene lattice ................................ ................................ 22 1 2 Electronic density of states of met allic an d s emicond u cting nanotube s ............. 22 1 3 Th in f ilms of SWNTs and their appli cations ................................ ........................ 23 2 1 DPLV grown SWNTs in flask immediately prior to nitric acid reflux. ................... 33 2 2 Nitric acid reflux in progress in the clean room fume hood. ................................ 33 2 3 Crossflow filtration assembly ................................ ................................ .............. 34 2 4 Samplings of the eliminated supernatant during crossflow filtration ................... 34 2 5 Demo nstr a tions o f shor t ing pat h wa y s in S WNT / organic d evices ..................... 38 2 6 Characteristic TEM images of aggregated bucky onions ................................ .... 39 2 7 SWN T fil ms af ter n i tric a cid r e fl u x ................................ ................................ ....... 40 2 8 Schematic representat ion of co nformal transfer tech nique ................................ 41 2 9 AFM image of transferred SWNT film w ith shr u nke n the r mopl a stic ................... 42 2 10 AFM image of inwards protruding ridge in transferred SWNT ............................ 43 2 11 Amplitude image of flooded SWNT surface with p e netrate d po l ymer ................. 45 2 12 Current density of OLEDs with confrom al la y e rs ................................ ................ 46 2 13 A F M i m a g es of SWNT ma t erial before and aft e r ma gn etic puri f i ca tion .............. 49 2 14 Charactaristic TEM images following magnetic purification ............................... 50 2 15 Ho llow bucky onions remain fol l o w ing ma gnetic purif ication .............................. 51 2 16 AFM image of 17000 RPM centrifuged material. ................................ ................ 54 2 17 Cross section view of Contifuge Stratos continuous flow centrifuge. .................. 56 2 18 Optical microscope (5X ) image of cont i nuous flow c entri fuge d SWNT s ............. 58 2 19 Optical microscope (20X ) i mag e of con t inuous flow centrif u ged SWN Ts ........... 59 2 20 AFM images of SWNT fi l m after co n tinu o us fl o w centr i fugatio n ........................ 60
13 2 21 UV Vis spe ctra for S WNT films after con t inuous flow cen trifu gation .................. 66 3 1 Schematic of later al ch annel T FT and CN VFET ................................ ............... 68 3 2 E lectros tatic s im u lation o f o peratio nal me c hanism in CN VFE T ......................... 73 3 3 Labelled phot o graph of CN VFET on silicon substrate ................................ ...... 76 3 4 AFM scans a nd resis t a nce of increasing film densities ................................ ...... 78 3 5 Structure of DNTT CN VFE T on silicon. ................................ ............................. 79 3 6 CN VFETs with a DNTT channel using SWNT films of varying thicknesses. ..... 80 3 7 G VFET architecture drive scheme and tra n sfer t e chniq ue ............................... 83 3 8 A s say of qual ity of improve d trans fer te c hn i que ................................ ................. 85 3 9 Per forman c e of G V FET fo r gr a p h ene sour ce ele c tro des of var ying p orosit y .... 87 3 10 Comparison of graphene and carbon nanotube enabled VFETs ........................ 90 4 1 Structure of NPD and NPD CN VFETs ................................ ............................... 93 4 2 Perfo r mance of CN VFETs with an var iabl e th ick ness NPD channel layer ....... 95 4 3 Transfer curves for N PD based CN V FET ................................ .......................... 96 4 4 Schematic of P3HT based CN VFET ................................ ................................ 98 4 5 Per f ormance of CN VFETs using a P3HT channel layer. ................................ ... 99 4 6 Schematic of C 60 based CN VFET ................................ ................................ .. 106 4 7 P er f orm ance of C60 based CN VFET wi th Al and A u dra i n elect r odes ........... 107 4 8 Reverse injection in C60 b ased C N V FET w ith A l and A u drain electro d es ..... 107 4 9 Perf ormance of C6 0 base d CN VFET with a d edope d SWNT f ilm ................... 108 5 1 Structure of I TO and SWNT based MEH PPV OLEDs. ................................ .... 115 5 2 Luminance and current density of ITO OLE D wi t h a 100 nm MEH PPV layer. 116 5 3 Perfo rmance o f SWN T and ITO OLEDs with th ick a c tive la yers ....................... 118 5 4 P e r forma nce f o r S WNT and ITO OLEDs us i ng high purif ied SWNTs ............... 121 6 1 Operational mechanism for a light emitting electrochemical cell. ..................... 123
14 6 2 Device schematic for ITO and SWNT film based L E Cs ................................ .... 127 6 3 Per for m ance of ITO and S WNT b as ed LECs ................................ ................... 128 6 4 Device schematic of a dual emissive LEC. ................................ ....................... 130 6 5 P hoto g raph o f d ual emissive LEC in light and dark with emission ................... 131
15 LIST OF ABBREVIATION S AFM Atomic force microscope ALD Atomic layer deposition BCB Benzocyclobutene CMOS Complementary metal oxide semiconductor CN VFET Carbon nanotube enable vertical field effect transistor CV D Chemical vapor deposition DNTT D inap htho [2,3 f]thieno[3,2 b] thiophene DOS Density of electronic states DPLV Dual pulsed laser vaporization EDS Energy dispersive spectroscopy FE Field effect FET Field effect transistor HOMO Highest occupied molecular orbital ITO Tin doped indium oxi de LEC Light emitting electrochemical cell LiOTf Lithium triflate LUMO Lowest unoccupied molecular orbital MCE Mixed cellulose ester MEH PPV Poly[2 methoxy 5 (2' ethyl hexyloxy) 1,4 phenylene vinylene] MoOx Molybdenum oxide M w Molecular weight NPD N,N' di( 1 naphthyl) N,N' diphenyl 1,1' diphenyl 1,4' diamine) OLED Organic light emitting diode P3HT poly(3 hexylthiophene)
16 PAA Poly(acrylic acid) PDMS Polydimethylsiloxane PEDOT:PSS poly(3,4 ethylenedioxythiophene) poly(styrenesulfonate PEO Poly(ethylene oxide) P ET Poly(ethylene terephthalate) PLV Pulsed laser vaporization PMMA Polymethylmethacrylate RPM Revolutions per minute SCL Space charge limited SEM Scanning electron microscope SWNT Single walled carbon nanotube TEM Transmission electron microscope TFT Thin film transistor VFET Vertical field effect transistor
17 Abstract 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 ORGANIC ELECTRO NIC DEVICES USING GRAPHENE AND HIGHLY PURIFIED THIN FILMS OF CARBON NANOTUBE S AS TRANSPARENT CONDUCTIVE ELECTRODES By Evan Peter Donoghue August 2012 Chair: Name Andrew G. Rinzler Major: Physics The impressive electrical, optical and mechanical propertie s of thin films of single walled carbon nanotubes (SWNTs) and graphene have sparked intense interest and extensive research into these materials, with significant recent efforts seeking to incorporate them into organic electronic devices. Generally, this work ha s not taken full advantage of the unique properties of these materials, such as a l ow density of electronic states, mechanical flexibility and an enhanced surf ace area for charge injection Progress has been further stymied by parti culates in the S WNT material that create s vertical protrusions into the thin organic active layer. This dissertation will discuss applications in which the unique properties of these materials can be tested or exploited in practical organic electronic devices. The low density of electronic states found in SWNTs and graphene allows for modulation of the ir Fermi level providing a new degree of freedom for tuning electronic transport that was r ecently demonstrated in carbon nanotube enabled vertical field effect transisto rs (CN VFETs) T hin films of SWNTs or graphene w ere used to probe this Schottky barrier height and width modulation and demonstrate the first graphene enabled VFET as well
18 as demonstrating solution processable and n type CN VFETs A dditionally, thin films of SWNTs were incorporated into organic light emitting diodes and organic light emitting electrochemical cells to study whether the properties of the carbon nanotube films offer any intrinsic advantages over more conventional electr odes. The mechanical flexibi l i ty of the SWNT film also makes possible a new dual emissive device structure in which a light emitting electrochemical cell that incorporates transparent SWNT film s as both anode and cathode to emit light in both the f orward an d reverse direction In addition to this device based work, extensive research into carbon nanotube purification techniques will be discussed including the adaptation of a scalable purification technique not previously demonstrated with materials on this l ength scale. Material made available by this large scale purification technique wer e incorporated into CN VFETs that use the thinnest organic channels ever achieved in these devices These projects offer insights into the special role that SWNTs can play in organic electronic devices.
1 9 CHAPTER 1 INTRODUCTION AND BAC KGROUND Physical and Electronic Structure of Graphene The two dimensional honeycomb carbon lattice of graphene offers a material system that couples near ballistic charge transport (with mobilit ies exceeding 200,000 cm 2 /V s) 1 across a planar material with high optical transparency. Though graphene has been the subject of theoretical inquiry for over sixty years 2 4 it was not considered to occur as physically stable structure until 2004 when it was accessed experimentally for the first time through the mechanical exfoliation of graphene from graphite 5, 6 Since this initial realization, graphene synthesis has been demonstrated using a range of methods beyond mechanical exfoliation such as epitaxial growth 7 9 chemical vapor deposition (CVD) 10, 11 and reduction from graphite oxide flakes 12, 13 These techniques, and others, have allowed for the creation of large area films of graphene 11, 14, 15 and wide ranging studies into its fundamental physical properties 16 19 2 bonded carbon is the source of many of its interesting electrical properties that we seek to take advantage of in this dissertation. physical structure allows us to approximate its electronic structure using an orthogonal nearest neighbor tight binding model where its electronic states can be approximated by a linear combination of 2p z orbitals 19 Solving using the Schrodinger equation yields the dispersion relations for the bonding ( ) and anti bonding ( *) bands as: ( Eq. 1 1)
20 where k x and k y define the components of the k vector in the first Brillioun zone and is the nearest ne ighbor hoping integral (2.75 eV). This results in symmetr ic conduction and valence bands with respect to the Fermi energy that meet at six K and points also known as the Dirac points The density of states is zero at the Dirac point and increases li nearly as energy increases away from the Fermi energy 20 This low density of states will be exploited in the Chapter 3 discussion of vertical field effect transistors where we will use graphene to disentangle the op erational mechanisms of this new device architecture. Physical and Electronic Structure of Single Walled Carbon Nanotubes An individual single walled carbon nanotube (SWNT) can be conceptually visualized as a single graphene lattice rolled into a seamle ss cylindrical tube that can be microns to millimeters in length and between 0.6 nm and 10 nm in diameter. The wrapping of this tube perpendicular to the tube axis is defined by the chiral vector, C h which itself is composed of two non orthogonal vectors a 1 and a 2 that access each site of the graphene sheet (Figure 1 1 ) 21, 22 The chiral vector, C h is defined as: C h = n a 1 + m a 2 ( ( Eq. 1 2) which is more commonl y abbreviated by its vector indices (n,m) to define the structure, or chirality, of any given SWNT. The physical structure of the SWNTs allows us to consider them a quasi 1D material offering near ballistic on tube transport that remains defect tolerant. While a single graphene sheet is considered a zero gap semiconductor, the confinement along C h introduced in SWNTs has a significant impact on the electronic structure by imparting periodic boundary conditions in the circumferential direction and quantizin g the wave vector along the chiral vector, while the wave vector along the axis
21 of the nanotube remains continuous 21, 22 The 1D energy dispersion relationships of SWNTs are effectively cross sectio nal cuts of the 2D graphene energy dispersion relation. If this cut crosses one of the six K points where the and bands touch in the first Brillouin zone of graphene, the nanotube will also have zero bandgap making it metallic. However, if this cut avoids the K point, there will be a nonzero band gap and the nanotube will be semiconducting. This relation can also be calculated by applying the aforementioned periodic boundary conditions to the wave vector and solving for the energy relations in the Schrdinger equation to yield the electronic density of states as shown for a demonstrative metallic and s emicond ucting SWNT in Figure 1 2 21, 22 A non zero density of states at the Fermi energy (E=0), renders one third of all carbon nanotubes metallic while the remaining two thirds possess a finite energy gap between the valence and conduction bands and are semiconducting. This relationship can be formalized and the type of any given nanotube can be determined from the vector indices where any tube for which (n m)/3 is equal to an integer will act as a metal a nd all others will act as a semiconductor. The reduced dimensionality of SWNTs leads to a very low density of states near the Fermi level in all carbon nanotubes 21, 22 As a result, the Fermi level of a SWNT can be widely modulated through charge transfer chemical doping 23 26 as well as through gating by an external field 27 30 The implications of these Fermi levels shifts will be discussed in Chapter 3 where tuning of the Schottky barrier at the SWNT / organic semiconductor will admit device mech anisms not available for higher DOS conventional metals.
22 Figure 1 1 Schematic indicating graphene lattice with a 1, a 2 and the chiral vector, C h as shown. Parallel dashed lines indicate nanotube axis. In this example, a (5,3) tube is demonstrated. Rep rinted with permission from Dresselhaus et al 31 Figure 1 2 Electronic density of states of A) a (10,0) nanotube and B) a (9,0) nanotube. Reprinted with permission from Saito et al. 21 Transparent Conductive Thin Films of Single Walled Carbon Nanotubes While i ndividual SWNTs have demonstrated impressive characteristics and have been utilized to demonstrate a range of devices 27, 32 37 separation and manipulation of individual SWNTs is challenging and can be difficult to incorporate into practical devices. However, by assembling these SWNT into a thin film, they can act as the transparent con ductive electrodes required for many organic electronic device technologies (Figure 1 3 ). In the eight years since their introduction by Wu et al. 38 thin films of SWNTs have been incorporated into numerous devices across a range of
23 applications 29, 30, 39 47 some of which will be discussed in greater detail in this dissertation. Figure 1 3. A) Thin film of single walled carbon nanotubes taken by a scanning electron microscope. B) Transmittance of ITO and SWNT films of varying thicknesses. C) Diagram of various application enabled by thin films of SWNTs in the Rinzler laboratory. Though thin films of SWNTs preserve many of the properties of individual SWNTs, there are a few distinctions worth noting. Individ ual carbon nanotubes have been shown to be near ballistic conductors with intrinsic carrier mobilities in an individual semiconducting tube on the order of 100,000 cm 2 /V s, among the highest values reported in a semiconducting material 48 ; however, when a nanotube network is formed the conductance is dominated by the impedance to charge transport across tube tube junctions 49 While the resistance along an individual SWNT is o n the order of tens of k s (approaching the 6k limit predicted for length independent ballistic transport) 33, 50, 51 the resistance at a crossed metal metal or semiconducting semiconducting SWNT junct ion was found to be approximately 200 k while a metal semiconducting SWNT A B C
24 junction was several orders of magnitude worse due to the presence of a Schotky barrier 52 Thus, to first order, the conductivity of thin fil m networks of SWNTs is heavily dependent on average tube length, as longer tubes require fewer tube tube junctions, with the network conductivity ( DC ) proportional to average tube length (L) to the 1.46 power ( ) 52, 53 Likewise, the sheet resistance is inversely proportional to film thickness as thicker films offer a greater probability of the more conductive pathways across similar nanotube junctions. This relation goes as where t is the film thickness and the dc conductivity, is strongly dependent upon on factors such as the connectivity of nanotube nanotube junctions doping, and tube length which will vary widely with preparat ion technique 53 While clearly the impedance at SWNT SWNT junction plays a major role in thin films of SWNTs, the conductivity loss should not be overstated. Highly conductive and transparent films of single walled c arbon nanotubes have been produced, with conductivities as high as 6600 S/cm 2 thick films that are 70% transparent through the visible spectra 38 This level of performance is below that of conventional transparent conductive oxides, such as t in doped indium oxide (ITO) which at 90% transparency can posses sheet resistances approaching in some applications where their similar work functions (4.6 4.9 eV for SWNTs 40 vs. 4.4 4.9 eV for ITO 54 ) make SWNT films a promising alternative. In other devices, the SWNT film can offer benefits and access modes of operation that are not achievable by ITO or conventional metals. For example, unlike conventional electrodes which offer only a planar surface for charge injection, the 3D
25 structure and porosity offered by a SWNT film offers an enhanced surface area, particularly when used with solution process able active layers that can penetrate into the porous SWNT network and access charge injection from all sides of a SWNT. In a 50 nm thick film, where the inter SWNT porosity is on the order of the length scale of the SWNTs th emselves, it was estimated via double layer capacitance measurements that the surface area was 2.5 times greater than a planar palladium electrode 55 Other advantages of SWNT films over conventional metals include their mechanical flexibility, room temperature and pressure film fabrication, and low density of electronic states allowing for Fermi level shifting (expanded upon in Chapter 3). Fabrication of these thin films of SWNTs can be achieved through a variety of methods 42, 56 59 though one of the most common techniques is through vacuum filtration as described by Wu et al 38 Purified SWN Ts (growth and purification of SWNTs will be covered in Chapter 2) are suspended in a surfactant solution which is subsequently vacuum filtered in a dead ended filtration through a porous, mixed cellulose ester (MCE) membrane, trapping the SWNTs on the sur face of the membrane as the liquid is sucked through the membrane. After allowing the membrane to dry, washing with deionized water removes the surfactant shell coating the SWNTs. These films can be adhered to the surface of a substrate through heat and pressure, with the evaporation of a small amount of water or isopropanol helping to pull the film into more intimate contact and forming van der Waals bonds between the SWNT film and the substrate. An acetone vapor bath gently dissolves the MCE membrane, leaving a pristine SWNT film adhered to the substrate and ready for device preparation. This technique has enabled SWNT films transferred to materials including glass, silicon, sapphire, quartz, Teflon,
26 poly(ethylene) and poly(ethylene terephthalate ) and offers facile control over film thickness / density simply through manipulation of the concentration and volume of SWNT solution that is filtered. These films will form the foundation of much of the work explored in this dissertation. Organic Semiconduct ors Though the semiconductor electronics industry has been dominated by research into inorganic materials such as silicon or III V semiconductors, developments in organic semiconductors over recent decades has revealed the important role that these materia ls can play in practical devices. The ability to widely tune the optoelectronic properties of organic semiconductors through synthetic structural modification enables their incorporation into an array of organic electronic devices that span a similarly wi de range of device properties and functions. While inorganic materials have typically demonstrated higher levels of performance, organic semiconductors and the organic electronic devices that they make possible offer advantages in their potential for low energy, inexpensive manufacture with a high throughput for flexible and lightweight devices that may offset the performance gap by enabling affordable, widespread electronics. Organic semiconductors are conjugated carbon based materials tha t can either b e small molecules ( typically deposite d through thermally evaporation) or polymers with repeated structural units which are traditionally dissolved in solution and depos i ted through techniques such as spin coating ink jet printing or spray casting. These materials derive their conductive properties from the delocalization of their electrons across the molecular chain, enabling long range electronic mobility. Materials are often
27 characterized by their highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) which develops into a band like structure as the number of energetic states increases, analogous to the valence and conduction band, respectively, of inorganic materials. Charge transport within organic semiconducting layers is controlled by two processes: intramolecular mobility and intermolecula r transport. Though mobilities within individual organic semiconductors molecules can be quite large, because the organic semiconducting layers are composed discrete molecular units held together only by relatively weak van der Waals bonds, the mobilities of these materials in bulk are orders of magnitude below their inorganic counterparts. Transport across molecules occurs via a hopping of free carriers from molecule to molecule, a process that is heavily influenced by factors such as the crystallinity, e nergetic disorder, conjugation length and charge traps present in these materials as well as the direction of charge transport as many of these materials can have large anisotropies in their mobilities. As will be seen throughout this dissertation, the lo w mobilities of organic semiconductors affects the performance of electronic devices by restricting organic layers to be just a few hundred nanometers thick motivating the purification work of Chapter Two as well as the novel vertical field effect transi stor of Chapters Three and Four.
28 CHAPTER 2 PURIFICATION OF SING LE WALLED CARBON NAN OTUBES Overview The use of single walled carbon nanotubes (SWNTs) in both organic electronic devices and general studies is largely reliant on access to pristine carbon n anotubes. As the bulk growth of SWNTs generates amorphous carbon, fullerenes and other carbonaceous structures in addition to residual metallic catalyst particles 60 62 careful p urification of the SWNT growth product is required. The importance of this purification is especially relevant in organic electronic devices where the relatively low mobility of organic materials limits organic layers to the order of a few hundred nanomet ers or thinner. The presence of particulates in the SWNT electrode can create shorting pathways that extend through the organic layer, severely limiting device performance. On a laboratory scale these issues have been overcome through the use of thick or ganic active layers 43 or a highly doped, conductive hole transport / planarizing layer 41, 63 but, while such workarounds allow devices to function at reaso nable levels, performance still remains below what can be achieved in devices with thinner organic layers. Fabrication is simplified and material usage reduced if these layers can be eliminated. The purification of carbon nanotubes has progressed signific antly since early efforts focusing on oxidative means 64 71 or low speed centrifugation 72, 73 Density gradient centrifugation has recently enabled impressive advances allowing selection by electronic type, band gap, diameter or even chirality 74 76 While this level of control is indeed impressive, it relies on separation of the bundled SWNTs by prolonged ultrasonication 37 Sonochemistry done by high energy cavitation bubbles during this
29 process can introduce defect sites into the SWNT sidewall and even cut the nanotubes 71, 77 79 to significantly shorter lengths. As a result, this technique tend s to yield lower conductivity materials that are not ideal for use as transparent conductive electrodes. Further, the scalability of such a technique has not been studied and is currently limited by what can be a ccommodated by the centrifuge rotor and vials (though schemes to overcome this have been suggested 74, 80 ). While batch processing can provide enough material to accommodate present laborator y needs, real world manufacturing is likely to require a significantly higher throughput. In this chapter, I will give a brief overview of SWNT growth as it pertains to purification and review the conventional purification techniques that I have carried out in the Rinzler Laboratory. After discussing the sources of shorts found in films made from this material after standard purifications, I will present several approaches, some novel and some adapted from known literature, towards further purification o f the nanotubes and discuss how I have implemented these procedures to eliminate shorting pathways. Most notably, I will introduce the first demonstration of the use of a continuous flow centrifuge to purify materials of a SWNT length scale and show how t he extremely high levels of purity achieved through this method has enabled organic active layers as thin as 100 nm in this readily scalable technique Single Walled Carbon Nanotube Synthesis by Dual Pulsed Laser Vaporization Sumio Iijima discovered carbon nanotubes in 1991 as the residual by product remaining on the cathode of his arc reactor during fullerene growth 81 In the ensuing years, much effort w as focused on the development of a means of synthesizing high quali ty, defect free SWNTs at low cost for the studies and potential applications that require their use. Numerous growth methods and techniques have been developed in
30 this pursuit, most notably arc discharge 82, 83 chemical vapor deposition (CVD) 84, 85 and pulsed laser vaporization (PLV) 61, 86, 87 Though each t echnique has its own merits and disadvantages, PLV grown SWNTs are typically considered to have the lowest defect density while maintaining long tube length, rendering the high conductivity material desired for organic electronic devices. As such, all stu dies discussed in this dissertation use PLV grown SWNTs synthesized either in the Smalley Laboratory at Rice University or using an in house built and operated system within the Rinzler Laboratory at the University of Florida. Though synthesis of carbon n anotubes was carried out by colleagues and is not part of this dissertation, the growth of high quality SWNTs is an essential part of all experiments and as such, it will be briefly summarized here. In this technique (also known as dual pulsed laser vapori zation, or DPLV), two high powered laser pulses strike a carbon target (98 wt % carbon, 1 wt % each Co and Ni) in a quartz tube furnace at high temperature (1150 1200 C) under a partial argon environment (500 mTorr). The first pulse is at 532 nm (15 W) to preheat the target followed by a second pulse 39 ns later, after the heat pulse has penetrated into the target, at 1064 nm (25 W) to locally vaporize the target and form a plume of ablated carbonaceous precursor s composed of mono atomic carbon as well as its dim ers, trimers and longer chains that is confined by a smaller diameter inner tube. The plume lasts just microseconds and in the absence of metal catalyst particles these structures would rapidly close and form self satisfied fullerenes; however, in the presence of ~1% catalyst material, SWNTs are formed 61 In this manner high quality SWNTs can be grown at a rate of a few hundred mg/hr and collected as a wispy, tumbleweed like soot at the end of each run.
31 Convent ional Purification of Carbon Nanotubes in the Rinzler Laboratory The SWNTs in the PLV grown soot has by products including metal catalyst satisfied carbon shells with a di ameter of 1 5 nm that can either be hollow or can have a metal catalyst core. These assorted contaminants intrinsic to the growth process must be eliminated for most applications and there are a variety of techniques used and studied indeed a few novel techniques will be explored here. In this section I will introduce typical techniques used in the Rinzler laboratory, that constitute refinements of methods developed by Rinzler et al. in the Smalley Laboratory 69 before examining additional techniques explored in my work. Nitric Acid Reflux The PLV grown, SWNT soot (Figure 2 1) is dispersed in 1.6 M nitric acid which then undergoes a 45 hour nitric acid reflux in which the solution is boiled in a round bottom flask with a condenser returning evaporated liquids back to the solution (Figure 2 2). During this process, the nitric acid readily attacks and dissolves any loose metallic catalyst particles spalled during the growth as well as any amorphous carbons which are dissolved as humic acids. While most carbon structures containing self satisfied carbon bonds (such as fullerenes, bucky onions or the pristine SWNT sidewall) are not attacked, any dangling bonds and even the nanotube end cap, which is less stable than t he nanotube sidewall, can be attacked and gradually eaten away followed by the gradual destruction of the nanotube from the end. Similarly, defect sites on the carbon nanotube sidewall offer dangling bonds that permit dissolution by nitric acid and can cu t the carbon nanotubes into pieces 71 All this has the detrimental effect of substantially lowering the post reflux SWNT yield but with the benefit that only the long, pristine,
32 defect free SWNTs remain; yielding a final product that may have better electrical properties than would be achieved if these lower quality tubes were not destroyed. The nitric acid reflux has an added benefit beyond dissolution of the various D PLV by products; it also serves to substantially dope the SWNTs p type. This doping helps give high conductivities to the semiconducting SWNTs, rendering them near metallic in their behavior (and indeed, we shall treat them largely as such here). The SWNT doping achieved here is a relatively stable process and remains essentially unchanged even after years in solution. After 45 hours of refluxing in nitric acid, the SWNTs are re moved from the 1.6 M acid solution by repeated dilution (with deio nized water) and centrifugation/ decantation steps in which the nan otubes settle as sediment. This process is repeated until the solution reaches pH 4 5 at which point the nanotubes are suspended in a slightly basic surfactant solution (1% Triton X 100 with dissolved NaOH to achieve pH 10 or 11). At this stage, many of the associated growth products have been broken down but remain in solution either as humic acid or partially dissolved solids. Cross Flow Filtration Following the nitric acid reflux, some of the dissolved by products have be en removed during the dilution and centrifugation steps; however, it becomes necessary to remove the remaining fraction in addition to the assorted self satisfied carbon structures such as fullerenes and bucky onions that are not desired in the final product. This is partially achieved through a cross flow filtration in which the nanotube solution is cycled through a hollow fiber filter with 200 nm pores. The nanotubes, microns in length, cannot permeate through these pores and remain trapped within the hollow fibers while
33 Figure 2 1. DPLV grown SWNTs in flask immediately prior to nitric acid reflux. Figure 2 2. Nitric acid reflux in progress in the clean room fume hood. The SWNTs are dispersed in 2.6 M nitric acid in a round bottom flask which is submerged in a hot oil bath mai ntaining the temperature at 230 C. Chilled water is flowed the cylindrical condenser on top to prevent escape of gaseous nitric acid.
34 Figure 2 3. Cross flow filtration assembly. The peristaltic pump is enclosed by a large box and connected to a fan to remove particulates before they can escape to the clean room environment. Fig ure 2 4. Samplings of the eliminated supernatant during crossflow filtration with every sample representing an additional hour of crossflow filtration. Far left is sampled 1 hour after beginning crossflow filtration. Far right is sampled after 22 hours of filtration.
35 the dissolved by products and structures smaller than 200 nm are flushed out and replaced with fresh surfactant solution (Figure 2 3). Periodic reversals of the flow prevent the SWNTs from accumulating on the inside of the hollow fibers and clogging the pores. Similarly, brief sonication once an hour prevents the SWNTs from excessive bundling which might clog the fibers. The cross flow filtration is run until t he permeate runs clear (Figure 2 4), indicating that the majority of the dissolved by products and structures smaller than 200nm have been removed. This process can take over 60 hours to complete and involve flushing more than 75 L of pure surfactant buff er solution. At this stage, having dissolved and eliminated amorphous carbons and removed particulates smaller than 200 nm, the material can go in a variety of directions for future steps to eliminate contaminants that did not fit through the 200 nm pore s ize of the hollow fiber filter. The traditional processes within the Rinzler Laboratory have involve d a 6000 RPM centrifugation of the concentrated nanotube solution to remove many of the larger particulates as sediment and then a dead ended filtration of a diluted fraction of the solution through 650 nm pores. At this stage, having eliminated particles greater than 650 nm and smaller than 200 nm, the solution is used to make films for general use; however, the remaining particles in the 200 nm 650 nm r ange remain highly problematic for the performance of thin organic devices Because we are searching for a universal means of eliminating particulates, these final two steps (low speed centrifugation and 650 nm filtration) are not incorporated into the pu rification processes discussed below Shorting Pathways from SWNT Thin Films through Organic Layers Shorting occurs in SWNT/organic devices when the conductive SWNT thin film extends upwards and protrudes through a relatively thin (100 500 nm) organic la yer
36 deposited on top of the SWNT film (Figure 2 5A). These protrusions can arise from a range of issues some intrinsic to SWNT growth and others from sample handling. A common problem is particulates on the surface of the substrate, which can come from an array of sources such as shards generated from dicing the substrate or airborne particulates landing on the sample. Though these particles themselves may not be conductive, a SWNT film transferred on top of the particle makes this protrusion of the fi lm a potential electrical short (Figure 2 5B). These problems can largely be overcome through proper handling and cleaning. To this end, one early project was a comparison of different cleaning techniques on glass, with characterization by atomic force m icroscopy (AFM) to assay efficacy. It was determined that dicing glass under soapy water (alconox) followed by scrubbing with an ultrasonic toothbrush was more effective than simple ultrasonication in cleaning substrates. Through the adoption of this cle aning technique in addition to the development of other methods and appropriate sample preparation in a cleanroom environment, shorting from external particulates can largely be avoided. Another source of shorting that was discovered in early SWNT films a rose from defects and scratches that existed on the surface of the mixed cellulose ester (MCE) membrane, likely introduced during the membrane manufacturing process. SWNTs could accumulate or deposit in these scratches and when the SWNT film was transferr ed to a substrate, the accumulated SWNTs would create ridges and folds in the nanotube film (Figure 2 5C). These features could extend through the organic active layer. To avoid this problem, membranes were ordered and tested from 5 different
37 m anufacture rs and it was determined that Sterlitech membranes largely avoided the scratches seen in membranes from other manufacturers. These are two examples of shorting pathways arising from extrinsic issues which are avoid able with appropriate care and handling. Much more prevalent and difficult to remove are the growth products intrinsic to the synthesis of SWNTs. Though exposed metal catalyst particles and amorphous carbons are dissolved during the nitric acid reflux, there are many other carbonaceous structure s that have self satisfied carbon bonds. These materials are so difficult to remove because they are structurally and chemically similar to the SWNTs and can possess a similar density. Most techniques that would chemically target these structures would l ikely also simultaneously attack the SWNTs. Figure 2 6 shows high resolution transmission electron microscope (TEM) images are formed around metal cores that were revealed by energy dispersive spect roscopy (EDS) to be the cobalt and nickel catalysts from SWNT synthesis Individually, these bucky onions are small (~5 nm) and should be flushed out during the cross flow filtration but we observe d that they frequently cluster together, forming extended structures that can be several hundred nanometers in size and do not seem to break up with moderate sonication. I t was found that the major fraction of the particulates were composed of aggregates of these bucky onions, both hollow and with a metal cataly st core. Sampling by EDS found very few external contaminants present in the sample, that is to say very few particles composed of elements other than carbon, cobalt or nickel, which indicates that these particulates are indeed being generated during the SWNT synthesis.
38 Extensive TEM, done over wide regions of the nanotube film, indicates that these bucky onion agglomerates are the predominant particulate that remains after a nitric acid reflux and cross flow filtration. Atomic force microscope images rev eal how problematic these can be (Figure 2 7A & B) in thin SWNT films. A secondary nitric acid reflux and cross flow filtration was tested to break up and filter out these agglomerates with some initial success (Figure 2 7C & D); however, it did not appea r that we would be able to fully eliminate these bucky onion clusters using such means. Further, the secondary nitric acid reflux did increase the sheet resistance of these films by 10 20% so it seems that additional nitric acid refluxes will further da mage the SWNTs consistent with reports that nitric acid refluxes of already purified material attacks the SWNTs more aggressively than in an initial reflux where the oxidation debris products actually protect the SWNTs from damage 88 T he remainder of this chapter will explore methods to overcome and avoid these particulates intrinsic to the SWNT suspension. Figure 2 5. A) AFM image of SWNT film ridge protruding through organic material with shadowed region behind ridge, B) SEM image of particle trapped under SWNT film, C) AFM image of ridge in SWNT film caused by scratches in MCE membrane. B C A
39 Figure 2 6 Characteristic TEM images of aggregated bucky onions and metallic catalyst particles which are indicated by the dark, high contrast regions. The scale varies across the images as indicated In many images, particularly at high magnification, the graphitic walls of the bucky onions and SWNTs can be resolved.
40 Figure 2 7. A) Height and B) ampl itude AFM image of film fabricated from PLV grown SWNTs after undergoing a 45 hour nitric acid reflux (2.6 M) and 27 hours of crossflow filtration. C) Height and D) amplitude image after secondary nitric acid reflux (21 hours, 66 hours total) and secondar y crossflow filtration (11 hours, 61 hours total). Full height scale in each image is 300nm. Conformal Layers to Achieve Flat Films Though this chapter largely discusses purification in the sense of eliminating particulates, one early effort focused on ove rcoming rather than removing these brought into intimate contact with the substrate and, while particulates or defects may cause protrusions on the side away from the substrate, the side of the film in contact with the substrate is forced to be as flat as the substrate itself. It is this side of the film that we attempted to access. A C B D
41 To access this side of the SWNT film, a m) water soluble layer was spunc a s glass itself (root mean square roughness is 0.16 nm vs. 0.21 nm for glass) ; here, poly(acrylic acid) (PAA) wa s used, thou gh other water soluble materials can b e used as well. The SWNT film wa s transferred onto this surface (Figure 2 8A) with care taken during the transfer process to use orthogonal solvents that do not dissolve the PAA. Though the PAA layer was initially e xtremely flat, it is so water soluble that a surface layer was dissolved by the condensation of water vapor from the ambient lab environment during the evaporation of solvents used in the membrane dissolution step of the film transfer. The resolution to t his problem was to perform these transfers in a glove bag purged with nitrogen to eliminate the major source of atmospheric water condensation. Following the film transfer from the membrane to the PAA layer, t he new top surface of the SWNT film wa s then c oated with a thermoset or a thermoplastic by sandwiching this material between the SWNT film and a second glass substrate (Figure 2 8B ). The thermoset or molten thermoplastic conforms to the SWNT film morphology including particulate protrusions followed by its solidif ication The glass/PAA/SWNT film/thermoset/glass sandwich wa s then submerged into water and the PAA layer dissolved away, exposing the flat SWNT film side bonded via its other side to th e thermoset (or thermoplastic) layer on glass Figure 2 8. Schematic representation of A) SWNT film transferred to PAA layer, B) with thermoset covering SWNT film and sandwiched with second glass substrate, C) after PAA is washed away and first glass sub strate is removed. Note that in C) the setup is rotated. Image is not to scale.
42 A number of thermoplastic and thermosetting systems were explored as the conformal layer to conform to the protruding features in the SWNT films. Most of these exhibited pro blems. For example, polydimethylsiloxane (PDMS) underwent significant shrinkage upon curing that built stresses into the material around the particulates in the SWNT films. Combined with the high flexibility of the cured PDMS these stresses displace d the p articles into the region where the flatness is desired (Figure 2 9). While this system was a failure, it highlighted the need for a low shrinkage, high durometer system. Other thermosets and thermoplastics worked well, eliminating large wrinkles that were visible on the optical microscope, but proved to be incompatible with the solvents required for subsequently deposited solution processable polymers (Figure 2 10). Figure 2 9. AFM image of transferred SWNT film A) using a thermoplastic that had too high a coefficient of thermal expansion, causing wrinkling and >100 nm variations and B) using a thermoset that shrank during curing, causing stresses and forcing inset particles upwards. After consulting with numerous manufacturers and experimenting with approximately 10 different thermosets and thermoplastics to determine material B A
43 properties, a UV cured polymer system was identified that was not dissolved in chlorobenzene and seemed to have all the other desired characteristics. Films were transferred as described using this polymer system and it was found that the UV cured polymer penetrated through the pores of the SWNT film as is shown in Figure 2 11. Organic light emitting diodes made from these samples exhibited extremely low current densities and l ittle luminous output suggesting that the flooded polymer was preventing charge injection into the organic layer. To overcome this issue, the transfer technique was modified by the deposition of a Parylene C protective layer on top of the SWNT film prior to deposition of the UV cure polymer, attempting to form a barrier to prevent the polymer from flooding the surface. This appeared to succeed and the flooding seemed solved; however, devices fabricated on such substrates continued to fail to achieve appre ciable current densities or any light emission. Figure 2 10. A) AFM image of inwards protruding ridge in transferred SWNT with B) a line scan indicating features less than 30 nm. C) The ridge is shown in the optical microscope (with AFM tip during s can) to demonstrate that the size of this feature that can be overcome. A B C
44 Another possible source of this failure was the presence of water vapor. The dissolution of the PAA layer is a diffusion limited process that took weeks for the 1 inch square substrat es used during which time water diffuse d into the voids within the SWNT film and the conformal layers. If water vapor remained and later escaped into the organic layers or calcium contact of an OLED, it could poison the contact and inhibit charge injecti on. To test this, the substrates were baked for 7 days at 130C under vacuum (~1 Torr) after PAA dissolution in water but prior to OLED fabrication. There was no improvement in device performance. Discussions with the manufacturer of the UV cure epoxy revealed that there is an oxygen sensitive component that does not cure after exposure to oxygen The a mbient laboratory environment may be creating an incurable component that (despite the Parylene C barrier) slowly work s its way into bulk of SWNT film either limiting charge injection or poisoning the MEH PPV layer. To avoid this issue, the assembly of the substrate onto the conformal layer and subsequent UV exposure was moved into the inert argon environment of a glove box. This finally enabled some l evel of luminance and charge injection, but both metrics were 1 2 order of magnitudes lower than in conventional devices. This demonstration of charge injection and luminance was a promising indicator that we were on the right path but perhaps there stil l was some barrier layer forming. I tested both a brief oxygen plasma ashing as well the local application of high currents through a mercury drop contacting the SWNT film to attempt to remove this barrier, to no avail. Finally, I attempted using a thick er (150 nm) SWNT film so that any problematic materials or epoxy components would have a more tortuous path to the
45 SWNT surface though the thicker film would admittedly absorb more light. OLEDs built on these devices demonstrated no shorting, even on de vices using 100 nm active layers. The I V characteristics looked similar to that observed in control ITO OLEDs (Figure 2 10); however, there was no light emission from these devices. At this stage, we have demonstrated proof of principle in a new method o f fabricating flat SWNT films that avoid shorting, even in devices using thin active layers. However, after one year of testing approximately ten different thermosets and thermoplastics and overcoming many obstacles, we had failed to find a material that was compatible with both the PAA layer used in the sacrificial transfer and the organic active layer. Though it may be possible that such a material exists, the time intensive optimization process made further pursuit of this technique an ineffective use of effort and this approach was abandoned. Further efforts were focused on means to eliminate the problematic particulates. Figure 2 11. Amplitude (V) image of flooded SWNT surface demonstrating penetration of UV cured polymer surrounding SWNTs.
46 Figu re 2 12. Current density of OLEDs with 150 nm MEH PPV organic layer for SWNT using conformal UV cure polymer and brief plasma clean to remove barrier (orange) vs ITO control device (green). It is clear that the devices are not shorted; however while the ITO device achieves >1000 cd/m 2 luminance, the SWNT device emits no light. This comparison is made to show the similarities in the devices and lack of shorting a more in depth description of OLED performance will be discussed in Chapter 5. Magnetic Pu rification After the failure of the conformal layers (which sought to overcome all particulates, regardless of source) in working devices, the remaining efforts were targeted at the specific particulates that remain. Based on observation that many of the remaining particulates were bucky onions with a metallic catalyst core, methods were sought to preferentially eliminate these metallic bucky onions. Attack by acid in this case was not an option because the carbon shell is self satisfied and any means tha t would damage these objects would likewise damage the SWNTs themselves. A method published by Kim and Luzzi to magnetically purify the sample consisted of filtering the SWNT suspension through iron granules in a strong magnetic field to trap and eliminat e the impurities in the strong local magnetic field gradients formed around protrusions in the iron granules 89
47 To test the magnetic purification scheme, DPLV grown SWNTs were used that had been pre purified through two nitric acid reflux and cross flow filtration cycles to eliminate and break up the bulk of these bucky onions (Figure 2 13A B). A magnetic purification was attempted to remove the remaining metal/bucky onion aggregations. A buret (with stopcock) that was 15 mm in diameter was filled to a height of 15 cm with iron granules ~1 2 mm in size. Seven pairs of strong rare earth magnets were placed on the outside of the buret with an alternating polarity such that a locally inhomogeneous field was created th rough the iron granules. After thoroughly flushing these iron granules with deionized water the S WNT solution was flowed through the buret at a rate of 4 mL/min. Dilute f ilms were made from this solution before (Figure 2 13A D) and after magnetic purif ication (Figure 2 13E F). These films were subsequently transferred to copper TEM grids and studied under TEM (Figure 2 14. A F) imaging and by EDS (Figure 2 14. G J). These studies demonstrated that most of the metal catalyst particles had been successf ully removed by the magnetic purification. The few catalyst particles remaining were typically singular (not in an agglomerate). There did exist many other carbonaceous materials as well as significant levels of iron contaminants that had been introduced during the magnetic purification. To eliminate these iron contaminants, baths in concentrated nitric or hydrochloric acid were used to dissolve the metals with relatively high success as indicated by both TEM imaging and EDS which reveal no evidence of r esidual iron. Hydrochloric acid can, in theory, contribute to the dedoping of the heavily p doped SWNTs but these effects were not observed here in UV Vis spectra or resistance measurements.
48 The ultimate fatal flaw in these experiments became excessive ma terial loss as well as too many non metal filled bucky onions that remained. In initial experiments, the SWNT concentration in the solution decreased by several orders of magnitude during the magnetic purification. This was attributed to two causes. Fir st, the use of Triton X as a non ionic surfactant is potentially problematic as we were using it at a pH of 10 11. Free ions can interfere with the suspendability of SWNTs in Triton X and the SWNTs can flock out. To test this idea, an ionic surfactant, s odium cholate, was tested with losses similar to those observed in Triton X suggesting that the surfactant type does not play a major role. The second loss mechanism is from the bundling of SWNTs in solution, held together by van der Waals bonds. If bucky onions are bound or trapped in these bundles, as they are magnetically removed the associated nanotube bundles may likewise be removed. Further, as additional SWNTs flow past these trapped bundles, they may also bind to the structures. With many potentia l trapped bundles, there are opportunities for such losses to occur which could explain the significant reduction in material. Sonication can break apart these bundles but carries the risk of further damaging the SWNTs due the sonochemistry done by the ve ry high energy pulses locally present in cavitation bubbles 71, 77 79 Magnetic purification could effectively eliminate many of the problematic catalyst filled bu cky onions but it proved labor intensive and led to problematic material loss. These limitations might be acceptable if the resultant material was highly pure; however, even after magnetic purification enough hollow bucky onions remain ed (Figure 2 15)
49 tha t devices made from films of this material continued to exhibit electrical shorts A more universal purification scheme wa s needed Figure 2 13. A) Height and B) amplitude AFM image of film fabricated from PLV grown SWNTs after underg oing a 45 hour nitric acid reflux (2.6 M) and 27 hours of crossflow filtration. C) Height and D) amplitude image after secondary nitric acid reflux (21 hours, 66 hours total) and secondary crossflow filtration (11 hours, 61 hours total). E) Height and F) amplitude image following magnetic purification. All images show material that has not been centrifuged or filtered (beyond cross flow). Full height scale in each image is 300nm. A C E B D F
50 Figure 2 14. A F) Charactaristic TEM images following magnetic purification. Dark particles can still be seen but these are no longer bucky onions. G J) Examination through energy dispersive spectroscopy (H, J) at points labeled rification process however this can be removed easily through an acid bath. A D B C E F G H J I
51 Figure 2 15. Hollow bucky onions remain after the metallic, cobalt and nickel catalyst filled bucky onions are removed during the magnetic purification. In these images, t he residual iron has been dissolved away in an acid bath. H igh Speed Centrifugation Many of the particulates remaining after the nitric acid reflux and cross flow filtration could be centrifuged out using a relatively low speed centrifugation at 6000 RPM ( 2500 g ) in a fixed angle rotor centrifuge; however, there remained a significant
52 fraction that were not be removed. Early efforts performed twelve, hour long centrifugations at 6000 RPM, collecting the supernatant after each run. While the first two centr ifugation passes showed significant improvement, the process quickly began yielding diminishing returns with little to no improvement in particulate removal by the twelfth pass. Material that had been centrifuged twice was used to create PLEDs that operat ed using a 250nm spun cast active layer, though shorts still remained and additional centrifugations (up to ten ) yielded no additional improvements in terms of working pixel yield. With repeated centrifugations at 6000 RPM offering little improvement beyon d modest initial gains, it became natural to wonder how centrifugation at higher speeds could affect material quality. Ultrahigh speed centrifugation (>100,000 g ) has been used to sediment SWNT bundles and nanotube aggregates and isolate individualized carb on nanotubes though such methods require aggressive sonication to debundle the nanotubes which can be damaging 37 More recently, moderately high speed centrifugations (20,000 g ) have been used to remove carbon n anoparticles and amorphous carbons 90, 91 By using the same Sorvall centrifuge used for previous centrifugations but changing the rotor out for a different model, rotational speeds of 17,000 RPM (23,000 g ) can be achieved though less than 70 mL can be centrifuged at a time enough for just a few SWNT films. To test this high speed centrifugation, material that had been nitric acid refluxed and cross flow filtered twice was then centrifuge d at 17,000 RPM in sequential 1 hour runs with the resultant material being assayed through characterization of thin films of the material largely via optical and atomic force microscopy (AFM) as well as further
53 characterization by transmission electron mi croscopy and energy dispersive spectroscopy (EDS). After the first pass, there was a drastic improvement, though a non negligible fraction of large impurities remained as could be seen in an optical microscope. The second and third centrifugation passes succeeded in eliminating the major fraction of impurities, leaving a film that was particulate free in ~90% of the 5 x 5 m 2 AFM scans taken (Figure 2 16). High resolution TEM images of the few remaining particulates showed that the major fraction were agglomerates of hollow bucky onions, like those shown in Figure 2 6. As these hollow bucky onions are empty, they are rela tively buoyant in solution and are not readily centrifuged away in a standard centrifuge scheme where the sediment, in this case assorted denser carbonaceous impurities and catalyst filled bucky onions, is trapped against the bottom side wall of the centri fuge bottle while the SWNT rich supernatant with buoyant bucky onions is collected for further us e (the sediment is discarded). An e ffort was made to preferentially collect the middle of the suspension, avoiding the top portion where more buoyant obje cts might be found, but no improvement in high resolution TEM images of these films was found relative to a control sample where both fractions were collected equally. Despite the few remaining hollow bucky onions, the gains made in this ultrahigh speed ce ntrifugation were impressive and would be beneficial to implement into our standard operating procedure; however, several factors limited its benefit The largest issue was that the centrifuge rotor was limited in the total mass that could be accelerated to high speeds, meaning that at 17,000 RPM, less th an 70 mL (enough for just a few dilute films) can be centrifuged. Larger volumes would require reduction in
54 centrifugal speeds. Even for lab scale applications, and certainly for industrial scale uses, a means of increasing the throughput was necessary. The second issue with this small batch centrifugation lay in the handling. After centrifugation was complete, the centrifug e gradually slows to a stop under its own friction over 20 minutes and the cent rifuge is opened. The vials must then be carefully removed and moved to the cleanroom where the supernatant is manually pipetted out and collected Each of these steps introduces external forces and turbulence into the solution in the vial, potentially d islodging the sediment collected at the bottom of the vial and resuspending it with the supernatant. This can be minimized, but not avoided, with repeated centrifugations. For these reasons, a centrifuge that could accelerate a larger volume to high cent rifugal speeds while also avoiding the handling issues that limit the efficacy of the fixed angle rotor centrifuge was desired. Figure 2 16. AFM image of 17000 RPM centrifuged material. Continuous Flow Centrifugation Continuous flow centrifugation, con sisting of a hollow rotor spinning relative to stationary inlet and outlet tubing, allows for the high speed centrifugation of an effective ly limitless volume of material. It has long been used in micron scale biological
55 separation applications but has ne ver been demonstrated to be useful in nanoscale separations. Th e technique offers the advantage of a high throughput, restricted only by the amount of sediment that can be collected in the rotor, while still attaining high centrifugal forces to e ffect the separation of a sediment from a supernatant solution. The cross section of the Thermofisher Contifuge Stratos continuous flow centrifuge rotor is shown in Figure 2 17. The head of the rotor, colored in shades of gray, remains stationary while the remaini ng components are accelerated in the centrifuge housing and rotate at 17,000 RPM (24,000 g ). Injection of glycerin into the outer housing forms a liquid seal, creating an air pocket that prevents solution from taking a pathway to escape out of the necessar y gaps between the rotating and stationary components. This glycerin seal requires the rotor to be brought to speed empty, followed by injection of solution from the inlet tubing, as indicated in the center of the rotor. The solution flows through the bo dy, with sediment trapped at the outer walls of the rotor body, and the purified solution then flows up and through the outlet where it can be collected and, if desired, passed through again. The volume of material that can be centrifuged is only limited by the lifetime of the bearings and the sediment capacity of the rotor which in these tests was never appreciably filled. Tests on the Contifuge continuous flow centrifuge were carried out using material that had undergone two nitric acid refluxes in 2.6 M nitric acid for a total of 70 hours 50 hours for the first reflux and 20 hours for the second. The material then spent a total of 26 hours in crossflow filtration (20 hours following the first reflux; 5 hours following the second), flushing the materi al with a total of 73 L of Triton X buffer (55 L; 18 L). This material was then diluted to a concentration of 7.5 g/mL five times more
56 concentrated than the standard film making concentration and a film was made to assess initial material quality. As can be seen in Figure 2 18A and 2 19A, this material contains many large particulates that would make successful device fabrication using thin organic layers impossible though it should be noted that a standard 6000 RPM centrifugation was not used here to provide a more severe test for the new technique Figure 2 17. Cross section view of Contifuge Stratos conti nuous flow centrifuge. Two liters of this solution was used as the starting material and was flowed through the continuous flow centrifuge at a speed of 15 mL/min, meaning that it took 16 minutes and 40 seconds for each infinitesimal volume fraction of the material to pass through the 250 mL centrifuge rotor, or put another way, on each pass the material was centrifuged for 16 minutes and 40 seconds (with the exception of the final pass which was carried out at 7.5 mL/min). Following each pass, the materia l was sonicated for 10
57 minutes to degas the solution and debundle the nanotubes. Each day (for three days), the solution was passed through the centrifuge four times and after the fourth pass each day, the centrifuge was stopped, the sediment collected an d the rotor cleaned. Twelve passes were selected so that the total centrifugation time was on the same order of magnitude as our standard fixed angle rotation time (3 x 1 hour). To assay material purity, small films were made using a constant volume (1mL on a 17mm diameter film membrane) of the sampled material at the end of selected runs (1 st 3 rd 6 th 12 th ) and transferred to glass and silicon substrates. Initial studies on these films were carried out using optical microscop y atomic force microscop y (AFM), UV Vis spectrophotomet y and resistivity measurements via a van der Pauw four probe station. As can be seen in Figure 2 18 and Figure 2 19, a dramatic improvement is observed after a single pass through the continuous flow centrifuge with further i mprovement upon further centrifugation cycles. These images, taken using an optical microscope, are intended to provide a feel for the improvement; however with so few particles, a single image cannot convey the cleanliness of the sample as a whole. AFM images are shown in Figure 2 20; however, this is a significantly more local technique and should be viewed in conjunction with the larger area studies of Figures 2 18 and 2 19. While such assays were reassuring, a more meaningful test of the efficacy of t he purification was developed: devices were made to test the frequency of shorts using a simple SWNT / organic semiconductor / gold electrode stack. By applying a voltage between the SWNT film and the gold electrode, it could be determined whether shortin g pathways exist between the two electrodes. By varying the thickness of the organic layer, the frequency of these shorting pathways could be determined.
58 Figure 2 18. Optical microscope (5X objective) image of film made from constant volume of solution (A) before centrifugation and after (B) one pass, (C) three passes, (D) six passes, and (E) twelve passes. One of our most promising applications, and the focus of Chapters 3 and 4 of this dissertation, is the carbon nanotube enabled vertical organic field effect transistor (CN VFET). For this reason, the shorting frequency at varying thicknesses was tested using the standard dilute CN VFET film that is effectively 2 nm thick (the thickness is A Befor e Centrifugation B 1 st pass C 3 rd pass D 6 th pass E 12 th pass
59 estimated by scaling by volume a solution used fro m a thicker test sample). These films were transferred to p + silicon / SiO 2 (200 nm) / benzocyclobutane (7 nm this layer, not relevant to these measurements, is discussed further in Chapter 3) substrates with amorphous small molecule (( N,N' di(1 naphthy l) N,N' diphenyl 1,1' diphenyl 1,4' Figure 2 19. Optical microscope image (20X objective) of film made from constant volume of solution (A) before centrifugation and after (B) one pass, (C) three passes, (D) six passes, and (E) twelve passe s. A Before Centrifugation B 1 st pass C 3 rd pass D 6 th pass E 12 th pass
60 Figure 2 20. AFM images of material after (A) 3 passes through continuous flow centrifuge and (B) 12 passes through continuous flow centrifuge with height scale bar shown to right. Because AFM only investigates relatively small areas, it is limited in its ability to assess film purity over a large sample however the lack of large particulates is promising. diamine) or (NPD) ) was evaporated at thicknesses ranging from 100 nm to 500 nm 92 The use of an amorphous material is important because more crystalline organics tend to grow via an island growth mode which leads to pinholes at their grain bounderies and shorting pathways in thin layers that may not be related to the quality of the bottom electrode. NPD wa s also selected because it possesses a HOMO of ~ 5.4 eV, creating A 3 rd pass B 3 rd pass C 12 th pass D 12 th pass
61 an initial Schottky barrier of ~0.6 eV with the SWNTs which means that there should be little initial charge injection. This barrier means that any significant current is indicative of shorting. Following the NPD layer, gold top electrodes, with an area of 0.035 mm 2 were evaporated on top of the organic layer through a TEM grid shadowmask, yielding 40 50 testable pixels per device. A standard probe station probe was used to cont act the SWNT/gold bottom contact and a gold wire gently contacts the top electrode. For each pixel, approximately 1 V wa s applied between the electrodes to test whether a shorting current was h a current that measured beyond a few tens of nanoamps wa s characterized as shorted. Films that had been centrifuged two times at 6000 RPM and filtered twice through 650 nm pore membranes were used as a control sample to test the conventional purification techniques. In making these films, fabricated on our standard mixed cellulose ester membranes, every precaution was taken to ensure cleanliness in handling including deposition of gold electrodes after transferring the SWNT film to minimize underlying pa rticulates. D espite the se precautions, this material possesse d a large number of shorts that ma de the use of organic layers below 400 nm difficult and inconsistent even when a 2 nm thick SWNT film wa s used. There were no working pixels at 200nm (0 pixe ls out of 32) and even at 300 nm, less than half the pixels work (15 pixels out of 32 or 47%) Fortunately, and consistent with previous publications from the group the yields do improve appreciably when 400 nm (27 out of 32 or 84%) and 500 nm (31 out of 32 or 97%) thick organic layers are used Material fabricated from the continuous flow centrifuged material performed considerably better. Table 2 1 shows the working pixel yields for films following the 6 th
62 and 12 th pass of the continuous flow centrifu ge on standard mixed cellulose ester (MCE) membranes as well as a polycarbonate membrane ( with ion etched pores) to test whether membrane defects were still playing a role in the presence of shorting pathways. While the yields were quite good (70 75% succ ess rate at 200 nm vs 0 % previously), a few anomalous results stood out. Namely, the similar yields with 200 nm and 300 nm NPD layers in the 12 th pass material on an MCE membrane, as well as the seemingly low yield using a 300 nm NPD layer with this mate rial. This suggested that perhaps the particulates were relatively large ones that had fallen on the substrate, an effect that might manifest itself as increased shorting across a range of thicknesses, similar to what was see n here. Table 2 1. Characteriz ation of SWNT material following continuous flow centrifugation with predeposited electrodes NP D Thickness (nm) 6 th pass MCE membrane 12 th pass MCE membrane 12 th pass Isopore Polycarbonate membrane Good Short Yield Good Short Yield Good Short Yield 200 25 38 40% 34 15 70 % 39 13 75% 300 64 9 88% 46 15 75% 56 5 92% 400 62 0 100% There was at least one step in which the substrate s were exposed to a non cleanroom environment. These samples had been fabricated following our standard operating procedures which was to predeposit the Cr / Au SWNT contacts. This exposes the substrates to the potentially dirty glovebox before the conductive nanotube film is adhered, allowing for particulates to land on the substrate and force the SWNT film upwards to create shorting pathways. To avoid this potential issue, the experiment was repeated with Au electrodes deposited after transfer of the SWNT film so that any particles would be on top of, rather than underneath, the conductive SWNT film. This
63 was don e with both a dilute, VFET density (~2 nm effectiv e thickness) SWNT film and a thicker 20 nm thick SWNT film. Between 32 and 70 pixels were measured on devices for NPD layers ranging from 100 250 nm thick and as can be seen in Table 2 2 we really have achieved remarkable levels of purity relative to our previous standards. For NPD layers of 250 nm, we did not observe a single shorted pixel in both the dilute film or in the 20 nm thick film, containing 10 times the quantity of material (and thus 10 tim es the number of intrinsic particulat e s per unit area) At 200 nm, where previously no working pixels were observed, we now achieve d a >95% success rate in both dilu t e and 20 nm thick films. The yields d id drop with thinner active layers, indicating room for improvement, but even at 100 nm thick organic layers the yield was still 35% in the VFET film allowing us to build VFETs across a previously inaccessible range of channel thicknesses to probe the physics at work these thinner layers as discussed in Chapter 4. Table 2 2. Characterization of SWNT material following continuous flow centrifugation with postdeposited electrodes NP D Thickness (nm) 2 nm thick SWNT Film 20 nm Thick SWNT Film Good Short Yield Good Short Yield 100 24 43 35 % 0 50 0 % 150 38 8 83 % 19 15 56% 200 65 3 96 % 41 2 95% 250 46 0 97% 33 0 100% Further characterization of the material via AFM, TEM and optical microscopy coupled with the high pixel yields of the highly centrifuged material demonstrated that we were eliminating many of the bucky onions and other contaminants from the SWNT material but it did not yield much information on SWNT material losse s. To study this, films were made both to determine material losses and to measure the conductivity of
64 the SWNT films as the mate rial underwent increasing centrifugation (Table 2 3 ). Films were made on a 17 mm diameter mixed cellulose ester membrane using a controlled amount of solution (collected after the number of centrifugation passes indicated). Four probe van der Pauw measure ments were made on these films to measure sheet resistance followed by optical characterization via UV Vis transmission spectra and thickness measurements were taken by atomic force microscope (AFM) in which a reliable stepheight was achieved by protecting the nanotube film with photoresist while the surrounding film was removed by ashing in an oxygen barrel asher (followed by dissolution of the photoresist in acetone). This allows us to study material loss and the conductivity of the film as defined by = 1/(R s The results of these studies are shown in Table 2 3 and Figure 2 21. Material loss is significant, exceeding 50% by the 6 th pass through the centrifuge. In early centrifugations, this material loss does not seem to have a significant impact o n conductivity. Indeed, after the first centrifugation we even see a s light increase in conductivity This increase is perhaps not surprising as it can be observed from Figure 2 18 that this first centrifugation pass removed many impurities that may hav e impeded the conduction pathways between SWNTs. Table 2 3 Characterization of SWNT material following continuous flow centrifugation Centrifugation Passes Volume of Solution (mL) Thickness ( nm) Loss (%) Sheet Resistance ) Conductivity (S/cm) Before 1 43 n/a 70 3322 1 st 1 30 30% 97 3436 3 rd 1.8 48 38% 65 3205 6 th 2.5 48 55% 69 3019 12 th 2.8 52 57% 85 2262
65 I n further centrifugations, as the beneficial effects of removing particulates are redu ced and material loss is coming almost entirely from the elimination of nanotubes, the conductivity does begin to suffer. This effect is somewhat surprising; as it implies a preferential loss of long SWNTs or bundles which contribute the most to the condu ctivity. W hile material losses were expected, we anticipated that thes e losses would be distributed evenly and not imp act the conductivity of the bul k film. However, i t has been reported that bundles of long SWNTs could have an increased density relative to individualized nanotubes if the surfactant molecules are only coating the outer tubes in the bundle, having a less buoyant effect on their density leading to their preferential removal 37 There are applicati ons where this preferential bundle removal might be advantageous, most notably the carbon nanotube enabled vertical field effect transistor (Chapter 3 4) where bundled nanotubes can screen the gate field and limit device performance. The conductivity decre ase should not be overstated and it is unlikely that this will be a limiting factor. After six centrifugation passes, the conductivity decrease was less than 10% and analysis of this material shows a high quality film, free of particulates. The improved pixel yield (with an associated performance enhancement that will be detailed in Chapter 4) allows for studies and large area devices that otherwise would be difficult to achieve and as such, the slight loss in conductivity is more than acceptable. While there certainly is room for improvement in the process, the continuous flow centrifuge can serve as a viable technique to achieve highly purified SWNT material on an industrial scale for organic electronic devices.
66 Figure 2 21. UV Vis spectra for films made from a constant volume 1mL solution on a 17mm diameter film membrane.
67 CHAPTER 3 S C HOTTKY BARRIER MODUL ATION IN VERTICAL FI ELD EFFECT TRANSISTO RS ENABLED BY LOW DENSI TY OF STATES ELECTRO DES Overview The low density of electronic states (DOS) found in carbon nanotubes and graphene near its Dirac point allows for modulation of the Fermi level of these materials in a manner not accessible using conventional metals that possess a high DOS. Recently, such Fermi level tuning was exploited in a novel dev ice architecture: the carbon nanotube enabled vertical field effect transistor (CN VFET), which demonstrated state of the art current densities at low operating voltages from comparatively low mobility organic semiconductors 40 Unlike in conventional thin film transistors (TFTs), transconductance arises from a gate field modulation of the contact barrier at the organic semiconductor/nanotube interface an effect that will be discussed in this chapter. Figure 3 1A shows a conventional organic TFT. A thin organic channel layer is deposited on top of a gate and gate dielectric with patterned source and drain electrodes defining the channel length. With the source at ground, a potential is applied between the so urce and gate electrodes inducing charge carriers in a very thin layer within the organic semiconductor, adjacent to the dielectric surface. As a source drain voltage is applied, these charges move through the organic channel layer giving rise to the devic e current. T he low mobility of organic semiconductors, typically on the order of 10 1 10 3 cm 2 limits the on currents that can be achieved by these conventional lateral channel TFTs. This can be overcome by the use of long channel widths or short channel lengths but the former makes for large devices, reducing the packing density
68 and increas ing parasitic capacitances, while the latter requires high resolution patterning to define the narrow channel, which becomes cost prohibitive. Unfortunately despite advancements in organic TFTs, the best published on currents in these devices remain about two orders of magnitude below what can be achieved using alternative competitive technologies such as polycrystalline silicon TFTs 93, 94 The low mobility of organics can be overcome through the novel architecture of the CN VFET which offers facile control over the channel length. As shown in Figure 3 1B, the device again is built upon a gate and gate dielectric but here the source, channel and drain are all stacked vertically. The s ource electrode used in the CN VFET is a dilute (but random) nanotube network that is porous but well above the p ercolation threshold (Figure 3 1C) For those more familiar with nanotube transistors in which the nanotube itself is the channel 27, 95, 96 it may be important to emphasize that the dilute nanotube network here serves only as the source electrode so that no separation of metallic and semiconducting nanotubes is req uired. The channel length in the CN VFET becomes simply thickness of the organic material which can be readily controlled by thermal evaporation or spin coating, eliminating the need for lithographic patterning to achieve short channel lengths. Figure 3 1. Schematic of A) conventional, lateral channel thin film transistor and B) vertical field effect transistor. C) AFM image of dilute SWNT source electrode.
69 In 2008, Liu et al. published the first CN VFET that used a dilute nanotube source electrode 30 Two years later, with an improved understanding of the essential device parameters, McCarthy et al. reported a high performance CN VFET that achieved on currents that were nearly four times better than those reported in conventional TFTs using the same channel material 40, 93 In 2011, McCarthy et al. integrat ed the optically transparent drive transistor as a VFET in a vertical stack with the OLED pixel, to demonstrate a vertical organic light emitting transistor emitting light across the full aperture with minimal performance degradation in power consumption, luminance or efficiency 41 The CN VFET has dem onstrated high levels of performance and in this chapter we discuss the underlying device physics. Because the device operates as a Schottky barrier transistor, I will first discuss energy band alignment in the Schottky Mott model and how low density of s tate metals (such as carbon nanotubes and graphene) enable a new device mechanism for barrier height modulation and charge injection. As further studied in my work, the device performance also benefits from a barrier width modulation enhanced by the direct gate field access to the low DOS metal/organic semiconductor interface. This is shown for nanotube source electrode based devices in which I changed the nanotube surface density, thereby modifying the screening of the gate field from the relevant interfac e. I then further elucidated the phenomenon in the first graphene source electrode based VFETs (G VFETs), in which both continuous graphene and graphene with a variable density of micron scale holes had been created by a novel process.
70 Energy Band Align ment at a Metal Semiconductor Interface Schottky Barrier Height Modulation To first approximation, the work function difference between a metal and a semiconductor dictates the transport across their junction, determining whether the contact will be Ohmic or have a Schottky barrier to electrical transport 97 99 When two such materials are brought into contact, charge flows between them allowing the Fermi levels to align as the sy stem reaches thermal equilibrium. For a conventional metal, possessing a high density of electronic states, the carrier concentration is so large that the Fermi level shift occurs almost entirely in the semiconductor (possessing a comparatively low carrie r concentration). In the Schottky Mott model, this shift creates a Schottky barrier that is equal to the difference between the work function of the metal and the electron affinity of the semiconductor. The large DOS in the metallic electrode serves as a reservoir, requiring the addition/subtraction of large amounts of charge to induce appreciable shifts in its work functions (like the water level in a large lake, much water must be added to change the level perceptibly). This picture changes dramatica lly for low DOS metals like carbon nanotubes and graphene for which charge addition/subtraction induces much more dramatic work function, or equivalently, Fermi level shifts (little water must be added to a tall narrow glass to change the level accordingly ). Since the Fermi level in these materials can be changed in response to gating fields, these low DOS carbon based metals, placed in contact with a semiconductor, admit a new mechanism for current modulation by the gate field control of their trans juncti on transport through tuning of the Schottky barrier height. This additional means of control allows for new high performance device architectures.
71 Experimental evidence for this mode of barrier height modulation in low DOS materials was first reported by L ongeran in 1997 in a tunable diode featuring a hybrid inorganic / organic structure ( n indium phosphide / poly (pyrrole )) 100 Longeran was able to demonstrate electrochemical manipulation of the work function of th e poly(pyrrole) polymer resulting in tuning of the turn on voltage of this device by more than 0.6 V. In 2004, Wu et al. demonstrated Fermi level control in a SWNT film electrode where the application of a gate field led to the depletion of the S1 and M1 van Hove singularities, as measured by optical spectrophotometry, providing strong evidence for a Fermi level shift of more than 0.7 eV 38 More recently, Wadhwa et al. succeeded in exploiting the low DOS of SWNTs t o actively modulate their Fermi level in a gated SWNT / Si Schottky junction solar cell 29 Schottky Barrier Width Modulation It is important to note that an applied gate field can modulate transport across a metal / semiconductor junction even in high DOS metals through band bending, which modulates the Schottky barrier width 99, 101 104 The application of an appropriate field causes the valence band (highest occupied molecular orbital, or HOMO, in an organic semiconductor) to bend towards its Fermi level, causing the Schottky barrier formed with the metal to thin. The thinning of this barri er in turn enhances the tunneling current across the barrier. Of course, the opposite is also true, by reversing the bias the applied field will induce a widening of the barrier and a reduction of the tunneling current thus allowing for modulation of the d evice current independent of the barrier height modulation discussed previously. Such a mechanism was exploited by Yang and co workers in related VFETs that used a partly oxidized aluminum source electrode and required a very high capacitance (supercapa citor) gate dielectric to achieve the reported
72 performance and others have since extended this work 101 106 The CN VFETs and G VFETs (with holes) use both mechanisms and thus dramatically outperform metal source electrode based VFET devices. Schottky Barrier Modulation in the CN VFET The carbon nanotube enabled vertical field effect transistor (CN VF ET) acts as a Schottky barrier transistor transconductance arises via gate field modulation of the Schottky barrier height and width. When there is no applied field, the device sits at thermal equilibrium with the Fermi level of the nanotubes aligned wi th that of the organic semiconductor. To turn on a CN VFET with a p type organic channel material, the source electrode is held at ground as a negative voltage is applied to the gate, shifting the Fermi level of the SWNT film down and diminishing the heig ht of the barrier at the SWNT / organic semiconductor interface. Simultaneous ly, the porous source electrode allows penetration of the gate field to access the nanotube/organic semiconductor interface, bending the highest occupied molecular orb ital (HOMO) towards its Fermi level and thinning the Schottky barrier. When a source drain voltage is applied concurrently, the lowered and thinned barrier allows for high on currents across the organic channel. Conversely, when a positive gate voltage i s applied, the barrier is enhanced and the device is turned off. The energy band alignment and Schottky barrier modulation was previously modeled at the interface of a single, metallic SWNT and a semiconducting polymer channel and the resulting simulation demonstrating the aforementioned barrier height and width modulat ion, is reproduced in Figure 3 2 30 It is this combination of both barrier height and width modulation that gives the CN VFET its dramatic performance ad vantage over metal source electrode VFETs 101 105 The high DOS of the metals used in those devices precludes barrier height modulation
73 of the Fermi energy of the metal electrode so that all current modulation arises via gate field modulation of the barrier width. Figure 3 2 Electrostatic simulation of energy bands at the interface of a single SWNT and the organic channel l ayer on gate / gate dielectric (inset) under increasing gate voltages. As the gate voltage is increased, the SWNT Fermi energy can be seen to be decreasing in addition to the thinning of the Schottky barrier. Reprinted with permission from Liu et al 30 Fabrication of Vertical Field Effect Transistors Though the materials used for the CN VFETs and G VFETs discussed in Chapters 3 and 4 of this dissertation were va r ied to match the desired experiments, the procedure to fabricate these devices wa s similar for all and will be briefly described here A labeled photograph of the structure for visual reference follows at the end of this section (Figure 3 3) Changes to these procedures for individual applications will be noted in the section that the ir implementation is discussed VFETs were fabricated using p + doped silicon as the gate electrode with a 200 nm SiO 2
74 use of a protective photoresist layer during dicin g to eliminate adhesion of Si chips to the surface. The use of this relatively thick gate dielectric necessitates large gate fields, with gate voltage ranges from 40 V to +40 V, but it has been demonstrated in earlier work that the use of a thinner and/o r higher k dielectric can minimize the gate voltage range required 40, 41 In some devices, such as the solution processable VFETs in Chapter 4, t his is done using an atomic layer depos ition (ALD) grown Al 2 O 3 dielectric over a metal gate deposited on glass ; however, for cost reasons the rest of the VFETs demonstrated here will be fabricated on the Si/SiO 2 substrates. A thin, hydrophobizing benzocyclobutane (BCB) layer 107, 108 has been shown to minimize charge trapping in organic FETs and has reduced hysteresis in previous CN VFETs 41 and is incorporated in all devices described here. B CB (Dow Chemical Co., Cyclotene 3022 35), diluted to 1 part in 50 in trimethylbenzene is spuncast on the Si/SiO2 at 4000 RPM, resulting in a 7 8 nm thick layer which is hard baked at 225 C to cross link the material and render it resistant to solvents. Go ld source contacts (40 nm) with a chromium layer (10 nm) to promote adhesion to the BCB are thermally evaporated onto the substrate through a shadow mask. SWNT films are fabricated on MCE membranes that are then cut using razor blades into 2 mm strips (tw o strips per substrate) and transferred to the substrates by wetting with isopropanol, using the evaporation of the isopropanol to pull the film into more intimate contact with the surface to generate strong adhesion via van der Waals forces. The MCE memb ranes are then dissolved away with an acetone vapor bath followed by 4 acetone baths with a final isopropanol bath. Samples are transferred to the glovebox and baked at 225 C to dedope the p doped SWNTs (doped during the nitric acid reflux) which shifts t heir Fermi
75 level to approximately 4.7 eV and generates a larger initial Schottky barrier with most p type organics. 41 The organic semiconductors are then deposited over the entire nanotube strip, by thermal evapor ation unless otherwise noted. Gold drain electrodes are evaporated through a transmission electron microscope (TEM) grid with 100 m hexagonal grid spacings to define a VFET area that is 0.035 mm 2 with over 160 VFET VFET pixels each) per device. Gate contact is made by scratching through the dielectric with a scribe and making contact with an indium dot. This indium dot and the gold source contact are contacted by needle probes and the gold drain electrode is gently contacted via a gold wire. Electrical measurement is done on a custom built probe station using a two channel Keithley 2612A System Sourcemeter controlled by a program written in LabVIEW (both probe station and LabVIEW program were designed by collaborators in the Rinzler Laboratory). Though a current is directly measured, this disse rtation will give values as a current density because, unlike conventional lateral channel TFTs where the current scales linearly with the channel width, the current in CN VFETs scales with the total channel area. Areal current density represents a ratio nal and important figure of merit to for insight into device performance as in practical applications, such as active matrix organic light emitting diode displays, the pixel area is limited by the real estate occupied by the driving transistor, impacting d evice properties such as pixel lifetime and power consumption.
76 Figure 3 3. Labelled layo u t of CN VFET on silicon substrate. The dashed line indicated the approximate outline of the SWNT source electrode as the film as the very dilute film is diff icult to see (though the reorientation of the pentacene org an ic la yer us e d here can be observed o n the strip to the right). Effect of SWNT Film Porosity on VFET Performance In the VFET architecture, the use of a porous source electrode allows the gate field to penetr ate and access the contact barrier at the source / organic interface. In CN VFETs, this gate field penetration is permitted by the low areal surface density of the SWNTs on the substrate which can be readily controlled during fabrication of the SWNT film by varying the concentration and/or volume of the SWNT suspension used. The nanotubes are typically bundled and screen the gate field, severely limiting Schottky barrier modulation at increasing SWNT densities 109 and man dating the use of dilute SWNT networks. By lowering the SWNT density, we anticipate seeing an
77 improvement in the device current on off ratio; however, as the film conductivity is inversely proportional to film density, the increased series resistance may negate the improved gate field penetration and limit on currents when a dilute network is used. To further elucidate the role that the porosity in the SWNT thin film plays, as well as to fabricate an optimized SWNT film for future device studies, I studi ed CN VFET performance as a function of SWNT density. SWNT films were made at four different characteristic densities / thickness (Figure 3 4) by varying the volume of a standard, dilute SWNT solution used to fabricate the films via vacuum filtration. It was found, via atomic force microscopy (AFM), that 120mL of this solution made a film that was approximately 40 nm thick. Films were then made using 4.8 mL, 7.2 mL, 9.6 mL, and 12 mL of this standard solution for SWNT films that had an effective thickness of 1.6 nm, 2.4 nm, 3.2 nm and 4.0 nm respectively. It should dilute to cover the entire substrate and vary locally from 0 nm to over 10 nm. Nevertheless, an equivalent properties and as such it will be used here. Devices were fabricated as shown in Figure 3 5. Two probe resistance measurements were taken and the sheet resistance of each film was estimated based on the sample geometry (displayed in Figure 3 4) prior to baking and dedoping the film. 460 nm of dinaphtho [2,3 b f ]thieno[3,2 b ] t hiophene ( DNTT ) doped with MoO 3 was grown on the SWNT film to serve as the channel layer. DNTT is a recently develop ed, air stable small molecule that is chosen for its relatively high mobility (2.9 cm 2 /V s) and deep HOMO level of 5.4 eV which creates an initial energy barrier of ~0.7 eV 110
78 Figure 3 4. AFM scan s of increasing film densities used to probe effect of SWNT film porosity with estimated sheet resistance measurement. The same S W N T mater i al wa s used f o r all f ilms. A ll SWNT films are limi ted by imp edance a t tub e t ube junc t ions ; however, as th e number o f long c ond u c tive p athwa ys a cross t h e s u b strate is dimi nis he d, t his im ped ena nc e begi n s to dra s t ical l y limit conduc t ivit y in ver y dilut e films.
79 Figure 3 5. Structure of DNTT CN VFET on silicon. For each film density, a minimum of s even pixels were measured with representative data plotted in Figure 3 6 At high SWNT densities, screening limits device performance and current modulation in both the on and off state current. As the SWNT film density is decreased, the on state current initially increases as the gate field is allowed to penetrate the porous SWNT film. This on current is maximized at the 2.4 nm equivalent film, after which it becomes limited by the high resistivity of the nanotube film due to the limited conductive path ways across the device indicated by the linear on current even at low source drain voltages. Though the on current becomes limited, the on off ratio is maximized due to the enhanced gating and the lower off state in this dilute network. As the order of magnitude gain in on off ratio is more important for potential device applications than the fractional reduction in on current, we chose to use the more dilute film to explore to explore future CN VFET devices. These results are indicative of the balance that must be struck in VFET source electrodes. As source electrodes become more porous, the enhanced gate field penetration leads to improved device performance; however, past a certain threshold, we begin to pay a high cost with regards to series resista nce. More conductive
80 electrodes, such as graphene, should enable high levels of performance by avoiding the resistivity and screening effects found in dilute networks. Figure 3 6. C N VFET s with a 5 00 DNTT channel usin g SWNT films of vary i ng thic k n esse s A) Transfer curves, B) output curves on a linear scale, C) current on/off ratio, and D) output curves on a logarithmic scale for films of varying densities. Devices with thin SWNT films turn on faster and are more fully off but may become limited by the high restance of the SWNT film. Graphene Based Vertical Field Effect Transistors The use of a graphene electrode in vertical field effect transistors should offer advantages over thin films of carbon nanotubes that may afford not only improved performance bu t also a solid test bed in which to study the physics that occur at the A B C D
81 source channel interface. Similar to carbon nanotubes, graphene has a low density of states near its Dirac point that should lend itself well to modulation by an external gate field. The monolithic nature of graphene also offers a natural advantage in sheet resistance relative to dilute SWNT networks where impedance at tube tube junctions dominates performance, particularly as the films are made more dilute. The planar structure of g raphene also can afford many other advantages by allowing thinner channel layers (and proportional increases in on currents) and enhanced reorientation of the small molecule channel layers. It has been demonstrated that some organic molecules will prefere stack on the pristine sidewall of a SWNT, reorienting it such that the high mobility direction is vertical, facilitating high performance CN VFETs 111 As the anisotropy with molecular orientation c an lead to orders of magnitude differences in the mobility of an organic material 108 this reorientation can yield essential improvements in device performance. Graphene is structurally similar to the nanotube sid ewall and should induce similar reorientation, but where the bundling, curvature and high porosity of nanotube networks contributed to mixed crystalline phases, the planar graphene should lead to a more uniform structure in the channel layer. Graphene also offers a unique window into the device physics at play in the VFET architecture. As has been discussed, both Schottky barrier height and width modulation contribute to current modulation in a VFET. The nanotube based devices operate in a mixed fashion, taking advantage of both mo des. As depicted in Figure 3 7B b the continuous electrode provided by graphene can probe (principally) the effect of the barrier height modulation by screening the gate field and preventing it from inducing barrier width modulat ion at graphene / organic interface. By then purposely i n t r oduci n g
82 holes in the continuous graphene electrode to allow gate field access(Figure 3 7Bc), it becomes possible to probe both modes in a single material system. The G VFET device architecture is sho wn in Figure 3 7A (discussed below). The thin channel layer between the graphene source and top drain electrodes imposes rather severe requirements on the quality of the graphene layer, at least in terms of minimizing vertical protrusions. The conventional polymethylmethacrylate (PMMA) transfer method 10 for copper based chemical vapor deposition (CVD) grown graphene was found to be unreliable, with numerous tears and wrinkles causing frequent shorting pathways 15, 112 Techniques described in the literature to improve the polymethylmethacrylate ( PMMA Microchem 11% in anisole) transfer were tried, however, these did not fully eliminate damage induced by the swelling o f the PMMA during acetone dissolution. 15, 112 A modified PMMA transfer method was developed that gave much better results. Transfer of CVD grown graphene from the copper growth substrate was improved by depositing a thin layer ( 100 nm) of Au as a protective layer before spinning the PMMA support film ensuring a post transfer surface that is free of difficult to remove polymeric residue. The thin metallic layer avoids strain induced by the swelling o f the PMMA film during the more conventional transfer process. This is especially important at domain boundaries where chemical bonding between the polymeric chains and the graphene is favorable. 113 Figure 3 7C illustrates the procedure for trans ferring and patterning the graphene using an Au thin film as a protective layer and etch mask. Gold was thermally evaporated at a thickness ranging from 20 to 100 nm through a rectangular shadow mask onto the graphene grown on polished copper foils by Max Lemaitre at the
83 FIGURE 3 7 (A) G VFET architecture and drive scheme. (B) Energy level diagram for a GVFET at the graphene semiconducting channel interface for constant drain voltage and three distinct gate voltages. The black line depicts the hole injection barrier and depletion layer in the semiconductor for a continuous graphene electrode. The red dashed line depicts the same features for the case of a graphene electrode perforated with holes. For the continuous graphene case current modulation is due principally to barrier height l owering (thermionic emission). For the perforated graphene case the barrier also thins (enhancing tunneling). (a) Initial Schottky barrier and band bending induced by the offset of the graphene work function with the HOMO level of the organic semiconductor (b) Moderately reduced Schottky barrier height resulting from the shift in the graphene work function. Band bending is less pronounced for the continuous graphene. (c) At high gate voltage the barrier height is significantly reduced and the depletion width thinned for the case of perforated graphene. (C) Schematic of the graphene source electrode fabrication process using a protective evaporated Au layer. (i) the Au film evaporated onto the as grown graphene on Cu, (ii) the PMMA spin coated onto the go ld layer, (iii) Cu etched away, (iv) graphene/Au/PMMA stack adhered to SiO 2 substrate, (v) PMMA etched away in O 2 plasma, (vi) Au film etched away leaving behind a residue free, perforated graphene sheet.
84 Nanoscale Research Facility followed by spin coating the PMMA and baking After copper dissolution in a perchloric solution (Transene, APD 100) the Au coated section of the graphene/Au/PMMA sandwich was adhered, graphene side down, to the p + Si/SiO 2 /BCB substrate through pressure applied by handclamps for four hours in an 80 C oven. After baking in a tube furnace for one hour, t he Au served as an etch mask while the PMMA and excess graphene around the gold mask were dry etched in an O 2 p lasma (1 hour, 600sccm, 600 W) thus defining the edge of the graphene source electrode. An iodide based gold etchant (Transene, Au TFA Etchant) subsequently removed the mask layer. Finally, a gold source contact was evaporated along one edge of the graphe ne layer completing the source electrode. Note than unlike in the CN VFET, here the gold source contacts are evaporated after the graphene source electrode was transferred as the etchant that removes the graphene mask would also attack these electrodes. M icron scale holes with a crudely controlled density were produced in the graphene by varying the thickness of the Au mask layer. Thin Au layers possess sub micron pinholes, with a through hole density that depends on the layer thickness. During the dry et ch of the PMMA, reactive oxygen radicals penetrate these holes to etch the graphene and underlying BCB, leaving behind circular holes in the graphene with an average diameter of 2 3 microns The diameters of these holes are self limiting due to the increas ing diffusion path length for counter propagating oxygen and reaction products in the confined space between the Au and the SiO 2 as the etched region grows. The measured areal hole densities in the graphene used to build the G VFETs
85 discussed below were 0 2, 13 and 20% with average hole diameters of 2.2 0.6m, 2.3 0.6m, 2.5 1.0m, respectively. Figure 3 8 compares the quality of the graphene transferred with and without the use of a 100 nm (pinhole free) protective Au layer. Raman spectroscopy provides a comparative measure of graphitic materials, capable of distinguishing single layer graphene from multi layer graphene and graphite 18 and characterizing disorder, crystalline grain size, stacking symmetr y, and doping of the graphene films 114 116 The D to G band intensity ratios are shown versus the FWHM of the 2D band overtone for 100 distinct points for the graphene film s transferred with and without the use of the gold FIGURE 3 8. (A) Raman spectral data in the form of a cluster plot of D/G peak ratios versus the 2D band FWHMs for a graphene layer transferred using the Au protected process (black squares) and a graphen e layer transferred using the conventional PMMA process (red triangles). One hundered point were recorded on each layer in a square array of points having a pitch of approximately 50 m A smaller D/G ratio and FWHM are desirable, as seen for the majority of points recorded for the Au transferred layer. (B) SEM (scale bars: 2um) and (C) AFM images (15 x 15 m) of graphene layers transferred to SiO 2 using the Au protected and the conventional PMMA processes, as indicated by the labels. protective layer. The D band was below the noise threshold, and the 2D FWHM was substantially reduced in the majority of measured spots for the Au protected films. Scanning electron (SEM) and atomic force micrographs (AFM) of films transferred by
86 the two methods are shown in Figures 3 8B & 3 8C, respectively. The films transferred using the gold protective layer are continuous, without the polymer residue, micro tears, and wrinkles characteristic of the standard PMMA transfer. The organic semiconductor channel layer e vaporated onto the graphene was dinaphtho [2,3 b f ]thieno[3,2 b ] t hiophene ( DNTT ) 110 The flatness of a single layer of graphene should in principle permit even sub 100 nm channel layer thickness (with corresponding performance enhancement) without incurri ng electrical shorts to the top drain electrode. We found however that device yields suffered when the DNTT thickness was below 250 nm. This may be a consequence of the low surface energy of graphene and crystallinity of DNTT that results in island growth incorporating pinholes and shorting paths to the subsequently deposited Au drain electrode, for thin channel layers. To ensure effectively 100% yields and to permit a direct performance comparison against comparable channel thickness CN VFETs, a DNTT chann el thickness of 500 nm was used. G VFET devices were tested with the graphene source electrode contact held at ground potential, while the drain and gate were biased relative to ground. Figure 3 9 shows typical output curves for the G VFETs with graphene source electrode areal hole densities of 0, 2, 13 and 20% (Figure 3 9A). Both the on (V G = 40 V) and off (V G =+40V) states are shown. The advantage of the short channel length in the vertical architecture is seen in the high on current densities at low drai n voltages ( | 5| V). The on current densities clearly scale with the density of holes in the graphene source electrode. Figure 3 9C plots the on/off current ratio of the devices as a function of the on current density (as the drain voltages are swept from 0 to 5 V). The 20% areal hole density electrode
87 yields on/off ratios exceeding 10 6 Attempts to get higher hole densities (>20%) by making the protective Au layer thinner resulted in discontinuous graphene sheets. Ordered hole arrays would avoid this pr oblem and provide a path for further device optimization. A summary of the device characteristics versus areal hole density is plotted in Figure 3 9D. FIGURE 3 9 (A) 50 x 50 m SEM images of the transferred graphene films having the indicated hole densit ies. (B) G VFET output curves for the off state (V G =+40V) and the on state (V G = 40V) for the se graphene source electrode hole densities. (C) On/Off current ratio versus on state current density up to a drain voltage of 5V for each hole density. (D) On current densities and On/Off current ratios for drain voltage up to 10 V versus source electrode hole density. 3 orders of magnitude current modulation is achieved in the pore free g VFET with an additional 3 orders of magnitude improvement u pon introduction of pores. The current modulation seen to occur in the continuous graphene electrode (over three orders of magnitude) provides strong support for the ant icipated Schottky barrier height modulation (changing principally the thermionic emission) in the low DOS metal. Introducing 20% holes into the graphene source electrode yields a further 2 3 decades
88 of transconductance, resulting from tunneling through the barrier. Extrapolating from the Kevin probe measurements of Yu et al. 117 we estimate that our gate sweep results in a 0.4 0.5eV shift of the graphene work function and a commensurate modulation of the barrier hei ght. Since tunneling currents also depend strongly on barrier height (e.g. within the WKB approximation ), such modulation can explain the dramatic performance advantage these low DOS metals have over conventional metal source e lectrode devices. Comparison between the G VFET and CN VFET reveals differences that can be attributed to the morphological differences between the respective source electrodes. Figure 3 10A compares transfer curves for a G VFET with a 20% areal hole den sity graphene source electrode and a typical CN VFET fabricated on the same p + Si/SiO 2 /BCB gate electrode/gate dielectric substrates. The drain voltages are adjusted to yield comparable on currents at a gate voltage of 40 V. The large hysteresis seen in t he nanotube based device has been explained by ambipolar charge traps in the BCB having a well defined critical field for charge exchange with the electrode 118 This hysteresis can be minimized by restricting the gate vol tage range but here it is interesting to observe the significantly smaller hysteresis for the graphene source electrode over the same large voltage range. This is likely due to the enhanced field concentration around the nanometer width nanotube electrodes versus the semi planar graphene electrode in the vicinity of a hole. Output curves for the two devices are plotted in Figure 3 10B for gate voltages of 40V and drain voltages out to 10V. Compared to the nanotube device the graphene devices exhibits a d rain voltage delay of ~500 mV before current begins to flow. This is
89 likely due to the work function difference between the graphene ( 4.6 eV) and the DNTT ( 5.4 eV) generating a larger initial barrier than that between the DNTT and the nanotubes. The nitr ic acid purification of the nanotubes charge transfer dopes them, placing their work function around 4.9 eV, after which a heating step dedopes them to an estimated 4.7 to 4.8 eV. The field concentration around the nanotubes may also give them an advant age in terms of this lower turn on voltage. As the drain voltage continues to grow, the nanotube device off current begins to suffer as the drain field concentration around the nanotubes begins to extract charge despite the off state (+40 V) gate voltage. The planar graphene does not exhibit such degradation in the off state. The graphene electrode also excels in the on state at high drain voltage. At V D = 10V the CN VFET on current density is ~300mA/cm 2 compared to an astounding ~1200mA/cm 2 for the grap hene device. We attribute this to the lower impedance of the monolithic graphene layer versus the dilute nanotube electrode. Figure 3 10C compares the on/off ratios for the two devices as a function of their on currents (V G = 40V) as the drain voltage is swept from 0 to 10 V. The impedance limited on current and increasing off current degrades the on/off ratio of the nanotubes device while that of the graphene device remains above 10 6 out to 10 V.
90 FIGURE 3 10 Comparison of graphene and carbon n anotube enabled VFETs with all other device layers the same (A) Transfer curves for a CN VFET and the G VFET with a 20% area l hole density electrode. (B) Output curves for both devices in the on (V G = 40 V) and off (V G = +40 V) states up to V D = 10V (C) On/Off ratios versus on current density for drain voltages to 10V Morphological differences in the two de vices allow for faster tu rn on by the CN VF ET but higher on currents in the G V FET due to the lower sheet resistance.
91 CHAPTER 4 DEVICE PHSICS OF VER TICAL FIELD EFFECT T RANSISTORS WITH VARY ING CHANNEL PROPERTIES Overview In Chapter 3 I introduced the structure of the vertical field effect tr ansistor (VFET) and discussed its operation as a Schottky barrier transistor using porous, low density of state materials. In this chapter, I detail investigations into the role of the organic channel using highly purified SWNT material to achieve the thi nnest channels yet in a CN VFET. Additionally, I have achieved state of the art performance in a low voltage, P3HT transistor by demonstrating the first solution processable CN VFET which could open pathways to manufactur e via inexpensive inkjet printing of the devices. Finally, I used C 60 to create an n type VFET, with potential applications in CMOS like inverters, and test our understanding of how variation of device parameters, such as the work function of the drain electrode and of the SWNT film, can alter the energy band alignment and impact device performance. Role of Channel Layer Thickness in Vertical Field Effect Transistors Though the VFET architecture allows facile fabrication of channel lengths down to the nanometer scale, particulates in the SWNT source electrode have traditionally led to poor device yields below 400 nm thick channels and effectively zero working VFETs sub 250 nm (see Chapter 2) While the incorporation of 500 nm channel lengths has enabled state of the art organic FET performance 40 even thinner channel layers should offer further enhancements in the device on currents. With the use of the continuo us flow centrifuge purification technique discussed in Chapter 2, high purity nanotube fil ms were available that enable workin g VFETs with channel layers as thin as 100 nm. The
92 high quality of this material offers the opportunity to study the effect of channel thickness in a range that was previously inaccessible. The selection of an appr opriate organic small molecule channel material is important as the channel is thinned. Many of the crystalline organic materials possessing (relatively) high mobilities have been shown to reorient on the nanotube sidewall, leading to a high surface rough ness and non uniform channel thicknesses 111 For very thin channel layers, the low surface energy surface offered by the SWNT sidewall can also yield island growth that does not create a uniform, pinhole fre e film until relatively thick films are grown. For this experiment, N,N' di(1 naphthyl) N,N' diphenyl 1,1' diphenyl 1,4' diamine) (NPD) was used as for the channel layer (Figure 4 1). NPD is a small molecule organic material with a HOMO of 5.4 eV 92 that is often used as a hole transport layer. As an amorphous material, NPD should create a relatively uniform layer on the BCB / SWNT surface. The NPD CN VFET device structure is shown in Figure 4 1B. Dilute SWNT film s were fabricated using the continuous flow centrifuged material discussed in Chapter 2 (12 passes through the continuous flow centrifuge); however to achieve the very sparse / porous films required here, one part of the SWNT solution was diluted in ten pa rts deionized water allowing for more uniform films yet very thin films. Though the initial material was in a 1% Triton X solution, deionized water (filtered through a 50nm pore, hollow fiber filter) was used for the dilution to avoid the introduction of new contaminants that might come with a fresh Triton X solution that had not been centrifuged. Despite the dilution, the solution remained about an order of magnitude above the critical micelle
93 concentration required to suspend SWNTs in Triton X which was calculated to be around 0.013%. Films were estimated by atomic force microscope (AFM) and sheet resistance measurements to be of approximately the same nanotube density as our standard VFET films though a small deviation from the standard film density is difficult to avoid. Figure 4 1. Structure of A) NPD and B) NPD CN VFETs with varying channel thicknesses. Contrary to standard procedure, the dilute SWNT films were transferred to Si/SiO 2 /BCB substrates prior to growing the gold source elect rode so that any contamination during processing steps in the glove box would be on top of, rather than below, the nanotube film and would be less likely to short Nonconductive particles underneath the conductive SWNT film can create shorting path ways through thin devices; however, if these particle land on top of an already transferred SWNT film, shorting can likely be avoided (though shadowed regions may still prove problematic ). In devices using a 200 nm NPD active layer, transferring the SWNT films prior to growing gold source electrodes increased the working pixel yield from 70% (with predeposited Au source contacts) to 96% (with postdeposited Au source contacts). A B
94 Following transfer of SWNT films, NPD was thermally evaporated onto the device. Each Si/SiO 2 /BCB/SWNT substrate can be divided into four quadrants (see Figure 3 3 for reference) and the channel can be grown independently in each quadrant by masking off the other regions. Therefore, 100 nm, 150 nm, 200 nm and 250 nm NPD VFETs were al l fabricated on a single substrate while 300 nm and 500 nm NPD VFETs were grown on a second substrate (and fabricated on a different day). The VFET stack was capped by a 30 nm Au thermally evaporated drain electrode with 40 nm Au source contacts grown on the SWNT film at the end of device fabrication. Output curves are shown for these devices are shown in Figure 4 2. As can be seen, thinning the channel layer yields significant enhancements in the device on currents (Figure 4 2A, B ). The device with a 100 nm channel thickness reaches 110 mA/cm 2 at 5 V, 50 times more than the current density achieved by the 500 nm device (2.2 mA/cm 2 ) and 2.5 times more than the 150 nm device (44 mA/cm 2 ) despite a reduction in the channel thickness of only a third. However, this increase in the on current density also coincides with a reduction in the off current density for thin channel layers, though the off state does not begin to suffer until the channel reaches approximately 150 nm. The on/off current ratio is shown in Figure 4 2D where the balance between these high on currents and reduced off currents becomes more evident. Transfer curves for these devices, where the source drain voltage was held constant while gate source voltage was swept across an 80 V range, are shown in Figure 4 3 These drastic improvements demonstrate the value of the purification techniques discussed in Chapter 2. With the low mobility of organic semiconductors, output
95 currents in organic TFTs are still 1 2 orders of magnitude below those a chievable by polycrystalline silicon 93, 111, 119 the thin channel lengths of VFETs offer a means to bridge this performance gap. In future VFETs, the use of th is highly purified material in conjunction with higher mobility organic materials should facilitate the development of even higher performing CN and graphene enabled VFETs. Figu re 4 2. CN VFETs with an NPD channel layer of varying thicknesses. Output curves in the on state (V G = 40 V) on a A) logarithmic and B) linear scale as well as in the C) off state (V G = 40 V) plotted on a logarithmic scale D) On/off current ratio found by dividing the most on state (V G = 40 V) output curve by the most off state (V G = 40 V) output curve plotted ag ainst the on current density. Significant enhancements in on currents can be achieved through the use of thinner channel layers though off currents may suffer.
96 Figure 4 3 Transfer curves in which the gate voltage is swept while the source drain voltage is held constant at the voltage indicate. On current density (V G = 40 V) was meant to be equal for all thicknesses to allow fo r a more meaningful comparison. CN VFETs with a Solution Processable Channel Layer Previous efforts involving CN VFETs have focused primarily on the use of small organic molecules as the organic channel with early efforts to use polymeric materials yieldi ng devices operating at levels below those required for use in commercial applications. 30 In this work, we seek to extend the materials accessible for use in the CN VFET architecture to include solution processable polyme rs. Such devices would enable the advancement of this technology to a state in which printable CN VFETs could be readily processed for mass market electronics. Poly(3 hexylthiophene) (P3HT) is a readily available, solution processable polymer frequently used in polymer based TFTs 102, 120 128 The mob ility of P3HT, though heavily dependent on factors such as molecular weight, regioregularity and ordering of the polymer, can reach values of up to 0.3 cm 2 127 129 Further, the HOMO level of P3HT ( 5.2 eV 130 ) is sufficiently deep to create a relatively large initial Schottky barrier
97 with a dedoped S WNT film so that the device can be suffi ci ently turned off, even as bundles of SWNTs may introduce screening effects. P3HT has been used previously in the VFET architecture in devices that used a thin, partly oxidized aluminum source electrode and an ultra high capacitance gate that relied on mobile ions 102 Switching speed is limited in such devices by the ionic mobility and the best on/off current ratio reported was 10 3 As shown below, the P3HT channel CN VFET do es considerably better. The device layout is as depicted in Figure 4 4 The gate dielectric can have a large impact on device characteristics; here we use an aluminum gate with a atomic layer deposited (ALD) Al 2 O 3 dielectric layer of approximately 25 nm, prepared by thermally evaporating 40nm of aluminum onto glass substrates followed by oxygen plasma ashing to create an approximately 5 nm thick alumina oxide layer. To ensure a robust dielectric layer free of leakage pathways, an additional 20nm of alumi na oxide was grown via atomic layer deposition. The resultant dielectric layer had a capacitance of 300 nF / cm 2 The BCB layer and transfer of SWNT films (using material that had been centrifuged at 17000 RPM in a fixed angle rotor) was carried out as d escribed previously. Solutions were made by dissolving P3HT (Sepiolid, Rieke Metals, > 98% regioregular, M W < 50,000) in 1,2 dichlorobenzene at a concentration of 30 mg/mL and stirring overnight. Under an inert argon environment, the solutions were then s puncast on the substrates at 600 RPM while the sample was simultaneously heated to approximately 80 C by illumination from a heat lamp, yielding a film that measured 325 nm thick. There was no thermal anneal following the spin coating. To complete the
98 device, 40 nm of gold was evaporated on top of this stack through a TEM grid (100 hexagonal mesh), defining a .035mm 2 pad to serve as the drain electrode. Figure 4 4. Schematic of P3HT based CN VFET with AFM image of SWNT film superimposed over BCB l ayer to depict transferred SWNT film. Transfer and output curves were measured under an argon environment using a Keithley dual channel sourcemeter (Model 2612) controlled by a custom LabView program. Transfer curves for a P3HT CN VFET held at a constant source drain voltage while sweeping the gate voltage from 3 V to +3 V and back are shown in Figure 4 6A. There is very little hysteresis shown in these devices. Over this 6 V range in gate voltage, the on off ratio remains above 10 4 for all drain volt ages up to 2 V. Figure 4 5 shows output characteristics for this same device, measured by sweeping the source drain voltage from 0 V to 3 V at gate voltages from +3 V to 3 V, stepping in 1 V increments with a maximum output current density of 84.5 mA/c m 2 at V SD = 3 V in the most on state. The on/off ratio, found by dividing the most on output curve ( 3 V) by the most off output curve (+3 V), is plotted versus on current density as the drain voltage is swept from 0 to 3 V in Figure 4 5 C. Though it doe s drop with
99 Figure 4 5. CN VFETs using a P3HT channel layer. (A) Transfer curve of the CN VFET over a 6 V range, beginning at V G = 3V, at decreasing drain voltages. Little hysteresis is observed but follows the path indicated by the arrows. (B) Output curves at gate voltages from 3 V to 3 V in 1 V increments. (C) On / Off ratio found by dividing the most on current (V G = 3V) by the most off current (V G =3V) and plotted against the most on current. On / Off ratio stays above 10 4 past 40 mA/cm 2 increasing voltage, the on/off rati o exceeds 10 4 through 40 mA/cm 2 falling to approximately 4 x 10 3 at the peak current of 84.5 mA/cm 2 (V SD = 3 V). Though the vast majority of published P3HT TFTs require voltages that are more than an order of magnitude grea ter than this CN VFET, Table 4 1 compares device performance for published P3HT devices operating at less than 5 V. As the difference B A C D
100 in how current is measured in conventional TFTs and VFETs makes direct comparison difficult, it is necessary to convert the linear current density of c onventional TFTs into an effective areal current density ( J eff ), as introduced previously by McCarthy et al. 40 J eff is calculated from by assuming an interdigitated source drain electrode pattern in which the source and drain electrodes are patterned to the same width as the channel length in the TFT, thereby maintaining the minimum feature size of the device, such that where W and L represent the channel width and length respectively. The P3HT CN VFET achieves the highest areal current density of these low voltage devices with roughly comparable on off ratio. This current density is more than enough for practical applications; if the P3HT CN VFET were used to drive a OLED with a luminance effic iency of 5 cd/A, then a current density of just 20 mA/cm 2 achievable at a drain voltage of 1.2 V (V G = 3 V), would be required achieve a luminance of 1000 cd/m 2 Table 4 1. Comparison of P3HT CN VFET to published, low voltage (<5 V) P3HT TFTs Ref Devi ce Type Capacitance (nF/cm 2 ) Operating V SD (V) Channel Length ( m) J EFF (mA/cm 2 ) I ON / I OFF 7 TFT 392 5 7 20 10 4 8 TFT 750 5 5 38 10 3 9 TFT 250 2.2 60 0.025 10 TFT 188 3 10 10 <10 2 11 TFT 332 4 30 1.1 150 1 2 Metal Base n/a 1.2 0.12 10 10 2 10 3 3 VFET 1000 5 17 10 3 This Work CN VFET 150 3 0.325 84.5 10 3 10 4
101 An important aspect of the operation of the CN VFET is that, while current in conventional lateral channel TFTs flows parallel t o the substrate, in a CN VFET current flows in the direction perpendicular to the problem. This is potentially problematic as many organic molecules and polymers demonstrate an anisotropy in their mobility with conventional TFT materials taking advantage of molecules in which the high mobility direction is along the substrate surface. Fortunately, it has been demonstrated that stack on the pristine sidewall of a SWNT, reorienting it such that the high mobility direction is now vertical, facilitating high performance CN VFETs 111 Such a reorientation is desired in P3HT on the SWNT film; however, while the stack ing on the SWNT sidewall, it is unclear if such a process will occur in a spuncast polymer. Previous studies have shown that the anisotropy between the mobility in P3HT when the (100) plane is oriented parallel to the substrate surface measured along and w hen the (010) plane is oriented in this direction can give rise to two orders of magnitude difference in carrier mobility between the two directions 108 For regioregular P3HT, the (100) plane typically lays parall el to the substrate surface and can have mobilities on the order of several tenths of a cm 2 stated field effect mobility of 0.2 0.3 cm 2 To study whether there is any reorientation of the P3HT as has been observed in small molecule devices, we extract the P3HT mobility using the Mott Gurney equation in the space charge limited (SCL) regime following methods described previously 111 though the extracted SCL mobility may vary from the field effect (FE ) mobility measured
102 in conventional TFTs 131 and make comparison to literature difficult. As the resistance of the dilute CNT film dominates the relationship between drain voltage and current density at high voltages, the mobility cannot be obtained directly from CN VFETs. Instead, the CNT film was transferred onto a more highly conducting ITO electrode thereby eliminating the film series resistance. Since the device possessed no gate electrode to reduce the contact b arrier between the nanotubes and the subsequent P3HT layer alternative means was needed to attain the Ohmic contact between them. To provide the needed Ohmic contact a thin, 1 nm interfacial layer of molybdenum oxide (MoOx) was evaporated between the CNT f ilm and the spuncast P3HT layer. A second 1 nm MoOx layer was evaporated on top of the P3HT followed by a 40 nm Au drain electrode. J V curves were recorded and fit to the space charge limited Mott Gurney equation: (Eq 4 1) where is the mobility, L r is the relative permittivity of the P3HT. The latter was determined by measuring the parallel plate capacitance of the P3HT film sandwiched between Al electrodes using an HP 4284A Precision LCR meter. In contrast with previous studi es involving thermally evaporated small molecule VFETs, we do not see any evidence of a reorientation of P3HT on the SWNT film though it is important to note that the MoOx may interfere with the reorientation. Unlike those systems, where conjugated orga nic molecules have bee the carbon nanotube side wall such that the high mobility direction grows vertically rather than horizontally, there is no evidence of such a process occurring here. Indeed,
103 measurement of the mobility of P3HT films spuncast on both ITO and dilute SWNT networks, yielded comparable values of approximately 3 x 10 3 cm 2 devices. At two orders of magnitude less than the manufacturer stated mobility of 0.2 0.3 cm 2 ult may confirm that there is no reorientation of P3HT lamellae on a SWNT film though more study into the difference between SCL and FE mobilities might prove insightful. Efforts to induce reorientation via thermal annealing yielded no improvement in devi ce performance. Despite the low mobility of P3HT in the vertical direction, we have fabricated solution processable CN VFET channel layers that achieve high currents at low operating voltages. With future materials that either reorient on the nanotube sid ewall or naturally possess high mobility in the vertical direction, we anticipate that solution processable channel layers will prove to be a viable option for future VFETs and VOLETs, opening manufacturing avenues such as facile inkjet printing fabricatio n. n Type Vertical Field Effect Transistors Despite significant research in organic TFTs, the majority of the materials studied have largely (though certainly not exclusively) focused on p type materials. There are many reasons for this, largely due to t he lack of appropriate, stable n type materials tied to the poor electrochemical properties of known electron accepting materials. Despite this lack of attention, n type transistors are essential to the development of complementary logic circuits and as s uch it is important to develop these materials. There have been previous forays into n type VFETs. The first published VFET, featuring a copper / aluminum source, demonstrated strong performance using a C 60 channel layer to achieve high on currents and on off ratios of more than 10 6 101 Another patterned VFET using a gold source electrode demonstrated much lower levels
104 of performance, achieving on current densities of less than 1 mA/cm 2 at 10V 103, 104 Additionally Bo Liu demonstrated ambipolar behavior in a CN VFET using a poly(9,9 dioctyl fluorene co N (4 butylphenyl) diphenylamine) / [6,6] phenyl C61 butyric acid methyl ester (PCBM) blend in whic h the PCBM acted as the n type channel. The n type performance in this CN VFET only achieved on off ratios of ~100 so an improved n type material was desired as a first step towards a VFET based inverter 109 CN VFETs wer e made using triply sublimed C 60 as the channel material thermally evaporated in a 450 500 nm thick layer to make VFETs of the structure shown in Figure 4 6 As C 60 is an n type semiconductor (LUMO = 3.7 eV, HOMO = 7.0 eV) 94, 132 the energy band alignment varies from the more typical CN VFET structures which have largely been studied using p type materials offering a new system to test and advance our understanding of device parameter s. It has been demonstrated that for p type channels, baking of the SWNT film prior to device fabrication dedopes the intrinsically p doped SWNTs, shifting the nanotube work function from approximately 4.9 eV to approximately 4.7 eV and generating an en hanced initial barrier to turn the device more fully off 40 However, for an n type material such as C 60 baking may make it more difficult to turn the device off as the barrier here will be diminished by bak ing though the very large initial barrier with C 60 may make the small shifts(0.1 0.2 eV) irrelevant. In a separate experiment published in his dissertation 109 Bo Liu demonstrated that for devices using a p type cha nnel ( poly[(9,9 dioctyl fluorenyl 2,7 diyl) alt co (9 hexyl 3,6 carbazole)] or (PF 9HK) ), the use of a higher work function metal enhanced device performance by increasing device on currents and decreasing off currents at a given
105 voltage. The enhanced on currents were attributed to a diminished hole extraction barrier at the organic / metal interface with increasing work function metals. It was not clear why the off currents were increased in lower work function metals, though it was proposed that the inc reased reverse injection of electrons into the LUMO may have played a role. Here I revisit both of these experiments (regarding the effects of baking and drain electrode selection) using C 60 to explore these parameters in an n type VFET. Figure 4 7 shows typical output and transfer curves for CN VFETs that have been baked to 225 C followed by deposition of a 450 nm C 60 channel and either an aluminum (work function = 4.3 eV 133 ) or gold drain electrode (work function = 5.1 eV 133 ) Though both devices perform well, devices using an aluminum drain electrode achieve d higher on current s and lower off currents. The enhanced on currents can be explained by a smaller hole extraction barrier at the C 60 / Al contact; however, similar to the studies by Liu, this makes it difficult to explain the higher off currents in the Au based device. The reverse injection in these d evices were measured (Figure 4 8 ) by applying a negative source dra in bias at varying gate fields and a higher current in the reverse direction was observed perhaps a contributing factor. To probe the effects of baking on device performance, CN VFETs were made with a baked (dedoped) SWNT film as well as a non baked film These devices again had a 450 nm C 60 layer and a 40 nm Al top electrode was used. In this experiment, the baked film generally performed better except at high voltages (Figure 4 9 ). This may be explained by t h e large initial barrier between the SWNTs and C 60 allowing the device to be reliably off so the slightly diminished barrier is not a major issue in the dedoped film until at higher voltages, where we do begin to see the off current creep up. The device
106 did turn on a little faster so it seems that the reduced barrier may contribute in this regard At high temperatures (>600 C) baking has been demonstrated to help to remove contaminants and coatings on the SWNT sidewall, leaving a more pristine surface that could also benefit charge injection. This may have b e e n occurring here as well, though at 225 C it seems unlikely. These C 60 based devices should enable the development of CN VFET based CMOS like inverters and ring oscillators. The large initial Schottky barrier to electron injection does appear to inhibit device performance and require relatively large source drain voltages, particularly compared to earlier VFET work by Ma and Yang 101 which incorporated a porous aluminum (work function = 4.1 eV 134 ) source electrode for a smaller initial barrier. Though these devices do not benefit from Schottky barrier height modulation, the smaller initial barrier appears to compensate for this. This idea is further strengthened by comparis ons to work by Tessler 103, 104 where the use of a patterned Au source (work function = 5.1 eV 134 ) with an even larger initial barrier achieves two orders of magnitude less current at 10 V than in the CN VFET. Future studies of materials with smaller initial barriers should enhance device performance. Figure 4 6. Schematic of C 60 based CN VFET with AFM image of SWNT film superimposed over BCB layer to depict transferred SWNT film.
107 Figure 4 7 Device data for CN VFETs with a 450 nm C 60 active layer and either Au or Al drain electrodes, as indicated. The smaller barrier formed with Al demonstrates improved performance in these devices, though the difference is small Figure 4 8 Reverse injection in CN VFETs with a 450 nm C 60 active layer and either Au or Al drain electrodes, as indicated.
108 Figure 4 9 Device data for CN VFETs with a 450 nm C 60 active layer and 40 nm Al drain electrode. Devices were either not baked or baked to 225 C prior to C 60 deposition. Here baking improves device performance.
109 CHAPTER 5 ORGANIC LIGHT EMITTI NG DIODES USING THIN FILMS OF SINGLE WALLED CARBON NANOTUBES AS ANODES Introduction While the principle of electroluminescence in organic materials has been known since the early 1950s 135, 136 it was the d iscovery of the modern, efficient organic light emitting diode (OLED) by Tang and Van Slyke in 1987 137 that sparked a flurry of research into this organic electronic device structure 138 Research in the field has yielded insights into organic semiconductors as well as into the charge injection and transport processes in organic electronic devices that have benefited not only the development of OLEDs but also our understanding of organic semiconductors. Progress in OLEDs has led to the development of an array of consumer technologies that incorporate OLED displays including phones, watches, media players and televisions that offer richer colors, wider viewing angles and lower pow er consumption. Research in OLEDs has focused largely on the organic active layers and electron injecting cathode as the stringent requirements demanded by the transparent, hole injecting anode has restricted the range of materials available for this ele ctrode. To date, no clear cut alternative to the same transparent conducting oxides (such as tin doped indium oxide (ITO)) used in the first modern OLED has been found. As the e materials has grown in recent years. Thin films of single walled carbon nanotubes possess many similar optical and electrical properties and offer a potential alternative to ITO. Further, the large surface area of the porous SWNT film and low density of states suggest ways in which the properties of the SWNT film might contribute to increased charge injection
110 and better band alignment. In this chapter I will discuss organic light emitting diodes that incorporate these SWNT thin films and compare these d evices to ITO. Theoretical Background All OLEDs operate on the same fundamental principles requiring the simultaneous injection of oppositely charged carriers (electrons and holes) into an organic electroluminescent layer where they recombine to form an ex citon which may radiatively decay into an emitted photon 139 While more advanced designs may include one or more additional hole/electron transport/blocking layers to improve the efficiency of this process throug h charge confinement, the most basic design requires only an anode for hole injection and a cathode for electron injection sandwiching the electroluminescent layer (Figure 5 1A). This is basic the structure that will be the focus of this chapter. There are several important physical properties of these materials involved that will dictate both the efficiency of these processes and the voltages required to drive this to remove an electron from the interior of a solid to the vacuum level (Figure 5 1B) 140 The electroluminescent material is characterized by the energy of its lowest unoccupied molecular orbital (LUMO) in its co nduction band and its highest occupied molecular orbital (HOMO) in its valence band, analogous to the electron affinity and ionization potential respectively in inorganic crystalline semiconductors. In an ideal OLED, it is desirable to align the HOMO with the work function of the anode and the LUMO with the work function of the cathode. Differences between these can create a potential barrier to charge injection that must be overcome by tunneling through the barrier as dictated by Fowler Nordheim tunnelin g theory 139, 141, 142
111 When the materials are brought together, thermodynamic equilibrium requires charge exchange that brings the work function of the anode, cathode an d the Fermi energy of the electroluminescent layer into alignment (as discussed in great depth in Chapter 3) 143 Under the application of a bias voltage, an electric field is created which reduces the barrier to charge injection. At sufficiently high field strengths, holes are injected from the anode while electrons are injected simultaneously from the cathode into the HOMO and LUMO, respectively, of the electroluminescent layer (Fig 5 1C) 142 For polymeric electroluminescent layers, charge transport occurs via site to site hopping along the polymer chain through the active layer (Fig 5 1D). This process occurs rapidly and the holes and electrons combine within the active l ayer to form an exciton (Fig 5 1E). This exicton can decay radiatively to emit a photon of light (Fig 5 1F) or non radiatively in which case the excess energy is released as heat 139 Figure 5 1. Schematic of OLED operation 144
112 The electroluminescent and charge transporting layers can be either polymeric or small molecule layers and significant research has gone into the study and optimization of these materials as well as the ele ctron injection cathode (for clarification, polymer light emitting diodes are sometimes referred to as PLEDs; however, here I will use the more general term of OLEDs for both device types). Despite the wealth of research in these areas, limited progress h as been made in the study of novel anodes beyond transparent conductive oxides despite known limitations with these materials 145 The lack of progress here may be largely due to a lack of materials that can meet the stringent set of requirements imposed upon the anode, such as high conductivity, optical transparency and a high work function, which has limited the opportunity for anodes composed of most materials other than the transparent conducting oxides such as ti n doped indium oxide (ITO) conventionally used in OLEDs. Some progress has been made in the use of conductive polymeric anodes ; however the low mobility of most organics, typically less than 1 cm 2 /V s, has limited their viability 146 149 Thin films of single walled carbon nanotubes (SWNTs) have been proposed as alternative anode materials for their high work function, considerable conductivity and trans parency throughout the visible spectra 38 In 2006, three papers 46, 63, 150 were published reporting the performance of OLEDs usi ng thin films of SWNTs as the anode. Each study noted the difficulty in dealing with the large surface roughness of the SWNT film and attempted to overcome this roughness by planarizing the nanostructured film. Two of the studies applied a water based pl anarizing layer p oly(3,4 ethylenedioxythiophene) poly(styrenesulfonate) (PEDOT:PSS) to the SWNT film 63, 150 This decreased the surf the surface roughness but the aqueous solution likely did not
113 penetrate well into the hydrophobic nanotube film, though in one of the papers the authors did blend the aqueous PEDOT:PSS solution with methanol to enhance this wetting. The third effort used an evaporated insulating layer which may degrade charge inj ection and also used an evaporated small molecule organic layer which would only coat the surface of the SWNT film 46 In these reports, and others published since 43, 45, 144, 151 153 carbon nanotube based devices have failed to outperform ITO control samples (except with regards to f lexibility) but these tests have often treated SWNT films similarly to ITO, not fully exploiting the large surface and low density of states of SWNT thin films. It has been demonstrated in double layer capacitance measurements that the porosity of a 50 n m SWNT films enhances the accessible surface area by a factor of 2.5 relative to a planar electrode 55 which could significantly enhance performance and light emission. The use of a single layer OLED structure that inc orporates a polymeric active layer should permit penetration of the organic material into the SWNT film to access this surface area. In this work, I will examine what effect, if any, a SWNT thin film has on single layer organic light emitting diodes. Techn ical Approach Prepatterned ITO on glass substrates, with an ITO thickness of 150 nm and a slides were first cleaned by lightly scrubbing with alconox followed by aceton e and methanol rinses. The ITO slides then underwent an oxygen plasma ashing (3 min, 300 W, 300 sccm) to render it hydrophilic, eliminate any organic contaminants and modify the surface layer to improve hole injection 54 A thin (~40 nm) p type hole conducting polymer, p oly(3,4 ethylenedioxythiophene) poly(styrenesulfonate) (PEDOT:PSS,
114 Baytron P AL 4083, 0.45 micron filtered ) was spuncast on the ITO surface at 4000 RPM to prevent shorting by ITO spikes and also to furthe r improve the hole injection from the ITO into the organic layer by lowering the barrier height at this interface (work function of 5.1 eV) 154 156 Following a 2 hour bak e at 120C to dry the PEDOT:PSS layer, these ITO substrates were transferred to an argon glovebox for OLED fabrication (Figure 5 2A). SWNT films were fabricated at a thickness of 50 nm (unless otherwise noted) on mixed cellulose ester membranes using methods des cribed previously in this document and elsewhere 38 and transferred to glass slides with predeposited gold or palladium electrodes (40 nm) on top of a chromium adhesion layer (10 nm). The SWNT films were cut to the same pattern as the ITO and transferred to glass substrates and brought into the glovebox for device fabrication (Figure 5 2B). The PEDOT:PSS layer was intentionally omitted from these SWNT film based devices to allow the polymeric layers direct access to the large surface area of the SWNT film. Poly[2 methoxy 5 (2' ethyl hexyloxy) 1,4 phenylene vinylene] (MEH PPV) is a conductive, light emitting polymer that has been thoroughly studied for over two decades 157 and was chosen as the electroactive layer for its well known properties on ITO which allowed for a meaningful comparison with the properties of the SWNT film. With a HOMO around 5.3 eV, MEH PPV offers a relatively small barrier to charge injection and has a band gap of ~2.3 eV for an orange emission 158 Aromatic solvents have been shown to improve polymer/anode contact and electrical conduction in MEH PPV and so MEH PPV was dissolved in chlorobenzene 159 This solution was then spun onto the SWNT or ITO/PEDOT PSS surface, forming an electroactive layer of a
115 thickness that was controlled by varying either the spin speed or polymer concentration. Calcium electrodes (10nm, evaporation rate: 1 /s), serving as the cathode with a work function of 2.9 eV 142 were thermally evaporated onto the MEH PPV and topped by a thermally evaporated aluminum layer (75 nm, 2 /s) to prevent oxidation of the calcium. Th e area of the Ca/Al pixel was defined by a shadow mask during thermal evaporation to be 8 mm 2 with eight pixels per device (Figure 5 2C). Organic light emitting diodes were fabricated and measured under the inert argon environment of a glove box (Figure 5 2D). MEH PPV was wiped off of the gold electrodes and contact was made to the sample using a sample holder custom made sample holder. Voltage was supplied by a Keithley 2400 power meter and measurement was carried out by a handheld Minolta colorimeter (m easured through the glass) or using a n Ocean Optics fiber optic spectrometer that was manipulated and aligned with a micrometer inside of a custom built, light tight box that was anodized to minimize reflected light. For these OLEDs data was taken by hand. Figure 5 2. Structure of A) ITO and B) SWNT based MEH PPV OLEDs. Photograph of OLED layout from C) top view and D) emitting light under applied voltage. A B D C
116 Results and Discussion Initial MEH PPV OLEDs were built both on ITO /PEDOT:PSS an d SWNT films using MEH PPV dissolved in chlorobenzene at a concentration of 5 mg/mL and spuncast at 1000 RPM to achieve film thicknesses of 100 nm. The ITO based devices performed very well under such conditions, achieving maximum luminance values of over 14,000 cd/m 2 (Figure 5 3), as measured by the Ocean Optics fiber optic spectrometer, far exceeding the requirements of most applications which is approximately 100 500 cd/m 2 for displays. Unfortunately, every SWNT based device fabricated with 100 nm or ganic active layers shorted without any light emission. Figure 5 3. Luminance (blue, left axis) and current density (orange, right axis) of ITO device using a 100 nm MEH PPV layer. Post failure optical microscopy revealed localized, clearly demarcate d circular patterns with black spots at their centers and in many cases tracks typical for electrical breakdown. These suggested the existence of localized conductive high spots in the SWNT anode leading to locally thin anode to cathode distances in the de vices causing these failures. AFM of the bare SWNT anodes indeed showed locally high features sufficiently tall to penetrate through the relatively thin MEH PPV organic layer and
117 directly contact the Ca cathode, creating a direct shorting pathway through w hich the current could entirely bypass the photoluminescent layer. These features arose from an array of sources (discussed in Chapter 2) including particulates and ridges in the SWNT film where material had accumulated in scratches in the membrane (as sh own in Figure 2 5B C). These shorts made fabrication of SWNT based OLEDs using a 100 nm active layer impossible and prompted the purification work that is the focus of Chapter 2. The complexity of the purification of SWNTs made it clear that highly pure SWNT material would not be immediately available (indeed, it wound up taking three years significantly longer than initially anticipated). In order to allow the OLED project to move forward, a thicker MEH PPV layer was used to overcome these local prot rusions To cast layers that could be up to 500 nm thick, the concentration of MEH PPV in chlorobenzene was increased to 10 mg/mL (twice the initial concentration). The significant viscosity of this solution prompted concerns that this new solution would not penetrate the SWNT film and fail to take advantage of the large available surface area. To promote this penetration and help build a thick MEH PPV layer that more completely covered all parts of the SWNT film, the spincoating procedure was modified f rom a single spin to three sequential spins at increasing concentrations 1 mg/mL, 5 mg/mL and 7.5 mg/mL or 10 mg/mL with a 15 minute bake at 55 C (T g of MEH PPV = 65 C 160 ) in between spins to dry the organic active layer and drive off residual solvent. The use of these thicker layers largely overcame the observed shorting pathways and permitted the fabrication of both SWNT and ITO based OLEDs that used these thicker layers. On a typical SWNT device with eight pixel s, there still were frequently 3
118 4 pixels that were either partially or fully shorted but enough working devices could be fabricated to test the films and gain some preliminary experience with the OLED. The SWNT film for these devices was made using mater ial that had been centrifuged twice at 6000 RPM and filtered twice through 0.65 micron filtration membranes and then vacuum filtered on Sterlitech membranes and transferred to glass. Spins of MEH PPV in chlorobenzene at increasing concentration (1 mg/mL, 5 mg/mL and 7.5 mg/mL) created a film that was roughly 200 nm thick. As shown in Figure 5 4, the best SWNT based devices consistently achieved nearly twice the luminance achieved in the ITO control OLED at comparable current densities yielding more than a factor of two improvement in the maximum current efficiency of the SWNT based OLED relative to ITO (1.5 cd/A vs. 0.74 cd/A). These results appeared to be a confirmation of the potential benefits of the SWNT film Figure 5 4. A) Luminance and B) Cu rrent Density for MEH PPV OLEDs using thick active layers and SWNT material that had been twice centrifuged at 6000 RPM and filtered. Here the SWNT based OLED s achieve twice the luminance at comparable current densities as in the ITO device. There were two major caveats to these results that could not be neglected. First, though the SWNT film based OLED outperf ormed its ITO counterpart when thick MEH A B
119 PPV layers were used, the best ITO devices using thinner MEH PPV layers were still a factor of 3 4 better. Secondly, the extremely rough SWNT film has many locations where local high spots thin the organic layer so that the effective MEH PPV thickness might be substantially thinner on the SWNT based device than in OLEDs using ITO, which could account for the improved performance observed. While these initial results were promising, it became clear that a true compa rison would require the use of optimized, thin organic active layers. The success of the continuous flow centrifuge (Chapter 2) in generating highly purified material allowed for a more meaningful comparison to be attempted SWNT films using this material were made at thicknesses of 20 nm and 40 nm and transferred as described previously (with the exception of the gold contacts to the SWNT film which were deposited after transfer of the SWNT film to avoid the introduction of particulates in the glove box e nvironment). Following the same procedures as were used previously, thick MEH PPV OLEDs were made on these films and on an ITO substrate. The MEH PPV thickness was measured just adjacent to the SWNT film by scratching through the MEH PPV / Ca / Al electr ode using a razor blade. The metal coated region was used to limit stretching of the MEH PPV film during scratching. An AFM stepheight was taken and it was found that the film was approximately 200 nm thick above each anode that is to say that the MEH PPV layer was 240 nm thick on the 40 nm SWNT substrate which suggests that the cathode should sit 200 nm above the 40 nm SWNT film, and similarly for the 20 nm SWNT film and ITO electrode. It should be noted that there is significant variability in the me asured thicknesses and in a few regions the MEH PPV layer on the 40 nm SWNT film was 40 50 nm thicker.
120 In these devices that incorporate the use of highly purified SWNT material, OLEDs using ITO and both 20 nm and 40 nm SWNT films performed nearly identi cally. Representative pixels are plotted in Figure 5 5 and, in contrast to previous experiments, show nearly identical luminance values and current densities in all three devices. Accepting that the MEH PPV film thicknesses are indeed comparable, there d oes not appear to be any improvement arising from the enhanced surface area of the 40 nm thick film. Similarly, the roughly comparable turn on voltage that can be observed in Figure 5 5D suggests a lack of enhanced energy alignment in these SWNT based dev ices (the earlier turn on that appears present in the 20 nm thick SWNT device is likely due to a leakage current that remains in shorting pathways). Additionally, even with this more highly purified material, the SWNT based devices still begin s to show shorting currents when organic layers below 200 nm are used while the performance of the ITO based devices improve. This suggests that the roughness of the SWNT films may still be contributing to an effective thinning of the organic layer even in the devices using a 200 nm thick active layer shown in Figure 5 5. Conclusions Though SWNT based OLEDs have achieved results comparable to ITO devices using thick organic layers, a fair comparison in optimized devices is made impossible by particulates in the SWNT film. T hough further purification may improve these results, at present, it appears that SWNT based OLEDs must incorporate the planarizing layers demonstrated in literature. Though SWNT films may enabled some unique or novel designs such as flexible OLEDs and, i f sufficiently n doped, may ultimate prove useful for cathode applications in transparent OLEDs, there does not appear to be an advantage to the use of SWNT films for traditional OLEDs.
121 Figure 5 5. Results for ITO a nd SWNT based MEH PPV OLED using highly centrifuge d material including A) luminance, B) current density (linear scale), C) current efficiency and D) current density (log scale). As c a n be s e en, t he de vices perf o r m comp a rab ly w h en th is more high l y purifi e d mate ri al is u sed.
122 CHAPTER 6 LIGHT EMITTING ELECT ROCHEMICAL CELLS USI NG THIN FILMS OF SIN GLE WALLED CARBON NANOTU BES AS ELECTRODES Overview A ligh t emitting electrochemical cell (LEC) is a solid state electrochemical device in which light emission occurs via the formation of p n junction in a redox reaction. Though LECs share an apparently similar structure and performance characteristics to those found in organic light emitting diodes, the device mechanisms are fundamentally different and insensitive to either the active layer thickness or electrode work functions. This offers an exciting opportunity for transparent thin films of SWNTs serve as bo th electrodes in a dual emissive device. Here I discuss the fabrication of two types of LECs using thin conductive films of SWNTs as the transparent electrode. First I studied a conventional LEC using a SWNT film to replace the more traditional ITO elect rode that emits light in a single direction in a structure similar to the OLEDs discussed previously. After this single emissive device, I fabricated a more revolutionary structure in which I took advantage of the flexibility of a thin SWNT film on plasti c to demonstrate a dual emissive device that emits light in both the forward and reverse directions. Operating Principles and Scientific Background Light emitting electrochemical cells (LECs) are made up of two electrodes sandwiching a n electroactive lay er that is comprised of a n intermixed blend of a luminescent conjugated polymer and a n electrolyte (Figure 6 1A) 161 165 As voltage is ap plied across the two electrodes, the mobile ions contained in the electrolyte redistribute and form an electric double layers at the electrode / polymer interface. LUMO gap electrons can be injected int o the conjugated polymer and are electrostatically compensated by
123 counterions from the electrolyte, p doping the region adjacent to the anode and doping the opposite region adjacent to the cathode to be n type. As the electrical conductivity of polymeric materials increases with doping concentration, these layers create a low resistance contact at the polymer/electrode interface. This highly conductive doped region grows with time and allows for the migration of holes and electrons, driven by the voltage induced electric field towards the center of the cell where they meet in a thin insulating layer to create an electrochemically induced p i n junction in which electrons and holes combine and radiatively decay to emit light (Figure 6 1C) Figure 6 1. O perational mechanism for a light emitting electrochemical cell. a) Structure of typical LEC with two electrodes sandwiching an electroactive layer comprised of a blend of conjugated luminescent polymer and a solid electrolyte. b) After bias voltage is appl ied the electroluminescent polymer is oxidized or reduced adjacent to opposite electrodes, leading to the introduction of p and n type carriers, compensated by counterions from the electrolyte. c) Charge carriers are driven by the induced electric field t owards the opposite electrode and meet to create an electrochemically induced p i n junction in which electrons and holes combine and radiatively decay to emit light As the LECs rely on the growth and electrochemical doping of these p and n regions (whic h can be thought of as the extension of an optimized anode and cathode)
124 to drive light emission, the device itself is relatively insensitive to the work function of either the anode or the cathode 161, 163, 164, 166 This allows an LEC to operate under either forward or reverse bias and furthermore allows for the same electrode material to act as both anode and cathode. This has been demonstrated in seve ral lateral channel LECs with coplanar electrodes 166 but most readily deposited materials accessible in a standard LEC architecture do not possess sufficient transparency for light emission. Materials that do posse ss this transparency, such as transparent conducting oxides like ITO, are typically sputtered and become difficult to deposit on top of a polymer layer without damaging or overheating the polymer unless advanced techniques are used 167, 168 SWNTs do not require energetic deposition techniques and their mechanical flexibility 169, 170 should allow thin films of SWNTs to be laminated on top of the polymer layer. This enables a device in which the SWNT thin film can act as both transmissive anode and cathode to allow light emission through each for a dual emitting device. The highly conductive, doped regions also make LECs relatively ins ensitive to film thickness as has been demonstrated in devices that used a millimeter sized lateral gap between coplanar electrodes 166, 171 Though the turn on time, limited by ionic mobilitie s which are significantly lower than electronic mobilities, does increase substantially in these thicker devices (over 5 minutes in devices with a 1 mm gap), it does not appear that other device metrics suffer by using thicker active layers. This offers an opportunity to overcome the shorting pathways in carbon nanotube thin films by using thicker active layers without compromising performance. Though LECs do not have as extensive a background of scientific research as can be found in the OLED literature the field has progressed significantly since pioneering
125 161, 162, 171 Since this time, there has been steady research into the operational mechanisms, device structure and interplay of the components that make up the electroactive layer. Due to the insensitivity of the LEC to the work function of the electrodes, the area of alternative electrode s has been the subject of greater inquiry than was observed in OLEDs where the demands are more restrictive and limit material selection. Though at the time that this research project began there was no existing literature on either SWNT based LECs or dua l emissive devices, within six months of the project commencing and shortly after we achieved our first working devices, two independent reports were published of devices discussing dual emissive LECs incorporating carbon based electrodes one on SWNT fil ms and one using graphene. In November of 2009 the group of Qibing Pei (one of the inventors of the LEC) published a transparent, dual emissive device formed by the lamination of two thin SWNT films on PET sandwiching the electroluminescent layer 43 A few months later, Matyba et al. from the group of Ludvig Edman, published dual emissive devices using graphene and PEDOT:PSS as the two transparent electrodes 172 Incidentally, the first authors from both papers presented also presented my preliminary work on dual emissive devices. Since this sudden burst of independent publication, there have been several related follow ups 173 175 limiting the uniqueness of the research done here. Single Emissive Devices Experimental Methods Polymer light emitting electrochemical cells that emit light in a single direction were fabricated in a device structure similar to that used for organic light emitting diodes and
126 shown in Figure 6 2. 50 nm thick SWNT films were fabricated on a mixed cellulose ester (MCE) membrane via vacuum filtration as described in Chapter 1 and transferred to glass substrates with prede posited gold electrodes (with a chromium adhesion layer). Prepatterned ITO substrates were purchased from VisionTek with an ITO thickness of followed by acetone and methanol rinses. As a hole transport layer serves little purpose in a light emitting electrochemical cell, no PEDOT:PSS layer was spun cast on the ITO substrate. All steps after substrate cleaning / preparation were carried out in the inert argon environment of a glovebox. For the electrochemiluminescent layer, MEH PPV was again used as the emissive conjugated polymer and was admixed with poly(ethylene oxide) (PEO) complexed wi th a salt, lithium triflate (Li O T f), to serve as the polyelectrolyte. MEH PPV was used as re ceived; however, the PEO and Li O T f underwent a vacuum anneal to remove any adsorbed water. Master solutions of each were made in cyclohexanone at concentrations of 10 mg/mL for MEH PPV and 20 mg/mL each for PEO and LiOTf with heating of the PEO so lution to 50 C necessary to get the material to go into solution. From these master solutions, blend solutions were made at a 1:1:0.20 (MEH PPV:PEO:LiOTf) weight ratio. This blend solution is further split and diluted with pure cyclohexanone by a factor of 10 for one solution and a factor of 2 for the second with some solution remaining at full concentration. These solutions were spuncast at 1000 RPM onto the ITO and SWNT substrates in three sequential steps, starting with the most dilute solution and in creasing in concentration, with a 50 C thermal anneal between each step to dry the film. These multiple spins were used to build up the film
127 thickness to 300 350 nm (to overcome protrusions in the SWNT film) with the initial dilute spins enhancing penet ration into the porous SWNT film. On top of this polymer / electrolyte layer was evaporated 75 nm of aluminum with a shadowmask defining an 8 mm 2 active area. Devices were measured in a custom made sample holder using a Keithley 2400 Sourcemeter and a Mi nolta Colorimeter inside a light tight box. Figure 6 2. Device schematic for ITO and SWNT film based light emitting electrochemical cells. Results In direct comparisons between ITO based and SWNT film based LECs, it was found that though both devices a chieved light emission, the ITO based devices achieved nearly five times the luminance observed in SWNT based devices with an improved current efficiency (in terms of luminance per current required) in the ITO based devices (Figure 6 3). One phenomenon th at was observed but is not shown in the data is that many SWNT based LECs that were initially non shorted developed shorts after application of moderate to low voltages (often less than 4 V). A literature review prompted by this observation discover e d that PEO is a key component in many SWNT surfactants including Triton X 100, the material that we use in our SWNT suspensions 176, 177 This suggests that the PEO may be acting to lift loosely bound SWNTs from the surface and into the active region which may explain the relatively high currents at low luminance in the SWNT based devices. To avoid this, a pyrene
128 functionalized polyfluorene based polymer (Sticky PF) was used to coat the SWNT film by soaking in solution and appeared to reduce, but not eliminate the shorting issue, though no significant change in performance was observed 47 Another possible explanation for the di minished performance is that the nanotubes may induce preferential phase segregation in the active layer around the SWNT surface. Such phase segregation could alter the MEH PPV / PEO ratio in this region and inhibit balanced charge transport as well as th e formation of a highly doped region. Phase segregation, arising from the mixing behavior of the non polar MEH PPV with the polar PEO solid electrolyte, is an important factor in LECs and can be seen in both ITO and SWNT LECs via AFM at the top surface of this polymer film in Figure 6 3. The phase segregation seen here is not uncommon and can be controlled either through additives or the use of a crown ether based electrolyte in place of the PEO. How this phase segregation changes at the SWNT film surface is difficult to discover without generating cross sectional slices of the device for TEM via etching by focused ion beam which is both expensive and challenging and the heating induced in this process could also damage the polymer itself. Though the us e of crown et her alternatives might have Figure 6 3. A) Luminance and B) Current density data for a single emissive SWNT and ITO based LEC. C) Phase segregation observed in AFM imaging of the LEC surface.
129 b een i n t eresting they would involve starting over from scratch and, in light of other publications on both SWNT based LECs and dual emissive LECs, these avenues were not pursued. Dual Emissive Light Emitting Electrochemical Cells Experimental Methods Though the performance of ITO b ased LECs exceeded observed the performance in devices using a thin SWNT film, the mechanical flexibility of SWNT films offered an opportunity to create a novel, dual emissive device that was not available using ITO. For this device, SWNT films were trans ferred to both glass and p olyethylene terephthalate (PET) substrates with predeposited gold electrodes using a thin chromium adhesion layer. PET is a transparent flexible material (tradename: Mylar) that tolerates many chemical solvents and did not requir e any major modifications of the film transfer process. The PET was cut into rectangular strips that were longer and narrower than the glass substrates used so the substrates would not fully overlap, allowing contact to be made to the gold electrode / SWN T film. Solutions were mixed as described for the single emissive devices. These solutions were spuncast onto the SWNT films on both PET and glass substrates, again increasing from dilute solutions to a more concentrated ones. Immediately after the fin al spin, while the solutions were still wet, the PET and glass were placed in contact and clamped between metal plates using hand held spring clamps to supply the force. Samples were left in clamps overnight forming the structure shown in Figure 6 4. Devi ce yield upon removal from clamps was low and appeared dependent on how wet the solution coated substrates were at time of clamping. If the substrate was either too wet (including when the solution was dropcast immediately prior to clamping) or too
130 dry, t he stack did not adhere; controlled exposure to solvent vapor also proved ineffective. By trial and error, successful dual emissive devices were fabricated, though their stability if torque or stressed was not very good. Electrical contact was made to the gold electrodes using alligator clips and voltage supplied by a Keithey 2400 Sourcemeter. Figure 6 4. Device schematic of a dual emissive LEC. Results The lack of stability in the adhesion of the substrates made the measurement of the dual emissive dev ices difficult. The devices were measured in air to avoid mishandling in the glovebox. Lifetimes were short (minutes) and devices appeared to locally burn out, possibly due to inhomogeneities in the polymer layers from clamping. The instability of the d evice made accurate quantitative measurements impossible; however a picture showing device operation is shown in Figure 6 5. The device is photographed against a mirror to show emission in both directions. Though clearly room for improvement exists in th is device, it shows the potential for use of SWNT films in light emitting devices that emit light in both directions.
13 1 Figure 6 5. Dual emissive LEC in light (left) and dark with emission (right).
132 CHAPTER 7 CONCLUSIONS AND PATH S FORWARD This dissertatio n has demonstrated organic electronic devices that take advantage of the special properties of SWNTs in addition to new purification technique that will enable further progress in SWNT based organic electronic devices by allowing for thinner organic active layers to be used. The continuous flow centr ifuge technique demonstrated in Chapter 2 not only offers high levels of purity but also the throughput of material on a scale that is appropriate for industrial application. Though further studies may fine tu ne the parameters and optimize this process the ability to purify on this large scale allows future students to spend less time focusing on small batch purification of SWNTs and more time studying the fundamental physics that this material makes possible. Substantial progress was also made in understanding Schottky barrier modulation at the interface of low density of state materials and organic semiconductors but there is room for even more growth in our understanding. Projects are underway already withi n the Rinzler lab to pattern both graphene and metal electrodes and compare the VFET architecture in low DOS state metals such as graphene directly to those using high DOS materials with a comparable work function (such as Ag). Similarly, though the funda mentals of graphene enabled VFETs have been demonstrated here, there is much still to be learned in the role that pore size, density and pore ordering plays. With further understanding, and the incorporation of still thinner layers as allowed by the plana r graphene, the outlook for these G VFETs is promising This dissertation has also demonstrated the versatility in the CN VFET by incorporating solution processable materials that achieved state of the art performance and n type C60 based VFETs.
133 This pro gress should enable the incorporation of the VFETs into new devices such as CMOS like inverters work that is already underway. Three different light emitting devices were als o demonstrated within this work. T hough SWNT based OLEDs and LECs still cann ot exceed the performance of ITO, in conventional devices there is potential for their incorporation in novel designs such as flexible OLEDs or the dual emissive LEC. Further investigation into these devices may also yield important insight for future a dv ancements in novel designs such as the vertical organic light emitting transistor that incorporates a CN VFET with an organic light emitting diode. Low density of states electrodes offer exciting possibilities for device designs that help to move organic electronics beyond being existing technologies. There is significant room for exciting growth in the field and perhaps this work offers some hints at the future paths that these materials make possible I am excited to see what the future holds for orga nic electronics.
134 LIST OF REFERENCES 1. Bolotin, K. I. et al Ultrahigh electron mobility in suspended graphene. Solid State Commun. 146 351 355 (2008). 2. Wallace, P. The band theory of graphite. Phys Rev 71 622 (1947). 3. McClure, J. Diamagnetism of graphite. Phys Rev 104 666 (1956). 4. Slonczewski, J. & Weiss, P. Band structure of graphite. Phys Rev 109 272 (1958). 5. Novoselov, K. et al Electric field effect in atomically thin carbon films. Science 306 666 669 (2004). 6. Geim, A. K. & Nov oselov, K. S. The rise of graphene. Nat M ate r. 6 183 191 (2007). 7. Berger, C. et al Electronic confinement and coherence in patterned epitaxial graphene. Science 312 1191 1196 (2006). 8. Sutter, P. W., Flege, J. I. & Sutter, E. A. Epitaxial gra phene on ruthenium. Nat. Mater. 7 406 411 (2008). 9. Berger, C. et al Ultrathin epitaxial graphite: 2D electron gas properties and a route toward graphene based nanoelectronics. The J. of Physical Chemistry B 108 19912 19916 (2004). 10. Rein a, A. et al Large area, few layer graphene films on arbitrary substrates by chemical vapor deposition. Nano letters 9 30 35 (2008). 11. Li, X. et al Large area synthesis of high quality and uniform graphene films on copper foils. Science 324 1312 1314 (2009). 12. Stankovich, S. et al Synthesis of graphene based nanosheets via chemical reduction of exfoliated graphite oxide. Carbon 45 1558 1565 (2007). 13. Gmez Navarro, C. et al Electronic transport properties of individual chemically reduced grap hene oxide sheets. Nano letters 7 3499 3503 (2007). 14. Eda, G., Fanchini, G. & Chhowalla, M. Large area ultrathin films of reduced graphene oxide as a transparent and flexible electronic material. Nature nanotechnology 3 270 274 (2008). 15. Li, X. et al Transfer of large area graphene films for high performance transparent conductive electrodes. Nano letters 9 4359 4363 (2009).
135 16. Novoselov, K. et al Two dimensional gas of massless Dirac fermions in graphene. Nature 438 197 200 (2005). 17. Zhang Y., Tan, Y. W., Stormer, H. L. & Kim, P. Experimental observation of the quantum Hall effect and Berry's phase in graphene. Nature 438 201 204 (2005). 18. Ferrari, A. et al Raman spectrum of graphene and graphene layers. Phys. Rev. Lett. 97 187401 (2 006). 19. Neto, A. H. C., Guinea, F., Peres, N., Novoselov, K. & Geim, A. The electronic properties of graphene. Reviews of Modern Physics 81 109 (2009). 20. First, P. N. et al Epitaxial graphenes on silicon carbide. MRS Bull 35 296 305 (2010). 21. S aito, R., Fujita, M., Dresselhaus, G. & Dresselhaus, M. S. Electronic structure of chiral graphene tubules. Appl. Phys. Lett. 60 2204 2206 (1992). 22. Saito, R., Dresselhaus, G. & Dresselhaus, M. S. in Physical Properties of Carbon Nanotubes 259 (Imperia l College Press, London, 1998). 23. Fischer, J. E. Chemical doping of single wall carbon nanotubes. Acc. Chem. Res. 35 1079 1086 (2002). 24. Zhou, W. et al Charge transfer and Fermi level shift in p doped single walled carbon nanotubes. Physical Review B 71 205423 (2005). 25. Kong, J. et al Nanotube molecular wires as chemical sensors. Science 287 622 625 (2000). 26. Bockrath, M. et al Chemical doping of individual semiconducting carbon nanotube ropes. Physical Review B 61 10606 10 608 (2000). 27. Tans, S. J., Verschueren, A. R. M. & Dekker, C. Room temperature transistor based on a single carbon nanotube. Nature 393 49 52 (1998). 28. Rinzler, A. G. & Donoghue, E. P. All the Dope on Nanotube Films. ACS Nano 5 3425 3427 (2011). 29. Wadhwa, P., Liu, B., McCarthy, M. A., Wu, Z. & Rinzler, A. G. Electronic Junction Control in a Nanotube Semiconductor Schottky Junction Solar Cell. Nano Letters 10 5001 5005 (2010). 30. Liu, B. et al Carbon Nanotube Enabled Vertical Field Effect and Light Emitting Transistors. Adv Mater 20 3605 3609 (2008).
136 31. Dresselhaus, M., Dresselhaus, G., Charlier, J. C. & Hernandez, E. Electronic, thermal and mechanical properties of carbon nanotubes. Philosophical Transactions of the Royal Society of London .Series A: Mathematical, Physical and Engineering Sciences 362 2065 2098 (2004). 32. Derycke, V., Martel, R., Appenzeller, J. & Avouris, P. Carbon nanotube inter and intramolecular logic gates. Nano Letters 1 453 456 (2001). 33. McEuen, P. L., Fuhrer, M. S. & Park, H. Single walled carbon nanotube electronics. Nanotechnology, IEEE Transactions on 1 78 85 (2002). 34. Misewich, J. et al Electrically induced optical emission from a carbon nanotube FET. Science 300 783 786 (2003). 35. Sazonova, V. et a l A tunable carbon nanotube electromechanical oscillator. Nature 431 284 287 (2004). 36. Qi, P. et al Toward large arrays of multiplex functionalized carbon nanotube sensors for highly sensitive and selective molecular detection. Nano Letters 3 347 35 1 (2003). 37. O'connell, M. J. et al Band gap fluorescence from individual single walled carbon nanotubes. Science 297 593 596 (2002). 38. Wu, Z. et al Transparent, Conductive Carbon Nanotube Films. Science 305 1273 1276 (2004). 39. Lee, K. et al S ingle Wall Carbon Nanotubes for p Type Ohmic Contacts to GaN Light Emitting Diodes. Nano Lett. 4 911 914 (2004). 40. McCarthy, M. A., Liu, B. & Rinzler, A. G. High Current, Low Voltage Carbon Nanotube Enabled Vertical Organic Field Effect Transistors. Na no Letters 10 3467 3472 (2010). 41. McCarthy, M. A. et al Low Voltage, Low Power, Organic Light Emitting Transistors for Active Matrix Displays. Science 332 570 573 (2011). 42. Gruner, G. Carbon nanotube films for transparent and plastic electronics. J. Mater. Chem. 16 3533 3539 (2006). 43. Yu, Z. et al Fully bendable polymer light emitting devices with carbon nanotubes as cathode and anode. Appl. Phys. Lett. 95 203304 3 (2009). 44. Ou, E. C. et al Surface modified nanotube anodes for high perfor mance organic light emitting diode ACS Nano 3 2258 2264 (2009).
137 45. Hu, L., Li, J., Liu, J., Gruner, G. & Marks, T. Flexible organic light emitting diodes with transparent carbon nanotube electrodes: problems and solutions Nanotechnology 21 155202 (2010). 46. Aguirre, C. M. et al Carbon nanotube sheets as electrodes in organic light emitting diodes. Appl. Phys. Lett. 88 183104 (2006). 47. Vasilyeva, S. V. et al Color Purity in Polymer Electrochromic Window Devices on and Single Walled Carbon Nanotube Electrodes. ACS Applied Materials & Interfaces 1 2288 2297 (2009). 48. Drkop, T., Getty, S., Cobas, E. & Fuhrer, M. Extraordinary mobility in semiconducting carbon nanotubes. Nano letters 4 35 39 (2004). 49. Staderman n, M. et al Nanoscale study of conduction through carbon nanotube networks (2004). 50. Li, S., Yu, Z., Rutherglen, C. & Burke, P. J. Electrical properties of 0.4 cm long single walled carbon nanotubes. Nano L ett 4 2003 2007 (2004). 51. Javey, A., Gu o, J., Wang, Q., Lundstrom, M. & Dai, H. Ballistic carbon nanotube field effect transistors. Nature 424 654 657 (2003). 52. Fuhrer, M. et al Crossed nanotube junctions. Science 288 494 497 (2000). 53. Hecht, D., Hu, L. & Grner, G. Conductivity scalin g with bundle length and diameter in single walled carbon nanotube networks. Appl. Phys. Lett. 89 133112 (2006). 54. Wu, C. C., Wu, C. I., Sturm, J. C. & Kahn, A. Surface modification of indium tin oxide by plasma treatment: An effective method to improv e the efficiency, brightness, and reliability of organic light emitting devices. Appl. Phys. Lett. 70 1348 1350 (1997). 55. Das, R. K., Liu, B., Reynolds, J. R. & Rinzler, A. G. Engineered macroporosity in single wall carbon nanotube films. Nano Lett 9 677 683 (2009). 56. Tenent, R. C. et al Ultrasmooth, Large Area, High Uniformity, Conductive Transparent Single Walled Carbon Nanotube Films for Photovoltaics Produced by Ultrasonic Spraying. Adv Mater 21 3210 3216 (2009). 57. Meitl, M. A. et al So lution casting and transfer printing single walled carbon nanotube films. Nano Lett 4 1643 1647 (2004). 58. Hur, S. H., Khang, D. Y., Kocabas, C. & Rogers, J. A. Nanotransfer printing by use of noncovalent surface forces: Applications to thin film transistors that use single walled carbon nanotube networks and semiconducting polymers. Appl. Phys. Lett. 85 5730 5732 (2004).
138 59. Dan, B., Irvin, G. C. & Pasquali, M. Continuous and scalable fabrication of transparent conducting carbon nanotube films. A C S Nano 3 835 843 (2009). 60. Yudasaka, M., Komatsu, T., Ichihashi, T., Achiba, Y. & Iijima, S. Pressure dependence of the structures of carbonaceous deposits formed b y laser ablation on targets composed of carbon, nickel, and cobalt. J. Phys Chem B 102 4892 4896 (1998). 61. Guo, T., Nikolaev, P., Thess, A., Colbert, D. T. & Smalley, R. E. Catalytic growth of single walled manotubes by laser vapor ization. Chem. Phys. Lett. 243 49 54 (1995). 62. Giles, J. Growing nanotech trade hit by questions over quality. Nature 432 791 791 (2004). 63. Li, J. et al Organic l ight e mitting d iodes h aving c arbon n anotube a nodes. Nano Lett. 6 2472 2477 (2006). 64. Ebbesen, T. W., Ajayan, P. M., Hiura, H. & Tanigaki, K. Purification of nanotubes. Nature 367 519 519 (1994). 65. Dujardin, E., Ebbesen, T. W., Krishnan, A. & Treacy, M. M. J. Purification of single shell nanotubes. Adv Mater 10 611 613 (1998). 66. Dillon, A. et al Storage of Hydrogen in Single Walled Carbon Nanotubes. Nature 386 377 379 (1997). 67. Zimmerman, J. L., Bradley, R. K., Huffman, C. B., Hauge, R. H. & Margrave, J. L. Gas phase purification of single wall carbon nanotubes. Chem M ater 12 1361 1366 (2000). 68. Xu, Y. Q., Peng, H., Hauge, R. H. & Smalley, R. E. Controlled multistep purification of single walled carbon nanotubes. Nano Lett. 5 163 168 (2005). 69. Rinzler, A. G. et al Large scale purification of single wall carbon nanotubes: process, product, and characterization. Appl. Phys. A 67 29 37 (1998). 70. Hu, H., Zhao, B., Itkis, M. E. & Haddon, R. C. Nitric acid purification of single walled carbon nanotubes. J. Phys. Chem B 107 13838 13842 (2003). 71. Liu, J. et al Fullerene Pipes. Science 280 1253 1256 (1998). 72. Hu, H. et al Influence of the zeta potential on the dispersability and purification of single walled carbon nanotubes. J. Phys Chem B 109 11520 11524 (2005). 73. Bonard, J. M. et al Purification and size selection of carbon nanotubes. Adv Mater 9 827 831 (1997).
139 74. Arnold, M. S., Green, A. A., Hulvat, J. F., Stupp, S. I. & Hersam, M. C. Sorting carbon nanotubes by electronic structure u sing density differentiation. Nature Nan o. 1 60 65 (2006). 75. Arnold, M. S., Stupp, S. I. & Hersam, M. C. Enrichment of single walled carbon nanotubes by diameter in density gradients. Nano Le tt. 5 713 718 (2005). 76. Ghosh, S., Bachilo, S. M. & Weisman, R. B. Advanced sorting of single walled carbon nanotubes by nonlinear density gradient ultracentrifugation. Nature Nano. 5 443 450 (2010). 77. Lu, K. Mechanical damage of carbon nanotubes by ultrasound. Carbon 34 814 816 (1996). 78. Lucas, A. et al Kinetics of nanotube and microfiber scission under sonication. J. Ph ys Ch em. C 113 20599 20605 (2009). 79. Ahir, S., Huang, Y. & Terentjev, E. Polymers with aligned carbon nanotubes: Active composite materials. P olymer 49 3841 3854 (2008). 80. Anderson, N. G. et al K Series Centrifuges I. Development of the K II continuous sample flow with banding centrifuge system for vaccine purification. Anal. Biochem. 32 460 494 (1969). 81. Iijima, S. Helical microtubules of graphitic carbon. Nature 354 56 58 (1991). 82. Ebbesen, T. W. & Ajayan, P. M. Large scale synthesis of carbon nanotubes. Nature 358 220 222 (1992). 83. Iijima, S. & Ichihashi, T. Single shell carbon nanotubes of 1 nm diameter. Nature 363 603 605 ( 1993). 84. Jose Yacaman, M., Miki Yoshida, M., Rendon, L. & Santiesteban, J. G. Catalytic growth of carbon microtubules with fullerene structure. Appl. Phys. Lett. 62 657 659 (1993). 85. Li, W. Z. et al Large Scale Synthesis of Aligned Carbon Nanotubes Science 274 1701 1703 (1996). 86. Thess, A. et al Crystalline Ropes of Metallic Carbon Nanotubes. Science 273 483 487 (1996). 87. Rinzler, A. G. et al Large scale purification of single wall carbon nanotubes: process, product, and characterization. Appl. Phys. A 67 29 37 (1998).
140 88. Shao, L. et al Removal of amorphous carbon for the efficient sidewall functionalisation of single walled carbon nanotubes. Chem Comm 5090 5092 (2007). 89. Kim, Y. & Luzzi, D. E. Purification of Pulsed Laser Synthesized Single Wall Carbon Nanotubes by Magnetic Filtration. J Phys Chem B 109 16636 16643 (2005). 90. Yu, A. et al Application of Centrifugation to the Large Scale Purification of Electric Arc Produced Single Walled Carbon Nanotubes. J. Am. C hem. Soc. 128 9902 9908 (2006). 91. Anson Casaos, A., Gonzalez Dominguez, J. & Martinez, M. Separation of single walled carbon nanotubes from graphite by centrifugation in a surfactant or in polymer solutions. Carbon 48 2917 2924 (2010). 92. Forsythe, E., Choong, V. E., Le, T. & Gao, Y. Interface analysis of naphthyl substituted benzidine derivative and tris 8 (hydroxyquinoline) aluminum using ultraviolet and x ray photoemission spectroscopy. J. of Vac Sci Techno A 17 3429 (1999). 93. Zschieschang, U. et al Flexible Low Voltage Organic Transistors and Circuits Based on a High Mobility Organic Semiconductor with Good Air Stability. Adv Mater 22 982 985 (2010). 94. Dodabalapur, A., Katz, H., Torsi, L. & Haddon, R. Organic heterostructure field effect transistors. Science 269 1560 1562 (1995). 95. Martel, R., Schmidt, T., Shea, H., Hertel, T. & Avouris, P. Single and multi wall carbon nanotube field effect transistors. Appl. Phys. Lett. 73 2447 (1998). 96. Tans, S. J. et al Individual single wall carbon nanotubes as quantum wires. Nature 386 474 477 (1997). 97. Schottky, W. Halbleitertheorie der sperrschicht. Naturwissenschaften 26 843 843 (1938). 98. Mott, N. Note on the contact between a metal an d an insulator or semi conductor Mathematical Proceedings of the Cambridge Philosophical Society Ser. 34 Cambridge Univ Press, 1938 99. Sze, S. & Ng, K. K. Metal Semiconductor Contacts. Physics of S emiconductor D evices 2 245 311 (1981). 100. Lonerga n, M. C. A tunable diode based on an inorganic semiconductor / conjugated polymer interface. Science 278 2103 (1997). 101. Ma, L. Unique architecture and concept for high performance organic transistors. Appl. Phys. Lett. 85 5084 3 (2004).
141 102. Li, S., Xu, Z., Yang, G., Ma, L. & Yang, Y. Solution processed poly(3 hexylthiophene) vertical organic transistor. Appl. Phys. Lett. 93 213301 3 (2008). 103. Ben Sasson, A. J. et al Patterned electrode vertical field effect transistor fabricated using block cop olymer nanotemplates. Appl. Phys. Lett. 95 213301 (2009). 104. Ben Sasson, A. J. & Tessler, N. Patterned electrode vertical field effect transistor: Theory and experiment. J Appl Ph y s. 110 044501 044501 12 (2011). 105. Ben Sasson, A. J. & Tessler, N. Patterned electrode vertical OFET: analytical description, switching mechanisms, and optimization rules Proc SPIE 8117 ( 2011). 106. Ben Sasson, A. J., Chen, Z., Facchetti, A. & Tessler, N. Solution processed ambipolar vertical organic field effect transistor. Appl. Phys. Lett. 100 263306 (2012). 107. Chua, L. L., Ho, P. K. H., Sirringhaus, H. & Friend, R. H. High stability ultrathin spin on benzocyclobutene gate dielectric for polymer field effect transistors. Appl. Phys. Lett 84 3400 (2004). 108. Sirringhaus, H. et al Two dimensional charge transport in self organized, high mobility conjugated polymers. Nature 401 685 688 (1999). 109. Liu, B. Car b on Na n otube E nabled Verti cal Field Eff ect T rans istors a n d thei r Deri v ati v e s Thesi s Un i versit y of F l orida (2010). Extended Heteroarenes, Dinaphtho [2, 3 b: 2', 3' f] chalcogenopheno [3, 2 b] chalcogenophenes, and Their Application to Field Effect Transistors. J. Am. Chem. Soc. 129 2224 22 25 (2007). 111. McCarthy, M. A. et al Reorientation of the h igh m obility p lane in p entacene b ased c arbon n anotube e nabled v ertical f ield e ffect t ransistors. ACS Nano 5 291 298 (2011). 112. Cheng, Z. et al Toward i ntrinsic g raphene s urfaces: A s ystemat ic s tudy on t hermal a nnealing and w et c hemical t reatment of Si O2 s upported g raphene d evices. Nano L ett (2011). 113. Sharma, R., Baik, J. H., Perera, C. J. & Strano, M. S. Anomalously large reactivity of single graphene layers and edges toward electron transfer chemistries. Nano Le tt. 10 398 405 (2010). 114. Pimenta, M. et al Studying disorder in graphite based systems by Raman spectroscopy. Phys.Chem.Chem.Phys. 9 1276 1290 (2007).
142 115. Ferrari, A. C. Raman spectroscopy of graphene and graphite: D isorder, electron phonon coupling, doping and nonadiabatic effects. Solid State Commun. 143 47 57 (2007). 116. Wang, Y. et al Raman studies of monolayer graphene: the substrate effect. J. Phys Chem. C 112 10637 10640 (2008). 117. Yu, Y. et al Tuning the g raphene w ork f unction by e lectric f ield e ffect. Nano Lett. 9 3430 3434 (2009). 118. Liu, B., McCart hy, M. A. & Rinzler, A. G. Non v olatile o rganic m emory e lements b ased on c arbon n anotube e nabled v ertical f ield e ffect t rans istors. Adv. Funct. Mater. 20 3440 3445 (2010). 119. Kohno, A., Sameshima, T., Sano, N., Sekiya, M. & Hara, M. High performance poly Si TFTs fabricated using pulsed laser annealing and remote plasma CVD with low temperature processing. IEEE T rans. O n Elec. Dev 42 251 257 (1995). 120. Bao, Z. Soluble and processable regioregular poly(3 hexy lthiophene) for thin film field effect transistor applications with high mobility. Appl. Phys. Lett. 69 4108 3 (1996). 121. Yoon, W. & Berger, P. R. Atomic layer deposited HfO2 gate dielectrics for low voltage operating, high performance poly (3 hexythiophene) organic thin film transistors. Org Elec 11 1719 1722 (2010). 122. Wang, G. et al Poly(3 hexylthiophene) field effe ct transistors with high dielectric constant gate insulator. J. Appl. Phys. 95 316 7 (2004). 123. Majewski, L., Schroeder, R., Grell, M., Glarvey, P. & Turner, M. High capacitance organic field effect transistors with modified gate insulator surface. J. Appl. Phys. 96 5781 7 (2004). 124. Bartic, C., Jansen, H., Campitelli, A. & Borghs, S. Ta2O5 as gate dielectric material for low voltage organic thin film transistors. Org Elec 3 65 72 (2002). 125. Liu, Y. R., Deng, L. F., Yao, R. H. & Lai, P. T. Low Operating Voltage Polymer Thin Film Transistors Based on Poly(3 Hexylthiophene) With Hafnium Oxide as the Gate Dielectric IEEE T r ans D e v M a ter R el. 10 233 (2010). 126. Chao, Y., Lin, Y., Dai, M., Zan, H. & Meng, H. R educed hole injection barrier for achieving ultralow voltage polymer space charge limited transistor with a high on/off current ratio. Appl. Phys. Lett. 95 203305 3 (2009). 127. Zen, A. et al Effect of Molecular Weight and Annealing of Poly(3 hexylthiop hene)s on the Performance of Organic Field Effect Transistors. Adv. Funct. Mater. 14 757 764 (2004).
143 128. Wang, G., Swensen, J., Moses, D. & Heeger, A. J. Increased mobility from regioregular poly(3 hexylthiophene) field effect transistors. J. Appl. Phys 93 6137 6141 (2003). 129. Kline, R. J. et al Dependence of Regioregular Poly(3 hexylthiophene) Film Morphology and Field Effect Mobility on Molecular Weight. Macromol 38 3312 3319 (2005). 130. Onoda, M., Tada, K., Zakhidov, A. A. & Yoshino, K. Photoinduced charge separation in photovoltaic cell with heterojunction of p and n type conjugated polymers. Thin Solid Films 331 76 81 (1998). 131. Tanase, C., Meijer, E., Blom, P. & De Leeuw, D. Unification of the hole transport in polymeric field ef fect transistors and light emitting diodes. Phys. Rev. Lett. 91 216601 (2003). 132. Ohno, T. R. et al C 60 bonding and energy level alignment on metal and semiconductor surfaces. Phys. Rev. B 44 13747 13755 (1991). 133. Tsumura, A., Koezuka, H. & An do, T. Polythiophene field effect transistor: Its characteristics and operation mechanism. Synth. Met. 25 11 23 (1988). 134. Huber, E. & Kirk, C. Work function changes due to the chemisorption of water and oxygen on aluminum. Surf. Sci. 5 447 465 (1966) 135. Bernanose, A., Comte, M. & Vouaux, P. A new method of emission of light by certain organic compounds. J. Chem. Phys. 50 64 68 (1953). 136. Kallmann, H. & Pope, M. Positive hole injection into organic crystals. J. Chem. Phys. 32 300 (1960). 137. Tang, C. W. & VanSlyke, S. A. Organic electroluminescent diodes. Appl. Phys. Lett. 51 913 915 (1987). 138. Kajikawa, Y. & Takeda, Y. Citation network analysis of organic LEDs. Tech Fore So c. C hange 76 1115 1123 (2009). 139. Li, Z R. & Meng, H. Organic Light Emitting Materials and Devices. 672 (2007). 140. Ashcroft, N. W. & Mermin, N. D. in Solid State Physics 826 (Brooks/Cole, United States, 1976). 141. Fowler, R. H. & Nordheim, L. Electron Emission in Intense Eletric Fields. Proc. R. So. London, Ser. A 119 173 (1928).
144 142. Parker, I. D. Carrier tunneling and device characteristics in polymer light emitting diodes. J. Appl. Phys. 75 1656 1666 (1994). 143. Braun, S., Salaneck, W. R. & Fahlman, M. Energy Level Alignment at O rganic/Metal and Organic/Organic Interfaces. Adv Mater 21 1450 1472 (2009). 144. Xu, F. et al Single walled carbon nanotube anodes based high performance organic light emitting diodes with enhanced contrast ratio. Org Elec (2011). 145. Kumar, A. & Zhou, C. The race to replace tin doped indium oxide: which material will win? ACS N ano 4 11 14 (2010). 146. Gustafsson, G. et al Flexible light emitting diodes made from soluble conducting polymers. Nature 357 477 479 (1992). 147. Carter, S. A., Angelopoulos, M., Karg, S., Brock, P. J. & Scott, J. C. Polymeric anodes for improved polymer light emitting diode performance. Appl. Phys. Lett. 70 2067 2069 (1997). 148. Scott, J. C., Carter, S. A., Karg, S. & Angelopoulos, M. Polym eric anodes for organic light emitting diodes. Synth. Met. 85 1197 1200 (1997). 149. Kim, W. H. et al Molecular organic light emitting diodes using highly conducting polymers as anodes. Appl. Phys. Lett. 80 3844 3846 (2002). 150. Zhang, D. et al Tran sparent, Conductive, and Flexible Carbon Nanotube Films and Their Application in Organic Light Emitting Diodes. Nano Lett. 6 1880 1886 (2006). 151. Williams, C. D. et al Multiwalled carbon nanotube sheets as transparent electrodes in high brightness org anic light emitting diodes Appl. Phys. Lett. 93 183506 (2008). 152. Chien, Y., Lefevre, F., Shih, I. & Izquierdo, R. A solution processed top emission OLED with transparent carbon nanotube electrodes Nanotec h 21 134020 (2010). 153. Li, J. et al Indium tin oxide modified transparent nanotube thin films as effective anodes for flexible organic light emitting diodes Appl. Phys. Lett. 93 083306 (2008). 154. Cao, Y., Yu, G., Zhang, C., Menon, R. & Heeger, A. J. Polymer light emitting diodes with pol yethylene dioxythiophene polystyrene sulfonate as the transparent anode. Synth. Met. 87 171 174 (1997). 155. Bharathan, J. M. & Yang, Y. Polymer/metal interfaces and the performance of polymer light emitting diodes. J. Appl. Phys. 84 3207 3211 (1998).
145 156. Brown, T. et al Built in field electroabsorption spectroscopy of polymer light emitting diodes incorporating a doped poly (3, 4 ethylene dioxythiophene) hole injection layer. Appl. Phys. Lett. 75 1679 (1999). 157. Braun, D. & Heeger, A. J. Visible light emission from semiconducting polymer diodes. Appl. Phys. Lett. 58 1982 1984 (1991). 158. Campbell, I., Hagler, T., Smith, D. & Ferraris, J. Direct measurement of conjugated polymer electronic excitation energies using metal/polymer/metal structures Phys. Rev. Lett. 76 1900 1903 (1996). 159. Shi, Y., Liu, J. & Yang, Y. Device performance and polymer morphology in polymer light emitting diodes: The control of thin film morphology and device quantum efficiency. J. Appl. Phys. 87 4254 4263 (2000). 160. Lee, T. W., Park, O. O., Do, L. M. & Zyung, T. Improvement of EL efficiency in polymer light emitting diodes by heat treatments. Synth. Met. 117 249 251 (2001). 161. Pei, Q., Yu, G., Zhang, C., Yang, Y. & Heeger, A. J. Polymer Light Emitting Electro chemical Cells. Science 269 1086 1088 (1995). 162. Pei, Q., Yang, Y., Yu, G., Zhang, C. & Heeger, A. J. Polymer light emitting electrochemical cells: In situ formation of a light emitting pn junction. J. Am. Chem. Soc. 118 3922 3929 (1996). 163. Leger, J. M. Electrochemical doping and the optical properties of light emitting polymer materials and devices. Dissertation (2005). 164. Dyer, A. L. Conjugated polymer electrochromic and light emitting devices D isse r tation, Univ er s ity o f Flor ida. 2007 165. Fang, J., Matyba, P. & Edman, L. The d esign and r ealization of f lexible, l ong l ived l ight e mitting e lectrochemical c ells. Adv Funct Mater 19 2671 2676 (2009). 166. Shin, J. H., Matyba, P., Robinson, N. D. & Edman, L. The influence of electrodes on the performance of light emitting electrochemical cells. Electrochim. Acta 52 6456 6462 (2007). 167. Herrero, J. & Guillen, C. Transparent films on polymers for photovoltaic applications. Vacuum 67 611 616 (2002). et al A surface emitting vacuum deposited or ganic light emitting device. Appl. Phys. Lett. 70 2954 (1997).
146 169. Saran, N. et al Fabrication and c haracterization of t hin f ilms of s ingle w alled c arbon n anotube b undles on f lexible p lastic s ubstrates. J. Am. Chem. Soc. 126 4462 4463 (2004). 170. Sk kalov, V., Kaiser, A. B., Woo, Y. & Roth, S. Electronic transport in carbon nanotubes: From individual nanotubes to thin and thick networks. Phys. Rev. B 74 085403 (2006). 171. Dick, D. J., Heeger, A. J., Yang, Y. & Pei, Q. Imaging the structure of the p n junction in polymer light emitting electrochemical cells. Adv Mater 8 985 987 (1996). 172. Matyba, P. et al Graphene and mobile ions: the key to all plastic, solution processed light emitting devices. ACS N ano 4 637 642 (2010). 173. Matyba, P., Yamaguchi, H., Chhowalla, M., Robinson, N. D. & Edman, L. Flexible and metal free light emitting electrochemical cells based on graphene and PEDOT PSS as the electrode materials. ACS N ano (2011). 174. Robinson, N. D., Edman, L. & Chhowalla, M. Graphene electrodes for organic metal free light emitting devices. Phys. Scripta 2012 014023 (2012). 175. Yu, Z., Niu, X., L iu, Z. & Pei, Q. Intrinsically s tretchable p olymer l ight em itting d evices u sing c arbon n anotube p olymer c omposite e l ectrodes. Adv Mater (2011). 176. Chen, R. J. et al Noncovalent functionalization of carbon nanotubes for highly specific electronic biosensors. Proc Nat Acad Sci 100 4984 (2003). 177. Vaisman, L., Wagner, H. D. & Marom, G. The role of surfactants in dispersion of carbon nanotubes. Adv. Colloid Interface Sci. 128 37 46 (2006).
147 BIOGRAPHICAL SKETCH Evan Peter Donoghue was born on July 21 st 1984 in Amherst, Massachusetts, the youngest sibling to two older sisters. With a physicist for a father and a librarian for a mother, he quickly became enamored with science and was bestowed with a love of family on vacatio n across the globe in conjunction with conferences and sabbaticals. As he watched his father working on his theoretical calculations while relaxing a beach in the south of France, Evan realized that a career in physics might be an enjoyable lifestyle. It was not until years later that he learned the catch: while theoretical physicists can bring their work to the beach, experimental physicists spend much of their day working in a windowless basement laboratory. Graduating from high aerospace engineering, he decided that his calling was in physics. Though his school year was often dominated by his efforts o n the Notre Dame Rowing Team, he spent his summers doing research both at the University of Massachusetts and the Jefferson National Accelerator Facility through the College of William and Mary. These efforts resulted in three publications and the opportu nity to present his work at the Conference on RF Superconducting Cavities. After graduating with a Bachelor of Science in Physics from the University of Notre Dame in 2006, Evan the University of Florida where he saw the opportunity to connect his passion for physics with real world applications. He began work in June of 2006, the summer before enrolling in his first year of studies and was quickly integrated into projects in collaboration with John
148 oup in the Department of Chemistry and began to prob e the effects of doping o n thin films of carbon nanotubes This work led to the efforts in organic electronics and carbo n nanotube purification that are seen here. Upon g r a duat i o n h e will j o i n e M a gin C orp o r a t i o n to f u r t her t h i s p u rsuit and co n t r ibute to the devel opme n t of O LED d ispla y s for wid e ranging applica t i o ns.