1 U LTRA THIN, HIGHLY UNIFORM, SPRA YED SINGLE WALL CARBON NANOTUBE FILMS FOR VERTICAL FIELD EFFECT TRANSISTORS By YU SHEN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2014
2 Â© 2014 Yu Shen
3 To my parents
4 ACKNOWLEDGMENTS I owe deepest gratitude to my dear advisor Prof. Andrew Rinzler, for his support and guidance during the past few years. His knowledge in both theoretical and experimental knowledge of physics research is comprehensive and inexhaustible . He is always present to help students with their issues, ranging from fundamental theoretical problem to equipment set up. I learnt knowledge and skills from him during conversations or discussion wit h him. The most important thing he taught me is the enthusiasm to learn from unknown and belief in seeking the truth from the facts. He dedicated his time and passion to what he truly believes in, which constantly encourages me in my research career. Besid es, he is open up to new ideas and always encourages us to develop our new adventures. I am truly grateful as one of his students. I want to thank my coworker Svetlana V. Vasilyeva . She is well knowledgeable and talented in both physical and chemical rese arch. I appreciate that she is always generous to share her ideas and skills. I am very thankful to her for her care and help during my thesis preparation. I also want to thank Dr. Mitchell McCarthy and Dr. Bo Liu , from whom I learnt so much lab skills and research methods. I am thankful to Xiao Chen and Stephen Matt Gilbert for their cooperation. I want to thank my other group members Dr. Maureen K. Petterson, Dr. Rajib Das, Dr. Max Lemaitre, Dr. Pooja Wadhwa, Dr. Evan Donoghue, Dr. Po Hsiang Wang, Nan Zha o, and Jie Hou . I learnt a lot of research ideas and fundamental knowledge from discussion with my excellent colleges. I owe great thanks to Prof. John Reynold s and his group for their help and collaborations. I am gratefu l to Prof. Weihong Tan and his student Cuichen Wu for being so fri endly and great collaborations . I want to thank my committee members for ongoing
5 support. Prof. Franky So was so kind and a great source of help. I am thankful for Prof. Art Hebard for being so friendly and alway s ready to assist. I want to thank Prof. Selman Hershfield and Prof . David Tanner for being so kind and helpful. I am very grateful to have them as my counselor over the past few years. I also want to thank Ms. Darlene Latimer and Mr. Eugene H atley for ta king care of my lab affairs. I owe great thanks to Ms. Pam Marlin for her help in wading through the academic bureaucracy on my behalf and ensuring I was on the right track to graduate. I am grateful to physics machine shop personnel and Nanoscale Research Facility researchers for their great help. I want to show great thanks to my beloved f amily. Thank you, my wonderful d ad, my beautiful m om and my d ear g randpa, my g randma. You are always the source of endless love and support that encourages me for my work and life. I am extremely grateful and proud of be one of this family. I also want to give special thanks to my dear boyfriend, Fei Weng. He brought me joy and love during my thesis preparation , and has always been encouraging and supportive.
6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 9 LIST OF FIGURES ................................ ................................ ................................ ........ 10 LIST OF ABBREVIATIONS ................................ ................................ ........................... 13 ABSTRACT ................................ ................................ ................................ ................... 15 CHAPTER 1 INTRODUCTION TO SINGLE WALL CARBON NANOTUBES .............................. 17 1.1 Fundamentals ................................ ................................ ................................ ... 17 1.2 SWNTs Synthesis ................................ ................................ ............................. 20 1.3 SWNTs Film Fabrication: Vacuum Filtration Method ................................ ........ 20 1.4 Other Solution based SWNTs Film Deposition Methods ................................ .. 21 2 INTRODUCTION TO SINGLE WALL CARBON NANOTUBE DISPERSI ON (SWNT INK) ................................ ................................ ................................ ............ 25 2.1SWNTs Ink Preparation Methods: Nanotube Modification ................................ . 25 2.2 Surfactant Functionalized SWNTs ................................ ................................ .... 27 2.2.1 Fundamentals of Surfactant ................................ ................................ .... 27 2.2.2 Surfactant functionalized SWNTs ................................ ............................ 28 2.3 Aromatic Compounds Pyrene Derivatives Functionalized SWNTs Inks ........... 30 2.3.1 Introduction to Pyrene Derivatives ................................ ........................... 30 220.127.116.11 Pyrene functionalized hydroxypropyl cellulose: Py HPC ................ 31 18.104.22.168 Small molecular pyrene derivatives ................................ ............... 33 2.3.2 Pyrene Fluorescence ................................ ................................ ............... 36 22.214.171.124 Introduction to fluorescence measurement ................................ .... 36 126.96.36.199 Fundamentals of fluorescence spectrum: emission and excitation ................................ ................................ ................................ . 37 188.8.131.52 Typical pyrene fluorescence spectr um ................................ ........... 41 3 SINGLE WALL CARBON NANOTUBE INK PREPARATION AND CHARACTERIZATION ................................ ................................ ........................... 43 3.1 Py HPC/SWNTs Ink ................................ ................................ ........................... 43 3.1.1 Py HPC /SWNTs ink Preparation ................................ ............................. 43 3.1.2 Py HPC/SWNTs Ink Characterization: UV vis Absorbance ..................... 45 3.2 PTSA/SWNTs Ink ................................ ................................ ............................. 47 3.2.1 PTSA/SWN Ts Ink Preparation ................................ ................................ 47
7 3.2.2 PTSA/SWNTs Ink Characterization ................................ ......................... 49 3.2.3 Optimizationof the PTSA/SWNTs Ink Preparation Procedure Using Fluorescence Measurements ................................ ................................ ........ 50 184.108.40.206 Typical fluorescence spectrum of PTSA in water ........................... 50 220.127.116.11 TX induced enhanced fluorescence of PTSA ................................ 51 18.104.22.168 Minimizingthe PTSA to SWNT weight ratio ................................ .... 52 22.214.171.124 Minimizing PTSA / SWNTs association time ................................ .... 57 4 ULTRASONIC SPRAY COATING ................................ ................................ .......... 62 4.1Ult rasonic Spray Coating and Home made Spray Station ................................ . 62 4.2 Spray from Py HPC/SWNTs ink ................................ ................................ ....... 65 4.2.1 Py HPC/SWNTs Ink Dilution ................................ ................................ ... 65 4.2.2 Post treatment of Sprayed Py HPC/SWNTs Films ................................ .. 68 4.2.3 Comparison between Sprayed Films and Filter transferred Films ........... 70 4.2.4 Homogeneity Comparisonof Sprayedand Filter transferred Films ........... 72 126.96.36.199 Homogeneity measurement set up ................................ ................ 72 188.8.131.52 Homogeneity discussion ................................ ................................ 74 4. 3 Error Analysis of Ultrasonic Spraying and Filter Transferring Process ............. 76 4.3.1 Calibration Curve: C SWNTs vs. Absorbance ................................ .............. 76 4.3.2 Spray depostion material losses ................................ .............................. 77 4.4 Spray coating from PTSA/SWNT ink ................................ ................................ 83 4.4.1 Sprayed Films from PTSA/SWNT ink ................................ ...................... 83 4.4.2 Post treatment ................................ ................................ ......................... 85 4.4.3 Homogeneity Test ................................ ................................ ................... 87 5 SPRAYED FILMS FOR VERTICAL FIELD EFFECT TRANSISTORS .................... 90 5 .1 Fundamentals ................................ ................................ ................................ ... 90 5.1.1 Introduction to Vertical Field Effect Transistor ................................ ......... 90 5 . 1. 2 Working Mechanism of VFETs ................................ ................................ 92 5.1.3Fabrication Process of Organic VFETs ................................ .................... 96 5.1.4 Typical VFET Curves ................................ ................................ ............... 98 5.1.5 Effect of different VFET parameters ................................ ...................... 101 184.108.40.206 Nanotube bundle effect ................................ ................................ 101 220.127.116.11 HOMO level of semiconducting channel and work function of electrodes ................................ ................................ ............................. 102 18.104.22.168 Surface treatment on dielectric layer ................................ ............ 103 5.2 VFETs with Ultrasonic Sprayed Py HPC/SWNTs Films ................................ ... 105 5.3 VFETsOptimization ................................ ................................ ......................... 108 5.3.1 CNT Film Cleaness and Optimization of Gate/Gate Dielectric Layer .... 108 5.3 .2 Sprayed Films with No Harsh Treatment PTSA/SWNTs Films ............. 111 6 CONCLUSIONS AND FUTURE WORK ................................ ............................... 115
8 REFERENCES ................................ ................................ ................................ ............ 117 BIOGRAPHIC SKETCH ................................ ................................ .............................. 126
9 LIST OF TABLES Table page 3 1 Quenching rate R quenching emission = 383 excitation = 348 nm ... 56 3 2 Decay constant s from fitting of the kinetic curves of the inks with different PTSA:SWNTs weight ratios ................................ ................................ ................ 59 4 1 Sheet resistance of the SWNT films sprayed the inks dilutedwith water and ethanol. ................................ ................................ ................................ ............... 68 4 2 S tatistics of nanotube pixel resistancefrom s prayed and filter transferred films with the same size of 21.82 mm x 19.28 mm . ................................ ..................... 74 4 3 Sheet resistance comparison between sprayed and filter transferred films with corresponding effective thickness. ................................ .............................. 83 4 4 Effect of the ethanol percentage in the PTSA based inks on the conductivity of the sprayed films. ................................ ................................ ........................... 84 4 5 Homogeneity s tatistics of R s of 2 block area nanotube pixels for sprayed and transferred films . ................................ ................................ ................................ . 89 5 1 Statistics of 9 out of 10 measured devices with transferred CNT and 8 out of 10 measured devices with sprayed CNTs. ................................ ....................... 107 5 2 Uniformity Statistics for 9 out of 10 pixels in transferred CNT VFETs and 10 out of 10 pixels in sprayed CNT VFETs. ................................ .......................... 114
10 LIST OF FIGURES Figure page 1 1 The density of states of semiconducting (10,0) single wall carbon nanotube (solid line) and graphene (dotted line) as a function of energy dispersion. ......... 19 2 1 Chemical structure of benzene ring and carbon nanotube sidewall. ................... 26 2 2 Micelle structure of surfactant in aqueous solution. ................................ ............ 28 2 3 Self organization structures of surfactant in surfactant/nanotube suspensions, divided into three association structure. ................................ ........ 29 2 4 Schematic representation of how three most widely studied surfactants associate with nanotube surface. ................................ ................................ ....... 30 2 5 Chemical structure of pyrene a flat aromatic system formed by four neighboring benzene rings. ................................ ................................ ................ 31 2 6 Py HPC/SWNT ink preparation.. ................................ ................................ ........ 33 2 7 Chemical structure of some small molecular pyrene derivativesstudied for CNT dispersion . ................................ ................................ ................................ .. 35 2 8 Schematic representation of Fluorescence Measurement Syste m ..................... 37 2 9 Energy level diagram showing why structure is seen in the absorption and emission spectrum ................................ ................................ .............................. 39 2 10 Mirror image of the emission (black line) and absorption spectra of the PTSA excitation = 348 nm. ................................ 40 2 11 Fluorescence spectrum of pyrene.. ................................ ................................ .... 42 3 1 Py HPC/SWNTs ink preparation process. ................................ .......................... 44 3 2 UV vis spectrum for Py HPC/SWNTs ink characterization. ................................ 46 3 3 UV Vis spectra of the PTSA/SWNTs/TX mixture before ( black line ) and after (red line) 48 h of dialysis . . ................................ ................................ ................... 49 3 4 Fluorescence spectra of PTSA in water. ................................ ............................ 51 3 5 Emission spectra of PTSA/TX (black line) and PTSA (red line) solutions in water. . ................................ ................................ ................................ ................. 52 3 6 P hotograph of the PTSA/SWNTs /TX dispersions with various PTSA:SWNTs weight ratio values ranging from (1 : 10) to (1 : 450). ................................ ......... 54
11 3 7 Fluorescence Quenching in PTSA/SWNT/TX mixture. . ................................ ...... 55 3 8 Fluorescence Measurement to investigate association kinetcs of PTSA/SWNTs. . ................................ ................................ ................................ ... 58 3 9 Decay curves smoothened and normalized for solutions with PTSA : SWNT ranging from 1:10 to 1:450. ................................ ................................ ................ 60 3 10 Fluorescence emission spectra of the PTSA/SWNT/TX mixture recorded on day 1 after 1.5 h of emission peak intensity decay measurement. ..................... 61 4 1 Ultrasonic spray station system.. ................................ ................................ ........ 63 4 2 Optical microscope images (x5 magnification) and AFM images of the sprayed SWNT films from diluted Py HPC based SWNT inks.. .......................... 67 4 3 Optical microscope images (x5 magnification) and AFM images of the sprayed Py HPC/SWNTs films.. ................................ ................................ ......... 70 4 4 AFM images of the SWNT films fabricated using different deposition processes.. ................................ ................................ ................................ ......... 71 4 5 Substrate array for the SWNT film homogeneity measurement. ......................... 73 4 6 Homogeneity comparison between sprayed and filiter transferred films. ........... 75 4 7 UV vis spectrum of diluted SWNT/1%TX solutions with different nano tubes and TX concentrations. ................................ ................................ ...... 76 4 8 L and C swnts, fitting equation ................................ ................................ ................... 77 4 9 UV Vis transmittance spectrum of filter transferred films (solid line) and sprayed film s (dashed line) . . ................................ ................................ ............... 80 4 10 Loss percentage during spray coating deposition as a function of effective thickness of filter transferred film. ................................ ................................ ....... 8 1 4 11 Sheet resistance of the sprayed (black line) and filter transferred (red line) films as a function of target thickness . ................................ ................................ 82 4 12 AFM images of the sprayed and filter transferred films before and after treatment.. ................................ ................................ ................................ .......... 86 4 13 Percent transmittance spectra of the acid treated, ethanol treated and filter transferred films, all showing T% ~ 97% at wavelength 500 nm. ................... 87 4 14 E lectrode array in 2 block area, 144 electrodes (black region) forms into 128 electrode pairs. ................................ ................................ ................................ ... 88
12 4 15 Homogeneity comparison between sprayed and filter transferred films. ............ 89 5 1 Schematic representations of transistors. . ................................ .......................... 91 5 2 Structure of a carbon nanotube enabled VFET. V g and V d are applied with respect to the grounded source. . ................................ ................................ ........ 93 5 3 Schottky barrier formation between a metal and a p type semiconductor. ......... 94 5 4 Schottky barrier between one intrinsic point on CNTs and semiconductor layer . ................................ ................................ ................................ ................... 96 5 5 Structure of Organic VFET. . ................................ ................................ ............... 97 5 6 Typical curves of VFET . ................................ ................................ ................... 100 5 7 . .................... 102 5 8 Transfer curves of VFETs with different dielectric surface treatment layer. ...... 104 5 9 Comparison of VFETs performance between sprayed Py HPC/SWNT films as source electrode and traditional tran sferred CNTs as source electrode.. .... 106 5 1 0 Performance comparison of three VFETs. ................................ ....................... 110 5 1 1 10 random pixels in the best performed device area from both transferred CNT VFETs and sprayed CNT VFETs were measured.. ................................ . 114
13 LIST OF ABBREVIATIONS A FM Atomic Force Microscopic AMOLED active matrix organic light emitting diode BCB Benzocyclobutene CMC Critical micelle concentration CNT Carbon nanotube CVD C hemical vapor deposition DCMD Double critical micelle concentration D t Target thickness D m Measured thickness DOS Density of states HOMO Highest occupied molecular orbital HPC Hydroxypropyl c ellulose I d Source drain current I g Gate source current J max Maximum current density LB Langmuir Blodgett NaDDBS S odium dodecylbenzenesulfonate NPD di(1 naphthyl) N,N diphenyl 1,1 0 diphenyl 1,4 0 diamine) PBA 1 pyrenebutyric acid PCA 1 pyrenecarboxylic acid PLV P ulsed laser vaporization PMA 1 pyrenemethylamine PTSA 1,3,6,8 Pyrenetetrasulfonic Acid Tetrasodium Salt Py HPC P yrenederivatized h ydroxypropyl c ellulose
14 R s Sheet resistance RSD Relative standard deviation SD Standard d eviation SDS S odium dodecyl sulfate SWNT Single wall carbon nanotube TX Triton X V d Drain voltage V g Gate voltage V s Source voltage V ss Subthreshold voltage V th Threshold voltage VFET Vertical field effect transistor
15 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 U LTRA THIN, HIGHLY UNIFORM, SPRA YED SINGLE WALL CARBON NANOTUBE FILMS FOR VERTICAL FIELD EFFECT TRANSISTORS By Yu Shen August 201 4 Chair: Andrew G. Rinzler Major: Physics The operating mechanism of the carbon nanotube enabled vertical field effect transistor (CN VFET) requires gate field access to the channel layer without too much screening from the intervening single wall carbon nanotube source electrode layer. Hence, whi le the nanotube surface density can be well above the electrical percolation threshold the layer must still be rather dilute (sub monolayer). For large area arrays of devices this places more severe constraints on the nanotube deposition method in terms of attaining homogeneous films over the entire array. To demonstrate nanotube layers exhibiting the requisite homogeneity I have explor ed u ltrasonic s pray deposition of the single wall carbon nanotubes in a home built spray station from single wall carbon na notube ink. To spray coat the generally insoluble nanotubes from solution the use of two distinct types of dispersants was developed. In contrast to surfactant based dispersions which require a large excess of surfactant for metastable suspension of the na notubes, each dispersant used here was to associate to the nanotubes with sufficient binding energy to permit elimination of the excess dispersant before the spray deposition. In this way there would be minimum dispersant to remove, post deposition, to yie ld the high
16 conductivity necessary for the devices to function well. In both cases the strong association of the dispersant with the nanotubes was via non covalent stacking of pyrene molecules to the nanotube sidewalls. One dispersant consisted of the py renes attached via flexible linkers to a hydroxypropyl cellulose polymer backbone. The other dispersant consisted of tetrasod ium pyrenetetrasulfonic acid . The resulting i nk s minimized the amount of dispersant reducing the necessary post treatment of ultras onic sprayed films, providing good stability. UV vis spectra l absorbance and fluorescence measurement s w ere used to optimize the ink recipe and simplify ink preparation. Film quality was characterized by atomic force microscopy , sheet resistance , UV Vis spectroscopy, a specially developed large r area uniformity test that samples VFETs . In all cases comparison wa s made between spray fabricated and the previously developed filtration b ased nanotube film fabricat ion methods . The results demonstrate a scalable method for film formation for such devices.
17 CHAPTER 1 INTRODUCTION TO SINGLE WALL CARBON NANOTUBES 1.1 Fundamentals Carbon nanotubes (CNTs) have been widely investigated because of their unique mechanical, thermal and electrical properties, including extremely high tensile strength [1, 2] , high thermal and electron conductivity [3, 4] . Extensive research has been done to study fundamental properties and incorporate carbon nanotubes int o various devices ranging from prosthetics [5, 6] to microelectronics, including thin film transistors [7 9] , nonvolatile memories [10, 11] and rechargeable batteries [1, 12, 13] . As a potential substitution for transparent conductor indium tin oxide (ITO) films that is used in virtually all commercial electro optic devices ranging from solar cells to displays, CNTs films are preferred because of their lower cost and higher flexibility  . Annual production capacity exceeding several thousand tons per year indicates the worldwide interest in this unique material . Single wall carbon nanotubes are categorized into three types: armchair, zigzag and chiral nanotubes, depending on the chiral or roll up vector C h (n,m) (m
18 at room temperature with high electrical conductivity [16 18] , and all other SWNTs are semiconducting with energy bandages that depend approximately inversely on the tube diameter  . Th ese rules lead to an overall 2:1 ratio of semiconducting to metallic nanotubes over all allowed n, m values. Diameters of SWNTs are typically between 0.8 to 2 nm with tube length ranges from less than 100 nm to a few centime ters, resulting in an extremely high aspect ratio. A carbon atom in SWNTs form the sp 2 bonds with three nearest neighboring atoms, forming 1.44Ã… long C=C bond s and resulting hexagonal lattice structure similar with graphene. This structure similarity results in their closely related electronic properties  . Tight b i nding theory calculates the el ectronic band structure of single layer graphene with and bands touching at six Dirac points on the boundary of the first Brillouin zone, resulting in a semi metallic (zero bandgap semiconductor) band structure for two dimensional graphene. For carbon nanotubes, because of their confinement and periodic boundary condition along circumferential direction and thus the restriction of wave vector circumferentially, continuous 2 D energy dispersion surfaces of graphe n e are cut by a set of parallel energy ba nds with corresponding reciprocal lattice vectors, therefore inducing the Van Hove Singularities observed in density of state (DOS) as shown in Figure 1 1.
19 F igure 1 1. The density of states of semiconducting (10,0) single wall carbon nanotube (solid li ne) and graphene (dotted line) as a function of energy disper sion. No density of states was near Fermi level (Ef=0 eV). Reprinted with permission from Riichiro Saito, Gene Dresselhaus, Mildred S Dresselhaus, Physical properties of Carbon Nanotubes (World Sc ientific Publishing Co, Inc.) R esearch has demonstrated that the Fermi level of carbon nantubes can easily be manipulated based on their relatively low density of states through either chemical charge transfer doping or electrical field effect gating [20 23] . For example, strong Bron st ed acids, namely H 2 SO 4 and HNO 3 shift the Fermi level of SWNTs towards the valence band by acceptor doping (p doping) . The so called intercalation , an energetically driven process here involves a negative charge transfer from the SWNTs to the intercalated species [24, 25] . Adsorption of acceptor/electron donating gas molecules on a nanotube causes cha rge transfer between nanotube and adsorbed molecules, leading to dramatic electron conductance change in nanotube films, which serves as basis for highly sensitive carbon nanotube sensors [ 26, 27] . SWNTs can easily be de doped through high temperature annealing to de adsorb any
20 attached/intercalated species and re shift Fermi level to its initial, dedoped position. Previous study in our lab  demonstrated appreciable Fermi level shit from electrical field gating via optical transmittance measurements. Fermi level shifting from electric fie ld gating effect serves as a main key for device working mechanism discussed later in this t hesis . 1.2 SWNTs Synthesis There are three major ways to produce SWNTs, including arc discharge, pulsed laser vaporization (PLV) and chemical vapor deposition (CVD ) [28 31] . With each having advantage s and drawbacks, PLV is the method producing SWNTs with high quality and low defect density [31, 32] , and is the technique use d for SWNTs growth in our lab. As prepared carbon nanotubes (CNTs) require multiple purification steps to remove contaminants residue, mainly including catalyst nanoparticles and amorphous carbon. Detailed information is described in previous literature  . After purification, SWNTs are suspended in a surfactant solution, polyoxyethylene tert octyphenyl ether, trade name Triton X100 surfactant (TX) using bath ultrasonication , and stored refrigerat ed at 4 Â° C. 1.3 SWNTs Film Fabrication: Vacuum Filtration Method SWNTs films can be prepared through a simple vacuum filtration method  . In a typical procedure, SWNTs/surfactant suspension is filtered through a mixed cellulose ester (MCE) membrane with 100 nm pore size, leaving nanotubes accumulated on the membrane forming a thin film. Film thickness is easily controlled by the concentration of SWNTs in suspension and the volume of suspension filtered. The SWNTs film on the membrane is washed with copious amount s of DI water to remove any surfactant residue, dried under lamp a nd stored in inert atmosphere for future use. Vacuum
21 filtration is usually followed by a film transfer process for film deposition onto desired substrates. W hen transferring the film to a substrate, the film is pressed against the substrate to form an inti mate contact and the membrane is graduall y dissolved in acetone vapor bath at 60 Â°C for 2 hours followed by thoroughly washed away with three acetone baths (each at least 2 h long) and one short isopropyl alcohol bath (5 10min long). This v acuum filtration method produce s SWNTs films with high uniformity and conductivity, which were widely used in electronic device ma de in our lab in the past few years and showed outstanding performance [10, 33 35] . However, the weakness of this vacuum filtration method is that film size produced is limited to the scale of filtration apparatus and available membrane size s and thus hard to scale up for mass production. Another disadvantage is that a transfer step usually is needed for vacuum filtered film s. During the transfer process, substrates must tolerate exposure to hot acetone vapor and multiple hours long acetone baths, which limits the substrate material s onto which the films can be transferred . To overcome this limitation another solution based S WNTs deposition technique was investigated in this thesis . 1.4 Other Solution based SWNTs Film Deposition Methods Solution based SWNT film deposition has been extensively investigated because of its substantial advantages compared to direct growth methods , for example, it is a low temperature process which is compatible with various substrates including plastics ; it does not need a high vacuum system, significantly lowering the fabrication cost s . Besides vacuum filtration transfer method discussed in the S ection 1.3 , solution based deposition includes Langmuir Blodgett deposition , s elf assembly method, dip coating, spin coating, drop casting, Mayer rod coating, electrophoretic deposition,
22 dielectrophoresis, and spray coating  . A brief descrip tion of these techniques and their advantages and draw backs follow s : Blodgett (LB) method consists of forming a SWNT film at the air / liquid interface in an LB trough and pushing the film via a motorized paddle onto a substrate that is simultaneously lifted at the same rate . SWNT films using the LB like method have been prepared based on the hydrophobic behavior of CNTs [36 39] . A k ey factor is the good surface spreadin g of CNTs at the air/water interface. Tube alignment can be observed, and is attributed to compressed induced or flow induced orientation  . The LB technique is useful for produc ing monolayer/submonolayer films; however, it is not readily scalable. Dip coating has been used to fabricate thin CNT films on various subst rates  . Solution viscosity, interaction betw een functionalized CNTs and substrates surface and coating speed are the main factors determ ining coated film qualities. Surfactants/dispersants are usually used to functionalize CNT surface with negatively or positively charged chemical groups. Substrate surface is modified by linking either polar or non polar functional groups with CNTs being attracted toward polar regions or by writing charge with CNTs being attracted to the charged region through Columbic interactions [41, 42] . However, during the dip coating process both sides of substrate are coated, which may not be preferred for some applicatio ns. Another disadvantage is that self organization of CNTs in evaporating droplets is complicated during drying process, which may lead to certain local tube alignments and thus effect film quality [43, 44] . Besides, substrates fu nctionalization may not be compatible with the functionality of the device to be built using this method.
23 Spin coating [45, 46] and drop castin g  methods rely on a high shear force during film fabrication. Small amount of solution is dropped o nto the substrate followed by high speed spinning of substrate (s pin coating) or air drying (drop casting), respectively. These two methods are easily processed; however, the non uniformity of CNT distribution caused by circumferential force during spin coating or drying process during drop casting remains problematic. The most wide spread solution based coating techniques in industry that have been applied to deposit CNTs films mainly include Mayer rod, Slot slide and Gravure [48, 49] followed by controlled drying. For example, Mayer rod is an easy processed method that dries the liqu id thin film with a wire wound heating bar that passes the liquid film with certain speed allowing liquid to be fully evaporated and leaving CNTs on substrates. This method can deposit films onto various substrates in a scalable way, however, it does have restrictions for coating fluid, which requires specific rheological behavior and wetting properties. Electrophoretic deposition (EPD) and dielectrophoresis (DEP) require external electrical field with certain magnitude and/or frequency to drive charged CN Ts moving towards conductive electrodes. These two process es are difficult to scale up and less applicable [50 53] . Spray coating method has been investigated to fabricate CNT films for further promising applications [54 56] . Advantages of this technique include: high uniformity can be achieved by using diluted nanotube/agent solutions, called nanotube inks and multiple spray steps; requirement of surface tension of CNTs dispersion and surface en ergy of substrates is less stringent, especially for ultrasonic spray coating ; less
24 dewetting effect on film quality because of extremely small droplet size after atomization in ultrasonic nozzle; spray coating is compatible with substrates having differen t topology, even extremely rough surface; spray coating is easy to scale up. Spray coating technique has been investigated in this work for fabrication of highly uniform, large area SW NT film based on its superiority to other solution based process es . How ever, subsequent removal of dispersant that is utilized to stabilize nanotube dispersion, without disturbi ng or damaging nanotubes in the sprayed film remains a major research objective. Challenges of this work include SWNT ink optimization by minimizing n ecessary amount of dispersant and therefore simplifying post treatment of sprayed films.
25 CHAPTER 2 INTRODUCTION TO SINGLE WALL CARBON NANOTUBE DISPERSION (SWNT INK ) 2.1SWNTs Ink Prepara tion Methods: Nanotube Modification As prepared SWNTs are hydrophobic and possess only minor solubility in organic solvents . Moreover, in high yield synthesis methods the nanotubes naturally form van der Waals bundles and larger assemblies that are difficult to break apart. Th is is due to th e inter tube Van Der Waals forces that caused by the large aspect ratio of nanotubes. Such strong forces keep CNTs bundled together in a form of big ropes, which are hard to separate and disperse, increasing process ing difficult y and imposing limit s o n CNT s applications. Nanotubes can be suspended in some solvents by simple ultrasonication. This type of nanotube suspension possesses poor stability with precipitation occurring immediately or after a short time when sonication is interrupted . Meanwhile, harsh ultrasonication induces defects on the ends and sidewalls of carbon nanotubes and even shortens the tube length causing low conductivity of subsequently fabricated CNT films. Superior to sonication only process, chemical modification of nanotube surface t o induce CNT solubility in organic solvents/water for a more stabilized CNT dispersion has attracted extensive interest [57 59] . Chemical m odification approaches can be divided i nto two categ ories: a) covalent attachment of chemical groups onto carbon nanotube surface ; b) noncovalent interaction with the aid of surfactants or dispersants (e.g., polymer wrapping ) [60 62] . Covalent bond formation a ffect s electronic band structure of nano tubes conjugation and changing their electronic , optical, and mechanical properties. On the other hand, n oncovalent fu nctionalization is based on Van der Waals forces and/or stacking bonds between chemical surfactant/dispersant and CNTs; it attaches functional
26 groups to CNTs without affecting the electronic properties of the nanotubes . Therefore, noncovalent functiona lization is superior to covalent nanotube functionalization methods and will be used in this report. Noncovalent functionalization is generally formed through interaction between carbon nanotube surface and various chemical compounds including polymers [62, 63] , polynuclear aromatic compounds  , surfactants [65, 66] or biomolecules  , among which aromatic compounds attract most attention. The reason is that aromatic compounds contain benzene rings as shown in Figure 2 1, likely as a unit of CNT s graphite structure, has an affinity to CNTs on basis of conjugated bonds, which is more stable than Van der Waals forces induced by other noncovalent functionalization. In our lab , SWNTs are originally suspended and stored in 1%TX surfactant suspension. Therefore, both surfactant /CNTs suspension and aromatic compounds/ CNTs suspension attract our interest and will be discussed. Figure 2 1 . Chemical structure of benzene ring and carbon nanotube sidewall. nanotube sidewall Benzene ring
27 2.2 Surfactant Functionalized SWNTs 2.2.1 Fundamentals of Surfactant Surfactants are compounds t hat greatly reduce the surface tension between two liquids or between a solid and a liquid. S urfactants are water soluble with both hydrophobic and hydrophilic functional groups, enhancing solubility of insoluble/low solubility matter in aqueous solution. This mechanism was used to prepare aqueous based carbon nanotube dispersion. Surfactant s tend to partition into water/air or water/oil interface with hydrophilic heads dissolve in water and hydrophobic tails extended into air/oil, reducing the system s free energy by removing contact between its hydrophobic part and water. As concentration inc reases, surfactant in bulk aqueous solutions aggregates into Micelles. Micelle structure is shown in Figure 2 2; hydrophobic tails start to associate with each other while hydrophilic heads helps to disperse in water  . Critical Micelle Concentration (CMC) is the point where micelles start to form. Above CMC, any addition al surfactant a dded to solution will increase the number of micelles. Previous studies [69 71] reported that CMC and double CMC (DCMC) of TX are 0.28 mM and 0.54 mM and aggregation number in one micelle of TX ranges from 105 to 139 at room temperatur e T = 25 depending substantially on the ionic strength in solution. As T increases, aggregation number increases. The micellar radius ofaround 3.7 nm has been determined when CTX is above CMC up to 9.6 mM.
28 F igure 2 2. Micelle structure of surfactant in aqueous solution . Reprinted with permission from Pasquali, Matteo , Gelation: Grow with the flow (Nature Materials, 2010). 2.2.2 Surfactant functionalized SWNTs Strong intertube v an der Waals forces cause CNTs to form large bundles and render them insoluble in common organic solvents and water . Stability of surfactant assisted nanotube suspension depends on a variety of factors, including pH level of the solution, temperature, ultrasonication power and time, nature of the surfactant, its concentration in suspension , an d concentration of CNTs . W ith all other external conditions the same, strong surfactant graphite interactions play the most important role in achieving high stability of the dispersion . The association structure between surfactants and nanotubes varies dep ending on the nature of the surfactant. Self organization of surfactant on nanotubes can be divided into three types as shown in Figure 2 3: cylindrical micelles, semi micelles and random [72 74] . Which self organization of surfactant gives better stability of ink still remains debatable.
29 Figure 2 3. Self organization structures of surfactant in surfactant/nanotube suspensions, divided into three association structure: cylindrical micelles, semi micelles &random. Reprinted with permission from Vaisman, L., H.D. Wagner, and G. Marom, The rol e of surfactants in dispersion of carbon nanotubes (Elsevier, 2006). Among surfactants that ha ve been applied for nanotube dispersion, TX, sodium dodecyl sulfate (SDS) and sodium dodecylbenzene sulfonate (NaDDBS) are most widely studied. Studies showed tha t the hydrophobic alkyl chain part of the surfactant molecules tends to lie flat on the graphite surface [75, 76] . Figure 2 4 gives a schematic representation of how these three surfactants associate with nanotubes and form semi micelle structures. The interaction streng th between the surfactant molecules that lie along the tube surface (parallel to the tube axis ) and nanotube sidewalls determines disper sion stability. It has been reported that surfactant with conjugated moiety, longer alkyl chain length, and smaller head groups tends to form stronger interactions with nanotube sidewalls and induce higher stability of the CNT dispersion  . In ionic surfactants the ionic head group ( e.g. SO 3 ) causes electros tatic repulsion leading to charge stabilization of nano tubes, which also improves stability of the CNT dispersion . Micelle Random Semi Micelle
30 F igure 2 4. Schematic representation of how three most widely studied surfactants associate with nanotube surface. Reprinted with permissi on from Islam, MF , et al., High weight fraction surfactant solubilization of single wall carbon nanotubes in water (American Chemical Society, 2003). 2.3 Aromatic Compounds Pyrene Derivatives Functionalized SWNTs Inks 2.3.1 Introduction to Pyrene Derivativ es Among aromatic compounds for dispersing CNTs, pyrene derivatives have draw n substantial attention. Pyrene (C 16 H 10 ) is a flat aromatic system formed by four neighboring benzene rings, as shown in Figure 2 5. This conjugated molecule has very similar structure with nanotube sidewalls, thus providing a very strong affinity with nanotube surface through bonding . By attaching to the pyrene moiety a water soluble group the entity can behave like a surfactant to sol ublize carbon nanotubes . Another advantage is that pyrene has characteristic fluorescence signals, which are dramatically quenched by Forster resonance energy transfer once the pyrene associat es with either graphene or nanotubes [77 80] . This feature can be can be used to monitor the association during the preparation process. In my work , we have exploited two types of pyrene derivatives: a pyrene tethered by a flexible linker along a long chain polymer backbone and small molecular pyrene derivatives.
31 Figure 2 5. Chemical structure of pyrene a flat aromatic system formed by four neighboring benzene rings. 22.214.171.124 Pyrene functionalized h ydroxypropyl cellulose: Py HPC Stable CNTs dispersions in water and organic solvents were obtained using pyrene deriva tized polymers [57, 81] . In this thesis , a long chain polymer Pyrene Hydroxypropyl Cellulose (HPC) functionalized by pyrene was used for nanotube dispersion and film formation . Cellulose is a long chain polysaccharide consisting of linked D glucose units. In HPC some of the hydroxyl hydrogens are replaced by the hydroxypropylated form, attaching OCH 2 CH(OH)CH 3 groups derived from propylene oxide . Commercially available HPC is soluble in water and co mmon organic solvents and biodegradable to some extent  . As shown in F igure 2 6 A , a pyrene derivatized HPC (Py HPC) was synthesized via a condensation substitution reaction ; the py renes are attached to oxypropyl pendant groups of the HPC polymer as described elsewhere  . The Py HPC used was synthesized by Prof. John Reynold s group formerly of UF but presently at the Georgia Institute of Technology.
32 Both Py HPC and HPC have been used to disperse CNTs [58, 79, 83] . HPC has an affinity to nanotubes presumably due to the Van der Waals interaction between its polysacharide structure and nanotubes while its polar groups promote water/ alcohol solubility of the CNTs . However, a high HPC concentration vs . SWNTs concentration is required to ensure a stable SWNTs/HPC suspension. Py HPC provides a stronger affinity to the nanotubes through stacking between the pyrene s and the nanotube sidewalls, resulting in more stable dispersions and the need for far less excess dispersant to stabilize the nanotubes providing the rationale behind its use in this work . Figure2 6 B shows a schematic representation of how Py HPC associates with nanotubes. It has been reported that pyrene containing polymers or conjugated polymers wrap around nanotube in the dispersion ; however the wrapping structure varies depending on the solvent and p H value [84, 85] . It should also depend upon the density of pyrene substitution along the polymer backbone. Figure 2 6 C compares two SWNT ink s that were dispersed using HPC and Py HPC under the same condition after six months of storag e in a refrigerator at 4 Â° C. Almost all of the nanotubes in HPC solution aggregate d, while the Py HPC/SWNTs ink remained stable with no visible aggregates, with the HPC concentration equivalent to the Py HPC concentration. Pyrene functionalized HPC provide s a dramatically enhanced SWNTs ink stability, a s will be shown below with minimum excess dispersant .
33 Figure 2 6. Py HPC/SWNT ink preparation. A ) synthe tic route for Py HPC from Pyrene and HPC, reprinted with permission from Yang, Q. and X. Pan, Preparation and characterization of water soluble single walled carbon nanotu bes by hybridization with hydroxypropyl cellulose derivatives (American Chemical Society, 2010), B ) schematic representation of how Py HPC absorb s on to the na notube surface, C ) Stability comparison of SWNTs dispersed with HPC &Py HPC after six months stora ge at 4 Â° C in refrigerator . 126.96.36.199 Small molecular pyrene d erivatives Pyrene molecules modified with polar hydrophilic functional groups attached to hydrophobic polycyclic aromatic system induce good solubility in water or other solvents (e.g., alcohols) . Numerous small molecular pyrene derivatives are commercially available and some of them have been used as dispersant s to create graphene or CNT dispersions , includ ing 1 pyrenesulfonic acid (PSA), 1 pyrenecarboxylic acid (PCA), 1 pyrenebutyric acid (PBA), 1 pyrenemethylamine (PMA) and 1,3,6,8 Pyrenetetrasulfonic Acid Tetrasodium Salt (PTSA) (chemical structure s are shown in Figure 2 7. Hydrophobic pyrene is functionalized with hydrophilic groups to induce their solubility in various solvents . Because of th eir amphiphilic nature, it was reported that functionalized A B C CNTs in HPC solution CNTs in Py HPC solution
34 pyrene molecules may form micell ar architecture of aggregates when pyrene concentration is as low as C pyrene =10 4 m M [78, 86] . For dispersing CNTs, the planar aromatic structure of the pyrene moiety anchor s it sel f onto hydro phobic sidewall surface of nanotubes through stacking while the functional hydrophilic groups provide dispensability in water or alcohol . Recent studies show ed that pyrenesulfonate based SWNTs dispersion exhibited higher stability compare d to that of the PCA and PBA based SWNTs suspensions  . Moreover, PTSA based graphene dispersions could be prepared using milder association conditions than in case of PBA dispersant  . These findings are consistent with our independen t studies. With two CNT dispersions prepared under the same condition s (same concentrations) , PCA based SWNTs ink started to aggregate after a week, showing a poorer stability while PTSA based SWNTs ink was stable for more than two months. This could be al so explained by charge stabilization of tubes. In general, introduction of multiple bulky sulfonic groups around the nanotube s should induce strong repulsive forces and thereby realize an enhanced PTSA/SWNTs ink stability compared to that of the PCA or PB A based molecular dispersions.
35 Figure 2 7 . Chemical structure of some small molecular pyrene derivativesstudied for CNT dispersion . 1 pyrenesulfonic acid (PSA) 1 pyrenecarboxylic acid (PCA) 1 pyrenebutyric acid (PBA) 1 pyrenemethylamine (PMA) 1,3,6,8 Pyrenetetrasulfonic Acid Tetrasodium Salt (PTSA)
36 2.3.2 Pyrene Fluorescence Pyrene chromoph o re s have been widely used as fluorescent probe s because of their pronounced characteristic fluorescen t signals, including both monomer and excimer emission [88 91] . 188.8.131.52 Introduction to fluorescence measurement Fluorescence is the emission of light by a substance that has absorbed light or other electromagnetic radiation. Except for circumstances in which the excitation has very high intensity (not considered in this work) the emitted light has longer wavelength ( lower energy ) than the absorbed radiation  . The typical fluorescence instrumentation is illustrated schematically in F igure 2 8. Incident light from an excitation source is filter ed through the first monochromator and strikes the sample. A proportion of the incident light is absorbed and some of the molecules fluoresce in all directions. To help avoid the intense incident light, only emitted light perpendicular to the incident dire ction is collected by the detector. In this work fluorescence emission, excitation and lifetime measurement were used. Emission spectra were obtained by keeping the excitation wavelength fixed and scanning across the emission wavelengths. Excitation spectr a were obtained by sweeping the excitation light across a wide range of wavelengths and recording the emission intensity at a particular wavelength . Fluorescence lifetime measurement records the emission intensity at a fixed excitation and emitting light w avelength as a function of time (with the broad band excitation replaced by pulsed excitation source of short duration). These studies were performed using a SpexFluorolog 3 model FL3 22 Spectrofluorometer (Jobin Yvon Horiba) in the laboratory of Prof. Weih ong Tan in the Chemistry Department at UF. The Spex Fluorolog 3 is equipped with a 450 W Xenon source,
37 double monochromators for excitation and emission, and a R928P photomultiplier tube in photon counting mode as the emission detector . Figure 2 8. Schematic representation of Fluorescence Measurement System 184.108.40.206 Fundamentals of fluorescence spectrum: emission and excitation Emission spectr a are the relative distribution of wavelengths emitted when the sample is excit ed a t a single wav elength. During the excitation, molecules that adsorb the incident radiation at a specific wavelength (energy) undergo a transit ion from the lowest vibrational level in the ground electronic state S 0 to some vibrational level in excited electronic states. The radiative deactivation of upper electronically excited state
38 is known as fluorescence and occurs very rarely as a mode of electronic transition between two electronic states other than the ground state and the lowest excited singlet state. Expl anation is the excited electronic states are much closer together than are the ground state and the lowest excited state in most molecules. Consequently, efficient vibrational relaxation and internal conversion almost always occurs and precludes the possib ility of fluorescence arising as the result of electronic transition between states other than the ground state and the lowest excited state. For this reason, only the lowest excited single t state S 1 is shown in Figure 2 9. After excited to the different v ibrational levels in S 1 , rapid vibrational relaxation (10 14 10 12 s) relax es the molecules to the lowest vibrational level of the excited electronic state S 1 , from where the fluorescence takes place. Molecule fluorescence is among the slowest processes in electronically excited states, requiring from 10 10 to a few seconds  . The fluorescing molecule will then arrive into the vibrational manifold of the ground electronic state , as shown in Fig ure 2 9. This explains why emission spectr aare p ractically independent of the excitation wavelength (so long as sufficient energy for the electronic transition is provided). Subsequent to emission, vibrational relaxation occurs, with molecule ultimately arriving in the lowest vibrational level of the gr ound state.
39 Figure 2 9. Energy level diagram showing why structure is seen in the absorption and emission spectrum and why these two spectrum are roughly mirror images to each other. Excitation spectra probe the absorption of the incident light. By monitoring the emission at a specific wavelength (within the emission band) as the excitation wavelength is swept, one gets a record of the relative distribution o f the absorbed intensities (Fig ure 2 10 , Absorption). Emission generally has lower energy ( lo nger wavelength ) compared to adsorption, with a displacement called the Stokes s hift , explained by the Franck Condon principle . The emission and absorption spectr a often appear as m irror images . Schematic explanation is shown in Figure 2 10. Emission w avel ength 0 corresponds to the transition to lowest vibrational level in excited state S 1 , representing the lowest absorption energy and longest wavelength in the absorption
40 spectrum. In emission, 0 represents the highest fluorescent energy with the lowest w avelength. The fact that the spacing between energy levels in the vibrational manifolds of the ground state and the first excited state are roughly equal and the transition probabilities are similar results in the approximately mirror images between absorp tion and emission spectr a . Figure 2 10 shows the absorption and emission spectra of the PTSA solution in water forming a m irror image Generally, increasing the size of the aromatic system results in lower frequencies (l onger wavelength) of fluorescence because the energy gap between the ground and excited states decreases in magnitude as the aromatic molecule increases. Figure 2 10. Mirror image of the emission (black line) and absorption spectr a of the PTSA solution in water , excitation = 348 nm . The e xcitation spectrum shows the relative efficiencies with which different wavelengths of incident light to excite fluorophores. If the fluorescent energy is
41 proportional to the absorbed energy, excitation spectrum will match the absorption spectrum. For more sensitive fluorescence detection the emission spectrum is generally excited at the wavelength of the maximum intensity signal in the absorption spectrum; while the exc itation spectrum is generally detected at the wavelength of the peak emission. 220.127.116.11 Typical pyrene fluorescence spectrum The p yrene fluorescence spectrum is characterized by both excimer and monomer emission band s . An e xcimer is a dimeric or heterodime ric molecule formed by two species, at least one of which is in electronic excited state. In general , excimers are formed by two atoms or molecules that would not combine if both were in ground state. Excimer formation increases along with the density of m onomer because of its bi molecule interaction mechanism. With small concentration, excited monomers decay to its ground state before interacting with a ground state monomer and no excimer band would be observed in the emission spectrum. However, studies [91, 94] have show n that pyrene excimers can be formed by pre associate ground state dimers or by the association of an excited and ground state pyrene molecule, depending on the structure of the pyrene derivative or the external gas/liquid environment. Figure 2 11 shows the typical spe ctr a of pyrene in a) emission and b) excitation . Pyrene exhibits a structured violet emission with three vibrational peaks located at 370 nm, 398 nm and 421 nm (monomer emission) and a broad structureless aqua blue emission which appears at around 470nm (e xcimer emission)  . Typically, the unstructured emission in blue region of the spectrum arises as concentration of pyrene is brought up to 10 2 M. The intensity of the blue fluorescence is proportional to the square of the pyrene concentration  .
42 As pyrene concentration increases, monomer emission essentially decreases and excimer emission occurs. Excimer emission, as part of pyrene characteristic spectrum, has been widely applied to investigate the interaction between pyrene functionalized species [88, 96, 97] . Figure 2 11.Fluorescence spectrum of p yrene. A ) emission curve excited at 328 nm for Py HPC in aqueous solution ,B ) excitation curve monitored at 376 nm for Py HPC in aqueous solution. Reprinted with permission from Yamazaki, I., et al., Excimer formation in pyrene labeled hydroxypropyl cellulo se in water: picosecond fluorescence studies (American Chemical Society, 1987) .
43 CHAPTER 3 SINGLE WALL CARBON NANOTUBE INK PREPAR ATION AND CHARACTERIZATION 3.1 Py HPC/SWNTs Ink 3.1.1 Py HPC/SWNTs ink Prepar ation Solution based deposition of CNTs films is usually followed by removal of insulating dispersant/surfactant residue to increase film conductivity. Therefore, minimizing amount of the dispersant used in the ink is desirable . To improve conductivity of CNT fi lms that are made from CNT inks, longer and highly individualized dispersed tubes are preferred  . Superior SWNTs ink is characterized by highly individualized tubes with less/smaller bundles, l onger tubes and small dispersant to CNTs ratio. A high stability, as another important feature of good CNT inks, also needs to be considered for potential scale up usage. In our lab, SWNTs are grown by dual pulsed laser vaporization, followed by a multi s tep purification and stored in a 1% (wt) non ionic surfactant TX  . To prepare Py HPC/SWNTs ink, the1% TX surfactant of the stock suspension was first reduced to a level approaching the CMC. The TX associated with the nanotubes was subsequently exchanged by the Py HPC dispersant. Finally, the excess surfactant and unassociated Py HPC were removed and the Py HPC dispersed SWNT ink collected for later use. These steps are illustrated in Figure 3 1 and elaborated below. 1) Minimizing the TX concentration in TX/ SWNT s dispersion . The SW NT s/ 1% TX suspension was bath sonicated for 20 min followed by vacuum filt ration of the suspension and excess TX through an Isopore VCPT polycarbonate membrane (0.1 Âµm) . In carrying out this step it was critical to avoid letting the SWNT film that forms at the filter surface as a filter cake dry since once dried the SWNTs associate in a mass
44 that becomes very difficult to redisperse. T his film was washed with an aqueous TX solution at its D CMC, in order to get rid of the excess of the TX but still ensure a good TX coating on SWNTs . The wet film was then scraped off the membrane into a n aqueous TX solution at double the critical micelle concentration ( D CMC ) and left stirri ng overnight. The SWNTs were re dispersed by a mild bath sonication (20 min) followed by centrifug ation at 6 , 000rpm for 1h to sediment any large tube aggregates and t he supernatant was collected. Figure 3 1. Py HPC/SWNTs ink preparation process. 2) TX exchange . Equal volumes of D CMC TX/SWNTs and Py HPC solution (0. 33mg/m L ) were mixed and stirred for 4 5 days. TX association with the nanotubes is a meta stable and dynamic process. During this incubation , because of strong stacking interaction between p yrene moieties and SWNTs, Py HPC gradually replac es the TX to as sociate with nanotube sidewalls .
45 3) Minimi zing excess TX surfactant and unassociated Py HPC dispersant . The suspension was centrifuged at 17 , 500 rpm for 5 hours, the supernatant was pipet t e out, and the Py HPC/SWNTs sediment was redispersed in DI water or ethanol via brief bath sonication. This step was repeated twice to thoroughly wash the Py HPC/SWNTs. 3.1.2 Py HPC/SWNTs Ink Characterization: UV vis Absorbance Minimiz ing excess dispersant through the centrifugation cycles was monitored using a Perkin Elmer Lambda 900 UV Vis spectrophotometer . Figure 3 2 A shows the absorbance spectr a of the TX and Py HPC aqueous solutions. The absorbance spectrum of the Py HPC (red line) has six characteris tic pyrene absorption peaks at 224, 267, 277, 314, 328 and 344 nm  . The spectrum of TX (black line) shows a characteristic absorption peak at 276 nm with a non pronounced shoulder at 287 nm and an intense absorption band at 223 nm. Absorbance spectr a of the Py HPC/SWNTs/TX mixture before and after dispersant minimizing (step 3 above) were record ed to verify whether the excess TX and Py HPC w as thoro ughly removed . As shown in Figure 3 2 B both the pyrene and TX absorption signals are greatly diminished , indicating that the surfactant and non associated dispersant have essentially been removed. After preparation, the ink was stored in a refrigerator at 4 Â° C ; no nanotube flocculation was observed even after six months of storage.
46 Figure 3 2 . UV vis spectrum for Py HPC/SWNTs ink characterization. A ) UV vis absorbance spectrum of the TX (black line) and Py HPC (red line) solutions in water ,B ) UV vis ab sorbance spectrum of mixture Py HPC/SWNTs /TX before (black line) and after (red line) purification through centrifugation process. Although the stability and puri ty of the prepared SWNT ink s w ere quite high , purification of the inks using the multi step centrifugation was labor intensive and would be difficult to scale u p. Another less labor intensive means for removing the TX and A B
47 excess Py HPC from the sample, effectively to replace step 3 above is dialysis. During Py HPC/SWNT dialysis the material is placed within a porous pouch membrane with pore s of a size significantly smaller than the size of nanotubes but large enough to let out surfacta nt/ dispersant. These pouches were sealed and then permitted to float in a large volume container of DI water. The driving force for dialysis is the concentration gradient between the aqueous solutions inside and outside the pouch membrane so the DI water i n the large container was periodically refreshed. Unfortunately , Py HPC , as a polymeric material, takes long time to be dialyzed compar ed to monomeric dispersants (discussed below) , due to agg regation in solution . Experiment s were performed t o dialyze the Py HPC/SWNTs/TX mixture . While this successful ly made usable Py HPC/SWNTs ink , the long dialysis time s and need for membrane replacement due to fouling of the relatively expensive membrane pores made this less promising . Another weak point of the Py HPC/SWNTs ink l ay in the post treatment of the SWNTs films fabricat ed from the ink using spray coating . Because of its long chain structure and complex wrapping association with nanotubes, it proved somewhat difficult to thoroughly remove. Multi ple nitric acid baths were need ed for this . 3.2 PTSA/SWNTs Ink 3.2.1 PTSA/SWNTs Ink Prepar ation The challenges of purifying the excess Py HPC from the Py HPC/SWNT mixture and removing the SWNT associated Py HPC from the films encouraged a search for alter native dispersants. To decrease the dialysis time and potentially avoid harsh acid post treatment for SWNTs films, small molecular pyrene derivative s w ere selected to try as SWNT dispersants . As discussed previously PTSA has a strong affinity to the graph e ne like structure of nanotubes . The multiple sulfonate groups promote SWNT
48 dispersability in water and alcohols and provide high columbic repulsive forces prevent tube to tube association and thus, induce good ink stability . The PTSA/SWNTs ink preparation include s three steps with the first two steps similar to the Py HPC based SWNTs ink preparat ion steps : 1) minimizing TX concentration to the DCMC level in TX /SWNTs suspension in water, 2) replacement of the TX surfactant with PTSA dispersant. During th is s tep specific volumes of aqueous TX /SWNTs suspension and PTSA solution were mixed to achieve the following concentrations: C TX DCMC of TX surfactant) and C PTSA = 10 4 m M. The concentration of the SWNTs C SWNTs in the TX /SWNT s suspension s was va ri ed depending on the target weight ratio s of PTSA to SWNTs. The mixture was then slowly stirred for 4 5 days to insure efficient association of the PTSA molecules with the carbon nanotubes. After that incubation time the PTSA/SWNT s /TX suspension was subjected to dialysis using b iotech cellulose ester membranes with a molecular weight cut off (MWCO) of 1,000 kD (Spectrum Labs) against water to remove unassociated PTSA and TX. The DI water in the exterior container was replaced every 24 h . Dialysis time depends on density of nanotubes in the mixture and rang ed from 48 to 96 h. The efficiency of the d ialysis in removing both excess surfactant and non associated dispersant was monitored using UV V is absorbance measurement s . After the dialysi s, the suspension was centrifuged at 6 , 000 rpm for 1 h to remove any impurities in tro duced during the dialysis and to get rid of large SWNT aggregates that formed during the process. Once complete the ink was stored in refrigerator at 4 Â° C for future use.
49 3.2.2 PTSA/SWNTs Ink Characterization Figure 3 3 shows effect of the dialysis during an early batch of PTSA based SWNT ink preparation. A quantity of 0.5mL of 0.4 Âµ g/mL PTSA aqueous solution were added to 3m L of 0.67 Âµ g/m L SWNT suspension in ~DCMC TX yielding a weight ratio of PTSA : SWNTs=1:10. SWNT suspension in ~DCMC TX was prepared from original store suspension SWNT in 1% TX as described in S ection 3.1.1. After 48 h of dialysis, characteristic TX absorption band s at 230 3 0 0 nm absorption region were deplete to a baseline intensity (red line) indicating an efficient removal of the TX surfactant. Since the PTSA concentration in the original PTSA/SWNTs/TX mixture was extremely low (C PTSA =10 4 m M), the characteristic absorbance peaks of the pyrene at 300 350 nm were barely observ able before the dialysis though a small decrease is observed after dialysis. After a quick centrifugation, UV V is absorption was measured again. No change was observed between the spectra of the inks before and after centrif ugation (overlapping with red line). This ink was stable for at least two months of refrigerated storage . Figure 3 3 . UV V is spectr a of the PTSA/SWNT s /TX mixture before ( black line ) and after (red line) 48 h of dialysis . In the PTSA/SWNTs/TX mixture C TX 0. 56 mM , C SWNTs =0.57 Âµ g/m L , C PTSA =10 4 m M.
50 3.2.3 Optimiz ationof the PTSA/SWNTs Ink Prepara tion Procedure Using Fluorescence Measurement s The objective of the ink optimization wa s to : 1) minimize the amount of dispersant so that less post processing treatment would be needed for sprayed films, and 2) simplif ication of the ink preparation process. The UV V is measurements were an effective monitor of the TX removal by dialysis based on dramatic decrease of the TX absorbance signals, however, it provide d little information about the PTSA dispersant because of the extremely low PTSA concentration in the initial PTSA/SWNTs/TX suspension . In order to quantify the effect of dialysis on removal of the excess of unbound PTSA dispersant , more sensitive fluorescence measurement s were conducted . 18.104.22.168 Typical fluorescence spectrum of PTSA in water T ypical fluorescence spectr a for PTSA in water (10 4 m M) are shown in Figure 3 4. Both emission ( A ) and excitation ( B ) curves are shown . The e mi ssion spectrum excited at 348 nm shows only the monomer emission peaks located at 383 nm, 402 nm and 425 nm ; no excimer signals were observed , as is typical for solutions with pyrene concentration s lower than 10 2 M. The e xcitation spectrum monitored at 402 nm shows three corresponding vibrational peaks located at 374 nm, 355 nm and 338 nm with three less resolved peaks located at 260 290 nm. Since in the initial PTSA/SWNT s /TX suspension we have two fluorophores (TX and PTSA) with emission and absorption maxima at similar wavelengths , energy transfer between these two fluorescent molecules could occur a ffect ing their fluorescent intensity values and obstructing analysis of the spectra [69, 77, 99] . Fluorescence measurements were performed to study and understand this ph enomenon.
51 Figure 3 4. Fluorescence spectr a of PTSA in water. A ) Emission spectrum recorded at excit ation wavelength = 348 nm, B ) excitation spectrum monitored at 402 nm , C PTSA = 10 4 m M . 22.214.171.124 TX induced enhanc ed fluorescence of PTSA It has previously been reported [69, 99] , that a strong ov erlap between the fluorescence emission of TX ( donor ) and the absorption of pyrene ( acceptor ) in pyrene and 1 pyrenebutyric acid aqueous solutions results in an efficient energy transfer that yields an increased intensity of pyrene emission (so called sens itization of pyrene) . T he donor and acceptor molecules must lie within quite close proximity for the energy transfer to take place . Such close proximity readily arises if the pyrene molecules are solubilized in the micellar interior of the surfactant molec ule . For this to occur t he concentration of T X surfactant should be higher than its critical micelle concentration (our circumstance) and w e indeed observed an enhanced intensity of pyrene emission in PTSA/TX s olution s compared to that of the PTSA in pure water ( F igure 3 5). The PTSA concentration in both solutions was C PTSA = 10 4 m M while the TX concentration in the PTSA /TX solution wa s slightly less than DCMC level ( C TX ~ 0. 56 m M ) . The e mission intensity for PTSA /TX solution (bla ck line) was almost 25% higher compared A B
52 that of the PTSA solution in pure water (red line) at the peak emission = 383 nm , indicating enhanced fluorescence of the pyrene acceptor induced via energy transfer from TX donor . This phenomenon cou ld contribute to an enhanced fluorescence emission in PTSA/SWNTs/TX suspensions and must be taken into consideration during analysis of the spectra. For accurate characterization in the following series of fluorescence measurements of the PTSA/SWNTs/TX sus pensions with various SWNT concentrations, PTSA/TX aqueous solution containing both fluorophores will be measured as a control solution (baseline). Figure 3 5 . Emission s pectr a of PTSA/TX (black line) and PTSA (red line) solutions in water.Both solutions excitation = 348 nm with PTSA concentration at C PTSA = 10 4 m M. In PTSA / TX solution C TX ~ 0. 56 mM . 126.96.36.199 Minimizingthe PTS A to SWNT weight ratio Pyrene based molecules exhibit intense fluorescence emission that is quenched in the presence of carbon nanotubes, which is understood to be due to energy transfer
53 between the pyrene moiet y and the SWNTs. The photoluminescence quenching phenomenon is well known and is often used as evidence of an efficient non covalent [77, 80] . Steady state and time resolved fluorescence measurements were used to characterize the association between PTSA and the SWNTs. In this set of experiments emission spectra of a series of suspension s containing different PT SA to SWNT weight ratios were recorded using Spex Fluorolog 3 spectro fluorometer at room temperature . In these suspension s the PTSA concentration was kept constant while the concentration of carbon nanotubes in TX was gradually increased. Specifically, 3 m L of TX at its DCMC with different SWNT conce ntration (ranging from 0.67ug/mL to 30ug/mL ) were added to 0.5ml 0.4 Âµg/mL PTSA solutions in water . Most concentrated SWNT/TX suspension with C SWNT = 30ug/mL was prepared as previously described in S ection 3.1.1 and then diluted with DCMC TX solution accordingly to lower the SWNT concentration but still ensuring a good TX coating on SWNTs . In all suspensions , the conc entration of PTSA and TX remain ed constant C PTSA = 10 4 m M, C TX ~ 0. 56 mM while concentrati on of nanotubes was gradually increase d , from 0.57 Âµ g/m L (PTSA : SWNTs =1:10) to 25.7 Âµ g/m L (PTSA : SWNTs =1:450) . Immediately after mixing , suspension s w ere ultrasonicated in a bath sonicator for 2s , and then the emission spectra were recorded. A p hotograph of the PTSA/SWNT s /TX suspension s, containing PTSA to SWNT s ratio values ranging from (1 : 10) to (1 : 450), is displayed in F igure 3 6A s pectrum of the PTSA/TX solution without SWNTs is given for comparison and was used as a baseline. Upon excita tion at 348 nm the characteristic vibronic structure of pyrene monomer emission with three distinct bands at 383, 402, and 425 nm is
54 observed. The emission intensity decreases with the increase of SWNT concentration in suspension until it essentially deple tes when the PTSA to SWNT s weight ratio reaches a (1 : 450) value ( F igure 3 7 A ). Figure 3 7B shows the rate with which the fluorescence emission peak intensity is quenched (3 1) Where I 0 and I are the emission intensity value of the baseline PTSA/TX solution and the PTSA/SWNT/TX suspension s , respectively, at emission = 383 nm . As shown in F igure 3 7 B , the rate of quenching increases sublinearly, reaching> 90 % quenching for the maximum amou nt of SWNTs (1:450). This sublinear quench rate dependence on the relative nanotube concentration is most likely due to the increasing absorbance by the nanotubes as their concentration is increased, i.e. the intensity available for PTSA excitation is decr eased by the SWNT absorbance , and the increasing energy transfer between nanotube and pyrene along with increase of nanotube concentration. F igure 3 6. P hotograph of the PTSA/SWNT s /TX dispersions with various PTSA:SWNTs weight ratio values ranging from (1 : 10) to (1 : 450) .
55 Figure 3 7 . Fluorescence Quenching in PTSA/SWNT/TX mixture. A ) Fluorescence emission spectr a of the PTSA/TX control solution used as a baseline (black line) and PTSA / SWNT/TX suspension s with various PTSA:SWNTs weigh ratios, excitation = 348 nm , B ) Quenching rate ((I 0 I)/I 0 ) at the emission peak at emission = 383 nm as a function of SWNT to PTSA weight excitation = 348 nm. In all solutions C PTSA = 10 4 mM , C TX 0. 56 mM .
56 Table 3 1 . Quench ing rate R quench ing of pyrene emission peak intensity at emission = 383 nm corresponding to different SWNT:PTSA weight ratios , excitation = 348 nm SWNT: PTSA 1:10 1:40 1:90 1:105 1:150 1:200 1:250 1:300 1:360 1:400 1:450 R quench ing 20.5 31.9 47.1 51.2 56.1 66.9 72.2 78.4 83.7 86.7 90.6 Quenching rate values corresponding to different SWNT s : PTSA r atios are given in Table 3 1. Clearly demonstrated is that the PTSA : SWNT = 1 : 10 weight ratio utilized in our initial SWNT ink preparation procedure, the quenching rate was quite low (~ 20 %), implying a large excess of unbound PTSA dispersant in the suspension , which indicated that the amount of PTSA necessary for SWNT dispersion could be reduced significantly. The SWNT dispersions containing different PTSA to SWNT ratios were purified using dialysis and stored in refrigerator. As the weight ratio of SWNTs : PTSA increases (weight r atio PTSA:SWNTs decreases), less excess of unbound PTSA dispersant is left in the mixture , which should facilitate removal of the dispersant after the SWNT film is fabricated . However, w hen the concentration of the nanotubes wa s as high as 25.7 Âµ g/m L for t he ink with weight ratio of PTSA : SWNT=1:450 , TX and PTSA diffusion in such concentrated suspension becomes quite slow , making purification of the ink via dialysis more time consuming (96 h) . The s tability of the PTSA based SWNT dispersions was monitored b y periodic inspection for visible flocculation . With increase of the SWNT concentration and decrease of the relative amount of dispersant the stability of the PTSA based SWNT dispersions was found to d ecrease . I nk with a PTSA : SWNT = 1 : 105 weight ratio was deemed to be optimal for further processing and storage. The concentration of PTSA dispersant in this case was reduced by about an order of magnitude compared to that used in itially without sacrificing stability of the dispersion. The PTSA : SWNT = 1 : 105 dispersions were stable for more than one and
57 a half month s of storage without visible nanotube flocculation and demonstrated no change in UV Vis absorption spectra before and after storage. Insignificant aggregation was observed after two months of s torage. Minimizing of dispersant concentration while allow ing for stable carbon nanotube dispersion should facilitate dispersant removal after film fabrication . 188.8.131.52 Minimiz ing PTSA / SWNTs association time As noted above, during the initial studies usin g PTSA as the dispersant, to associate the PTSA with the SWNTs (replacing the TX surfactant coating), the DCMC TX suspended SWNTs and PTSA were stirred for 4 5 days before the dialysis. To determine if such a long time was necessary for this association/re placement, kinetics studies were performed. In these studies immediately after mixing the DCMC TX suspended nanotubes with the PTSA the sample was loaded into the Fluorescence spectrometer and with excitation at 348 nm, decay of the emission peak intensity 383 nm was monitored as a function of time over the first 1.5 h . The emission spectra were then additionally recorded after 1, 7, and 10 days. Figure s 3 8 A K display the fluorescence decay data for the PTSA/SWNT s /TX solutions, containing PTSA to SW NT weight ratio s values ranging from (1 : 10) to (1 : 450) . Figure 3 8L shows all these decay curves on the same plot . After ~ 1000s, the PTSA fluorescence was mostly quenched , saturat ing to a distinct fluorescent baseline depending on the PTSA : SWNT ratio.
58 Figure 3 8 . Fluorescence Measurement to investigate association kinetcs of PTSA/SWNTs. A K ) Decay of emission peak intensity at 383 nm of various PTSA: SWNT weight ratios as a function of time (1.5hrs) , L ) Decay curves of all solutions with different PTSA : SWNT w eight ratios. E F C D A B
59 Figure 3 8. Continued . Table 3 2 . Decay constant s from fitting of the kinetic curves of the inks with different PTSA:SWNTs weight ratio s 1:10 1:40 1:90 1:105 1:150 1:200 1:250 1:300 1:360 1:400 Fas t decay 64 92 82 72 100 87 99 96 86 122 Slo w decay 1225 2622 3209 1416 1298 9108 4609 3624 2955 4353 The decay curves were fitted well by a biexponential decay of the form: G H I J K L
60 (3 2) These possessed a fast decay time constant t 1 and a slow decay time constant t 2 saturating to different baselines depe ndin g on the concentration (see Fig ure 3 8 L ). A re presentative fit is shown in Figure 3 8 D . The extracted time constants are listed in Table 3 2. All fast time constants were similar as were all the slow time constants, with some deviations, but with no ap parent relation to the nanotube concentrations. Figure 3 9 overlaps the kinetics plots each normalized by its first data point, showing the close correspondence of the curves. As speculation (without further studies), the fast time constant might be relate to the diffusion of PTSA to the immediately available sites (made available by the brief ultrasonication) while the long time constant might be related to the rate at which the metastable TX surfactant spontaneously desorbs from the nanotube surfaces (to be replaced by the PTSA). Figure 3 9 . Decay curves smoothened and normalized for solutions with PTSA : SWNT ranging f rom 1:10 to 1:450 .
61 As shown in Figure 3 10 , emission spectra recorded for the dispersion with PTSA : SWNT = 1 : 105 weight ratio after 7 and 10 days did not show any significant changes compared to that measured after 1.5 h. Thus, it could be concluded tha t the association of PTSA with carbon nanotubes is a fast process, and can be completed within several hours. This was corroborated experimentally. Two batches of the PTSA/SWNT/TX dispersions, containing identical concentrations of the dispersant and nanot ubes, were prepared. One batch was stirred for 4 5 days as was done initially while the other was stirred for 20 h . The dispersions were then dialyzed. Both batches of SWNT inks were stable for over one and a half month s without any visible flocculation. T h ese experiments met the goal of better optimizing the procedures for forming the PTSA based SWNT inks. Figure 3 10. Fluorescence emission spectra of the PTSA/SWNT/TX mixture recorded on day 1 after 1.5 h of emission peak intensity decay measurement, then on day 7 and day 10 . Emission spectrum of PTSA/TX base solution (black curve) excitation = 348 nm .
62 CHAPTER 4 ULTRASONIC SPRAY COATING 4.1Ultrasonic Spray Coating and Home made Spray Station Spray coating in general includes the formation of fine droplets of a coating solution and their deposition on a substrate. In ultr a sonic spray coating a thin layer of liquid driven by capillary force s to an ultrasonically vibrating nozzle is atomized into micron scale droplets. The generated droplets are directed towards atypically heated substrate where they form a coating. Advantages of spray coating are: simplicity and good reproducibility [100, 101] . Most commercially available ultrasonic spray systems move the spray head above the substrate to get areal coverage across the surface . However m oving parts tend to generate particulates . For our purpose of generating SWNT films for electronic and optoelectronic devices particulates can be highly problematic. Often the devices consist of the nanotube layer acting as one electrode separated from another electrode by a thin film measuring 10 s to hundreds of nanometers. If nanotubes run over particles co deposited during their deposition, the nanotubes can protrude above the substrate to tou ch the counter electrode in the final device and create an electrical short. To avoid particles from a mechanism that moves the spray head across the substrate a home built spray station was designed by Matt Gilbert and Prof. Rinzler that keeps the spray h ead stationary but moves the underlying substrate, keeping the mechanical parts in motion below the substrate. A p hotograph of the ultrasonic spray station built in the UF Physics machine shop is shown in F igure 4 1 A . In this system the ultrasonic nozzle sp ray head (Sonoaer 120K50T 120 KHz) and the directing air nozzle are stationary, while a heated sample platform executes the motions. Motion is controlled by linear
63 stage motors (Lexmark) that lie underneath the heated platform below a shield. The 12 x 12 i nch 2 travel of the heated sample platform (temperature is controlled by an Omega CN7500 temperature controller) permits uniform spraying across 10 x 10 inch substrates. A Gilson Pump 307 is used for delivering the SWNT ink to the ultrasonic nozzle . The ink mist is directed to the substrate by a flow of argon gas that entrains the untrasonic mist by the Venturi effect ( F igure 4 1 B ). Figure 4 1 C illustrates the ty p ical spray pattern. The s ubstrate moves along the path indicated, after which it retraces the path backwards . Th is motion pattern is repeated as many times as necessary to deposit the film thickness desired . Indicated in Figure 4 1C is the substrate ( blue block ) and an overspray region in both X and Y directions to ensure uniformity at the substra te edges . Figure 4 1. Ultrasonic spray station system. A) photograph of the home made spray station, B) schematic representation of how atomized ink was redirected and sprayed onto substrates, C) typical spray patterns with substrate moving relative to stationary ultrasonic nozzle. A
64 Figure 4 1. Continued . It is reported that the uniformity of a sprayed film is influenced by a a number ofparameters  : ink flow r ate through the system (R), temperature of the substrate (T), the distance between nozzle and substrate (L) and the moving velocity of the substrate (V) , n itrogen flow that carries the atomized solution to the underlying substrate surface (R ) and spacing b etween tw o adjacent spraying paths (D ) . In our initial sprayed films, the well known coffee ring effect was observed in which the material deposits inhomogeneously from each microdroplet as rings of high density around centers of low density. This was de scribed by Deegan et al. to be the result of capillary flow  where for a drop on a surface the liquid evaporates faster from the pinned contact line of the drople tto be replenished by a flow of the solution outwards from the center of the drop carrying entrained material that deposits in a ring at the boundary . If droplet evaporation from the surface is made rapid there is insufficient time for such flow to be established avoiding the effect. Low substrate temperature, short nozzle/substrate distance, fast ink flow rate and low nozzle sonication power were all fo und to result in such coffee ring deposition. T hese parameters were optimized through n umerous experiments to eventually set tle on R=0.06 m L /min, T=130 Â° C, L= 13 cm , V= 5 cm/s , B C
65 D=1.9 mm, R =3.5 L/h and nozzle sonication power setting at 69%, the X and Y ov erspray was set at X =0.5 cm, y=1.2 cm. The target nanotube surface density and sheet resistance of the sprayed nanotube films in these studies was set by those found to be useful in the carbon nanotube enabled vertical field effect transistor developed i n our group [20, 34, 104] that tradit ionally used filtration fabricated films. Details regarding the performance of devices made using the sprayed nanotube inks optimized as discussed in this chapter are discussed in Chapter 5. 4.2 Spray from Py HPC/SWNTs ink 4.2.1 Py HPC/SWNTs Ink Dilution Py HPC/SWNTs ink prepared through multiple centrifugation , washing, redispersion steps was sprayed onto Benzocyclobutene (BCB) coated glass substrates to form the nanotube film s with sheet resistance values of 4 6 . T o test for ink concentration dep endence on the uniformity of the sprayed films, the Py HPC/SWNTs ink was diluted with DI water , decreas ing the concentration of the ink and increas ing the number of spray repetitions. S olutions contained different nanotube concentration s : C SWNTs =0.1 Âµ g/m L , 0.2 Âµ g/m L , 0.5 Âµ g/m L and 1 .0 Âµ g/m L were made . The u niformity of the sprayed films was test ed using sheet resistance (R s ) measurements (derived from four probe conductivity measurements) and atomic force microscop y (AFM). For the same amount of original P y HPC/SWNTs ink, spray coating time for dilut ed solution with nanotube concentration of C SWNTs = 0.1 Âµ g/m L was ten times longer than that for the solution with C SWNTs = 1 .0 Âµ g/m L. No significant changes i n R s values and SWNT film uniformity on AFM images w ere observed for all tested films . These results
66 indicated that reducing the ink concentration below 1.0 Âµ g/m L SWNT ink does not improve the quality of the sprayed film. Given the hydrophobic ity of the BCB coat ed glass surface, another strategy to improve uniformity of the sprayed films is to decrease surface tension of the spray ed solution for better substrate surface wetting . The Py HPC/SWNTs ink was diluted to the concentration of C SWNTs =1 .0 Âµ g/m L using solvent s that ha ve different surface tension s : 1) DI water with the surface tension water ~ 72 N/m and 2) ethanol with ethanol ~ 22 N/m. Sprayed films fabricated on BCB coated glass substrates using these two diluted SWNT ink solutions with different surface tensions w ere characterized by R s measurement s, AFM , and optical microscopy . F igure 4 2 shows the optical micro graphs and AFM images of the SWNT films sprayed from the SWNT ink solutions diluted with water ( F igure 4 2 A ) and ethanol ( F igure 4 2 B ), respectiv e ly. SWNT film s sp ra yed from the SWNT ink diluted with water were covered with coffee ring shaped depositions . AFM images show that lighter area s found on the optical micrographs contained a high density of aggregat ed nanotube bundles while darker grey area s are barely covered with nan otubes. Black ring edge areas are covered with the excess of dispersant and some nanotubes burried underneath the rings formed during the SWNT ink evaporation. Dilution of the SWNT ink with ethanol yielded more homogeneous film, as shown in Figur e 4 2 B , uniformly covering the substrate (grey area s ) with fewer ring edge lines .
67 Figure 4 2. Optical microscope images (x 5 magnification) and AFM images of the sprayed SWNT films from diluted Py HPC based SWNT inks . A ) ink diluted with water, B ) ink diluted with ethanol . Arrows indicate certain areas on the micrographs that were measured by AFM. A B
68 Better uniformity of the films sprayed using the SWNT ink solutions diluted with ethanol was also supporte d by R s measurement s as shown in Table 4 1. For the same amount of solution sprayed, the films from sol u tions diluted with ethanol show ed five times lower sheet resistance, R s =2 . 98 than the film made from so lutions diluted with water , R s =16 . 6 5 . Besides the reduced surface tension im p roving the surface wetting, a nother possible reason for the impr oved homogeneity is that ethanol has a lower evaporation temperature compar ed to water thereby increasing the evaporation . Table 4 1. Sheet resistance of the SWNT films sprayed the inks dilut ed with water and ethanol . Concentration ( C SWNTs = 1 .0 Âµ g/m L) and volume of the diluted inks were the same. SWNT films sprayed from the inks Sheet resistance ( k ) diluted with water 16 . 6 5 diluted with ethanol 2 . 98 4.2.2 Post treatment of Sprayed Py HPC/SWNTs Films Sprayed Py HPC/SWNTs films were treated with a vari ety of acid s and organic solvent s , includ ing sulfuric acid (10 mM H 2 SO 4 to 1 M H 2 SO 4 ), nitric acid, methanol, acetone and isoproponal baths to attempt removal of the residual Py HPC . AFM images of sprayed films before and after treatment showed that film clean li ness was slightly improved after a 1 M H 2 SO 4 bath but with no visible change after organic solvents. It has been reported  that concentrated nitric acid treatment of the nanotube film s sprayed from a carboxymethyl cellulose based SWNT dispersions removes most of the dispersant (due to acid hydrolysis of the cellulose polymer backbone) . We put as sprayed Py HPC/ SWNT film to 4 M HNO 3 bath overnight then let the film dry in the air for 2 3 h, followed by a rinse with DI water . Figure 4 3 shows optical micro graphs and
69 AFM images of the film before and after cleaning. The a s sprayed film exhibits a considerable amount of the dispersant residue partially covering the nano tubes, resulting in a relative ly high sheet resistance R s =20 . 1 . After the HNO 3 treatment, the circular shape dark lines were mostly removed, leaving a clean grey surface uniforml y cover ed with nanotubes. AFM imag ing after the cleaning shows that the major fraction of the polymer dispersant residue w as removed . There was also some nano tube loss during this treatment. This is unsurprising since nitric acid is a known intercalant of nanotubes. Despite of the SWNT film density loss during acid treatment, the sheet resistance values decreased from the original 20 . 1 to 6 . 6 5 ing an efficient removal of the insulating Py HPC from the tube surfa ces improv ing the tube to tube contact and thus the conductivity.
70 Figure 4 3. Optical microscope images (x 5 magnification) and AFM images of the sprayed Py HPC/SWNTs film s . A ) as sprayed film with R s =20 . 1 k , B ) film after overnight in 4M HNO 3 bath treatment with R s =6 . 6 5 k . 4.2.3 Comparison between Sprayed Films and Filter transferred Films To compare the two solution based SWNTs deposition process es , ultrasonic spray coating and vacuum filtration transferring, the electronic and optical properties of the films fabricated using these two methods , were investigated. Figure 4 4 shows the AFM images of the acid treated sprayed film and traditional filt er transferred film. B oth films showed similar transmittance as high as ~97% at wavelength = 500 nm . Sprayed A Before Treatment R s =20 . 1 k B After Treatment R s =6 . 6 5 k
71 film with slightly higher nano tube densi ty shows higher sheet resistance R 6 than that of the typical filter transferred film with R s~ 4 . The reason for this is that during vacuum filtration the SWNTs tend to lie straighter and get into better contact with each other, coming tog ether in Y junctions as opposed to X junction s that appeared the more often in sprayed films, which induces more tube to tube contact resistance and thus causes the s heet resistance increase  . Another possible reason is tha t the residue of insulating di sp ersant prohibit s the tube to tube contact and affects conductivity of the sprayed film. Figure 4 4. AFM images of the SWNT films fabricated using different deposition process es . A ) sprayed film with Rs 6 , B ) traditional filter transferred film with Rs ~ 4 . Overall, sprayed Py HPC/SWNTs films after acid treatment show comparable electronic and optical properties with t ypical filter transferred film s . Further investigation will be performed by preparing CNTs based electronic devices from both sprayed and filter transferred films. A Sprayed Film B Filter transferred Film
72 4.2.4 Homogeneity Comparison of Sprayedand Filter transferred Films 184.108.40.206 H omogeneity measurement set up Uniformity of the ultrasonic spraying and filter transferring technique was test ed using a film resistance homogeneity measurement station designed and built by my colleges Dr. Bo Liu and Xiao Chen. The h omogeneity test process is as follow s : 1) Substrate Preparation . A 4 glass wafer was coated with 9 blocks of patterned 50 nm Au thick el ec trodes as shown in Figure 4 5 A . Enlargement of the electrode array in one block is shown in Figure 4 5 B . Each block occupies an area of 21.82 mm x 1 9.28mm, consisting of 72 electrodes (the black region s ) arranged as 64 electrode pairs. T he ma gnified area in Figure 4 5 B shows one electrode pair posses s ing a 400 Âµ m gap between the two electrodes . The substrate is subsequently spin coated with BCB to form a thin hydrophobic layer and patterned using lithography to only occupy the region s where there is no Au. 2) . After a blanket deposition of the nanotube film across th e array, the SW NT film was patterned using photo lithography such that it crosses the 4 00 Âµm gap between the electrode pair extending across each electrode pair as indicated by the orange area in the magnified region in Figure 4 5B as a 4 00 Âµm wide strip , r esulting in a 400 Âµ m x 400 Âµ m square, which comprises one nanotube pixel . 3) Homogeneity measurement . The two terminal r esistance each electrode pair was measured by a home made set up that allows recording the resista nce of 8 pairs at a time before the probe must be repositioned to measure the adjacent column of pixels. Because of the square shape of the pixel area, the resistance measured between paired electrodes equals the sheet resistance of nanotube film in that r egion (ignoring contact resistance) .
73 By acquiring statistics on the resistances across such blocks we obtain a measure of the film homogeneity for each method of film deposition . Figure 4 5 . Substrate array for the SWNT film homogeneity measurement. A ) A 4 glass wafer is covered with nine blocks of patterned 50 nm gold electrodes , B ) Enlargement of electrode array in one block area . 72 electrodes (black region) were paired into 64 electrode pairs. Each electrode pair is shown in the magnified area with a 400 Âµ m gap between. For the resistance measurements the SWNT film is patterned as 4 00 Âµm wide strips extending across each electrode pair as indicated by the orange area in magnified region . A B
74 220.127.116.11 Homogeneity discussion SWNTs films having the size of 1 block area were sprayed or filter transferred onto prepared substrates to compare the film uniformity from the two solution based deposition process es . Both films were prepa red targeting sheet resistance of Rs ~ 4 6 . Resistance of 64 pixels in both films were measured and plotted in color plots appearing in Figure 4 6, ranging from 3190 (blue) to 5710 (red). The sprayed film had a lower overall sheet resistance . This is consistent with AFM measurement s showing that the sprayed film has higher nano tube density than the transferred film. For the sprayed film the center of the spray pattern was not set with sufficient precision such that the turn around point ended up too close to one edge of the array. This meant that the last column on one side had less material sprayed on it, as indicated by the increased resistance observed in the right edge column (>4500 ) in Figure 4 6 A . The sheet resistance statistics for the sprayed film is thus based on 56 measured column . These gave a mean resistance of 3567 with a standard deviation (SD) of 176 yielding a relative standard deviation (RSD) that was 4.94% of the mean. The s tatistics are tab ulated in Table 4 2. Compar ed with transferred film with R SD% ~ 5.32% with mean sheet resistance in the similar range (3 5 ), the sprayed film shows slightly better homogeneity, indicating that the ultrasonic spray technique provides at least compara ble film uniformity compared with that of the traditional filter transfer ring technique. Table 4 2 . S tatistics of nanotube pixel resistance from s prayed and filter tran sferred films with the same size of 21.82 mm x 19.28 mm . Average R s ( ) SD ( ) R SD ( % ) Sprayed Film 3567 176 4.94 Transferred Film 4904 261 5.32
75 Figure 4 6. Homogeneity comparison between sprayed and filiter transferred films. A ) Color sheet resistance plot of acid treated sprayed film from Py HPC/SWNT ink, B ) AFM image of acid treated sprayed film , C ) A c olor sheet resistance graph of traditional filter transferred film , D ) AFM image of filter transferred film. Both color plots show resistance of 64 nanotube pixels ranging from 3190 (blue) to 5710 (red). Filter transferred Film Sprayed Film A B C D
76 4.3 Error Analysis of Ultrasonic Spraying and Filter Transferring Process 4.3.1 Calibration Curve: C SWNTs vs. Absorbance To achieve more precise control over ink and film fabrication process es , a calibration curve defining nanotube concentration C SWNTs on the basis of UV V is absorbance spectr a of SWN T solution s was plott ed. Experimental details are as follows: concentrated SWNT dispersion in 1%TX aqueous solution with a known nanotube concentration C SWNTs = 3.75 Âµ g/m L was diluted with different volume s of DI water resulting in solutions with a series of C SWNTs and C TX . UV vis absorption spectr a of these solutions are shown in Figure 4 7. As more DI water is added, characteristic TX signals a = 223 nm, 276 nm and 287 nm gradually decrease as well as t he absorbance baseline ( 300 nm 500 nm) that is mainly determined by C SWNTs . Figure 4 7. UV vis spectrum of diluted SWNT/1%TX solutions with different nanotube s and TX concentration s .
77 Figure 4 8 plots the absorbance intensity of tested solution s at 500 nm , sufficiently distant from absorption bands of TX, as a function of nanotube concentration C SWNTs , showing linear dependence: (4 1) Which means C SWNTs can be easily determined through following equation: (4 2) Th is linear relationship is consistent with Beer Lambert Law and will help to indentify nanotube concentration in solutions. Figure 4 8 . L inear dependence between absorbance of SWNT solution s and C swnts, fittin g equation of this relationship is . 4.3.2 Spray depostion material losses In the vacuum filtration transfer process all the SWNTs that are in the suspension deposit onto the filtration membrane and are subsequently transferred, without loss of SWNT material, to the substrate. This is not tru e for spray deposition where some
78 fraction of the material is lost by deposition beyond the target area or lost , as a condensed layer, on the ultrasonic spray nozzle . In this section these losses are estimated. SWNTs films with a series of effective thick nesses D e (between 4 nm ~ 20 nm) were filter transferred onto BCB coated glass substrates. All films were prepared with the same SWNT/1% TX solution with the same nanotube concentration. The v olume of the solution that was needed for film with target thick ness D t was calculated as follow s : As a first order estimate of the n anotube weight needed for certain target thickness (D t ) was decided by : (4 3) S is the area of the spray pattern , D t is the target thickness and is the nanotube film density ; = 0.71 g/cm 3 as previously reported  . Based on the amount of SW NTs needed, the v olume v of the original S WN Ts ink solution was calculated as: (4 4) where C SWNTs wa s obtained from UV vis absorbance spectr a . After filter transferring onto the substrate, film thickness was measured by AFM step height measurement. This process was only used for t he thickest film D t ~ 25 nm because the nanotube film density 0.71 g/ cm 3 was obtained from the calculation of a 1.5 Âµ m thick film, within which empty space between tubes is largely eliminated , something not true for sub monolayer films that do not fully cover the substrate . For thinner films, an effective film thickness was measured based on the UV Vis transmittance spectra using t he Beer Lambert Law (4 5)
79 W here T 0 and T are the intensity of i ncident radiation and transmitted radiation, respectively; is the absorption coefficient and N is the concentration of attenuating medium; D is the distance light travels through the material, which is the effective nanotube film thickness in our experim ents. For the same nanotube material, both and N are the same. 3 mL 10 Âµ g/mL of SWNTs in 1% TX solutions were filtered through membrane forming a film with the target thickness D t ~ 2 0 nm . AFM step height in multiple places were measured with an average e ffective thickness yielding D e = 24.3 nm . Thinner SWNT films thicknesses (<25 nm) were controlled by the volume of the SWNT/1% TX solutions that was filter deposited . Effective thickness of these films ( D e ) were obtained by calculation of ratios from the Beer Lambert law, (4 6) (4 7) UV Vis spectra of the filter transferred films with D e of 24 .3 nm, 18.9 nm, 13.5 nm, 9.6 nm and 4.2 nm are as shown as the solid lines in Figure 4 9, corresponding to 3 mL, 2.25 mL, 1.8 mL, 1.2 mL and 0.6 mL of the original volume of SWNTs in 1% TX solution, respectively. To estimate the loss es during the spray deposition, the same series of volume s of the original SWNT/1% TX solutions were sprayed onto BCB coated glass substrates. Sprayed films were washed with an overnight 4 M HNO 3 bath; film thicknesses were calculated using the Beer Lambert law ratio and film quality w as further characterized by AFM and sheet resistance measurement s . As shown in Figure 4 9, dashed lines are the transmittance spectra of sprayed films; the transmittance of
80 films with the same volume of SWNT/1% TX solutions are marked with the sa me color as the sprayed films . The latter cons istently show a higher transmittance, representing a substantial material loss during the spray deposition. For example, 3mL of SWNT/ 1% TX solution formed a film with D e = 24.3 nm by filtration, compar ed to an effective thickness of D e = 9.1 nm using spray deposition. Now even this effective thickness measurement is somewhat off because of the turn around points in the spray pattern where the deposited film ends up somewhat thicker, but this error wil l become les s significant as the spray area increases. Figure 4 9. UV V is transmittance spectrum of filter transferred films (solid line) and sprayed film s (dashed line) . Spectra of films with the same volume of original SWNT/1% TX solution were marked with the same color. Loss es during spray deposition was calculated as a percent of effective filter transferred thickness according to x100 (4 8)
81 Figure 4 10 shows the loss% of spray coating technique as a function of effective thickness of filter transferred films corresponding to the same volume of original SWNT/ 1% TX sol ution. The l oss% for the film with D e = 0.68 nm wa s 83.8%, significantly higher than loss% for other films, with a loss% typically of 50% ~ 60%. This is explained by the fact that loss during spray coating deposition is caused by operational / instrumental loss and the material loss beyond target area . The latter one is proportional to the spray repetition steps, and therefore should scale with the film thickness and be a constant after divided by the effective thickness of filter transferred film. With extr emely th in sprayed films, operational/ instrumental loss plays an important role in loss calculation. However, as sprayed thickness increases, influence of operational / instrumental loss diminishes and loss% tends to remain consistent. Figure 4 10. Loss percentage during spray coating deposition as a function of effective thickness of filter transferred film. Sheet resistance measurement s were also used to characterize sprayed/ filter transferred films. Figure 4 11 shows the sheet resistance of the sprayed and filter transferred films as a function of effective thickness. For the same effective thickness, R s of both films are reasonably close with sprayed films being slightly higher than
82 transferred films. This re sistance difference between two type s of films is likely due to two factors: 1) the possibility of minor residue of insulating dispersant left on the nanotube surface after acid treatment resulting in a lower conductivity, 2) vacuum filt e r ed SWNTs tend to lie straighter and get into better contact with each other, coming together in Y junctions, which maximizes tube tube contact area as opposed to X junction s in sprayed films which have more limited tube tube contact at the single crossing point, inducing m ore tube to tube contact resistance . Tube tube contact resistance is the main source of resistance in SWNT thin films  . Detailed i nformation is shown in Table 4 3 . R s vs. D e curve for the sprayed film (black line) was fitted to get the relation: (4 9) F igure 4 11. Sheet resistance of the sprayed (black line) and filter transferred (red line) films as a function of target thickness .
83 Table 4 3 . Sheet resistance comparison between sprayed and filter transferred films with corresponding effective thickness. Effective thickness D e spray (nm) S prayed R s ( ) Effective Thickness D e filter (nm) F ilter transferred R s ( ) 0.68 14641 2 5500 2.5 6640 4.2 1194.9 3.7 2980 9.6 473 6.3 1564 13.5 331.45 8.2 837 18.9 216.74 9.1 586 24.3 168.17 12.7 441 55 49 For nanotube enabled Vertical Field Effect Transistor s (VFETs), which will be discuss ed in the Chapter 5 , the filter tran s ferred nanotube film s that are typical ly used have an effective thickness ~ 2 nm , corresponding to R s ~ 4 Table 4 3 shows that an effective thickness between 2.5 ~3 nm is needed for sprayed films for this desired sheet resistance range, which is consistent with the idea that sprayed films ha ve a slightly poorer conductivity compared with filtered transferred films of the same thickness. This 2.5 ~ 3 nm effective thickness for sprayed films, as predicted in F i gure 4 10 with an loss% of ~70%, corresponds to the effective thickness of fi lter transferred film at D e ~ 7 nm according to Equation 4 8. 4.4 Spray coating from PTSA/SWNT ink 4.4.1 Sprayed Films from PTSA/SWNT ink Sprayed SWNT films were made from the PTSA/SWNTs ink s discussed in Chapter 3. A fter dialysis and brief centrifugation (to remove some macroscopic fibers associated with the dialysis membranes) the PTSA/SWNT s ink was diluted with ethanol so that the nanotube concentration in the mixture for spray coating was C SWNTs =1 Âµ g/m L . As previously discussed, the PTSA/SWNTs i nks wer e all prepared with the same PTSA concentration C PTSA =10 7 M. Lower PTSA : SWNTs weight ratio meant a
84 higher nanotube concentration, so to keep the nanotube spray concentration constant required varying volume s of ethanol for dilution. These dilut ed solutions were sprayed onto BCB coated glass substrates to form films with a target R s between 4 6 k . As shown in Table 4 4 , depend ing on different PTSA : SWNTs weight ratio values in the series of inks, the weight percentage of ethanol range d from 54 .8% to 92.3%. All dilut ed solutions were sprayed under the same spray conditions. Solutions were sprayed at 69% nozzle power onto the BCB coated glass substrate s at a temperature of 130 Â° C. Ink flow rate was R = 0.06 m L /min, nitrogen flow rate wa s R = 3L/h. Overspray areas in X and Y directions were x = 0.5 cm, y = 1.2 cm. The film sprayed from the PTSA : SWNTs = 1 : 105 ink containing larger amount of ethanol ( 82.8% ) show ed substantial conductivity enhancement with R s twice lower than that of the SWNT film sprayed fro m the PTSA : SWNTs = 1 : 40 ink containing less ethanol (54.8%) . This conductivity improvement is due to ethanol improved wetting characteristics , evaporate rate and less insulating dispersant. Surface tension of the mixture increases along with increase of ethanol weight%  . However, further ethanol weight% increases does not causes significant R s changes. Films sprayed from the inks with ratio values of 1:105, 1:200, 1:300 and 1:400 ha d sheet resistance ~ 4 8 k s (Table 4 4 ), which is within the range of typical R s values ne cessary for the SW NTs films used in VFET. Table 4 4 . Effect of the ethanol percentage in the PTSA based inks on the conductivi ty of the sprayed films. PTSA : SWNTs C SWNTs ( Âµ g/m L ) DI ( mL ) Ethanol (mL) Ethanol weight% R s ( ) 1:40 2.54 2.8 4.3 54.8% 12366 1:105 7.06 1 6.1 82.8% 6393 1:200 11.4 0.69 6.41 88.0% 8427 1:300 16.18 0.5 6.6 91.2% 4857 1:400 18.62 0.44 6.66 92.3% 5598
85 Optical m icroscopic images of as sprayed films were obtained to determine film cleanness. The a mount of dark lines caused by the coffee ring effect decrease d with the increase of ethanol mass%. However, a few dark lines still appeared on films that were sprayed from the ink with minim um PTSA : SWNTs weigh ratio 1:400, meaning that post treatment wa s unavoidable. 4.4.2 Post treatment Since PTSA dispersant ha s a good solubility in alcohols , we used ethanol (avoiding concentrated acid needed with Py HP C) for post treatment of the PTSA based sprayed films . The films were characterized using R s , AFM , and UV V is measurement s . Figure s 4 11A and B show the AFM images of the as sprayed and ethanol treated SWNT film s ( PTSA : SWNTs = 1 : 105 ink, target thickne ss of the film D t ~ 7 nm ). As seen from the AFM images most of the dispersant residue w as effectively removed by ethanol bath . This is in a good agreement with sheet resistance measurements: after ethanol treatment the R s dro p ped from >6 to 4 5 k . Another PTSA/SWNT film sprayed under the same condition s was treated with an overnight 4 M HNO 3 bath for comparison . The f ilm was th o roughly cleaned after acid bath, as shown in Figure 4 11C, giving R s ~ 6 k . It could be concluded that much m ilder eth anol treatment allows for removal of the dispersant residue as effici ently as a 4 M nitric acid bath . This could be attribute d to the small quantity of the PTSA dispersant in the SWNT inks and sprayed films and the good solubility of PTSA in ethanol. Comparing ethanol treated sprayed film with filter transferred film with a similar R s ~ 6 k , as shown in Figure 4 11B and D, filter transfer deposition seems to induce more local tube alignment while ultrasonic spray c oa ting shows randomly spread tubes over the whole substrate . The tube density in both films is comparable. The
86 transmittance of the acid treated, ethanol treated sprayed films and a f ilter transferred film is shown in F igure 4 12 . All three films give similar transmittance between 97 .0 ~ 97 .5 % at wavelength = 500 nm . Figure 4 12 . AFM images of the sprayed and filter transferred f ilms before and after treatment . A ) as s prayed PTSA/SWNT film with R s > 6 k , B ) PTSA/SWNT film after overnight ethanol treatment , R s ~ 4 5 k , C ) PTSA/SWNT film after overnight in 4 M HNO 3 bath, R s ~ 6 k , D ) transferred film used for homogeneity test, R s ~ 4 5 k . A B C D
87 F igure 4 13 . Percent t ransmittance spectr a of the acid treated, ethanol treated and filter transferred films, all showing T% ~ 97% at wavelength 500 nm . 4.4.3 Homogeneity Test The u niformity of a more extended area sprayed PTSA/SWNT film was investigated using homogeneity measurement set up discussed in Section 18.104.22.168and is compared w ith transferred films. Nanotube films were ultrasonic sprayed or filter transferred onto a 2 block area of the wafer. The h omogeneity measurement process, including substrate preparation, nanotube pixel deposition and sheet resistance measurement, w ere the same as discussed in Section 22.214.171.124.
88 Figure 4 14 . E lectrode array in 2 block area , 144 electrodes (black region) forms into 128 ele ctrode pairs. Each electrode pair is shown in the magnified area with a 400 Âµ m gap between. For the resistance measurements the SWNT film is patterned as 4 00 Âµm wide strips extending across each electrode pair as indicated by the orange area in magnified region The R s measurement of 128 pixels is shown in color plots in Figure 4 14, ranging betwe en 4240 6960 from blue to red. The majority of pixels in both films are blue/green with R s < 6 k with the exception of the left side column of the sprayed film, which was again due to a centering problem on initiating the spray pattern . The lef t s ide column was accordingly excluded from R s statistic s calculation. The h omogeneity statistics are shown in Table 4 5 . The s prayed film has a mean R s value of 5113 with a RSD that was 6.37% of the mean , compar ed to a mean R s of 4933 with a R SD of 6.82% for transferred film. Homogeneity of the sprayed film is slightly better than filter transferred films. Th ese result s are encouraging for future scale up of spray deposition .
89 Figure 4 15 . Homogeneity comparison between sprayed and filter transferred films. A ) color plot of the R s values of 128 nanotube pixels for sprayed PTSA/SWNT film, B ) a color plot of the R s values of 128 nanotube pixels for transferred film. Each film covers an area of 21.87 mm x 41.51mm with R s ranging from 4240 to 6960 from blue to re d. Table 4 5 . Homogeneity s tatistics of R s of 2 block area nanotube pixels for sprayed and transferred films . Average R s ( ) SD ( ) R SD% Sprayed Film 5113 326 6.37 Transferred Film 4933 336 6.82 A B
90 CHAPTER 5 SPRAYED FILMS FOR VERTICAL FIELD EFFECT TRANSISTORS 5 .1 Fundamentals 5.1.1 Introduction to Vertical Field Effect Transistor The thin film transistor (TFT) has been the dominant transistor choice for the array backplane in active matrix displays for a long time [107 109] . Fig ure 5 1 A s hows typical structure of a bottom gate top contact TFT. The source, drain and active layer are co planar above the dielectric layer and gate. S ource is always grounded while voltage is applied to the drain electrode (V d ) and gate electrode (V g ). Carriers are induced by the applied gate electric field in the active layer near the interface with the dielectric layer  . A simultaneously applied drain voltage generates a drift current that flows between the source and drain. Positive gate voltage V g induces negative charge (electron) injected from grounded electrodes while negative gate V g voltage induces positive charge (hole) injected. The amount of accumulated charges in the interface is proportional to applied gate voltage V g and capacitance of gate dielectric layer C according to capacitance equation (5 1) A ccumulated charges in the interface between the semiconductor and dielectric layer that are induced by V g w ill be driven to flow from source to drain in response to the horizontally applied electrical field V d thus forming the drain current I d . The V g applied must be larger in magnitude than the threshold voltage (V th ) in order to accumulate carriers. V th is af fected by the electronic properties of the active layer, gate/gate dielectric layer and charge traps on and around the dielectric layer . O utput current I d in a TFT , which is determined by many factors , is illustrated in equation  :
91 ( 5 2 ) Where C w is the channel width, C L is the channel length between source and drain electrodes , C i is dielectric layer capacitance, V g is gate voltage, V th is threshold voltage, V d is source drain voltage . I d increases as channel le ngth C L decreases. In order to achieve sufficient output current for active matrix organic light emitting diode ( AMOLED ) backplane application , for example, channel length needs to be short , which requires high resolution lithography that is high ly cost ly and complex . An effective way to resolve this problem is to change the architecture of transistors in an innovative way, as shown in F ig ure 5 1 B  . Transistors with this vertically oriented structure are called vertical field effect transistor s (VFET s ). Fi gure 5 1. Schematic representations of transistors. A ) Structure of t op contact bottom gate thin film tra nsistor. Source, active layer and drain are all co planar. Output current depends on active layer length C L and width C W , B ) Structure of v ertical field effect transistor . A ctive layer is sandwiched by source and drain. The VFET source is designed to be perforated to avoid screening of the gate electric field. Channel length C L now c an be visualized as the thickness of the active layer which is thin film (note however, E q uation 5 2 does not apply to VFET s ). Reprinted with permission from Carbon Nanotube En abled Vertical Field Effect and Light Emitting Transistors (John Wiley and Sons, 2008).
92 VFETs have been found to be a promising solution to both avoid high fabrication expenses and expand the range of potential channel materials. In a VFET, as shown in Fig ure 5 1B , s ource, active layer and drain are all stacked on the dielectric layer one after another. The overlapping region between the source, active layer and drain is the device area C A , with which the output current scales. The active layer length C L that sc ales inversely with current in E quation 5 2 is simply defined by the semiconductor thickness in VFETs . Therefore, instead of using costly high resolution patterning to define sub micron channel lengths (as would be required in the conventional TFT a rchitecture to provide the current required in an AMOLED backplane) , VFET s achieve sufficiently high output current by simply depositing the channel layer as a thin film , which readily results in a nanometer scale channel length . 5 . 1. 2 Working Mechanism o f VFETs Fig ure 5 2 shows the schematic of a working carbon nanotube enabled VFET (CN VFET)  . The source elec trode is grounded while the gate voltage (V g ) is applied to the gate electrode and the drain voltage (V d ) applied to the drain electrode . In operation , a constant V d is applied and V g is being swept between a range of voltage to switch the output current I d on and off. Current between source and gate is referred to as gate leakage current (I g ) and best if kept below 1Ã—10 6 A/cm 2 . Gate leakage current through the dielectric layer usually occurs through defects such as pinholes or contaminants or from the uni ntended application of a n overly large gate voltage resulting in an electric field across the dielectric above its breakdown value.
93 Fig ure 5 2 . Structure of a carbon nanotube enabled VFET . V g and V d are applied with respect to the grounded source. Reprinted with permission from Carbon Nanotube Enabled Vertical Field Effect and Light Emitting Transistors (John Wiley and Sons, 2008). Distinct from the mechanism of a TFT, a VFET is a schottky barrier gated device  . It controls output current by modulating a schottky barrier that exists between the source electrode (CNTs in the case of a CN VFET) and the sem iconductor layer. This is where CNTs as the source electrode became the key innovation in our devices. Because of the low DOS of CNTs, their Fermi level can be easily manipulated by either chemical charge transfer doping or electrical field effect gating [20 , 22, 23] , and therefore modulating the schottky barrier height between source and channel layer . As shown in Figure 5 3, w hen a metal is put into contact with a p type semiconductor, the Fermi level of the semiconductor line s up with the work function of the metal for thermal equilibrium at the interface.
94 Fig ure 5 3 . Schottky barrier formation between a metal and a p type semiconductor. A ) Flat band diagram of the metal (work function q m) and semiconductor (electron affinity q g ) before contact , B ) after co ntacting, Fermi level of semiconducting material line up with work function of metal forming schottky barrier at the interface Charge rearrangement during this process forms a potential gradient within the semiconductors that prohibits any further transfer of charges, with band bending occurring for both conduction and valence bands. This creates a schottky barrier that resists holes moving from the metal to the semiconductor . Assuming the metal work functio n is q m , the electron affinity of t he p type semiconductor is q and its band gap is E g . After contact, the schottky barrier height for holes to move from the metal to the semiconductor is q b = E g ( q m q ) (5 3 ) In CN VFETs, the CNT source serves as a metal whose work function can be shifted in response to the gating electric field to control the transconductance through source semiconductor drain . Both the schottky barrier height and width are modulated A B
95 at the CNT/semiconductor interface to turn the device ON or OFF . The modulation of the barrier height , which is due to the shift of Fermi level of CNTs upon electric field gating, modulates the thermionic emission current . This is in addition to a thinning of the barrier as would be expected when the semiconductor is brought into accumulation under negative gate voltages (for a p channel device). Modulation of the barrier width results in the modulation of the tunneling current . These two phenomena work together t o make the CN VFET very effective as a transistor. Electrostatic simulations have been done in previous work  and confirmed these phenomena. Fig ure 5 4 shows the simulation of gating effect on schottky barrier height at one point (green spot) on CNTs in the CNT/semiconductor ( p type) interface. The barrier height decreases as Vg increases in magnitude from 1v t o 20v along with the thinning of the barrier. N egative gate voltage with increasing magnitude reduces schottky barrier height and width, induc ing drain current and turn s the device ON. On the other hand, increasing positive gate voltage increases barrier height and width and turn s the device OFF. Typical metals have a much larger DOS than CNTs and therefore would not experience Fermi level modulation under electric field gating . This in turn would prevent schottky barrier height modulation , limiting the so ur c e semiconductor drain transconductance control to modulation of schottky barrier width only and making conventional metal based source VFETs inferior to CN VFETs [10, 112] .
96 Fig ure 5 4 . Schottky barrier between one intrinsic point on CNTs and semiconductor layer , both schottky barrier height and width changes along with applied gate voltage. Reprinted with permission from Car bon Nanotube Enabled Vertical Field Effect and Light Emitting Transistors (John Wiley and Sons, 2008). 5 . 1.3Fabrication Process of Organic VFETs Figure 5 5 A shows the typical d evice structure . Fabrication procedure is as follow s: Heavily p doped silicon wi th a 200 nm thermally oxidized layer of SiO 2 was used as received for the gate electrode and dielectric layer, respectively. BCB , as a commercially available, easy to process polymer, was used as the dielectric surface treatment layer. It was diluted with trimethylbenzene and spin coated to form a 5 nm thin film as a dielectric surface treatment layer , which was subsequently hard baked on a hot plate in an argon glovebox at 250 for 1 h to be cross linked and resistant to organic solvents. D ilute CNT soluti on was spray coated or filter transferred onto the substrate as the source electrode with vapor deposited Cr/Au (10/30 nm) at ends as source contact. The substrate was placed on a hot plate in an argon glovebox at 225 for 1 h to dedope the CNTs ( Fermi lev el from 4.9 eV to 4.6 eV) . As purchased di(1 naphthyl) diphenyl 1,10 diphenyl 1,40 diamine)(NPD) with HOMO level ~5.5 eV ,
97 st ructure as shown in Figure 5 5B , was used as the channel material, forming initial schottky barrier height b ~0.9 eV in th e NPD/CNT interface. NPD was thermally evaporated from an effusion cell at 220 with a growth rate of 4 Ã…/s under a pressure of ~7E 7 Torr to form a thickness of 450 nm film. T he substrate was subsequently transferred in an argon glovebox to a metal evapo ration chamber without exposure to ambient. A TEM grid shown in magnified area in Figure 5 5 C w as used as a shadow mask to define hexagonal 30 nm Au pixels (0.035 mm 2 ) as the drain electrode s on top . Figure 5 5 C shows the top view of a substrate with four sections (yellow circles) deposited with drain electrode pixels. The CNT source electrodes are shown as two grey strips. Pixel to pixel uniformity across the substrate was characterized by measuring numerous pixels across all four sections and wi ll be discussed later on in this chapter . Fig ure 5 5 . Structure of Organic VFET. A ) Schematic representation of an NPD channel VFET , B ) chemical structure of NPD, C) top view of a 0.6 " 0.6 substrate with four sections of working pixels. A B C
98 5.1.4 Typical VFET Curves Typical VFET curves include transfer curve, output curve and ON/OFF ratio. Cu rves of the device prepared in S ection 5. 1. 3 are shown in F ig ure 5 6. Fig ure s 5 6 A and B show the typical transfer curve s in linear and log scale. V g sweeps from 6 0 to +20 V while V d is held constant at 5 V . Threshold Voltage (V th ) is the interception of the fitted regression line with the X axis in linear transfer curve. It indicates when the transistor starts to turn on in the linear scale ( OFF to ON ). The log scale transfe r curve clearly shows the ON/OFF ratio of the device. Subthreshold slope (V ss ) is another feature of the device and has units of volts per decade representing how much V g is required to modulate I d one order of magnitude when the device is being turned on : (5 4) It is extracted by fitting a linear regression line with the steepest part of log scale transfer curves ( OFF to ON ). A small V ss exhibits a fast transition between OFF and ON state. The counter clockwise hysteresis results predominantly from charge traps in the dielectric surface treatment layer. According to electrostati c theory, in the ON state (V g = 6 0 V ), accumulated negative ch arge on the gate would induce a compensating amount of positive carriers in the VFET active region, compromised of CNT source and the region of the semic onductor layer nearby the CNTs and dielectric. In addition to the positive carriers a ccumulated in th is active region, trapped positive charge is also present on or near the dielectric surface (in this case BCB). Positive carriers in the active region reduce the schottky barrier turning the device on. When V g is swept from 6 0 V to 0 V, the amount of electr ons on the gate reduces. For electrostatic balance, the positive carriers in the active region should also decrease to the same level. Because of
99 the trapped positive charge in the BCB layer however, more positive charge is removed from the active region t urning the device off earlier. On the opposite, while transistor is fully OFF (V g = 20 V ), positive carriers accumulate in gate, leading corresponding amount of electrons in the active region and BCB. Accumulation of electrons in the active region increase s the schottky barrier, turning the device OFF . With V g sweeping from 20 V to 0 V, positive charge o n the gate decreases, followed by drastic reduction of electrons in the active region. The electric field from trapped electrons affects the active region a s more negative gate voltage than is actually being applied and turns the device ON quicker than without traps. Therefore, because the trapped charge is affecting the transfer curves on both ends of the sweep and in opposite directions, hysteresis is obser ved. Because of its good storage ability for both holes and electrons, BCB gives large hysteresis and has been used for memory devices  . Fig ure 5 6 C and D show the linear linear and log linear output curves of the device pictured in Fig ure 5 5 A . The o utput curve records source drain current with sweepi ng V d from 0 to 8 V and keeping V g constant. A series of sweeps at gate vol tages ranging from the ON (V g = 6 0 V ) to the OFF (V g = 20 V) state were taken. ON/OFF ratio curve shown in Figure 5 6E s hows the ON/OFF cur rent ratio as a function ofon state current density (current per unit drain area).
100 Fig ure 5 6. Typical curves of VFET . A ) T ransfer curve in linear scale , B ) transfer curve in log scale , C ) output curve in linear scale , D ) output curve in log scale , E) ON/OFF ratio VS o n c urrent d ensity . A B C D E
101 5.1.5 Effect of different VFET parameters As a schottky barrier modulated device, the VFET is influenced by multiple device parameters. T h is section talks about key features of some components: semiconducting channel layer and drain electrodes, nanotube networks and dielectric surface treatment la yer. 126.96.36.199 Nanotube bundle effect Because of nanotube aggregation in initial solutions or during deposition process, nanotube network as source electrodes is normally comprised of a bundle distribution ranging in diameter from 1 nm to 9 nm with peak cen tered at ~5 nm  . Some of the tubes, especially those sitting on top of bundles get screened from field gatin g. So if it is initially ON device like F ig ure 5 7 A , positive gate fielding can only switch Ferm i level of CNTs in the bottom ( F ig ure 5 7 B ) leaving a residual amount of current going through the channel between source and drain. In this case leading to a poor ON/OFF ratio. On the opposite side, in accumulation mode, initial barrier height is large enough; no current goes through active layer. Adding negative gate voltage would decrease q b between CNTs in the bottom and semiconductor, turning device ON as shown in F igure 5 7 C . When switching gate voltage to positive, schottky barrier between CNTs and semiconductor layer is increased, turning device fully OFF ( F igure 5 7 D ).Th erefore , h igher ON/OFF ratio is achieved. To get large initial barrier for lower OFF current , CNT films were baked in glovebox at 225 , changing its Fermi level from ~4.9 eV to ~4.6 eV thus increasing the schottky barrier q b .
10 2 Fig ure 5 7 . . A ) normally ON device in its ON state, B ) Positive gate voltage can only switch Fermi level of lower half CNTs, not turning device fully OFF , C ) normally OFF device in its ON state, D ) positiv e gate voltage can fully turn the device OFF . Reprinted with permission from Bo Liu: carbon nanotube enabled vertical field effect transistors and their device derivatives (University of Florida, 2010) . CNT networks need to be made dilute enough to avoid screening from near lying CNT bundles to allow the gate electrical field access to the organic semiconductor. However, the networks must be near or above the percolation threshold of metallic tubes so the sheet resistance is low enough to pass through appr eciable current. 188.8.131.52 HOMO level of semiconducting channel and work function of electrodes Previously built CN VFETs in our lab were n ormally OFF devices , meaning little source drain current occurs when V g equals 0 V . This requires an appreciably large
103 initial barrier height between CNT source and semiconductor layer. To achieve this, for p channel device, the HOMO level of the semiconductor must be several hundred meV (say > 400 meV in magnitude) deeper than the CNT work function. Otherwise, i f HOMO level of semiconductor is sufficiently close (say < 200 meV in magnitude) to the CNT work function , an ohmic contact between the CNT/semiconductor will be formed , transforming the CN VFET to a normally ON device. In our experiments, semiconductors are usua lly chosen with their HOMO level ~ 5.3 eV or even deeper so the initial schottky barrier is large enough that device is normally OFF [34, 104, 113] . The work function of the drain electrode is selected to be close to the HOMO level of the semiconductor to minimize the hole extraction barrier at the interface between the metal drain a nd semiconductor, reaching higher on currents. Au (work function ~ 5.1 eV) is generally used. 184.108.40.206 S urface treatment on dielectric layer Different surface treatments on dielectric layer highly affect device performance. This section discusses some previous experiments by earlier researchers in our group and insights of influence. Fig ure 5 8 shows transfer curves of VFETs with no surface treatment and treated with BCB and SiO 2 . BCB is very stable after cross linking and has excellent ambipolar charge storage capability  . VFET with BCB treated dielectric surface gives lower off current and higher ON/OFF ratio than with Rain X treated or bare SiO 2 . This phenomenon is due to the difference in the charge trap depths in the BCB and SiO 2 surfaces. SiO 2 layer has been proved to posse a large amount of electron traps states because of SiOH Silanol groups  . Electron traps in the BCB layer are deeper than those on the SiO 2 surface. When p ositive gate voltage is applied to turn the device OFF , electrons are induced in the dielectric surface treatment layer ,
104 semiconductor layer and CNTs f or electrostatic balance . Shallower traps in SiO 2 means that gate voltage starts to induce electron traps in SiO 2 at an earlier point when sweep ing gate voltage . For example, when V g sweeps from 100 V to 100 V , electrons could start to be trapped at V g tr ap = 70 V for SiO 2 and V g trap = 90 V for BCB. S tarting from V g = 70 V , negative charges start to be trapped in SiO 2 surface, partially compensating further positive gating and therefore device is less turned OFF . On the other side , with deeper traps in BCB layer, negative charges only start to be trapped when V g reaches to 90 V . Before this point, positive gate electric field is efficiently imposed on CNTs to turn the device more OFF , as shown a lower OFF current for BCB coated VFE T in Figure 5 8 . RainX treated device gives a lower OFF current because the functional groups in it can be bonded with hydroxyl groups in SiO 2 reducing the electron traps in SiO 2 when it is turning OFF . Figure 5 8. Transfer curves of VFETs with different dielectric surface treatment layer . Reprinted with permission from Bo Liu: carbon nanotube enabled vertical field effect transistors and their device derivatives (University of Florida, 2010) .
105 5.2 VFETs with Ultraso nic Sprayed Py HPC/SWNTs Films Electrical properties of acid treated sprayed Py HPC/SWNTs films were tested by building VFETs on top. Device uniformity was also measured as an indicator of uniformity of nanotube films used as the source electrode . The d evic e structure was shown in F igure 5 5 A and the same process ing procedure mentioned in S ection 5.1.3 was used to fabricate the device . Two devices were prepared in parallel for comparison, one with sprayed CNTs (sprayed CNT VFET) , another with transferred CNT s (transferred CNT VFET) , all other conditions were the same. For ultrasonic spraying technique, a hand made PET mask was used to define CNT strips. For filter transferring techniqu e, CNTs were cut into strips before transferring. Ten pixels were measured for both devices. At its first opportunity, 1 out of 10 measured pixels from transferred CNT VFET and 2 out of 10 measured pixels from sprayed showed deviated performance due to the source drain (S D) leakage. Figure 5 9 plotted all remaining pixels for bo th transferred CNT VFETs(black lines) and sprayed CNT VFETs (red lines) in linear A and log scale B , respectively . In one single substrate with either filter transferred or sprayed film, pixels gave highly uniform performance. Device uniformity was tested by analyzing statistics on maximum current density J max , ON/OFF ratio, threshold voltage V th and subthreshold slope V ss . For a conventional TFT the V th is calculated from a linear fit on an I d 1/2 vs. V g plot (based on the gradual channel approximation) a nd defined to be where the extracted regression line intersects the x axis  .
106 Figure 5 9. Comparison of VFETs performance between sprayed Py HPC/SWNT films as source electrode and traditional transferred CNTs as source electrode. 9 out of 10 measured pixels from transferred CNT VFETs (black lines) and 8 out 10 measured pixels from sprayed CNT V FETs (red lines) were plotted. A ) transfer curve in linear scale, V g s wept between 60 v ~ 20 v, V d = 5 v, B ) transfer curve in log s cale, C )output curve comparison between a random chosen pixel from both transferred CN T VFETs and sprayed CNT VFETs, D ) O N/OFF ratio as a function of on current density J, mostly between 5*10 3 ~5*10 4 . Because the CN VFET is a Schottky barrier device, the equations from the gradual channel approximation are not applicable . Instead, V th is extracted from linear J d vs. V g plot in its ON state transfer curve because the J d of the CN VFET follows a nearly linear dependence on V g  . Statistics information is shown in Table 5 1. Sprayed CNT VFETs have an average threshold voltage V th = 28.82 V with a standard A B C D
107 deviation SD = 1.35 V yielding an average deviation that was 4.68% of that mean. Comparing to SD % = 5.18% for transferred CNT VFETs, sprayed CNT VFETs gives slightly more uniform performance. This result is also consistent with subthreshold slope measurement which shows V ss of sprayed CNT VFETs is 4.56% whi ch is lower than that of transferred CNT VFETs V ss = 5.47%. T ransfer curves clearly show that the ON current of transferred CNT VFET is nearly twice the value of sprayed SWNTs VFET while OFF current s in both are similar. J max was measured in transfer curves where V d = 5 V and V g = 60 V . On current difference is also observed from output curve comparison between a randomly chosen pixel from both transferred CNT VFET and sprayed CNT VFET, as shown in F igure 5 9 C . This differen ce in on current leads to a higher ON/OFF ratio for transferred CNT VFETs, nearly 14K comparing to 7K for sprayed CNT VFETs as shown in F igure 5 9 D . Th e lower ON/OFF ratio of the sprayed device may be attributed to the association of the Py HPC residue wit h the CNT surface and may restrict hole injection from the CNT source into the channel layer. Table 5 1. Statistics of 9 out of 10 measured devices with transferred CNT and 8 out of 10 measured devices with sprayed CNTs. max J max (mA/cm 2 ) ON/OFF ratio V th (V) V ss (V/dec) T S T S T S T S av 9.04 4.79 13984 7090 31.29 28.82 5.67 4.39 SD 0.61 0.32 3548 2226 1.62 1.35 0.31 0.20 SD % 6.75% 6.68% 25.37% 31.40% 5.18% 4.68% 5.47% 4.56% * T means device with transferred CNTs and S means device with sprayed CNTs. ON/OFF ratio statistics is calculated at 5 mA/cm 2 on current density. Max J max is measured at V d = 5 V and V g = 60 V . Overall, sprayed Py HPC/SWNTs films maintain good electrical properties given comparable performance of sprayed CNT VFETs with transferred CNT VFETs. Overnight 4 M HNO 3 treatments clean the sprayed films without damaging nanotube properties or inducing extra observable impurities. That 80% of pixels from sprayed -
108 CN T VFET measured were functional without source drain leakage is encouraging comparing to 90% for transferred CNT VFET. With smaller SD % of both threshold voltage V th and subthreshold slope V ss , device with sprayed nanotubes shows slightly higher uniformity , which is promising for future large scale electronic device fabrication. 5.3 VFETsOptimization In this section , several optimizations regarding nanotube cleanness, gate/gate dielectric layer and sprayed film treatment will be investigated to further low er operating voltage and source higher output current. 5.3 .1 CNT Film Cleaness and Optimization of Gate/Gate Dielectric Layer Similar with lateral TFTs, the shorter the channel length, the higher the ON current is . However, the channel layer thickness can only be reduced so far before S D leakage current starts to arise due to impurities in the nanotube film. To solve this issue, oxygen plasma etching method for defining graphene perforated films  was applied to purify and planarize nanotube f ilm s by removing any large size impurities. Planarization steps are as follows: 100 nm Au was deposited onto nanotube film to fully cover the surface for protection. The s ubstrate was then etched in oxygen plasma at 600 sccm/600 w for 30mins to remove any impurities that w ere not covered by the Au layer . Following plasma treatment the Au was subsequently removed by submerging in gold etchant bath for 1min . With planarized CNTs, fewer impurities are expected to be present on the CNT source electrode and ther efore thinner channel layers become possible before S D leakage current takes over. Reducing the c hannel thickness is expected to lead to higher on current.
109 To decrease operating gate voltage, efforts have been focused on redu cing the dielectric thickness d and using a dielectric material with larger dielectric constant to increase gate dielectriccapacitance according to (5 5) Three devices were prep a red for the optimization experiments. To test the effect of planarization of the CNT source electrode two devices were made. The first was c ontrol device VFET1 and wa s comprised of Si/200 nm SiO x as gate/gate dielectric layer, with tra n sferred CNTs as source electrode . The channel material was NPD and was deposited to a thickness of 450 nm . The second device VFET2 was made with a reduced channel thicknes s of 360 nm and had planarized CNTs (processed as described above), with all other layers the same as VFET1. Figure 5 10 shows the transfer curves (A), output curves (B ) and ON/OFF ratio (C ) for VFETs 1 and 2 . As clearly shown in Figure 5 10 B , with slight improve ment in the ON current, VFET2 (red line) shows a dramatic decre a se in the OFF current, from 10 3 mA/cm 2 to 1 0 4 mA/cm 2 , boosting the ON/OFF ratio nearly 2 magnitude, from 10 4 to 10 6 comparing to VFET1 (black line). Planarization of CNTs allowed a thinner channel layer and lead to greatly improved device performance. The third device VFET3 was prepared with ITO/50 nm AlO x as gate/gate d ielectric layer with other conditions the same with VFET2. Aluminum oxide (AlO x ) with dielectric constant  was used to replace SiO x with as the dielectric layer to further decrease operating gate voltage. Glass substrates covered with ITO were used as purchased, respectively. 50 nm AlO x was deposited on ITO through Atomic Layer Dep osition technique. Figure 5 10A shows that t he required g ate
110 voltage sweep range to switch VFET3 (blue line) between fully ON and fully OFF states decreased dramatically from 60 V ~ 20 V to 5 V ~ 3 V with substantially reduced hysteresis. Comparing to VFET2, both ON and OFF current for VFET3 were increased leading to compromised ON/OFF ratio ~10 5 . In general, comparing VFET1 and VFET3, planarizing the CNT source and increasing the gate capacitance drastically reduce hysterisis, increased the ON/OFF ratio and ON currrent and greatly reduced the operating gate voltage. Figure 5 1 0 . Performance comparison of three VFETs. A ) transfer curves, B ) output curves, C ) ON/OFF ratio curves of three VFETs. VFET 1(black lines) is with Si/200 nm SiOx as gate/gate dielectric, 460 nm NPD channel layer and as transferred CNTs. VFET2 (red lines) is with Si/200 nm SiO x gate/gate dielectric, 360 nm NPD channel layer and planarized CNTs. VFET3 (blue lines) is with ITO/50 nm AlO x gate/gate dielectric, 360 nm NPD channel layer and planarized CNTs. A
111 Figure 5 10. Continued . 5.3 .2 Sprayed Films with No Harsh Treatment PTSA/SWNTs Films Another optimization is alleviating the need to use harsh acid treatment during sprayed film cleanness by replacing sprayed Py HPC/SWNT films with PTSA/SWNTs C B
112 films for CN VFET source electrodes deposition. Sprayed PTSA/SWNTs films that were simply purified through overnigh t ethanol bath as discussed in S ection 4.4 showed good film cleanness with comparable conductivity as transferred CNT films. Experiments investigating the electrical properties of films deposited in this way will be described in this section . The electrical properties were explored by using the deposited films as source electrodes in subsequently buil t VFETs . D evice uniformity experiments will also be discussed . Two VFETs were prepared for parallel comparison: one with transferred CNTs and another with sprayed PTSA/SWNTs films as source electrodes . VFETs were processed based on the op timized parameters obtained in S ection 5.3 .1. ITO/50 nm AlO x was used as gate/gate dielectric layer. Ultrasonic sprayed/filter transferred CNTs were planarized before grow ing 360 nm NPD channel materials on top. The d evice structure was shown in Figure 5 5 with the same pro cess ing procedure mentioned in section 5.1.3. 40 pixels distributed randomly in four sections with 10 pixels in each were measured. At its first opportunity, after eliminating pixels that did work due to S D shortage, 39/40 pixels for transferred CNT VFETs and 37/40 pixels for sprayed CNT VFETs were plotted. Overall, transferred CNT VFETs gives more variety and is less uniform compared with sprayed CNT VFETs. To clearly perform comparison between devices, only the best perform ing section from both transferr ed CNT VFETs and sprayed CNT VFETs were chosen and plotted. Figure 5 11 shows the comparison of 9 out of 10 pixels for transferred CNT VFETs (black lines) and 10 out of 10 pixels for sprayed CNT VFETs (red lines) in the best performed section. With optimiz ed device parameters, both VFETs operate at low gate voltage sweep range with small hysteresis.
113 Ethanol treated sprayed PTSA/SWNTs film shows good electrical properties with effectively turning device ON and OFF at ON/OFF ratio >10 4 . VFETs with transferred CNTs have slightly higher on current with half a magnitude lower off current, inducing a higher ON/OFF ratio. However, the result that transferred CNTs show higher ON/OFF ratio was not repeated cons is t e ntly . M ore investigation needs to be carried out . Uni formity analysis of the best perform ing section for both substrates is shown in Table 5 2. Both threshold voltage and subthreshold slope of sprayed CNT VFETs are slightly lower but comparable with those of transferred CNT VFETs. SD% for both devices is wit hin reasonable range which shows that both spray coating and filter transferring technique enable highly uniform VFETs devices . CN VFETs utilizing s prayed PTSA/SWNTs as source electrodes were optimized and improved performance was observed. Through eliminating harsh treatment (concentrated nitric acid baths) in the CNT post spray treatment steps, mechanical damage to the CNTs was avoided . Application of sprayed Py HPC films as source electrodes require d underneath substrate materials resistant to st rong acid treatment, evidently narrowing down qualified material range. For example organic top gate transistors, where nanotube is deposited on to channel materials [111, 117, 118] , most highly mobile organic channel materials are not tolerant in strong acid baths. Removing harsh acid treatment enhances practicability of spray c oating and dramatically increase s the range of compatible materials used for the dielectric surface layer of the CN VFET .
114 Figure 5 1 1 . 10 random pixels in the best performed device area from both transferred CNT VFETs and sprayed CNT VFETs were measured. 9/10 pixels for transferred CNT VFETs (black lines) and 10/10 pixels for sprayed CNT VFETs (red lines) were plotted. One pixel was randomly chosen from each best performed device area to show output curve and ON/OFF ratio charact eristics . A ) T r ansfer curves in linear scale, B ) transfer curves in log scale, C ) output curves, D ) ON/OFF ratio curves. Table 5 2 . Uniformity Statistics fo r 9 out of 10 pixels in transferred CNT VFETs and 10 out of 10 pixels in sprayed CNT VFETs. All pixels were chosen randomly in the substrates best performed section. Jmax (mA/cm2) Vth (v) Vss (v/dec) T S T S T S av 56.50 49.97 2.28 2.19 0.80 0.58 sd 1.64 2.3 0.12 0.11 0.03 0.03 sd% 2.9% 4.6% 5.3% 5% 3.75% 5.2% * T means transferred CNT VFETs, S means sprayed CNT VFETs. A B C D
115 CHAPTER 6 C ONCLUSIONS AND FUTURE WORK In this dissertation, a fully scalable process ultrasonic spray coating of SWNTs was studied to prepare SWNT films and these films compared with those made by the filtration transfer method (ref 2004 Science paper). Pyrene derivatives were chosen as part bonding between pyrene and nanotubes. SWNT inks with two different pyrene derivatives a long chain polymer Py HPC and a small molecule PTSA were studied. PTSA is likely to be preferred because its smaller size made for a less labor intensive dialysis to remove the excess PTSA from the ink before spraying while residual PTSA in the resulting films can be easily and thoroughly washed away by a mild ethanol solvent bath. The disp ersant concentration in PTSA/SWNT ink was successfully minimized using the PTSA fluorescence quenching to monitor the PTSA/SWNT association. The PTSA : SWNT weight ratio could be as low as 1 : 450 with PTSA quenching rate reaches as high as 90%. The PTSA/S WNT association time was greatly shortened from 4 5 days to a few hours based on kinetics measurement. Considering ink stability and dialysis time, the optimized PTSA/SWNT ink had a weight ratio of PTSA : SWNT = 1: 105, dialysis time of 72 h, exhibiting st ability for at least 1.5 months. Sprayed films were characterized by sheet resistance, AFM images and UV Vis absorbance spectra measurement. Sprayed CNT films showed comparable conductivity and optical properties compared with filter transferred films. El ectrical properties of the sprayed CNT films were further investigated by their incorporation into CN VFETs. These devices proved to be comparable in their performance compared to such devices made using filter transferred CNT films. The spray coating tech nique showed
116 comparable uniformity compared to filtration transfer method. This result was concluded from both homogeneity resistance measurement and device uniformity measurement of CN VFETs with sprayed and filter transferred films used as the source ele ctrodes. Future work should include further fluorescence measurements to more fully investigate the PTSA/SWNT fluorescence quenching. During the fluorescence measurements, rather than increasing the C swnt while C TX and C PTSA were held constant, the concen tration of PTSA could be decreased to reduce the PTSA : SWNT weight ratio while the nanotube concentration C swnt and the TX concentration C TX remain constant. In this manner, the influence on the PTSA fluorescence quenching rate caused by nanotube scatteri ng and absorption would be eliminated. Future work could also include Finally, the spray parameters (SWNT concentration, ink flow rate, ultrasonic nozzle power, redirection gas flow rate, substrate temperature and the number of repetitions) were optimized based upon AFM imaging combined with sheet resistance measurements. These parameters could be adjusted to look for their effect on the larger scale uniformity using the homo geneity test wafer and measurement apparatus.
117 REFERENCES 1. Lima, M.D., et al., Electrically, chemically, and photonically powered torsional and tensile actuation of hybrid carbon nanotube yarn muscles. Science. 338 (6109): p. 928 932. 2. Peng, B., et al., Measurements of near ultimate strength for multiwalled carbon nanotubes and irradiation induced crosslinking improvements. Nature nanotechnology, 2008. 3 (10): p. 626 631. 3. Wei, B., R. Vajtai, and P. Ajayan, Reliability and current ca rrying capacity of carbon nanotubes. Applied Physics Letters, 2001. 79 (8): p. 1172 1174. 4. Pop, E., et al., Thermal conductance of an individual single wall carbon nanotube above room temperature. Nano Letters, 2006. 6 (1): p. 96 100. 5. Lipomi, D.J., et al., Skin like pressure and strain sensors based on transparent elastic films of carbon nanotubes. Nature nanotechnology. 6 (12): p. 788 792. 6. Webster, T.J., et al., Nano biotechnology: carbon nanofibres as improved neural and orthopaedic implants. Nanotechnology, 2004. 15 (1): p. 48. 7. Ionescu, A.M. and H. Riel, Tunnel field effect transistors as energy efficient electronic switches. Nature. 479 (7373): p. 329 337. 8. Tans, S.J., A.R. Verschueren, and C. Dekker, Room temperature transistor based on a single carbon nanotube. Nature, 1998. 393 (6680): p. 49 52. 9. Appenzeller, J., et al., Band to band tunneling in carbon nanotube field effect transistors. Physical Review Letters, 2004. 93 (19): p. 196805. 10. Bo Liu, M.A.M., Andrew G. Rinzler, Non Volatile Organic Memory Elements Based on Carbon Nanotube Enabled Vertical Field Effect Transistors. Advanced Functional Materials, 2010. 20 (20): p. 3440 3445. 11. Rueckes, T., et al., Carbon nanotube based nonvolatile random access memory for molecula r computing. Science, 2000. 289 (5476): p. 94 97. 12. Dai, L., et al., Carbon nanomaterials for advanced energy conversion and storage. Small. 8 (8): p. 1130 1166. 13. Evanoff, K., et al., Towards ultrathick battery electrodes: Aligned carbon nanotube enable d architecture. Advanced Materials. 24 (4): p. 533 537. 14. De Volder, M.F., et al., Carbon nanotubes: present and future commercial applications. Science. 339 (6119): p. 535 539.
118 15. R. Saito, G.D.a.M.S.D., Physical Properties of Carbon Nanotubes. Imperial College Press, London, 1998. 16. Hamada, N., S. i. Sawada, and A. Oshiyama, New one dimensional conductors: graphitic microtubules. Physical Review Letters, 1992. 68 (10): p. 1579. 17. Mintmire, J., B. Dunlap, and C. White, Are fullerene tubules metallic? P hysical Review Letters, 1992. 68 (5): p. 631. 18. Ouyang, M., et al., Energy gaps in" metallic" single walled carbon nanotubes. Science, 2001. 292 (5517): p. 702 705. 19. S. G. Louie, M.S.D., G. Dresselhaus and P. Avouris, Springer Verlag, Berlin, Carbon Nan otubes: Synthesis, Structure, Properties, and Applications. 2001: p. 113. 20. Wu, Z., et al., Transparent, conductive carbon nanotube films. Science, 2004. 305 (5688): p. 1273 1276. 21. Wadhwa, P., et al., Electronic junction control in a nanotube semicondu ctor Schottky junction solar cell. Nano Letters, 2010. 10 (12): p. 5001 5005. 22. Graupner, R., et al., Doping of single walled carbon nanotube bundles by Bronsted acids. Physical Chemistry Chemical Physics, 2003. 5 (24): p. 5472 5476. 23. Blackburn, J.L., et al., Transparent Conductive Single Walled Carbon Nanotube Networks with Precisely Tunable Ratios of Semiconducting and Metallic Nanotubes. ACS Nano, 2008. 2 (6): p. 1266 1274. 24. Bower, C., et al., Intercalation and partial exfoliation of single walled ca rbon nanotubes by nitric acid. Chemical Physics Letters, 1998. 288 (2): p. 481 486. 25. Hennrich, F., et al., Reversible modification of the absorption properties of single walled carbon nanotube thin films via nitric acid exposure. Physical Chemistry Chemi cal Physics, 2003. 5 (1): p. 178 183. 26. Kong, J., M.G. Chapline, and H. Dai, Functionalized carbon nanotubes for molecular hydrogen sensors. Advanced Materials, 2001. 13 (18): p. 1384 1386. 27. Kong, J., et al., Nanotube molecular wires as chemical sensors . Science, 2000. 287 (5453): p. 622 625. 28. Hu, L., D.S. Hecht, and G. Gruner, Carbon nanotube thin films: fabrication, properties, and applications. Chemical reviews, 2010. 110 (10): p. 5790 5844. 29. Hata, K., et al., Water assisted highly efficient synthesis of impurity free single walled carbon nanotubes. Science, 2004. 306 (5700): p. 1362 1364.
119 30. Cinke, M., et al., Pore structure of raw and purified HiPco single walled carbon nanotubes. Chemical Physics Letters, 2002. 365 (1): p. 69 74. 31. Guo, T. , et al., Catalytic growth of single walled manotubes by laser vaporization. Chemical Physics Letters, 1995. 243 (1): p. 49 54. 32. Rinzler, A., et al., Large scale purification of single wall carbon nanotubes: process, product, and characterization. Applie d Physics A: Materials Science & Processing, 1998. 67 (1): p. 29 37. 33. McCarthy, M., et al., Low voltage, low power, organic light emitting transistors for active matrix displays. Science. 332 (6029): p. 570. 34. McCarthy, M.A., B. Liu, and A.G. Rinzler, H igh current, low voltage carbon nanotube enabled vertical organic field effect transistors. Nano Letters, 2010. 35. Wadhwa, P., et al., Electrolyte induced inversion layer Schottky junction solar cells. Nano Letters, 2011. 11 (6): p. 2419 2423. 36. Krstic, V. , et al., Langmuir Blodgett films of matrix diluted single walled carbon nanotubes. Chemistry of materials, 1998. 10 (9): p. 2338 2340. 37. Lim, S.C., et al., Extracting independently the work function and field enhancement factor from thermal field emissio n of multi walled carbon nanotube tips. Carbon, 2005. 43 (13): p. 2801 2807. 38. Yang, P., Nanotechnology: Wires on water. Nature, 2003. 425 (6955): p. 243 244. 39. Kim, Y., et al., Langmuir Blodgett films of single wall carbon nanotubes: layer by layer depo sition and in plane orientation of tubes. Japanese journal of applied physics, 2003. 42 (12R): p. 7629. 40. Saran, N., et al., Fabrication and characterization of thin films of single walled carbon nanotube bundles on flexible plastic substrates. Journal of the American Chemical Society, 2004. 126 (14): p. 4462 4463. 41. Rao, S.G., et al., Nanotube electronics: large scale assembly of carbon nanotubes. Nature, 2003. 425 (6953): p. 36 37. 42. Seemann, L., A. Stemmer, and N. Naujoks, Local surface charges direct the deposition of carbon nanotubes and fullerenes into nanoscale patterns. Nano Letters, 2007. 7 (10): p. 3007 3012. 43. Sharma, R., et al., Nanometer positioning, parallel alignment, and placement of single anisotropic nanoparticles using hydrodynamic forc es in cylindrical droplets. Nano Letters, 2007. 7 (9): p. 2693 2700.
120 44. De Gennes, P. G., Wetting: statics and dynamics. Reviews of modern physics, 1985. 57 (3): p. 827. 45. LeMieux, M.C., et al., Self sorted, aligned nanotube networks for thin film transist ors. Science, 2008. 321 (5885): p. 101 104. 46. Seo, M., et al., Terahertz electromagnetic interference shielding using single walled carbon nanotube flexible films. Applied Physics Letters, 2008. 93 (23): p. 231905. 47. Lay, M.D., J.P. Novak, and E.S. Snow, Simple route to large scale ordered arrays of liquid deposited carbon nanotubes. Nano Letters, 2004. 4 (4): p. 603 606. 48. Dan, B., G.C. Irvin, and M. Pasquali, Continuous and scalable fabrication of transparent conducting carbon nanotube films. ACS Nano, 2009. 3 (4): p. 835 843. 49. Hu, L., et al., Flash dry deposition of nanoscale material thin films. Journal of Materials Chemistry, 2009. 19 (32): p. 5845 5849. 50. Kamat, P.V., et al., Self assembled linear bundles of single wall carbon nanotubes and their alignment and deposition as a film in a dc field. Journal of the American Chemical Society, 2004. 126 (34): p. 10757 10762. 51. Lima, M.D., et al., Thin, conductive, carbon nanotube networks over transparent substrates by electrophoretic deposition. Journal of Materials Chemistry, 2008. 18 (7): p. 776 779. 52. Vijayaraghavan, A., et al., Ultra large scale directed assembly of single walled carbon nanotube devices. Nano Letters, 2007. 7 (6): p. 1556 1560. 53. Krupke, R., et al., Thin films of metallic ca rbon nanotubes prepared by dielectrophoresis. Advanced Materials, 2006. 18 (11): p. 1468 1470. 54. Tenent, R.C., et al., Ultrasmooth, Large Area, High Uniformity, Conductive Transparent Single Walled Carbon Nanotube Films for Photovoltaics Produced by Ultra sonic Spraying. Advanced Materials, 2009. 21 (31): p. 3210 3216. 55. Hsieh, Y. P., et al., Direct deposition of single walled carbon nanotube thin films via electrostatic spray assisted chemical vapor deposition. Nanotechnology, 2009. 20 (6): p. 065601. 56. Zhao, X., et al., Spray deposition of steam treated and functionalized single walled and multi walled carbon nanotube films for supercapacitors. Nanotechnology, 2009. 20 (6): p. 065605.
121 57. Bahun, G.J., C. Wang, and A. Adronov, Solubilizing single walled ca rbon nanotubes with pyrene functionalized block copolymers. Journal of Polymer Science Part A: Polymer Chemistry, 2006. 44 (6): p. 1941 1951. 58. Karthigeyan, A., N. Minami, and K. Iakoubovskii, Highly sensitive, room temperature gas sensors prepared from c ellulose derivative assisted dispersions of single wall carbon nanotubes. Japanese journal of applied physics, 2008. 47 (9R): p. 7440. 59. Piret, J.P., et al., Dispersion of multi walled carbon nanotubes in biocompatible dispersants. Journal of Nanoparticle Research, 2009. 12 (1): p. 75 82. 60. Tasis, D., et al., Chemistry of carbon nanotubes. Chemical reviews, 2006. 106 (3): p. 1105 1136. 61. Kudin, K.N., H.F. Bettinger, and G.E. Scuseria, Fluorinated single wall carbon nanotubes. Physical Review B, 2001. 63 ( 4): p. 045413. 62. Thostenson, E.T., Z. Ren, and T. W. Chou, Advances in the science and technology of carbon nanotubes and their composites: a review. Composites science and technology, 2001. 61 (13): p. 1899 1912. 63. Andrews, R., et al., Multiwall carbon nanotubes: synthesis and application. Accounts of Chemical Research, 2002. 35 (12): p. 1008 1017. 64. Zhang, J., et al., Photoluminescence and electronic interaction of anthracene derivatives adsorbed on sidewalls of single walled carbon nanotubes. Nano Let ters, 2003. 3 (3): p. 403 407. 65. Islam, M., et al., High weight fraction surfactant solubilization of single wall carbon nanotubes in water. Nano Letters, 2003. 3 (2): p. 269 273. 66. Moore, V.C., et al., Individually suspended single walled carbon nanotube s in various surfactants. Nano Letters, 2003. 3 (10): p. 1379 1382. 67. Katz, E. and I. Willner, Biomolecule functionalized carbon nanotubes: applications in nanobioelectronics. ChemPhysChem, 2004. 5 (8): p. 1084 1104. 68. Pasquali, M., Gelation: Grow with the flow. Nature materials, 2010. 9 (5): p. 381 382. 69. Saha, D., K. Ray, and T. Misra, Energy transfer in triton X 100 micelles: a fluorescence study. SpectrochimicaActa Part A: Molecular and Biomolecular Spectroscopy, 2000. 56 (4): p. 797 801. 70. Phillies, G .D. and J.E. Yambert, Solvent and solute effects on hydration and aggregation numbers of Triton X 100 micelles. Langmuir, 1996. 12 (14): p. 3431 3436.
122 71. Brown, W., et al., Static and dynamic properties of nonionic amphiphile micelles: Triton X 100 in aque ous solution. The Journal of Physical Chemistry, 1989. 93 (6): p. 2512 2519. 72. O'connell, M.J., et al., Band gap fluorescence from individual single walled carbon nanotubes. Science, 2002. 297 (5581): p. 593 596. 73. Richard, C., et al., Supramolecular self assembly of lipid derivatives on carbon nanotubes. Science, 2003. 300 (5620): p. 775 778. 74. Vaisman, L., H.D. Wagner, and G. Marom, The role of surfactants in dispersion of carbon nanotubes. Advances in colloid and interface science, 2006. 128 : p. 37 46. 75. Liu, J. F. and W.A. Ducker, Self assembled supramolecular structures of charged polymers at the graphite/liquid interface. Langmuir, 2000. 16 (7): p. 3467 3473. 76. Wanless, E.J. and W.A. Ducker, Organization of sodium dodecyl sulfate at the graphi te solution interface. The Journal of Physical Chemistry, 1996. 100 (8): p. 3207 3214. 77. Tasis, D., et al., Single walled carbon nanotubes decorated with a pyrene fluorenevinylene conjugate. Nanotechnology, 2009. 20 (13): p. 135606. 78. Zhang, M., et al., Production of graphene sheets by direct dispersion with aromatic healing agents. Small, 2010. 6 (10): p. 1100 1107. 79. Yang, Q., et al., Functionalization of multiwalled carbon nanotubes by pyrene labeled hydroxypropyl cellulose. The Journal of Physical Ch emistry B, 2008. 112 (41): p. 12934 12939. 80. Yang, Q. and X. Pan, Preparation and characterization of water soluble single walled carbon nanotubes by hybridization with hydroxypropyl cellulose derivatives. Industrial & Engineering Chemistry Research, 2010 . 49 (6): p. 2747 2751. 81. Lou, X., et al., Synthesis of pyrene containing polymers and noncovalent sidewall functionalization of multiwalled carbon nanotubes. Chemistry of materials, 2004. 16 (21): p. 4005 4011. 82. Schagerlof, H., et al., Characterization of chemical substitution of hydroxypropyl cellulose using enzymatic degradation. Biomacromolecules, 2006. 7 (1): p. 80 85. 83. Piret, J. P., et al., Dispersion of multi walled carbon nanotubes in biocompatible dispersants. Journal of Nanoparticle Research. 12 (1): p. 75 82.
123 84. Star, A., et al., Preparation and properties of polymer wrapped single walled carbon nanotubes. AngewandteChemie International Edition, 2001. 40 (9): p. 1721 1725. 85. Steuerman, D.W., et al., Interactions between conjugated polymers and single walled carbon nanotubes. The Journal of Physical Chemistry B, 2002. 106 (12): p. 3124 3130. 86. Ehli, C., et al., Interactions in single wall carbon nanotubes/pyrene/porphyrinnanohybrids. Journal of the American Chemical Society, 2006. 128 (34): p. 1 1222 11231. 87. Jonsson Niedziolka, M., A. Kaminska, and M. Opallo, Pyrene functionalised single walled carbon nanotubes for mediatorlessdioxygenbioelectrocatalysis. ElectrochimicaActa, 2010. 55 (28): p. 8744 8750. 88. Winnik, F.M., Association of hydrophobi c polymers in water: fluorescence studies with labeled (hydroxypropyl) celluloses. Macromolecules, 1989. 22 (2): p. 734 742. 89. Zachariasse, K.A., et al., The kinetic parameters for intramolecularexcimer formation from the meso and racemic diastereoisomers of 2, 4 DI (2 pyrenyl) pentane: evidence for rapid conformer interconversion. Chemical Physics Letters, 1987. 133 (5): p. 390 398. 90. Winnik, F.M., Photophysics of preassociatedpyrenes in aqueous polymer solutions and in other organized media. Chemical re views, 1993. 93 (2): p. 587 614. 91. Yamazaki, I., et al., Excimer formation in pyrene labeled hydroxypropyl cellulose in water: picosecond fluorescence studies. Journal of Physical Chemistry, 1987. 91 (16): p. 4213 4216. 92. Gilmore, F.R., R.R. Laher, and P .J. Espy, Franck Condon Factors, r Centroids, Electronic Transition Moments, and Einstein Coefficients for Many Nitrogen and Oxygen Band Systems. Journal of physical and chemical reference data, 1992. 21 (5): p. 1005 1107. 93. Ashutosh Sharma, S.G.S., Intro duction to Fluorescence Spectroscopy. 1999. 94. Zagrobelny, J., T.A. Betts, and F.V. Bright, Steady state and time resolved fluorescence investigations of pyreneexcimer formation in supercritical CO2. Journal of the American Chemical Society, 1992. 114 (13) : p. 5249 5257. 95. Birks, J.B., Photophysics of Aromatic Molecules. Wiley Interscience, New York, 1970.
124 96. Winnik, F.M., et al., Synthesis and characterization of pyrene labeled hydroxypropyl cellulose and its fluorescence in solution. Macromolecules, 1987. 20 (1): p. 38 44. 97. Winnik, F.M., M.A. Winnik, and S. Tazuke, Interaction of hydroxypropylcellulose with aqueous surfactants: fluorescence probe studies and a look at pyrene labeled polymer. Journal of Physical Chemistry, 1987. 91 (3 ): p. 594 597. 98. Hecht, D., L. Hu, and G. Gruner, Conductivity scaling with bundle length and diameter in single walled carbon nanotube networks. Applied Physics Letters, 2006. 89 (13): p. 133112 133112 3. 99. Cummings, E.A., et al., Spectroscopic study o f the ternary complex of beta cyclodextrin, pyrene, and triton X 100. Applied spectroscopy, 1993. 47 (12): p. 2129 2134. 100. Pham, N.P., J.N. Burghartz, and P.M. Sarro, Spray coating of photoresist for pattern transfer on high topography surfaces. Journal of Micromechanics and Microengineering, 2005. 15 (4): p. 691. 101. Pham, N.P., et al., Photoresist coating methods for the integration of novel 3 D RF microstructures. Microelectromechanical Systems, Journal of, 2004. 13 (3): p. 491 499. 102. Bose, S., et al. , Process optimization of ultrasonic spray coating of polymer films. Langmuir, 2013. 29 (23): p. 6911 6919. 103. Deegan, R.D., et al., Capillary flow as the cause of ring stains from dried liquid drops. Nature, 1997. 389 (6653): p. 827 829. 104. Liu, B., et al., Carbon Nanotube Enabled Vertical Field Effect and Light Emitting Transistors. Advanced Materials, 2008. 20 (19): p. 3605 3609. 105. Das, R.K., et al., Engineered Macroporosity in Single Wall Carbon Nanotube Films. Nano Letters, 2009. 9 (2): p. 677 683. 1 06. Vazquez, G., E. Alvarez, and J.M. Navaza, Surface tension of alcohol water+ water from 20 to 50. degree. C. Journal of chemical and engineering data, 1995. 40 (3): p. 611 614. 107. Brody, T.P., The birth and early childhood of active matrix A personal m emoir. Journal of the Society for Information Display, 1996. 4 (3): p. 113 127. 108. Sheraw, C., et al., Organic thin film transistor driven polymer dispersed liquid crystal displays on flexible polymeric substrates. Applied Physics Letters, 2002. 80 (6): p. 1088 1090.
125 109. Kawasaki, M., et al., Thin film transistor and matrix display device . 2003, Google Patents. 110. Zaumseil, J. and H. Sirringhaus, Electron and ambipolar transport in organic field ffect transistors. ChemInform, 2007. 38 (29): p. no no. 111. Zaumseil, J. and H. Sirringhaus, Electron and ambipolar transport in organic field effect transistors. Chemical reviews, 2007. 107 (4): p. 1296 1323. 112. Xu, Z., et al., Vertical organic light emitting transistor. Applied Physics Letters, 2007. 91 (9): p. 0 92911. 113. McCarthy, M.A., et al., Reorientation of the high mobility plane in pentacene based carbon nanotube enabled vertical field effect transistors. ACS Nano. 114. Chua, L. L., et al., General observation of n type field effect behaviour in organic s emiconductors. Nature, 2005. 434 (7030): p. 194 199. 115. Lemaitre, M.G., et al., Improved transfer of graphene for gated Schottky junction, vertical, organic, field effect transistors. ACS Nano, 2012. 6 (10): p. 9095 9102. 116. Guha, S., et al., High qualit y aluminum oxide gate dielectrics by ultra high vacuum reactive atomic beam deposition. Journal of applied physics, 2001. 90 (1): p. 512 514. 117. Zaumseil, J., et al., Efficient Top Gate, Ambipolar, Light Emitting Field Effect Transistors Based on a Green Light Emitting Polyfluorene. Advanced Materials, 2006. 18 (20): p. 2708 2712. 118. Facchetti, A., M.H. Yoon, and T.J. Marks, Gate dielectrics for organic field effect transistors: new opportunities for organic electronics. Advanced Materials, 2005. 17 (14): p. 1705 1725.
126 BIOGRAPHIC SKETCH Yu Shen was born in 198 7 in Nantong , Jiangsu , P. R. China. Yu has always been a keen kid with lots of passion for science and technology. Her father was a physics teacher who showed her great knowledge in science at her early age. Before joining high school, Yu spent her wonderful and happy childhood in a small town with her grandmoth er, an elegant and knowledgeable lady who Yu learned to be a spiritual and honest person as she grew up . Yu went to Rudong H igh School which was one of the best high schools in Jiangsu province at the age of 16 . She took part in the International Physics Olympiads during her high school and found herself felt in love with physics. After three years study Yu got enrolled in Nanjing University, one of the best universities in China. Yu chose physics as her major. Yu had a busy but very fulfilling life durin g her four years study in the university. As the vice president of student union in physics department, she organized the first national college physics conference. During the fall semester on her junior year Yu joined Prof. Shining Zhu s research group, w here she learned a lot of theoretical and experimental skills in condensed matter physics. I n 2009, Yu went to University of Florida to pursue her PhD degree in p hysics . During the first one and a half year s , Yu finished studying fundamentals of graduate level physics with top notched performance among her peers. Yu joined in Prof. Andrew G. Rinzler s lab in the spring of 2011 . There Yu worked on preparing and optimizing single wall carbon nanotube inks and f abrication of clean, ultra thin, and highly uniform sprayed nanotube films that subsequently used in electronic devices. Yu loved her research and enjoyed the graduate life in Florida.
127 At the summer of 2011, Yu also enrolled in Master of Science in Manage ment program in UF. Yu joined this program for her other personal interest. She learned multiple personal skills and obtained diverse knowledge background.