Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Controlling alignment and orientation in carbon nanotube films fabricated using inkjet printing Beyer, Simon Travis 2011

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


24-ubc_2012_spring_beyer_simon.pdf [ 42.57MB ]
JSON: 24-1.0072517.json
JSON-LD: 24-1.0072517-ld.json
RDF/XML (Pretty): 24-1.0072517-rdf.xml
RDF/JSON: 24-1.0072517-rdf.json
Turtle: 24-1.0072517-turtle.txt
N-Triples: 24-1.0072517-rdf-ntriples.txt
Original Record: 24-1.0072517-source.json
Full Text

Full Text

Controlling Alignment and Orientation in Carbon Nanotube Films Fabricated Using Inkjet Printing  by Simon Travis Beyer B.A.Sc., The University of British Columbia, 2009  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  Masters of Applied Science in THE FACULTY OF GRADUATE STUDIES (Electrical and Computer Engineering)  The University Of British Columbia (Vancouver) December 2011 c Simon Travis Beyer, 2011  Abstract Carbon nanotubes (CNTs) are a fascinating material with a diverse set of electrical and mechanical properties. In particular, they exhibit either metallic or semiconducting behaviour depending on their size and molecular configuration. This ability to tune their electrical properties makes CNTs applicable to a diverse range of electronics applications. One of the long-standing barriers of their application is in the difficulty of controlling their position and orientation on a substrate. This is not only true for applications utilizing individual CNTs, but also for large-scale electronics such as sensors and displays, where CNT thin films are of great interest. The CNTs in such films are typically randomly entangled and do not exhibit the same excellent properties observed in individual CNTs. However, when the CNTs possess long range mutual alignment the electrical properties of the film can be improved. The objective of this research was to design a process for fabricating films of mutually aligned CNTs with a controlled orientation using solely inkjet printing. The innate lyotropic liquid crystallinity of highly concentrated CNT suspensions was used here as a mechanism of achieving long-range mutual alignment. The CNT orientation was found to depend on the evaporation behaviour, which was dictated by the printed pattern. A unique printing scheme was developed in order to achieve the necessary high concentrations for a lyotropic liquid crystalline phase transition, and the morphology of the resulting films was studied using scanning electron microscope (SEM) and polarized light microscopy. It was found that the alignment does not necessarily persist throughout the depth of the film, but is strikingly evident on its surface. In order to both isolate the aligned surface layer and investigate the sub-surface morphology, a method was developed for removing thin consecutive layers from the film using a polydimethylsiloxane (PDMS) stamp. ii  SEM results indicated that the CNT film morphology was one of stacked layers, each exhibiting a decreasing degree of alignment. These results were supported by polarized light microscopy and are suggestive of a smectic liquid crystal structure.  iii  Table of Contents Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  ii  Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  iv  List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  vi  List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  vii  Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  x  Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  xii  1  2  Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  1  1.1  Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  3  1.2  Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  5  1.3  Prior Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  6  1.3.1  Suspension and Purification . . . . . . . . . . . . . . . .  6  1.3.2  Stabilization in Solution . . . . . . . . . . . . . . . . . .  6  1.3.3  Purification . . . . . . . . . . . . . . . . . . . . . . . . .  7  1.3.4  Alignment Mechanisms . . . . . . . . . . . . . . . . . .  11  Carbon Nanotube Film Fabrication . . . . . . . . . . . . . . . . . .  14  2.1  Liquid Crystal Fundamentals . . . . . . . . . . . . . . . . . . . .  15  2.2  Solution Processing of CNT Suspensions . . . . . . . . . . . . .  19  2.2.1  Aqueous Stabilization . . . . . . . . . . . . . . . . . . .  19  2.2.2  Ultrasonication . . . . . . . . . . . . . . . . . . . . . . .  20  2.2.3  Ultracentrifugation . . . . . . . . . . . . . . . . . . . . .  23  iv  2.3  Film Fabrication . . . . . . . . . . . . . . . . . . . . . . . . . . .  24  2.3.1  Evaporation Model . . . . . . . . . . . . . . . . . . . . .  27  2.3.2  Inkjet Printing Parameters . . . . . . . . . . . . . . . . .  35  2.4  Substrate Considerations . . . . . . . . . . . . . . . . . . . . . .  38  2.5  Peeling Procedure . . . . . . . . . . . . . . . . . . . . . . . . . .  41  Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . .  44  3.1  CNT Characterization Techniques . . . . . . . . . . . . . . . . .  44  3.1.1  SEM . . . . . . . . . . . . . . . . . . . . . . . . . . . .  45  3.1.2  AFM . . . . . . . . . . . . . . . . . . . . . . . . . . . .  46  3.1.3  Polarized Light Microscopy . . . . . . . . . . . . . . . .  46  3.2  Surface Morphology . . . . . . . . . . . . . . . . . . . . . . . .  48  3.3  Contributing Parameters . . . . . . . . . . . . . . . . . . . . . .  62  3.3.1  Contaminants . . . . . . . . . . . . . . . . . . . . . . . .  62  3.3.2  Length . . . . . . . . . . . . . . . . . . . . . . . . . . .  64  3.3.3  Concentration . . . . . . . . . . . . . . . . . . . . . . . .  66  Sub-Surface Morphology . . . . . . . . . . . . . . . . . . . . . .  70  Conclusion and Future Work . . . . . . . . . . . . . . . . . . . . . .  78  4.1  Summary of the Results . . . . . . . . . . . . . . . . . . . . . . .  78  4.2  Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  79  4.2.1  Process Optimization . . . . . . . . . . . . . . . . . . . .  79  4.2.2  Physical Understanding . . . . . . . . . . . . . . . . . .  81  4.2.3  Film Analysis . . . . . . . . . . . . . . . . . . . . . . . .  82  4.2.4  Peeling Procedure . . . . . . . . . . . . . . . . . . . . .  82  4.2.5  Applications . . . . . . . . . . . . . . . . . . . . . . . .  83  Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  85  A MATLAB Script . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  97  3  3.4 4  B Common Particles in SEM . . . . . . . . . . . . . . . . . . . . . . . 101  v  List of Tables Table 2.1  The sonication parameters used for each CNT length regime. .  Table 2.2  The centrifugation parameters used for each CNT length regime. 24  Table 2.3  Experimental details of each concentration regime. . . . . . . .  Table 2.4  The inkjet printing parameters used for different nozzle orifice sizes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  vi  22 35 38  List of Figures Figure 1.1  A graphene lattice showing the different chiral vectors of a CNT.  Figure 2.1  Common types of liquid crystals. . . . . . . . . . . . . . . .  15  Figure 2.2  Disclinations with different s values. . . . . . . . . . . . . . .  17  Figure 2.3  The theoretically calculated phase diagram for CNTs of different aspect ratios. . . . . . . . . . . . . . . . . . . . . . . . .  Figure 2.4  18  An SEM image showing the highly bundled nature of the CNT soot. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  Figure 2.5  2  20  The radial fluid velocity profile around an imploding cavitation center. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  21  Figure 2.6  CNTs before (a) and after (b) ultracentrifugation. . . . . . . .  23  Figure 2.7  A typical example of the images used to calculate the CNT length distributions. . . . . . . . . . . . . . . . . . . . . . . .  Figure 2.8  25  The distributions CNT lengths from the (a) long, (b) medium, and (c) short length regimes. . . . . . . . . . . . . . . . . . .  26  A printed line with a bulge deformation (a), and without (b). .  27  Figure 2.10 A schematic showing the boundary conditions of the model. .  29  Figure 2.9  Figure 2.11 The saturated water vapor concentration in air at a given temperature. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  30  Figure 2.12 The numerical solution for the total evaporation flux across the surface of a printed line at different mesh sizes. . . . . . . . .  31  Figure 2.13 The deposition rate required to match the spatially varying evaporation rate of a printed line. . . . . . . . . . . . . . . . .  32  Figure 2.14 The rastorized printing pattern used to create a CNT film . . .  32  Figure 2.15 A schematic of the self regulated cooling module. . . . . . . .  34  vii  Figure 2.16 The waveform used to actuate the piezoelectric printer nozzle. The percentages indicate the relative contribution of each region to the total pulse width. . . . . . . . . . . . . . . . . . .  38  Figure 2.17 A schematic of the custom inkjet printing system used. . . . .  39  Figure 2.18 The contact angle of a processed CNT sodium cholate (SC) suspension on polyethylene terephthalate (PET) (a) before, and (b) after a 10 s plasma treatment. . . . . . . . . . . . . . . . .  40  Figure 2.19 Film thickness before (grey) and after (green) a layer is peeled away and the resulting difference in film thickness (red). . . .  42  Figure 2.20 An SEM image of a CNT film which has been removed with scotch tape and laminated on PET. . . . . . . . . . . . . . . .  43  Figure 3.1  A schematic of the birefringence effect. . . . . . . . . . . . .  47  Figure 3.2  SEM images of CNT alignment in an inkjet printed line . . .  49  Figure 3.3  SEM images of CNT alignment in an inkjet printed circle. . .  50  Figure 3.4  Drop cast CNT droplets on copper and plasma treated PET. . .  51  Figure 3.5  Change in contact angle and CNT concentration in the pinned evaporation regime on plasma treated PET and copper substrates. 52  Figure 3.6  A 5 layer inkjet printed CNT line on plasma treated PET. . . .  Figure 3.7  A schematic of the observed evaporation behaviour of an inkjet  53  printed line and the resulting CNT orientation. . . . . . . . .  54  Figure 3.8  Several common disclinations in the CNT film surface. . . . .  55  Figure 3.9  Quantifying CNT orientation through automated image analysis. 56  Figure 3.10 The surface layer of an aligned CNT under crossed polarizers.  57  Figure 3.11 The end of the surface layer of a printed line under crossed polarizers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . Figure 3.12 The two possible  45◦ orientations  58  the CNTs can take with re-  spect to the incident polarized light. . . . . . . . . . . . . . .  59  Figure 3.13 The middle of the surface layer of a printed line under crossed polarizers. . . . . . . . . . . . . . . . . . . . . . . . . . . . .  60  Figure 3.14 A printed CNT line in which the ends fully receded before the edges, viewed under crossed polarizers. . . . . . . . . . . . .  viii  61  Figure 3.15 A schematic showing the observed evaporation behavior of a printed CNT line with a bulge defect and the expected CNT orientation. . . . . . . . . . . . . . . . . . . . . . . . . . . .  62  Figure 3.16 energy dispersive spectroscopy (EDS) spectrum of contaminated CNTs. . . . . . . . . . . . . . . . . . . . . . . . . . .  63  Figure 3.17 SEM images of shortened CNTs with an average length of 131 nm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  64  Figure 3.18 SEM image of long CNTs with an average length of 787 nm. .  65  Figure 3.19 The morphology of a 5 layer printed line using a dilute, 0.0125 wt%, CNT suspension. . . . . . . . . . . . . . . . . . . . . . . . .  66  Figure 3.20 The morphology of a 5 layer printed CNT line . . . . . . . .  68  Figure 3.21 Profile of a 20 layer printed CNT line taken with an optical profilometer. . . . . . . . . . . . . . . . . . . . . . . . . . .  69  Figure 3.22 Several SEM images of the sub-surface morphology of the aligned CNT films . . . . . . . . . . . . . . . . . . . . . . .  71  Figure 3.23 An SEM image of the underside of a peeled layer of CNTs. The film was peeled off using conductive tape. . . . . . . . .  72  Figure 3.24 A schematic showing the crusting phenomenon. . . . . . . . .  74  Figure 3.25 An SEM image showing an aligned layer of CNTs covering a highly bundled film. . . . . . . . . . . . . . . . . . . . . . .  75  Figure 3.26 An SEM image showing the cross section of an aligned CNT film . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  76  Figure 3.27 The second layer peeled from a printed CNT line under crossed polarizers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . Figure 4.1  Figure B.1  76  A proposed CNT transistor architecture which can be almost entirely inkjet printed. . . . . . . . . . . . . . . . . . . . . .  83  An example of SC agglomerates and adsorbed metal particles.  101  ix  Glossary AFM atomic force microscope CNT carbon nanotube COMOCat cobalt-molybdenum catalyst CVD chemical vapour deposition DAQ data acquisition unit DGU density gradient ultracentrifugation DOD drop on demand EDS energy dispersive spectroscopy FFF field flow fractionation HiPCO high-pressure decomposition of carbon monoxide MWNT multi-walled carbon nanotube PET polyethylene terephthalate PDMS polydimethylsiloxane PVDF polyvinylidene fluoride SC sodium cholate SDBS sodium dodecyl benzene sulfonate x  SDS sodium dodecyl sulfate SEC size-exclusion chromatography SEM scanning electron microscope SWNT single-walled carbon nanotube TFT thin film transistor  xi  Acknowledgments I express my sincere gratitude to my supervisor Dr. Konrad Walus for his constant encouragement and support, and for giving me the opportunity to pursue fascinating scientific research. I could not have asked for a better mentor. I extend my appreciation to Dr. John Madden, Dr. Boris Stoeber, and Dr. Edmund Cretu for their helpful insight during project meetings. I would additionally like to thank my lab mates Robert Busch, Derek Tsan, Lisheng Wang, and Dan Yu for countless stimulating conversations and making life in the laboratory fun. I also thank Chris Sherwood for allowing me to use an ultracentrifuge and always being very accomodating, and Dr. Maya Kopylova for allowing me to use her polarized light microscope. I would not be here if not for the unwavering support of my parents Heike and Kevin Beyer and my wife Ashley Beyer. Thank you all for your constant encouragement and love.  xii  Chapter 1  Introduction Since their discover nearly two decades ago [1], carbon nanotubes (CNTs) have been the focus of a tremendous global research effort. Their excellent intrinsic properties make them desirable for a broad variety of applications including: largearea and high-performance electronics [2], polymer composites [3], and chemical [4] and electro-mechanical sensors [5]. In the electrical domain, pristine singlewalled carbon nanotubes (SWNTs) have current carrying capacities three orders of magnitude higher than copper [6], and have structurally dependent electrical properties. The latter point is particularly what makes CNTs such a promising material, but it also increases the difficulty of using them effectively. A CNT may behave either as a metal, semi-metal, or semiconductor depending on its so called chirality. CNTs are essentially rolled up sheets of graphene; though in actuality they are grown as tubes. As depicted in Figure 1.1, a sheet of graphene may be rolled along many different vectors, producing tubes with slightly different atomic configurations. These configurations are defined by the chirality, described by the integers (m, n), which represents the total number of unit vectors it takes to transverse the circumference of a given CNT. Unfortunately, good control over the chirality during synthesis remains problematic and typically one is left with a mixture of metallic CNTs and semiconducting CNTs of different chiralities. State of the art synthesis techniques are capable of narrowing chirality distribution [7, 8], but absolute control has not been demonstrated to date. Isolating CNTs by electronic type is possible using post-synthesis techniques, but is tedious and suffers 1  Figure 1.1: A graphene lattice showing the different chiral vectors of a CNT. The majority of the CNTs used had a (6, 5) chirality, which is indicated by the red vector. The red arrows are parallel to the length of the CNT. from low yield [9]. Nevertheless, this unique feature of CNTs allows them to be used as both tunable semiconductors and high performance conductors. Lastly, because of their small diameter CNTs are considered one-dimensional materials. That is, electrons are confined to travel only along the length of a CNT and do so ballistically under optimal conditions [10]. In the mechanical domain, SWNTs have a tensile strength on the order of tens of gigapascals [11]. As such, much work is being done in the area of light-weight, high tensile strength CNT polymer composites [12, 13]. There is a fundamental challenge in employing CNTs in devices that has largely prevented their commercial application. The challenge is in controlling the position and orientation of deposited CNTs. Currently the most economical way of producing CNTs is through batch chemical vapour deposition (CVD) synthesis [14]. The result is a dry CNT powder that must be controllably deposited on a substrate, usually via suspension in a liquid. Though solution processing is perhaps an unrealistic means of fabricating single-CNT devices, it is a viable means of creating CNT films for use in micro-electronics and sensors. There are a variety of exist-  2  ing solution deposition technologies that lend themselves well to the deposition of CNT suspensions: spin-coating [15]; drop-casting [16]; dip coating [17]; stamping [18]; spinning [19]; and inkjet printing [20]. While some degree of control over the nanoscale position and orientation of CNTs using these techniques has been demonstrated, the resulting films are typically composed of randomly oriented networks with limited nanoscale control. In this thesis, a process has been developed for achieving a high degree of control over the nanoscale alignment and orientation of CNTs in macroscale films using solely inkjet printing and the natural forces present in solvent evaporation. The alignment mechanism utilized is the intrinsic lyotropic liquid crystalline behaviour of highly concentrated CNT suspensions and no external forces were applied. Different patterns were printed and it was found that the long range orientation of the aligned CNTs corresponded directly to the evaporation behaviour of the pattern, which in turn was determined by the pattern geometry. The morphological properties of these films are investigated and several important observations are made. This thesis is comprised of four chapters: • Chapter 1: An introduction to CNT suspension in solution, solution phase alignment, purification and analysis techniques, and a review of the most relevant literature on these topics. • Chapter 2: A detailed account of the experimental details involved in the suspension and purification of CNTs in solution, the inkjet printing procedure used to create the films, and the processes used to analyze the resulting films. • Chapter 3: An analysis and discussion of the observed experimental results. • Chapter 4: Concluding remarks and a discussion on the future direction of this research.  1.1  Motivation  In microfabrication, photolithography is the standard by which all other technologies are compared. The process is inherently high throughput and its resolution 3  capabilities are evident in examples such as integrated circuits. However, there are several drawbacks. First the photolithographic process involves a conglomeration of technologies, including several different thin film deposition and etching technologies, electron beam lithography, high vacuum and ultra-clean environments, and several others depending on the particular devices being fabricated. This required collection of equipment amounts to a significant initial capital cost. Because silicon fabrication technology is so mature silicon devices are relatively inexpensive, however using lithography in ways that deviate from silicon processing norms, such as depositing pre-synthesized CNTs, can be very expensive. Furthermore, the approach itself is wasteful as material is deposited without spatial consideration and etched away in selected areas to establish a pattern. This in itself provides an opening for the development of new fabrication technologies, but the door is opened even further by an emerging new class of large-area electronics. This class includes displays, sensing arrays, lighting, photovoltaics, and electronic paper; all of which do not require the incredible resolutions offered by photolithography. Photolithography becomes less economical when fabricating large components. Fewer components may be made in parallel and so the unit cost of each is increased. Inkjet printing is garnering much interest as an economical alternative for the fabrication of large-area electronics. It has high throughput capability and potentially requires no accompanying technologies. It is a “direct write” fabrication technology meaning that material is deposited only where desired, resulting in little waste. Furthermore, advances in colloidal chemistry have enabled the printing of materials such as metals, polysilicon, conducting and semiconducting polymers and countless others, making the prospect of all-inkjet printed devices an increasingly realistic one. The equipment and operational costs are thus incomparable to photolithography. Additionally, processes and device prototypes may be easily and inexpensively developed using laboratory scale inkjet printers, then scaled up for mass production without significant scaling induced technological challenges since the technology is fundamentally the same. In industrial technology development scaling from laboratory-scale prototypes to mass production comprises one third of the overall cost and technological challenge [21]. Inkjet printing does have its downsides, namely a large minimum feature size and low mobilities. The feature size is a surmountable challenge for large area electronics, but the low mobilities 4  result in limited operating frequencies. However, advances in colloidal and polymer chemistry may see a solution to the mobility problem. Because of the budding interest in inkjet printing as an economic alternative for the fabrication of electronics, strides in process development utilizing this technology are expected to have a high impact. The deposition of CNTs using inkjet printing is not a novel concept in itself, in fact it is a relatively widely used method of patterning CNT films from solution [22–25]. However, unless careful considerations are made the CNTs in these films will have a random and uncontrolled orientation. Even in a random network CNTs have exciting potential. Inkjet printed CNT devices have been previously manufactured using such random networks [24, 26]. However, both theoretical [27] and experimental [28] evidence has indicated that enhancements in the electrical properties of films is expected when the CNTs have a mutual, parallel alignment. For example, Kang et al. [28] showed that films of well aligned CNTs had an electron mobility of about 1000 cm2 /Vs while similar films of randomly oriented CNTs had mobilities of only 10 cm2 /Vs. The ability to both align CNTs in parallel and control their long-range orientation is thus a highly sought after objective. Being able to achieve such control using inkjet printing could mean great strides in the quality and diversity of large-area electronics, sensors, and other applications suitable to inkjet printing.  1.2  Objectives  The primary objective of this work was to develop a method of aligning CNTs using solely inkjet printing, and if successful, investigate the film properties and determine any enhancement gained by the alignment. Achieving this objective required the identification of a suitable alignment mechanism that was compatible with inkjet printing, and the processing of a CNT suspension that met the physical requirements for both inkjet printing and the chosen alignment mechanism. The morphological properties of the film were investigated to determine the degree of success achieved, and to explain the physical processes involved in the observed results. Simple devices would then be designed and fabricated in order to test the electrical properties of the film and compare those properties to that of a randomly  5  oriented film.  1.3 1.3.1  Prior Art Suspension and Purification  The processing surrounding the purification of CNTs and their suspension in liquid has made great advancements in recent years. Purification and stable suspension are two of the most important considerations when utilizing solution processing technologies to fabricate CNT films. In terms of purity it is now possible to purchase CNT material containing very little metallic catalyst or carbonaceous impurities, however vendor and batch variability remains a concern. It has been shown that rated purity and actual purity often vary dramatically [29]; a fact verified in this work, where a purchased CNT sample rated at greater than 95% purity actually contained significant impurities. Thus post-synthesis purification processes are still an active area of research. In terms of suspension, the challenge lies in breaking apart the CNT bundles and preventing them from agglomerating over time.  1.3.2  Stabilization in Solution  CNTs are superhydrophobic by nature. The carbon-carbon bonds that make up the structure of a CNT are chemically stable and lacking in polarity. Their suspension in aqueous solutions is thus impossible without some form of stabilization. Even in non-polar solvents the strong van der Waals forces between the CNTs cause them to rapidly agglomerate without some form of inter-tube repulsion. This repulsion may come about sterically or electrostatically. Long-chain, uncharged polymer wrapping is the primary form of steric stabilization [30]. Electrostatic repulsion has a number of forms, which are described below. In superacids, protonation of the CNTs creates a net positive charge along their surface, preventing agglomeration [31]. Very high concentrations, up to 8 wt%, have been created in this way [32]. Such strong acids require very careful handling and severely restrict the types of substrate and processing materials that may be used. Another possibility is chemical functionalization of the CNT surface. Using 6  a mix of concentrated sulfuric and nitric acids, in either a reflux [33] or under ultrasonication [34], it is possible to oxidize the surface of the CNTs, creating polar OH- and COOH- groups which provide both the electrostatic repulsion required to prevent agglomeration and the polarity for stabilization in an aqueous solution. The final and most common method of stabilization is through the use of surfactants. These amphiphilic molecules readily adsorb onto the surface of the CNTs. In principle, the hydrophobic (non-polar) end of the surfactant attaches to the CNT surface, allowing the hydrophilic (polar) end to interact with the surrounding aqueous solution. While this principle holds true, preliminary atomistic simulations have shown that the way in which the surfactant binds to the CNT can be quite complex, and is diameter dependent [9]. Ionic surfactants provide not only the polarity for aqueous suspension, but also the electrostatic repulsion. Three surfactants are commonly used for this purpose: sodium dodecyl sulfate (SDS), sodium cholate (SC), and sodium dodecyl benzene sulfonate (SDBS), though more exotic molecules such as gellan gum [35] and DNA [36] have also been used. One potential downside in using surfactants is their impact on the electrical properties of the CNT film. For example, the electrical resistance of a CNT film created using the procedures outlined in Chapter 2 of this thesis was beyond the measurement capability of the Keithley sourcemeter used (Keithley 2410), which is in the gigaohm range. However, after rinsing away the surfactant with ethanol the resistance dropped to approximately 300 ω. The dramatic disparity suggests that device properties will be a strong function of the effectiveness and repeatability of the surfactant removal procedure employed.  1.3.3  Purification  The importance of purification was recognized almost immediately upon the discovery of CNTs. As such, the body of work on this topic is immense, consisting of perhaps hundreds of journal articles. The importance of CNT purity warrants some attention here, but more rigorous reviews can be found elsewhere [37]. Impurities arise during the synthesis process and typically consist of metallic catalyst particles, amorphous carbon, carbon nanoparticles and undesired CNTs, such as multiwalled carbon nanotubes (MWNTs) in a batch of SWNTs. Impurities are typically  7  considered disadvantageous as they add uncertainty to, and degrade material properties. However they are particularly deleterious to self-assembly. Impurities act as binding sites for CNTs, facilitating bundling and random entanglements. Such bundles no longer possess the behavior expected of high aspect ratio particles and act as local sites for disorder [38]. As discussed in Chapter 3, alignment is seemingly impossible to achieve if the impurity content is too high. There are essentially four routes to purification: improved synthesis processes; wet chemical treatments; dry gaseous treatments; and physical treatments [39]. CNT Synthesis Ideally, as-synthesized CNTs would be free of impurities and post-purification treatments would be unnecessary. Currently this is not the case, though synthesis techniques have come a long way. The first methods of synthesizing CNTs were: arc-discharge, which is a method that utilizes the local heat generated by a plasma discharge against a carbon/catalyst target, and laser ablation, a similar method that uses the heat generated by a rapidly pulsed laser. Both of these methods typically result in material containing a high weight percent of carbonaceous and catalytic impurities. More recent approaches have improved the CNT yield greatly. The economical and high-throughput CVD methods [14], which operate broadly on the decomposition of carbon containing gases catalyzed by metallic particles, are now standard. In particular, the high-pressure decomposition of carbon monoxide (HiPCO) method produces high yields of CNTs with little carbonaceous impurities, and is perhaps the most common existing method. Another approach known as the cobalt-molybdenum catalyst (COMOCat) method has been shown to produce CNTs largely with a chirality of (6,5), and is a promising method of obtaining pure semiconducting CNTs. While CVD methods reduce carbonaceous impurities, metallic impurities remain problematic and must be removed using a post-synthesis purification process. Wet Chemical Purification Wet chemical treatments are typically designed to remove metallic catalyst particles through digestion in nitric acid (HNO3 ). A standard procedure is to reflux raw 8  CNTs in 2.6 M HNO3 for up to 45 h [40]. The HNO3 breaks the bonds that entangle the metallic impurities with the CNTs and digests the metal through oxidation [37]. This procedure is followed by vacuum filtration to collect the purified CNTs and rigorous rinsing to remove any dissolved particles. Several variations on this process have been proposed, but fundamentally it is always the same. Dry Chemical Purification Dry chemical purification processes rely on the principle of gaseous oxidation at high temperature. In the most basic procedure the more chemically reactive carbonaceous impurities burn at a lower temperature than the CNTs and may thus be selectively removed. For MWNTs it is possible to perform the procedure in still air, but more care is required for SWNTs, where mixtures of Ar with O2 [41], and Cl2 with H2 O [42] were found to be more effective. However, this procedure does not remove metallic particles. In fact the standard method of quantifying the metal content is through thermal gravimetric analysis, where all carbon material, including the CNTs is burned away at 1000 ◦ C, leaving any metal behind [37]. Furthermore it has been found that the reason air oxidation is ineffective for SWNTs is because the metallic particles catalyze oxidation reactions, causing them to burn at lower-than-expected temperatures. Thus, gaseous oxidation must always be accompanied by a wet chemical treatment for greatest effectiveness. Physical Purification Physical purification can be used both as a coarse, initial purification process to remove metallic and carbonaceous impurities, and as a finishing step to separate CNTs by length [43], and even chirality [44]. Four methods comprise this type of purification: ultracentrifugation, density gradient ultracentrifugation (DGU) [9], size-exclusion chromatography (SEC) [45], and field flow fractionation (FFF) [46]. Ultracentrifugation acts on the mass of suspended particles in a centrifugal field, forcing more massive particles to the base of the centrifuge tube at a greater velocity than less massive ones. Large CNT bundles thus comprise the pellet (bottom portion) of the tube, and singular CNTs the supernatant (top portion). It was found that the pellet contained significantly more impurity content than the supernatant  9  [47]. In this way ultracentrifugation provides a method of removing impurities. DGU utilizes a centrifugal field, but rather than acting on the mass of the particles, acts on minute differences in their densities when suspended in a density gradient medium, typically iodixanol. Material that is less dense than the local solution density will travel against the centrifugal field until it reaches a region of neutral buoyancy, at which point it remains stationary regardless of the strength or duration of the centrifugal field; the opposite is true of particles which are more dense than their surroundings. Centrifugation proceeds for at least 20h, ensuring that each particle is positioned in its region of neutral buoyancy. Density differences arise between CNTs of different chiralities due to slight changes in surfactant adsorption. Nonlinear density gradients may be created in which the gradient is less severe near values pertaining to a certain chirality, allowing highly specific separation of a mixed chiral species [9]. Yet further, by using two different surfactants (SC and SDS) enantiomers of a single chirality may be distinguished [44]. DGU, however, is not apt at sorting CNTs by length. Standard ultracentrifugation may be utilized to this end; longer, more massive CNTs sediment faster than short CNTs enabling some degree of separation. Batchelor et al. showed that the velocity of CNTs in a centrifugal field scales with the natural logarithm of their aspect ratio [48]. More precise methods include SEC and FFF. SEC is a commonly used technique in biology and polymer science, and bears similarity to a filtration system with a gradient in pore size. CNTs suspended in an aqueous solution are hydraulically driven through a gel-containing tube. The gel is typically composed of cross-linked polystyrene that has decreasing pore sizes along the length of the tube. Short CNTs are driven down the tube, unimpeded by the gel, whereas longer CNTs are less able to pass through the diminishing pores, resulting in highly specific length fractionation. FFF produces similar results, but using a different approach. The CNT suspension is driven into a narrow channel and a force field is applied perpendicular to the length of the channel. This force may be an electric field, a fluid flow, a centrifugal field or any other force field that will act upon the suspended particles in the desired way. Particles will physically rearrange themselves within the fluid based on their response to the applied force. Fluid is then driven down the length of the channel and because the velocity profile of the fluid is parabolic, the spatially separated particles will travel different dis10  tances based on where they are within the fluid [49]. In this way the fractions are separated without altering the medium in which they are suspended. Using FFF it is possible to separate CNTs both by electronic type, though not in such a refined way as DGU [39], and by their physical properties such as mass and length.  1.3.4  Alignment Mechanisms  For several applications it has been found that aligning the CNTs in a film such that they are all oriented in parallel to one another is highly beneficial. For example, Kang et al. showed that thin film transistors (TFTs) composed of well aligned CNTs had on-off ratios of 105 and mobilities of 480 cm2 /Vs [28]. These values far exceed those reported for randomly oriented CNT TFTs such as mobilities of 150 cm2 /Vs and on/off ratios of 100 [50]. Furthermore the alignment gives rise to electrical anisotropy along different spatial vectors [27], which may be beneficial for certain applications such as electromechanical sensors. Several methods of achieving alignment in CNT films have been conceived with varying degrees of practicality. The most well documented methods utilize CVD growth. Vertically aligned CNT forests are relatively easy to grow, though for thin film applications horizontally aligned CNTs are more desirable as electron travel is typical in-plane, rather than out of plane. When a single-crystal quartz substrate is used, horizontal CNT alignment during the CVD growth arises and is determined by the crystal lattice of the substrate, resulting in highly oriented CNTs [51]. Another recent approach is to grow a very tall (1 mm) vertically aligned forest and push it over mechanically with a roller, forming a horizontal film with a high degree of alignment [52]. Though undeniably powerful, the necessity of having such incredibly long CNTs makes this approach somewhat limited. Several other CVD alignment methods exist [53] but are somewhat limited by the high temperatures involved (up to 900◦ C) [28], which restricts the choice of substrate. Deposition of CNTs from the solution phase has many benefits in terms of application and process flexibility. Alignment from solution is highly desired for this reason. Out of the techniques developed to this end, dielectrophoresis has been given the most attention [54, 55]. One noteworthy drawback of this method is the difficulty in aligning CNTs over long ranges. Typically the spacing between the  11  two electrodes is on the order of the length of the CNTs [56]. The second drawback is the need for specifically shaped electrodes, which may limit or impede the geometry of the end device or circuit. Apart from magnetic field alignment, which has limited practicality forCNTs due to either the high fields required [57] or to the requirement of attached magnetic nanoparticles [58], most remaining methods rely on naturally occuring self-assembly mechanisms. These may include alignment induced by fluid flow in evaporating droplets [59, 60]; alignment due to the orientation of a liquid crystal medium in which the CNTs are suspended [61]; or alignment due to the innate lyotropic liquid crystallinity of CNTs in aqueous suspension [62]. CNTs suspended in a liquid crystal medium poses a significant challenge with respect to inkjet printing due to the physical restrictions imposed on the solution, such as viscosity limits, thus the most realistic method of aligning CNTs using inkjet printing is either through their innate liquid crystallinity, or by fluid flow. In fact, both mechanisms contribute to the method developed as part of this research and will be described fully in Chapters 2 and 3. Liquid Crystalline Alignment The liquid crystalline properties of CNTs have previously been used to facilitate alignment in several ways. As explained in Section 2.1, obtaining a high concentration is a necessary precursor to the liquid crystalline phase transition. Sufficiently high concentrations have been achieved when dispersing CNTs in superacids, but not in surfactant stabilized aqueous solutions. When using surfactants, others have achieved high concentrations in three ways: by allowing the liquid to evaporate to 1/20th of its initial volume [35]; by using ultracentrifugation, which will concentrate CNTs at the bottom of the centrifuge tube [63]; or by evaporation of individually deposited droplets [34]. In the former two methods, the highly concentrated (5 wt%) suspensions exhibited short ranged alignment with rapidly changing orientation over long ranges. It was found that expression of the liquid crystalline phase took several weeks to fully develop [64, 65], presumably due to the low CNT mobilities in the highly concentrated suspension. Long range alignment was then achieved by mechanically shearing the suspension. Unfortunately these two methods are incompatible with inkjet printing as a CNT suspension concentrated  12  by a factor of 20 is too thick to be printed; it is more gel-like than liquid. The droplet evaporation method takes advantage of a phenomenon that occurs during the evaporation of most liquids known commonly as the “coffee ring effect”, so named after the ring of solute commonly observed in dried coffee stains. When a droplet of liquid evaporates, it does so in one of two basic modes: (1) with a pinned perimeter and decreasing contact angle, or (2) with a receding perimeter and constant contact angle [66]. These will be referred to as pinned mode and receding mode respectively. In colloidal systems evaporation shows hysteresis between advancing and receding contact angles. Evaporation initially proceeds in the pinned mode, and is followed by a transition to the receding mode once the critical receding contact angle is reached. The pinned mode is most notably characterized by the existence of an internal hydrodynamic flow directed outwards from the center of the droplet. This flow acts to replenish liquid lost at the outermost edges due to evaporation and carries any suspended particles with it [67]. As such, suspended CNTs will be shuttled toward the edge of the droplet, accumulating there. If the concentration is sufficiently high, the “coffee ring” will exhibit liquid crystalline behaviour and the CNTs will self assemble into an aligned film. Their orientation will be determined by the geometric boundary condition imposed by the edge of the droplet [65]. Not all of the CNTs are deposited in the coffee ring and those remaining in suspension at the onset of the receding evaporation mode are deposited randomly across the middle region of the film [31, 65].  13  Chapter 2  Carbon Nanotube Film Fabrication The process for achieving alignment in inkjet printed CNT films involves two fundamental steps: preparation of the CNT suspension, and inkjet printing of that suspension. This chapter will explain, in detail, these two steps and all of the considerations involved. First, the basics of liquid crystals and their governing theory as it pertains to the alignment of CNTs will be discussed. Second, the process for creating a CNT suspension which meets the criteria for both inkjet printing and liquid crystalline behaviour will be detailed. This process consists of surfactant stabilization in water, ultrasonication, and ultracentrifugation. Third, the inkjet printing itself will be detailed. This process involves not only careful selection of inkjet printer parameters, but a specific printing technique also. The CNT concentration must be enhanced concurrently with the inkjet printing; a process that involves modeling the evaporation rate of the printed pattern, and controlling and monitoring the environmental conditions near the pattern. The considerations involved in substrate selection and preparation will also be discussed.  14  Figure 2.1: The typical molecular arrangement in: (a) a liquid, (b) a nematic liquid crystal, and (c) a smectic A liquid crystal.  2.1  Liquid Crystal Fundamentals  Liquid crystals are a form of matter that exhibit characteristic of both a liquid and a crystal. In particular, the molecules of which the material is composed have mobility similar to that of a liquid, allowing the material to flow, but also possess some degree of spatial order, as in a crystal. The number of degrees of order and how that order is expressed define the type of liquid crystal. The most basic liquid crystals, called nematic liquid crystals, possess orientational ordering but lack translational order, as seen in Figure 2.1. Smectic liquid crystals have both orientational order and some translational order. For example, the material may form planar layers. There may not be translational order within each layer, but the layers themselves constitute an additional degree of order. As will be shown in Chapter 3, this appears to be the liquid crystal phase expressed in the CNT films developed in this work. Several types of smectic liquid crystals exist and are differentiated by the exact orientation of the molecules. Cholesteric liquid crystals, which are typically composed of chiral molecules, possess a helical orientation. The types of liquid crystals mentioned above belong broadly to two classes: thermotropic and lyotropic liquid crystals. Thermotropic liquid crystals are those  15  materials which exhibit a temperature dependent liquid crystalline phase transition. Some polymer melts are a good example of a thermotropic liquid crystal. Lyotropic liquid crystals exhibit a concentration dependent liquid crystalline phase transition and are composed of a suspension of molecules or particles. Surfactants dispersed in water are one possible example of a lyotropic liquid crystal. Surfactant stabilized CNTs in solution are another. If the most basic pre-condition for liquid crystallinity is either a very high aspect ratio, for rod-like liquid crystals, or a very low aspect ratio, for disc-like liquid crystals, then with aspect ratios exceeding 1000, CNTs appear to be an ideal candidate for lyotropic liquid crystallinity. Indeed it has been shown in several publications that both MWNTs and SWNTs are lyotropic nematic liquid crystals under the correct conditions [31, 62, 68, 69]. A high level description of basic liquid crystal theory is presented here to give an understanding of how ordered liquid crystal phases arise. A more in-depth explanation may be found elsewhere [70]. The original theory of liquid crystals was described by Onsager [71] in an effort to explain anisotropic phases in certain colloidal solutions. The model he developed describes the phase transition from isotropic to nematic in terms of translational and orientational entropy. In a dilute suspension of slightly repulsive rods, translational entropy dominates. That is, the lowest energy configuration is one in which the particles are evenly spread out. Orientational entropy has a very minimal role in this situation and no particular orientation is preferred. When particle concentration is greatly increased energy can only be minimized by achieving a higher packing density. This is known as volume exclusion. A better packing density is achieved through changes in orientation, so at high concentrations orientational entropy dominates and ordered phases are possible. This model works under the assumption of repulsive rods, which is consistent with ionic surfactant wrapped CNTs. Mathematically this process is described using density functional theory, where a functional is a function of functions. The possible orientations a rod can take are described by a function, and a functional describes the free energy of the whole system based on the spatial distribution of these functions. Minima in the free energy corresponding to some distribution of orientations can be determined, though it is not a trivial mathematical problem and does not have an analytical solution [72]. The theory has since been refined in various ways, for example to include pa16  Figure 2.2: Disclinations with different s values. rameters describing disclinations (the so called Frank constants) [70]. Disclinations are topographical defects in the structure of the liquid crystal that appear spontaneously with the phase transition and diminish over time [73]. They consistently exhibit several configurations which are described by the orientation parameter s. If a vector is drawn through the center of a disclination, s is the change in orientation of the director in multiples of 2π along that vector, where the director is simply a vector describing the local orientation of the particles. Figure 2.2 shows several examples. Disclinations arise naturally due to temperature, applied forces and impurities [64] and are a commonly identified feature in liquid crystals. Using liquid crystal theory it is possible to create a phase diagram describing the material phase against certain important parameters, such as temperature and particle concentration. Somoza et al. [74] created such a phase diagram for CNTs and it is shown in Figure 2.3, where η is packing fraction, which is a function of concentration. This phase diagram was generated under the assumption of Van der Waals attraction between the CNTs, rather than ionic repulsion, which is why the required temperatures for a phase transition are so extreme. The shaded regions of the diagram are where the isotropic and nematic phases coexist. The sharp rise near the middle of the diagram is known as the Flory chimney and the area above the shaded region and to the left of the chimney represents the isotropic phase; the area above and to the right represents the pure nematic phase and presumably higher ordered phases such as smectic. The black vertical bar marks the isotropic-nematic 17  Figure 2.3: The theoretically calculated phase diagram for CNTs of different aspect ratios. L and D are the CNT length and diameter respectively. Reprinted with permission from [74]. Copyright 2001 The American Physical Society. phase coexistence region in the absence of Van der Waals forces, where the phase is isotropic to the left and nematic to the right. This is a better approximation for the ionically charged CNTs used here. What is interesting is that the packing fraction required to achieve the nematic phase decreases with increasing particle aspect ratio. Being lyotropic in nature, the concentration of CNT in suspension is of primary importance. The particular concentration at which a CNT suspension experiences a liquid crystalline phase transition is dependent on several factors, described collectively and qualitatively by the parameter known as mesogenicity. The higher the mesogenicity of the lyotropic liquid crystal, the lower the required concentration for a liquid crystal phase transition. In CNT suspensions, mesogenicity is determined by: the aspect ratio of the CNTs; their degree of bundling; their tortuosity; the CNT-CNT interaction, which may depend on chemical functionalization of the CNTs, on surfactant type and quantity, and on electronic type of the CNTs; and 18  the CNT-solvent interactions. In this thesis, the CNT length, which is related to tortuosity; the bundling; and the concentration have been explicitly studied. The impact of changing these variables is discussed in Chapter 3, and the experimental way in which they were controlled is detailed in Section 2.2 of this chapter.  2.2  Solution Processing of CNT Suspensions  The parameters involved in suspending the CNTs in solution have the greatest impact on the outcome of the CNT alignment. It is in these processing steps where all of the previously mentioned parameters affecting mesogenicity are controlled. The process used in this research consisted of three steps: stabilization in water, ultrasonication, and ultracentrifugation.  2.2.1  Aqueous Stabilization  Several suspension techniques are described in Section 1.3.2. Aqueous solutions were used here due to their compatibility with inkjet printing, and their ease of handling and processing. Polymer substrates were used which excludes the use of organic solvents and very strong acids, and surfactant stabilization is chosen over chemical functionalization as it allowed for more control over the physical properties of the CNTs. It is a known fact that the chemical functionalization of CNTs in the presence of strong acid causes a reduction in CNT length and an increase in surface defects [33]. In fact, the presence of functional groups is itself a defect. These defect sites weaken the pristine CNT structure and it is expected that the scission caused by ultrasonication would be more severe as a result. Experimental Details Both SDS and SC were used in this research, though SC was seen as the better option for two reasons. First, the ratio of surfactant required to suspend the CNTs was less for SC. At 0.5 wt%, SC was able to suspend 0.5 wt% CNT without significant precipitate, whereas considerable precipitate formed under the same conditions using SDS. The second reason is that SDS has a greater impact on the surface tension of the water, reducing it to a greater extent than SC. This has implications on inkjet printing which are discussed in Section 2.3. Thus throughout all alignment ex19  Figure 2.4: An SEM image showing the highly bundled nature of the CNT soot. periments the initial CNT suspension, before any additional processing, consisted of 0.5 wt% SC and 0.5 wt% CNTs in water. Here wt% refers to the weight of the given component divided by the total weight of the solution. A concentration of 1 wt% CNT with 1 wt% SC was attempted, but resulted in considerable precipitate.  2.2.2  Ultrasonication  Because the initial condition of the CNTs is one of extreme bundling as seen in Figure 2.4, some force must be applied to debundle and allow the surfactant to adsorb to the individual CNTs. Currently the most effective way of doing this is through ultrasonication [75]. Physically, the acoustic waves produced by the piezoelectrically driven sonication mechanism, typically a probe or a cup, cause cavitation in the liquid medium. Cavitation is the spontaneous formation and immediate collapse of bubbles within the liquid caused by the rapid change in acoustic pressure. The implosion of these bubbles imparts tremendous force on any particles within the liquid and causes significant friction induced heat, which can be used to drive chemical reactions. CNTs will be forced apart locally, and if cavitation continues for an extended period of time, the majority of the CNT bundles will come apart. An important effect under strong ultrasonication is cavitation induced scission.  20  Figure 2.5: The radial fluid velocity profile around an imploding cavitation center and the net force induced on a nearby CNT. Reprinted with permission from [76]. Copyright 2007 The American Chemical Society. Implosion of cavitation bubbles can cause reported strain rates as high as 109 /s in the surrounding fluid, which is transfered to any nearby CNTs [76]. It has been shown for both CNTs [76] and polymers [77] that the radial flow velocity around an imploding bubble increases exponentially towards the cavitation center as shown in Figure 2.5. The end of a CNT nearest to the cavitation center will move rapidly with the fluid around it, however the fluid surrounding the end of the CNT furthest away from the cavitation center will be moving slower, giving rise to friction induced drag force as shown in Figure 2.5 inset. These forces cause tensile stress along the CNT with a maximum net force at the middle of the tube. It was shown that the net force is proportional to the square of the CNT length, thus long CNTs readily break when subjected to cavitation, and there is a minimum length at which scission can no longer occur. It is possible that in surfactant wrapped CNTs the strain is more severe as the surfactant molecules mediate better physical coupling between the CNT and fluid, though this implication appears not to have been studied.  21  Table 2.1: The sonication parameters used for each CNT length regime. Length regime  Sonication duration (hours)  Sonication power (W)  Sonication type  Short Medium Long  20 20 20  120 20 130  Probe ultrasonication Probe ultrasonication Bath sonication  Experimental Details Three length regimes were studied; what will be known from here on as the short, medium, and long length regimes. Two different ultrasonication instruments were used to obtain the different lengths. For both the short and medium regimes a high power probe ultrasonicator was used (Heat Systems, Ultrasonicator), and for the long regime a low power bath sonicator (Branson, Bransonics 2510) was used. The sonication parameters associated with each regime are shown in Table 2.1. Note that the power of the bath sonicator appears to be quite high, however that power is distributed throughout 2.8 L of water, whereas in the probe ultrasonicator all of the power is delievered directly to the sample. The procedure used for each sonicator was slightly different. For the probe ultrasonicator, the probe was inserted to a depth of two thirds of the 10 mL liquid volume. Sonication was applied in 0.5 s long pulses, at a 50% duty cycle for the duration of the procedure. The vial containing the CNT suspension was partially submerged in a cold water bath to prevent overheating. The sonication frequency used was 20 kHz. It is important to note that the power ratings listed in Table 2.1 are the electrical power delivered to the piezo-driver, not the acoustic power imparted on the solution. The exact acoustic pressure imparted on the CNT suspension was not measured, but could be with an appropriate hydrophone. The bath sonicator operated at 40 kHz and was run continuously for the given duration. The sonication duration was notably longer than that observed in most publications, where 10 minutes to 1 hour is typical. It was found that such short durations did not effectively debundle the CNTs. This was tested very simply by filtering the suspension through a 5 µm pore size nylon syringe filter. 5 µm is well beyond the specified length of the CNTs, so 22  Figure 2.6: CNTs before (a) and after (b) ultracentrifugation. any retinate would have to be large bundles. After 1 hour of sonication at 120 W the suspension was not able to pass through the filter, clogging it almost immediately. After 3 hours the filter still readily clogged. After 9 hours of sonication all of the suspension easily passed through the filter. The 20 hour duration used here is likely significantly more than the minimum requirement, however Hennrich et al. [76] showed that the CNT length distribution plateaued after approximately 120 minutes of sonication, meaning that any additional sonication will assist in further debundling the CNTs without further contributing to scission.  2.2.3  Ultracentrifugation  Ultrasonication is never able to fully remove the CNT bundles in a given suspension. Any remaining bundles must be removed using ultracentrifugation. The concept here is simple but effective; higher mass particles sediment more quickly in a centrifugal field and lower mass particles more slowly. Thus large agglomerates populate the centrifugate and individually suspended CNTs the supernatant. Figure 2.6 shows the difference in film morphology before and after ultracentrifugation. Using this same principle it is also possible to separate short, individually suspended CNTs from long ones. Both sediment slowly compared to the agglomerates, so higher rotational speeds or longer durations are required.  23  Table 2.2: The centrifugation parameters used for each CNT length regime. Length regime  Centrifuge duration (hours)  Centrifugal force (x gravity)  Fraction removed  Short Medium Long  3 2 2  170,000 20,000 20,000  Top 50% of supernatant Top 80% of supernatant Top 80% of supernatant  Experimental Details The ultracentrifugation for this research was performed with a Beckman TL-100 ultracentrifuge using a 45◦ angle swinging bucket rotor with 1.2 mL centrifuge tubes. Two different sets of parameters were used for the three length regimes studied and are listed in Table 2.2. In the case of the medium and long CNTs, the top 80% of the supernatant was carefully extracted using a pipette and the pellet was discarded. In the case of the short CNTs only the top 50% of the supernatant was extracted. Ultracentrifugation represented the final step in the solution processing, after which the CNT lengths in each regime were investigated. A sample of each regime was diluted to 0.005 wt% using the 0.5 wt% SC solution and deposited on a silicon wafer. SEM images were taken of each sample and the lengths of many individual CNTs were recorded. Figure 2.7 shows an example of the SEM images used. The average CNT length in the short, medium and long length regimes were 131 nm, 337 nm, and 787 nm with standard deviations of 60 nm, 190 nm, and 360 nm respectively and are shown in Figure 2.8. This result supports the earlier statement that sonication and ultracentrifugation are more suited to isolating short length regimes, not long ones. Furthermore, after several days, precipitation was detected in suspensions which had not undergone an ultracentrifugation step, but suspensions subjected to ultracentrifugation were stable even after several months, with no noticeable precipitate or agglomeration.  2.3  Film Fabrication  The term inkjet printing actually describes several related technologies. There is continuous jet printing, thermal drop on demand (DOD) printing and piezoelectric 24  Figure 2.7: A typical example of the images used to calculate the CNT length distributions. DOD printing. In continuous inkjet printing, a pressure driven stream of electrostatically charged liquid is dispensed from a nozzle and breaks into individual droplets due to Rayleigh instability [78]. An electric field is used to deflect the droplets either into a gutter for recycling, or onto the substrate. This method has the advantage of speed, reaching deposition frequencies of up to 100 kHz [79], but requires more complicated hardware. Such high deposition rates were not a requirement of this research. DOD printing is different in that a droplet is dispensed only when it is required. In thermal DOD printing, a heating element within the nozzle vaporizes a small volume of ink. The creation and subsequent collapse of the vapor bubble causes a pressure wave within the fluid which propagates toward the nozzle orifice, ejecting a droplet upon arrival. Thermal printing places an additional restriction on the fluid’s thermal properties, namely the requirement of a high vapor pressure. In piezoelectric DOD printing a piezoelectric element is constricted, creating a pressure wave in the fluid. Though different architectures exist, the piezoelectric DOD system used in this research consisted of a glass capillary tube surrounded by a piezoelectric disc which squeezed the tube when actuated. The advantages of this method are that the pressure pulse is well controlled using the waveform sent to the piezo element, and the actuation process does not impose 25  Figure 2.8: The distributions CNT lengths from the (a) long, (b) medium, and (c) short length regimes. any additional restrictions on the fluid. It is thus the most versatile of the different inkjet technologies. As discussed in Section 1.3.4 CNT suspensions that are sufficiently concentrated to exhibit liquid crystal properties cannot be inkjet printed. Instead, a dilute suspension of CNTs is inkjet printed and the concentration is increased during the printing process. First a pattern is printed such that it has a continuous liquid form. Before it enters the receding evaporation mode and changes geometry, more CNT suspension is deposited at a rate matching the evaporation rate. In this way there 26  Figure 2.9: A printed line with a bulge deformation (a), and without (b). is a net increase in the CNT concentration within the printed pattern. Two patterns were explored in this research, a circle and a 0.75 mm by 7.5 mm line. Several different experiments were performed and will be described in Chapter 3, but the two patterns remained consistent. For the circular pattern, a 5µL droplet was deposited on a copper substrate and allowed to evaporate naturally on a level surface in a standard room environment. The dried CNT films on copper were continuous and did not possess a significant coffee ring region. For the line, deposition had to be distributed across the entire pattern. If fluid was delivered to only one location, bulging would occur at that point while areas further away would begin to recede. Figure 2.9 shows a comparison between a line fabricated by depositing only at one location and a line fabricated by distributing liquid across the entire surface. To achieve this, the evaporation rate of the line, or whatever arbitrary pattern is being printed, must be known prior to printing. This was accomplished using a mathematical model, the details of which are described in Section 2.3.1. In Section 2.3.2 the most important parameters involved in inkjet printing, their effect on the printing process, and how they pertain to CNT alignment are discussed.  2.3.1  Evaporation Model  Increasing the CNT concentration during printing requires the inkjet deposition rate and the evaporation rate of the liquid pattern to be matched. However, obtaining the evaporation rate of the liquid pattern is not trivial. The total evaporation rate could be measured gravimetrically, however this does not provide sufficient  27  information as the evaporation flux varies spatially across the film surface. As previously mentioned, bulging and receding are reduced by distributing the printed liquid across the surface of the pattern, however if the deposition rate is matched to the total average evaporation rate, receding at the ends will undoubtedly occur as they are not fully compensated, and bulging of the mid section will occur as it is receiving too much. Thus, the spatial evaporation profile must be known and the deposition rate changed to match it. The following is a brief explanation for the origin of the spatially varying evaporation flux. It has been shown that in a semi-spherical droplet of liquid, the evaporation flux is enhanced toward the droplet perimeter [66]. When a molecule evaporates from a surface under normal atmospheric conditions it performs a random walk as it diffuses into the surrounding atmosphere. There is a non-zero probability that its random walk will lead it back into the liquid. A molecule evaporating from the edge of the droplet has a larger available space to diffuse into, as quantified by the solid angle at that location, and thus a larger number of possible random walk paths that lead away from the droplet surface, compared to molecules evaporating from the center of the droplet. Thus, molecules leaving from the edge of a droplet are more likely to contribute to evaporation than molecules evaporating from the center, leading to a net increase in the evaporation flux near the edges. The effect also arises in semi-cylindrical droplets, where the evaporation flux is more pronounced toward the ends. The evaporation model is based on two key assumptions, following the work of others [80, 81]. The first assumption is that the evaporation rate is diffusion limited. This means that the ejection rate of the water molecules from the surface is greater than the diffusion rate of those molecules away from the surface. The consequence of this is that the air in immediate contact with the surface is at 100% relative humidity and the diffusion equation  δC(r,t) δt  =D  2 C(r,t),  may be used to  find the spatially varying vapor concentration profile in the area around the liquid pattern. D is the diffusion coefficient of water vapor and r is the spatial coordinate. The second assumption is that C(r,t) is stationary, meaning that the time dependence in the diffusion equation may be set to 0. This assumption is not necessarily valid for evaporating droplets because their shape is constantly changing, however, because the intension here is to prevent the liquid feature from changing shape, it 28  Figure 2.10: A schematic showing the boundary conditions of the model. Cs indicates the saturated vapour concentration on the surface of the liquid, CA is the ambient vapour concentration, and L is the length of the liquid line. is a reasonable assumption. The diffusion equation thus simplifies to the Laplace equation, given as: 2  C(r) = 0.  The Laplace equation may be solved numerically for arbitrary geometries using finite element analysis. The boundary conditions for the problem are depicted in Figure 2.10 and are described as follows: • The vapor concentration on the surface of the liquid feature is at 100% relative humidity. The value is temperature dependent and was extrapolated from a table [82]. Figure 2.11 shows the relationship between temperature and vapor concentration plotted with a best fit curve. • The vapor concentration far away from the surface of the liquid is taken from a physical measurement of the relative humidity given the atmospheric temperature. • The evaporation flux into the surface of the substrate,  C, is zero.  Lastly the shape of the liquid pattern was derived. The footprint, its length and width, was explicitly defined by the inkjet printer to be 0.75 mm by 7.5 mm. The 29  Figure 2.11: The saturated water vapor concentration in air at a given temperature. height of the feature was measured using an in-plane camera and was found to be 140 µm. The profile was assumed to be one of a cylindrical cap and so the exact profile was found using a simple geometric relation. The model was solved in COMSOL Multiphysics using a cubic environment that was 20 times larger than the length of the line on all sides. The size of the modeled environment is critical. If it is too small the evaporation rate will be significantly overestimated. The larger the environment the more accurate the result will be, assuming a proportional mesh size. Figure 2.12 shows the total evaporation flux across the surface of the line for increasing mesh size at different environment sizes. Environments 12,16 and 20 times larger than the length of the line appear to converge, while smaller environments give significantly different results. Ultimately a mesh size of 682 000 elements was used and took approximately 15 minutes to simulate on a standard desktop computer. It was not possible to use a larger mesh due to memory limits on the computer used. The modeled line was split into 10 segments and the total evaporation flux across the surface of each segment was calculated. In order to determine the deposition rate, the size of each ejected droplet, and the translational speed of the printer stage had to be measured. Droplets were measured prior to inkjet printing using a stroboscopic camera sys30  Figure 2.12: The numerical solution for the total evaporation flux across the surface of a printed line at different mesh sizes. Each line represents different cubic environment sizes of 2, 4, 8, 12, 16, and 20 times the length of the printed line. tem, which captured images of the droplets at different stages of their ejection. The details associated with droplet size and ejection are described in Section 2.3.2. The stage speed is not controllable in the custom printer used for this research, but the droplet ejection frequency is. A shortcoming of the software used to operate the inkjet printing system is that a small, approximately 0.5 s delay in printing occurs whenever the frequency of droplet ejection is changed. Thus the average speed was measured by taking the total time to print one layer of the liquid pattern and dividing it by the total distance traveled. The average speed, given the delays, was 1.5 mm/s. Knowing the required volume and the average speed, an appropriate droplet frequency was determined and is shown in Figure 2.13 under the typical environmental conditions of 26.5◦ C and 45.6% relative humidity. Because environmental conditions were not rigorously controlled, the evaporation and deposition rates were re-evaluated at the beginning of each experiment based on the environmental conditions at that time.  Figure 2.14 shows schematically how a  line was printed. Each line pattern consisted of 10 smaller lines printed in a rastorized fashion. These 10 smaller lines formed a single layer of the larger, rectangular line pattern. In the concentration enhancement scheme up to 65 consecutive overlapping layers were printed in the formation of a CNT film. The first 5 layers were deposited at a higher rate than the evaporation rate so that a continuous liquid 31  Figure 2.13: The deposition rate required to match the spatially varying evaporation rate of a printed line. The inset shows the evaporation flux (mol/s) of the COMSOL model.  Figure 2.14: The deposition rate required to match the spatially varying evaporation rate of a printed line. The inset shows the evaporation flux (mol/s) of the COMSOL model.  32  pattern was established. This 5 layer pattern acted as an initial condition for the evaporation model geometry, after which further deposition was performed at rates matching the spatially varying evaporation rate. Since the translational speed of the stage was constant, droplets were spaced closer together in regions with a high evaporation rate and further apart in regions with a low evaporation rate. Furthermore, because the modelled evaporation rate along the four middle segments of the line pattern was relatively constant, as shown in Figure 2.13, a single deposition rate was used there that was equal to the average over that region. Lastly, because the entire line pattern is evaporating, but the nozzle can only be at one location at a time, the actual deposition rate was multiplied by 10 for each region. This factor is included in Figure 2.13. In this way, the total amount of liquid deposited after 10 passes of the nozzle, which is one full layer, was equal to the total amount of liquid evaporated in each segment over that period of time. The fixed speed of the inkjet printer produced an additional problem. Depending on the particular environmental conditions, when the inkjet nozzle was positioned at one end of the pattern, the time it took to travel back to the other end was longer than the depinning time of the liquid contact line, resulting in the end furthest from the nozzle receding prematurely. To prevent this a Peltier element was used to cool the substrate to 20◦ C, which was sufficiently low to prevent the premature receding. The cooling module was self-regulating, using a temperature sensor mounted on the Peltier element, a power supply, and a data acquisition unit (DAQ) and CPU to communicate between each. Figure 2.15 shows a schematic of the module. The cooling was accounted for in the evaporation model by changing the vapor concentration at the liquid surface to match the new, lower temperature. The degree of cooling applied depends on the dimensions of the liquid pattern; the higher the aspect ratio, the cooler the substrate will need to be to prevent premature receding of the contact line. Experimental Error There are several sources of error in the experimental setup which produce a potential mismatch between the evaporation and deposition rates. These include: thermal losses between the Peltier element and the mounted temperature sensor,  33  Figure 2.15: A schematic of the self regulated cooling module. thermal losses between the Peltier element and the printed sample, temperature and humidity fluctuations in the surrounding atmosphere, and changes in evaporation rate with increasing CNT and SC concentration. Despite these errors the modeled evaporation rate and corresponding deposition rate were sufficiently matched to prevent bulging and receding over periods of up to one hour of continuous deposition. For higher resolution patterns a lower error tolerance would be required. Concentration Enhancement The concentration enhancement scheme described above allowed for precise control over the suspension concentration. Four concentration regimes were investigated, which will be referred to as very low, low, medium and high. For the very low regime, the initial CNT and concentration was 0.0125 wt% and for all other cases the initial concentration was 0.5 wt%, but the number of printed layers (ie. the concentration enhancement) was changed. The SC concentration for all cases was 0.5 wt%. The details pertaining to each concentration regime are listed in Table 2.3. The results described here are for medium length CNTs only. Alignment was not observed in the short and long length regimes regardless of the concentrations investigated.  34  Table 2.3: Experimental details of each concentration regime.  2.3.2  Concentration regime  Initial CNT concentration (wt%)  Concentration enhancement (printed layers)  Very Low Low Medium High  0.0125 0.5 0.5 0.5  5 5 20 and 40 65  Inkjet Printing Parameters  In DOD inkjet printing, the goal is typically to eject a single droplet with a stable trajectory upon each excitation of the nozzle. Several parameters contribute to the formation of a droplet, which may be split into two categories: those belonging to the ink, and those involved in actuation of the nozzle. Ink Parameters Only liquids within a narrow range of physical properties are suitable for inkjet printing. It is an additional reason for choosing an aqueous CNT suspension as water-based liquids typically fall within the appropriate range. The most notable characteristic is dynamic viscosity, but liquid-air surface tension must also be considered. Most of these parameters can be grouped into three dimensionless numbers known as the Reynolds, Weber, and Ohnesbourge numbers: Re =  vρα , η  We =  v2 ρα , γ  √ We η Oh = =√ , Re γρα where ρ is the density, v is the fluid velocity, η is dynamic viscosity, α is a characteristic length, and γ is the liquid-air surface tension. α is fixed for a given nozzle architecture, v is related to the applied voltage and ρ, and η and γ are solution 35  properties and have typical ranges of 0.5 cP - 40 cP, and 20 dy/cm - 70 dy/cm respectively [83]. Derby [78] outlined a general rule of thumb for determining whether a liquid is printable using inkjet: 10 > Z > 1 where Z = 1/Oh . The pressure wave in the nozzle must overcome the viscous damping of the fluid, related to its dynamic viscosity η, and the liquid-air surface tension, γ. Inks with a very low surface tension produced an additional problem. At some point during drop formation, a bubble is drawn up into the nozzle. These bubbles accumulate in the nozzle and eventually cause deposition to cease due to dampening of the pressure wave. This effect was studied rigorously by Arjan van der Bos [79] and others [84] and it was discovered that bubble formation occurred during the initial stages of droplet ejection, during formation of the tail end of the droplet [79]. The technical documentation from the nozzle manufacturer additionally warns of such an effect for liquids with a low surface tension [83]. The bubble formation was found to be a strong function of velocity, with a lower droplet velocity reducing bubble formation. This effect appears more prominent with SDS suspensions than with SC suspensions, even at low droplet velocities. Using a goniometer it was possible to determined that SDS decreases surface tension to a greater degree than SC. The following equation [85] was used to compare the liquid-air surface tension of a 0.5 wt% solution of each surfactant: γla =  h2 gρ , 2(1 − cosθ )  where h is the droplet height, γla is the liquid-air surface tension, θ is the contact angle, g is the acceleration due to gravity, and ρ is the solution density, which was assumed to be the same for both solutions. h and θ were 0.108 mm and 6.5◦ respectively for SDS, and 0.611 mm and 31◦ respectively for SC. The result is that the surface tension of the SC solution was 35% higher than the SDS solution. Interestingly the bubble formation was only evident in nozzles that had already been rigorously used, and was not evident in new nozzles. This may be due to wear at the nozzle orifice, or adsorption of CNTs at the orifice that disturbs the meniscus. Because adsorption of CNTs at the nozzle orifice is seemingly unavoidable, this was one of the primary reasons that SC was used over SDS. Sonic cleaning of the nozzles, as prescribed by the manufacturer, removes some, but not all of the 36  adsorbed CNTs. Nozzle Actuation Parameters The fluid physics involved in droplet generation in inkjet systems is complicated and continues to be an active area of research [78]. Comprehensive reviews on both experimental [86] and theoretical [87] results on the physical details of droplet formation can be found elsewhere; what follows is a brief account of the most notable parameters pertaining to this research. In any DOD system there are four controllable parameters that have an impact on droplet creation: the piezoelectric voltage, the shape of the actuation waveform, the pulse width of the waveform, and the fluid pressure at the nozzle orifice. The shape of the waveform sent to the piezoelectric element can have a significant impact on droplet formation. The waveform used can be seen in Figure 2.16. The initial rise of the waveform (tr ) causes an expansion of the glass walls of the nozzle resulting in a negative pressure wave that propagates both toward the nozzle orifice and the ink supply. The wave is reflected at the ink supply and returns as a positive pressure wave. The dwell time (tdwell ) is designed such that the reflected wave returns precisely when the nozzle is compressed (tc ), amplifying the pressure wave generated. The dwell time follows the relation: tdwell =  2L c  where L is the  length of the glass nozzle (11 mm) and c is the speed of sound in the fluid [88]. The reinforced pressure wave causes a droplet to be ejected from the nozzle orifice. The second dwell time (2tdwell ) and the following rise (tr2 ) is called the “echo” and helps to dampen out any remaining pressure oscillations in the nozzle. When the voltage amplitude is too high, secondary pressure waves can lead to the ejection of lower velocity droplets, called satellites; the echo phase helps to prevent this. Pressure in the nozzle must also be carefully controlled and maintained at a value slightly below atmospheric pressure. If the static pressure is equal to or greater than atmospheric pressure then ink will leak out of the nozzle uncontrollably due to gravity, since the nozzle orifice is situated below the ink resevoir. This is assuming that the gravitational force is sufficient to overcome the surface tension at the nozzle orifice, which is the case for aqeous CNT suspensions. A self-regulating pneumatic pressure setup was designed to maintain a constant pressure of 0.52 kPa  37  Figure 2.16: The waveform used to actuate the piezoelectric printer nozzle. The percentages indicate the relative contribution of each region to the total pulse width. Table 2.4: The inkjet printing parameters used for different nozzle orifice sizes. Nozzle orifice size (µm)  Peak-to-peak voltage (V)  Waveform pulse width (µs)  Typical droplet volume (pL)  60 80  30 41.5  75 75  39 92  below atmospheric pressure at the nozzle orifice, as measured in the air line preceding the ink chamber. A syringe pump was used to control the air pressure in an ink-containing vial. By increasing or reducing the air pressure, the corresponding liquid pressure at the nozzle orifice could be controlled. Table 2.4 shows the parameters used in this research for printing a CNT-SC suspension with both 80 µm and 60 µm orifice diameter nozzles, and Figure 2.17 shows a schematic of the entire inkjet printer setup.  2.4  Substrate Considerations  One of the great advantages of inkjet printing is in its ability to accommodate virtually any relatively planar substrates like polymers or silicon, to more exotic ones  38  Figure 2.17: A schematic of the custom inkjet printing system used. (a) is the two-dimensional stage, (b) is the piezoelectric nozzle, (c) is the ink reservoir, (d) is an in-line sensor for pressure regulation, (e) is the syringe pump pressure controller, and (f) is the custom made C++ software that controls the system. like leaves [89]. When aligning CNTs however, the criteria for substrate selection is more strict. The printed pattern must be a continuous liquid feature, meaning that the liquid contact line must be pinned to the substrate regardless of its geometry. The hydrophobicity of a given substrate is a good metric by which to determine compatibility. A copper substrate, for example, is too hydrophobic to create a continuous liquid pattern, which spontaneously separates into individual droplets rather than maintain its shape. When a substrate is too hydrophilic the time-scale for the pinned evaporation mode is small and the liquid edges readily recede. Fur39  Figure 2.18: The contact angle of a processed CNT SC suspension on PET (a) before, and (b) after a 10 s plasma treatment. ther, contact line pinning appears to be quite weak on hydrophilic materials and they are much more prone to bulging as liquid is added. For this reason, substrates such as glass and oxidized silicon wafers are not ideal candidates. It was found that the polymer polyethylene terephthalate (PET), subjected briefly to an oxygen plasma, had the desired properties. Prior to plasma treatment, the PET has a similar hydrophobicity to copper and is too hydrophobic to form continuous patterns. However, after a 10 s plasma treatment in air using a hand held corona discharge device, the the contact angle changed from 78◦ to 40◦ , as shown in Figure 2.18. Corona discharge is a common method of reducing the hydrophobicity of polymers and there is a diverse body of literature available on the mechanisms responsible [90]. In well controlled oxygen environments, the field produced by the corona device generates oxygen ions, ozone molecules and UV light. The UV light aids in breaking the polymer bonds [91], allowing the ions and ozone molecules to chemically react with the polymer, forming polar OH groups [92]. In an uncontrolled gaseous environment such as ambient air, the mechanisms may be more complex and the results less repeatable, however the plasma treated PET surface consistently demonstrated the required properties for printing continuous liquid features that maintained their shape. A notable feature of the corona treatment of polymers is that the contact angle returns to the untreated value  40  over time, likely due to the reorientation and redistribution of the polar functional groups, or through reactions with free radicals in the air [91]. Because of this, corona treatment was performed immediately prior to the inkjet deposition.  2.5  Peeling Procedure  In order to utilize transmissive polarized light microscopy, as well as to study the sub-surface morphology of the CNTs film, a simple procedure was developed to remove a thin layer from the film surface. Details regarding the results of this procedure are discussed in Section 3.2. It was found that merely applying a piece of adhesive tape to the surface of the film and peeling it off removed a thin layer of CNTs. Four tapes were found to do this successfully: Scotch tape (3M), wafer dicing tape (Epak Electronics), thermal release tape (Nitto Denko) and conductive carbon tape (Epak Electronics). In addition, a polymeric polydimethylsiloxane (PDMS) stamp (Sylgard) proved capable of removing a similarly thin layer. The PDMS was mixed at a 10:1 base to hardener ratio, degassed, and cast into a petri dish where it was then cured at 65◦ C for 3 h. PDMS has been used by others to selectively remove microscale objects [93]. When removed quickly, the PDMS showed strong adhesion, but when removed slowly the adhesion was weak. Thus by controlling the peeling rate it is possible to pick-and-place micro-scale objects. Here the PDMS was capable of removing CNTs but not at transferring them to another substrate. CNTs are far smaller than the micro-scale objects used in related studies and the Van der Waals forces are likely to prevent them from being removed, regardless of the peeling rate. This was tested by applying the PDMS stamp to a CNT film and removing it very slowly. In this case a layer of CNTs was removed the same as if the stamp was peeled away rapidly, suggesting that the inter-layer bonding is significantly weaker than the adhesive force of the stamp. The thickness of a layer, peeled using a PDMS stamp was determined by taking before and after images of its topography with an optical profilometer. Subtracting the height data, which was an average of ten evenly spaced cross sections, showed a reduction of approximately 80 nm after the peeling procedure, as shown in Figure 2.19. CNT samples fabricated on PET required a thin Cr layer for SEM imaging and it was found that once coated, CNTs could not be removed by peeling. Simi-  41  Figure 2.19: Film thickness before (grey) and after (green) a layer is peeled away and the resulting difference in film thickness (red). larly, peeling was not possible until the SC was rinsed away using ethanol. Rinsing involved simply swirling the sample in an ethanol bath for about 1 minute. Several simple experiments were performed to investigate this peeling procedure, primarily to see whether or not the CNTs film could be transfered from one substrate to another once peeled. This could be an important consideration as the adhesive tapes and PDMS used for peeling may not be the desired substrates for a given application. First, a piece of scotch tape containing a freshly peeled layer of CNTs was applied to a clean piece of PET, and by applying heat (140◦ C) and pressure, was successfully removed from the tape and laminated onto the PET. 140◦ C is above the 75◦ C glass transition temperature of PET, but well below its 250◦ C melting point. The scotch tape appeared unaffected by the heating. SEM imaging was attempted in order to determine whether all of the CNTs were transfered from the scotch tape to the PET, however the CNTs appeared to be embedded in a polymer matrix as shown in Figure 2.20 and it was not possible to conclusively determine the morphology. It is likely that this is the polymer adhesive from the scotch tape. Attempts to removed the adhesive using isopropanol, ethanol and a commercial adhesive remover (Elmer, Goo Gone) were unsuccessful and so it was 42  Figure 2.20: An SEM image of a CNT film which has been removed with scotch tape and laminated on PET. not possible to obtain a clear SEM image. The application of stronger organic solvents such as acetone damaged the PET substrate and only proved to worsen the image quality. Attempts at laminating the CNTs onto the more chemically resistive polyvinylidene fluoride (PVDF) were unsuccessful. A similar experiment was performed by applying a piece of scotch tape, that contained a peeled CNT layer, to the adhesive side of a piece of packing tape, which has a much stronger adhesive. As the two pieces of tape were pulled apart, the majority of the CNT film was transferred to the packing tape. Due to the adhesive problem described above, once again it was not possible to ascertain a clear SEM image of the CNT morphologies on either piece of tape.  43  Chapter 3  Results and Discussion In this chapter, the key results of the alignment process are analyzed and discussed. The morphology of the films was investigated primarily using polarized light microscopy and SEM, and these assessment techniques are discussed. The degree of alignment in the CNT films was found to depend on concentration, CNT length, and presence of contaminants. The long-range orientation was found to be a function of the geometry of the printed pattern, and the resulting evaporation behaviour. Interestingly the sub-surface morphology is different from that of the surface, and this phenomenon was also explored. Lastly, preliminary work was performed in isolating the controllably aligned CNTs on the film surface through a peeling procedure.  3.1  CNT Characterization Techniques  Numerous techniques have been utilized for determining the various physical properties of CNTs and CNT films. Only the techniques utilized in this research will be described in this thesis. For a review of several additional techniques, the reader is directed to the following reference [29]. The orientation of the CNTs was the parameter of greatest interest in this research, and the best methods of investigating the morphology are: SEM; atomic force microscope (AFM); polarized light microscopy; and polarized Raman spectroscopy. Unfortunately a polarized Raman spectrophotometer was not available and so its use was not possible. This tool  44  has been used to provide a quantitative metric for comparing the degree of longrange alignment in CNT films [38] and is commonplace in the study of CNT liquid crystallinity.  3.1.1  SEM  SEM provided the most meaningful method of investigation as it allowed for resolution of individual CNTs on the surface of a printed film. The orientation of the CNTs was thus directly observable. A drawback of the SEM, when investigating large macro-scale films, is that individual CNTs are only resolvable at a minimum of 30 k times magnification. For the printed lines described in Section 2.3 this implies that over 450 000 images would need to be taken to acquire a complete picture of the film surface. Clearly this represents an unreasonable time investment. had it been possible to obtain all of the required images, the orientation of the CNTs could be easily quantified, for example by measuring their angles with respect to the edge of the film as described in Section 3.2, but as it is not possible, SEM is only able to give a qualitative description of the global morphology, though locally the CNT angles may still be quantified. The SEM is also an appropriate tool for measuring CNT length as long as the CNTs are distinctly separated from one another. Experimental Details The SEM used in this research was a Hitachi S-4700 field emission SEM. It was found that CNTs do not image well above about 5 keV beam voltages and so imaging was performed between 2.3 keV to 5.0 keV. The ideal beam voltage will depend on the particular substrate. When imaging CNTs it is also necessary to use a secondary electron detector, rather than a primary electron detector. CNTs have very little bulk and so the deeply penetrating primary electrons will not yield useful information about the CNTs unless exquisite care is taken. The lower energy secondary electrons give clear topographical data. Lastly, for CNT films deposited on polymer substrates a 3 nm to 5 nm layer of chromium was sputter coated onto the surface to dissipate electrons and heat. Cr has a very small nucleation size and thus does not diminish feature resolution to the same extent as most other metals.  45  Without this step severe charging distortions and localized melting of the substrate prevent a reasonable resolution from being achieved.  3.1.2  AFM  AFM gives a topographical profile of the film surface with sub-nanometer height resolution. It is thus an ideal tool for measuring CNT diameter, but it can also give information similar to that obtained by an SEM, though without any substrate dependencies. AFM was not used extensively here due to the significant time-scales involved in collecting images. A high quality AFM scan may take 20 minutes, whereas a high quality SEM scan may take less than one minute.  3.1.3  Polarized Light Microscopy  Despite the comparatively low resolution, polarized light microscopy can provide very useful insight into the morphology of a liquid crystalline film. Most materials with long-range anisotropy in their molecular structure exhibit a behaviour known as birefringence. Examples of such materials are liquid crystals, including CNT suspensions and films, many extruded polymer films, such as tape backing, and crystals such as calcite [94]. Birefringent materials possess different indices of refraction along different crystallographic vectors. In the case of CNTs, if the polarization direction of transmitted light is parallel to the long axis of the CNTs it travels at a different speed than when it is perpendicular to the long axis. Because of this, the CNT film will cause a phase shift between the two components and the transmitted light will have a different polarization than the incident light. If the incident light is linearly polarized and is oriented either parallel or perpendicular to the CNTs, it will transmit without change. However, if the polarization direction of the light is at an angle to the CNTs, a portion of the light will travel slower, causing a phase shift, and thus a change in the polarity of the transmitted light. Figure 3.1 demonstrates this effect. Naturally, the thicker the CNT film, the greater the phase shift. Thus film thickness, molecular orientation and degree of birefringence all contribute to the change in polarization of the transmitted light.  46  Figure 3.1: A schematic of the birefringence effect. The unpolarized source light is transmitted through a linear polarizer. The linearly polarized light transmits through a birefringent material at a speed determined by the molecular orientation of the birefringent molecules. The molecular orientation is indicated with a red arrow. When the incident light is parallel or perpendicular to the molecular orientation (a, b) its polarization is maintained and it is extinguished by the analyzer. When the incident light is at an angle to the molecular orientation, the component parallel to the molecules travels slower than the component traveling perpendicular, and the two components become out of phase and the polarization of the transmitted light changes. The components of the transmitted light that are parallel to the analyzer pass through and can be visualized.  47  Experimental Details The polarized light microscope used (Nikon) consisted of, from bottom up: a white light source, a linear polarizer, a rotating sample stage, a second linear polarizer, called the analyzer, and a strain free magnification lens. The polarization axes of the polarizer and analyzer were perpendicular to one another, meaning that light transmitted through the polarizer will be extinguished by the analyzer and the visible field will be dark. If a birefringent sample is placed between the polarizer and analyzer, and its axis of molecular orientation is not parallel to either the polarizer or analyzer, the sample will re-polarize the light, allowing some to transmit through the analyzer. Thus the orientation of large regions of CNTs may be determined by investigating the magnitude of transmitted light. When the CNT orientation is at a 45◦ angle with respect to the polarizer and analyzer, the transmitted light will be the most intense.  3.2  Surface Morphology  The morphology of the CNT film surfaces was investigated using SEM. As previously mentioned, both line and a circle patterns were investigated. As shown in Figure 3.2 and Figure 3.3, under the correct conditions the surface of the CNT film is composed of densely packed and well aligned CNTs. The conditions required to achieve this are investigated below in Section 3.3. Furthermore, the orientation of the CNTs appears to be primarily parallel to the edge of the pattern and persists across the entire surface of the film. Thus it appears that the orientation of the CNTs corresponds directly to the geometry of the pattern and so some degree of control over orientation is achieved as the geometry is easily controlled with the inkjet printer. More specifically, the pattern appears to be related directly to the evaporation behaviour, which in turn is related to the geometry. This is a significant improvement on the results of others, where the orientation of the CNT film was parallel at the film edges, but perpendicular and isotropic towards the middle [23].  48  Figure 3.2: The surface morphology of an inkjet printed line: (a) shows the orientation at the outermost edge. (b) and (f) show the orientation partway across the film, (c) shows the orientation near the end of the line, (d) is a zoomed out image of the line where the inset is a photograph of the line, and (e) shows the morphology in the middle region. The white arrows indicate the angle of the geometrical edge of the line and the black arrows show the average CNT angle at that location.  49  Figure 3.3: The morphology of a printed circular CNT film. The orientation of the CNTs follows the edge in a circular pattern and persists across the entire surface Circular Patterns First the pattern formations and evaporation behaviour of 5 µL drop cast circular droplets of CNTs suspension were investigated. The results herein do not utilize the concentration enhancement scheme outlined in Section 2.3, but use instead the fully processed 0.5 wt% CNT-SC suspension without further changes. For the circle pattern, evaporation proceeded in the expected way, beginning in pinned mode, and upon reaching the receding contact angle, receding inwards. This was readily observed using a goniometer, where the critical contact angle was found to be 18◦ on plasma treated PET and 24◦ on copper, for a fully processed CNT-SC suspension. The pinned mode accounted for 47% of the total evaporation time on plasma treated PET and 70% of the total evaporation time on copper. As seen in Figure 3.4, the internal hydrodynamic flow driven by the contact line pinning caused the formation of a noticeable coffee ring on the plasma treated PET substrate, however on the copper substrate the film was continuous, despite there being a greater degree of contact line pinning. This is likely because copper is more hydrophobic than PET, a fact which has two implications. First, as mentioned in Section 2.3.1, droplets shaped as a spherical cap have a spatially varying evaporation flux profile which is enhanced towards the droplet perimeter. Internal hydrodynamic 50  Figure 3.4: (a) shows a 5 µL drop cast droplet of CNT suspension on plasma treated PET. (b) shows a 5 µL drop cast droplet on copper. flow is expected to arise anytime the contact line is pinned as the fluid lost at the edge must be replenished, however it is clear that this flow will be enhanced when evaporation flux at the edge is higher [95]. The evaporation flux gradient across the droplet surface decreases as contact angle approaches 90◦ , at which point evaporation is constant across the surface and there is very little internal hydrodynamic flow. Thus in the initial stages of evaporation there is a less significant internal flow on the copper surface as compared to the plasma treated PET surface and so fewer CNTs are expected to be shuttled to the edge. The critical receding contact angle is higher on the copper substrate and so the internal hydrodynamic flow velocity is never as high as on the plasma treated PET sample. This, coupled with the theoretical calculation by Hu and Larsen [95] that radial flow velocity is not a linear function of contact angle, but increases more rapidly with decreasing contact angle, supports the more pronounced coffee ring effect on plasma treated PET. In addition, the Marangoni flow that arises due to the temperature gradient across the liquid surface will cause some of the CNTs to flow back toward the middle. Hu and Larsen [96] showed theoretically that this recirculating flow diminishes with decreasing contact angle, so fewer CNTs would be expected to recirculate away from the droplet edge on the plasma treated PET substrate. The second implication of the higher contact angle on copper is that the volume to footprint ratio is higher and so the CNT concentration rises more rapidly with evaporation. As seen in Figure 3.5, at the onset of the receding mode the CNT concentration is 1.3 wt% for the plasma treated PET sample and 2.7 wt% for the copper sample. As concentration increases, CNT mobility within the suspension decreases and so the internal flow 51  Figure 3.5: Change in contact angle and CNT concentration in the pinned evaporation regime on plasma treated PET and copper substrates. on the copper substrate sample will have a lesser effect on the movement of the CNTs than in the plasma treated PET sample, despite the magnitudes of the flow being similar. This hydrophobicity based concentration enhancement was sufficient to cause a liquid crystalline phase transition and resulted in the aligned film seen in Figure 3.3, making copper a good substrate for circular droplet patterns, though given its metallic nature, a poor substrate for electronics applications. Line Patterns Circular patterns are useful for investigating the behaviour of CNTs in suspension and how that behaviour yields the film morphologies observed, however they are not practical for most applications. This is particularly true in that the orientation of the aligned CNTs is also circular, making them difficult to utilize. A much more useful feature is a line in which the CNTs are oriented along the direction of the line. Unfortunately, high aspect ratio features such as lines are not stable on copper and spontaneously segregate into individual droplets upon evaporation. This suggests that the forces responsible for pinning the contact line are weak on copper.  52  Figure 3.6: A 5 layer inkjet printed CNT line on plasma treated PET. Much stronger pinning is evident on plasma treated PET and it is possible to create long, continuous liquid lines. As stated above, coffee rings, rather than continuous films, form on plasma treated PET without concentration enhancement. This is also true for line patterns as shown in Figure 3.6. To form a continuous film the initial concentration of the deposited suspension must be increased such that the CNT concentration at the onset of the receding mode is sufficient for a liquid crystalline phase transition to occur. In this situation, the central portion of the line, which in a more dilute sample would lack alignment, will be composed of aligned CNTs. The experimental details for achieving this concentration enhancement can be found in Section 2.3 and the resulting morphology seen in Figure 3.2. By taking several SEM images at different locations along the line it is possible to obtain a qualitative understanding of the overall CNT behaviour. As in the circle, the orientation of the aligned CNTs appears to correspond with the evaporation behaviour of the liquid line and is shown schematically in Figure 3.7. This suggests that the physical processes leading to long-range orientation are the same regardless of the pattern geometry and substrate. In the initial stages of evaporation all edges are pinned and a buildup of CNTs forms around the perimeter. Next the ends of the line begin to recede while the sides remain pinned. This is because there is an inward directed, surface tension driven net force on the ends of the line. Once the receding contact angle is reached, the sides of the line will also begin to recede. As a result of this evaporation behaviour, the CNTs at the ends of the line are oriented in an arc. In the middle region of the line, which corresponds to the receding sides, the CNTs are oriented primarily in parallel to the edges of the pattern. The orientation is not exactly as expected across the entire film surface. Disclinations characteristic of a liquid crystal are infrequent, but evident. Figure 3.8  53  Figure 3.7: A schematic of the observed evaporation behaviour of an inkjet printed line and the resulting CNT orientation. shows some examples of these disclinations. The perturbation in orientation caused by a disclination appears to affect only the local orientation, the region surrounding the disclination contains CNTs aligned and oriented as expected following the evaporation behaviour. Disclinations will need to be minimized when applying these films to devices as they will undoubtedly have a negative impact on performance. Quantitative Analysis It is possible to obtain a modest quantitative analysis for the CNT orientation from SEM images by calculating the angle of each CNT with respect to the geometrical edge of the pattern. Were it possible to obtain enough SEM images to view the entire area of the film this would be a very good quantitative measurement. A simple MATLAB script was written using the image processing toolbox to automatically calculate the distribution of angles between the CNTs in an SEM image and the 54  Figure 3.8: Several common disclinations on the CNT film. (a) shows a bend deformation, (b) shows an s = - 12 disclination, (c) shows two s = 12 disclinations and one s = - 12 disclination, and (d) shows an s = - 12 disclination leading to an extended bend deformation. x-axis of the image, which could then be compared to any arbitrary vector, presumably the edge of the line pattern. In actuality there is not sufficient distinction between each CNT to separate them out individually. It was observed that a common feature in the CNT films is the formation of regular cracks, which possess an orientation corresponding to the CNTs around them. Sharp physical edges often appear brighter in SEM images and the same is true for the edges of these cracks as seen in Figure 3.9 a. Several filters were applied to help increase the contrast of the edges, which were then isolated as shown in Figure 3.9 b. Once isolated, each edge was turned into an individual object so that it could be analyzed separately. Objects below a specified size (15 pixels) were ignored because it was not possible to determine a relevant orientation with small objects. A best fit line was drawn for  55  Figure 3.9: The steps taken in calculating the spatially dependent orientation of CNTs in an SEM image. (a) filters are applied to emphasize regions of high contrast (ie. edges). (b) edges are identified and isolated as individual objects. (c) A best fit line is drawn for each edge object and overlayed on the original image. (d) The angles of the CNTs with respect to the x-axis, within 3 standard deviations, are plotted as a histogram. each edge object as shown in Figure 3.9 c and the angle of this line with respect to the x-axis was then easily determined. The drawbacks of this method are: the need to reconfigure the filter parameters as brightness and contrast are not always constant across a set of SEM images; the inability to measure individual CNTs; and errors arising due to unexpected features such as contaminants or a strangely shaped cracks. It does however give a statistically relevant picture of the spatially varying orientation angle of the CNTs in an SEM image. The MATLAB code may be found in Appendix A.  56  Figure 3.10: (a) shows the surface layer of an aligned CNT film without crossed polarizers. (b) shows the film under crossed polarizers at a -3◦ angle and (c) at a 71◦ angle. The black cross indicates the orientation of the polarizer and analyzer. Polarized Optical Microscopy Data While SEM images provide direct insight into the orientation of the CNTs over very small areas, and several may be combined to gain a general understanding of the long-range behaviour, it does not provide a complete picture. Polarized light microscopy provides excellent complementary information to the SEM in that it gives a picture of the whole film, however not at the resolution of individual CNTs. A description of light microscopy using crossed linear polarizers is given in Section 3.1.3. When the CNTs are at a 45◦ angle with respect to the polarizer and analyzer, that region will appear bright. The long-range orientation may thus be determined by rotating the sample until the peak in intensity is seen. The polarized light microscope setup used relies on transmissive light, however the as-printed CNT film is too thick, and thus too opaque to transmit light. To circumvent this problem a thin layer of the CNT film was peeled off using a PDMS stamp. The experimental details regarding the procedure may be found in Section 2.5. PDMS was used preferentially over adhesive tape, as the extruded backing of most tapes is birefringent and would skew the results; cast PDMS, however, is not birefringent. Figure 3.10 shows the surface layer of a printed CNT line under crossed linear polarizers, printed under conditions identical to the line from Figure 3.2. Figure 3.11 shows the end of the line under crossed polarizers. When the edge of the line is oriented in parallel to the crossed polarizers, the bottom edge appears bright. This suggests that the orientation of the CNTs along the bottom edge of the line is 45◦ with respect to the polarizers. When the pattern is rotated 57  Figure 3.11: The end of the surface layer of a printed line under crossed polarizers at (a) 0◦ , (b) 45◦ , (c) 90◦ , and (d) 135◦ with respect to the linear polarizers. The bottom left arrows indicate the polarization direction of the polarizers and the bottom right arrows indicate the possible orientations of the CNTs in the bright regions. to 90◦ , the middle region appears bright, suggesting that the CNTs in the middle are oriented either parallel or perpendicular to the edge of the pattern. Another 45◦ rotation causes the other side of the film to appear bright, suggesting that the CNTs there have an orientation of 45◦ . A further 45◦ rotation results in full optical extinction. These results raise an interesting question. The periodicity of the birefringent behaviour is 180◦ rather than the expected 90◦ . That is, when the linearly polarized light is at a 45◦ angle with respect to the CNT direction, there are two possible orientations as shown in Figure 3.12. Typically these two orientations are equivalent and the light intensity transmitted through the analyzer would be the same for both, however this does not appear to be the case for aligned CNT films.  58  Figure 3.12: The two possible 45◦ orientations the CNTs can take with respect to the incident polarized light. The arrow indicates the polarization direction of the light and the lines represent the orientation of the CNTs. For example, in Figure 3.11 b the incident light is at a 45◦ to the CNTs at the bottom edge, which is why that edge appears bright, but it is also at the same angle to the CNTs along the top edge, as discovered when rotated to 135◦ . Thus it is expected that both regions should appear bright at 45◦ and 135◦ ; however this is clearly not the case. This suggests that it is possible to uniquely differentiate the two orientations seen in Figure 3.12 and determine the exact orientation of the CNTs. This result is supported by the findings of Lefebvre and Finnie who showed that single CNTs under polarized light show similar orientational uniqueness. It is tempting to suggest that the effect is due to the changing orientation of layers beneath the surface, however no possible combination of orientations would yield a result that is able to differentiate between 45◦ and 135◦ over a full 360◦ rotation using linearly polarized light. To verify whether or not the results are indeed due to birefringence rather than some unpolarized lighting artifact, several considerations were made. First, the results were carried out in a dark enclosure, ruling out the possibility of reflected ambient light. Secondly, tilt caused by possible unevenness in the PDMS stamp may alter the way light is transmitted, however the results were identical even under conditions of severe tilt. Other artifcats due to contaminants and bubbles within the PDMS were ruled out as their effect does not appear to depend on orientation. Ultimately, the observed birefringence corresponds as expected to the SEM results, but its unique behaviour warrants further investigation. The same analysis may be performed for the central region of the CNT film. 59  Figure 3.13: The middle of the surface layer of a printed line under crossed polarizers at (a) 0◦ , (b) 45◦ , and (c) 135◦ . As shown in Figure 3.13, when the film is parallel to the crossed polarizers it is dark, but becomes bright when rotated by 45◦ . The film stays bright until rotated to approximately 75◦ . This suggests that the middle region of the film contains an intimate mixture of CNT orientations, ranging from parallel/perpendicular to 30◦ /60◦ to the line edge. Following the analysis above, it is likely that only one of the orientations, parallel or perpendicular, is present. Indeed, when the angle is changed to 135◦ the field is dark, as shown in Figure 3.13 c. By the SEM results it is reasonable to assume that the CNTs are relatively parallel to the edge of the film and that this behaviour persists across the entire mid-section of the film. This also means that the configuration shown in Figure 3.12 b leads to transmission through the analyzer, while the configuration in Figure 3.12 a leads to extinction, and so the orientation in all regions can be determined. Between the SEM and polarized optical microscopy 60  Figure 3.14: A printed CNT line in which the ends fully receded before the edges, viewed under crossed polarizers at (a) 0◦ , (b) 45◦ , and (c) 90◦ . data, the evidence suggests that the orientation of the CNTs across the entire film surface corresponds well to the proposed schematic in Figure 3.7. What is also clear is that the CNTs around the outer edge of the line (”coffee ring” region) possess a wide distribution of orientations. This may be related to the extremely high concentrations in that region, which will be discussed in Section 3.3.3. Alternatively it may be due to the geometrical boundary condition of the edge, which is scalloped because of a slightly incorrect droplet spacing in the first printed layer. Polarized optical data also shows the distinction between geometry and evaporation behaviour. In the sample shown in Figure 3.14, the deposition and evaporation rates were poorly matched and a bulge formed in the middle region of the line. This bulge caused evaporation to proceed according to Figure 3.15, yielding the expected long-range CNT orientation shown. The polarized optical microscope  61  Figure 3.15: A schematic showing the observed evaporation behavior of a printed CNT line with a bulge defect and the expected CNT orientation. images shown in Figure 3.14 clearly show that the CNT orientation is as expected. It also shows that while this film has a very similar shape to the other printed lines, the evaporation behaviour and resulting long-range CNT orientation was significantly changed.  3.3  Contributing Parameters  Several parameters were found to contribute to the degree of alignment observed in the printed films.  3.3.1  Contaminants  Contaminants contribute to the bundling of CNTs and prevent liquid crystalline behaviour. When bundled in this way, sonication does not appear to be an effective means of debundling. The CNTs initially used for this research (NanoLab, CVD grown COOH functionalized SWNT) were highly contaminated, despite being la-  62  Figure 3.16: EDS spectrum of contaminated CNTs. beled as greater than 95% purity. The cause of the contamination was unknown at the time and it was systematically attributed to the CNTs. The surfactant, solvent, and physical devices used in the processing of the the CNT suspension were investigated individually and found not to contribute to the observed contamination. Energy dispersive spectroscopy (EDSs) analysis was used on the contaminated CNTs and only a small quantity of metallic impurity was detected, as shown in the EDS spectrum seen in Figure 3.16, suggesting that the contamination was primarily carbonaceous particles. An appropriate purification treatment could thus be identified. It is expected that the 95% rated purity referred only to carbon content as opposed to SWNT content. Ultimately the problem was solved by purchasing new CNTs (Southwest Nanotechnologies, SWeNT SG65) which contained significantly fewer impurities and did not require any purification outside of the procedure described in Section 2.2.  63  Figure 3.17: SEM images of shortened CNTs with an average length of 131 nm.  3.3.2  Length  As predicted by liquid crystal theory, the mesogenicity of a particle is a strong function of its aspect ratio. Zhang et al. [64] showed that very short multi-walled CNTs did not form a nematic phase while longer ones did, and that the quality of the nematic phase was improved when the short CNTs were selectively removed. However, work done by Puech et al. [63] suggested that the opposite was true for single-walled CNTs. By this discrepancy it was clear that the variable warranted investigation. As detailed in Section 2.2.2 three length regimes were studied: short; medium; and long. Short CNTs did not undergo a long-range liquid crystalline phase transition regardless of the concentration tested. The morphology typical of the short CNTs may be seen in Figure 3.17. This is in agreement with the phase diagram seen in Figure 2.3 which shows that the CNT concentration required for a nematic phase transition increases dramatically with decreasing CNT aspect ratio. It is likely that below a certain length the concentration required for a phase transition is so high that the CNTs would not have the mobility necessary to rearrange themselves into an ordered phase. Based on the results here the short CNTs may be below such a limit. The medium length CNTs readily formed an aligned phase as seen in Section 3.2. This result is expected given the phase diagram. Despite the theoretical prediction that mesogenicity should continue to increase with particle aspect ratio, it was found that the long CNTs used in this study did not 64  Figure 3.18: SEM image of long CNTs with an average length of 787 nm. exhibit a liquid crystal phase transition, as shown in Figure 3.18, the morphology of which is representative of the entire film. There are a number of possible explanations for this: The longer, more massive CNTs may have a nemaitc phase transition time constant which is much greater than the evaporation time; low power bath sonication may not sufficiently debundle the CNTs; or longer CNTs may have a lower mesogenicity. To test the phase transition time, a 5 µL droplet of long CNTs was cooled to marginally above the dew point of the local atmosphere (approximately 16 ◦ C), allowing it to evaporate over approximately 8 hours. The resulting film did not possess any alignment, suggesting that if the transition time is the cause it is too long to be compatible with inkjet processes. Bath sonication has been shown to be an effective method of suspending and dispersing CNTs and further, the centrifugation step would have removed any bundles, suggesting that it is an intrinsic property of long CNTs not to form a nematic phase. As suggested by others, it may be a result of their tortuosity. Increased flexibility causes long CNTs to deviate from predictions based in rigid rod theory [97] and increases the likelihood of entanglements, which have a severely negative impact on alignment as discussed previously in Section 3.3.1. This experiment suggests that medium length CNTs with aspect ratios of approximately 450 are the most mesogenic and that a low power setting on a high power ultrasonicator is the optimal suspension technique.  65  Figure 3.19: The morphology of a 5 layer printed line using a dilute, 0.0125 wt%, CNT suspension. (a) shows some alignment is evident along the edge of the line, and (b) shows that the middle of the line contains sparse and randomly oriented CNTs.  3.3.3  Concentration  In lyotropic liquid crystals, the liquid crystal phase transition is fundamentally dependent on the concentration of particles and so it is expected that this parameter has a significant impact on the CNT alignment. The very low concentration sample exhibited alignment around the droplet periphery, but the rest of the film contained randomly oriented CNTs as shown in Figure 3.19. In such a dilute suspension it is possible that the alignment is not due to a liquid crystalline phase transition, but rather by the internal hydrodynamic flow. As a CNT approaches the edge of the droplet, one end may become pinned by the edge. The internal flow will induce a torque at the unpinned end, turning the CNT and aligning it with the edge. This method of alignment has been explained by Sharma and Strano [60] and is expected only to hold true when the CNT-CNT interactions are minimal. The low concentration sample diverged somewhat from the expected behaviour in that the coffee ring did not contain aligned CNTs. This somewhat contradicts observations by others [34], where under similar conditions significant coffee ring alignment was noted. The morphology of the material in the coffee ring suggests a composition of carbonaceous particles rather than CNTs, and these particles may be interfering with the alignment there. This may be true in the very low concen-  66  tration suspension also, however the particle-CNT interaction is minimal owing to the low concentration. The evaporation behaviour of this sample was such that one end quickly receded while the other remained pinned, forming a pool of suspension at one end. The reason one end depinned preferentially was likely due to inhomogeneity in surface properties, possibly roughness or non-uniform plasma treatment. A slight tilt could have also caused such behaviour however it was observed from different samples that each end had approximately the same likelihood of preferentially receding. The resulting film morphology may be seen in Figure 3.20. The end of the film where a pool formed was continuous and contained large regions of alignment. The orientation appeared to be parallel to the edge of this region, however an insufficient number of SEM images were taken to say so conclusively. As interesting as the aligned region is, the resulting film is not useful for applications since the majority of the film is sparsely populated with randomly oriented CNTs. The concentration must be increased such that the CNT film is continuous across its entire footprint. The medium concentration regime produced the results seen in Section 3.2. What differentiates these results from those of others, and from the low concentration results, is that the alignment effect spans the entire surface of the film, and is not just isolated to the coffee ring region. This clearly suggests that the concentration of CNTs remaining in suspension (those not deposited in the coffee ring) at the onset of the receding evaporation mode is sufficiently high to exhibit a liquid crystalline phase transition. The following analysis gives an estimate of the CNTs concentration during the receding evaporation mode. In a 20 layer inkjet printed line a total volume of 1.98 µL was deposited, corresponding to 10 µg of CNTs. Observing the evaporation behavior of the CNT suspension on plasma treated PET using a goniometer, the contact angle at the onset of the receding mode was found to be 15.5◦ . The per-unit-length volume of a cylindrical cap can be determined with the relation: VL = R2  θc − cosθc sinθc , sin2 θc  where the radius R, defined as half the width of the central region of the line, was 0.24 mm as determined by the profilometer data shown in Figure 3.21, and θc is the contact angle [98]. Assuming that the CNTs are invariant along the length of 67  Figure 3.20: The morphology of a 5 layer printed CNT line. (d) shows a photograph of the printed line, (a) shows the morphology of the coffee ring, and (b) and (c) show the morphology of the film at the pinned end. the printed line, the expected concentration in a small segment of 1 mm in length may be estimated. The expected volume in such an area is 12.6 nL and the portion of CNTs corresponding to that segment has a mass of 1.33 µg. It can be seen by the profilometer data that approximately 62% of the CNTs make up the coffee ring, leaving 38% in the central region. The CNTs in the coffee ring have already been deposited at the onset of the receding mode leaving an approximate concentration of 4 wt% in the central area. This value is often reported as the concentration corresponding to a liquid crystalline phase transition in CNT suspensions [38, 99]. Interestingly, the polarized optical microscopy results in Section 3.2 suggest that the coffee ring does not contain well oriented CNTs, despite the concentration being undeniably high there suggesting that beyond a certain concentration the  68  Figure 3.21: Profile of a 20 layer printed CNT line taken with an optical profilometer. The inset shows a segment of the profilometer image corresponding to the thickness data below. orientation is diminished. Indeed when 65 layers were printed, the long range CNT orientation was diminished across the entire film. That is, large regions contained CNTs which were oriented perpendicular to the edge of the pattern, where in a lower concentration sample they would have been parallel. This suggests there is a trade-off in increasing concentration. At higher concentrations the coffee ring does not possess the desired CNT orientation, while the central region does, but at lower concentrations the central region is not continuous. Thus, the ideal concentration may lie somewhere between 5 and 20 layers. Based on the low concentration results it may not be possible to obtain a well oriented coffee ring, however it may be possible to minimize its size. At 40 layers the results were qualitatively identical to 20 layers meaning that no apparent benefit is gained from increasing concentration beyond what is minimally required for the liquid crystalline phase transition, at least not in the context of achieving long range alignment and mutual CNT orientation. Clearly at yet higher concentrations the impact on orientation is detrimental. The same qualitative results were observed for concentration changes in circular films.  69  3.4  Sub-Surface Morphology  Many of the prior published studies investigating liquid crystallinity in CNTs rely solely on surface characterization techniques such as SEM, AFM, polarized Raman spectroscopy, and reflective polarized light microscopy. These measurements do not provide direct insight into the film properties below the surface. As such, information in literature about the sub-surface film morphology is scarce. Huang et al. [100] gave a brief description of the effect but did not provide clear experimental evidence. Bravo-Sanchez et al. [69] included the isotropic sub-surface morphology in the supplementary material but did not provide an explanation. Furthermore, this study was regarding a CNT-polymer composite material in which it was the liquid crystalline properties of the polymer that were attributed to the observed CNT alignment. Here the sub-surface film morphology was investigated by peeling off the top layer of CNTs with either an elastomeric PDMS stamp, or adhesive tape as described in Section 2.5. Peeling off the surface layer allowed both SEM investigation of the sub-surface morphology as well as the transmissive polarized light microscopy of the aligned surface layer and subsequent layers, which is discussed in Section 3.2. Interestingly it was found that the alignment and long range orientation observed on the surface does not persist throughout the depth of the film. Figure 3.22 shows several SEM micrographs of the subsurface morphology of films which had well aligned and oriented surfaces. Furthermore, the underside of a peeled film was investigated to determine whether the peeled layer contained solely aligned CNTs; in which case the underside would contain well aligned CNTs. This was found not to be the case, and the underside of the peeled film had a similar morphology to the sub-surface of the film it was peeled from, as seen in Figure 3.23. The film was removed with conductive carbon tape to make SEM imaging possible.  From the SEM images much of the sub-surface morphology appears to be  isotropic, however some regions clearly still contain well aligned CNTs. There is a startling combination of textures in these images. Regions of aligned CNTs appear to be covered in a dense, somewhat amorphous material. Upon closer examination this material appears to be a mixture of randomly oriented CNTs and nanoparticles. The nanoparticles are likely carbonaceous in nature as there is no significant  70  Figure 3.22: Several SEM images of the sub-surface morphology of the aligned CNT films. All images show the sub-surface morphology of a drop cast circle on copper which had a well aligned surface. (a) and (b) show the morphology after a single peeled layer has been removed. (c) and (d) show the morphology after two consecutive layers have been removed, and (e) and (f) show a crack in a film that did not have its surface removed.  71  Figure 3.23: An SEM image of the underside of a peeled layer of CNTs. The film was peeled off using conductive tape. contrast between them and the CNTs under SEM. Surfactant agglomerates and metallic particles both appear bright in the SEM compared to the CNTs, and are the only other possible particles to exist in the suspension (Appendix B). The carbonaceous particles are likely a result of the intense ultrasonication procedure, or may have been present in the CNT soot. The solution processing scheme used was designed only to remove large CNT agglomerates, not individually suspended nanoparticles or shortened CNTs. These results raise the question as to why the well oriented and aligned CNTs are constrained to the surface of the film, and why none of the particles found below are evident on the surface. There are several possible explanations for this phenomenon. First, it has been theorized that for lyotropic liquid crystals, a difference in phase may arise between the bulk and surface regions of the fluid. This prediction by Matsuyama and Kato [101] takes into account the differences in free energy of the two regions and is based on Onsager’s theory [71] for rigid rods in solution. For particles that are attracted to a surface rather than repelled by it, this will lead to a nematic liquid crystalline phase transition across the surface while the bulk remains isotropic. The effect is well known for thermotropic liquid crystals and has been determined both theoretically and experimentally though x-ray reflectivity studies [102–104]. The theory also predicts that further increasing the bulk  72  concentration will cause a continual increase in the thickness of the nematic phase until the entire bulk has undergone the phase transition. Important assumptions are made as part of this theory, namely that the particle aspect ratio is constant, which is not the case here, and that the particles are attracted to the surface (liquid-air interface). Some support exists for the latter assumption and also helps explain why the short CNTs and nanoparticles are not seen on the film surface. CNTs would be attracted to the liquid-vapour surface if they retained some hydrophobicity. If the longer CNTs were more hydrophobic than the short CNTs, it may explain why they are preferentially driven to the surface. Such a difference in hydrophobicity could result from differences in surfactant adsorption, be it in wrapping morphology or overall quantity. Preliminary atomistic simulations of surfactant wrapping on CNTs have studied the surfactants SC [105] and SDS [106, 107] and suggest the possibility of non-uniform surfactant wrapping at different loading values, although only diameter is taken into consideration and not tube length. Sonication induced scission was found to be only minimally dependent on diameter [76] so it is assumed that each length regime contains a similar distribution of tube diameters. Differences in surfactant wrapping and loading of long and short CNTs may be caused by preferential surfactant binding at the ends of the tubes. Using EDS the ratio between carbon and sodium atoms was investigated for the two length regimes. To account for any surfactant gradients resulting from the centrifugation process and ensure initial conditions were the same, a sample of each length regime was dispersed in a 15 wt% SC solution at a 5:1 dilution and mixed using bath sonication for 10 minutes followed by a post evaporation rinse in ethanol. The results showed that short CNTs had a C/Na ratio of 29, and the medium length CNTs had a ratio of 48, suggesting that medium length CNTs have less bound surfactant and may have a greater affinity to the liquid-vapor surface as a result. An additional explanation arises from a common phenomenon in polymer and colloidal chemistry known as “crusting” [108]. When a sessile droplet of polymer or highly concentrated colloidal suspension evaporates in the pinned evaporation mode, the particle concentration near the the droplet surface increases [109]. At a critical concentration a crust forms on the surface of the droplet as shown schematically in Figure 3.24. This may be a solid, elastic crust in the case of a polymer solution, or a gel crust in the case of a colloidal system. The formation of such 73  Figure 3.24: A schematic showing the crusting phenomenon. (a) shows the initial state of the droplet, where the particles are evenly distributed. The particles represent CNTs in this case. (b) shows the influences involved in the redistribution of the particles as evaporation proceeds. The particle concentration is increased by the number of particles shown outside the droplet. The red arrows indicate the internal hydrodynamic flow direction and the green arrows indicate the downward direction of the liquid surface. (c) shows the expected distribution of particles partway through the evaporation process. a crust is a result of the competition between diffusion kinetics and evaporation rate. If the downward velocity of the evaporating surface is greater than the velocity of CNTs diffusing away from the surface, the concentration of CNTs near the surface will increase. If the CNTs have an affinity for the liquid-vapour interface, as suggested above, then this phenomenon would be more pronounced. Several mathematical models exist describing the crusting effect and the resulting morphological effects on the dried film. In the SC-CNT suspensions used here, no crusting was directly observed, that is, no undulations appeared on the droplet surface during evaporation, however the underlying principle leading to the crusting phenomenon, a particle concentration enhancement near the surface, may still be evident. There are two pieces of evidence in support of the crusting phenomenon. First, if the CNT distribution shown in Figure 3.24 is accurate, there would be a reduction in CNT density in the middle of the droplet. Figure 3.2e shows the middle of a printed line, where large gaps between regions of alignment can be seen, suggesting the density of aligned CNTs is lower there than in the rest of the film  74  Figure 3.25: An SEM image showing an aligned layer of CNTs covering a highly bundled film. (Figure 3.2b). Secondly, some films created from suspensions which were not centrifuged, and thus contained agglomerates, still exhibited regions of alignment. The aligned CNTs in these regions appear to be draped over the rest of the deposited film meaning they were the last thing to be deposited. Oddly the orientation of the surface is not impacted by the presence of the large agglomerates, as shown in Figure 3.25 suggesting that there was a physical disconnect between the two during the liquid crystalline phase transition of the surface CNTs. This phenomenon is not mutually exclusive to the theoretical prediction based on free energy described above, and both may contribute to the observed surface alignment. Even though there is clear evidence of aligned CNTs beneath the film surface, the long-range orientation is not maintained. This is shown in Figure 3.22e and Figure 3.22f, where it is clear that the orientation shifts between layers. It is unclear whether this shift in orientation is the common behaviour, or if it is a chiral deformation. This figure also shows another interesting morphological feature of the CNTs film. The film is composed of stacked planar layers. Indeed, the cross section seen in Figure 3.26 shows these layers clearly and suggests a smectic liquid crystalline structure rather than a nematic one. Further analysis must be done to validate such a claim. This explains why it is possible to peel thin layers off the surface of the film. Were the CNTs entangled three dimensionally throughout the  75  Figure 3.26: An SEM image showing the cross section of an aligned CNT film  Figure 3.27: The second layer peeled from a printed CNT line under crossed polarizers at (a) 0◦ , (b) 45◦ , and (c) 135◦ .  76  film, it would not be possible to peel off consecutive layers. Figure 3.27 shows a polarized light microscopy image of the second peeled layer from an aligned CNT film. It is clear that the birefringence is not as vibrant as in the first peeled layer, seen in Figure 3.13, but that it is still evident and suggestive of a long range CNT orientation which resembles that of the first layer. SEM images after two consecutive peels, seen in Figure 3.22c and Figure 3.22d, similarly show that aligned CNTs exist at that depth. Based on the cross sectional image this is not unexpected. By the SEM images of the surface, the strong birefringence of the first peeled layer, and the weak birefringence of the second peeled layer, it can be concluded that the first peeled layer contains primarily well aligned and mutually oriented CNTs. Thus, regardless of the morphology beneath the surface layer, it is possible to isolate the aligned CNTs for further use through the peeling procedure.  77  Chapter 4  Conclusion and Future Work 4.1  Summary of the Results  A process was designed for inkjet printing CNT films with long-range mutual alignment. The orientation of the CNTs was found to correspond to the geometry of the pattern, which was dictated by the inkjet printer. The mechanism of alignment utilized was the intrinsic lyotropic liquid crystallinity evident in highly concentrated CNT suspensions and so no processes outside of inkjet printing were required to fabricate the films. The high concentration required results in a suspension which is too viscous to deposit using inkjet printing and so a unique process was designed to facilitate the use of this technology. A dilute suspension was used to print a continuous liquid pattern, and the CNT concentration within that pattern was incrimentally increased by further, overlapping deposition. By matching the deposition rate to the evaporation rate of the pattern, its shape will not change, but there will be a net increase in the concentration of CNTs. The concentration can be raised sufficiently to drive a liquid crystalline phase transition causing long-range alignment of the CNTs across the pattern. This is a significant improvement over other published liquid crystal based alignment strategies which either do not possess a consistent long-range orientation, or require the CNTs to sit undesturbed for several weeks followed by mechanical shearing, making pattern control challenging. Here the fabrication takes only several minutes. The morphology of the aligned films was investigated using SEM and polar78  ized light microscopy and several interesting features were identified. First, it was discovered that the surface of the film held the greatest degree of alignment, and that the alignment degraded throughout the depth of the film. To investigate the sub-surface morphology the surface layer of the film was removed using either adhesive tape or an elastomeric stamp, exposing the morphology beneath for investigation using SEM. The sub-surface of the film contained a mixture of aligned CNT bundles, randomly oriented CNTs, and what appeared to be CNT fragments, and carbonaceous nanoparticles. The curious nature of the surface-isolated alignment was investigated, and while it warrants a more thorough investigation, several explanations were proposed. The morphology of the material depth was found to be one of stacked planar layers and is indicative of a smectic liquid crystal rather than a nematic one. In addition, the effects of CNT aspect ratio and concentration on alignment quality were investigated and it was found that when CNTs are too short, too long, at too low a concentration, or at too high a concentration, that alignment was diminished or lost. The optimal length regime had an average aspect ratio of 450. The optimal concentration is somewhat more complex to determine quantitatively as it is spatially varying within the liquid feature, however the estimated CNT concentration at the onset of the receding mode of evaporation, which is the stage in which most of the film area is deposited, was 4 wt%.  4.2  Future Work  This thesis has outlined a process for successfully and repeatably creating aligned CNT films, but there is still considerable room for growth in several aspects of this research: Optimization of the alignment process, furthering the physical understanding of the system, refinement and further characterization of the peeling process, and application and electrical characterization of the films.  4.2.1  Process Optimization  Suspension Optimization Ultimately it would be ideal to achieve a film in which alignment persists throughout its depth and avoid the complexity involved in removing the surface layer. By 79  looking at the sub-surface morphology results, it appears that much of the area is composed of CNT fragments and carbonaceous nanoparticles. It was shown that short CNTs did not exhibit long-range alignment and so it follows that the presence of these short CNTs and nanoparticles diminish alignment when present in longer CNT length regimes. These poorly mesogenic particles may be responsible for the poor sub-surface alignment observed and selectively removing them could increase the degree of alignment throughout the film. Having a high degree of control over the CNT length is also important in further characterizing the physics of this system as it clearly plays a pivotal role in the alignment process. There are several ways of narrowing the CNT length distributions. Arguably the simplest is by using several rounds of ultracentrifugation. Intense ultracentrifugation moves long CNTs and bundles toward the bottom of the centrifuge tube, while shorter CNTs remain near the top. This was the method used to create the short length regime, but in the same manner it can be used to remove the short CNTs. Two rounds of ultracentrifugation would be required: A low intensity round where the supernatant is reseverd in order to eliminate any bundles, followed by a high intensity round where the supernatant is removed. By repeating the second round several times it should be possible to remove the majority of the short CNTs and low mass particles. Length fractionation is also possible using both SEC and FFF and these will be more effective than ultracentrifugation at isolating highly specific length fractions, though the yield of both methods is considerably lower than that of ultracentrifugation. Printing Optimization The inkjet process itself can be further optimized also. The process used in this research focussed on low resolution patterns which are appropriate for investigative and proof of concept purposes, and some applications, however increasing the resolution will likely be important for many electronics applications. Sharma et al. [59] showed that the internal hydrodynamics involved in droplet evaporation are not impacted by scaling above 3 µm line widths. Inkjet technology cannot currently achieve such fine resolutions and so there is no physical reason why this process cannot be scaled within the limits of inkjet printing (approximately 20 µm line  80  widths), though this boundary is being continually pushed. There are some technological challenges associated with scaling down the resolution of this method. As feature size is reduced, the deposition rate is aliased. Deposited droplets are fixed, discrete units of volume, meaning that as pattern features are reduced, control over the deposition rate becomes more coarse. The extreme case being a pattern composed of a single droplet, where it would be impossible to match the evaporation rate without allowing the pattern to evaporate completely.  4.2.2  Physical Understanding  The physical problem of droplet evaporation and all of the processes involved is not a trivial one. Indeed, the analytical solution to the evaporation of a pure solvent was solved little more than a decade ago [66]. Since then, the internal hydrodynamic flow in a pure solvent has been solved for [95] under certain conditions, and the movement of innert particles within that flow has also been investigated analytically given some basic assumptions [110]. However, the real physical problem is considerably more complex. In the system described in this research, several physical phenomena are at play. Some of the following points would need to be considered in a comprehensive analysis: • The internal hydrodynamic flow will cause particle redistribution. How anisotropic particles, particularly ones with varying dimensionality respond to the flow must be considered. • The particle redistribution will give rise to particle concentration gradients and thus particle diffusion. • The surfactant will also be redistributed, creating a surface tension gradient and giving rise to a recirculating Marangoni flow. • Evaporation varies across the droplet surface, giving rise to temperature gradients due to the latent heat of evaporation. Surface tension is temperature dependent and will give rise to Marangoni flow. • The reduction in height of the liquid surface can cause a local increase in particle concentration there, particularly if the particles are attracted to the 81  surface. The accumulation of these particles can lead to a reduction in the evaporation rate. The degree to which the evaporation rate changes may depend on the phase structure of the particles near the surface. A highly packed liquid crystalline phase may allow less solvent to evaporate than a porous isotropic phase. • The shape of the pattern is both semi-spherical and semi-cylindrical in this research, but ideally should be arbitrary.  4.2.3  Film Analysis  The underlying physical nature of the surface-isolated alignment is still very much in question. Whether it is an effect brought on by the underlying phenomenon leading to crusting, differences in orientational entropy between the surface and the bulk, selective surface adsorption of the most mesogenic CNTs, or a combination of those phenomena and others, is not conclusively understood. What is clear, based on observations such as that seen in Figure 3.25, is that it is a surface effect, and it may be possible to investigate the evolution of the suface morphology using reflective polarized light microscopy. Reflective polarized light microscopy gives similar information to that of transmissive polarized light microscopy, however reflective light is used and so the surface of opaque samples may be investigated. By watching a droplet evaporate under a reflective polarized light microscope the evolution of the birefringence may show how the aligned surface layer forms. This technique has been used by others in the investigation of evaporating CNT-polymer composite materials [111].  4.2.4  Peeling Procedure  If achieving alignment that persists throughout the film depth is found not to be possible, the peeling procedure must be further refined. In particular it must be possible to remove the aligned surface layer and transfer it completely to another substrate. Much of the peeling techniques utilized here were based on the competing strengths of different adhesives. A better approach may be lamination, where both heat and pressure are applied to transfer the CNTs. Lamination was investi-  82  Figure 4.1: A proposed CNT transistor architecture which can be almost entirely inkjet printed. gated in this research to a limited extent but a more rigorous experimental approach may yield better results.  4.2.5  Applications  The application of aligned CNT films extends to many areas of thin film electronics and sensors, but two possibilities include strain sensors and thin film transistors. The gauge factor of a single CNT has been found to be very high under axial strain (600 to 1000) [112]. This has sparked several different strain gauge architectures based on CNTs [113–115]. In the cases of thin film strain gauges based on randomly oriented CNT films, the reported gauge factors were quite low (4.5) [116]. A film composed of well aligned CNTs may exhibit a marked increase in gauge factor compared to a randomly oriented film. The peeled surface layers from the films developed in this research may be ideal candidates for such a study. A second possible application is a thin film transistor. Others have shown that CNT transistors composed of well aligned CNTs have better performance compared to those made of randomly oriented CNTs [2]. In this application the film may be beneficial even without isolating the aligned surface layer. In the transistor architecture shown in Figure 4.1, the surface of the CNT film forms the channel. The channel of a transistor is typically very thin so it is likely that channel conduction would take place primarily through the aligned film surface. The proposed architecture can be fabricated entirely using inkjet printing, following Li et al. [117], except for the metal oxide deposition step. In addition, density gradient ultracentrifugation would need to be used to remove the metallic CNTs in the printed suspension and eliminate the possibility of a percolating metallic network,  83  which would be detrimental to the on/off current ratio of the transistor.  84  Bibliography [1] S. Iijima, “Helical microtubules of graphitic carbon,” Nature, vol. 354, no. 6348, pp. 56–58, 1991. [2] M. Engel, J. Small, M. Steiner, M. Freitag, A. Green, M. Hersam, and P. Avouris, “Thin film nanotube transistors based on self-assembled, aligned, semiconducting carbon nanotube arrays,” Acs Nano, vol. 2, no. 12, pp. 2445–2452, 2008. [3] Y. Jung, S. Kar, S. Talapatra, C. Soldano, G. Viswanathan, X. Li, Z. Yao, F. Ou, A. Avadhanula, and R. Vajtai, “Aligned carbon nanotube-polymer hybrid architectures for diverse flexible electronic applications,” Nano Letters, vol. 6, no. 3, pp. 413–418, 2006. [4] N. Sinha, J. Ma, and J. Yeow, “Carbon nanotube-based sensors,” Journal of Nanoscience and Nanotechnology, vol. 6, no. 3, pp. 573–590, 2006. [5] C. Li and T. Chou, “Strain and pressure sensing using single-walled carbon nanotubes,” Nanotechnology, vol. 15, p. 1493, 2004. [6] S. Hong and S. Myung, “A flexible approach to mobility,” Nature nanotechnology, vol. 2, pp. 207–8, 2007. [7] Q. Wang, S. Yang, Y. Yang, M. Chan-Park, and Y. Chen, “Charge transfer between metal clusters and growing carbon structures in chirality-controlled single-walled carbon nanotube growth,” The Journal of Physical Chemistry Letters, vol. 2, pp. 1009–1014, 2011. [8] R. Sankaran, “Towards chirality-controlled swcnts: can a plasma help?,” Journal of Physics D: Applied Physics, vol. 44, p. 174005, 2011. [9] M. Arnold, A. Green, J. Hulvat, S. Stupp, and M. Hersam, “Sorting carbon nanotubes by electronic structure using density differentiation,” Nature nanotechnology, vol. 1, no. 1, pp. 60–65, 2006. 85  [10] A. Javey, J. Guo, Q. Wang, M. Lundstrom, and H. Dai, “Ballistic carbon nanotube field-effect transistors,” supernovae, vol. 118, pp. 1766–1776, 1999. [11] J. Salvetat, J. Bonard, N. Thomson, A. Kulik, L. Forro, W. Benoit, and L. Zuppiroli, “Mechanical properties of carbon nanotubes,” Applied Physics A: Materials Science & Processing, vol. 69, no. 3, pp. 255–260, 1999. [12] P. Ajayan, L. Schadler, C. Giannaris, and A. Rubio, “Single-walled carbon nanotube polymer composites: strength and weakness,” Advanced Materials, vol. 12, no. 10, pp. 750–753, 2000. [13] J. Coleman, U. Khan, W. Blau, and Y. Gun’ko, “Small but strong: A review of the mechanical properties of carbon nanotube-polymer composites,” Carbon, vol. 44, no. 9, pp. 1624–1652, 2006. [14] J. Isaacs, A. Tanwani, M. Healy, and L. Dahlben, “Economic assessment of single-walled carbon nanotube processes,” Journal of Nanoparticle Research, vol. 12, no. 2, pp. 551–562, 2010. [15] M. LeMieux, M. Roberts, S. Barman, Y. Jin, J. Kim, and Z. Bao, “Self-sorted, aligned nanotube networks for thin-film transistors,” Science, vol. 321, no. 5885, p. 101, 2008. [16] R. Duggal, F. Hussain, and M. Pasquali, “Self assembly of single walled carbon nanotubes into a sheet by drop drying,” Advanced Materials, vol. 18, no. 1, pp. 29–34, 2006. [17] H. Shimoda, S. Oh, H. Geng, R. Walker, X. Zhang, L. McNeil, and O. Zhou, “Self-assembly of carbon nanotubes,” Advanced Materials, vol. 14, no. 12, p. 899, 2002. [18] Y. Zhou, L. Hu, and G. Grner, “A method of printing carbon nanotube thin films,” Applied Physics Letters, vol. 88, p. 123109, 2006. [19] Y. Li, I. Kinloch, and A. Windle, “Direct spinning of carbon nanotube fibers from chemical vapor deposition synthesis,” Science, vol. 304, no. 5668, p. 276, 2004. [20] C. S. Han, J. W. Song, J. Kim, Y. H. Yoon, B. S. Choi, and J. H. Kim, “Inkjet printing of single-walled carbon nanotubes and electrical characterization of the line pattern,” Nanotechnology, vol. 19, no. 9, p. 095702, 2008. 86  [21] B. Merritt, “Semiconductor industry in transition,” in CMOS Emerging Technologies, 2011. [22] T. Takenobu, H. Okimoto, K. Yanagi, Y. Miyata, H. Shimotani, H. Kataura, and Y. Iwasa, “Tunable carbon nanotube thin-film transistors produced exclusively via inkjet printing,” Advanced Materials, vol. 22, no. 36, pp. 3981–3986, 2010. [23] A. Blayo, A. Denneulin, J. Bras, F. Carcone, and C. Neuman, “Impact of ink formulation on carbon nanotube network organization within inkjet printed conductive films,” Carbon, vol. 49, no. 8, pp. 2603–2614, 2011. [24] T. Mustonen, “Inkjet printing of carbon nanotubes for electronic applications,” phd thesis, University of Oulu, 2009. [25] A. C. Ferrari, P. Beecher, P. Servati, A. Rozhin, A. Colli, V. Scardaci, S. Pisana, T. Hasan, A. J. Flewitt, J. Robertson, G. W. Hsieh, F. M. Li, A. Nathan, and W. I. Milne, “Ink-jet printing of carbon nanotube thin film transistors,” Journal of applied physics, vol. 102, no. 4, p. 043710, 2007. [26] T. Takenobu, N. Miura, S. Y. Lu, H. Okimoto, T. Asano, M. Shiraishi, and Y. Iwasa, “Ink-jet printing of carbon nanotube thin-film transistors on flexible plastic substrates,” Applied Physics Express, vol. 2, no. 2, p. 025005, 2009. [27] A. Behnam, J. Guo, and A. Ural, “Effects of nanotube alignment and measurement direction on percolation resistivity in single-walled carbon nanotube films,” Journal of applied physics, vol. 102, p. 044313, 2007. [28] S. Kang, C. Kocabas, T. Ozel, M. Shim, N. Pimparkar, M. Alam, S. Rotkin, and J. Rogers, “High performance electronics based on dense, perfectly aligned arrays of single walled carbon nanotubes,” Nature Nanotechnology, vol. 2, pp. 230–235, 2007. [29] J. Fagan, B. Bauer, E. Hobbie, M. Becker, A. Hight Walker, J. Simpson, J. Chun, J. Obrzut, V. Bajpai, and F. Phelan, “Carbon nanotubes: Measuring dispersion and length,” Advanced Materials, vol. 23, pp. 338–348, 2011. [30] T. Hasan, V. Scardaci, P. Tan, A. Rozhin, W. Milne, and A. Ferrari, “Stabilization and debundling of single-wall carbon nanotube dispersions in n-methyl-2-pyrrolidone (nmp) by polyvinylpyrrolidone (pvp),” The Journal of Physical Chemistry C, vol. 111, no. 34, pp. 12594–12602, 2007.  87  [31] S. Zhang and S. Kumar, “Carbon nanotubes as liquid crystals,” Small, vol. 4, no. 9, pp. 1270–1283, 2008. [32] S. Ramesh, L. Ericson, V. Davis, R. Saini, C. Kittrell, M. Pasquali, W. Billups, W. Adams, R. Hauge, and R. Smalley, “Dissolution of pristine single walled carbon nanotubes in superacids by direct protonation,” The Journal of Physical Chemistry B, vol. 108, no. 26, pp. 8794–8798, 2004. [33] A. Dillon, T. Gennett, K. Jones, J. Alleman, P. Parilla, and M. Heben, “A simple and complete purification of single-walled carbon nanotube materials,” Advanced Materials, vol. 11, no. 16, pp. 1354–1358, 1999. [34] S. Zhang, Q. Li, I. Kinloch, and A. Windle, “Ordering in a droplet of an aqueous suspension of single-wall carbon nanotubes on a solid substrate,” Langmuir, vol. 26, no. 3, pp. 2107–2112, 2009. [35] L. Lu and W. Chen, “Large-scale aligned carbon nanotubes from their purified, highly concentrated suspension,” ACSNano, vol. 4, no. 2, pp. 1042–1048, 2010. [36] N. Nakashima, S. Okuzono, H. Murakami, T. Nakai, and K. Yoshikawa, “Dna dissolves single-walled carbon nanotubes in water,” Chemistry Letters, vol. 32, no. 5, pp. 456–457, 2003. [37] R. C. Haddon, J. Sippel, A. G. Rinzler, and F. Papadimitrakopoulos, “Purification and separation of carbon nanotubes,” MRS Bulletin, vol. 29, no. 4, pp. 252–259, 2004. [38] P. Poulin, N. Puech, C. Blanc, E. Grelet, C. Zamora-Ledezma, M. Maugey, C. Zakri, and E. Anglaret, “Highly ordered carbon nanotube nematic liquid crystals,” Journal of Physical Chemistry C, vol. 115, no. 8, pp. 3272–3278, 2011. [39] E. Joselevich, H. Dai, J. Liu, K. Hata, and A. H. Windle, “Carbon nanotube synthesis and organization,” Carbon Nanotubes, pp. 101–164, 2008. [40] A. Rinzler, J. Liu, H. Dai, P. Nikolaev, C. Huffman, F. Rodriguez-Macias, P. Boul, A. Lu, D. Heymann, and D. Colbert, “Large-scale purification of single-wall carbon nanotubes: process, product, and characterization,” Applied Physics A: Materials Science & Processing, vol. 67, no. 1, pp. 29–37, 1998.  88  [41] I. Chiang, B. Brinson, A. Huang, P. Willis, M. Bronikowski, J. Margrave, R. Smalley, and R. Hauge, “Purification and characterization of single-wall carbon nanotubes (swnts) obtained from the gas-phase decomposition of co (hipco process),” The Journal of Physical Chemistry B, vol. 105, no. 35, pp. 8297–8301, 2001. [42] J. Zimmerman, R. Bradley, C. Huffman, R. Hauge, and J. Margrave, “Gas-phase purification of single-wall carbon nanotubes,” Chemistry of Materials, vol. 12, no. 5, pp. 1361–1366, 2000. [43] J. Fagan, M. Becker, J. Chun, and E. Hobbie, “Length fractionation of carbon nanotubes using centrifugation,” Advanced Materials, vol. 20, no. 9, pp. 1609–1613, 2008. [44] Y. Feng, Y. Miyata, K. Matsuishi, and H. Kataura, “High-efficiency separation of single-wall carbon nanotubes by self-generated density gradient ultracentrifugation,” The Journal of Physical Chemistry C, vol. 115, pp. 1752–1756, 2011. [45] Y. L. Yang, L. M. Xie, Z. Chen, M. H. Liu, T. Zhu, and Z. F. Liu, “Purification and length separation of single-walled carbon nanotubes using chromatographic method,” Synthetic Metals, vol. 155, no. 3, pp. 455–460, 2005. [46] J. Chun, J. Fagan, E. Hobbie, and B. Bauer, “Size separation of single-wall carbon nanotubes by flow-field flow fractionation,” Analytical Chemistry, vol. 80, no. 7, pp. 2514–2523, 2008. [47] A. Yu, E. Bekyarova, M. Itkis, D. Fakhrutdinov, R. Webster, and R. Haddon, “Application of centrifugation to the large-scale purification of electric arc-produced single-walled carbon nanotubes,” Journal of the American Chemical Society, vol. 128, no. 30, pp. 9902–9908, 2006. [48] G. Batchelor, “Slender-body theory for particles of arbitrary cross-section in stokes flow,” Journal of Fluid Mechanics, vol. 44, no. 03, pp. 419–440, 1970. [49] X. Wang, J. Yang, Y. Huang, J. Vykoukal, F. Becker, and P. Gascoyne, “Cell separation by dielectrophoretic field-flow-fractionation,” Analytical Chemistry, vol. 72, no. 4, pp. 832–839, 2000. [50] E. Snow, P. Campbell, M. Ancona, and J. Novak, “High-mobility carbon-nanotube thin-film transistors on a polymeric substrate,” Applied Physics Letters, vol. 86, p. 033105, 2005. 89  [51] C. Kocabas, S. Hur, A. Gaur, M. Meitl, M. Shim, and J. Rogers, “Guided growth of large scale, horizontally aligned arrays of single walled carbon nanotubes and their use in thin film transistors,” Small, vol. 1, no. 11, pp. 1110–1116, 2005. [52] S. Tawfick, K. O’Brien, and A. Hart, “Flexible high-conductivity carbon-nanotube interconnects made by rolling and printing,” Small, vol. 5, no. 21, pp. 2467–2473, 2009. [53] Y. Yan, M. Chan Park, and Q. Zhang, “Advances in carbon nanotube assembly,” Small, vol. 3, no. 1, pp. 24–42, 2007. [54] P. Poulin, B. Vigolo, and P. Launois, “Films and fibers of oriented single wall nanotubes,” Carbon, vol. 40, no. 10, pp. 1741–1749, 2002. [55] L. Dong, V. Chirayos, J. Bush, J. Jiao, V. Dubin, R. Chebian, Y. Ono, J. Conley Jr, and B. Ulrich, “Floating-potential dielectrophoresis-controlled fabrication of single-carbon-nanotube transistors and their electrical properties,” The Journal of Physical Chemistry B, vol. 109, no. 27, pp. 13148–13153, 2005. [56] D. Jain, N. Rouhi, C. Rutherglen, C. Densmore, S. Doorn, and P. Burke, “Effect of source, surfactant, and deposition process on electronic properties of nanotube arrays,” Journal of Nanomaterials, vol. 2011, p. 7, 2011. [57] D. Walters, M. Casavant, X. Qin, C. Huffman, P. Boul, L. Ericson, E. Haroz, M. O’Connell, K. Smith, and D. Colbert, “In-plane-aligned membranes of carbon nanotubes,” Chemical Physics Letters, vol. 338, no. 1, pp. 14–20, 2001. [58] M. Correa-Duarte, M. Grzelczak, V. Salgueirio-Maceira, M. Giersig, L. Liz-Marzn, M. Farle, K. Sierazdki, and R. Diaz, “Alignment of carbon nanotubes under low magnetic fields through attachment of magnetic nanoparticles,” The Journal of Physical Chemistry B, vol. 109, no. 41, pp. 19060–19063, 2005. [59] R. Sharma and M. Strano, “Centerline placement and alignment of anisotropic nanotubes in high aspect ratio cylindrical droplets of nanometer diameter,” Advanced Materials, vol. 21, no. 1, pp. 60–65, 2009. [60] R. Sharma, C. Lee, J. Choi, K. Chen, and M. Strano, “Nanometer positioning, parallel alignment, and placement of single anisotropic 90  nanoparticles using hydrodynamic forces in cylindrical droplets,” Nano Letters, vol. 7, no. 9, pp. 2693–2700, 2007. [61] J. Lagerwall and G. Scalia, “Carbon nanotubes in liquid crystals,” J. Mater. Chem., vol. 18, no. 25, pp. 2890–2898, 2008. [62] W. Song, I. Kinloch, and A. Windle, “Nematic liquid crystallinity of multiwall carbon nanotubes,” Science, vol. 302, no. 5649, p. 1363, 2003. [63] N. Puech, C. Blanc, E. Grelet, C. Zamora-Ledezma, M. Maugey, C. Zakri, E. Anglaret, and P. Poulin, “Highly ordered carbon nanotube nematic liquid crystals,” The Journal of Physical Chemistry C, vol. 115, no. 8, pp. 3272–3278, 2011. [64] S. Zhang, I. Kinloch, and A. Windle, “Mesogenicity drives fractionation in lyotropic aqueous suspensions of multiwall carbon nanotubes,” Nano Letters, vol. 6, no. 3, pp. 568–572, 2006. [65] Q. W. Li, Y. T. Zhu, I. A. Kinloch, and A. H. Windle, “Self-organization of carbon nanotubes in evaporating droplets,” Journal of Physical Chemistry B, vol. 110, no. 28, pp. 13926–13930, 2006. [66] R. D. Deegan, O. Bakajin, T. F. Dupont, G. Huber, S. R. Nagel, and T. A. Witten, “Capillary flow as the cause of ring stains from dried liquid drops,” Nature, vol. 389, no. 6653, pp. 827–829, 1997. [67] R. D. Deegan, O. Bakajin, T. F. Dupont, G. Huber, S. R. Nagel, and T. A. Witten, “Contact line deposits in an evaporating drop,” Physical Review E, vol. 62, no. 1, pp. 756–765, 2000. [68] W. Song and A. Windle, “Isotropic-nematic phase transition of dispersions of multiwall carbon nanotubes,” Macromolecules, vol. 38, no. 14, pp. 6181–6188, 2005. [69] M. Bravo-Sanchez, T. Simmons, and M. Vidal, “Liquid crystal behavior of single wall carbon nanotubes,” Carbon, vol. 48, no. 12, pp. 3531–3542, 2010. [70] P. de Gennes and J. Prost, The physics of liquid crystals, vol. 268. Clarendon press Oxford, 1993. [71] L. Onsager, “The effects of shape on the interaction of colloidal particles,” Annals of the New York Academy of Sciences, vol. 51, no. 4, pp. 627–659, 1949. 91  [72] D. Frenkel, “Perspective on the effect of shape on the interaction of colloidal particles?,” Theoretical Chemistry Accounts: Theory, Computation, and Modeling (Theoretica Chimica Acta), vol. 103, no. 3, pp. 212–213, 2000. [73] C. Liu and M. Muthukumar, “Annihilation kinetics of liquid crystal defects,” The Journal of chemical physics, vol. 106, p. 7822, 1997. [74] A. Somoza, C. Sagui, and C. Roland, “Liquid-crystal phases of capped carbon nanotubes,” Physical Review B, vol. 63, no. 8, p. 081403, 2001. [75] C. Furtado, U. Kim, H. Gutierrez, L. Pan, E. Dickey, and P. Eklund, “Debundling and dissolution of single-walled carbon nanotubes in amide solvents,” Journal of the American Chemical Society, vol. 126, no. 19, pp. 6095–6105, 2004. [76] F. Hennrich, R. Krupke, K. Arnold, J. A. R. Stutz, S. Lebedkin, T. Koch, T. Schimmel, and M. M. Kappes, “The mechanism of cavitation-induced scission of single-walled carbon nanotubes,” Journal of Physical Chemistry B, vol. 111, no. 8, pp. 1932–1937, 2007. [77] M. Kuijpers, P. Iedema, M. Kemmere, and J. Keurentjes, “The mechanism of cavitation-induced polymer scission; experimental and computational verification,” Polymer, vol. 45, no. 19, pp. 6461–6467, 2004. [78] B. Derby, “Inkjet printing of functional and structural materials: Fluid property requirements, feature stability, and resolution,” Annual Review of Materials Research, vol. 40, pp. 395–414, 2010. [79] A. van der Bos, “Air entrapment and drop formation in piezo inkjet printing,” phd thesis, University of Twente, 2011. [80] L. Y. Barash, T. P. Bigioni, V. M. Vinokur, and L. N. Shchur, “Evaporation and fluid dynamics of a sessile drop of capillary size,” Physical Review E, vol. 79, no. 4, p. 046301, 2009. [81] H. Hu and R. G. Larson, “Evaporation of a sessile droplet on a substrate,” Journal of Physical Chemistry B, vol. 106, no. 6, pp. 1334–1344, 2002. [82] G. I. Association, “climate/humidity table,” 2011. [83] I. MicroFab Technologies, “Microfab technote 99-02 fluid properties effects on ink-jet device performance,” tech. rep., MicroFab Technologies, Inc., 1999. 92  [84] J. de Jong, G. de Bruin, H. Reinten, M. van den Berg, H. Wijshoff, M. Versluis, and D. Lohse, “Air entrapment in piezo-driven inkjet printheads,” The Journal of the Acoustical Society of America, vol. 120, p. 1257, 2006. [85] P. De Gennes, F. Brochard-Wyart, and D. Quere, Capillarity and wetting phenomena: drops, bubbles, pearls, waves. Springer Verlag, 2004. [86] E. Tekin, P. Smith, and U. Schubert, “Inkjet printing as a deposition and patterning tool for polymers and inorganic particles,” Soft Matter, vol. 4, no. 4, pp. 703–713, 2008. [87] H. Wijshoff, “The dynamics of the piezo inkjet printhead operation,” Physics reports, vol. 491, no. 4-5, pp. 77–177, 2010. [88] I. MicroFab Technologies, “Satellites occurrence and approaches to eliminate them,” tech. rep., MicroFab Technologies, Inc., 2007. [89] L. Wang, S. Beyer, Q. Cronk, and K. Walus, “Delivering high-resolution landmarks using inkjet micropatterning for spatial monitoring of leaf expansion,” Plant Methods, vol. 7, no. 1, p. 1, 2011. [90] J. Carley and P. Kitze, “Corona discharge treatment of polymeric films, ii: Chemical studies,” Polymer Engineering & Science, vol. 20, no. 5, pp. 330–338, 1980. [91] S. Yang and M. Gupta, “Surface modification of polyethyleneterephthalate by an atmospheric-pressure plasma source,” Surface and Coatings Technology, vol. 187, no. 2-3, pp. 172–176, 2004. [92] B. Leclercq, M. Sotton, A. Baszkin, and L. Ter-Minassian-Saraga, “Surface modification of corona treated poly (ethylene terephthalate) film: adsorption and wettability studies,” Polymer, vol. 18, no. 7, pp. 675–680, 1977. [93] M. Meitl, Z. Zhu, V. Kumar, K. Lee, X. Feng, Y. Huang, I. Adesida, R. Nuzzo, and J. Rogers, “Transfer printing by kinetic control of adhesion to an elastomeric stamp,” Nature Materials, vol. 5, no. 1, pp. 33–38, 2006. [94] P. J. Collings and J. S. Patel, Handbook of liquid crystal research. New York: Oxford University Press, 1997. [95] H. Hu and R. Larson, “Analysis of the microfluid flow in an evaporating sessile droplet,” Langmuir, vol. 21, no. 9, pp. 3963–3971, 2005. 93  [96] H. Hu and R. G. Larson, “Analysis of the effects of marangoni stresses on the microflow in an evaporating sessile droplet,” Langmuir, vol. 21, no. 9, pp. 3972–3980, 2005. [97] I. Chuang, B. Yurke, A. Pargellis, and N. Turok, “Coarsening dynamics in uniaxial nematic liquid crystals,” Physical Review E, vol. 47, no. 5, p. 3343, 1993. [98] A. J. Petsi and V. N. Burganos, “Potential flow inside an evaporating cylindrical line,” Physical Review E, vol. 72, no. 4, p. 047301, 2005. [99] A. H. Windle, S. J. Zhang, Q. W. Li, and I. A. Kinloch, “Ordering in a droplet of an aqueous suspension of single-wall carbon nanotubes on a solid substrate,” Langmuir, vol. 26, no. 3, pp. 2107–2112, 2010. [100] L. Huang, X. Cui, G. Dukovic, and S. OBrien, “Self-organizing high-density single-walled carbon nanotube arrays from surfactant suspensions,” Nanotechnology, vol. 15, p. 1450, 2004. [101] A. Matsuyama and T. Kato, “Theory of surface ordering on solutions of rigid-rod-like molecules,” Macromolecules, vol. 28, no. 1, pp. 131–135, 1995. [102] B. Tjiptomargo, A. K. Sen, L. Mederos, and D. E. Sullivan, “The liquid-vapor interface of nematic liquid-crystals,” Molecular Physics, vol. 67, no. 3, pp. 601–614, 1989. [103] H. Kasten and G. Strobl, “Nematic wetting at the free surface of 4 cyano 4 n alkyl biphenyls,” The Journal of chemical physics, vol. 103, p. 6768, 1995. [104] R. Lucht, P. Marczuk, C. Bahr, and G. Findenegg, “X-ray reflectivity study of smectic wetting and prewetting at the free surface of isotropic liquid crystals,” Physical Review E, vol. 63, no. 4, p. 041704, 2001. [105] S. Lin and D. Blankschtein, “Role of the bile salt surfactant sodium cholate in enhancing the aqueous dispersion stability of single-walled carbon nanotubes: A molecular dynamics simulation study,” The Journal of Physical Chemistry B, vol. 114, no. 47, pp. 15616–15625, 2010. [106] F. Zerbetto, M. Calvaresi, and M. Dallavalle, “Wrapping nanotubles with micelles, hemimicelles, and cylindrical micelles,” Small, vol. 5, no. 19, pp. 2191–2198, 2009.  94  [107] A. Striolo and N. R. Tummala, “Sds surfactants on carbon nanotubes: Aggregate morphology,” Acs Nano, vol. 3, no. 3, pp. 595–602, 2009. [108] D. Head, “Volume-controlled buckling of thin elastic shells: application to crusts formed on evaporating partially wetted droplets,” Journal of Physics: Condensed Matter, vol. 18, p. L485, 2006. [109] R. Style and S. Peppin, “Crust formation in drying colloidal suspensions,” Proceedings of the Royal Society A: Mathematical, Physical and Engineering Science, vol. 467, no. 2125, p. 174, 2011. [110] R. Bhardwaj, X. Fang, and D. Attinger, “Pattern formation during the evaporation of a colloidal nanoliter drop: a numerical and experimental study,” New Journal of Physics, vol. 11, p. 075020, 2009. [111] R. Duggal, F. Hussain, and M. Pasquali, “Self-assembly of single-walled carbon nanotubes into a sheet by drop drying,” Advanced Materials, vol. 18, no. 1, pp. 29–34, 2006. [112] J. Cao, Q. Wang, and H. Dai, “Electromechanical properties of metallic, quasimetallic, and semiconducting carbon nanotubes under stretching,” Physical Review Letters, vol. 90, no. 15, p. 157601, 2003. [113] P. Dharap, Z. Li, S. Nagarajaiah, and E. Barrera, “Nanotube film based on single-wall carbon nanotubes for strain sensing,” Nanotechnology, vol. 15, p. 379, 2004. [114] T. Yamada, Y. Hayamizu, Y. Yamamoto, Y. Yomogida, A. Izadi-Najafabadi, D. Futaba, and K. Hata, “A stretchable carbon nanotube strain sensor for human-motion detection,” Nature nanotechnology, vol. 6, no. 5, pp. 296–301, 2011. [115] C. Hsu, R. Chao, C. Liu, and S. Liang, “Evaluation of the gauge factor for single-walled carbon nanonets on the flexible plastic substrates by nano-transfer-printing,” Journal of Micromechanics and Microengineering, vol. 21, p. 075012, 2011. [116] K. Loh, J. Lynch, and N. Kotov, “Conformable single-walled carbon nanotube thin film strain sensors for structural monitoring,” in The 5th International Workshop on Structural Health Monitoring, vol. 14, 2005. [117] J. Li, T. Unander, A. Cabezas, B. Shao, Z. Liu, Y. Feng, E. Forsberg, Z. Zhang, I. Jgi, and X. Gao, “Ink-jet printed thin-film transistors with 95  carbon nanotube channels shaped in long strips,” Journal of applied physics, vol. 109, p. 084915, 2011.  96  Appendix A  MATLAB Script The following MATLAB script is used to determine the spatially distributed orientations of CNTs from an SEM image. The script finds the brightest areas in the image, which typically correspond to the edges of cracks in the aligned film. It then isolates each discovered edge as a distinct object, fits a linear curve to each object and calculates the angle with respect to the x axis. The user must input a greyscale image, and may alter the brightness threshhold depending on the initial contrast of the image. The script outputs several images at varying stages of the calculation for convenient investigation of the quality of the brightness threshhold, and arrays containing the mid point coordinates (xVal and yVal) and the associated contact angles (Theta) for each best fit lines.  %CNT image processing program % %This program takes an SEM image of highly aligned carbon % nanotubes and calculates the spatially dependent angle %of the CNTs with respect to the x-axis close all clear all InStr = input(’Input file name with extension: ’, ’s’); 97  I1 = imread(InStr); %The brightness threshold when converting to binary brightthresh = 188; %Remove the banner at the bottom, setting it to a value %that will be easily ignored when turning into binary [y x]=size(I1); for i=1:(x) for j=1:(y) if (I1(j,i))>=250 I1(j,i)=0; end end end I1 = adapthisteq(I1); I1 = medfilt2(I1, [3 3]); figure, imshow(I1)  %Make the image binary using a threshold meant to isolate %bright line edges I2=I1; for i=1:(x) for j=1:(y) if ((I1(j,i))>brightthresh) I2(j,i)=255; else I2(j,i)=0; end 98  end end I2 = medfilt2(I2, [7 7]); I2 = bwmorph(I2,’fill’); I2 = bwmorph(I2,’thin’, inf); I2 = bwmorph(I2,’clean’); figure, imshow(I2) figure, imshow(I1) cc=bwconncomp(I2); labeled = labelmatrix(cc); x=1:2; count=0; hold on for t=1:cc.NumObjects [r c]=find(labeled==t); [row col]=size(r); if row>15 count=count+1; %best fit a linear curve x=min(c):max(c); [row2 col2] = size(x); if col2 > 1 p=polyfit(c,r,1); FofX=p(1).*x+p(2); %calculate the angle the line makes with the x axis %y|  \  % |__(\___x theta(count)=(180/pi).*(atan((FofX(2)-FofX(1))/(x(2)-x(1)))); 99  if theta(count)<1 theta(count)=180+theta(count); end xVal(count)=median(x); yVal(count)=median(FofX); plot(x,FofX, ’linewidth’, 3); h = text(xVal(count), yVal(count), num2str(theta(count), 3)); set(h,’BackgroundColor’,[1 1 .6]); end end end hold off deviation=std(theta); mean=mean(theta); count2=1; for i=1:count if (theta(i) >= (mean-3*deviation)) ... && (theta(i) <= (3*deviation+mean)) theta2(count2)=theta(i); count2=count2+1; end end figure, hist(theta2, count2/4);  100  Appendix B  Common Particles in SEM SEM images of the only two types of non-carbonaceous particles found in CNT suspensions: SC agglomerates and metallic nanoparticles. Both particles appear bright compared to the CNTs.  Figure B.1: An example of SC agglomerates and adsorbed metal particles. (a) The CNTs in this sample were highly contaminated with metal catalyst particles (36 wt%). (b) This sample was not rinsed with ethanol and the SC residue is clearly visible.  101  


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items