<%BANNER%>

Charge Induced Actuation in Carbon Nanotubes and Resistance Changes in Carbon Nanotube Networks


PAGE 1

CHARGE-INDUCED ACTUATION IN CA RBON NANOTUBES AND RESISTANCE CHANGES IN CARBON NANOTUBE NETWORKS By JENNIFER ANN SIPPEL-OAKLEY A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2005

PAGE 2

Copyright 2005 by Jennifer Ann Sippel-Oakley

PAGE 3

To my parents, Richard and Cynthia Sippel

PAGE 4

iv ACKNOWLEDGMENTS I would like to extend my deepest gratitude to my advisor, Dr. Andrew Rinzler. He has an amazing amount of enthusiasm for the pursuit of knowledge in the field of carbon nanotubes. He was always encouraging th roughout the course of a frustrating and difficult experiment. As a teacher, he was patient and thorough. For these things I am extremely grateful. His support and encour agement are the reason I am receiving this degree. If I could make the same choice over again I would not hesitate to pick him as my advisor. I would also particularly like to thank former group members Dr. Zhihong Chen and Dr. Amol Patil for close collaboration a nd friendship during our graduate careers. Addtionaly I am grateful to current an d former group members Dr. Hidenori Tashiro, Jacob Alldredge, Zhuangchun Wu, Lex Kemper Daniel Ranken, Daniel Barrow and Jonathon Logan. I owe thanks to Dr. Art Hebard for use of laboratory equipment and also for always being a friendly and helpful resource. Thanks are due to Hung-Ta Wang, Byoung Sam Kang, Dr. Fan Ren and Dr. Steve Pearton fo r collaboration on the nanotube hydrogen sensor. Many thanks go to my parents, Rick a nd Cindy Sippel, who have always supported me in all aspects of my life. I am eternally grateful for that an d for their unconditional love.

PAGE 5

v Finally I would like to thank my husband, Ga rrett Oakley. He is a fellow scientist and a great dancer. His love and suppor t have guided and uplifted me during my graduate career.

PAGE 6

vi TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES...........................................................................................................viii LIST OF FIGURES...........................................................................................................ix ABSTRACT......................................................................................................................x ii CHAPTER 1 INTRODUCTION...................................................................................................1 2 CARBON NANOTUBE BACKGROUND.............................................................3 2.1 Growth Methods of Carbon Nanotubes.............................................................3 2.2 Carbon Nanotube Structure................................................................................4 3 MOTIVATION AND THEORY FOR INDIVIDUAL CARBON NANOTUBE ACTUATION........................................................................................................11 3.1 Electromechanical Actuation...........................................................................11 3.2 Macro-Scale Carbon Nanotube Actuators.......................................................11 3.3 Bond Length Changes in Intercalated Graphite...............................................13 3.4 Theoretical Work for Bond Length Changes in Carbon Nanotubes................17 4 PREPARATION OF AND EXPERIMENTAL MEASUREMENTS ON SUSPENDED CARBON NA NOTUBE SAMPLES.............................................22 4.1 Fabrication of Suspended Carbon Nanotube Structures..................................22 4.1.1 Substrate Preparation........................................................................23 4.1.2 Carbon Nanotube Deposition and Growth........................................24 4.1.2.1 Deposition of Laser Ablation Grown Nanotubes from Solution..............................................................................24 4.1.2.2 Chemical Vapor Deposition Grown Nanotubes................27 4.1.3 Electrode Patterning..........................................................................29 5.1.4 Etching..............................................................................................33 5.1.5 Release Procedure.............................................................................34

PAGE 7

vii 4.1.6 Identification of Suspended Nanotubes............................................37 5.1.7 Final Sample Preparations................................................................38 4.2 Experimental Procedure...................................................................................39 5 RESULTS AND DISCUSSION OF CH ARGE INDUCED ACTUATION OF SUSPENDED CARBO N NANOTUBES..............................................................51 5.1 Results of Actuation Measurements................................................................51 5.2 Discussion of Results.......................................................................................58 6 CONDUCTANCE CHANGES IN CARBON NANOTUBES DUE TO HYDROGEN.........................................................................................................66 6.1 Carbon Nanotube Sensor Background.............................................................66 6.2 Carbon Nanotube Sensor Fabrication..............................................................69 6.3 Resistance Changes from Metal Deposition....................................................72 6.4 Conductance Changes due to Hydrogen Exposure..........................................76 6.4.1 Pure SWCNT Samples......................................................................76 6.4.2 SWCNT Samples Coated w ith Sputtered Pd....................................78 6.4.3 SWCNT Samples Coated w ith Thermally Evaporated Pd................80 6.4.4 Thin Pd Film.....................................................................................82 6.4.5 Conclusion........................................................................................82 APPENDIX SPECTROELECTROCHEMICAL STUDY OF CARBON NANOTUBE THIN FILMS.............................................................................................84 A.1 Carbon Nanotube Experiment.........................................................................85 A.2 Experimental Section......................................................................................87 A.3 Results and Discussion....................................................................................88 A.4 Conclusions.....................................................................................................89 LIST OF REFERENCES...................................................................................................91 BIOGRAPHICAL SKETCH.............................................................................................96

PAGE 8

viii LIST OF TABLES Table page 4-1 Resistance before and after processing steps...........................................................36 5-1 Summary of data from sample 1..............................................................................53 5-2 Summary of data for sample 2.................................................................................55 5-3 Summary of all data from sample 3.........................................................................56 6-4 Summary of the actuator data...................................................................................58 6-1 Resistance changes of nanotube samples with sputtered Pd....................................72 6-2 Resistance changes on SWCNT film s with thermally evaporated Pd.....................73 6-3 Conductance changes in sputtered Pd/SWCNT.......................................................80

PAGE 9

ix LIST OF FIGURES Figure page 2-1 Hexagonal graphite lattic e showing the unit vectors a1 and a2..................................7 2-2 Unit cells of graphite..................................................................................................8 3-1 Macro-scale actuator................................................................................................12 3-2 Strain values for bucky paper actuat ors films versus applied potential...................13 3-3 Strain versus charge transfer curve for graphite......................................................15 4-1 Actuation experimental setup...................................................................................23 4-2 Aligned carbon nanotubes from Triton-X solution..................................................26 4-3 Nanotubes on silicon................................................................................................28 4-4 Metal deposition on resists.......................................................................................29 4-5 Scope trace of electrodes..........................................................................................30 4-6 Nanotubes after lithography.....................................................................................32 4-7 Processing steps for suspe nded carbon nanotube samples.......................................35 4-8 Suspended carbon nanotube images.........................................................................37 4-9 Map of the created marks.........................................................................................38 4-10 Cross sectional schematic of chip mounted on AFM puck......................................40 4-11 Free-space drift of a bare tip and a paralyne coated tip in aqueous 1M NaNO3......43 4-12 Identifying mark created in the SEM.......................................................................43 4-13 Force calibration curves against two surfaces..........................................................47 4-14 Illustration of a force calibration curv e against a hard surface and nanotube..........48 4-15 Nanotube under tension............................................................................................49

PAGE 10

x 4-16 Plot of dy versus y forL0=430nm and L/L=0.0001................................................50 5-1 Nanotube sample 1...................................................................................................52 5-2 Sample 1 data...........................................................................................................52 5-3 Nanotube sample 2...................................................................................................54 5-4 Sample 2 data...........................................................................................................54 5-5 Nanotube sample 3...................................................................................................55 5-6 Sample 3 data...........................................................................................................56 5-7 Data of the Z Movement and applied voltage while the AFM was suspended in free space..................................................................................................................57 5-8 Data of Z Movement and applied voltage while the AFM tip was in contact with the bottom of the trench...........................................................................................57 5-9 Hydrated Cland Na+ ions........................................................................................60 5-10 Density of states for a (10,11) SWCNT...................................................................63 6-1 Electrode pattern on micr o-device nanotube sensor................................................69 6-2 Nanotube thin films..................................................................................................70 6-3 SWCNT thin film sensor wired for measurement....................................................71 6-4 Pd coated nanotube films.........................................................................................75 6-5 Micro-sensor device curre nt vs. time measurement.................................................77 6-6 7nm film current vs time measurement...................................................................77 6-7 25nm film current vs time measurements...............................................................78 6-8 Micro-device sensor coated with spu ttered Pd current vs. time measurements.......79 6-9 7nm film with sputtered Pd current vs. time measurements....................................79 6-10 25nm film with sputtered Pd current vs. time measurements..................................80 6-11 7nm film with thermally evaporat ed Pd current vs. time measurement...................81 6-12 Thin Pd film current vs. time measurements............................................................82 A-1 Spectroelectrochemical cell......................................................................................86

PAGE 11

xi A-2 Fermi Level of a semiconducting (12,8) SWCNT..................................................87 A-3 Percent Transmittance vs. Wavelength for SWCNT thin-film at various positive applied potentials......................................................................................................89 A-4 Percent Transmittance vs. Wavelength for SWCNT thin-film at various negative applied potentials......................................................................................................90

PAGE 12

xii Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy CHARGE-INDUCED ACTUATION OF SUSPENDED CARBON NANOTUBES AND RESISTANCE CHANGES IN CA RBON NANOTUBE NETWORKS By Jennifer A. Sippel-Oakley May 2005 Chair: Andrew G. Rinzler Major Department: Physics In 1999 it was demonstrated that macros copic films comprised of single wall carbon nanotubes exhibited dimensional changes with charge injection onto the films. A fundamental mechanism was proposed for this effect related to the dimensional changes observed in graphite inter calation complexes upon charge transfer doping with the intercalant species. The major fr action of this thesis concerns experiments at the single nanotube level designed to test the validity of this mechanism. The metals compatible with our fabrication processes inevitably p-dope the nanotubes resulting in smaller dimensional changes. Additionally, there are co ntact barriers that prevent the injection of electrons onto the nanotubes. Although the proposed mechanism may still be responsible for the results seen in the nanotube films, the effect is too small to be consistently measured in individual nanotubes. The conductivity of a carbon nanotube can be varied by exposure to various chemicals having utility in chemical sensing applications. We use thin films of carbon

PAGE 13

xiii nanotubes to exploit this eff ect. The films are made sens itive to hydrogen by association with palladium metal. Such sensors operate at room temperature with very low power dissipation of ~0.25 mV.

PAGE 14

1 CHAPTER 1 INTRODUCTION Carbon nanotubes are quasi one-dimensional structures with the high mechanical strength of graphite and have the useful attribute of occurri ng in either semiconducting or metallic variants. These qualities have cause d an explosion of research since their discovery in the soot of fullerene production by Ijima1 in 1991. These first nanotubes were multi-walled carbon nanotubes (MWCNTs) consisting of at least two graphene cylinders nested one within the next. This discovery prompted efforts to produce single wa ll carbon nanotubes (SWCNTs) and prompted much theoretical effo rt on the properties of such tubes. The addition of metal catalyst partic les to the carbon starting mate rial was found to be the key and production of SWCNTs was first established by Bethune et al .2 and Ijima and Ichihashi3 in 1993. All the nanotubes used in our work were SWCNTs. In a standard description, a carbon nanotube is a graphene sheet rolled into a seamless cylinder. As the name implies, the nanotubes have a diameter around a nanometer. In general, they have a length of a few microns but there have been recent reports of nanotubes with le ngths of up to centimeters.4 Much of the fascination with nanotubes arises from the fact that they can be either semiconducting or metallic depending on the orientation the hexagonal grap hene lattice relative to the nanotube axis. Coupled with this property is the extremely high Young’s modulus (1 TPa)5 and tensile strength (estimated at 45 Gpa)6,7 of nanotubes. They are also stable at high temperatures and show strong resistance to most chemicals.

PAGE 15

2 Our study addresses two specific effects in carbon nanotubes that could lead to useful applications. One proposed applica tion for carbon nanotubes is electromechanical actuators. In 1999, a carbon nanotube film sh owed dimensional changes with charge injection.8 This behavior was attributed to an effect seen in intercalated graphite compounds whereby the carbon-carbon bond length is modified by charge transfer to graphite’s orbital system. The first part of th is thesis describes efforts to observe dimensional changes due to charge injection in an individual nanotube and determine if the proposed mechanism was really responsible for the effect seen in the nanotube films. Conductance of nanotubes can be manipulat ed by the interaction with other molecular species. This led to the second focus of our study, resistance changes in nanotube films as a way of hydrogen detection. Nanotube films were made sensitive to hydrogen by the addition of palladium particles and H2 levels as low as 10 ppm were detected.

PAGE 16

3 CHAPTER 2 CARBON NANOTUBE BACKGROUND 2.1 Growth Methods of Carbon Nanotubes As the interest in carbon nanotube resear ch grew, so did the need for increased quantities of high-quality nanotubes. Seve ral methods were developed for their production. The first method is arc-discharge or carbon arc. In this setup, two carbon rods are used as electrodes with a small se paration (about 1mm) and a high dc current is passed between them while in a helium atmo sphere. The high currents passing between the carbon electrodes ignite plasma of th e helium gas, and temperatures exceed 3000 C. Carbon is evaporated from the anode and th en condenses to form nanotubes. This method can be used to grow MWNT or SWNT if a metal catalyst (usually a transition metal) is added to the carbon electrode. The high temperatures at which growth occurs ensure high-quality nanotubes with few defects. However, many byproducts (such as fullerenes) and amorphous carbons are produced and must be removed by purification. The second method of nanotube growth is laser ablation (first demonstrated by Thess et al. ).9 This was the first method to pr oduce nanotubes on the scale of several grams per run. Early work in the actuation pr oject and all work with nanotube sensors used SWCNTs grown by this process. With this method, a pulsed laser is used to vaporize a graphite target within a heated tube furnace (temp 1100-1200 C). To achieve SWNT growth metal catalyst material is mixed with the graphite target. A flow of inert gas carries the nanotubes downstream to a c ooled copper collector. This method also requires purification to remove the byproducts generated during growth. Both the arc-

PAGE 17

4 discharge method and laser ablation method form nanotubes that tend to be aggregated in ropes (several nanotubes bundled together by van der Waals forces). Although vigorous sonication of solutions containing these nanot ubes can help break up the bundles, it is difficult to isolate individual nanotubes usi ng material from these growth processes. This leads to the next method of growing nanotubes: chemical vapor deposition (CVD). Most nanotubes used to measure nanotube actuation were grown by CVD. The method involves placing a sample with me tal catalysts into a furnace (growth temperatures between 500-1000 C) and flowing a carbon fe edstock gas (usually a hydrocarbon gas) over the sample. The pres ence of the metal catalysts causes the dissociation of the hydrocarbon feedstock gas. The metal particles become supersaturated with carbon and precipitate na notubes. It has been found by transmission electron microscopy (TEM) that the nanotube diameter co rresponds to the diameter of the catalyst particles.10 Catalyst islands can even be lithogra phically patterned onto silicon substrates for more controlled placem ent of the nanotube growth.11 A fourth growth method (developed recen tly) involves decomposition of carbon monoxide at high pressures (between 1 and 10 atm) and is commonly referred to as HiPCO.12 Catalyst particles are provided by the thermal decomposition of iron pentacarbonyl Fe(CO)5 which produces iron clusters in the gas phase. This process produces tubes with smaller av erage diameters and less vari ation then tubes produced by the first three methods. 2.2 Carbon Nanotube Structure Carbon nanotubes were predicted to be either semiconducting or metallic in 1992 soon after their discovery.13,14,15 However, this was only veri fied experimentally 6 years

PAGE 18

5 later by scanning tunne ling microscopy (STM)16,17. Many techniques (such as STM, TEM and scanning electron microscopy) have ve rified nanotube structure as that of a seamless graphene-like cylinder. STM has pr oved a useful tool to study nanotubes since it can reveal a nanotube’s chirality. This section gives some basic background on the physical and resulting electronic structure of SWCNTs. The treatment here is a summary of that developed by Saito et al. in Chapter 3 of reference 18. Nanotubes are mo st easily described by their chiral vector, Ch. Ch= n a1 + m a2 (2-1) The chiral vector denotes the circumfe rence of the nanotube and connects two crystallographically equivalent sites on the planar graphite sheet. In other words, a nanotube can be envisioned as the cylinder fo rmed from rolling up a sheet of graphene by connecting the tip to the tail of the chiral vector. As shown in Figure 2-1, the vectors a1 and a2 are two non-orthogonal unit vect ors, the combination of which can span any point on the hexagonal graphite lattice and have the following relations a1 a1 = a2 a2 = a2 a1 a2 = a2/3 (2-2) where a=1.44 x 31/2 = aC-C x 31/2. a is the lattice constant of the graphite sheet and aC-C is the carbon-carbon bond distance. In x and y coordina tes, the real-space unit vectors are represented by a1=(31/2a/2, a/2) and a2=(31/2a/2, -a/2). Thus the diameter of the nanotube is given by dt = L/ = ( Ch Ch)1/2/ = 31/2aC-C(m2 + mn + n2)1/2/ (2-3) The translation vector T is orthogonal to Ch and parallel to the na notube axis. It is the shortest repeat distance along th e axis. It is defined to be the unit vector of a nanotube and is represented as

PAGE 19

6 T = t1a1 + t2a2 (t1 and t2 are integers) (2-4) Since T and Ch are orthogonal by definition, we use Ch T = 0 to calculate t1 and t2. (n a1 + m a2) (t1a1 + t2a2) = (2n + m)t1 + (2m + n)t2 = 0 (2-5) t1 = (2m + n)/dR t2 = -(2n + m)/dR where dR is the greatest common di visor of (2m + n) and (2n + m). Introducing d as the greatest common divisor of n and m gives dR as dR = d if n-m is not a multiple of 3d (2-6) = 3d if n-m is a multiple of 3d The unit cell of a nanotube is th e rectangle defined by the vectors T and Ch. The number of hexagons contained in this unit cell is determined by the area of the unit cell (| Ch x T |) divided by the area of a hexagon (| a1 x a2|) and is given in terms of (n,m) by the equation N = 2(m2 + nm + n2)/dR (2-7) Each hexagon contains two carbon atoms, so the total number of carbon atoms in the nanotube unit cell is thus 2N. The construction of the nanotube along with its electrical propert ies is completely determined by the indices (n,m). Nanotubes are grouped into three categories: zigzag, armchair, and chiral. Both zigzag and armchair nanotubes have special symmetry directions and earn their name from the pattern of the chain of carbon atoms around the circumference. Zigzag nanotubes are those wi th one of the indices equal to zero, either (n,0) or (0,m). Armchair nanotubes have iden tical indices, (n,n). They are both achiral, meaning that the structure of the mirror image is identical to that of the original. Chiral nanotubes are those that fall anywhere betw een the two types. The chiral angle represents the tilt of the hexagons with respect to the na notube axis. Figure 2-1 shows see that can be found by taking the dot product of Ch and a1.

PAGE 20

7 cos Ch a1/| Ch|| a1| = (2n + m)/2(n2 + nm + m2)1/2 (2-8) Zigzag nanotubes have = 0 armchair have = 30 and chiral fall anywhere in between that range. Figure 2-1: Hexagonal graphite lattice showing the unit vectors a1 and a2. This image shows the chiral vector, Ch, and translation vector T for a (6,3) nanotube As mentioned, the indices (n,m) determin e a nanotubes electrical properties. The electronic structure of carbon nanotubes can be obtained by using the two-dimensional structure of graphite and applying period boundary conditions along the circumferential direction Ch. The unit structure of graphite contai ns two inequivalent carbon sites within the hexagonal lattice called points A and B, Fi gure 2-2A. Real-space lattice vectors were previously defined in x,y coordinates. In reciprocal space, the lattice vectors are b1=(2 /31/2a, 2 a) and b2=(2 /31/2a, -2 a), Figure 2-4B. When the first Brillouin zone is chosen as the shaded region in Figure 2-2B the highest symmetry is obtained for gra phite. In 2-D graphite, the three bonds hybridize in an sp2 configuration. The third 2pz orbital forms covalent bonds. Generally, only the

PAGE 21

8 bands are considered when deriving th e band structure of graphite. Saito et al.18 used a tight binding model that only considers neares t neighbor interactions to A and B, to get the following dispersion relations for the band of 2D graphite E2Dgraph( k )= 2p t ( k )/(1 s ( k )) (2-9) Figure 2-2:Unit cells of graphite. A) The real space unit cell of graph ite is defined by the dotted line and contains point s A and B. The unit vectors a1 and a2 are also shown. B) the Brillouin zone in k space for graphite and the reciprocal lattice vectors b1 and b2. The + signs in Equation (2-9) give the values for the bonding band while, the – signs are for the anitbonding band. The parameter 2p is the orbital energy for the 2p level, s is the overlap integral between the nearest A a nd B atoms, and t is the transfer integral. The function ( k ) has the form ( k ) = {1 + 4cos(31/2kxa/2) cos(kya/2) + 4cos2(kya/2)}1/2 (2-10) To get a simple approximation for the elec tronic structure of graphene the overlap integral s is usually taken to be zero with 2p also being set to zer o. This yields the following expression for the energy dispersion relations. E2Dgraph(kx,ky) = t{1 + 4cos(31/2kxa/2) cos(kya/2) + 4cos2(kya/2)}1/2 (2-11)

PAGE 22

9 Plotting the energy dispersion relations for gr aphite reveal that the bonding and antibonding bands meet at the K point; thus graphite is considered a zero-gap semiconductor (the density of stat es is zero at the Fermi level). Previously the chiral vector Ch and translation vector T for a carbon nanotube were defined. The corresponding reciprocal lattice vectors K1 (in the circumfere ntial direction) and K2 (along the direction of the nanotube axis) are calculated by the expression Ri Ki = 2 ij ( Ri are the lattice vectors in real-space corresponding to Ch and T ) giving Ch K1 = 2 T K1 = 0 (2-11) Ch K2 = 0 T K2 = 2 This leads to K1 = (-t2b1 + t1b2)/N and K2 = (m b1 – n b2)/N (2-12) Because of the translational symmetry of T wave vectors in the direction of K2 are continuous for an infinitely long nanotube. In the circumferential direction there are N wave vectors K1 ( = 0,1. . N-1), which means N disc rete k vectors. When the allowed k vectors in the circumferential direction in clude the K point (where the valence band and conduction band meet), the nanotube will be metal lic. This occurs for the condition (n-m)=3i (where i is an integer). Ot herwise the nanotube will be semiconducting. The armchair tubes (n,n) are all metals with a finite density of states at the Fermi level. Chiral tubes with (n-m)=3i technicall y have a tiny gap that opens up at the Fermi level due to tube curvature effects. However th is gap is so small that at room temperature these tubes can be considered metallic. The band gaps of semiconducting nanotubes are inversely proportional to the nanotube diamet er. The size of the band gap depends on how close the allowed k values are the to spec ial K point. The larger the diameter of a nanotube the more k values there are. Thus the spacing between the k points decreases

PAGE 23

10 and the values get closer to the K point resulting in a smaller band gap. The onedimensional nature of the nanotubes leads to peaks in the density of states known as van Hove singularities. Semiconducting tubes that are slightly doped (a s is often the case) can have their Fermi levels shifted into the first Van Hove singular ity. Because of this these doped semiconducting nanotubes are of ten more conductive than the intrinsic metallic nanotubes at room temperature b ecause of the large nu mber of carriers.

PAGE 24

11 CHAPTER 3 MOTIVATION AND THEORY FOR INDIVIDUAL CARBON NANOTUBE ACTUATION 3.1 Electromechanical Actuation Electromechanical actuators convert electrical energy in to mechanical work. The fine motor movement possible with these device s is used in various fields such as laser tuning, vibration cancellation, adaptive optics and micromanipulation. Piezoelectric actuators are the most commonly used devices of this type (one example is the scanner used in atomic force microscopes). Many actuators can only be operated in a limited temperature range or have ot her physical limitation (piezoel ectric actuators must be operated below their Curie temperature and suffer from hysteresis). In 1999, Baughman et al.8 reported an experiment in which they attempted to use carbon nanotubes (in bulk form) as actuators. The chemical robustne ss and small dimensions of SWCNTs made them seem like a promising addition to the repe rtoire of actuator materials. The intriguing results of these macro-scale actuators led to our study aiming to observe actuation in individual SWCNTs.. 3.2 Macro-Scale Carbon Nanotube Actuators Baughman et al.8 used a macro-scale composite of SWCNTs know as “bucky paper” as a demonstration of their actuator. Bucky paper is formed by vacuum filtering a suspension of carbon nanotubes ont o Teflon filter paper. This forms a sheet of tangled nanotubes, which can be peeled off the filter as a freestanding film. One version of the actuator was fashioned by applying equal sized st rips of bucky paper to both sides of a

PAGE 25

12 slightly larger piece of doubl e-sided tape as shown in Figure 3-1. Electrodes were attached to both strips in order to inject charge and the bucky paper-tape composite was immersed in an electrolyte solution. The ions in the electrolyte served to screen the charges being injected onto the nanotube s by forming a double layer. Without it, Coulombic repulsion would prevent the inje ction of significant amounts of charge. Figure 3-1: Macro-scale actuator. Voltages of opposite polarity were applied to the two sheets and motion of the trilayer structure was observed. It appeared th at the strip on the ne gative terminal would elongate and conversely the strip on the posit ive terminal would contract causing an overall wagging motion as the polarity was f lipped back and forth. To actually measure the stress and strain of th e bucky paper another setup was employed. The bottom of a freestanding bucky paper was fixed to the bot tom of the apparatus while the top was attached to a cantilever. Changes in the length of the nanotube f ilm would deflect the cantilever and an optical sensor detect ed displacements of the cantilever. The strain values calculated from th e experimental data from this second experiment setup are shown in Figure 3-2. The behavior shown in Figure 3-2 was interpreted as arising from a change in the la ttice constant of the nanotube atomic lattice

PAGE 26

13 with charge injection into the nanotube pi-orb ital system. This is an effect known to occur in chemically intercalated graphite (described in section 3.3), and Baughman et al.8 attributed their observations to it. Figure 3-2: Strain values for bucky paper actu ators films versus appl ied potential in 1M NaCl electrolyte solution. The applie d potential was symmetric about V=0. 3.3 Bond Length Changes in Intercalated Graphite Graphite intercalations compounds (GICs) are formed by the spontaneous chargetransfer driven insertion of a chemical species, called the in tercalant, between layers of the graphite matrix. They are classified by a stage index, n, which re presents the number of graphite layers in between intercalant la yers. Thus a large stage number would denote a dilute GIC and a stage number of 1 would represent one layer of intercalant for each layer of graphite. The interclants are labele d as acceptor (donor) com pounds if they take

PAGE 27

14 (give) electrons from the graphite lattice and this process can vary the free carrier concentration greatly. This in turn can va ry numerous electrical and physical properties of the host graphite. Intercalation compounds undergo not only an expansion along the c-axis due to the additional volume of the intercalants but more interestingly (in the present context) they also show in plane dimensional changes. In 1969 Nixon and Parry19 showed, through xray diffraction, that the C-C bond length expanded due to the in tercalation of graphite by the donor species potassium. Following that, measurements on other graphite intercalation compounds were performed using such dopants as Na, Li, Ba, MnCl2, and AsF5 and a strain versus estimated charge transfer curve was generated, Figure 34).20,21,22,23,24 Doping by donor compounds, i.e. additional electrons in the carbon lattice, resulted in lengthening of the C-C bond, while doping by acceptor compounds resulted in bond shortening (although smaller in magnitude) That the bonds would shorten at all is counterintuitive. Simple Coulombic consider ations from the additional charges imply an expansion regardless of the charge sign (since the Coulomb for ce is quadratic). Coulombic repulsion of the additional charges w ould try to expand the host in an effort to separate the charges. Howe ver, a quantum mechanical effect shifts the minimum graphite lattice constant from the point of zero charge transfer to the hole doped side. After the compilation of the charge transfer vs. strain curve some models were proposed to explain this behavior.25,26,27,28 It is these changes in the carbon-carbon bond length as a result of charge transfer that is the proposed mechanism for carbon nanotube actuation. In 1981, Pietronero and Strssler first atte mpted to model the effect of additional charge in a single graphite layer us ing tight binding energy calculations.25 They noted

PAGE 28

15 that two effects would occur from this addi tional charge per carbon atom. First, this additional charge would cause a change in the occupancy of the states and thus a modification of the bond order (in molecular or bital theory the bond order is defined as the number of bonding electrons less the numbe r of anti-bonding electrons all divided by 2). Second, since the atomic potential has chan ged, a modification of all the tight binding matrix elements will arise. Additional electrons in the anti-bonding orbitals weaken the C-C bond and causing bond lengthening. Also incl uded in their work is the change in total energy due to Coulomb re pulsion of the charged atoms (the dominating effect). Figure 3-3: Strain versus char ge transfer curve for graphite This work was soon followed by Kertesz26 and Kertesz et al27. who also used tight binding theory to ascribe the bi polar nature of the strain of graphite intercalations

PAGE 29

16 compounds to second neighbor an ti-bonding interactions. The inclusion of only nearest neighbor contributions would re sult in expansion for both elect ron and hole injection so this interaction was needed to explain the observed results. The next significant work was done by Chan et al.28 who used density functional theory to focus on the changes induced in gr aphite solely as a result of the charge exchange with the intercalants. For the higher stage GIC’s they assumed that the bounding layers had the same in pl ane lattice parameter as thos e next to the intercalants. In other words, the strain is shared uniformly among all th e graphite layers. This was noted in experiment19 and supported by energy consideratio ns. It would take far greater energy to assume different la ttice parameters for the bounding and interior layers than would be required to make all the layers assume the same parameter. Using this approach, Chan et al. also noticed a previously unr ecognized mechanism that could not be obtained from rigid-band models. Not onl y does the donated charge reside in the orbitals but their theory also sugge sts a depletion of charge from the bonds. The external potentials from donor species appear to cause electron transfer from the to bonds further weakening the C-C bond and increasing the bond length. All previous experimental work estimated the amount of charge transfer from the intercalants to the graphite and the theoretica l calculations were ba sed upon this estimate. In most instances, it was assumed that the ch arge transfer would be nearly complete. This work performed an experiment in wh ich the intercalation compounds were placed on a charge transfer scale and thus the actua l charge transfer could be measured and correlated with the associated bond length chan ges. Although their experimental setup dictated that they use neutron diffraction, which is lower in resolution than x-ray

PAGE 30

17 diffraction, the experiment showed that the previous assumptions about the amount of charge transfer were within reason. 3.4 Theoretical Work for Bond Length Changes in Carbon Nanotubes Since carbon nanotubes have the structure of a graphene sheet rolled up into a tube it was only a matter of time before theoretical calculations were perf ormed to predict the strain from charge transfer to nanotubes. Just as a SWCNT’s electrical properties differ from that of graphite as a function of ch irality, the predicted dimensional changes are predicted to differ from graphite as a function of chirality. In genera l, it appears that the exception is the armchair nanotubes (n,n). For this flavor of nanotube one of the allowed k values always passes through th e K point where the occupied and unoccupied bands meet, just like in graphite. Since the el ectronic properties are si milar to graphite so should be the dimensional changes due to tran sferred charge. Seve ral groups differed on the predicted behaviors. The following pa ragraphs summarize th ese predictions and highlight some of the salient differences in the predicted response of SWCNTs compared to graphite. Gartstein et al.29 studied bond length changes in nanotubes from the modulation of electron hopping integrals. As tight binding models from graphite and carbon nanotubes are charge conjugated symmetric in the n earest neighbor hopping approximation they included second order hopping to account fo r the asymmetric actuation response going from positive to negative charge injection. They predicted oscillating dimensional changes as a function of chirality, with most nanotubes showing typical bond length expansion upon electron in jection but some small diameter SWCNTs showing contraction for electron injection. The magn itude of the dimensional change (for the

PAGE 31

18 same amount of charge transfer) also varies wi th the chirality. Their results predict some semiconducting nanotubes will show strains 3-4 times larger than those in graphite and other tubes will show substa ntially decreased responses. A subsequent paper by Gartstein et al .30 extended their previous work on the effect of charge injection on the ge ometries of nanotubes. Thos e results are surprising since they are markedly different than the response in graphite. They predict that a (10,0) and a (16,0) will show contraction upon both hole a nd electron injection a nd conversely, that a (11,0) and a (17,0) tube will e xpand for both signs of doping. Also at a certain charge injection level the response appears to re verse for the (16,0) and (17,0) nanotubes. It should be noted that Coulombic effects, which are quadratic in their effect and certainly play the dominant role at large values of charge tran sfer, were not taken into consideration in thes e calculations. Verissimo-Alves et al.31 also examined charge transfer induced dimensional changes for a limited number of nanotubes. They carried out ab initio calculations with the density functional theory and pseudopotential frameworks using a numerical-atomicorbitals basis set. They used those calculatio ns to generate strain versus charge transfer curves for graphite, metalli c (5,5) and (12,0) nanotubes, and a semiconducting (11,0) nanotube. The predictions for the (12,0) tube and graphite appear very similar but the (5,5) nanotube differs in that it appears to show a signifi cantly increased elongation for negative charge injection but diminished cont raction for electron withdrawal. The results for the (11,0) nanotube predict expansion for both signs of charge injection. Strain calculations (at charge tr ansfer of 0.01 electrons/C atom) for several zigzag semiconducting nanotubes with n=3i-1 (i is an integer) predict that SWCNT’s with a

PAGE 32

19 diameter smaller than that of the (23,0) tube (diameter =1.8nm) will expand for both positive and negative charge injection. Larger diameters switch back to the behavior seen in graphite (contraction for hole injection). Their analysis is that this behavior can be explained by the position of the atomic level of the orbital state of carbon, 2p, with respect to the Fermi level. According to tight binding calculations, inclusion of next nearest neighbor effects rais es the energy eigenvalues at the K points above the 2p level. Anything above 2p is anti-bonding and anything below is bonding. Thus, (if the Fermi level is above 2p) adding electrons causes expans ion and withdrawing them causes contraction. If enough elect rons are withdrawn the Fe rmi level will fall below 2p and into the bonding regime. At th is point if further electrons are withdrawn the bond length will expand. For large band gaps, the Fermi level is below 2p. So for small diameter (large band gap) nanotubes, they pred ict the Fermi level will lay below 2p and demonstrate only expansion from charge inject ion. As the diameter increases, and the band gap decreases, and the semiconducting tube s should revert back to graphene-like behavior. In 2002, a work by Sun et al.32 calculated strain versus charge curves for different varieties of achiral nanot ubes (armchair (n,n) and zig zag (n,0)) using a uniform background charge to represent counterions in order to more closely reflect the conditions in the actuation experiments. They used density functional theory and the generalized gradient approximation. Nanotubes were groupe d into four categories: (n,n), (n=3i,0), (n=3i + 1,0) and (n=3i + 2,0). These result s predicted ambipolar behavior for all SWCNTs with the exception of th e extremely small diameter (5,0) nanotube. In that case the response was expansion for both positive and negative charge injection.

PAGE 33

20 The following year (2003) these authors used density functional theory to calculate the effect of charge injection on the geom etries of achiral tubes as a function of diameter.33 Their approach in this case differe ntiated between two different types of bonds, b1, which are those orientated ma inly around the diameter and b2, which are those orientated mainly along the nanotube axis. In armchair nanotubes b1 is orthogonal to the tube ax is (parallel to the diameter) and b2 has components both parallel and orthogo nal to the axis. Conversely, for the zigzag nanotubes b2 is parallel to the nanotube axis while b1 has the mixed components. Unlike graphene, their calculations predicted that the bond length’s b1 and b2 would be different from each other with b1 > b2 in almost all cases. The larger value of b1 is a result of the curvature and the differences in lengths between the two bonds decreases with increasing diameter. The differences between b1 and b2 are smaller for the armchair tubes than the zigzag tubes of the same diam eter and decrease with increasing nanotube diameter. They calculated how these differi ng bonds would change with the addition of charges and then computed th e overall strain for several charged nanotubes at charge injection levels of q= 0.01e/carbon These results show some contradictions to their previous work. One example is that the small diameter (3i + 2, 0) nanotubes will show expansion for both signs of charge injection, something not shown in their earlier publication bu t similar to the predictions of Verissimo-Alves et al.31 However, these results indicate the (3i + 2,0) family will revert to bipolar behavior st arting with diameters greater th an or equal to that of the (14,0) nanotube while Verissimo-Alves et al. predict this behavior wont resume until the nanotube diameter is at least as large as th e that of the(23,0) SWCN T. Another interesting

PAGE 34

21 result of this work is that the predictions for the (3i + 1, 0) series of nanotubes show a larger magnitude of contraction for electr on withdrawal compared to expansion for electron donation. This asymmetry goe s the opposite way in graphite. There are quite a few discrepancies between the different theoretical calculations published in the literatu re. As another example, the resu lts for the (10,0) tube with q=+/0.1 e/carbon in the work Garstein et al.30 predict contraction fo r both signs of charge injection (something not seen in any other theo retical work) while the calculations of Sun et al.32,33 predict a response for the (10,0) tube mo re similar to graphite with contraction for positive charge injection and expansi on for electron injection, also for q=0.01 e/carbon. These papers show th at there is much interest in nanotube dimensional changes caused by charge injection. Our work was started prior to the publication of these theoretical works and was motivated by the buc ky paper actuator experiment. The goal was to observe charge induced actuation in in dividual nanotubes to determine if that was the mechanism responsible for the effect witnessed by Baughman et al.8

PAGE 35

22 CHAPTER 4 PREPARATION OF AND EXPERIMENT AL MEASUREMENTS ON SUSPENDED CARBON NANOTUBE SAMPLES To measure dimensional changes in a SW CNT we devised an experiment that would allow injection of ch arge onto an individual na notube while measuring the resulting small changes in its length. The e xperimental arrangement devised do this is illustrated in Fig. 4-1. Shown there is a nanotube suspended over a micro-machined trench in a silicon substrate where the nanotube is pinned at its ends by metal electrodes.7 Like the bucky paper actuator setup, this experi ment was also preformed in an electrolyte solution. Charge injection ont o the nanotube occurred via a potential applied to the pinning electrodes versus a counter electrode in the electrolyte solution. To measure length changes, the suspended nanotube was pr e-tensioned by a modest force applied at its center with the tip of an AFM cantilever This deflected the center of the nanotube from the top of the trench while simultane ously deflecting the AF M cantilever from its set point equilibrium position. If the nanotube lengthened, its center would deflect further into the trench to be detected by the corres ponding relaxation of th e AFM cantilever. If the nanotube shortened, the tip would in c ontrast be forced upwards increasing the deflection of the cantilever from its set point position. 4.1 Fabrication of Suspended Carbon Nanotube Structures Fabrication of samples consisted of depos iting nanotubes (either out of solution or growing by CVD) onto silicon and lithographi cally patterning electrodes on top of the SWCNTs. The nanotubes were suspended by a wet chemical etch of the underlying silicon. Details are describe d in the following sections.

PAGE 36

23 4.1.1 Substrate Preparation Commercially available p-t ype <100> silicon wafers with a thermal oxide layer were used as the substrates. Wafers were coated with Shipley S 1813 photoresist to avoid silicon dust contamination prior to dicing into 1cm2 pieces used for suspended nanotube sample fabrication. To strip photoresist from the 1 cm2 pieces the chips were rinsed with acetone, bathed in acetone for 10 minutes, ri nsed with acetone again, rinsed with methanol, rinsed with 18 MOhm deionized water (DI) and finally blown dry with N2 gas. Figure 4-1: Actuation experimental setup.

PAGE 37

24 4.1.2 Carbon Nanotube Deposition and Growth Early samples in this work were fabri cated by depositing purified laser ablation grown SWCNT’s out of solution onto a silic on substrate. However, longer nanotubes were required and later samples used SW CNT’s grown by CVD. Both methods are described here. 4.1.2.1 Deposition of Laser Ablation Grown Nanotubes from Solution The first step in this approach is to get the nanotubes on the silicon surface. Walters et al.7, who had also fabricated suspende d nanotube structures, used purified34 pulsed laser vaporization grown nanot ubes that were suspended in N,Ndimethylformamide and deposited this onto si licon chips that had a 100nm thermal oxide layer. To date, there are no known solvents for SWCNTs. Some solvents, such as N,Ndimethylformamide or 1,2 –dichlorobenzene, can form quasi stable solutions which will suspend the nanotubes for short periods of time35 but the density of the nanotubes is very limited. The nanotubes tend to form bundles with each other while in solution. As Brownian motion causes indivi dual nanotubes to encounter each other in liquid, the strong van der Waals interactions between the sidewalls causes the nanotubes to aggregate into bundles. Thus, deposition from these metastable suspensions rarely results in individual nanotubes. Additionally, if the tubes are deposited in a manner that allows the solvent they are suspended in to evapor ate, any impurities within the solvent will be left on the sample. Nanotubes can be suspended in aqueous so lution with the aid of surfactants. Sodium dodecyl sulfate, Triton X-100 a nd most recently sodium dodecylbenzene sulfonate have been used for this purpose.36,37. Originally I used purified34 Tubes@Rice material that was suspended in a 1% Triton X-100 solution and depos ited on bare silicon

PAGE 38

25 surfaces from solution. Following purification, the nanotubes are kept in a slightly basic solution of Triton-X in dionized water. Th is solution was vacuum filtered onto a Teflon membrane, forming a film of nanotubes on the membrane. This film was washed with 1% Triton X-100 in DI water and allowed to dry. A small amount of this film was torn off and immersed in an aqueous 0.5% Triton X-100 mixture. The solution was placed in an ultra-sonicator bath for typically 10 minutes This action broke up the strip of bucky paper into individual and small bundles of nanotubes. The 1 cm2 silicon chips were subjected to a buffered oxide etch (BOE) diluted 1:1 with deionized water to remove the thermal oxi de layer. This left the samples with a hydrophobic H-terminated silicon surface. Thes e chips were subsequently place in a small vial and covered with a few milliliters of the nanotube/Triton-X solution. The addition of a drop of concentrated acid (u sually nitric acid) interfered with the surfactant’s ability to suspend the nanotubes such that after a few minutes the initially homogeneous, particle free solution contai ned flocculated nanotubes. The nanotubes (which are themselves hydrophobic) that lay cl ose to the silicon su rface adhered to the hydrophobic silicon as they exited the suspension. Various techniques in which acid was us ed to remove the nanotubes from the Triton-X suspension were explored. One me thod involved placing a silicon chip face down onto several drops of nanotube/TritonX solution and sliding the chip across a nearby drop of acid. This technique produced depositions in which the nanotubes were aligned in the same directi on, as example of which is s hown in Figure 4-2. A comparison of the alignment direction and the direction of the acid front as it propagated through the nanotube/Triton-X solution showed the two to be the same. Once the acid removed the Triton-X micelle from the nanotube, the hydrophobic SWCNTs aligned along the acid

PAGE 39

26 front so that as much surface area as possibl e of the tubes would be out of the water solution. Figure 4-2: AFM image of aligned ca rbon nanotubes from Triton-X solution Even though these samples would show high concentrations of nanotubes on the silicon surface, the density of suspended nanot ubes after etching was low. As mentioned, nanotubes in solution tend to form bundles. This was certainly the case when the acid was added to the surfactant solution. After being deposited on the surface the laser grown nanotubes were grouped in bundles with very few (if any) individual nanotubes. Moreover, the purification these nanotubes we re subjected to can shorten the SWCNTs. During the experiment, samples with nanotubes deposited by this method would often pull apart once tensioned by the AFM probe. This evidence, along w ith the low density of surviving suspended nanotubes, indicated that even though the ropes of bundled nanotubes would span the width of the tren ches, oftentimes the individual nanotubes

PAGE 40

27 comprising that bundle would not. This problem led to a switch to CVD grown nanotubes. 4.1.2.2 Chemical Vapor Deposition Grown Nanotubes Substrates for CVD growth were diced a nd cleaned by the same method mentioned above but the buffered oxide etch was omitted so that the samples re tained their thermal oxide layer. Additionally the chips were calcined at 1000 C in air for 7 minutes. Iron nitrate nonahydrate (Fisher Scie ntific, certified A.C.S.) wa s used as the catalyst.38 and dissolved in 2-propanol (Fisher Scientific Optima) in concentrations of about 10 g/ml. Several drops, enough to flood the surface, we re dropped onto the chip and the solution was spun dried at 3000 rpm. Growth took place in a 1” diameter tube furnace (Thermolyne F79300, 12” heating zone ) using conditions reported by Li et al .39 Hydrogen (200 sccm) and argon (300 sccm) were flowed across the chips placed aproximately 2 cm downstream of the fu rnace center while the furnace heated to 900 C. Once at temperature, the argon flow was turned off and methane (200 sccm) and hydrogen (200 sccm) were introduced for 10 minutes. Following growth, the methane and hydrogen flow were stopped, the furnace h eat switched off, and the samples were allowed to cooled under flowi ng argon (300 sccm). The density of the nanotubes grown on the samples was characterized by atomic force microscopy imaging. The CVD grown nanotube samples were superior to those made with deposited laser ablation grown tubes. The CVD grown nanotubes were much longer, sometimes with lengths over 10 microns, and were often individual nanotubes instead of bundles. Since all of the SWCNT’s grown by CVD were likely to survive the pr ocessing steps a CVD gr own sample with a lower beginning density, like that in Figure 4-3A, would result in higher density of

PAGE 41

28 suspended nanotubes than a sample with a hi gh starting density of nanotubes deposited out of Triton-X suspension, like that in Figure 4-3B. Figure 4-3: AFM images of nanotubes on si licon. A) Nanotubes grown by CVD. B) Nanotubes deposited out of solution. A B

PAGE 42

29 4.1.3 Electrode Patterning To make electrical contact to the nanotubes and also pin th em in place (so that they could be suspended) metal electrodes were de posited on top of the nanotubes. Patterning was done by photolithography. A recent development in the field has been the formulation of lift off resists (LORs). These resists are not photosen sitive but dissolve in the same developer used for the primary resist. The LOR is applied in a thin layer to the substrate to underlie the subsequently spun on photosensitive resist. During the photoresist development step, the exposed edge of the LOR underlying the unexposed (not removed) photoresist is dissolved back resulting in an eff ective undercut to the photoresist layer. This undercut means that when metal is depo sited it is done so discontinuously as illu strated in Figure 4-4. Figure 4-4: Metal deposition on re sists. A) Discontinuous metal deposition as a result of using a lift off resist. B) Continuous metal laye r using only a standard photoresist. When using only conventional photoresist, the metal tears at the points where the resist is lifted away causing rough edges on th e pattern. The use of LOR permits smooth electrode edges and easier lift off procedure requiring minimal ultrasonication for the liftoff. AFM line scans (essentially cross-secti ons )of electrodes fabr icated both with and without an LOR layer are shown in Figure 4-5.

PAGE 43

30 Figure 4-5: AFM scope trace of electrodes. A) No LOR, the flaps of metal at the edges are clearly visible. B) With LOR During the course of this work, a particul ar benefit of using LOR with regard to using carbon nanotubes in lit hography applications was found. After all the lithography steps, concluding with lift off, the sample s were AFM imaged to inspect the surface quality, nanotube density, and metal height. Samples that used an LOR layer contacting the nanotubes had much cleaner carbon na notubes than samples that only used the Shipley S1813 contacting the nanotubes. It seems that the organi c materials in the photoresist have a particular affinity for the nanotubes. Although the surface of the A B

PAGE 44

31 underlying silicon was very clean the SWNTs app eared to have globs of material stuck to them. Despite submersion in various solvents (photoresist remover, acetone, methanol, etc.), sometimes with sonication and at el evated temperatures, the globules remained attached to the nanotubes. Subjecting the nanotubes to an extended ozone cleaning step (~15 minutes) did help remove much of the mate rial but there is the possibility this could be harmful to the nanotubes and even that di d not remove all of the extraneous material. The components in the LOR do not seem to have the same affinity towards the nanotubes and leave an uncontaminated surfac e without the need for additional cleaning. So there are two benefits to using LOR duri ng electrode fabrication. The electrodes have smooth edges and standard photolithography can be performed on SWCNTs without being hampered by additional contaminants. Figure 4-6A shows the image of a nanotube on which Shipley S1813 photoresist was used without the LOR layer protecting the nanotubes. The contaminants are clearly vi sible on the surface of the nanotube. For comparison, Figure 4-6B shows two intersec ting nanotubes on a sample where LOR and the same Shipley S1813 photoresist were use d. The latter sample is obviously much cleaner. To prepare the silicon chips with nanotubes on the surface for lithography, the samples were heated in a 110 C oven for 30 minutes to drive off any surface water. Microchem LOR 3B was diluted 1:1 with type G thinner and spun onto the silicon surface at 5000 rpm for 30 seconds. The LOR layer was baked at 170 C for 45 minutes. Shipley S1813 was used as the photoresist and was mixed 3:1 with Shipley Type P thinner. It was spun on at 5000 rpm for 30 s econds which yielded a layer of photoresist ~800nm thick. The photoresist-coated sa mples were then soft baked at 90 C for 30 minutes. UV exposure was performed with a Karl Suss MA6 mask aligner with a 365nm

PAGE 45

32 UV light source (11 seconds exposure time, 7.4 mW/cm2 intensity). Following exposure, the samples were developed in Shipley MF -319 developer (22 seconds), rinsed in dionized water and blown dry with a stream of clean nitrogen gas. Figure 4-6: AFM images of nanotubes af ter lithography. A) Nanotube coated with photoresist. B) Nanotubes coated with a layer of LOR under the photoresist A B

PAGE 46

33 Prior to metallization the samples were s ubjected to a UV ozone cleaning step for 1 minute. This treatment helped remove any residual organic contamination left on the nanotubes and ensures good adhesion of the meta l to the nanotubes. Since the metal acted as the etch mask for the later SiO2/Si etch, without this treatment, those etchants would chew through the residual photoresist layer, ge tting under the metal el ectrodes, to attack the underlying silicon. The next step in sample preparation was metallization. Chrome/Gold and Chrome/Platinum-Iridium we re both used as the pinning electrode metals. The chrome/gold layers were grown by thermal evaporation while chrome/platinum-iridium samples were grown by RF magnetron sputtering. After metallization, a lift off step to remove the rest of the photoresist and unwanted metal was employed. Lift off occu rred in Michrochem Nanoremover PG. The samples were immersed in the Nanoremover and place in a 60 C bath for 30 minutes. During the last 10 minutes of that the samp les underwent low power ultra sonication to aid the lift off process. High powered or pr olonged sonication was found to remove the nanotubes from the surface. AFM images of samples subjected to more powerful sonication show many nanotubes lying unde r the surface of the electrodes but no nanotubes in the space between. Following the first heated bath, the samples were immersed in fresh nanoremover solution and placed in a 60 C bath for an additional 30 minutes. After lift off, the samples were ri nsed in 2-propanol and blown dry with clean nitrogen gas. 5.1.4 Etching To suspend the nanotubes the underlying sili con was etched away. Samples with an oxide layer were first etched in a 1:1 buffe red oxide: dionized water solution. The next step for these samples, and the first step for samples whose oxide layer had previously

PAGE 47

34 been removed, was an anisotropic wet silicon etch. The etchant so lution was an aqueous 30% (by weight) potassium hydroxide soluti on. The KOH solution was mixed 4 parts to 1 with 2-propanol and heated to 60 C. Samples were etched for 90 seconds, which gave a trench depth of around 600 nm. To quench the etching, the samples were moved from the KOH bath to a 60 C deionized water bath and then to a room temperature dionized water bath but were kept submerged during ea ch transfer. A schematic of all processing steps described so far is illustrated in Figure 4-7. 5.1.5 Release Procedure Location of the nanotubes for the experiment first required imaging the samples in a scanning electron microscope to find the na notubes. This required that the samples be dry. If the nanotubes samples were removed directly from water to air, the surface tension of the water would pull down the suspen ded tubes to the bottom of the trenches. Originally we tried the method used by Walters et al.7. This technique involves slowly exchanging the liquid for successively lower surface tension solvents while keeping the sample immersed the entire time. After etching the water was slowly exchanged for acetone, which was in turn slowly replaced wi th tetramethylsilane (TMS). TMS has a low surface tension of 10.2 mN/m, or roughly 1/7 the value of water. Once the TMS exchange was complete the samples were removed from the liquid and the TMS was allowed to evaporate. Unfortunately, TMS is difficult to purify and the impurities in the solvent were left on the nanotubes. To avoid this contamination we develope d an alternative approach in which the samples were “freeze dried”. After etching a chip was transferred to a small copper boat filled with water just sufficient to cover th e sample. The water was then flash frozen by

PAGE 48

35 placing the boat directly into li quid nitrogen. These frozen sa mples were then placed in a freeze drier (home built) where the ice was slowly sublimed away Spin coat LOR Figure 4-7: Illustra tion of processing steps for su spended carbon nanotube samples. A method to determine if these processing steps had an effect on the samples was to measure the resistance between adjacent el ectrodes (across the nanotubes) before and after the etching steps. For the actual sa mples numerous nanotubes did not survive the processing, ending up at the bottom of the trench Since the doped silicon then provided a current path between the adjacent electrodes via these downed tubes these samples could not be used for such measurements. Instead nanotubes were grown by CVD on silicon

PAGE 49

36 with a 600nm oxide layer and a 14 electrode in terdigitated pattern (made of thermally evaporated Cr/Pd) deposited on top of the nanotubes. The spacing between electrodes was the same as the pattern used for the ac tuation measurement samples (~ 1 micron). These test chips went through identical pr ocessing steps (HF etch, KOH etch, and TMS release or freeze drying). Since the oxide was much thicker on these samples the HF step did not etch through the oxide layer. The resistance was ch ecked before and after the processing steps for samples that went th rough the liquid release procedure and freeze dry procedure, Table 4-1. The resistances of the samples did increase but that is likely due to fewer numbers of tubes surviving to connect across electr odes at this stage. Table 4-1: Resistance before and after processing steps Sample Release Procedure Rbefore Rafter Rafter/Rbefore 1 Freeze dry 7.5 k 152 k 20 2 Freeze dry 12.3 k 1470 k 120 3 Freeze dry 3.72 k 23.5 k 6.3 4 Liquid 3.17 k 11.7 k 3.7 5 Liquid 3.54 k 46 k 13 6 Liquid 3.05 k 16.6 k 5.4 Following either of these “release” procedures, samples were imaged by a Hitachi S-4000 field emission scanning electron micr oscope (SEM) to identify suspended nanotubes.The suspended nanotubes are fairly robust. In addition to imaging by SEM they can also be imaged by tapping mode AF M with no damage. Shown in Figure 4-8 is tapping mode image of a suspended nanotube sample and the corresponding SEM image of the same nanotube.

PAGE 50

37 Figure 4-8: Suspended carbon nanotube images A) Height tapping mode AFM image of a suspended nanotube. It can be clearly seen that the nanotube is suspended and intact. B) AFM deflection image -m ore detail on the surrounding area can be distinguished. C) SEM image of the same nanotube 4.1.6 Identification of Suspended Nanotubes One of the difficulties of an experiment such as this is being able to locate the suspended tubes with the AFM tip. Fortunately we were able to exploit an unwanted by product from the SEM to make this process easier. Prolonged exposure of a substrate to

PAGE 51

38 the electron beam in a SEM usua lly results in the build up of a deposit generated from the breakdown of residual hydrocarbons in the ch amber under the electr on beam irradiation. After finding a suitable suspended nanotube, we moved directly over to the electrode pinning it and did a prolonged line scan on the electrode. After the mark was deposited a SEM image was taken of the mark and its re lative position to the suspended tube. Thus once the mark was found by the AFM tip, the distance to the nanotube could be found by referring to the position of the mark in re lation to the nanotube in the SEM image. Another SEM image was taken referencing the created marks to some type of larger fiduciary marking (such as a corner or defect in the lithography). Thus a map of the marks is produced. An example is shown in Figure 4-9. Figure 4-9: SEM image of a map of the created marks. 5.1.7 Final Sample Preparations After marking the sample, the chip was permanently attached to a polished AFM puck with conducting silver epoxy. The epoxy wa s carefully molded around the edges of the chip to create smooth sloping walls reachi ng from the top of the sample down to the

PAGE 52

39 puck. A shadow mask masked off the center electrode area, containing the suspended nanotubes, and metal was deposit ed over the entire puck/chip assembly. This extra layer of metal electrically connect ed the large electrode pads (and consequently the small electrodes pinning the nanotubes) to the AFM puck. Finally, the experiment itself was done in the liquid electrolyte environment and thus, the samples had to be carefully submerged again. In this case, we immersed the chips into the low surface tension TMS. As no liquid evaporates in this procedure, the purity of the TMS is not an issue. In the re verse of the previously mentioned “release” procedure, the liquids were slowly exchange d for those with higher surface tensions with the last liquid being the electrolyte solution used in the experiment. Experiments were performed using 0.1 M to 1 M solutions of NaCl in water, NaNO3 in water, LiClO4 in acetonitrile, and LiBF4 in acetonitrile. 4.2 Experimental Procedure All experiments were done in an AF M electrochemical cell and a Digital Instruments Nanoscope III AFM. This speci al AFM tip holder is manufactured out of glass so the laser can still reflect off of the cantilever and reach the split photodiode. The cell has a center chamber with an o-ring gr oove surrounding it. As the cell is lowered closer to the sample surface the o-ring compresses and forms a seal between the glass sidewalls of the electrochemical cell and also the AFM sample puck. To load the sample in the electrochemical cell, the greased o -ring was submerged into a beaker containing both the immersed sample and the electroly te. The o-ring was carefully placed on the AFM puck so that it surrounded th e mounted silicon chip and is illustrated in Figure 4-10. This allowed removal of the puck from liqui d while still keeping the suspended SWCNTs submerged in the small pool of electrolyte within the o-ring walls.

PAGE 53

40 Figure 4-10: Cross sectional schematic of ch ip mounted on AFM puck during transfer to the AFM. During the actuator experiment voltages we re applied using a three terminal setup with a Princeton Applied Research Pote ntiostat/Galvanostat model 283. The three terminal arrangement includes a working electr ode, a counter electrode, and a reference electrode immerse in an elec trolyte. The reaction of inte rest occurs at the working electrode. Whenever a metal is immersed in a solution a potentia l drop occurs at the boundary of the electrode and electrolyte. Ther e is no way to independently measure this without introducing another electrode into the solution, which in turn has its own potential drop at the interface. In a two term inal measurement chemical reactions at the working and counter electrodes can cause th e potential drop at the interface to be different from the applied potental. Thus the actual difference between the two electrodes is unknown. A reference electrode has high impeda nce and is chosen so that it provides a stable and reproducible potent ial against which the worki ng electrode potential can be controlled. Thus any changes that occur in the cell potential occur at the working electrode and not the reference. The refere nce electrode is desi gned to reproduce the same potential regardless of the solution the working electrode is in. The counter electrode is an inert metal (usually Pt) that completes the circuit.

PAGE 54

41 The electrochemical cell had 3 ports whic h we made use of. The first port was connected to a syringe that could be use for adding a nd exchanging the electrolyte solution. While not in use, the syringe was disconnected and the por t sealed. The second port connected to a small glass tube that hous ed the reference electr ode (Ag/AgCl for the aqueous experiments and Ag/Ag+ for the aceton itrile runs). Between the glass tube and AFM port was a porous Vycor frit. This frit allo wed the ions in solution to pass while not allowing solvent exchange. Fi nally, the third port contained the counter electrode. We opted to use a strip of bucky paper due to its high surface area and thus large capacitance. The working electrode was the nanotube sample and voltage was applied to it through the AFM piezo cap. Prior to use in the electrochemical cell, the electrolyte solution was sparged with helium. Sparging removes dissolved gases fr om liquids by bubbling an inert gas through the liquid. Air bubbles have the possibility of in terfering with the experiment in a variety of ways. An air bubble above the cantilever can interfere with the laser reaching the photodiode and disrupt the measured signals. An air bubble in either the reference or counter electrode port causes a discontinuity in the electrolyte so lution and thus would break the conduction path. In the case of a bubbl e in the counter elec trode line the sample simply would not get the potential we were tr ying to apply. In the case of a bubble in the reference electrode line the potentiostat w ould apply the maximum voltage between the counter and working electrode and destroy the nanotube samp le. The electrolyte solution was sparged for about 15 minutes before loading. Upon loading the sample and tip into the liquid environment we observed an enourmous drift in the free space deflection sign al of the cantilever. After an initial huge drift the deflection signal change would slow down but not completely dissipate. This

PAGE 55

42 has often been attributed to thermal drift in the literature. However, this explanation is completely inadequate since the cantilever is better thermally coupled to its environment in a liquid than it is in air. This drift wa s far too large to allow performance of the experiment requiring that its source be determ ined to continue. We ultimately determined that this drift was due to chemical interactions between the tip and the liquid surroundings. The heat from the laser reflec ting off the metal-coated backside of the cantilever would increase the rate of these r eactions and would ofte n etch the reflective coating. To fix this problem, we depos ited (in a home built system) a thin (~50nm) conformal layer of parylene C on the tip.40 Parylene C is a polymer deposited from the vapor phase that is both chemi cal inert and electrically in sulating. The basic procedure for parylene deposition involves subliming th e dimer, di-para-Xylylene, in a moderate temperature zone (95 C), decomposing the dimer into the monomer in a high temperature zone (680 C), and spontaneous polymerization of the polymer on a cooled surface (20 C). By isolating the tip from the solution we eradicated the drif t, yielding a stable deflection signal. Figure 4-11 shows the drif t of the free space de flection signal over 3 hours for two tips, one bare tip and one co ated with parylene, in aqueous 1M NaNO3. The AFM tips were calibrated by the reference cantilever method41 after application of the parylene coating. To find the marked nanotubes with the AFM, the tip was engaged on the large electrode pads in the near a fiduciary mark. Once the end of the electrode was found, the tip was translated down the electrode (by th e amount indicated from the SEM map) until the mark was located. The marks are quite large (heights of several hundreds of nanometers and lengths around a micron) and ar e easily imaged with the AFM, Figure 412

PAGE 56

43 Figure 4-11: Free-space drift of a bare tip and a paralyne coated tip in aqueous 1M NaNO3 Figure 4-12: AFM image of identif ying mark created in the SEM

PAGE 57

44 AFM imaging performed in the fluid e nvironment was done in contact mode. Contact mode AFM uses a soft cantilever that is in intimate contact with the surface while the sample is rastered back and forth. In the Digital Instruments Multimode AFM the piezo motion is all that of the sample st age (the tip is stationary except for its Z deflection caused by interaction with the sample ). A laser is bounced off the back of the cantilever (which has been coated with a reflective metal) to a split photo diode detector. Before the tip is contacted with the surface to be imaged, the position of the detector is usually adjusted so that the measured laser intensity striking the top portion of the split photo diode equals that striking the bottom por tion (the free space value is set to zero). Once in contact, the stage will move upward deflecting the tip until the signal on the split photodiode reaches the deflection setpoint set by the user. The amount of force exerted by the cantilever on the sample is determined by the difference between the free space setpoint and the deflection set point. While engaged with the surface, vertical deflections of the cantilever are detected as differences in the laser intensity on the top and bottom halves of the split photodiode. The force applied by the cantilever is equivalently that of a spring, F=Kx, with the bending force constant, K, of the cantilevers being calibrated prior to use and x being the actual tip deflection. After located the mark, we used the AFM in an unconventional way to make contact with the nanotube. If the deflection setpoint is set to a value below the free space deflection the sample will move down away from the tip in an effort to decrease the value measured by the photodiode to the user defi ned setpoint. Once the mark was found the scan size was set to zero, the feedback ga in was reduced (to make the rate of piezo response to differences between the setpoint and the signal slow) and the deflection setpoint was chosen to be below the free space signal in order to move the sample away

PAGE 58

45 from the tip. To stop the Z motion once th e tip was disengaged from the surface the feedback gain was set to zero. At this poi nt the X and Y positions of the stage were adjusted such that the tip would be placed directly over the suspended SWCNT. To guarantee minimal stress to the nanotube a low setpoint just above the free space value was chosen and the gains were set to a very low level. This caused the stage to be brought up very slowly as the nanotube came in contact with the tip. Once the tip reached its setpoint while tensioning the nanot ube, the gains were raised back to their original value. To accurately find the nanotube, the tip wa s lifted and lowered several times while making small increments along the length of the trench until the highest spot, and thus the position of the nanotube, was located. On ce found by this method, force calibration curves against the nanotube were taken to extr act essential information and to verify that the object the tip located was tr uly a nanotube. Force calibrati on mode is a useful feature within contact AFM. In this mode the sa mple stage is repeatedly moved up and down while plotting the tip deflection against the Z position of the stage as shown in Figure 413A. The right side of the plot corresponds to the stage being furthest from the tip. The white line is the deflection signal of the cantil ever as the stage moves toward the tip and the yellow is the deflection as the stage retr eats. The deflection signal remains flat until a surface is hit, at which point the tip begins to deflect. The stage will move until the tip reaches the user-defined setpoint. Against a hard surface the tip deflection will equal the Z movement, as any movement by the stage will cause the same amo unt of deflection in the tip (this permits calibration of the tips Z deflection signal in volts to a distance sin ce the displacement of

PAGE 59

46 the piezoelectric stage, per volt applied to it is known). However, against an object that stretches, like a nanotube, the Z movement w ill be greater than the tip deflection, the difference being the amount of vertical deform ation of the object. A schematic drawing of the tip against both a ha rd surface and a nanotube is shown in Figure 4-14. This difference made it easy to recognize a nanotube being tensioned versus say the bottom of the trench. Because they are elastic, when th e tip first comes in cont act with the tube is stretches easily. The nanotube becomes harder to stretch as the tip continues to press on it and the curve begins to straighten. Figur e 4-13A is a force calibration curve against a metal electrode while Figure 4-13B shows a force calibration curve against a suspended SWCNT using the same tip as in Figure 4-13A. The force causing this stretch is calculated by multiplying the force constant of the cantilever (measured before loading the samp le) by the deflection of the tip. The number of nanotubes in a rope can be approximated from parameters derived from the force calibration curve.

PAGE 60

47 Figure 4-13: Force calibration curves against two surfaces. A) Against a hard surface. B) Against a suspended nanotube A B

PAGE 61

48 Figure 4-14: Illustration of a force calibration curve against a hard surface and nanotube. The original vertical stretc h of the nanotube, the initial deflection designated by y and illustrated in Figure 4-15, is calculated by subt racting the deflection of the tip from the Z movement of the sample stage while the tip and nanotube are in contact. Simple statics considerations give the force needed to deflect the nanotube a vertical distance y as, k is defined as k = Y A/L0 (4.2) Y is the Young’s modulus (1012 Pa) for graphene. A is the cross sectional area of the nanotubes under tension. The cross sectional area for one nanotube is the van der Waals thickness for graphite (t = 0.34x10-9 m) multiplied by the circumference of the nanotube ( d where d is the diameter of the tube) Thus the total cross sectional area for a rope of F2k1 L o y2L o2 1 2 y ( 4.1)

PAGE 62

49 nanotubes is n t d where n is the total number of nanot ubes in the rope. This gives the following expression for k. Figure 4-15: Diagram of nanotube under tension. The parameter y is essentia l to calculating the observe d actuation. The tensioned length of the nanotube is L1=(L0 2 + y2)1/2 where L0 is the same as mentioned above. Any actuation of the nanotube will cause a deflection of the cantilever by dy as the tube lengthens or contracts. Using dy we can calcu late the new length of the tensioned tube as L2=(L0 2 + (y+dy)2)1/2. Thus the actuation strain L/L is defined as L/L=(L2 – L1)/L1. It is instructive to plot dy as a function of y for a given amount of actuation, i.e. setting L/L to a constant, as shown in Figure 4-16. The parameters L0 of 430nm and L/L of 0.0001 (corresponding to an actuation of 0.01%) were used. The value of dy decreases sharply with incr easing y. Once this particular tube has been stretched past y=18nm the value of dy becomes less than 1nm. In order to observe actuation with the AFM tip, the nanotube must be under some tension but the amount of initial stretch needs to be small in orde r to have any sensitivity to the effect. k nYtd L o ( 4.3)

PAGE 63

50 Figure 4-16: Plot of dy versus y forL0=430nm and L/L=0.0001. Both y and dy are plotted in nanometers. After the tip tensioned the nanotube by the amount y, as measured by the force calibration curve, voltage was applied to th e electrodes through the piezo cap of the AFM. Elongation or contraction of the nanot ube resulting from the applied voltage will cause a deflection of the cantileve r. A Princeton Applied Research Potentiostat/Galvanostat model 283 controlled all applied voltages to th e three electrodes. A square wave potential with a low frequency of either 0.5Hz or 1 Hz was used. Data of the applied voltage, current, a nd piezo movement in the Z di rection was recorded with a Labview program. Usually, data was taken fo r different values of pre-tension y and different applied voltages on each nanotube. Seve ral runs were taken with each variation in parameter.

PAGE 64

51 CHAPTER 5 RESULTS AND DISCUSSION OF CHAR GE INDUCED ACTUATION OF SUSPENDED CARBON NANOTUBES 5.1 Results of Actuation Measurements The experiment was performed on 35 su spended nanotubes. Several distinct electrolytes were used. Aqueous NaCl, aqueous NaNO3, LiClO4 in acetonitrile, and LiBF4 in acetonitrile were employed in concentr ations ranging from 0.1 M to 1 M. The organic solvents were used to obtain a larger voltage window than the hydrolysis of water permits. Square wave voltages were applie d up to 2 Volts peak to peak. During some runs only negative voltage was applied to th e sample since the bond length changes are expected to be greater for electron injec tion compared to electron withdrawal. The parameter monitored to detect dimensional changes in the nanotube was the Z voltage supplied to the piezo, a change of 1 V corres ponded to 12.9nm. Changes in the nanotube length caused the AFM tip to either relax be low or be pushed above the setpoint. The sample stage would move to compensate for this change in length so that the setpoint was restored. Out of the 35 samples only 3 nanotub es displayed length changes that could be duplicated in more than one run. Sample 1 is shown in Figure 5-1. As measured from the SEM image the parameter L0 is 484nm. This sample was immersed in 1M aqueous NaCl solution. An example data run is shown in Figure 5-2 and a summary of all the data runs for this nanotube is listed in Table 5-1. The paramete r y is the initial vertical deflection of the tube as measured by the force calibration curves dZ is the is the voltage change in the AFM peizo corresponding to the vertical nanotub e changes, dy is the dZ voltage from the

PAGE 65

52 piezo converted into nanometers (12.9nm/Volt) and % dL/L is the percentage strain of the nanotube. Figure 5-1: SEM image of nanotube sample 1. Th e nanotube of interest is the thin tube directly to the right of the mark Figure 5-2: A data file from sample 1. The applied voltage is shown in black with its scale on the left. The Z piezo movement is shown in blue

PAGE 66

53 Table 5-1: Summary of data from sample 1 Run Y (nm) dZ (V) Dy (nm) % dL/L 1 36.9 0.0269 0.346 0.00544 2 36.9 0.0175 0.225 0.00354 3 14.3 0.0206 0.265 0.00163 4 14.3 0.0189 0.243 0.00150 Nanotube sample 2 is shown in Figure 5-3. This sample was exposed to polyethylene imine (PEI) prior to loading in the AFM electrochemical cell. PEI has been shown to n-dope the nanotubes.42 In graphite, bond length ch anges for the same amount of charge transfer are smaller if the sample is already p-doped but larger if the sample is n-doped. PEI is an extremely viscous liqui d and was dissolved in methanol from concentrations starting at 2.5% an increasing to 20%. The suspended nanotube sample was first carefully submerged in methanol which was slowly exchanged for the 2.5% PEI solution. This was repeated 4 times with subsequently higher c oncentrations of PEI solution with the final step resulting in the sample residing in the 20% PEI mixture. To adsorb the PEI onto the nanotubes the samp le was left submerged overnight. The following morning the sample went through the reverse of the process, with the liquids being exchanged for lower concentration so lutions of PEI. The PEI solution was exchanged for pure methanol and that was ex changed for water at which point another freeze dry was performed. From the SEM image of this sample the parameter L0 was determined to be 429nm. This experiment was preformed in 0.5M LiBF4 in acetonitrile. As sample of a data file is show in Figure 5-4 and a summary of the parameters from all data files for sample 2 is shown in Table 5-2

PAGE 67

54 Figure 5-3: SEM image of nanotube sample 2. The nanotube is to the left of the mark. Figure 5-4: Sample 2 data

PAGE 68

55 Table 5-2: Summary of data for sample 2. Run y (nm)dZ (V) Dy (nm) % dL/L 1 18.4 0.0134 0.172 0.00178 2 18.4 0.0104 0.134 0.00134 3 11.4 0.0151 0.194 0.00121 4 11.4 0.0230 0.296 0.00186 Sample 3 was also submerged in 0.5 LiBF4 in acetonitrile. From the SEM image shown in Figure 5-5 the parameter L0 was determined to be 444nm. Figure 5-6 displays the data from one run and Table 5-3 is a su mmary of all the data runs for sample 3. Figure 5-5: Image of sample 3. The nanotube m easure is the one to the left of the lower marking.

PAGE 69

56 Figure 5-6: Sample data file from nanotube 3. Table 5-3: Summary of a ll data from sample 3 Run y (nm) dZ (V) dy (nm) % dL/L 1 16.8 0.0102 0.131 0.00112 2 16.8 0.0112 0.144 0.00123 3 16.8 0.00800 0.103 0.000878 4 14.6 0.0115 0.148 0.00110 5 14.6 0.0119 0.153 0.00114 6 14.6 0.0109 0.140 0.00104 To ascertain whether the discrete jumps observed in the in the Z movement from the application of voltage to the nanotubes where a result of nanotube actuation or from some type of chemical reaction in the cell othe r data files were also taken. These include monitoring the Z movement while the tip was suspended in the electrolyte but not in contact with anything as shown in Figure 5-7. Additionally data was taken with the tip in contact with the bottom of th e silicon trench, Figure 5-8.

PAGE 70

57 Figure 5-7: Data of the Z Movement and applied voltage while the AFM was suspended in free space Figure 5-8: Data of Z Movement and applie d voltage while the AFM tip was in contact with the bottom of the trench.

PAGE 71

58 A summary of all the data with average actuation values is listed in Table 5-4. Table 6-4: Summary of the actuator data Sample Electrolyte Applied Voltage (peak to peak) L0 (nm) Y (nm) Dy (nm) % Actuation 1 1M NaCl in H2O 1.5 V 484 36.9 0.285 0.0045 2 0.5M LiBF4 in CH3CN 1.2 V 429 11.4 0.245 0.0015 3 0.5M LiBF4 in CH3CN 1.4 V 444 14.6 0.147 0.0011 5.2 Discussion of Results The experiment perform by Baughman et al .8 observed much higher actuation strains in the macroscopic sheets. For an appl ied voltage of 0.9 Vpp the strain was nearly .1%. Since the nanotubes were bundled, with only the outer nanotubes presumed to be undergoing actuations, these valu es were estimated to be lower limits to nanotube actuation. Based on their results, the author s speculated that the maximum strain for individual nanotubes could re ach ~1%/V. Our measurements fall far short of that prediction and are smaller than that observed for the nanotube sheets. This lack of evidence for charge-indu ced actuation in our work led us to reexamine of some of our fundamental assumptio ns. The first question that needed to be asked was “how much charge can a nanotube hold?” And once that has been determined, “how much strain is predicte d for this amount of charge in jection. To answer the first question we calculated the capacitance of a carbon nanot ube under these conditions. Previously, Krger et al.43 and Rosenblatt et al.44 had used MWCT and SWNT in electrolyte gated transistors. Both groups had calculated the capacitance per unit length of the nanotube within the electrolyte soluti on using the geometry of two concentric cylinders. C= 2 0 / ln(R2/R1) (5.1)

PAGE 72

59 The inner cylinder is the nanotube whose radius is given as R1 in the equation. The value they used for the outer cylinder diameter R2 was that of the nanotube, R1, plus the Debye length, D, for the ionic concentrati on of the solution. They also used the accepted value of =80 for dielectric constant of wate r. Both groups used electrolyte concentrations of 1mM whic h gives a Debye length of D =1nm. Using these values the two groups calculated gate capacita nce values of ~ 10 nF/m or 10-17 F/nm (62.5 e-/nm/V). This use of the bulk dielectric constant of water responsible for such large capacitance values is however not correct in the case of electrolytes where the effective spacing between electrodes is the width of the electric double layer. In electrolyte solutions, when a voltage a ppears on an immersed electrode (in this case the nanotube) the ions of the opposite pola rity will migrate towa rd the electrode to neutralize the charge. Thus th e composition of the electrolyt e solution near the interface is different than the rest of the bulk solution generating the outer half of the electrochemical double layer (the inner half being the charge on the electrode). The layer of charge at the surface of the electrode a nd the concentration of opposite charges in the solution surrounding the electrode generate a capacitance. However calculation of the double layer capacitance is non trivial and remain s an active area of research and debate (e.g. see chapter 2 reference 45). For example, what is the locati on of the counterions surrounding the electrode? The most accepted arrangement is the Gouy-Chapman-SternGrahame model. In this mode l part of the potential drop oc curs from a layer of fairly closely packed counterions near the surface, the Helmholtz layer. The rest of the potential drop occurs over a diffuse section of ions in which electrostatic forces compete with Brownian motion. For solutions with a co ncentration of 1M or higher generally the entire drop takes place over the compact He lmholtz layer and the influence from the

PAGE 73

60 diffuse layer is only a minor correction. The positions of the ions near the electrode are also influenced by the dipole moments of the water molecules surrounding them. Conversely, the electric fields generated by the ions in solution and the electrode also affect the water dipole moments and can change the dielectric constant. This effect was noticed as early as 1948 by Hasted et al .46 who tried measuring the modified dielectric constant in aqueous salt soluti ons. The ions in solution orient the water molecule dipoles. The water molecules in the immediate vicinity of an ion are immobilized by the ion’s electric field. The ion is referred to as hydrated. Figure 5-9 shows an illustration of hydrated Na and Cl ions Figure 5-9: Hydrated Cland Na+ ions The polar water molecules are no longer fr ee to rotate in response to the applied external electric field, which is what impart s to water its large bulk dielectric constant. The presence of ions in the water breaks up the water network and decrease the overall dielectric strength, something known as dielectri c saturation. The dielec tric strength for a water-counterion complex for Na is Na~2 and the radius of a hydrated Na ion is r=0.36nm.47,48 If we use this radius as the di stance to the outer cylinder in our capacitance equation we have to use the drastically reduced value of Na=2 for the dielectric constant. Us ing these values (and R1=1.36nm, the diameter of a (10,10) nanotube) the calculated cap acitance goes to C=2.6x10-19 F/nm (1.6 e-/nm/V). This value

PAGE 74

61 is more than order of magnitude smaller than what was estimated by Krger et al.43 and Rosenblatt et al.44. In one nanometer of length fo r a SWCNT with a diameter of d=1.36nm we have 160 C atoms. Using this in formation we can convert to electrons per carbon atom. C=(1.6 e-/nm/V)(1 nm/160 C atoms)= 0.01 e-/C-atom/V. (5.2) So for sample 1 that had 1.5 volts peak to p eak applied to it, we can expect that the amount of charge injection was q = 0.0075 e-/C-atom. The theoretical work on SWCNT dimensional changes discussed in Chapter 3 give an average of 0.1% strain for the addition of 0.01 e-/C. Hence for 0.0075 e-/C-atom calculated from the capacitance we should expect a strain of 0.075% .While acetonitrile (bulk =37.5) was used as the solvent in a number of the experiments the same considerations can be expected to apply. Results from a spectro-electrochemical st udy on carbon nanotube thin films in the Appendix give further support to this cap acitance estimate. Carbon nanotubes have optical absorbance peaks cen tered around 1650 nm and 900 nm. These peaks are due to photoinduced electronic transitions between valence to conduction band van Hove singularities. The 1650 nm and 900 nm peaks correspond to the 1st and 2nd van Hove singularities for semiconducting nanotubes. Th e peaks are broad becau se of the range of nanotube diameters in the sample and also be cause of perturbations to the electronic structure from tube to tube interactions. The absorbance peaks can be reduced or even eliminated by shifting the Fermi level. If electrons (holes) are added then the singularity in the conduction (valence) ba nd will be filled (depleted) and no electronic transitions, and thus absorption of phot ons, will be possible. From these measurements we can determin e at what applied voltage the first van Hove singularity is either depleted or filled. At an applied voltage of negative -0.7 V the

PAGE 75

62 first peak has started to shif t but the second has not, Figure A-4. Note that the peak has not been completely removed. That is to be expected the sample contains bundles of nanotubes. At low applied voltages the Ferm i level of an outlying nanotube has been shifted while the Fermi level of an inner ly ing tube remains untouched. Since there is no change in the second peak we assume that ev en the outer tubes have not had their Fermi levels shifted into the second singularity yet. We determined the additional e-/C atom when the Fermi level was shifted to the poi nt just before the second semiconducting van Hove singularity by using a program that calcul ated the density of st ates for nanotubes of any indices (n,m). The method was based on zone folding of the tight binding 2D energy dispersion relations of graphite (Equation 2-11) and a simple rota tion transformation of kx and ky into k and k|| (relative to the nanotube axis).49 A graph of the density of states for a (10,11) semiconducting nanotube and the shifted Fermi level are shown in Figure 510. Integrating the density of states up to the dashed line in this figure gives 0.005 e-/Catom. From the results in the appendix an applied voltage of -0.7Volts therefore corresponds to an addi tional charge of 0.005 e-/C. We also did the same calculation for a (10,10) metallic nanotube with the Fermi le vel shifted the same amount and found an additional 0.007 e-/C-atom. Using V=0.7V in our cap acitance calculations (Equation 5.3) yields .007 e-/C-atom which is in reas onable agreement for the two values. It should be noted that the calculation for the capacitan ce (Equation 5.3) assumes the nanotube is a perfect metal able to accommodate unlimited additional electrons. Since the nanotubes are limited in the number of additional electrons they can contain by the density of states, the derived formula for capacitance is likely an overestimate. Despite the reduced capacitance, the results of Baughman et al.8 are around what would expected so capacitance alone cannot explain our small results. However, their

PAGE 76

63 sample also contained bundled nanotubes. Si nce the strain would be shared by the inner nanotubes, which would not be undergoing dimensional changes, the actual strain of the outer nanotubes would be gr eater than that displayed by the nanotube sheet. Additionally, their results disp lay some reversals in the actuations (Figure 3-2) at voltages of 0.5V and up that cannot be expl ained by bond length changes from additional charges in the orbital system. That behavior is very similar to an effect caused by double layer charging in porous graphite in which changes in the interfacial tension within the pores cause dimensional changes in the graphite.57 . Figure 5-10: Density of states for a (10,11) SWCNT. The SWCNT film used in the spectroelec trochemical experiment in the Appendix was baked to remove any dopants acquire d during purification. However, the transmittance minimum for the peak associat ed with the first semiconducting optical transitions occurred between –0.2V and –0.4V. That is where the point of zero charge (PZC) on the nanotube film would occur. This indicates that the nanotubes were p-doped

PAGE 77

64 by their exposure to electrolyte (which ha s its own chemical potential) atmosphere despite the desorbing bake. The experiment by Baughman et al.8 used aqueous 1M NaCl. For graphite in 1M NaCl the PZC is –0.2 V55,56,57 so it is likely the PZC of the nanotube film would also be negative. As mentioned, the dimensional changes in graphite (for the same amount of charge transfer) that is al ready p-doped will be smaller than the changes in un-doped graphite. Even accounting for the reduced dielectric c onstant, the predicted strain is still larger than what we measured. One of the difficulties within the nanotube field has been making good contacts to the one dimensional SWCNTs. SWCNT field effect ransistors (FETs) made with metals that were consiste nt with our fabricati on process have been found to be consistently p-type, even with good ohmic contacts for the on state. 50,51,52,53,54 It has been suggested that st rong dipole moments of adsorb ed gases (particularly oxygen) at the contacts modifies the barriers and this effect leads p-type behavior.53 Additionally, there is a barrier introduced by mismatch in work functions of the contact metal and the nanotube. Rosenblatt et al.44 formulated water gated nanotube FETs. Like our work, they used an electrolyte solution to gate the nanotube and also co ntacted the tubes with Cr/Au electrodes. They found that small diameter nanotubes (3nm and under) were still p-type. Even using the electrolyte gate they coul d not push across the ba nd gap to get electron transport. The results in the Appendix show that nanotube films can be pushed across the gap. However only a very small portion of the nanotubes in a film are in contact with the metal contact electrode and thus only th at small portion would be p-doped by the contacts. But individual nanotubes (like the ones used in our experiment and that of

PAGE 78

65 Rosenblatt et al.44) are sufficiently p-doped by the contacts so that the Fermi level cannot be shifted into the conduction band. These issues result in two factors that explain why charge induced dimensional changes in these experiments is so small. Contact barriers with nanotubes prevent electron injection in semiconduc ting nanotubes. And perhaps more importantly, the fact that the nanotubes are p-doped puts the nanotubes into the portion of the strain versus charge transfer curve where the length changes are smaller (see Figure 3-3). Attempts to inject sufficient electrons to take the nanotubes to the steepe r part of the curve require pushing the semiconducting tubes across the gap while the metal elec trode contact with the nanotubes result in Schottky barriers that further impede such electron injection. Using an extremely low work function meta l (such as Ca) would n-dope the nanotubes and cause barriers favorable to electron in jection. However, those metals are incompatible with the processes required to the suspended nanotube samples. Thus we are prevented from measuri ng significant charge induced dimensional changes in individual nanotube because of eff ects resulting from the contacts.

PAGE 79

66 CHAPTER 6 CONDUCTANCE CHANGES IN CARBON NANOTUBES DUE TO HYDROGEN The work in this chapter demonstrates the utility of car bon nanotubes as hydrogen gas sensors. Parallel effo rts demonstrated that sputtering of metals onto SWCNT decreases the intrinsic conduc tance of the nanotubes. 6.1 Carbon Nanotube Sensor Background The first demonstrated use of nanot ubes as chemical sensors was by Kong et al. .58 They monitored the charge transport changes of semiconducting SWCNTs when exposed to NH3 and NO2. These early sensors patterned sour ce and drain electrodes on individual nanotubes in a field effect tr ansistor (NFET) configurati on Upon exposure to electronwithdrawing NO2 the turn on voltage of the NFET was shifted by +4V. When exposed to electron donating NH3 the turn on voltage was shifted by V. The interaction of these molecules with the SWCNT shifted the Fermi le vel, through the density of states of the nanotubes, modulating their res ponse to the gate voltage. Other groups have since then tested the effect of vari ous gases on the conductance of carbon nanotubes using either individual tube s, bundles of tubes, or thin films of SWCNTs.59,60,61,62 The observation has been that electron donating species cause a decreased conductance of the p-type semiconducting nanotube s and electron withdrawing molecules increase the conductiv ity. These sensors all had long recovery times (hours) although exposure to UV light or elevated te mperatures helped speed up the desorption process.

PAGE 80

67 Recent efforts have focused more on using thin films of nanotubes as sensors. Novak et al. 63 used chemical vapor deposition to grow a nanotube network an SiO2/Si substrate that was just above the percolation threshold. Thes e films were used to detect several electron donating mo lecules including dimethyl methylphosphonate, DMMP (a simulant for the nerve agent sarin). This gr oup was the first to use the clever method of applying a positive back gate to refresh the device. They surmised that the Coulombic repulsion between the negative charges induced by the gate and the negative charge donated by the DMMP was enough to lower th e desorption barrier and drive off the DMMP. Although these devices are in some se nse easier to manufacture than individual nanotube FETs, consistent growth of a nanotube film near the percolation threshold can be difficult. The number of chemicals forming charge tr ansfer complexes with the nanotubes, and which may therefore be detected by this mechanism, is rather limited. Y. Lu et al.64 first demonstrated the detection of methan e, which does not by itself undergo charge transfer with SWCNTs, by lo ading the SWCNT film with palladium particles. The sensors were formed by first sputtering a 10nm thick layer of Pd onto SWCNT powder and mixed by shaking this concoction. Th e metal-coated powder was dispersed in dionized water, sonicated, and dr op-dried onto an interdigitat ed electrode pattern. This work certainly showed the utility of load ing SWCNTs with metals but the method of associating the nanotubes with the metal was poo rly controlled as was the density of the nanotubes bridging the electrodes. The work presented here focuses on using SWCNTs loaded with Pd particles for the detection of hydrogen gas. During the co arse of this work another group published a

PAGE 81

68 paper also using Pd-SWCNTs as H2 sensors.65 This group deposited Pd onto the nanotubes by RF sputtering and also by usi ng a technique in which a Pd salt solution (containing SWCNTs) was reacted with toluene to deposit Pd particles on the nanotubes. After Pd loading, the SWCNT films were deposited by airbrush ing a suspension of nanotubes in ethanol onto an aluminum substr ate. Contrary to the high sensitivity of sputtered Pd-SWCNT presented here, this group had no sensitivity to H2 with the sputtered Pd-SWNT samples. Additionally, we observed higher R/R ratios for amounts of H2 at least an order of magnitude smaller. Palladium is an interesting choi ce for metal loading of SWCNT H2 sensors for a couple of reasons. At room temperature and atmospheric pressure palladium can adsorb up to 900 times its own volume of hydrogen. A dditionally, palladium has been shown to make ohmic contacts with carbon nanotubes. Javey et al.54 used Pd electrodes to form semiconducting nanotube FETs. Up on exposing the devices to H2 they demonstrated a decrease in the conductance of the p-type semiconducting nanotubes. It was speculated that this was because the adsorbed H2 lowers the work function of the palladium. Earlier experiments testing the change in resistan ce of discontinuous Pd films upon exposure to hydrogen claimed an increase in the palladium work function.66,67,68 As pointed out by Barr66 the relevant mechanism is that hydrogen adsorption lowers the Fermi level of the palladium, thus causing a increased Scho ttky barrier for hole transport. Hydrogen detection is important in a va riety of areas including semiconductor fabrication clean rooms, space missions where H2 is used as fuel and wherever H2 could cause an explosive mixture. The current efforts toward a hydrogen based fuel economy

PAGE 82

69 further emphasize the need to find dependable low power H2 sensors. The goal of this work was to produce a reliable, easy to fabricate, H2 sensor made from carbon nanotubes. 6.2 Carbon Nanotube Sensor Fabrication We fabricated two types of SWCNT sensors, both coated with a thin layer of Pd metal. The first involved tens of CV D grown nanotubes wire d between parallel electrodes spaced roughly 1 micron apart. The second type involved ul tra thin nanotube films made from pulsed laser vapo rization bulk produced SWCNTs The first type, from hereon called the micr o-device sensor, was fabricated using p type <100> silicon chips with a 600nm therma l oxide layer. Using the same techniques discussed in Chapter 4, the chips were cleaned, nanotubes grown by CVD, and photoligthography used to form an interdigitat ed14 electrode pattern. The lengths of the electrodes are 500 m so there are many nanotubes in parallel between two adjacent electrodes. The pattern is shown in Figure 6-1. Figure 6-1: Electrode pattern on micro-device nanotube sensor. The ultra thin nanotube film sensors were manufactured using a technique described by Wu et al. .69 Purified SWCNTs grown by la ser ablation were suspended in an aqueous 0.5% Triton-X solution with a nanotube density of 0.001 mg/mL. The nanotube film was formed on a membrane by vacuum filtering the solution onto a mixed

PAGE 83

70 cellulose ester membrane (0.1 m pore size, Millipore). After the film was formed the Triton-X was washed away with DI water. The film thickness was easily controlled by the amount of the nanotube suspension filtered. To transfer the nanotube film to the desired substrate the film was adhered to the substrate by wetting with water and pressing between metal plates until the water dried. The mixed cellulose ester membrane was then dissolved away by acetone leavi ng the ultra thin nanotube fi lm behind. Films made with this technique can be as thin as a few nanometers. The nanotube thin film sensors were fabri cated on cleaned silicon substrates with a 600 nm SiO2 layer. Nanotube films with nomin al thickness of 7nm and 25nm were used are shown in Figure 6-2. Step height s of thicker films (50 nanometers) were measured by AFM and correlated to the amount of nanotube solution used. The 7nm and 25nm heights of the films are estimates f ound by scaling down the amount of solution. As seen in Figure 6-2b, the 7nm film effectively sub-monolayer. Figure 6-2: AFM images of a) 25nm SWCNT film and b) 7nm SWCNT film The films were deposited on the silicon chips as follows. After washing the nanotube film on the membrane it was allowed to dry. Pieces, of the ultimately desired

PAGE 84

71 film size were then cut from the membrane wetted with DI water and placed nanotube side down onto the silicon chip. The substr ate was sandwiched betw een several pieces of filter paper for cushioning and pressed by spring clamps between two metal plates. After the water evaporated, aided by placing the samples in a 95 C oven for an hour, the nanotube film/filter membrane was attached to the substrate. To remove the mixed cellulose ester filter, the silicon chips were transferred to an acet one vapor bath. The membranes dissolved in the acetone vapors le aving only the nanotube film. The samples subsequently went through five additional a cetone baths followed by a methanol wash to ensure the complete removal of the membrane polymer. The samples were baked in an inert gas to desorb remaining contaminants and nanotube dopants. To give volatiles a chance to escape before they re acted with the nanotubes the oven was ramped at a slow rate under the following schedule: 5 C/min to 110 C, held for 30 min, then 1 C/min to 600 C, held at temperature for 2 hours and and powered off to cool to room temperature. Contact pads of Cr/Pd were deposited acro ss the ends of the SW CNT film by either sputtering or thermal evaporation. A photogra ph of the SWCNT film sensor device is shown in Figure 6-3. Figure 6-3: SWCNT thin film sensor wire d for measurement. Br ight regions labeled source and drain are Pd metal pads. Volta ge is applied across the source and drain while the current, passed thr ough the SWNT film, is measured

PAGE 85

72 6.3 Resistance Changes from Metal Deposition As will be discussed in the next sect ion, both the micro-device and the film sensors show miniscule response to H2 at this stage. To be sensitive to the gas we deposited thin, in some cases non-percolat ing, layers of palladium on top of the nanotubes by either sputtering or thermal ev aporation. The samples were masked such that Pd was deposited only onto the nanotube f ilms. Surprisingly, the resistance of all the nanotube samples coated with sputtered Pd in creased. As this situation is analogous to two parallel conductance paths (the nanotube film and the metal layer) the increase in resistance is contrary to expectations, whic h would have the resistance decreasing. Table 6-1 shows the increase in the resistances of three samples, a micro-device sample, a 7nm film and a 25nm film, (AJA Internationa l ST10 magnetron source, 3 W power, 5 mT Ar, 5 s deposition time). 600 nm oxide-on-sili con witness chips, having the same source/drain electrodes but no nanotube film s rode along with each nanotube sample during the thin Pd layer deposition. These show ed the Pd films to be sub-percolation in all these cases. Table 6-1: Resistance changes of nanotube samples with sputtered Pd Sample Rbefore Rafter Rafter/Rbefore Micro-device 1.33 k 5.32 k 4.0 7nm film 3.18 k 6.69 k 2.1 25nm film 829 1.34 k 1.6 The resistance of several SWCNT films was monitored before and after the deposition of thermally evaporated Pd incorpor ating similar witness chips. In several of these cases the witness chips exhibited finite resistances demonstrati ng that the amount of metal deposited was above the percolati on threshold. This point and the resulting resistance will of course depend on the s ubstrate upon which the metal is deposited

PAGE 86

73 (nanotubes for the films and SiO2 for the chips), nevertheless to first order, we take the resistance of the metal layer measured on the witness chips to be that of the metal layer on the nanotubes. Table 6-2 lists the before and after resistance of several nanotube film samples with varying amounts of thermally eva porated Pd, as well as the resistan ce of the witness chips. In each instance the resist ance of the nanotube films decreased however the decrease varied with the amount of Pd. Films 1 and 4 had roughly a 5nm thick Pd film deposited on top (measured in situ by an Inficon quartz crys tal thickness monitor). The resistance of the Pd layer on top of Film 1 is smaller th an the resistance of Film 1 itself and is comparable in height. The final resistance is greater than the value calculated by combing the resistance of the nanot ube and the Pd film in parallel (R||=1.24 k compared to Rafter=1.72 k ). Similarly the final resistance of Film 4 is also larger than the calculated parallel resistance (R||=0.785 k compared to Rafter=0.993 k ). Films 2 and 3 had Pd layers between 2nm and 2.5nm deposited. The final resistance of these films is smaller than the calcu lated parallel resistance. Table 6-2: Resistance changes on SWCNT films with thermally evaporated Pd Sample Nanotube film height Rbefore Rafter RPd film Rafter/Rbefore 1 7nm 5.72 k 1.72 k 1.58 k 0.30 2 7nm 8.86 k 3.45 k 90.0 k 0.39 3 7nm 5.76 k 4.82 k non-percolating 0.84 4 25nm 1.46 k 0.993 k 1.70 k 0.69 Both metallic and semiconducting nanotube s have been shown to be ballistic conductors with resistances per nanotube approaching the theore tical limit of 6.5 k .70,54 However, the resistance acr oss nanotube films between macroscopically spaced electrodes is dominated by the resistance acro ss tube to tube contacts, which constitute tunnel junctions.71,72,73 The data shows that when less metal is evaporated on the

PAGE 87

74 nanotube films the resistance across the films is lowered by an amount greater than consideration of the parallel (nanotube/metal) conductance while when more is evaporated the resistance (while still lowe red) is lowered by an amount less than the parallel conductance. Give n the ballistic on tube conduc tance a simple view would suggest that by increasing the electrical contact area across tube-tube contacts the resistance across the films should be greatly lowered. We can make sense of the data by recognizing that it is nave to assume that a singled walled nanot ube associated with metal along its length retains its ballistic conduc tance. We conclude that the better than parallel path resistances of fi lms 2 and 3 are due to lowered tu be to tube resistance arising from increased contact area at the junctions due to the Pd particles which however are discontinuous and do not coat long sections of the nanotubes. Too much metal, which coats long sections of the nanotube sidewalls, interferes with their ballistic conductance. Both theoretical calculations and expe rimental evidence s upport this view.74,75 Regardless of the method of deposition the palladium coated the nanotubes. Figure 6-4 shows AFM images of two 7nm film s, one coated with sputtered Pd and the other by thermally evaporated Pd. The mor phological differences between the two films are likely due to the variation in the ra tes of deposition. Bo th films are ~10-30 Angstroms thick but the sputtered film was deposited in 10 seconds while the thermally evaporated film was grown in 80 seconds. Higher deposition fluxes result in smaller grain sizes and more even cove rage while slower rates allow the metal time to diffuse along the nanotubes and coalesce into the larger grains seen in Figure 6-4b.76

PAGE 88

75 Figure 6-4: AFM images of Pd coated nanot ube films a) sputtered and b) thermally evaporated To determine if the increase in nanot ube resistance after sp utter deposition was confined to Pd we also spu tter deposited a thin film of Au onto similar samples (in a different sputtering system), again obtaining an increased resistance. In light of these results we additionally conclude that sputtering damages the nanotubes changing their intrinsic conductance. The three sputtered sa mples in Table 6-1 had the Pd deposited at the same time so the amount of Pd was the same for each sample. The 25nm film had the smallest amount of resistance increase, a fact or of 1.6, while the micro-device resistance increased by a factor of four. Every nanot ube in the micro-device would be exposed to and damaged by the sputtering plasma causing th e largest change in resistance. In the thicker sample (25nm film) many of the unde rlying nanotubes would be shielded from the plasma by the tubes on top. Thus only a portion of the nanotubes would be damaged resulting in a smaller increase in resistance.

PAGE 89

76 6.4 Conductance Changes due to Hydrogen Exposure Resistance changes in the nanot ube devices upon exposure to H2 were monitored for several different samples. Data was take n for samples both with and without the thin Pd layer. We also compare the response of a 7nm film coated with thermally evaporated Pd to that of a sputter coated film. Additionally, the response to H2 of a thin Pd film without nanotubes was monitored. The samples were exposed to H2 at room temperature and atmospheric pressure in a quartz flow tube with electr ical feed-throughs for voltage a nd current leads. The gasses were fed via mass flow controllers to mainta in a total flow of 450 sccm of either pure nitrogen, 500 ppm hydrogen in nitrogen or a mixture of the two to obtain reduced concentrations of hydrogen, or compressed air humidified by pa ssing through a water bubbler. Electrical measurements were performed with an HP4156B source-meter. 6.4.1 Pure SWCNT Samples A micro-device sensor, a 7n m film and a 25nm film w ithout any additional metal besides the contact pads were al l tested for sensitivity to H2 (Figures 6-5, 6-6 and 6-7 respectively). The current was monitored as a f unction of time at constant bias (0.5V) for varying H2 concentrations. At t=14minutes the sa mples were exposed to air to recover their initial conductance. The pure nanotubes in the micro-device sens or show absolutely no sensitivity to H2. At the highest H2 concentration of 500ppm the unc oated 7nm and 25nm films show decreases in the current of 1.5% and 1.6% respectively. These comparatively small changes (compare below) are probably entire ly due to changes occurring within the nanotubes at the Pd/nanotube contacts as oppos ed to the bulk unc oated nanotube film.

PAGE 90

77 micro-device without Pd Figure 6-5: Current vs. time measurement for micro-device sensor exposed to different H2 concentrations. Figure 6-6: Current vs. time measuremen t for 7 nm film exposed to different H2 concentrations.

PAGE 91

78 Figure 6-7: Current vs. time measurement for 25nm sensor exposed to different H2 concentrations.. 6.4.2 SWCNT Samples Coated with Sputtered Pd The three sputter coated samp les listed in Table 6-1 were tested for their response to hydrogen (Figures 6-8, 6-9, and 6-10). Ag ain, the current vs. time for a 0.5 V bias was monitored during exposure to hydrogen. Th e samples were exposed to air at t=14 min to recover their nearly or iginal conductance. A summary of I/I at t=14 min and 500 ppm H2 concentration is listed in Table 6-3. The Pd coated sensors all show d ecreased conductance upon exposure to H2. The 7 nm film shows the highest sensitivity while the micro-device sensor shows the lowest. Optimization of the sensitivity would occur by finding the optimum ratio of Pd to SWCNTs. The micro-device sensor appears to have too much Pd for the amount of nanotubes while the 25 nm film pr obably doesnt have enough.

PAGE 92

79 micro-device w/ sputtered Pd Figure 6-8: Current vs. time for micro-device sensor w ith sputtered Pd for different H2 concentrations. Figure 6-9: Current vs. time for 7 nm film with sputtered Pd for different H2 concentrations

PAGE 93

80 Figure 6-10: Current for time for 25 nm film with sputtered Pd for different H2 concentrations. Table 6-3: Conductance changes in sputte red Pd/SWCNT samples at t=14 min for 500 ppm H2. Sample I/I (%) Micro-device 5.6 7 nm film 30. 25 nm film 25.9 6.4.3 SWCNT Samples Coated with Thermally Evaporated Pd A 7nm film with a non-percola ting layer of thermally evaporated Pd was tested for H2 sensitivity. The current vs. time graph for 500ppm H2 and 0.5V bias voltage is shown in Figure 6-11. The total change in resistance is 24%, which is smaller than the sputtered Pd/7nm film sample (30.4%). However, the response rate is much faster. Within one minute this sample shows a change in conductan ce of 12% (50% of the total change). At t=5 minutes the current has changed by 23% (9 6% of the total change). By comparison

PAGE 94

81 the sputtered 7nm film shows a I/I of only 1.2% (or 3.9% of the total change) at t=1 minute. After 5 minutes the change is 18.5% (60.9% of the total resistance change). The recovery rate is also much faster. One minute after exposure to air the thermally evaporated Pd/SWCNT sample recovers 77.4% of the H2 induced conductance change. After 5 minutes the recovery is 95%. The sputtered Pd/SWCNT sample only recovers 35.9% after 1 minute and 45.7% af ter 5 minutes. The differences in the total change between the sputtered and evaporated samples might be due to diffe rent amounts of Pd. The rapid response of the evaporated Pd /SWCNT film makes it a more promising candidate for real world applications. Optim izing the amount of Pd might increase its sensitivity to match the sputter Pd/SWCNT sensor. Figure 6-11: Current vs. time for 7nm film with thermally evaporated Pd exposed to 500ppm of H2.

PAGE 95

82 6.4.4 Thin Pd Film The current vs. time was monitored for a very thin layer of thermally evaporated Pd to 500ppm of H2, Figure 6-12. The initial re sistance of the film was 55.6 k The current first increased and th en decreased upon exposure to H2. After the sample was exposed to air the current increased again, w ithout recovery of the initial resistance. These results are very different than those exhibited by the Pd/SWCNT samples. From this we can conclude that the effects seen in the Pd/SWCNT samples are due to changes in the resistance of the nanotubes and are not s imply an effect cause by changes in the Pd resistance. Figure 6-12: Current versus time for a thin Pd film. 6.4.5 Conclusion Carbon nanotubes coated with Pd have been shown to be sensitive to H2 levels as low as 10ppm within 10 minutes exposure. A lthough nanotube samples with Pd deposited by

PAGE 96

83 both sputtering and thermal evaporation show sensitivity to H2, the latter shows a much faster response and recovery. It has been shown that nanotubes subjected to sputtered deposition of metals exhibit a decrease in thei r conductance, likely due to damage.

PAGE 97

84 APPENDIX SPECTROELECTROCHEMICAL STUDY OF CARBON NANOTUBE THIN FILMS Spectroelectrochemistry is a field of study which combines spectroscopy with electrochemistry. Spectroelectrochemistry can provide valuable information about thin films, such as the effects of different app lied voltages on the optical transmittance of the thin film. Aqueous electrolyte is commonly used in many electrochemical applications. However, water is not ideal in spectroelec trochemistry because it has strong absorption bands which make it almost impossible to gain information about a thin film at certain wavelengths. Additionally, water has a narro w potential window. Water can only be taken to -0.81V or +1.23V before it breaks down into its byproducts (H2 gas and OHions at negative voltage, O2 gas and H+ ions at positive voltage).77 To circumvent these types of problems, a nonaqueous solvent must be used. The experiment was performed using th e nonaqueous solvent acetonitrile. And involved a SWCNT thin film. In this experime nt potentials were applied to the thin film (versus a platinum wire counter electrode us ing a silver wire pse udoreference electrode) while changes in the transmittance spectrum were monitored. We infer that these changes in the spectrum occur as a consequence of changes in the Fermi level. From this inference, we are able to locate the in trinsic doping level of the semiconducting nanotubes. A chemically resistant cell and identical reference cell were developed for use in these experiments. One of these cells is disp layed in Figure A-1. The bodies of the cells are composed of Teflon. They use Simr iz SZ485 O-rings and have quartz optical

PAGE 98

85 windows. All of these components have high ch emical resistances to most substances. Nylon tubing was used for getting solutions into and out of the cells. Additionally, Teflon valves were used to control the fl ow of the solution, a nd a steel poppet valve was attached so that any gasses produced would be able to leave the cells without allowing atmospheric contaminants to enter. There were three types of electrodes used in the cell. The working electro de was the film under study co ntacted by a graphite rod, the counter electrode was platinum, and the refere nce electrode was silv er. A Perkin-Elmer {model} potentiostat supplied the potentials in a three terminal set-up, where in the counter electrode voltage is varied to make the voltage difference between the working electrode and reference electrode the desired value. The benefit of this is that the desired potential on the working electrode is main tained versus a well-defined reference independent of any Faradaic electron transf er occurring at the working electrode. The cells were designed in an attempt to keep out water vapor and air. The cells served their purpose well, and substantial impurity peaks were only observed after the cells were exposed to the environment for periods of over 24 hours. A.1 Carbon Nanotube Experiment When taking a transmittance spectrum of a thin film of SWCNTs one observes three broad peaks. Our study focuses on th e peak centered at 1650nm. This peak corresponds to photons absorbed during electronic transitions from the highest occupied semiconducting valence band (V1) to the lo west unoccupied semiconducting conduction band (C1). Figure A-2 shows that the valence and conduction states for carbon nanotubes have sharp peaks in the density of states known as Van Hove singularities. This would imply very sharp absorption bands. However, a thin film of SWCNTs is composed of

PAGE 99

86 semiconducting nanotubes of different diameter s and chiralities. As a result of this assortment of types of nanotubes, the peak s in the transmittance spectrum end up being broad and smooth. Figure A-1: The spectroelectrochemi cal cell used in the experiments. By applying a voltage to the SWCNT sample, one can shift the Fermi level. One can see in Figure A-2a that if the valence state (V1) becomes more depleted, then fewer electrons will be available to undergo the transition to the conduction state (C1). Therefore there will be less absorption of photon energy, and the absorbance peak will diminish. Similarly, if the Fermi level is increased, as in Figure A-2b, so that some electrons already occupy the conduction state (C1), then there will be fewer available states for the remaining valence state electr ons to move into and the peak will also diminish. Based on this knowledge, one can th en determine the intrinsic doping level of the semiconducting nanotubes. If the peak is largest at 0V then the semiconducting

PAGE 100

87 nanotubes will be intrinsically undoped. Using this method, one can only determine the amount of doping qualitatively. Figure A-2: Diagram of the Fermi Level of a semiconducting (12,8) SWCNT which is a) p-doped or b) n-doped A.2 Experimental Section This experiment is similar in nature to an experiment performed by Kavan et al ..78 Our experiment mainly differs from that one in that the SWCNT thin film that we use is mounted directly on quartz instead of ITO (Indium Oxide doped with Tin Oxide). We can do this because the nanotube samples also contain metallic nanotubes and we have learned how to make ultrathin (and hence tr ansparent) films of SWCNTs that are in contiguous electrical contact with each other.69 A SWCNT film was prepared and attached to a piece of quartz. This sample was baked in flowing argon gas in a Thermolyne 79300 tube furnace to drive off chemical dopants. The temperature was ramped at 5C per minute until it reached 110C, stayed at 110C for 30 minutes, and then ramped at 1C per minute to 600C. The temperature remained here for 2 hours. A reference piec e of quartz was also baked in the same way to be used in a reference cell.

PAGE 101

88 Once the two pieces of quartz were allowed to cool, they were removed from the tube furnace and attached to the graphite working electrodes inside the electrochemical cell and reference cell described in the introduction. Next, both cells were heated to 75C under flowing argon atmosphere for approxima tely 24 hours. Upon removal from the tube furnace, the cells were sealed by closi ng their Teflon valves. The cells were then placed inside of an argon glove box and allo wed to remain in there for 24 more hours with their valves open. This argon glove box had 1-2ppm oxygen. Inside the argon glove box, the solvent and el ectrolyte were added to the two cells. The solvent we used was Acetonitrile obtained from Fisher Scientific and the electrolyte was 0.5 M LiBF4. This solution was also placed in side glass tubes which held the reference and counter electrodes. Once the so lvent and electrolyte were added, the cells were resealed, and removed from the argon gl ove box. Potentials were applied with a Perkin-Elmer Model 283 Potentiostat, while transmission spectra were recorded with a Perkin-Elmer Lambda 900 UV/Vis/NIR spect rometer. With the reference cell (no nanotube film but including the quartz plat e, cell windows and electrolyte) in the reference beam of the spectrometer, and the sample cell in the sample beam, the absorbance of the sample cell itself (window s, electrolyte, quartz substrate) could be normalized out, leaving only the sp ectrum of the nanotube film. A.3 Results and Discussion Voltage was only applied to the cell with the nanotube thin film. We applied voltages ranging from -1.0V to +0.8V and reco rded their spectra. The spectra observed at positive voltages are displayed in Figure A3, and the spectra at negative voltages are displayed in Figure A-4. As can be seen in Figure A-3, the transmittance peak is smaller at +0.2V than it is at 0.0V. However, one can see in Figure A-4 that the peak size is the

PAGE 102

89 greatest at -0.4V. After +0.2V and -0.6V th e transmittance peaks change dramatically with a change in voltage. One problem encountered with the two cells was that they were not exactly the same thickness. We could only minimize the di fference to within 0.2mm. This variance caused some acetonitrile peaks to show up in the spectra. Since these peaks were the same in all spectra, they were removed and the figures were smoothed prior to presentation in Figures. A-3 and A-4. A.4 Conclusions We determined that the maximum size of the transmittance peak occurs at -0.4V. At this voltage, the Fermi level is in the ba nd gap between the first valence state and first conduction state of the most semiconducting nanot ubes. This conclusion indicates that at 0.0V, the Fermi level has been shifted toward the positive side. Therefore the nanotube film is intrinsically slightly p-doped. Figure A-3 Percent Transmittance vs. Wavelength for SWCNT thin-film at various positive applied potentials.

PAGE 103

90 Figure A-4: Percent Transmittance vs. Wavelength for SWCNT thin-film at various negative applied potentials.

PAGE 104

91 LIST OF REFERENCES 1. S. Iijima, Nature 354 56 (1991) 2. D.S. Bethune, C.H. Kiang, M.S. de Vries, G. Gorman, R. Savoy, J. Vazques. and R. Beyers,. Nature 363 605 (1993) 3. S. Iijima and T. Ichihashi, Nature 363 603 (1993) 4. L.X. Zheng, M.J. O’Connel, S.K. Door n, X.Z. Liao, Y.H. Zhao, E.A. Akhadov, M.A. Hoffbauer, B.J. Roop, Q.X. Jia, R.C. Dye, D.E. Peterson, S.M. Huang, J. Liu, and Y.T. Zhu, Nature Mat. 3 673 (2004) 5. M.M.J. Treacy, T.W. Ebbesen, and J.M. Gibson, Nature 381 678 (1996) 6. H.D. Wagner, O. Lourie, Y. Feldma n, and R. Tenne, Appl. Phys. Lett. 72 188 (1998) 7. D.A. Walters, L.M. Ericson, M.J. Casavant J. Liu, D.T. Colbert, K.A. Smith, and R.E. Smalley, Appl. Phys. Lett. 74 3803 (1999) 8. R. Baughman, C. Cui, A. Zakidov, Iqbal, J. Bari sci, G. Spinks, G. Wallace, A. Mazzoldi, D. De Rossi, A. Rinzler, O. Jasc hinski, S. Roth, and M. Kertesz, Science 284 1340 (1999) 9. A. Thess, R. Lee, P. Nikolaev, H.J. Dai, P. Petit, J. Robert, C.H. Xu, Y.H. Lee, S.G. Kim, A.G. Rinzler, D.T. Colbert, G.E. Scuseria, D. Tomanek, J.E. Fischer, and R.E. Smalley, Science 280 483 (1996) 10. Y. Li, W. Kim, Y. Zhang, M. Rolandi, D. Wang, and H. Dai, J. Phys. Chem. B. 105 11424 (2001) 11. N.R. Franklin, Y. Li, R.J. Chen, A. Javey, and H. Dai, Appl. Phys. Lett. 79 4571 (2001) 12. P. Nikolaev, M.J. Bronikowski, R.K. Br adley, F. Rohmund, D.T. Colbert, K.A. Smith, and R.E. Smalley, Chem. Phys. Lett. 313 91 (1999) 13. M.S. Dresselhaus, G. Dresselhaus, and R. Saito, Phys. Rev. B. 45 6234 (1992) 14. J.W. Mintmire, B.I. Dunlap, and C.T. White, Phys. Rev. Lett. 68 631 (1992) 15. N. Hamada, S. Sawada, and A. Oshiyama, Phys. Rev. Lett. 68 1579 (1992)

PAGE 105

92 16. J.W.G Wildoer, L.C. Venema, C. Dekker, A.G. Rinzler, and R.E. Smalley, Nature 391 59 (1998) 17. T. Odom, P. Kim, J.-L. Huang, and Ch. Lieber, Nature 391 62 (1998) 18. R. Saito, G. Dresselhaus, and M.S. Dresselhaus, Physics Properties of Carbon Nanotubes (Imperial College Press, London 1998) 19. D. Nixon and G. Parry, J Phys C 2 1732 (1969) 20. M.E. Preil, J.E. Fischer, S.B. Dicen zo, and G. K. Wertheim, Phys. Rev. B 30 3536 (1984) 21. J.E. Fischer, H.J. Kim, and V.B. Cajipe, Phys. Rev. B 36 4449 (1987) 22. F. Baron, S. Flandrois, C. Hauw, and J. Gaultier, Solid State Commun. 42 759 (1982) 23. R.S. Markiewicz, J.S. Kasper, and V. Interrante, Synth. Met. 2 363 (1980) 24. Y. Murakami, T. Kishimoto, H. Suematsu, J. Phys. Soc. Jpn. 59 571 (1990) 25. L. Pietronero and S. Strassler, Phys. Rev. Lett. 47 593 (1981) 26. M. Kertesz, Mol.Cryst. Liq. Cryst. 125 103 (1983) 27. M. Kertesz, F. Vonderviszt, and R. Hoffman, Intercalated Graphite (Elsevier, New York, 1983) p. 141 28. C. Chan, W. Kamitakahara, K. Ho, and P. Eklund, Phys. Rev. Lett. 58 1528 (1987) 29. Y.N. Gartstein, A.A Zakhidov, and R.H. Baughman, Phys. Rev. Lett. 89 045503 (2002) 30. Y.N. Gartstein, A.A. Zakhidov a nd R.H. Baughman, Phys. Rev. B 68 115415 (2003) 31. M. Verissimo-Alves, B. Koiller, H. Chacham, and R.B. Capaz, Phys. Rev. B 67 121401 (2003) 32. G. Sun, J. Krti, M. Kertesz, and R. Baughman, J. Am. Chem. Soc. 124 15076 (2002) 33. G. Sun, J. Krti, M. Kertesz, and R. Baughman, J. Phys. Chem. B 107 6924 (2003)) 34. A. G. Rinzler, J. Liu, H. Dai, P. Nikol aev, C.B. Huffman, F.J. Rodrigues-Macia, P.J. Boul, A.H. Lu, D. Heymann, D.T. Colber t, R.S. Lee, J.E. Fischer, A.M. Roa, P.C. Eklund, and R.E. Smalley, Appl. Phys. A 67, 29 (1998)

PAGE 106

93 35. J. Bahr, E. Mickleson, M. Broniko wski, R. Smalley, and J. Tour, Chem Commun., 193 (2001) 36. J. Liu, A.G. Rinzler, H. Dai, J. Hafner, R.K. Bradley, P.J. Boul, A. Lu, T. Iverson, K. Shelimove, C.B.Huffman, F. Rodrigue z-Macias, Y.S. Shon, T.R. Lee, D.T. Colbert, and R.E. Smalley, Science 280 1253 (1998) 37. M.F. Islam, E. Rojas, D.M. Bergey, A.T. Johnson, and A.G. Yodh, Nano Lett. 3 269 (2003) 38. J. Hafner, C. Cheung, T. Oosterkamp and C. Lieber, J. Phys. Chem. B 105 743 (2001) 39. Y. Li, W. Kim, Y. Zhang, M. Rolandi, D. Wang, and H. Dai, J. Phys. Chem. B 105 11424 (2001) 40. A. Patil, J. Sippel, G.W. Martin, and A.G. Rinzler, Nano Lett. 4 303 (2004) 41. M. Tortonese and M. Kirk, SPIE Micromachining and Imaging 3009 (1997) 42. M. Shim, A. Javey, N.W.S. Kam, and H. Dai, J. Am. Chem. Soc. 123 11512 (2001) 43. M. Krger, M.R. Buitelaar, T. Nussba umer, and C. Schnenberger, Appl. Phys. Lett. 78 1291 (2001) 44. S. Rosenblatt, Y. Yaish, J. Park, J. Gore, V. Sazonova, and P. McEuen, Nano Lett., 2 869 (2002) 45. E. Gileadi and M. Urbakh, Thermodynamics and Electrified Interfaces (WileyVCH, 2002) 46. J.B. Hasted, D.M. Ritson, and C.H. Collie, J. Chem. Phys. 16 1 (1948) 47. F. Booth, J. Chem. Phys. 19 391 (1951) 48. R. Pottel, Water, A Comprehensive Treatise edited by F. Franks (Plenum Press, New York, 1973) Chapter 8. 49. A. Kemper, H.H.P. Gommans and A.G. Rinzler, unpublished result 50. S. Tans, A. Verschueren, and C. Dekker, Nature 393 49 (1998) 51. R. Martel, T. Schmidt, H. Shea, T. Hertel, and Ph. Avouris, Appl. Phys. Lett 73 2447 (1998) 52. C. Zhou, J. Kong, and H. Dai, Appl. Phys. Lett 76 1597 (2001) 53. R. Martel, V. Derycke, C. Lavoie, J. Appe nzeller, K. K. Chan, J. Tersoff, and Ph. Avouris, Phys. Rev. Lett 87 256805 (2001)

PAGE 107

94 54. A. Javey, J. Guo, Q. Wang, M. Lundstrom and H. Dai, Nature 424 654 (2003) 55. J.-P. Radin and E. Yeager, Electroana l. Chem. Interfacial Electrochem. 36 257 (1972) 56. H. Gerischer, R. McIntyre, D. Sche rson, and W. Storck, J. Phys. Chem. 91 1930 (1987) 57. Y. Oren, I. Glatt, A. Livnat, O. Kafr i, and A. Soffer, J. Electroanal. Chem. 187 59 (1985) 58. J. Kong, N.R. Franklin, C. Zhou, M.G. Ch apline, S. Peng, K. Cho, and H. Dai, Science 287 622 (2000) 59. P.G. Collins, K. Bradley, M. Ishigami, and A. Zettl, Science 287 1801(2000). 60. J. Kong and H. Dai, J. Phys. Chem. B 105 2890 (2001) 61. G.U. Sumanasekera, B.K. Pradhan, H.E. Romero, K.W. Adu, P.C. Eklund, Phys. Rev. Lett. 89 166801 (2002) 62. K. Bradley, J-C P. Gabriel, A. Star, and G. Grner, Appl. Phys. Lett. 83, 3821 (2003) 63. J.P Novak, E.S. Snow, E.J. Houser, D. Park, J.L. Stepnowski, and R.A. McGill, Appl. Phys. Lett. 83 4026 (2003) 64. Y. Lu, J. Li, J. Han, H.-T. Ng, C. Bind er, C. Partridge and M. Meyyappan, Chem. Phys. Lett 391 344 (2004) 65. I. Sayago, E. Terrado, E. Lafuente, M.C. Horrillo, W.K. Maser, A.M. Benito, R. Navarro, E.P. Urriolabeitia, M.T. Ma rtinez and J. Gutierrez, Synth. Met. 148 15 (2005) 66. A. Barr, Thin Solid Films 41 217 (1977) 67. F. Wu and J.E. Morris, Thin Solid Films 246 17 (1994) 68. J.E. Morris, A. Kiesow, M. Hong and F. Wu, Int. J. Electronics 81 441 (1996) 69. Z. Wu, Z. Chen, X. Du, J.M. Logan, J. Sippel, M. Nikolou, K. Kamaras, J.R. Reynolds, D.B. Tanner, A.F. Hebard and A.G. Rinzler, Science 305 ,1273 (2004) 70. W. Liang, M. Bockrath, D. Bozovic, J. Hafner, M. Tinkham, and H. Park, Nature 411 665 (2001)

PAGE 108

95 71. M.S. Fuhrer, J. Nygard, L. Shih, M. Fo rero, Y.-G. Yoon, M.S.C. Mazzoni, H. J. Choi, J. Ihm, S. G. Louie, A. Zettl, and P.L. McEuen, Science 288 494 (2000)H.W.Ch. Postma, M. de Jonge, Z. Yao, and C. Dekker, Phys. Rev. B. 62 R10653 (2000) J.W. Park, J. Kim and K.-H. Yoo, J. Appl. Phys. 93 4191 (2003) 74. S. Dag, O. Glseren, S. Ciraci, a nd T. Yildirim, Appl. Phys. Lett. 83 3180 (2003) 75. J. Nygard, D.H. Cobden, M. Bockrath, P.L. McEuen, and P.E. Lindelof, Appl. Phys. A. 69 297 (1999) 76. R. Anton, Thin Metal Films and Gas Chemisorption edited by P. Wissmann (Elsevier, New York, 1987) Chap. 1, p. 3. 77. K. Izutsu. Electrochemistry in Nonaqueous Solutions (Wiley-Vich, Weinheim, 2002) 78. L. Kavan, P. Rapta, L. Dunsch, M.J. Br onikowski, P. Willis, and R.E. Smalley, J. Phys Chem. B. 105 10764(2001)

PAGE 109

96 BIOGRAPHICAL SKETCH Jennifer was born May 28th, 1977 in Spencer, Iowa to Rick and Cindy Sippel. Jennifer developed an interest in science early on and was greatly influenced by her father’s interest in astronomy. She has al ways been curious about understanding how the universe, and everything in it, works. This led her to the University of Iowa after graduating from Ames High School. At the University of Iowa she earned a double major B.S. in both physics and astronomy. As an undergraduate she studied eclipsing binary star systems under Dr. Robert Mutel a nd also worked from him as an operator of the university’s Automated Telescope Facility. She also did research for Dr. Steven Spangler and operated a 4.5m radio telescope under his tutelage. In between her junior and senior years in college Jennifer participated in the Walt Disney World college program. She took a brea k from research for a summer to work at the Magic Kingdom in Orlando, Florida telling bad jokes on the Jungle Cruise attraction. Jennifer immensely enjoyed this experience an d it prompted her to look into coming to Florida for graduate school. Jennifer started her graduate program in Physics at the Un iversity of Florida in the fall of 1999. She began working for Dr. A ndrew Rinzler during the summer of 2000 studying carbon nanotubes. She has enjoyed wo rking with these fascinating materials and learning many instrumentation skills during the course of her research.

PAGE 110

97 While at the University of Florida Jennifer met Garrett Oakley. He was a fellow graduate student but in the field of Chem istry. They were married on the beach in October of 2004.


Permanent Link: http://ufdc.ufl.edu/UFE0010052/00001

Material Information

Title: Charge Induced Actuation in Carbon Nanotubes and Resistance Changes in Carbon Nanotube Networks
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0010052:00001

Permanent Link: http://ufdc.ufl.edu/UFE0010052/00001

Material Information

Title: Charge Induced Actuation in Carbon Nanotubes and Resistance Changes in Carbon Nanotube Networks
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0010052:00001


This item has the following downloads:


Full Text












CHARGE-INDUCED ACTUATION IN CARBON NANOTUBES AND RESISTANCE
CHANGES IN CARBON NANOTUBE NETWORKS















By

JENNIFER ANN SIPPEL-OAKLEY


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2005

































Copyright 2005

by

Jennifer Ann Sippel-Oakley

































To my parents, Richard and Cynthia Sippel
















ACKNOWLEDGMENTS

I would like to extend my deepest gratitude to my advisor, Dr. Andrew Rinzler. He

has an amazing amount of enthusiasm for the pursuit of knowledge in the field of carbon

nanotubes. He was always encouraging throughout the course of a frustrating and

difficult experiment. As a teacher, he was patient and thorough. For these things I am

extremely grateful. His support and encouragement are the reason I am receiving this

degree. If I could make the same choice over again I would not hesitate to pick him as

my advisor.

I would also particularly like to thank former group members Dr. Zhihong Chen

and Dr. Amol Patil for close collaboration and friendship during our graduate careers.

Addtionaly I am grateful to current and former group members Dr. Hidenori Tashiro,

Jacob Alldredge, Zhuangchun Wu, Lex Kemper, Daniel Ranken, Daniel Barrow and

Jonathon Logan.

I owe thanks to Dr. Art Hebard for use of laboratory equipment and also for always

being a friendly and helpful resource. Thanks are due to Hung-Ta Wang, Byoung Sam

Kang, Dr. Fan Ren and Dr. Steve Pearton for collaboration on the nanotube hydrogen

sensor.

Many thanks go to my parents, Rick and Cindy Sippel, who have always supported

me in all aspects of my life. I am eternally grateful for that and for their unconditional

love.









Finally I would like to thank my husband, Garrett Oakley. He is a fellow scientist

and a great dancer. His love and support have guided and uplifted me during my

graduate career.
















TABLE OF CONTENTS

page


A C K N O W L E D G M E N T S ................................................................................................. iv

LIST OF TA BLE S .................. .................................. .... .. ........ ........ ....... viii

LIST OF FIGURES ......... ......................... ...... ........ ............ ix

ABSTRACT ........ ........................... .. ...... .......... .......... xii

CHAPTER

1 IN T R O D U C T IO N ........................................................................ .......... .. .. ... 1

2 CARBON NANOTUBE BACKGROUND..........................................................3

2.1 Growth Methods of Carbon Nanotubes..........................................................3
2.2 Carbon N anotube Structure................................................... ..................4

3 MOTIVATION AND THEORY FOR INDIVIDUAL CARBON NANOTUBE
A CTU A TION ........................................................... .. ... ......... 11

3.1 Electrom echanical A ctuation .................................... .................................... 11
3.2 M acro-Scale Carbon Nanotube Actuators.....................................................11
3.3 Bond Length Changes in Intercalated Graphite...........................................13
3.4 Theoretical Work for Bond Length Changes in Carbon Nanotubes ..............17

4 PREPARATION OF AND EXPERIMENTAL MEASUREMENTS ON
SUSPENDED CARBON NANOTUBE SAMPLES...........................................22

4.1 Fabrication of Suspended Carbon Nanotube Structures.............................22
4.1.1 Substrate Preparation .................... .... ................. ............... .... 23
4.1.2 Carbon Nanotube Deposition and Growth....................................24
4.1.2.1 Deposition of Laser Ablation Grown Nanotubes from
Solution ............................. .... .................. ............. 24
4.1.2.2 Chemical Vapor Deposition Grown Nanotubes ................27
4.1.3 Electrode Patterning...................................... ........................ 29
5.1.4 E thing ................................................. ...................... ...... 33
5.1.5 R release Procedure ...................................... .................. .... ........... 34









4.1.6 Identification of Suspended Nanotubes ........................................ 37
5.1.7 Final Sam ple Preparations ..................................... ............... ..38
4.2 Experimental Procedure ............... ........... .... ............... 39

5 RESULTS AND DISCUSSION OF CHARGE INDUCED ACTUATION OF
SUSPENDED CARBON NANOTUBES............................................................51

5.1 Results of Actuation M easurements ................ ................ ....................51
5.2 D discussion of R esults................ .......................... .................. ............... 58

6 CONDUCTANCE CHANGES IN CARBON NANOTUBES DUE TO
H YD R O GEN ................................................................ .... ........66

6.1 Carbon Nanotube Sensor Background.........................................................66
6.2 Carbon Nanotube Sensor Fabrication............................................................69
6.3 Resistance Changes from Metal Deposition.................... ..................72
6.4 Conductance Changes due to Hydrogen Exposure........................................76
6.4.1 Pure SW CN T Sam ples.................. ................... ............... .... 76
6.4.2 SWCNT Samples Coated with Sputtered Pd ....................................78
6.4.3 SWCNT Samples Coated with Thermally Evaporated Pd................80
6.4.4 Thin Pd Film ........................... ............. .................. ......82
6.4.5 Conclusion ............................ ..... ... ...... .. .... ............... 82

APPENDIX SPECTROELECTROCHEMICAL STUDY OF CARBON NANOTUBE
THIN FILM S .................. .......................... ...... ................... 84

A 1 Carbon N anotube Experim ent..................................... ......... ............... 85
A .2 Experim ental Section ......................................................... .............. 87
A .3 R results and D iscussion......................................................... ............... 88
A .4 Conclusions .................................... ............................... .......89

LIST OF REFEREN CES ............................................................................. 91

BIOGRAPH ICAL SKETCH ...................................................... 96
















LIST OF TABLES


Table pge

4-1 Resistance before and after processing steps ................................. ............... 36

5-1 Sum m ary of data from sam ple 1 ........................................ ......................... 53

5-2 Sum m ary of data for sam ple 2. ........................................ .......................... 55

5-3 Sum m ary of all data from sam ple 3 .............................................. ............... 56

6-4 Sum m ary of the actuator data ........ ................. ..................................................58

6-1 Resistance changes of nanotube samples with sputtered Pd.............. .....................72

6-2 Resistance changes on SWCNT films with thermally evaporated Pd ...................73

6-3 Conductance changes in sputtered Pd/SWCNT ..................................................80
















LIST OF FIGURES

Figure page

2-1 Hexagonal graphite lattice showing the unit vectors ai and a2..............................7

2-2 U nit cells of graphite ........ ........ ..................................... .................. ............ .... ... .8

3-1 M acro-scale actuator. ......................................... ................... .. .... .. 12

3-2 Strain values for bucky paper actuators films versus applied potential .................13

3-3 Strain versus charge transfer curve for graphite.................. ................................15

4-1 A ctuation experim ental setup ........................................................ ............... 23

4-2 Aligned carbon nanotubes from Triton-X solution...............................................26

4-3 N anotubes on silicon.. .................................... ................. .... ....... 28

4-4 M etal deposition on resists. ........................................ .......................................29

4-5 Scope trace of electrodes ............................................................................ ... .... 30

4-6 N anotubes after lithography .............................................................................. 32

4-7 Processing steps for suspended carbon nanotube samples .....................................35

4-8 Suspended carbon nanotube im ages..................................... ........................ 37

4-9 M ap of the created m arks. ........................................................................ 38

4-10 Cross sectional schematic of chip mounted on AFM puck ............................... 40

4-11 Free-space drift of a bare tip and a paralyne coated tip in aqueous 1M NaNO3 ......43

4-12 Identifying mark created in the SEM ................................... ......... ...... ......... 43

4-13 Force calibration curves against two surfaces.......... .....................................47

4-14 Illustration of a force calibration curve against a hard surface and nanotube..........48

4-15 N anotube under tension ........................................................................ 49









4-16 Plot of dy versus y forLo=430nm and 6L/L=0.0001 ..................... ................ 50

5-1 N anotube sam ple 1 ............................................. ............... ...........52

5-2 Sam ple 1 data. ........................................................................52

5-3 N anotube sam ple 2. ......................................... ... .... ..................54

5-4 S am p le 2 d ata ...................................................................... 54

5-5 N anotube sam ple 3 ............ ............... ....................55

5 -6 S am p le 3 d ata ...................................................................... 5 6

5-7 Data of the Z Movement and applied voltage while the AFM was suspended in
free space ............................................................... .... ..... ........ 57

5-8 Data of Z Movement and applied voltage while the AFM tip was in contact with
the bottom of the trench. ................................................................. ..................... 57

5-9 H ydrated C and N a ions.................................. ........................ ............... 60

5-10 Density of states for a (10,11) SWCNT. ........................................ ...............63

6-1 Electrode pattern on micro-device nanotube sensor. .............................................69

6-2 N anotube thin film s ....................................................... ........ .......70

6-3 SWCNT thin film sensor wired for measurement.................................................71

6-4 P d coated nanotube fi lm s .............................................................. .....................75

6-5 Micro-sensor device current vs. time measurement..........................................77

6-6 7nm film current vs. time measurement................................ ............... 77

6-7 25nm film current vs. time measurements.. .....................................................78

6-8 Micro-device sensor coated with sputtered Pd current vs. time measurements.......79

6-9 7nm film with sputtered Pd current vs. time measurements ..................................79

6-10 25nm film with sputtered Pd current vs. time measurements. ................................80

6-11 7nm film with thermally evaporated Pd current vs. time measurement.................. 81

6-12 Thin Pd film current vs. time measurements ................ .................................82

A -1 Spectroelectrochem ical cell........................................................... ............... 86









A-2 Fermi Level of a semiconducting (12,8) SWCNT ..............................................87

A-3 Percent Transmittance vs. Wavelength for SWCNT thin-film at various positive
ap p lied p otentials ................................................ ................ 89

A-4 Percent Transmittance vs. Wavelength for SWCNT thin-film at various negative
ap p lied p otentials ................................................ ................ 9 0















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

CHARGE-INDUCED ACTUATION OF SUSPENDED CARBON NANOTUBES AND
RESISTANCE CHANGES IN CARBON NANOTUBE NETWORKS

By

Jennifer A. Sippel-Oakley

May 2005

Chair: Andrew G. Rinzler
Major Department: Physics

In 1999 it was demonstrated that macroscopic films comprised of single wall

carbon nanotubes exhibited dimensional changes with charge injection onto the films. A

fundamental mechanism was proposed for this effect related to the dimensional changes

observed in graphite intercalation complexes upon charge transfer doping with the

intercalant species. The major fraction of this thesis concerns experiments at the single

nanotube level designed to test the validity of this mechanism. The metals compatible

with our fabrication processes inevitably p-dope the nanotubes resulting in smaller

dimensional changes. Additionally, there are contact barriers that prevent the injection of

electrons onto the nanotubes. Although the proposed mechanism may still be responsible

for the results seen in the nanotube films, the effect is too small to be consistently

measured in individual nanotubes.

The conductivity of a carbon nanotube can be varied by exposure to various

chemicals having utility in chemical sensing applications. We use thin films of carbon









nanotubes to exploit this effect. The films are made sensitive to hydrogen by association

with palladium metal. Such sensors operate at room temperature with very low power

dissipation of -0.25 mV.














CHAPTER 1
INTRODUCTION

Carbon nanotubes are quasi one-dimensional structures with the high mechanical

strength of graphite and have the useful attribute of occurring in either semiconducting or

metallic variants. These qualities have caused an explosion of research since their

discovery in the soot of fullerene production by Ijima1 in 1991. These first nanotubes

were multi-walled carbon nanotubes (MWCNTs) consisting of at least two graphene

cylinders nested one within the next.

This discovery prompted efforts to produce single wall carbon nanotubes

(SWCNTs) and prompted much theoretical effort on the properties of such tubes. The

addition of metal catalyst particles to the carbon starting material was found to be the key

and production of SWCNTs was first established by Bethune et al.2 and Ijima and

Ichihashi3 in 1993. All the nanotubes used in our work were SWCNTs.

In a standard description, a carbon nanotube is a graphene sheet rolled into a

seamless cylinder. As the name implies, the nanotubes have a diameter around a

nanometer. In general, they have a length of a few microns but there have been recent

reports of nanotubes with lengths of up to centimeters.4 Much of the fascination with

nanotubes arises from the fact that they can be either semiconducting or metallic

depending on the orientation the hexagonal graphene lattice relative to the nanotube axis.

Coupled with this property is the extremely high Young's modulus (1 TPa)5 and tensile

strength (estimated at 45 Gpa)6'7 of nanotubes. They are also stable at high temperatures

and show strong resistance to most chemicals.









Our study addresses two specific effects in carbon nanotubes that could lead to

useful applications. One proposed application for carbon nanotubes is electromechanical

actuators. In 1999, a carbon nanotube film showed dimensional changes with charge

injection.8 This behavior was attributed to an effect seen in intercalated graphite

compounds whereby the carbon-carbon bond length is modified by charge transfer to

graphite's 7t orbital system. The first part of this thesis describes efforts to observe

dimensional changes due to charge injection in an individual nanotube and determine if

the proposed mechanism was really responsible for the effect seen in the nanotube films.

Conductance of nanotubes can be manipulated by the interaction with other

molecular species. This led to the second focus of our study, resistance changes in

nanotube films as a way of hydrogen detection. Nanotube films were made sensitive to

hydrogen by the addition of palladium particles and H2 levels as low as 10 ppm were

detected.














CHAPTER 2
CARBON NANOTUBE BACKGROUND

2.1 Growth Methods of Carbon Nanotubes

As the interest in carbon nanotube research grew, so did the need for increased

quantities of high-quality nanotubes. Several methods were developed for their

production. The first method is arc-discharge or carbon arc. In this setup, two carbon

rods are used as electrodes with a small separation (about 1mm) and a high dc current is

passed between them while in a helium atmosphere. The high currents passing between

the carbon electrodes ignite plasma of the helium gas, and temperatures exceed 30000C.

Carbon is evaporated from the anode and then condenses to form nanotubes. This

method can be used to grow MWNT or SWNT if a metal catalyst (usually a transition

metal) is added to the carbon electrode. The high temperatures at which growth occurs

ensure high-quality nanotubes with few defects. However, many byproducts (such as

fullerenes) and amorphous carbons are produced and must be removed by purification.

The second method of nanotube growth is laser ablation (first demonstrated by

Thess et al.).9 This was the first method to produce nanotubes on the scale of several

grams per run. Early work in the actuation project and all work with nanotube sensors

used SWCNTs grown by this process. With this method, a pulsed laser is used to

vaporize a graphite target within a heated tube furnace (temp 1100-12000C). To achieve

SWNT growth metal catalyst material is mixed with the graphite target. A flow of inert

gas carries the nanotubes downstream to a cooled copper collector. This method also

requires purification to remove the byproducts generated during growth. Both the arc-









discharge method and laser ablation method form nanotubes that tend to be aggregated in

ropes (several nanotubes bundled together by van der Waals forces). Although vigorous

sonication of solutions containing these nanotubes can help break up the bundles, it is

difficult to isolate individual nanotubes using material from these growth processes.

This leads to the next method of growing nanotubes: chemical vapor deposition

(CVD). Most nanotubes used to measure nanotube actuation were grown by CVD. The

method involves placing a sample with metal catalysts into a furnace (growth

temperatures between 500-10000C) and flowing a carbon feedstock gas (usually a

hydrocarbon gas) over the sample. The presence of the metal catalysts causes the

dissociation of the hydrocarbon feedstock gas. The metal particles become supersaturated

with carbon and precipitate nanotubes. It has been found by transmission electron

microscopy (TEM) that the nanotube diameter corresponds to the diameter of the catalyst

particles.10 Catalyst islands can even be lithographically patterned onto silicon substrates

for more controlled placement of the nanotube growth.1

A fourth growth method (developed recently) involves decomposition of carbon

monoxide at high pressures (between 1 and 10 atm) and is commonly referred to as

HiPCO.12 Catalyst particles are provided by the thermal decomposition of iron

pentacarbonyl Fe(CO)5 which produces iron clusters in the gas phase. This process

produces tubes with smaller average diameters and less variation then tubes produced by

the first three methods.

2.2 Carbon Nanotube Structure

Carbon nanotubes were predicted to be either semiconducting or metallic in 1992

soon after their discovery.13,14,15 However, this was only verified experimentally 6 years









later by scanning tunneling microscopy (STM)16,17. Many techniques (such as STM,

TEM and scanning electron microscopy) have verified nanotube structure as that of a

seamless graphene-like cylinder. STM has proved a useful tool to study nanotubes since

it can reveal a nanotube's chirality.

This section gives some basic background on the physical and resulting electronic

structure of SWCNTs. The treatment here is a summary of that developed by Saito et al.

in Chapter 3 of reference 18. Nanotubes are most easily described by their chiral vector,

Ch.

Ch nal + ma2 (2-1)
The chiral vector denotes the circumference of the nanotube and connects two

crystallographically equivalent sites on the planar graphite sheet. In other words, a

nanotube can be envisioned as the cylinder formed from rolling up a sheet of graphene by

connecting the tip to the tail of the chiral vector. As shown in Figure 2-1, the vectors al

and a2 are two non-orthogonal unit vectors, the combination of which can span any point

on the hexagonal graphite lattice and have the following relations

a al = a2 a2 a2, al a2 = a2/3 (2-2)
where a=1.44A x 312 = ac-c x 312. a is the lattice constant of the graphite sheet and ac-c

is the carbon-carbon bond distance. In x and y coordinates, the real-space unit vectors

are represented by al=(31/2a/2, a/2) and a2=(31/2a/2, -a/2). Thus the diameter of the

nanotube is given by

dt = L/TT = (Ch Ch)1/2/7 = 31/2ac-c(m2 + mn + n2)1/2/ (2-3)

The translation vector T is orthogonal to Ch and parallel to the nanotube axis. It is the

shortest repeat distance along the axis. It is defined to be the unit vector of a nanotube

and is represented as









T = tiai + t2a2 (ti and t2 are integers) (2-4)

Since T and Ch are orthogonal by definition, we use Ch T = 0 to calculate ti and t2.

(nal + ma2) (tial + t2a2) = (2n + m)ti + (2m + n)t2 = 0 (2-5)
ti = (2m + n)/dR t2 = -(2n + m)/dR

where dR is the greatest common divisor of (2m + n) and (2n + m). Introducing d as the

greatest common divisor of n and m gives dR as

dR = d if n-m is not a multiple of 3d (2-6)
= 3d if n-m is a multiple of 3d

The unit cell of a nanotube is the rectangle defined by the vectors T and Ch. The number

of hexagons contained in this unit cell is determined by the area of the unit cell (|Ch x TI)

divided by the area of a hexagon (lai x a2l) and is given in terms of(n,m) by the equation

N = 2(m2 + nm + n2)/dR (2-7)

Each hexagon contains two carbon atoms, so the total number of carbon atoms in the

nanotube unit cell is thus 2N.

The construction of the nanotube along with its electrical properties is completely

determined by the indices (n,m). Nanotubes are grouped into three categories: zigzag,

armchair, and chiral. Both zigzag and armchair nanotubes have special symmetry

directions and earn their name from the pattern of the chain of carbon atoms around the

circumference. Zigzag nanotubes are those with one of the indices equal to zero, either

(n,0) or (0,m). Armchair nanotubes have identical indices, (n,n). They are both achiral,

meaning that the structure of the mirror image is identical to that of the original. Chiral

nanotubes are those that fall anywhere between the two types. The chiral angle 0

represents the tilt of the hexagons with respect to the nanotube axis. Figure 2-1 shows see

that 0 can be found by taking the dot product of Ch and al.









cos 0 = Ch al/|Chhal1 = (2n + m)/2(n2 + nm + m2)1/2 (2-8)

Zigzag nanotubes have 0 = 0, armchair have 0 = 300, and chiral fall anywhere in

between that range.




















Figure 2-1: Hexagonal graphite lattice showing the unit vectors al and a2. This image
shows the chiral vector, Ch, and translation vector T for a (6,3) nanotube

As mentioned, the indices (n,m) determine a nanotubes electrical properties. The

electronic structure of carbon nanotubes can be obtained by using the two-dimensional

structure of graphite and applying period boundary conditions along the circumferential

direction Ch. The unit structure of graphite contains two inequivalent carbon sites within

the hexagonal lattice called points A and B, Figure 2-2A. Real-space lattice vectors were

previously defined in x,y coordinates. In reciprocal space, the lattice vectors are

bl=(2t/31/2a, 27T/a) and b2=(2T/31/2a, -27T/a), Figure 2-4B.

When the first Brillouin zone is chosen as the shaded region in Figure 2-2B the

highest symmetry is obtained for graphite. In 2-D graphite, the three o bonds hybridize in

an sp2 configuration. The third 2pz orbital forms covalent 7t bonds. Generally, only the 7t








bands are considered when deriving the band structure of graphite. Saito et al. 18 used a
tight binding model that only considers nearest neighbor interactions to A and B, to get
the following dispersion relations for the 7t band of 2D graphite

E2Dgraph(k)=S2p t.o(k)/(l + sco(k)) (2-9)

(a) Y (b)

x b



r "' : :* M



a2 kx b2

Figure 2-2:Unit cells of graphite. A) The real space unit cell of graphite is defined by the
dotted line and contains points A and B. The unit vectors al and a2 are also
shown. B) the Brillouin zone in k space for graphite and the reciprocal lattice
vectors bl and b2.
The + signs in Equation (2-9) give the values for the bonding 7 band while, the signs

are for the anitbonding 7* band. The parameter g2p is the orbital energy for the 2p level, s
is the overlap integral between the nearest A and B atoms, and t is the transfer integral.
The function -o(k) has the form

o (k)= {1 + 4cos(3 2kxa/2) cos(kya/2) + 4cos2(kya/2)}1/2 (2-10)
To get a simple approximation for the electronic structure of graphene the overlap
integral s is usually taken to be zero with 62p also being set to zero. This yields the
following expression for the energy dispersion relations.

E2Dgraph(kx,ky)= t{ 1 + 4cos(31 2kxa/2) cos(kya/2) + 4cos2(kya/2)}1/2 (2-11)









Plotting the energy dispersion relations for graphite reveal that the bonding and anti-

bonding 7T bands meet at the K point; thus graphite is considered a zero-gap

semiconductor (the density of states is zero at the Fermi level).

Previously the chiral vector Ch and translation vector T for a carbon nanotube were

defined. The corresponding reciprocal lattice vectors K1 (in the circumferential direction)

and K2 (along the direction of the nanotube axis) are calculated by the expression Ri Ki

= 276ij (Ri are the lattice vectors in real-space corresponding to Ch and T) giving

Ch Ki = 27 T K1 = 0 (2-11)
Ch K2 = 0 T K2 = 27T

This leads to

Ki = (-t2bi + tib2)/N and K2 = (mbi nb2)/N (2-12)

Because of the translational symmetry of T, wave vectors in the direction of K2 are

continuous for an infinitely long nanotube. In the circumferential direction there are N

wave vectors [jKI (t = 0,1. N-1), which means N discrete k vectors. When the allowed

k vectors in the circumferential direction include the K point (where the valence band and

conduction band meet), the nanotube will be metallic. This occurs for the condition

(n-m)=3i (where i is an integer). Otherwise the nanotube will be semiconducting.

The armchair tubes (n,n) are all metals with a finite density of states at the Fermi

level. Chiral tubes with (n-m)=3i technically have a tiny gap that opens up at the Fermi

level due to tube curvature effects. However this gap is so small that at room temperature

these tubes can be considered metallic. The band gaps of semiconducting nanotubes are

inversely proportional to the nanotube diameter. The size of the band gap depends on

how close the allowed k values are the to special K point. The larger the diameter of a

nanotube the more k values there are. Thus the spacing between the k points decreases






10


and the values get closer to the K point resulting in a smaller band gap. The one-

dimensional nature of the nanotubes leads to peaks in the density of states known as van

Hove singularities. Semiconducting tubes that are slightly doped (as is often the case)

can have their Fermi levels shifted into the first Van Hove singularity. Because of this

these doped semiconducting nanotubes are often more conductive than the intrinsic

metallic nanotubes at room temperature because of the large number of carriers.














CHAPTER 3
MOTIVATION AND THEORY FOR INDIVIDUAL CARBON NANOTUBE
ACTUATION

3.1 Electromechanical Actuation

Electromechanical actuators convert electrical energy into mechanical work. The

fine motor movement possible with these devices is used in various fields such as laser

tuning, vibration cancellation, adaptive optics and micromanipulation. Piezoelectric

actuators are the most commonly used devices of this type (one example is the scanner

used in atomic force microscopes). Many actuators can only be operated in a limited

temperature range or have other physical limitation piezoelectricc actuators must be

operated below their Curie temperature and suffer from hysteresis). In 1999, Baughman

et al.8 reported an experiment in which they attempted to use carbon nanotubes (in bulk

form) as actuators. The chemical robustness and small dimensions of SWCNTs made

them seem like a promising addition to the repertoire of actuator materials. The intriguing

results of these macro-scale actuators led to our study aiming to observe actuation in

individual SWCNTs..

3.2 Macro-Scale Carbon Nanotube Actuators

Baughman et al.8 used a macro-scale composite of SWCNTs know as buckyy

paper" as a demonstration of their actuator. Bucky paper is formed by vacuum filtering a

suspension of carbon nanotubes onto Teflon filter paper. This forms a sheet of tangled

nanotubes, which can be peeled off the filter as a freestanding film. One version of the

actuator was fashioned by applying equal sized strips of bucky paper to both sides of a









slightly larger piece of double-sided tape as shown in Figure 3-1. Electrodes were

attached to both strips in order to inject charge and the bucky paper-tape composite was

immersed in an electrolyte solution. The ions in the electrolyte served to screen the

charges being injected onto the nanotubes by forming a double layer. Without it,

Coulombic repulsion would prevent the injection of significant amounts of charge.




+ -7





+ X+




Figure 3-1: Macro-scale actuator.

Voltages of opposite polarity were applied to the two sheets and motion of the tri-

layer structure was observed. It appeared that the strip on the negative terminal would

elongate and conversely the strip on the positive terminal would contract causing an

overall wagging motion as the polarity was flipped back and forth. To actually measure

the stress and strain of the bucky paper another setup was employed. The bottom of a

freestanding bucky paper was fixed to the bottom of the apparatus while the top was

attached to a cantilever. Changes in the length of the nanotube film would deflect the

cantilever and an optical sensor detected displacements of the cantilever.

The strain values calculated from the experimental data from this second

experiment setup are shown in Figure 3-2. The behavior shown in Figure 3-2 was

interpreted as arising from a change in the lattice constant of the nanotube atomic lattice








with charge injection into the nanotube pi-orbital system. This is an effect known to

occur in chemically intercalated graphite (described in section 3.3), and Baughman et al.

attributed their observations to it.



0.06-

0.8
0.04 0.6
0.6

S 0.
0.02 0-
5 0.2

0.00 -



0.9
0.7 0

-0.04

Applied potential (V vs. SCE)

Figure 3-2: Strain values for bucky paper actuators films versus applied potential in 1M
NaCl electrolyte solution. The applied potential was symmetric about V=0.
3.3 Bond Length Changes in Intercalated Graphite
Graphite intercalations compounds (GICs) are formed by the spontaneous charge-

transfer driven insertion of a chemical species, called the intercalant, between layers of

the graphite matrix. They are classified by a stage index, n, which represents the number

of graphite layers in between intercalant layers. Thus a large stage number would denote

a dilute GIC and a stage number of 1 would represent one layer of intercalant for each

layer of graphite. The interclants are labeled as acceptor (donor) compounds if they take









(give) electrons from the graphite lattice and this process can vary the free carrier

concentration greatly. This in turn can vary numerous electrical and physical properties

of the host graphite.

Intercalation compounds undergo not only an expansion along the c-axis due to the

additional volume of the intercalants but more interestingly (in the present context) they

also show in plane dimensional changes. In 1969 Nixon and Parry19 showed, through x-

ray diffraction, that the C-C bond length expanded due to the intercalation of graphite by

the donor species potassium. Following that, measurements on other graphite

intercalation compounds were performed using such dopants as Na, Li, Ba, MnC12, and

AsF5 and a strain versus estimated charge transfer curve was generated, Figure 3-

4).20,21,22,23,24 Doping by donor compounds, i.e. additional electrons in the carbon lattice,

resulted in lengthening of the C-C bond, while doping by acceptor compounds resulted in

bond shortening (although smaller in magnitude) That the bonds would shorten at all is

counterintuitive. Simple Coulombic considerations from the additional charges imply an

expansion regardless of the charge sign (since the Coulomb force is quadratic).

Coulombic repulsion of the additional charges would try to expand the host in an effort to

separate the charges. However, a quantum mechanical effect shifts the minimum

graphite lattice constant from the point of zero charge transfer to the hole doped side.

After the compilation of the charge transfer vs. strain curve some models were proposed

to explain this behavior.25'26'27'28 It is these changes in the carbon-carbon bond length as a

result of charge transfer that is the proposed mechanism for carbon nanotube actuation.

In 1981, Pietronero and Strassler first attempted to model the effect of additional

charge in a single graphite layer using tight binding energy calculations.25 They noted









that two effects would occur from this additional charge per carbon atom. First, this

additional charge would cause a change in the occupancy of the 7t states and thus a

modification of the bond order (in molecular orbital theory the bond order is defined as

the number of bonding electrons less the number of anti-bonding electrons all divided by

2). Second, since the atomic potential has changed, a modification of all the tight binding

matrix elements will arise. Additional electrons in the anti-bonding 7t orbitals weaken the

C-C bond and causing bond lengthening. Also included in their work is the change in

total energy due to Coulomb repulsion of the charged atoms (the dominating effect).

Graphite C-C Bond Length Vs. Charge Transfer

Donor Compounds Aepor Componds
MCt, 5R (A) Change in C-C
SYb\ 0.02 bond length

Ba. KCS
Li CaC
Ca. KC24 0.01

LiCi -Ij- KC48
I18 KC_ Transfered charge('
-o -0. '5 -q (e)
G-NiCIZO "..0-
G-AsF3"-,
G-FeD03
3-MnC12 -0.01



-0.02



Figure 3-3: Strain versus charge transfer curve for graphite

This work was soon followed by Kertesz26 and Kertesz et a!t7. who also used tight

binding theory to ascribe the bipolar nature of the strain of graphite intercalations









compounds to second neighbor anti-bonding interactions. The inclusion of only nearest

neighbor contributions would result in expansion for both electron and hole injection so

this interaction was needed to explain the observed results.

The next significant work was done by Chan et al.28 who used density functional

theory to focus on the changes induced in graphite solely as a result of the charge

exchange with the intercalants. For the higher stage GIC's they assumed that the

bounding layers had the same in plane lattice parameter as those next to the intercalants.

In other words, the strain is shared uniformly among all the graphite layers. This was

noted in experiment19 and supported by energy considerations. It would take far greater

energy to assume different lattice parameters for the bounding and interior layers than

would be required to make all the layers assume the same parameter. Using this

approach, Chan et al. also noticed a previously unrecognized mechanism that could not

be obtained from rigid-band models. Not only does the donated charge reside in the i7

orbitals but their theory also suggests a depletion of charge from the C bonds. The

external potentials from donor species appear to cause electron transfer from the a to 7t

bonds further weakening the C-C bond and increasing the bond length.

All previous experimental work estimated the amount of charge transfer from the

intercalants to the graphite and the theoretical calculations were based upon this estimate.

In most instances, it was assumed that the charge transfer would be nearly complete.

This work performed an experiment in which the intercalation compounds were placed

on a charge transfer scale and thus the actual charge transfer could be measured and

correlated with the associated bond length changes. Although their experimental setup

dictated that they use neutron diffraction, which is lower in resolution than x-ray









diffraction, the experiment showed that the previous assumptions about the amount of

charge transfer were within reason.

3.4 Theoretical Work for Bond Length Changes in Carbon Nanotubes

Since carbon nanotubes have the structure of a graphene sheet rolled up into a tube

it was only a matter of time before theoretical calculations were performed to predict the

strain from charge transfer to nanotubes. Just as a SWCNT's electrical properties differ

from that of graphite as a function of chirality, the predicted dimensional changes are

predicted to differ from graphite as a function of chirality. In general, it appears that the

exception is the armchair nanotubes (n,n). For this flavor of nanotube one of the allowed

k values always passes through the K point where the occupied 7r and unoccupied rT*

bands meet, just like in graphite. Since the electronic properties are similar to graphite so

should be the dimensional changes due to transferred charge. Several groups differed on

the predicted behaviors. The following paragraphs summarize these predictions and

highlight some of the salient differences in the predicted response of SWCNTs compared

to graphite.

Gartstein et a.29 studied bond length changes in nanotubes from the modulation of

electron hopping integrals. As tight binding models from graphite and carbon nanotubes

are charge conjugated symmetric in the nearest neighbor hopping approximation they

included second order hopping to account for the asymmetric actuation response going

from positive to negative charge injection. They predicted oscillating dimensional

changes as a function of chirality, with most nanotubes showing typical bond length

expansion upon electron injection but some small diameter SWCNTs showing

contraction for electron injection. The magnitude of the dimensional change (for the









same amount of charge transfer) also varies with the chirality. Their results predict some

semiconducting nanotubes will show strains 3-4 times larger than those in graphite and

other tubes will show substantially decreased responses.

A subsequent paper by Gartstein et al.30 extended their previous work on the effect

of charge injection on the geometries of nanotubes. Those results are surprising since

they are markedly different than the response in graphite. They predict that a (10,0) and a

(16,0) will show contraction upon both hole and electron injection and conversely, that a

(11,0) and a (17,0) tube will expand for both signs of doping. Also at a certain charge

injection level the response appears to reverse for the (16,0) and (17,0) nanotubes. It

should be noted that Coulombic effects, which are quadratic in their effect and certainly

play the dominant role at large values of charge transfer, were not taken into

consideration in these calculations.

Verissimo-Alves et al.31 also examined charge transfer induced dimensional

changes for a limited number of nanotubes. They carried out ab initio calculations with

the density functional theory and pseudopotential frameworks using a numerical-atomic-

orbitals basis set. They used those calculations to generate strain versus charge transfer

curves for graphite, metallic (5,5) and (12,0) nanotubes, and a semiconducting (11,0)

nanotube. The predictions for the (12,0) tube and graphite appear very similar but the

(5,5) nanotube differs in that it appears to show a significantly increased elongation for

negative charge injection but diminished contraction for electron withdrawal. The results

for the (11,0) nanotube predict expansion for both signs of charge injection. Strain

calculations (at charge transfer of 0.01 electrons/C atom) for several zigzag

semiconducting nanotubes with n=3i-1 (i is an integer) predict that SWCNT's with a









diameter smaller than that of the (23,0) tube (diameter =1.8nm) will expand for both

positive and negative charge injection. Larger diameters switch back to the behavior seen

in graphite (contraction for hole injection). Their analysis is that this behavior can be

explained by the position of the atomic level of the 7t orbital state of carbon, 62p, with

respect to the Fermi level. According to tight binding calculations, inclusion of next

nearest neighbor effects raises the energy eigenvalues at the K points above the 82p level.

Anything above 82p is anti-bonding and anything below is bonding. Thus, (if the Fermi

level is above 82p) adding electrons causes expansion and withdrawing them causes

contraction. If enough electrons are withdrawn the Fermi level will fall below 82p and

into the bonding regime. At this point if further electrons are withdrawn the bond length

will expand. For large band gaps, the Fermi level is below 82p. So for small diameter

(large band gap) nanotubes, they predict the Fermi level will lay below ;2p and

demonstrate only expansion from charge injection. As the diameter increases, and the

band gap decreases, and the semiconducting tubes should revert back to graphene-like

behavior.

In 2002, a work by Sun et al.32 calculated strain versus charge curves for different

varieties of achiral nanotubes (armchair (n,n) and zigzag (n,0)) using a uniform

background charge to represent counterions in order to more closely reflect the conditions

in the actuation experiments. They used density functional theory and the generalized

gradient approximation. Nanotubes were grouped into four categories: (n,n), (n=3i,0),

(n=3i + 1,0) and (n=3i + 2,0). These results predicted ambipolar behavior for all

SWCNTs with the exception of the extremely small diameter (5,0) nanotube. In that case

the response was expansion for both positive and negative charge injection.









The following year (2003) these authors used density functional theory to calculate

the effect of charge injection on the geometries of achiral tubes as a function of

diameter.33 Their approach in this case differentiated between two different types of

bonds, bi, which are those orientated mainly around the diameter and b2, which are those

orientated mainly along the nanotube axis.

In armchair nanotubes bl is orthogonal to the tube axis (parallel to the diameter)

and b2 has components both parallel and orthogonal to the axis. Conversely, for the

zigzag nanotubes b2 is parallel to the nanotube axis while b, has the mixed components.

Unlike graphene, their calculations predicted that the bond length's b, and b2 would be

different from each other with bi > b2 in almost all cases. The larger value of b is a

result of the curvature and the differences in lengths between the two bonds decreases

with increasing diameter. The differences between bl and b2 are smaller for the armchair

tubes than the zigzag tubes of the same diameter and decrease with increasing nanotube

diameter. They calculated how these differing bonds would change with the addition of

charges and then computed the overall strain for several charged nanotubes at charge

injection levels of q=+0.01 e/carbon

These results show some contradictions to their previous work. One example is

that the small diameter (3i + 2, 0) nanotubes will show expansion for both signs of charge

injection, something not shown in their earlier publication but similar to the predictions

of Verissimo-Alves et al.31 However, these results indicate the (3i + 2,0) family will

revert to bipolar behavior starting with diameters greater than or equal to that of the

(14,0) nanotube while Verissimo-Alves et al. predict this behavior wont resume until the

nanotube diameter is at least as large as the that of the(23,0) SWCNT. Another interesting









result of this work is that the predictions for the (3i + 1, 0) series of nanotubes show a

larger magnitude of contraction for electron withdrawal compared to expansion for

electron donation. This asymmetry goes the opposite way in graphite.

There are quite a few discrepancies between the different theoretical calculations

published in the literature. As another example, the results for the (10,0) tube with q=+/-

0.1 e/carbon in the work Garstein et al.30 predict contraction for both signs of charge

injection (something not seen in any other theoretical work) while the calculations of Sun

et al.32'33 predict a response for the (10,0) tube more similar to graphite with contraction

for positive charge injection and expansion for electron injection, also for q=0.01

e/carbon. These papers show that there is much interest in nanotube dimensional changes

caused by charge injection. Our work was started prior to the publication of these

theoretical works and was motivated by the bucky paper actuator experiment. The goal

was to observe charge induced actuation in individual nanotubes to determine if that was

the mechanism responsible for the effect witnessed by Baughman et al.8














CHAPTER 4
PREPARATION OF AND EXPERIMENTAL MEASUREMENTS ON SUSPENDED
CARBON NANOTUBE SAMPLES

To measure dimensional changes in a SWCNT we devised an experiment that

would allow injection of charge onto an individual nanotube while measuring the

resulting small changes in its length. The experimental arrangement devised do this is

illustrated in Fig. 4-1. Shown there is a nanotube suspended over a micro-machined

trench in a silicon substrate where the nanotube is pinned at its ends by metal electrodes.7

Like the bucky paper actuator setup, this experiment was also preformed in an electrolyte

solution. Charge injection onto the nanotube occurred via a potential applied to the

pinning electrodes versus a counter electrode in the electrolyte solution. To measure

length changes, the suspended nanotube was pre-tensioned by a modest force applied at

its center with the tip of an AFM cantilever. This deflected the center of the nanotube

from the top of the trench while simultaneously deflecting the AFM cantilever from its

set point equilibrium position. If the nanotube lengthened, its center would deflect further

into the trench to be detected by the corresponding relaxation of the AFM cantilever. If

the nanotube shortened, the tip would in contrast be forced upwards increasing the

deflection of the cantilever from its set point position.

4.1 Fabrication of Suspended Carbon Nanotube Structures

Fabrication of samples consisted of depositing nanotubes (either out of solution or

growing by CVD) onto silicon and lithographically patterning electrodes on top of the

SWCNTs. The nanotubes were suspended by a wet chemical etch of the underlying

silicon. Details are described in the following sections.

22








4.1.1 Substrate Preparation
Commercially available p-type <100> silicon wafers with a thermal oxide layer

were used as the substrates. Wafers were coated with Shipley S1813 photoresist to avoid

silicon dust contamination prior to dicing into 1cm2 pieces used for suspended nanotube

sample fabrication. To strip photoresist from the 1 cm2 pieces the chips were rinsed with

acetone, bathed in acetone for 10 minutes, rinsed with acetone again, rinsed with

methanol, rinsed with 18 MOhm deionized water (DI) and finally blown dry with N2 gas.

PSD










Au pinning
nanotube &
electrical
contact



i on
Si substrate


-Figure 4-1 Actuation experimental setup.
Figure 4-1: Actuation experimental setup.









4.1.2 Carbon Nanotube Deposition and Growth

Early samples in this work were fabricated by depositing purified laser ablation

grown SWCNT's out of solution onto a silicon substrate. However, longer nanotubes

were required and later samples used SWCNT's grown by CVD. Both methods are

described here.

4.1.2.1 Deposition of Laser Ablation Grown Nanotubes from Solution

The first step in this approach is to get the nanotubes on the silicon surface.

Walters et al.7, who had also fabricated suspended nanotube structures, used purified34

pulsed laser vaporization grown nanotubes that were suspended in N,N-

dimethylformamide and deposited this onto silicon chips that had a 100nm thermal oxide

layer. To date, there are no known solvents for SWCNTs. Some solvents, such as N,N-

dimethylformamide or 1,2 -dichlorobenzene, can form quasi stable solutions which will

suspend the nanotubes for short periods of time35 but the density of the nanotubes is very

limited. The nanotubes tend to form bundles with each other while in solution. As

Brownian motion causes individual nanotubes to encounter each other in liquid, the

strong van der Waals interactions between the sidewalls causes the nanotubes to

aggregate into bundles. Thus, deposition from these metastable suspensions rarely results

in individual nanotubes. Additionally, if the tubes are deposited in a manner that allows

the solvent they are suspended in to evaporate, any impurities within the solvent will be

left on the sample.

Nanotubes can be suspended in aqueous solution with the aid of surfactants.

Sodium dodecyl sulfate, Triton X-100 and most recently sodium dodecylbenzene

sulfonate have been used for this purpose.36 37. Originally I used purified34 Tubes@Rice

material that was suspended in a 1% Triton X-100 solution and deposited on bare silicon









surfaces from solution. Following purification, the nanotubes are kept in a slightly basic

solution of Triton-X in dionized water. This solution was vacuum filtered onto a Teflon

membrane, forming a film of nanotubes on the membrane. This film was washed with 1%

Triton X-100 in DI water and allowed to dry. A small amount of this film was torn off

and immersed in an aqueous 0.5% Triton X-100 mixture. The solution was placed in an

ultra-sonicator bath for typically 10 minutes. This action broke up the strip of bucky

paper into individual and small bundles of nanotubes.

The 1 cm2 silicon chips were subjected to a buffered oxide etch (BOE) diluted 1:1

with deionized water to remove the thermal oxide layer. This left the samples with a

hydrophobic H-terminated silicon surface. These chips were subsequently place in a

small vial and covered with a few milliliters of the nanotube/Triton-X solution. The

addition of a drop of concentrated acid (usually nitric acid) interfered with the

surfactant's ability to suspend the nanotubes such that after a few minutes the initially

homogeneous, particle free solution contained flocculated nanotubes. The nanotubes

(which are themselves hydrophobic) that lay close to the silicon surface adhered to the

hydrophobic silicon as they exited the suspension.

Various techniques in which acid was used to remove the nanotubes from the

Triton-X suspension were explored. One method involved placing a silicon chip face

down onto several drops of nanotube/Triton-X solution and sliding the chip across a

nearby drop of acid. This technique produced depositions in which the nanotubes were

aligned in the same direction, as example of which is shown in Figure 4-2. A comparison

of the alignment direction and the direction of the acid front as it propagated through the

nanotube/Triton-X solution showed the two to be the same. Once the acid removed the

Triton-X micelle from the nanotube, the hydrophobic SWCNTs aligned along the acid











front so that as much surface area as possible of the tubes would be out of the water

solution.

20.0 30.0 nm




15.0 nm




0.0 nm
10.0




Digital Instruments NanoScope
Scan size 20.00 pm
Scan rate 1.001 Hz
Number of samples 512
Image Data Height
Data scale 30.00 nm
0
0 10.0 20.0
pm


Figure 4-2: AFM image of aligned carbon nanotubes from Triton-X solution

Even though these samples would show high concentrations of nanotubes on the

silicon surface, the density of suspended nanotubes after etching was low. As mentioned,

nanotubes in solution tend to form bundles. This was certainly the case when the acid

was added to the surfactant solution. After being deposited on the surface the laser grown

nanotubes were grouped in bundles with very few (if any) individual nanotubes.

Moreover, the purification these nanotubes were subjected to can shorten the SWCNTs.

During the experiment, samples with nanotubes deposited by this method would often

pull apart once tensioned by the AFM probe. This evidence, along with the low density

of surviving suspended nanotubes, indicated that even though the ropes of bundled

nanotubes would span the width of the trenches, oftentimes the individual nanotubes









comprising that bundle would not. This problem led to a switch to CVD grown

nanotubes.

4.1.2.2 Chemical Vapor Deposition Grown Nanotubes

Substrates for CVD growth were diced and cleaned by the same method mentioned

above but the buffered oxide etch was omitted so that the samples retained their thermal

oxide layer. Additionally the chips were calcined at 10000 C in air for 7 minutes. Iron

nitrate nonahydrate (Fisher Scientific, certified A.C.S.) was used as the catalyst.38 and

dissolved in 2-propanol (Fisher Scientific, Optima) in concentrations of about 10 [tg/ml.

Several drops, enough to flood the surface, were dropped onto the chip and the solution

was spun dried at 3000 rpm. Growth took place in a 1" diameter tube furnace

(Thermolyne F79300, 12" heating zone) using conditions reported by Li et a.39

Hydrogen (200 sccm) and argon (300 sccm) were flowed across the chips placed

approximately 2 cm downstream of the furnace center while the furnace heated to 9000 C.

Once at temperature, the argon flow was turned off and methane (200 sccm) and

hydrogen (200 sccm) were introduced for 10 minutes. Following growth, the methane

and hydrogen flow were stopped, the furnace heat switched off, and the samples were

allowed to cooled under flowing argon (300 sccm). The density of the nanotubes grown

on the samples was characterized by atomic force microscopy imaging. The CVD grown

nanotube samples were superior to those made with deposited laser ablation grown tubes.

The CVD grown nanotubes were much longer, sometimes with lengths over 10 microns,

and were often individual nanotubes instead of bundles. Since all of the SWCNT's

grown by CVD were likely to survive the processing steps a CVD grown sample with a

lower beginning density, like that in Figure 4-3A, would result in higher density of












suspended nanotubes than a sample with a high starting density of nanotubes deposited


out of Triton-X suspension, like that in Figure 4-3B.


7.5 1 0.0
7.5 10.0


-5.0


2.5


n0
0.0
pm


Figure 4-3: AFM images of nanotubes on silicon.
Nanotubes deposited out of solution.


A) Nanotubes grown by CVD. B)









4.1.3 Electrode Patterning

To make electrical contact to the nanotubes and also pin them in place (so that they

could be suspended) metal electrodes were deposited on top of the nanotubes. Patterning

was done by photolithography. A recent development in the field has been the

formulation of lift off resists (LORs). These resists are not photosensitive but dissolve in

the same developer used for the primary resist. The LOR is applied in a thin layer to the

substrate to underlie the subsequently spun on photosensitive resist. During the

photoresist development step, the exposed edge of the LOR underlying the unexposed

(not removed) photoresist is dissolved back resulting in an effective undercut to the

photoresist layer. This undercut means that when metal is deposited it is done so

discontinuously as illustrated in Figure 4-4.

a. b.

LOR

Si Si


Figure 4-4: Metal deposition on resists. A) Discontinuous metal deposition as a result of
using a lift off resist. B) Continuous metal layer using only a standard
photoresist.

When using only conventional photoresist, the metal tears at the points where the

resist is lifted away causing rough edges on the pattern. The use of LOR permits smooth

electrode edges and easier lift off procedure requiring minimal ultrasonication for the lift-

off. AFM line scans (essentially cross-sections )of electrodes fabricated both with and

without an LOR layer are shown in Figure 4-5.










































Figure 4-5: AFM scope trace of electrodes. A) No LOR, the flaps of metal at the edges
are clearly visible. B) With LOR

During the course of this work, a particular benefit of using LOR with regard to

using carbon nanotubes in lithography applications was found. After all the lithography

steps, concluding with lift off, the samples were AFM imaged to inspect the surface

quality, nanotube density, and metal height. Samples that used an LOR layer contacting

the nanotubes had much cleaner carbon nanotubes than samples that only used the

Shipley S1813 contacting the nanotubes. It seems that the organic materials in the

photoresist have a particular affinity for the nanotubes. Although the surface of the









underlying silicon was very clean the SWNTs appeared to have globs of material stuck to

them. Despite submersion in various solvents photoresistt remover, acetone, methanol,

etc.), sometimes with sonication and at elevated temperatures, the globules remained

attached to the nanotubes. Subjecting the nanotubes to an extended ozone cleaning step

(-15 minutes) did help remove much of the material but there is the possibility this could

be harmful to the nanotubes and even that did not remove all of the extraneous material.

The components in the LOR do not seem to have the same affinity towards the

nanotubes and leave an uncontaminated surface without the need for additional cleaning.

So there are two benefits to using LOR during electrode fabrication. The electrodes have

smooth edges and standard photolithography can be performed on SWCNTs without

being hampered by additional contaminants. Figure 4-6A shows the image of a nanotube

on which Shipley S1813 photoresist was used without the LOR layer protecting the

nanotubes. The contaminants are clearly visible on the surface of the nanotube. For

comparison, Figure 4-6B shows two intersecting nanotubes on a sample where LOR and

the same Shipley S1813 photoresist were used. The latter sample is obviously much

cleaner.

To prepare the silicon chips with nanotubes on the surface for lithography, the

samples were heated in a 1100 C oven for 30 minutes to drive off any surface water.

Microchem LOR 3B was diluted 1:1 with type G thinner and spun onto the silicon

surface at 5000 rpm for 30 seconds. The LOR layer was baked at 1700 C for 45 minutes.

Shipley S1813 was used as the photoresist and was mixed 3:1 with Shipley Type P

thinner. It was spun on at 5000 rpm for 30 seconds which yielded a layer of photoresist

-800nm thick. The photoresist-coated samples were then soft baked at 900 C for 30

minutes. UV exposure was performed with a Karl Suss MA6 mask aligner with a 365nm












UV light source (11 seconds exposure time, 7.4 mW/cm2 intensity). Following exposure,


the samples were developed in Shipley MF-319 developer (22 seconds), rinsed in


dionized water and blown dry with a stream of clean nitrogen gas.


500 5.0 nm





2.5 nm





0.0 nm
250


Digital Instruments NanoScope
Scan size 500.0 nm
Scan rate 1.001 Hz
Number of samples 512
Image Data Height
Data scale 5.000 nm


-90
500
nm


- 400


300


5.0 nm





2.5 nm





0.0 nm


0 100 200 300 400
nm


Figure 4-6: AFM images of nanotubes after lithography. A) Nanotube coated with
photoresist. B) Nanotubes coated with a layer of LOR under the photoresist









Prior to metallization the samples were subjected to a UV ozone cleaning step for 1

minute. This treatment helped remove any residual organic contamination left on the

nanotubes and ensures good adhesion of the metal to the nanotubes. Since the metal acted

as the etch mask for the later Si02/Si etch, without this treatment, those etchants would

chew through the residual photoresist layer, getting under the metal electrodes, to attack

the underlying silicon. The next step in sample preparation was metallization.

Chrome/Gold and Chrome/Platinum-Iridium were both used as the pinning electrode

metals. The chrome/gold layers were grown by thermal evaporation while

chrome/platinum-iridium samples were grown by RF magnetron sputtering.

After metallization, a lift off step to remove the rest of the photoresist and

unwanted metal was employed. Lift off occurred in Michrochem Nanoremover PG. The

samples were immersed in the Nanoremover and place in a 600 C bath for 30 minutes.

During the last 10 minutes of that the samples underwent low power ultra sonication to

aid the lift off process. High powered or prolonged sonication was found to remove the

nanotubes from the surface. AFM images of samples subjected to more powerful

sonication show many nanotubes lying under the surface of the electrodes but no

nanotubes in the space between. Following the first heated bath, the samples were

immersed in fresh nanoremover solution and placed in a 600 C bath for an additional 30

minutes. After lift off, the samples were rinsed in 2-propanol and blown dry with clean

nitrogen gas.

5.1.4 Etching

To suspend the nanotubes the underlying silicon was etched away. Samples with an

oxide layer were first etched in a 1:1 buffered oxide: dionized water solution. The next

step for these samples, and the first step for samples whose oxide layer had previously









been removed, was an anisotropic wet silicon etch. The etchant solution was an aqueous

30% (by weight) potassium hydroxide solution. The KOH solution was mixed 4 parts to

1 with 2-propanol and heated to 600 C. Samples were etched for 90 seconds, which gave

a trench depth of around 600 nm. To quench the etching, the samples were moved from

the KOH bath to a 600 C deionized water bath and then to a room temperature dionized

water bath but were kept submerged during each transfer. A schematic of all processing

steps described so far is illustrated in Figure 4-7.

5.1.5 Release Procedure

Location of the nanotubes for the experiment first required imaging the samples in

a scanning electron microscope to find the nanotubes. This required that the samples be

dry. If the nanotubes samples were removed directly from water to air, the surface

tension of the water would pull down the suspended tubes to the bottom of the trenches.

Originally we tried the method used by Walters et al.7. This technique involves slowly

exchanging the liquid for successively lower surface tension solvents while keeping the

sample immersed the entire time. After etching the water was slowly exchanged for

acetone, which was in turn slowly replaced with tetramethylsilane (TMS). TMS has a

low surface tension of 10.2 mN/m, or roughly 1/7 the value of water. Once the TMS

exchange was complete the samples were removed from the liquid and the TMS was

allowed to evaporate. Unfortunately, TMS is difficult to purify and the impurities in the

solvent were left on the nanotubes.

To avoid this contamination we developed an alternative approach in which the

samples were "freeze dried". After etching a chip was transferred to a small copper boat

filled with water just sufficient to cover the sample. The water was then flash frozen by










placing the boat directly into liquid nitrogen. These frozen samples were then placed in a

freeze drier (home built) where the ice was slowly sublimed away


Grow nanotubes


Spin coat photoresist


Spin coat LOR
1--- .k


I Expose and develop
photoresist


Deposit metal

I---


HF etch SiO
----


SLift off

---


SKOH etch Si

P.


Figure 4-7: Illustration of processing steps for suspended carbon nanotube samples.

A method to determine if these processing steps had an effect on the samples was

to measure the resistance between adjacent electrodes (across the nanotubes) before and

after the etching steps. For the actual samples numerous nanotubes did not survive the

processing, ending up at the bottom of the trench. Since the doped silicon then provided a

current path between the adjacent electrodes via these downed tubes these samples could

not be used for such measurements. Instead nanotubes were grown by CVD on silicon


j





i








I ....) I
i
i i









with a 600nm oxide layer and a 14 electrode interdigitated pattern (made of thermally

evaporated Cr/Pd) deposited on top of the nanotubes. The spacing between electrodes

was the same as the pattern used for the actuation measurement samples (- 1 micron).

These test chips went through identical processing steps (HF etch, KOH etch, and TMS

release or freeze drying). Since the oxide was much thicker on these samples the HF step

did not etch through the oxide layer. The resistance was checked before and after the

processing steps for samples that went through the liquid release procedure and freeze

dry procedure, Table 4-1. The resistances of the samples did increase but that is likely

due to fewer numbers of tubes surviving to connect across electrodes at this stage.

Table 4-1: Resistance before and after processing steps
Sample Release Procedure Rbefore Rafter Rafter/Rbefore
1 Freeze dry 7.5 kM 152 kM 20
2 Freeze dry 12.3 kM 1470 kM 120
3 Freeze dry 3.72 kM 23.5 kM 6.3
4 Liquid 3.17 k 11.7 k 3.7
5 Liquid 3.54 k 46 kM 13
6 Liquid 3.05 kM 16.6 kM 5.4


Following either of these "release" procedures, samples were imaged by a Hitachi

S-4000 field emission scanning electron microscope (SEM) to identify suspended

nanotubes.The suspended nanotubes are fairly robust. In addition to imaging by SEM

they can also be imaged by tapping mode AFM with no damage. Shown in Figure 4-8 is

tapping mode image of a suspended nanotube sample and the corresponding SEM image

of the same nanotube.















Data type
Z race


Height
100.00 nm


4.00 um 0


Data type Anplitude
Z range 0.2094 V


Figure 4-8: Suspended carbon nanotube images. A) Height tapping mode AFM image of
a suspended nanotube. It can be clearly seen that the nanotube is suspended
and intact. B) AFM deflection image -more detail on the surrounding area can
be distinguished. C) SEM image of the same nanotube
4.1.6 Identification of Suspended Nanotubes
One of the difficulties of an experiment such as this is being able to locate the
suspended tubes with the AFM tip. Fortunately we were able to exploit an unwanted by
product from the SEM to make this process easier. Prolonged exposure of a substrate to


4.00 pm


Lkl









the electron beam in a SEM usually results in the build up of a deposit generated from the

breakdown of residual hydrocarbons in the chamber under the electron beam irradiation.

After finding a suitable suspended nanotube, we moved directly over to the electrode

pinning it and did a prolonged line scan on the electrode. After the mark was deposited a

SEM image was taken of the mark and its relative position to the suspended tube. Thus

once the mark was found by the AFM tip, the distance to the nanotube could be found by

referring to the position of the mark in relation to the nanotube in the SEM image.

Another SEM image was taken referencing the created marks to some type of

larger fiduciary marking (such as a corner or defect in the lithography). Thus a map of the

marks is produced. An example is shown in Figure 4-9.





















Figure 4-9: SEM image of a map of the created marks.

5.1.7 Final Sample Preparations

After marking the sample, the chip was permanently attached to a polished AFM

puck with conducting silver epoxy. The epoxy was carefully molded around the edges of

the chip to create smooth sloping walls reaching from the top of the sample down to the









puck. A shadow mask masked off the center electrode area, containing the suspended

nanotubes, and metal was deposited over the entire puck/chip assembly. This extra layer

of metal electrically connected the large electrode pads (and consequently the small

electrodes pinning the nanotubes) to the AFM puck.

Finally, the experiment itself was done in the liquid electrolyte environment and

thus, the samples had to be carefully submerged again. In this case, we immersed the

chips into the low surface tension TMS. As no liquid evaporates in this procedure, the

purity of the TMS is not an issue. In the reverse of the previously mentioned "release"

procedure, the liquids were slowly exchanged for those with higher surface tensions with

the last liquid being the electrolyte solution used in the experiment. Experiments were

performed using 0.1 M to 1 M solutions of NaCl in water, NaNO3 in water, LiC104 in

acetonitrile, and LiBF4 in acetonitrile.

4.2 Experimental Procedure

All experiments were done in an AFM electrochemical cell and a Digital

Instruments Nanoscope III AFM. This special AFM tip holder is manufactured out of

glass so the laser can still reflect off of the cantilever and reach the split photodiode. The

cell has a center chamber with an o-ring groove surrounding it. As the cell is lowered

closer to the sample surface the o-ring compresses and forms a seal between the glass

sidewalls of the electrochemical cell and also the AFM sample puck. To load the sample

in the electrochemical cell, the greased o-ring was submerged into a beaker containing

both the immersed sample and the electrolyte. The o-ring was carefully placed on the

AFM puck so that it surrounded the mounted silicon chip and is illustrated in Figure 4-10.

This allowed removal of the puck from liquid while still keeping the suspended SWCNTs

submerged in the small pool of electrolyte within the o-ring walls.









Chip with suspended
Electrolyte nanotubes



*+O-Ring




AFM Puck Epoxy

Figure 4-10: Cross sectional schematic of chip mounted on AFM puck during transfer to
the AFM.

During the actuator experiment voltages were applied using a three terminal setup

with a Princeton Applied Research Potentiostat/Galvanostat model 283. The three

terminal arrangement includes a working electrode, a counter electrode, and a reference

electrode immerse in an electrolyte. The reaction of interest occurs at the working

electrode. Whenever a metal is immersed in a solution a potential drop occurs at the

boundary of the electrode and electrolyte. There is no way to independently measure this

without introducing another electrode into the solution, which in turn has its own

potential drop at the interface. In a two terminal measurement chemical reactions at the

working and counter electrodes can cause the potential drop at the interface to be

different from the applied potential. Thus the actual difference between the two electrodes

is unknown. A reference electrode has high impedance and is chosen so that it provides a

stable and reproducible potential against which the working electrode potential can be

controlled. Thus any changes that occur in the cell potential occur at the working

electrode and not the reference. The reference electrode is designed to reproduce the

same potential regardless of the solution the working electrode is in. The counter

electrode is an inert metal (usually Pt) that completes the circuit.









The electrochemical cell had 3 ports which we made use of. The first port was

connected to a syringe that could be use for adding and exchanging the electrolyte

solution. While not in use, the syringe was disconnected and the port sealed. The second

port connected to a small glass tube that housed the reference electrode (Ag/AgCl for the

aqueous experiments and Ag/Ag+ for the acetonitrile runs). Between the glass tube and

AFM port was a porous Vycor frit. This frit allowed the ions in solution to pass while not

allowing solvent exchange. Finally, the third port contained the counter electrode. We

opted to use a strip of bucky paper due to its high surface area and thus large capacitance.

The working electrode was the nanotube sample and voltage was applied to it through the

AFM piezo cap.

Prior to use in the electrochemical cell, the electrolyte solution was sparged with

helium. Sparging removes dissolved gases from liquids by bubbling an inert gas through

the liquid. Air bubbles have the possibility of interfering with the experiment in a variety

of ways. An air bubble above the cantilever can interfere with the laser reaching the

photodiode and disrupt the measured signals. An air bubble in either the reference or

counter electrode port causes a discontinuity in the electrolyte solution and thus would

break the conduction path. In the case of a bubble in the counter electrode line the sample

simply would not get the potential we were trying to apply. In the case of a bubble in the

reference electrode line the potentiostat would apply the maximum voltage between the

counter and working electrode and destroy the nanotube sample. The electrolyte solution

was sparged for about 15 minutes before loading.

Upon loading the sample and tip into the liquid environment we observed an

enormous drift in the free space deflection signal of the cantilever. After an initial huge

drift the deflection signal change would slow down but not completely dissipate. This









has often been attributed to thermal drift in the literature. However, this explanation is

completely inadequate since the cantilever is better thermally coupled to its environment

in a liquid than it is in air. This drift was far too large to allow performance of the

experiment requiring that its source be determined to continue. We ultimately determined

that this drift was due to chemical interactions between the tip and the liquid

surroundings. The heat from the laser reflecting off the metal-coated backside of the

cantilever would increase the rate of these reactions and would often etch the reflective

coating. To fix this problem, we deposited (in a home built system) a thin (-50nm)

conformal layer of parylene C on the tip.40 Parylene C is a polymer deposited from the

vapor phase that is both chemical inert and electrically insulating. The basic procedure

for parylene deposition involves subliming the dimer, di-para-Xylylene, in a moderate

temperature zone (950C), decomposing the dimer into the monomer in a high temperature

zone (6800C), and spontaneous polymerization of the polymer on a cooled surface

(200C). By isolating the tip from the solution we eradicated the drift, yielding a stable

deflection signal. Figure 4-11 shows the drift of the free space deflection signal over 3

hours for two tips, one bare tip and one coated with parylene, in aqueous 1M NaNO3.

The AFM tips were calibrated by the reference cantilever method41 after application of

the parylene coating.

To find the marked nanotubes with the AFM, the tip was engaged on the large

electrode pads in the near a fiduciary mark. Once the end of the electrode was found, the

tip was translated down the electrode (by the amount indicated from the SEM map) until

the mark was located. The marks are quite large (heights of several hundreds of

nanometers and lengths around a micron) and are easily imaged with the AFM, Figure 4-

12



















SBare Tip
-- Parylene Tip


40 60


Time (minutes)


Figure 4-11: Free-space drift of a bare tip and a paralyne coated tip in aqueous 1M
NaNO3


I 00


300.0 nm





150.0 nm





0.0 nm


Digital Instrume
Scan size
Scan rate
Number of sample
Image Data
Data scale


nts NanoScope
1.750 pm
1.001 Hz
s 512
Height
300.0 nm


Figure 4-12: AFM image of identifying mark created in the SEM









AFM imaging performed in the fluid environment was done in contact mode.

Contact mode AFM uses a soft cantilever that is in intimate contact with the surface

while the sample is rastered back and forth. In the Digital Instruments Multimode AFM

the piezo motion is all that of the sample stage (the tip is stationary except for its Z

deflection caused by interaction with the sample). A laser is bounced off the back of the

cantilever (which has been coated with a reflective metal) to a split photo diode detector.

Before the tip is contacted with the surface to be imaged, the position of the detector is

usually adjusted so that the measured laser intensity striking the top portion of the split

photo diode equals that striking the bottom portion (the free space value is set to zero).

Once in contact, the stage will move upward deflecting the tip until the signal on the split

photodiode reaches the deflection setpoint set by the user. The amount of force exerted

by the cantilever on the sample is determined by the difference between the free space

setpoint and the deflection setpoint. While engaged with the surface, vertical deflections

of the cantilever are detected as differences in the laser intensity on the top and bottom

halves of the split photodiode. The force applied by the cantilever is equivalently that of a

spring, F=Kx, with the bending force constant, K, of the cantilevers being calibrated prior

to use and x being the actual tip deflection.

After located the mark, we used the AFM in an unconventional way to make

contact with the nanotube. If the deflection setpoint is set to a value below the free space

deflection the sample will move down away from the tip in an effort to decrease the value

measured by the photodiode to the user defined setpoint. Once the mark was found the

scan size was set to zero, the feedback gain was reduced (to make the rate of piezo

response to differences between the setpoint and the signal slow) and the deflection

setpoint was chosen to be below the free space signal in order to move the sample away









from the tip. To stop the Z motion once the tip was disengaged from the surface the

feedback gain was set to zero. At this point the X and Y positions of the stage were

adjusted such that the tip would be placed directly over the suspended SWCNT. To

guarantee minimal stress to the nanotube a low setpoint just above the free space value

was chosen and the gains were set to a very low level. This caused the stage to be

brought up very slowly as the nanotube came in contact with the tip. Once the tip

reached its setpoint while tensioning the nanotube, the gains were raised back to their

original value.

To accurately find the nanotube, the tip was lifted and lowered several times while

making small increments along the length of the trench until the highest spot, and thus the

position of the nanotube, was located. Once found by this method, force calibration

curves against the nanotube were taken to extract essential information and to verify that

the object the tip located was truly a nanotube. Force calibration mode is a useful feature

within contact AFM. In this mode the sample stage is repeatedly moved up and down

while plotting the tip deflection against the Z position of the stage as shown in Figure 4-

13A.

The right side of the plot corresponds to the stage being furthest from the tip. The

white line is the deflection signal of the cantilever as the stage moves toward the tip and

the yellow is the deflection as the stage retreats. The deflection signal remains flat until a

surface is hit, at which point the tip begins to deflect. The stage will move until the tip

reaches the user-defined setpoint.

Against a hard surface the tip deflection will equal the Z movement, as any

movement by the stage will cause the same amount of deflection in the tip (this permits

calibration of the tips Z deflection signal in volts to a distance since the displacement of









the piezoelectric stage, per volt applied to it is known). However, against an object that

stretches, like a nanotube, the Z movement will be greater than the tip deflection, the

difference being the amount of vertical deformation of the object. A schematic drawing

of the tip against both a hard surface and a nanotube is shown in Figure 4-14. This

difference made it easy to recognize a nanotube being tensioned versus say the bottom of

the trench. Because they are elastic, when the tip first comes in contact with the tube is

stretches easily. The nanotube becomes harder to stretch as the tip continues to press on

it and the curve begins to straighten. Figure 4-13A is a force calibration curve against a

metal electrode while Figure 4-13B shows a force calibration curve against a suspended

SWCNT using the same tip as in Figure 4-13A.

The force causing this stretch is calculated by multiplying the force constant of the

cantilever (measured before loading the sample) by the deflection of the tip. The number

of nanotubes in a rope can be approximated from parameters derived from the force

calibration curve.













-- Extending
Retractina


Deflection
20.77 nm/div


Setpoi n


Digital Instruments NanoScope
Ramp channel Z
Ramp size 200.0 nm
Scan rate 1.969 Hz
Z scale 207.7 nm


<- Extending
Retractinq


Deflection
20.34 nm/div


Setpoi r


Digital Instruments NanoScope
Ramp channel Z


Ramp size
Scan rate
Z scale


Figure 4-13: Force calibration curves against two surfaces. A) Against a hard surface. B)
Against a suspended nanotube


200.0 nm
1.001 Hz
203.4 nm









I Ic




Hard surface => dZ = dx



Sample stage moses : dy
upw ard until tip is -
deflected to setpoint dZ

Flexible surface => dZ = dx + dy

Figure 4-14: Illustration of a force calibration curve against a hard surface and nanotube.

The original vertical stretch of the nanotube, the initial deflection designated by y

and illustrated in Figure 4-15, is calculated by subtracting the deflection of the tip from

the Z movement of the sample stage while the tip and nanotube are in contact. Simple

statics considerations give the force needed to deflect the nanotube a vertical distance y

as,

Lo
F:=2-k- 1-- -y

y2+ L 2)2 (4.1)


k is defined as

k = Y-A/Lo (4.2)

Y is the Young's modulus (1012 Pa) for graphene. A is the cross sectional area of the

nanotubes under tension. The cross sectional area for one nanotube is the van der Waals

thickness for graphite (t = 0.34x10-9 m) multiplied by the circumference of the nanotube

(7r d where d is the diameter of the tube) Thus the total cross sectional area for a rope of









nanotubes is n-t--.d where n is the total number of nanotubes in the rope. This gives the

following expression for k.

n-Y-t-d-.
k -
S L (4.3)



Au L Au Lo 2 length of unstressed tube
.- -. L,- 2 length of tube under tension by .M tip
Y, L- 1/2 length of tube under tension with
Si Si voltage applied
dy 2*Lo = width of the trench

Figure 4-15: Diagram of nanotube under tension.

The parameter y is essential to calculating the observed actuation. The tensioned

length of the nanotube is Li=(L02 + y2)1/2 where Lo is the same as mentioned above. Any

actuation of the nanotube will cause a deflection of the cantilever by dy as the tube

lengthens or contracts. Using dy we can calculate the new length of the tensioned tube as

L2=(Lo2 + (y+dy)2)1/2. Thus the actuation strain 6L/L is defined as 6L/L=(L2 L1)/L1. It

is instructive to plot dy as a function of y for a given amount of actuation, i.e. setting

6L/L to a constant, as shown in Figure 4-16. The parameters Lo of 430nm and 6L/L of

0.0001 (corresponding to an actuation of 0.01%) were used.

The value of dy decreases sharply with increasing y. Once this particular tube has

been stretched past y=18nm the value of dy becomes less than Inm. In order to observe

actuation with the AFM tip, the nanotube must be under some tension but the amount of

initial stretch needs to be small in order to have any sensitivity to the effect.










6 I I I I I I I








Dy(y) 3









,0.464, I
0 5 10 15 20 25 30 35 40
0 y 40

Figure 4-16: Plot of dy versus y forLo=430nm and 6L/L=0.0001. Both y and dy are
plotted in nanometers.

After the tip tensioned the nanotube by the amount y, as measured by the force

calibration curve, voltage was applied to the electrodes through the piezo cap of the

AFM. Elongation or contraction of the nanotube resulting from the applied voltage will

cause a deflection of the cantilever. A Princeton Applied Research

Potentiostat/Galvanostat model 283 controlled all applied voltages to the three electrodes.

A square wave potential with a low frequency of either 0.5Hz or 1 Hz was used. Data of

the applied voltage, current, and piezo movement in the Z direction was recorded with a

Labview program. Usually, data was taken for different values of pre-tension y and

different applied voltages on each nanotube. Several runs were taken with each variation

in parameter.














CHAPTER 5
RESULTS AND DISCUSSION OF CHARGE INDUCED ACTUATION OF
SUSPENDED CARBON NANOTUBES

5.1 Results of Actuation Measurements

The experiment was performed on 35 suspended nanotubes. Several distinct

electrolytes were used. Aqueous NaC1, aqueous NaNO3, LiC104 in acetonitrile, and

LiBF4 in acetonitrile were employed in concentrations ranging from 0.1 M to 1 M. The

organic solvents were used to obtain a larger voltage window than the hydrolysis of water

permits. Square wave voltages were applied up to 2 Volts peak to peak. During some

runs only negative voltage was applied to the sample since the bond length changes are

expected to be greater for electron injection compared to electron withdrawal. The

parameter monitored to detect dimensional changes in the nanotube was the Z voltage

supplied to the piezo, a change of 1 V corresponded to 12.9nm. Changes in the nanotube

length caused the AFM tip to either relax below or be pushed above the setpoint. The

sample stage would move to compensate for this change in length so that the setpoint was

restored. Out of the 35 samples only 3 nanotubes displayed length changes that could be

duplicated in more than one run.

Sample 1 is shown in Figure 5-1. As measured from the SEM image the

parameter Lo is 484nm. This sample was immersed in 1M aqueous NaCl solution. An

example data run is shown in Figure 5-2 and a summary of all the data runs for this

nanotube is listed in Table 5-1. The parameter y is the initial vertical deflection of the

tube as measured by the force calibration curves, dZ is the is the voltage change in the

AFM peizo corresponding to the vertical nanotube changes, dy is the dZ voltage from the

51







52


piezo converted into nanometers (12.9nm/Volt), and % dL/L is the percentage strain of

the nanotube.


Figure 5-1: SEM image of nanotube sample 1. The nanotube of interest is the thin tube
directly to the right of the mark


- Applied Voltage


- Z Movement


I U .. UL


0 1000 2000 3000 4000 5000
Time (milliseconds)


Figure 5-2: A data file from sample 1. The applied voltage is shown in black with its
scale on the left. The Z piezo movement is shown in blue









Table 5-1: Summary of data from sample 1
Run Y (nm) dZ (V) Dy (nm) % dL/L
1 36.9 0.0269 0.346 0.00544
2 36.9 0.0175 0.225 0.00354
3 14.3 0.0206 0.265 0.00163
4 14.3 0.0189 0.243 0.00150


Nanotube sample 2 is shown in Figure 5-3. This sample was exposed to

polyethylene imine (PEI) prior to loading in the AFM electrochemical cell. PEI has been

shown to n-dope the nanotubes.42 In graphite, bond length changes for the same amount

of charge transfer are smaller if the sample is already p-doped but larger if the sample is

n-doped. PEI is an extremely viscous liquid and was dissolved in methanol from

concentrations starting at 2.5% an increasing to 20%. The suspended nanotube sample

was first carefully submerged in methanol which was slowly exchanged for the 2.5% PEI

solution. This was repeated 4 times with subsequently higher concentrations of PEI

solution with the final step resulting in the sample residing in the 20% PEI mixture. To

adsorb the PEI onto the nanotubes the sample was left submerged overnight. The

following morning the sample went through the reverse of the process, with the liquids

being exchanged for lower concentration solutions of PEI. The PEI solution was

exchanged for pure methanol and that was exchanged for water at which point another

freeze dry was performed. From the SEM image of this sample the parameter Lo was

determined to be 429nm. This experiment was preformed in 0.5M LiBF4 in acetonitrile.

As sample of a data file is show in Figure 5-4 and a summary of the parameters from all

data files for sample 2 is shown in Table 5-2






























Figure 5-3: SEM image of nanotube sample 2. The nanotube is to the left of the mark.


SApplied Voltage Z movement


1.U -

0.8-

0.6-

0.4-

0.2-

0.0-

-0.2-

-0.4-

-0.6 -

-0.8-

-1.0


0 1000 2000 3000 4000 5000
Time (milliseconds)


-4.4


-4.2


-4.0


-3.8


-3.6


-3.4


-3.2


Figure 5-4: Sample 2 data


r 1 1


I 11. 1


LI L,









Table 5-2: Summary of data for sample 2.
Run y (nm) dZ (V) Dy (nm) % dL/L
1 18.4 0.0134 0.172 0.00178
2 18.4 0.0104 0.134 0.00134
3 11.4 0.0151 0.194 0.00121
4 11.4 0.0230 0.296 0.00186


Sample 3 was also submerged in 0.5 LiBF4 in acetonitrile. From the SEM image

shown in Figure 5-5 the parameter Lo was determined to be 444nm. Figure 5-6 displays

the data from one run and Table 5-3 is a summary of all the data runs for sample 3.


Figure 5-5: Image of sample 3. The nanotube measure is the one to the left of the lower
marking.


0 0 1 ;1 7 9 E; 0 k Y X 2 5 6 k' i











A- Z Movement


O I I I I I*

8

6

4

2-

0

2-

4

6-

8


0 1000 2000 3000 4000 5000
Time (milliseconds)


Figure 5-6: Sample data file from nanotube 3.

Table 5-3: Summary of all data from sample 3
Run v (nm) dZ (V) dv (nm) % dL/L


16.8
16.8
16.8
14.6
14.6
14.6


0.0102
0.0112
0.00800
0.0115
0.0119
0.0109


0.131
0.144
0.103
0.148
0.153
0.140


0.00112
0.00123
0.000878
0.00110
0.00114
0.00104


To ascertain whether the discrete jumps observed in the in the Z movement from

the application of voltage to the nanotubes where a result of nanotube actuation or from

some type of chemical reaction in the cell other data files were also taken. These include

monitoring the Z movement while the tip was suspended in the electrolyte but not in

contact with anything as shown in Figure 5-7. Additionally data was taken with the tip in

contact with the bottom of the silicon trench, Figure 5-8.


- Applied Voltage








57




-- Applied Voltage Z Movement
.0 -0.4


-I.u .. .


0 1000 2000 3000

Time (milliseconds)


4000 5000


Figure 5-7: Data of the Z Movement and applied voltage while the AFM was suspended
in free space


--Applied Voltage


0 1000 2000 3000

Time (milliseconds)


- Z Movement


4000 5000


Figure 5-8: Data of Z Movement and applied voltage while the AFM tip was in contact
with the bottom of the trench.


6-



4-

2-

0

2

4-

6


Nt *









A summary of all the data with average actuation values is listed in Table 5-4.

Table 6-4: Summary of the actuator data
Sample Electrolyte Applied Voltage Lo Y (nm) Dy % Actuation
(peak to peak) (nm) (nm)
1 1M NaCl in 1.5 V 484 36.9 0.285 0.0045
H20
2 0.5M LiBF4 1.2 V 429 11.4 0.245 0.0015
in CH3CN
3 0.5M LiBF4 1.4 V 444 14.6 0.147 0.0011
in CH3CN

5.2 Discussion of Results

The experiment perform by Baughman et al.8 observed much higher actuation

strains in the macroscopic sheets. For an applied voltage of 0.9 Vpp the strain was nearly

.1%. Since the nanotubes were bundled, with only the outer nanotubes presumed to be

undergoing actuations, these values were estimated to be lower limits to nanotube

actuation. Based on their results, the authors speculated that the maximum strain for

individual nanotubes could reach -1%/V. Our measurements fall far short of that

prediction and are smaller than that observed for the nanotube sheets.

This lack of evidence for charge-induced actuation in our work led us to

reexamine of some of our fundamental assumptions. The first question that needed to be

asked was "how much charge can a nanotube hold?" And once that has been determined,

"how much strain is predicted for this amount of charge injection. To answer the first

question we calculated the capacitance of a carbon nanotube under these conditions.

Previously, Kruger et al.43 and Rosenblatt et a.44 had used MWCT and SWNT in

electrolyte gated transistors. Both groups had calculated the capacitance per unit length of

the nanotube within the electrolyte solution using the geometry of two concentric

cylinders.

C= 2 7 s0 s/ ln(R2/Ri) (5.1)









The inner cylinder is the nanotube whose radius is given as R1 in the equation. The

value they used for the outer cylinder diameter R2 was that of the nanotube, R1, plus the

Debye length, 4D, for the ionic concentration of the solution. They also used the accepted

value of s=80 for dielectric constant of water. Both groups used electrolyte

concentrations of ImM which gives a Debye length of XD =lnm. Using these values the

two groups calculated gate capacitance values of- 10 nF/m or 10-17 F/nm (62.5 e-/nm/V).

This use of the bulk dielectric constant of water responsible for such large capacitance

values is however not correct in the case of electrolytes where the effective spacing

between electrodes is the width of the electric double layer.

In electrolyte solutions, when a voltage appears on an immersed electrode (in this

case the nanotube) the ions of the opposite polarity will migrate toward the electrode to

neutralize the charge. Thus the composition of the electrolyte solution near the interface

is different than the rest of the bulk solution generating the outer half of the

electrochemical double layer (the inner half being the charge on the electrode). The layer

of charge at the surface of the electrode and the concentration of opposite charges in the

solution surrounding the electrode generate a capacitance. However calculation of the

double layer capacitance is non trivial and remains an active area of research and debate

(e.g. see chapter 2 reference 45). For example, what is the location of the counterions

surrounding the electrode? The most accepted arrangement is the Gouy-Chapman-Stern-

Grahame model. In this model part of the potential drop occurs from a layer of fairly

closely packed counterions near the surface, the Helmholtz layer. The rest of the

potential drop occurs over a diffuse section of ions in which electrostatic forces compete

with Brownian motion. For solutions with a concentration of 1M or higher generally the

entire drop takes place over the compact Helmholtz layer and the influence from the









diffuse layer is only a minor correction. The positions of the ions near the electrode are

also influenced by the dipole moments of the water molecules surrounding them.

Conversely, the electric fields generated by the ions in solution and the electrode

also affect the water dipole moments and can change the dielectric constant. This effect

was noticed as early as 1948 by Hasted et al.46 who tried measuring the modified

dielectric constant in aqueous salt solutions. The ions in solution orient the water

molecule dipoles. The water molecules in the immediate vicinity of an ion are

immobilized by the ion's electric field. The ion is referred to as hydrated. Figure 5-9

shows an illustration of hydrated Na and Cl ions

H









Figure 5-9: Hydrated C1- and Na ions

The polar water molecules are no longer free to rotate in response to the applied

external electric field, which is what imparts to water its large bulk dielectric constant.

The presence of ions in the water breaks up the water network and decrease the overall

dielectric strength, something known as dielectric saturation. The dielectric strength for a

water-counterion complex for Na is SNa~2 and the radius of a hydrated Na ion is

r=0.36nm.47'48 If we use this radius as the distance to the outer cylinder in our

capacitance equation we have to use the drastically reduced value of gNa=2 for the

dielectric constant. Using these values (and R1=1.36nm, the diameter of a (10,10)

nanotube) the calculated capacitance goes to C=2.6x10-19 F/nm (1.6 e-/nm/V). This value









is more than order of magnitude smaller than what was estimated by Kruger et al.43 and

Rosenblatt et al.44. In one nanometer of length for a SWCNT with a diameter of

d=1.36nm we have 160 C atoms. Using this information we can convert to electrons per

carbon atom.

C=(1.6 e-/nm/V)(1 nm/160 C atoms)= 0.01 e-/C-atom/V. (5.2)

So for sample 1 that had 1.5 volts peak to peak applied to it, we can expect that the

amount of charge injection was q = +0.0075 e-/C-atom. The theoretical work on SWCNT

dimensional changes discussed in Chapter 3 give an average of 0.1% strain for the

addition of 0.01 e-/C. Hence for 0.0075 e-/C-atom calculated from the capacitance we

should expect a strain of 0.075%.While acetonitrile (bulk s=37.5) was used as the solvent

in a number of the experiments the same considerations can be expected to apply.

Results from a spectro-electrochemical study on carbon nanotube thin films in the

Appendix give further support to this capacitance estimate. Carbon nanotubes have

optical absorbance peaks centered around 1650 nm and 900 nm. These peaks are due to

photoinduced electronic transitions between valence to conduction band van Hove

singularities. The 1650 nm and 900 nm peaks correspond to the 1st and 2nd van Hove

singularities for semiconducting nanotubes. The peaks are broad because of the range of

nanotube diameters in the sample and also because of perturbations to the electronic

structure from tube to tube interactions. The absorbance peaks can be reduced or even

eliminated by shifting the Fermi level. If electrons (holes) are added then the singularity

in the conduction (valence) band will be filled (depleted) and no electronic transitions,

and thus absorption of photons, will be possible.

From these measurements we can determine at what applied voltage the first van

Hove singularity is either depleted or filled. At an applied voltage of negative -0.7 V the









first peak has started to shift but the second has not, Figure A-4. Note that the peak has

not been completely removed. That is to be expected the sample contains bundles of

nanotubes. At low applied voltages the Fermi level of an outlying nanotube has been

shifted while the Fermi level of an inner lying tube remains untouched. Since there is no

change in the second peak we assume that even the outer tubes have not had their Fermi

levels shifted into the second singularity yet. We determined the additional e-/C atom

when the Fermi level was shifted to the point just before the second semiconducting van

Hove singularity by using a program that calculated the density of states for nanotubes of

any indices (n,m). The method was based on zone folding of the tight binding 2D energy

dispersion relations of graphite (Equation 2-11) and a simple rotation transformation of

kx and ky into k_ and k| (relative to the nanotube axis).49 A graph of the density of states

for a (10,11) semiconducting nanotube and the shifted Fermi level are shown in Figure 5-

10. Integrating the density of states up to the dashed line in this figure gives 0.005 e-/C-

atom. From the results in the appendix an applied voltage of-0.7Volts therefore

corresponds to an additional charge of 0.005 e-/C. We also did the same calculation for a

(10,10) metallic nanotube with the Fermi level shifted the same amount and found an

additional 0.007 e-/C-atom. Using V=0.7V in our capacitance calculations (Equation 5.3)

yields .007 e-/C-atom which is in reasonable agreement for the two values. It should be

noted that the calculation for the capacitance (Equation 5.3) assumes the nanotube is a

perfect metal able to accommodate unlimited additional electrons. Since the nanotubes

are limited in the number of additional electrons they can contain by the density of states,

the derived formula for capacitance is likely an overestimate.

Despite the reduced capacitance, the results of Baughman et al. are around what

would expected so capacitance alone cannot explain our small results. However, their










sample also contained bundled nanotubes. Since the strain would be shared by the inner

nanotubes, which would not be undergoing dimensional changes, the actual strain of the

outer nanotubes would be greater than that displayed by the nanotube sheet.

Additionally, their results display some reversals in the actuations (Figure 3-2) at

voltages of 0.5V and up that cannot be explained by bond length changes from additional

charges in the 7r orbital system. That behavior is very similar to an effect caused by

double layer charging in porous graphite in which changes in the interfacial tension

within the pores cause dimensional changes in the graphite.5

40 I--1--1 I
n= 10
35 m= 1

30 -

25 EF
E,
DOS 20 -
(e/nm)
15 -

10

5

0 0.2 0.4 0.6 0.8 1 1.2 1.4
E


Figure 5-10: Density of states for a (10,11) SWCNT.

The SWCNT film used in the spectroelectrochemical experiment in the Appendix

was baked to remove any dopants acquired during purification. However, the

transmittance minimum for the peak associated with the first semiconducting optical

transitions occurred between -0.2V and -0.4V. That is where the point of zero charge

(PZC) on the nanotube film would occur. This indicates that the nanotubes were p-doped









by their exposure to electrolyte (which has its own chemical potential) atmosphere

despite the desorbing bake. The experiment by Baughman et al.8 used aqueous 1M NaC1.

For graphite in 1M NaCl the PZC is -0.2 V55,56,57 so it is likely the PZC of the nanotube

film would also be negative. As mentioned, the dimensional changes in graphite (for the

same amount of charge transfer) that is already p-doped will be smaller than the changes

in un-doped graphite.

Even accounting for the reduced dielectric constant, the predicted strain is still

larger than what we measured. One of the difficulties within the nanotube field has been

making good contacts to the one dimensional SWCNTs. SWCNT field effect ransistors

(FETs) made with metals that were consistent with our fabrication process have been

found to be consistently p-type, even with good ohmic contacts for the on state. 50,51,52,5354

It has been suggested that strong dipole moments of adsorbed gases (particularly oxygen)

at the contacts modifies the barriers and this effect leads p-type behavior.53 Additionally,

there is a barrier introduced by mismatch in work functions of the contact metal and the

nanotube.

Rosenblatt et al.44 formulated water gated nanotube FETs. Like our work, they

used an electrolyte solution to gate the nanotube and also contacted the tubes with Cr/Au

electrodes. They found that small diameter nanotubes (3nm and under) were still p-type.

Even using the electrolyte gate they could not push across the band gap to get electron

transport. The results in the Appendix show that nanotube films can be pushed across the

gap. However only a very small portion of the nanotubes in a film are in contact with the

metal contact electrode and thus only that small portion would be p-doped by the

contacts. But individual nanotubes (like the ones used in our experiment and that of









Rosenblatt et al.44) are sufficiently p-doped by the contacts so that the Fermi level cannot

be shifted into the conduction band.

These issues result in two factors that explain why charge induced dimensional

changes in these experiments is so small. Contact barriers with nanotubes prevent

electron injection in semiconducting nanotubes. And perhaps more importantly, the fact

that the nanotubes are p-doped puts the nanotubes into the portion of the strain versus

charge transfer curve where the length changes are smaller (see Figure 3-3). Attempts to

inject sufficient electrons to take the nanotubes to the steeper part of the curve require

pushing the semiconducting tubes across the gap while the metal electrode contact with

the nanotubes result in Schottky barriers that further impede such electron injection.

Using an extremely low work function metal (such as Ca) would n-dope the nanotubes

and cause barriers favorable to electron injection. However, those metals are

incompatible with the processes required to the suspended nanotube samples. Thus we

are prevented from measuring significant charge induced dimensional changes in

individual nanotube because of effects resulting from the contacts.















CHAPTER 6
CONDUCTANCE CHANGES IN CARBON NANOTUBES DUE TO HYDROGEN

The work in this chapter demonstrates the utility of carbon nanotubes as hydrogen

gas sensors. Parallel efforts demonstrated that sputtering of metals onto SWCNT

decreases the intrinsic conductance of the nanotubes.

6.1 Carbon Nanotube Sensor Background

The first demonstrated use of nanotubes as chemical sensors was by Kong et al..58

They monitored the charge transport changes of semiconducting SWCNTs when exposed

to NH3 and NO2. These early sensors patterned source and drain electrodes on individual

nanotubes in a field effect transistor (NFET) configuration Upon exposure to electron-

withdrawing NO2 the turn on voltage of the NFET was shifted by +4V. When exposed to

electron donating NH3 the turn on voltage was shifted by -4V. The interaction of these

molecules with the SWCNT shifted the Fermi level, through the density of states of the

nanotubes, modulating their response to the gate voltage.

Other groups have since then tested the effect of various gases on the conductance

of carbon nanotubes using either individual tubes, bundles of tubes, or thin films of

SWCNTs.59,60,61,62 The observation has been that electron donating species cause a

decreased conductance of the p-type semiconducting nanotubes and electron withdrawing

molecules increase the conductivity. These sensors all had long recovery times (hours)

although exposure to UV light or elevated temperatures helped speed up the desorption

process.









Recent efforts have focused more on using thin films of nanotubes as sensors.

Novak et al. 63 used chemical vapor deposition to grow a nanotube network an SiO2/Si

substrate that was just above the percolation threshold. These films were used to detect

several electron donating molecules including dimethyl methylphosphonate, DMMP (a

simulant for the nerve agent sarin). This group was the first to use the clever method of

applying a positive back gate to refresh the device. They surmised that the Coulombic

repulsion between the negative charges induced by the gate and the negative charge

donated by the DMMP was enough to lower the desorption barrier and drive off the

DMMP. Although these devices are in some sense easier to manufacture than individual

nanotube FETs, consistent growth of a nanotube film near the percolation threshold can

be difficult.

The number of chemicals forming charge transfer complexes with the nanotubes,

and which may therefore be detected by this mechanism, is rather limited. Y. Lu et al.64

first demonstrated the detection of methane, which does not by itself undergo charge

transfer with SWCNTs, by loading the SWCNT film with palladium particles. The

sensors were formed by first sputtering a 10nm thick layer of Pd onto SWCNT powder

and mixed by shaking this concoction. The metal-coated powder was dispersed in

dionized water, sonicated, and drop-dried onto an interdigitated electrode pattern. This

work certainly showed the utility of loading SWCNTs with metals but the method of

associating the nanotubes with the metal was poorly controlled as was the density of the

nanotubes bridging the electrodes.

The work presented here focuses on using SWCNTs loaded with Pd particles for

the detection of hydrogen gas. During the coarse of this work another group published a









paper also using Pd-SWCNTs as H2 sensors.65 This group deposited Pd onto the

nanotubes by RF sputtering and also by using a technique in which a Pd salt solution

(containing SWCNTs) was reacted with toluene to deposit Pd particles on the nanotubes.

After Pd loading, the SWCNT films were deposited by airbrushing a suspension of

nanotubes in ethanol onto an aluminum substrate. Contrary to the high sensitivity of

sputtered Pd-SWCNT presented here, this group had no sensitivity to H2 with the

sputtered Pd-SWNT samples. Additionally, we observed higher 6R/R ratios for amounts

of H2 at least an order of magnitude smaller.

Palladium is an interesting choice for metal loading of SWCNT H2 sensors for a

couple of reasons. At room temperature and atmospheric pressure palladium can adsorb

up to 900 times its own volume of hydrogen. Additionally, palladium has been shown to

make ohmic contacts with carbon nanotubes. Javey et al.54 used Pd electrodes to form

semiconducting nanotube FETs. Upon exposing the devices to H2 they demonstrated a

decrease in the conductance of the p-type semiconducting nanotubes. It was speculated

that this was because the adsorbed H2 lowers the work function of the palladium. Earlier

experiments testing the change in resistance of discontinuous Pd films upon exposure to

hydrogen claimed an increase in the palladium work function.66,67,68 As pointed out by

Barr66 the relevant mechanism is that hydrogen adsorption lowers the Fermi level of the

palladium, thus causing a increased Schottky barrier for hole transport.

Hydrogen detection is important in a variety of areas including semiconductor

fabrication clean rooms, space missions where H2 is used as fuel and wherever H2 could

cause an explosive mixture. The current efforts toward a hydrogen based fuel economy









further emphasize the need to find dependable low power H2 sensors. The goal of this

work was to produce a reliable, easy to fabricate, H2 sensor made from carbon nanotubes.

6.2 Carbon Nanotube Sensor Fabrication

We fabricated two types of SWCNT sensors, both coated with a thin layer of Pd

metal. The first involved tens of CVD grown nanotubes wired between parallel

electrodes spaced roughly 1 micron apart. The second type involved ultra thin nanotube

films made from pulsed laser vaporization bulk produced SWCNTs .

The first type, from hereon called the micro-device sensor, was fabricated using p

type <100> silicon chips with a 600nm thermal oxide layer. Using the same techniques

discussed in Chapter 4, the chips were cleaned, nanotubes grown by CVD, and

photoligthography used to form an interdigitatedl4 electrode pattern. The lengths of the

electrodes are 500 |tm so there are many nanotubes in parallel between two adjacent

electrodes. The pattern is shown in Figure 6-1.












Figure 6-1: Electrode pattern on micro-device nanotube sensor.

The ultra thin nanotube film sensors were manufactured using a technique

described by Wu et al..69 Purified SWCNTs grown by laser ablation were suspended in

an aqueous 0.5% Triton-X solution with a nanotube density of 0.001 mg/mL. The

nanotube film was formed on a membrane by vacuum filtering the solution onto a mixed









cellulose ester membrane (0.1 jtm pore size, Millipore). After the film was formed the

Triton-X was washed away with DI water. The film thickness was easily controlled by

the amount of the nanotube suspension filtered. To transfer the nanotube film to the

desired substrate the film was adhered to the substrate by wetting with water and pressing

between metal plates until the water dried. The mixed cellulose ester membrane was then

dissolved away by acetone leaving the ultra thin nanotube film behind. Films made with

this technique can be as thin as a few nanometers.

The nanotube thin film sensors were fabricated on cleaned silicon substrates with

a 600 nm SiO2 layer. Nanotube films with nominal thickness of 7nm and 25nm were

used are shown in Figure 6-2. Step heights of thicker films (50 nanometers) were

measured by AFM and correlated to the amount of nanotube solution used. The 7nm and

25nm heights of the films are estimates found by scaling down the amount of solution.

As seen in Figure 6-2b, the 7nm film effectively sub-monolayer.















0 1.00 2,00 0 1.00 2.00 3.00

Figure 6-2: AFM images of a) 25nm SWCNT film and b) 7nm SWCNT film

The films were deposited on the silicon chips as follows. After washing the

nanotube film on the membrane it was allowed to dry. Pieces, of the ultimately desired









film size were then cut from the membrane, wetted with DI water and placed nanotube

side down onto the silicon chip. The substrate was sandwiched between several pieces of

filter paper for cushioning and pressed by spring clamps between two metal plates. After

the water evaporated, aided by placing the samples in a 950 C oven for an hour, the

nanotube film/filter membrane was attached to the substrate. To remove the mixed

cellulose ester filter, the silicon chips were transferred to an acetone vapor bath. The

membranes dissolved in the acetone vapors leaving only the nanotube film. The samples

subsequently went through five additional acetone baths followed by a methanol wash to

ensure the complete removal of the membrane polymer. The samples were baked in an

inert gas to desorb remaining contaminants and nanotube dopants. To give volatiles a

chance to escape before they reacted with the nanotubes the oven was ramped at a slow

rate under the following schedule: 50 C/min to 1100 C, held for 30 min, then 1 C/min to

6000 C, held at temperature for 2 hours and and powered off to cool to room temperature.

Contact pads of Cr/Pd were deposited across the ends of the SWCNT film by either

sputtering or thermal evaporation. A photograph of the SWCNT film sensor device is

shown in Figure 6-3.














Figure 6-3: SWCNT thin film sensor wired for measurement. Bright regions labeled
source and drain are Pd metal pads. Voltage is applied across the source and
drain while the current, passed through the SWNT film, is measured









6.3 Resistance Changes from Metal Deposition

As will be discussed in the next section, both the micro-device and the film

sensors show miniscule response to H2 at this stage. To be sensitive to the gas we

deposited thin, in some cases non-percolating, layers of palladium on top of the

nanotubes by either sputtering or thermal evaporation. The samples were masked such

that Pd was deposited only onto the nanotube films. Surprisingly, the resistance of all the

nanotube samples coated with sputtered Pd increased. As this situation is analogous to

two parallel conductance paths (the nanotube film and the metal layer) the increase in

resistance is contrary to expectations, which would have the resistance decreasing. Table

6-1 shows the increase in the resistances of three samples, a micro-device sample, a 7nm

film and a 25nm film, (AJA International ST10 magnetron source, 3 W power, 5 mT Ar,

5 s deposition time). 600 nm oxide-on-silicon witness chips, having the same

source/drain electrodes but no nanotube films rode along with each nanotube sample

during the thin Pd layer deposition. These showed the Pd films to be sub-percolation in

all these cases.

Table 6-1: Resistance changes of nanotube samples with sputtered Pd
Sample Rbefore Rafter Rafter/Rbefore
Micro-device 1.33 kf 5.32 kf 4.0
7nm film 3.18 kf 6.69 kf 2.1
25nm film 829 Q 1.34 kQ 1.6


The resistance of several SWCNT films was monitored before and after the

deposition of thermally evaporated Pd incorporating similar witness chips. In several of

these cases the witness chips exhibited finite resistances demonstrating that the amount of

metal deposited was above the percolation threshold. This point and the resulting

resistance will of course depend on the substrate upon which the metal is deposited









(nanotubes for the films and Si02 for the chips), nevertheless to first order, we take the

resistance of the metal layer measured on the witness chips to be that of the metal layer

on the nanotubes. Table 6-2 lists the before and after resistance of several nanotube film

samples with varying amounts of thermally evaporated Pd, as well as the resistance of the

witness chips. In each instance the resistance of the nanotube films decreased however

the decrease varied with the amount of Pd. Films 1 and 4 had roughly a 5nm thick Pd

film deposited on top (measured in situ by an Inficon quartz crystal thickness monitor).

The resistance of the Pd layer on top of Film 1 is smaller than the resistance of Film 1

itself and is comparable in height. The final resistance is greater than the value calculated

by combing the resistance of the nanotube and the Pd film in parallel (R =1.24 kM

compared to Rafter=1.72 kM). Similarly the final resistance of Film 4 is also larger than

the calculated parallel resistance (R =0.785 kM compared to Rafter=0.993 kM). Films 2

and 3 had Pd layers between 2nm and 2.5nm deposited. The final resistance of these

films is smaller than the calculated parallel resistance.

Table 6-2: Resistance changes on SWCNT films with thermally evaporated Pd
Sample Nanotube film Rbefore Rafter Rpd film Rafter/Rbefore
height
1 7nm 5.72 kM 1.72 k 1.58 k 0.30
2 7nm 8.86 kM 3.45 kM 90.0 kM 0.39
3 7nm 5.76 kM 4.82 kQ non-percolating 0.84
4 25nm 1.46 kQ 0.993 kM 1.70 kQ 0.69

Both metallic and semiconducting nanotubes have been shown to be ballistic

conductors with resistances per nanotube approaching the theoretical limit of 6.5 kM.70'54

However, the resistance across nanotube films between macroscopically spaced

electrodes is dominated by the resistance across tube to tube contacts, which constitute

tunnel junctions.71,72,73 The data shows that when less metal is evaporated on the









nanotube films the resistance across the films is lowered by an amount greater than

consideration of the parallel (nanotube/metal) conductance while when more is

evaporated the resistance (while still lowered) is lowered by an amount less than the

parallel conductance. Given the ballistic on tube conductance a simple view would

suggest that by increasing the electrical contact area across tube-tube contacts the

resistance across the films should be greatly lowered. We can make sense of the data by

recognizing that it is naive to assume that a singled walled nanotube associated with

metal along its length retains its ballistic conductance. We conclude that the better than

parallel path resistances of films 2 and 3 are due to lowered tube to tube resistance arising

from increased contact area at the junctions due to the Pd particles which however are

discontinuous and do not coat long sections of the nanotubes. Too much metal, which

coats long sections of the nanotube sidewalls, interferes with their ballistic conductance.

Both theoretical calculations and experimental evidence support this view.74'75

Regardless of the method of deposition the palladium coated the nanotubes.

Figure 6-4 shows AFM images of two 7nm films, one coated with sputtered Pd and the

other by thermally evaporated Pd. The morphological differences between the two films

are likely due to the variation in the rates of deposition. Both films are -10-30

Angstroms thick but the sputtered film was deposited in 10 seconds while the thermally

evaporated film was grown in 80 seconds. Higher deposition fluxes result in smaller

grain sizes and more even coverage while slower rates allow the metal time to diffuse

along the nanotubes and coalesce into the larger grains seen in Figure 6-4b.76
























0 50 500 0 250 500

Figure 6-4: AFM images of Pd coated nanotube films a) sputtered and b) thermally
evaporated

To determine if the increase in nanotube resistance after sputter deposition was

confined to Pd we also sputter deposited a thin film of Au onto similar samples (in a

different sputtering system), again obtaining an increased resistance. In light of these

results we additionally conclude that sputtering damages the nanotubes changing their

intrinsic conductance. The three sputtered samples in Table 6-1 had the Pd deposited at

the same time so the amount of Pd was the same for each sample. The 25nm film had the

smallest amount of resistance increase, a factor of 1.6, while the micro-device resistance

increased by a factor of four. Every nanotube in the micro-device would be exposed to

and damaged by the sputtering plasma causing the largest change in resistance. In the

thicker sample (25nm film) many of the underlying nanotubes would be shielded from

the plasma by the tubes on top. Thus only a portion of the nanotubes would be damaged

resulting in a smaller increase in resistance.









6.4 Conductance Changes due to Hydrogen Exposure

Resistance changes in the nanotube devices upon exposure to H2 were monitored

for several different samples. Data was taken for samples both with and without the thin

Pd layer. We also compare the response of a 7nm film coated with thermally evaporated

Pd to that of a sputter coated film. Additionally, the response to H2 of a thin Pd film

without nanotubes was monitored.

The samples were exposed to H2 at room temperature and atmospheric pressure in

a quartz flow tube with electrical feed-throughs for voltage and current leads. The gasses

were fed via mass flow controllers to maintain a total flow of 450 sccm of either pure

nitrogen, 500 ppm hydrogen in nitrogen or a mixture of the two to obtain reduced

concentrations of hydrogen, or compressed air humidified by passing through a water

bubbler. Electrical measurements were performed with an HP4156B source-meter.

6.4.1 Pure SWCNT Samples

A micro-device sensor, a 7nm film and a 25nm film without any additional metal

besides the contact pads were all tested for sensitivity to H2 (Figures 6-5, 6-6 and 6-7

respectively). The current was monitored as a function of time at constant bias (0.5V) for

varying H2 concentrations. At t=14minutes the samples were exposed to air to recover

their initial conductance.

The pure nanotubes in the micro-device sensor show absolutely no sensitivity to

H2. At the highest H2 concentration of 500ppm the uncoated 7nm and 25nm films show

decreases in the current of 1.5% and 1.6% respectively. These comparatively small

changes (compare below) are probably entirely due to changes occurring within the

nanotubes at the Pd/nanotube contacts as opposed to the bulk uncoated nanotube film.











0.350

0.345

n ~An


I.


micro-device without Pd






-V- 10ppm
A-- lOOppm
-*- 250ppm
.-- 500ppm


Time(min)

Figure 6-5: Current vs. time measurement for micro-device sensor exposed to different
H2 concentrations.


0.179


0.178


0.177


0.176


0.175


0.174


5 10 15


Time(min)

Figure 6-6: Current vs. time measurement for 7 nm film exposed to different H2
concentrations.


S0.335

0.330
U


0.325

0.320


I














0.94





) 0.93
-V- lOppm
-- 100ppm
--- 250ppmr
--- 500ppm
0.92 I ,
0 5 10 15
Timne(min)

Figure 6-7: Current vs. time measurement for 25nm sensor exposed to different H2
concentrations..

6.4.2 SWCNT Samples Coated with Sputtered Pd

The three sputter coated samples listed in Table 6-1 were tested for their response

to hydrogen (Figures 6-8, 6-9, and 6-10). Again, the current vs. time for a 0.5 V bias was

monitored during exposure to hydrogen. The samples were exposed to air at t=14 min

to recover their nearly original conductance. A summary of AI/I at t=14 min

and 500 ppm H2 concentration is listed in Table 6-3.

The Pd coated sensors all show decreased conductance upon exposure to H2. The

7 nm film shows the highest sensitivity while the micro-device sensor shows the lowest.

Optimization of the sensitivity would occur by finding the optimum ratio of Pd to

SWCNTs. The micro-device sensor appears to have too much Pd for the amount of

nanotubes while the 25 nm film probably doesn't have enough.
















0.105






0.100


5 10 15


Time(min)

Figure 6-8: Current vs. time for micro-device sensor with sputtered Pd for different H2
concentrations.


0.09


0.08



0.07


0.06


5 10 15


Time(min)

Figure 6-9: Current vs. time for 7 nm film with sputtered Pd for different H2
concentrations















0.45




0.40

.v 10Oppm
-A-- OOppm
0,35 -- 250ppm
500ppm
I*I I
0 5 10 15

Time(min)

Figure 6-10: Current for time for 25 nm film with sputtered Pd for different H2
concentrations.

Table 6-3: Conductance changes in sputtered Pd/SWCNT samples at t=14 min for
500 ppm H2.
Sample AI/I (%)
Micro-device 5.6
7 nm film 30.
25 nm film 25.9


6.4.3 SWCNT Samples Coated with Thermally Evaporated Pd

A 7nm film with a non-percolating layer of thermally evaporated Pd was tested for H2

sensitivity. The current vs. time graph for 500ppm H2 and 0.5V bias voltage is shown in

Figure 6-11. The total change in resistance is 24%, which is smaller than the sputtered

Pd/7nm film sample (30.4%). However, the response rate is much faster. Within one

minute this sample shows a change in conductance of 12% (50% of the total change). At

t=5 minutes the current has changed by 23% (96% of the total change). By comparison









the sputtered 7nm film shows a AI/I of only 1.2% (or 3.9% of the total change) at t=l

minute. After 5 minutes the change is 18.5% (60.9% of the total resistance change).

The recovery rate is also much faster. One minute after exposure to air the thermally

evaporated Pd/SWCNT sample recovers 77.4% of the H2 induced conductance change.

After 5 minutes the recovery is 95%. The sputtered Pd/SWCNT sample only recovers

35.9% after 1 minute and 45.7% after 5 minutes. The differences in the total change

between the sputtered and evaporated samples might be due to different amounts of Pd.

The rapid response of the evaporated Pd/SWCNT film makes it a more promising

candidate for real world applications. Optimizing the amount of Pd might increase its

sensitivity to match the sputter Pd/SWCNT sensor.




500 ppm H2
0.14 -A- 7nm film with thcnnall evaporated Pd




0.12 A




0.10 -


0 5 10 15 20

Time(min)

Figure 6-11: Current vs. time for 7nm film with thermally evaporated Pd exposed to
500ppm of H2.









6.4.4 Thin Pd Film

The current vs. time was monitored for a very thin layer of thermally evaporated Pd

to 500ppm of H2, Figure 6-12. The initial resistance of the film was 55.6 kM. The

current first increased and then decreased upon exposure to H2. After the sample was

exposed to air the current increased again, without recovery of the initial resistance.

These results are very different than those exhibited by the Pd/SWCNT samples. From

this we can conclude that the effects seen in the Pd/SWCNT samples are due to changes

in the resistance of the nanotubes and are not simply an effect cause by changes in the Pd

resistance.





11 500 ppm H2
Air







9- 1
-*-- Pd thin film

0 5 10 15 20 25 30
Tirne(min)

Figure 6-12: Current versus time for a thin Pd film.

6.4.5 Conclusion

Carbon nanotubes coated with Pd have been shown to be sensitive to H2 levels as low

as 10ppm within 10 minutes exposure. Although nanotube samples with Pd deposited by






83


both sputtering and thermal evaporation show sensitivity to H2, the latter shows a much

faster response and recovery. It has been shown that nanotubes subjected to sputtered

deposition of metals exhibit a decrease in their conductance, likely due to damage.














APPENDIX
SPECTROELECTROCHEMICAL STUDY OF CARBON NANOTUBE THIN FILMS

Spectroelectrochemistry is a field of study which combines spectroscopy with

electrochemistry. Spectroelectrochemistry can provide valuable information about thin

films, such as the effects of different applied voltages on the optical transmittance of the

thin film. Aqueous electrolyte is commonly used in many electrochemical applications.

However, water is not ideal in spectroelectrochemistry because it has strong absorption

bands which make it almost impossible to gain information about a thin film at certain

wavelengths. Additionally, water has a narrow potential window. Water can only be

taken to -0.81V or +1.23V before it breaks down into its byproducts (H2 gas and OH-

ions at negative voltage, 02 gas and H+ ions at positive voltage).77 To circumvent these

types of problems, a nonaqueous solvent must be used.

The experiment was performed using the nonaqueous solvent acetonitrile. And

involved a SWCNT thin film. In this experiment potentials were applied to the thin film

(versus a platinum wire counter electrode using a silver wire pseudoreference electrode)

while changes in the transmittance spectrum were monitored. We infer that these

changes in the spectrum occur as a consequence of changes in the Fermi level. From this

inference, we are able to locate the intrinsic doping level of the semiconducting

nanotubes.

A chemically resistant cell and identical reference cell were developed for use in

these experiments. One of these cells is displayed in Figure A-1. The bodies of the cells

are composed of Teflon. They use Simriz SZ485 O-rings and have quartz optical









windows. All of these components have high chemical resistances to most substances.

Nylon tubing was used for getting solutions into and out of the cells. Additionally,

Teflon valves were used to control the flow of the solution, and a steel poppet valve

was attached so that any gasses produced would be able to leave the cells without

allowing atmospheric contaminants to enter. There were three types of electrodes used in

the cell. The working electrode was the film under study contacted by a graphite rod, the

counter electrode was platinum, and the reference electrode was silver. A Perkin-Elmer

{model} potentiostat supplied the potentials in a three terminal set-up, where in the

counter electrode voltage is varied to make the voltage difference between the working

electrode and reference electrode the desired value. The benefit of this is that the desired

potential on the working electrode is maintained versus a well-defined reference

independent of any Faradaic electron transfer occurring at the working electrode. The

cells were designed in an attempt to keep out water vapor and air. The cells served their

purpose well, and substantial impurity peaks were only observed after the cells were

exposed to the environment for periods of over 24 hours.

A.1 Carbon Nanotube Experiment

When taking a transmittance spectrum of a thin film of SWCNTs one observes

three broad peaks. Our study focuses on the peak centered at 1650nm. This peak

corresponds to photons absorbed during electronic transitions from the highest occupied

semiconducting valence band (VI) to the lowest unoccupied semiconducting conduction

band (Cl).

Figure A-2 shows that the valence and conduction states for carbon nanotubes have

sharp peaks in the density of states known as Van Hove singularities. This would imply

very sharp absorption bands. However, a thin film of SWCNTs is composed of









semiconducting nanotubes of different diameters and chiralities. As a result of this

assortment of types of nanotubes, the peaks in the transmittance spectrum end up being

broad and smooth.











w "










Figure A-l: The spectroelectrochemical cell used in the experiments.


By applying a voltage to the SWCNT sample, one can shift the Fermi level. One

can see in Figure A-2a that if the valence state (VI) becomes more depleted, then fewer

electrons will be available to undergo the transition to the conduction state (Cl).

Therefore there will be less absorption of photon energy, and the absorbance peak will

diminish. Similarly, if the Fermi level is increased, as in Figure A-2b, so that some

electrons already occupy the conduction state (C ), then there will be fewer available

states for the remaining valence state electrons to move into and the peak will also

diminish. Based on this knowledge, one can then determine the intrinsic doping level of

the semiconducting nanotubes. If the peak is largest at OV then the semiconducting










nanotubes will be intrinsically undoped. Using this method, one can only determine the

amount of doping qualitatively.



S (12,8 I ) (12,8
FmiFm Lewl


V2 C2 V2 C2
vi CI
*I I I I
Q I !~---~---^ --------



-1 0 1 -1 0
a. Energy (eV) bEnergy (eV)


Figure A-2: Diagram of the Fermi Level of a semiconducting (12,8) SWCNT which is a)
p-doped or b) n-doped

A.2 Experimental Section

This experiment is similar in nature to an experiment performed by Kavan et a/..78

Our experiment mainly differs from that one in that the SWCNT thin film that we use is

mounted directly on quartz instead of ITO (Indium Oxide doped with Tin Oxide). We

can do this because the nanotube samples also contain metallic nanotubes and we have

learned how to make ultrathin (and hence transparent) films of SWCNTs that are in

contiguous electrical contact with each other.69

A SWCNT film was prepared and attached to a piece of quartz. This sample was

baked in flowing argon gas in a Thermolyne 79300 tube furnace to drive off chemical

dopants. The temperature was ramped at 50C per minute until it reached 1100C, stayed at

1100C for 30 minutes, and then ramped at 10C per minute to 6000C. The temperature

remained here for 2 hours. A reference piece of quartz was also baked in the same way

to be used in a reference cell.