The Pennsylvania State University. The Graduate School. Department of Physics BORON DOPING OF SINGLE WALLED CARBON NANOTUBES.

Transcription

1 The Pennsylvania State University The Graduate School Department of Physics BORON DOPING OF SINGLE WALLED CARBON NANOTUBES A Dissertation in Physics by Xiaoming Liu 2009 Xiaoming Liu Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy December 2009

2 The dissertation of Xiaoming Liu was reviewed and approved* by the following: Nitin Samarth Professor of Physics Dissertation Advisor Chair of Committee Jorge O. Sofo Associate Professor of Physics Associate Professor of Materials Science and Engineering Vincent H. Crespi Professor of Physics Professor of Materials Science and Engineering John V. Badding Professor of Chemistry Jayanth R. Banavar Professor of Physics Head of the Department of Physics *Signatures are on file in the Graduate School

3 iii ABSTRACT This thesis addresses the improvement of electrical properties of single walled carbon nanotubes (SWNTs), grown by different techniques, by a stable substitutional boron doping. The B-SWNTs material was created in four different ways, and <1 at.% boron was found in the SWNTs bundles with electron energy loss spectroscopy (EELS). The p-type doping is confirmed by Raman spectroscopy. The Raman D-band results and the line shape of EELS spectra show that boron is fitted into the sp 2 lattice of the tube wall. B-doping was found to downshift the positions of the optical absorption bands associated with van Hove singularities (E11s E22s and E11m ) by ~40 mev relative to their positions in the acid treated and annealed SWNTs. Optical transmission studies show that B-doping increases the number of free carriers, and that it does not significantly affect the optical transmittance of the B-SWNTs films in the visible region. We found that boron-doping lowers the sheet resistance by 75% of the pristine SWNTs films. Besides, boron-doped SWNTs may provide a better modification of the semi-conducting SWNTs in a stable way. We suggest that B-SWNTs films may be a potential material for future electronic devices, such as touch-screens, organic electronics device, organic solar cells and flexible display screens, because of its strengths, including transparency, high T/R Ñ and flexibility, low cost, and easy deposition on plastic substrates.

4 iv TABLE OF CONTENTS LIST OF FIGURES... vi LIST OF TABLES... xii ACKNOWLEDGEMENTS... xiii Chapter 1 Introduction Motivation Structure of carbon nanotubes Electronic and phonon states of SWNTs Raman scattering from carbon nanotubes Chapter 2 Structural, Optical and Electrical Characterization Techniques Transmission electron microscopy (TEM) Electron energy loss spectroscopy (EELS) Fourier transform infrared spectroscopy (FTIR) λ 950 double beam spectrometer Thermogravimetric ananlysis (TGA) Micro-Raman spectrometer: Jobin-Yvon T Renishaw Invia Raman microscope Four-probe resistance measurements Chapter 3 Synthesis and Purification Methods Vapour-Liquid-Solid (VLS) model Arc-discharge method Pulse laser vaporization (PLV) Chemical vapor deposition (CVD) Sample purification Boron doping of SWNTs Oxidation of Carbon by B 2 O Electronic structure of boron-doped SWNTs Chapter 4 Simultaneous Growth and Doping of SWNTs by ARC Discharge Sample growth Temperature programmed oxidation (TPO) study Raman study Resistance measurement Summary Chapter 5 Post-Growth B-Doping in Reactive Environments (NH 3 )... 68

5 v 5.1 Sample preparation TEM and EELS studies Raman measurement Transmission spectrum Summary Chapter 6 Post-Growth B-Doping by Thermal Evaporation of B 2 O Post-growth boron doping EELS measurement Raman spectroscopy Resistance measurement Chapter 7 Post-Growth B-Doping by High Temperature Treatment of Boron Oxide Decorated SWNTs Sample preparation Sheet resistance measurment Summary Chapter 8 Correlation Between Optical and Electrical Properties UV-Vis Transmission measurement Sheet resistance measurement Correlation between optical and electrical properties Chapter 9 Temperature Dependent Resistance of SWNTs and B-SWNTs Introduction Fluctuation induced tunneling (FIT) Mott variable range hopping (VRH) Summary Chapter 10 Far IR Measurement Introduction to free carrier absorption Far IR transmission measurement Chapter 11 Conclusion and Future Direction BIBLIOGRAPHY

6 vi LIST OF FIGURES Fig. 1-1: Formation of an ideal single wall nanotube (SWNT). OA is the chiral vector C h for a (4,2) tubule. OB is the translational vector T orthogonal to C h. θ is the chiral angle for (4,2) tubule and OBB A is a unit cell of a (4,2) tubule.[25]... 3 Fig. 1-2: Different kinds of nanotubes: armchair, zigzag, and chiral nanotubes.[27]... 5 Fig. 1-3: 2-D graphite band structure in the first Brillouin zone, constructed using Eq The conduction and valence bands touch at the six Fermi points K indicated at E = Fig. 1-4: Energy dispersion of grapheme and electronic band structure of a SWNT (a) energy dispersion relation for graphene with N discrete parallel cutting planes in K 2 direction. (b)electronic band structure for (10,10) armchair carbon nanotube[29] Fig. 1-5: Correlation of grapheme and SWNTs band structure. Tight-binding band structure of graphene, showing the main symmetry points and the allowed k-vectors of (5,5), (7,1) and (8,0) tubes (solid lines) mapped onto a graphene Brillouin zone Fig. 1-6: One-dimensional energy dispersion relations for (a) armchair (10,10) tubes, (b) zigzag (10,0) tubes, and (c) chiral (6,4) tubes and their unit cell, from zone-folded tight-binding dispersion relations.[30] Fig. 1-7: STM imaging and IV spectroscopy of SWNTs. (a) A comparison of the DOS obtained from an experiment (upper curve) and a π-only tight-binding calculation for a (13,7) SWNTs (lower curve). The broken vertical lines indicate the positions of van Hove singularity in the tunnelling spectra after consideration of thermal broadening convolution. The inset shows an atomic resolution image of the (13,7) tube. (b) A comparison of the DOS obtained from an experiment (upper curve) and a calculation for a (10,0) SWNTs (lower curve). The inset shows an atomically resolved image of the (10,0) SWNTs[34] Fig. 1-8: Kataura plot of the interband transition energies Ejj Vs nanotube diameter. The two vertical lines represent our diameter distribution (1.2~1.8nm). the horizontal lines are the excitation laser lines Fig. 1-9: Raman spectra, (a) first-order and (b) one-phonon second-order, (c) two-phonon second order resonance Raman spectral processes. (top) incident photon resonance and (bottom) scattered photon resonance

7 vii conditions. For one-phonon, second-order transitions, one of the two scattering events is an elastic scattering event (dashed lines). Resonance points are shown as solid circles[36] Fig. 1-10: Typical Raman spectra of bundles of SWNTs on quartz substrate Fig. 1-11: Atomic displacements of carbon related to RBM feature of SWNTs Fig. 1-12: The G-band for one semiconducting SWNT and one metallic SWNT.[42] Fig. 1-13: Atomic displacements of carbon related to the tangential mode (Gband) of SWNTs Fig. 2-1: Idealised schematics of an EELS spectrum, indicating Zero-Loss peak, Plasomon resonance, and core-loss electron peak[45] Fig. 2-2: Schematic representation of a Michelson interferometer Fig. 2-3: Schematics of Bruker V80 FTIR spectrometer Fig. 2-4: Transmission spectra of a silicon window Fig. 2-5: Transmission spectra of a slab of ZnSe windlow. The inset is a picture of the ZnSe substrate used in this work Fig. 2-6: Transmission of 1mm thick quartz substrate Fig. 2-7: Schematic diagram of the λ950 optical system Fig. 2-8: Picture of TA Q5000 system Fig. 2-9: Schematic diagram of T64000 emphasizing the optics of the confocal microscope used to collect Raman spectra of SWNTs. The objective spot size is around 1.0μm. Confocal aperture is used to define the scattering volume and remove stray light [49] Fig. 2-10: Schematic representation of the T6400 double subtraction premonochromator with spectrograph[49] Fig. 2-11: Schematic of the Renishaw Invia Raman microscope laser beam pass Fig. 2-12: Four probe inline measurement Fig. 2-13: Picture of the automated temperature dependent sheet resistance measurement setup Fig. 2-14: Program measuring the sheet resistance as a function of temperature for the SWNTs film

8 Fig. 3-1: Illustration of the concentration gradient across the catalytic metal particle during SWNT growth. Incorporation of atoms into the SWNT structure lowers the C concentration near the open SWNT end.[55] Fig. 3-2: Schematic diagram of the arc-discharge apparatus[57] Fig. 3-3: TEM image of raw arc SWNTs Fig. 3-4: Schematic diagram of the pulsed laser vaporization (PLV) apparatus Fig. 3-5: Purification setup (a) reflux for wet oxidation with H 2 O 2 and acid treatment. (b) vacuum filtration setup (c) cross-flow filtration Fig. 3-6: (a)tem image if raw PLV SWNTs and purified SWNTs (b)tpo analysis of purified SWNTs Fig. 3-7: Formation energy E as a function of nanotube diameter. (m,0) zigzag tubes of m=4-12 (filled circles) and (m,m)armchair tubes of m=4,5, and 6 (open circles) are plotted.[77] viii Fig. 3-8: Band structures and DOSs of pristine and B-doped (10,0) SWNT: (a) C 40, (b) BC 39, (c)) BC 79,(d)) BC 119, and (e) ) B 2 C 78.[77] Fig. 4-1: TEM images of the material produced from electrodes with boron concentration of (a)1at.%, (b)2 at.% and (c)3 at.% Fig. 4-2: TEM image of material produced from electrodes with boron concentration of 5 at.% Fig. 4-3: TEM image of material produced from electrodes with boron concentration of 10 at.% Fig. 4-4: HRTEM image and EELS spectrum of B-SWNTs. (left) HRTEM image of B-SWNTs bundles treated with H 2 O 2 for amorphous carbon removal. (right) EELS spectrum of the bundle shown in the left image Fig. 4-5: TPO and DTPO data of (a)swnts and (b)1%b-swnts Fig. 4-6: Room temperature Raman spectra showing (a) D-band(~1340 cm -1 ),Gband (~1590 cm -1 ) and (b) second-order G band. The excitation laser is 514.5nm and spectra were normalized to the tangential G+ band intensity Fig. 4-7: Room temperature Raman spectra showing Radio breathing mode (RBM) using laser with excitation energy of (a) 2.54eV (488nm) and (b) 2.41eV (514.5nm)... 61

9 ix Fig. 4-8: Room temperature Raman spectra showing (a) D-band (~1350 cm -1 ), G-band (~1590 cm -1 ) and (b) second-order G -band(~2680cm -1 ). The excitation laser is at 488nm Fig. 4-9: Raman spectra (T=300K) from various sp2 carbon using Ar-ion laser excitation: (a)highly ordered pyrolytic graphite (HOPG), (b) boron-doped HOPG (BHOPG), (c) carbon nano particles derived from the pyrolysis of benzene and graphitized at 2820 o C, (d) as-synthesized carbon nanoparticles (~850 o C), (e) glassy carbon[84] Fig. 4-10: Room temperature Raman spectra showing R-band(~160cm-1), D- band (~1270 cm-1), G-band (~1590 cm-1) and (b) second-order G - band(~2530 cm-1). The excitation laser is at 1064nm and spectra were normalized to the tangential G+ band intensity Fig. 5-1: TEM images of B-doped and undoped SWNTs bundles, (a) TEM image of undoped SWNTs bundles. (b) TEM image of B-SWNTs. Dark dots are residual NiY catalyst (c) HRTEM image of one B-SWNTs bundle (d) EELS spectrum of the bundle shown in (c) Fig. 5-2: Room-temperature Raman spectra showing radial R-band (~160cm -1 ), D-ban (~1350 cm -1 ) and G-band (~1590 cm -1 ) of (a) as-prepared (b) after HNO 3 reflux (c) after HNO3 reflux NaOH wash to ph=7 and 1100 o C vacuum anneal (d) after B 2 O 3 -NH 3 treatment (e) afte B 2 O 3 -NH 3 treatment and subsequent 1100 o C anneal Fig. 5-3: Optical density of thin SWNTs films on ZnSe (a) p-swnts (b) after B 2 O 3 -NH 3 processing and a vacuum anneal at 1100 o C Fig. 6-1: (a) HRTEM image of purified undoped SWNTs bundles. (b) low resolution TEM image of purified SWNTs Fig. 6-2: Schematics of the CVD setup for doping pre-deposited SWNTs. A half inch tube is placed in a 1 inch diameter quartz tube. Substrate with nantubes pre deposited on is placed closed to the open end of the tube. B 2 O 3 powder was located at the closed end of the quartz to have a higher gas pressure. A flow of 100 sccm of Argon with 10% Hydrogen was used to have a pressure gradient along the tube Fig. 6-3: Schematics of a nantube with some vacancies on the tube wall in interacting with boron oxide. The final produce is a nanotube with boron substituting the carbon on the tube wall Fig. 6-4: (a) HRTEM image of B-SWNTs bundle (b) EELS spectrum of the bundle.. 84 Fig. 6-5: Room-temperature Raman spectra showing radial R-band (~184cm-1), D-band (~1350 cm-1) and G-band (~1590 cm-1) of annealed SWNTs and after B2O3 treatment B-SWNTs

10 x Fig. 7-1: Schematics of the CVD setup for doping SWNTs. A flow of 100 sccm of Argon with 10% Hydrogen was used Fig. 7-2: TEM and HRTEM images of the SWNTs bundles, (a) TEM image of bundled SWNTs. (b)hrtem image of several p-swnt bundles.(c) EELS spectra of B-SWNT bundle Fig. 8-1: Spray setup and films of nanotube deposited on quartz substrate Fig. 8-2: optical transmittance of SWNT films with the same thickness. One went through a doping process the other went through a high temperature annealing(1000 o C) Fig. 8-3: Schematic diagram of resistance measurements probe for temperature range 4-500K Fig. 8-4: Schematic view of van der Pauw contacts (1,2,3,4)to SWNT film Fig. 8-5: Sheet resistance VS degassing time (200 o C in 10-7 torr vacuum) for both B-SWNTs and undoped SWNsT samples Fig. 8-6: Transmittance (at 550nm) and sheet resistance of B-doped SWNT (doped as in chapter 5 with the presence of NH 3 )and undoped p-swnt films (processed Arc SWNTs). Insert: Transmittance in the visible range for undoped SWNT p-films Fig. 8-7: Transmittance (at 550nm) and sheet resistance of B-doped (doped as in chapter 6) SWNsT and undoped p-swnts films (purified PLV SWNTs) Fig. 8-8: Transmittance (at 550nm) and sheet resistance of B-doped SWNT (doped as in chapter 7) and undoped p-swnt films (purified PLV SWNTs) Fig. 8-9: Reflection and transmission through a slab with thickness d Fig. 8-10: Schematic percolation of a two-dimensional nanotube network made of SWNT bundles Fig. 9-1: Schematic diagram of carbon nanotube network and bundle-bundle junction. M stands for metallic tubes and S stands for semiconducting tubes Fig. 9-2: Closed cycle refrigerator (CCR) Fig. 9-3: Temperature dependent data of boron doped SWNTs (black dots) and undoped SWNTs (red diamond) the solid line through the data is the fitting using just fluctuation induced tunneling. Both film has a transmission of 63% at 550nm

11 Fig. 9-4: Normalized Sheet resistance data ( K) films of B-SWNTs and SWNTs film with 63% transmission at 550nm Fig. 9-5: Fitting of doped (black) and undoped(red) temperature dependent resistance data with VRH (blue) and a linear term (bottom red line) Fig. 9-6: Plot of metallic tube contribution in the film sheet resistance using fitting parameters in Tab Fig. 10-1: Optical density of p-swnts and B-SWNTs, (a) Optical density of p- SWNT (red) and B-SWNT(black) thin films on Si substrate. The green line is the difference between the two samples and the thin red line is the Drude model fitting (b) comparison betweenthe O.D. of p-swnt and p-swnt decorated with BnOm+ ion xi

12 xii LIST OF TABLES Tab. 3-1: Standard Gibbs energy, standard entropy ( S 0 298) and equilibrium temperature of B 2 O 3 consumption reactions[72] Tab. 3-2: Electronc structure of a B-doped SWNTs[77] Tab. 4-1: Sheet Resistance of undoped, 1%B-SWNTs, 2%B-SWNTs and 3%B- SWNTs Tab. 6-1: Sheet Resistance of B-SWNTs and SWNTs films Tab. 7-1: Sheet Resistance of undoped and B-SWNTs of difference thickness Tab. 9-1: Fitting parameters of SWNTs films with different optical transmission using fluctuation-induced tunneling Tab. 9-2: VRH Fitting parameters of B-SWNTs and SWNTs

13 xiii ACKNOWLEDGEMENTS I would like to give special thank to my advisor, Peter for his great care and constant encouragement during the past five years. I thank Peter for giving me a chance to enjoy great science. It was his patience, encouragement and support guided me through my research adventure. I also have to thank my committee numbers: Prof. Nitin Samarth, Prof. Vincent H. Crespi, Prof. Jorge O. Sofo and Prof. John V. Badding for their help and suggestions in my research. I wanted to thank my colleagues, Dr. Humberto R. Gutierrez and Dr.Hugo E. Romero, for helping me, sharing with me their experience and giving me helpful advice. I must also thank my group members who gave me a lot of help. They are Dr. Kofi C. Adu, Dr. Un J.Kim, Dr. Jian Wu, Dr. Timothy J. Russin, Mr. Qiujie Lv, Ms. Bei Wang, Mr. Qingzhen Hao and Mr. Duming Zhang. Without such amazing colleagues my research would not have been nearly as fun nor as productive. Finally, I would like to thank my parents for their constant support, love and encouragement. I dedicate my thesis to my father who himself is a good physicist. I chose to follow his path mainly because I had so much fun playing in his lab and learning physics.

14 1 Chapter 1 Introduction 1.1 Motivation A SWNT[1] can be viewed as a graphene sheet rolled into a cylinder. The SWNTs derived from large batch processes, such as the Arc or Plasma CVD, are composed of bundles of tubes with a somewhat random distribution of chiral integers (n,m) in which ~1/3 are metallic and ~2/3 are semiconducting nanotubes. SWNT is a material with high electron mobility at room temperature. Mobility of semiconducting SWNTs is reported to be an order of magnitude larger than that of crystalline Si [2, 3]. This makes SWNTs a suitable candidate for electronic devices. Field-effect transistors with individual p-type semiconducting SWNTs have been reported [4-6]. But one major obstacle to the mass production of these devices is the lack of a method to deposit SWNTs on a wide variety of substrates with precise control over the density, position, and orientation of the SWNTs. One way to circumvent this is the use of electrically continuous network of SWNTs, as reported by Snow et al.[7]. After the density of the bundles exceeds the percolation threshold the nanotubes will be interconnected and form a continuous electrical path. Each filament in the percolating network can be comprised of a bundle of 10 s to 100 s of SWNTs, as is the case in this study, or the filament might be an individual tube. The networks made by these SWNTs bundles behave like a p-type semiconductor with a field-effect mobility of ~12cm 2 /Vs and a transistor on-to-off ratio~10 4 [8]. The p-type behavior is due to adsorbed O 2 [9]. Obtaining p-type material and controlling their charge carrier

15 2 densities is crucial to current microelectronics. Substitutionally doping boron will make these SWNTs a stable p-type semiconductor more suitable for such applications. Other advantages of this material include its high conductivity, its mechanical flexibility and its ability to fabricate films at room temperature on a variety of substrates. These make it perfect for transparent electrode applications. The goal is to improve the figure of merit FOM=T/R Ñ of these percolating bundled SWNTs films via boron doping, where T is the optical transmission 550 nm for visible applications) and R Ñ is the sheet electrical resistance. For a network made of highly one dimensional tubes, the transparency approaches 100% as the aspect ratio of the wire approaches infinity. The electrical conductance of these networks is related to two factors: one is the SWNTs bundle conductivity, since the higher bundle conductivity the better the network conductance; the other is the junction resistance between bundle-bundle contacts. P-type doping by boron will help improve the electrical conductance of the network. Chemical doping of various forms of sp 2 carbon have been the subject of considerable interest over the past 3 decades beginning with graphite intercalation compounds, where doping by reaction with ASF 5 produced an electrical conductivity better than copper[10]. This startling discovery was followed by chemical doping studies in carbon fibers[11]. Later, similar studies in the more modern forms of sp 2 carbon, i.e., fullerenes and single-wall carbon nanotubes were done[12]. The present work emphasizes the conductivity effects of the substitutional doping in SWNTs, where the dopant must substitute for carbon in the sp 2 lattice. Boron and nitrogen are the only two elements that can be incorporated into an sp 2 carbon network without significantly affecting the atomic arrangement in the hexagonal 2D lattice. Previous experimental studies regarding the possibility of B-doping in SWNTs have been

16 3 made[13-22]. Boron substitution has been reported to occur during[13-16, 22] or after[17-21] the growth of the SWNTs. Early studies have shown that boron can be substituted at the few % level into the graphene sheets in graphite[23]. Thus, it is reasonable to expect that SWNTs can be made p-type by substituting boron for carbon in the nanotube wall[13]. 1.2 Structure of carbon nanotubes SWNTs were first observed in 1993 by Iijima[24]. SWNTs are cylindrical shaped with the two ends open or capped with hemispherical fullerens[12]. The structure of the tube body can be viewed as a rolled up graphene sheet depicted in Fig The unit vectors of the lattice are defined by a 1 and a 2 with the C-C bond length being 1.42 Å. The geometry of a SWNT can be specified by a pair of integers (n,m), denoting the relative position C h (chiral Vector) of the pair of atoms on a graphene sheet which rolled onto each other to form a tube wall. C h = na 1 + ma 2 (n,m) 1-1 Fig. 1-1: Formation of an ideal single wall nanotube (SWNT). OA is the chiral vector C h for a (4,2) tubule. OB is the translational vector T orthogonal to C h. θ is the chiral angle for (4,2) tubule and OBB A is a unit cell of a (4,2) tubule.[25]

17 4 Here n and m are integers with n>m 0. The SWNT is constructed by rolling the graphene layer along a certain direction C h = n a 1 + m a 2 making OB and AB' coincide. Perpendicular to C h is the translational vector T given by: T=t 1 a 1 +t 2 a This vector points to the long axis of the SWNT. C h and T are referred to as the chiral vector and translational vector, respectively. Together they define the unit cell of a SWNT as the rectangle OAB'B. SWNTs can be classified according to how they are mapped into a single graphene layer. The indices (n, m) are commonly used to label SWNTs. Chiral angle θ (Eq.1-3) between a 1 and C h can be used to classify SWNTs: cos Due to the C 6 symmetry of the hexagonal lattice, the chiral angle θ 30 o. SWNTs with θ =30 o or θ =0 o have the highest symmetry (Fig. 1-2), known as armchair or zigzag nanotubes[26]. The rest are called chiral nanotubes. The Diameter of a SWNT is defined as: d 3, 1-4 where is the circumference of the tube.

18 5 Fig. 1-2: Different kinds of nanotubes: armchair, zigzag, and chiral nanotubes.[27] 1.3 Electronic and phonon states of SWNTs The electrical properties of nanotubes depend sensitively on the geometrical structure (n,m). Although graphene is a zero-gap semimetal, the calculated electronic structure of SWNTs show that carbon nanotubes can be either metallic or semiconducting depending on the choice of (n,m) [25]. A simple approximation for the band structure of nanotubes is obtained using the band structure of 2D graphite. From

19 6 the nearest-neighbor tight-binding approximation one can get the energy dispersion relation Eq.1-5: cos cos 2 2 4cos 2, 1-5 where Å and γ ~2.9eV is chosen to compute the π bands of graphene in the first Brillouin zone and the results are shown in Fig The π* antibonding band and the π bonding band, respectively, form the conduction and the valence bands of graphene. Since there are two atoms per unit cell in a graphene sheet, the valence band is completely filled. Only the π electrons contribute to the graphene electrical conduction. The conduction and the valence bands touch each other at the six corners points (K points) of the hexagonal Brillouin zone, where the Fermi energy E F = 0. A graphene sheet is therefore semimetallic. Fig. 1-3: 2-D graphite band structure in the first Brillouin zone, constructed using Eq The conduction and valence bands touch at the six Fermi points K indicated at E = 0.

20 7 From Eq.1-5, 1D dispersion relations for carbon nanotube (n,m) can be calculated based on a simple zone folding consideration, i.e., by imposing a periodic boundary condition around the waist of a SWNT. The allowed wave vectors k in the direction parallel to the chiral vector, resulting from radial confinement are: 2, 1-6 where q is an integer. The 1D energy dispersion curves of a nanotube correspond to the cross-section of the 2D energy dispersion surface shown in Fig.1-3, where the cuts are made on along parallel lines corresponding to the particular set of allowed states Fig. 1-4 [25]. In Fig. 1-5 several cutting lines, representing the allowed subbands of a nanotube, are shown[28]. The Brillouin zone of SWNT is a collection of N (Eq.1-6) 1D slices through the k space of graphene, each with a length of 2π/T, where T (Eq.1-6) is the length of the nanotube unit cell and N is the number of grapheme unit cells and d R is the greatest common divisor (gcd) of (2m+n) and (2n+m):,

21 8 Fig. 1-4: Energy dispersion of grapheme and electronic band structure of a SWNT (a) energy dispersion relation for graphene with N discrete parallel cutting planes in K 2 direction. (b)electronic band structure for (10,10) armchair carbon nanotube[29]. Fig. 1-5: Correlation of grapheme and SWNTs band structure. Tight-binding band structure of graphene, showing the main symmetry points and the allowed k-vectors of (5,5), (7,1) and (8,0) tubes (solid lines) mapped onto a graphene Brillouin zone.

22 9 On the basis of this simple scheme, if one of the allowed wave vectors passes through a Fermi point of the graphene sheet, the SWNTs should be metallic with a nonzero density of states at the Fermi level. If the K point of the 2D Brillouin zone is located between two cutting lines, the position is always one-third of the distance between two adjacent lines. This will give us a semiconducting nanotube with a finite energy gap. It is important to note that the states near the Fermi energy in both metallic and semiconducting tubes result from states near the K point, and their transport and other properties are related to the properties of the states on the allowed lines. The 1D energy dispersion relation of a (n,m) nanotube is given by Eq.1-9, 1 4cos 2 cos 2 4 cos For armchair nanotubes the energy dispersion is as Eq. 1-10, 1,, 1 4cos cos 4 cos For zigzag nanotubes the energy dispersion is as Eq. 1-11, 1,, 1 4cos 3 cos 2 4 cos Fig. 1-6 The band structures of (10,10), (10,10) and (6,4) are constructed and shown in

23 10 Fig. 1-6: One-dimensional energy dispersion relations for (a) armchair (10,10) tubes, (b) zigzag (10,0) tubes, and (c) chiral (6,4) tubes and their unit cell, from zone-folded tight-binding dispersion relations.[30] The electronic densities of states calculated from a simple tight-binding model shows characteristic singularities (van Hove singularities (vhss)) in Fig This is also determined by experimental results. The DOS in a (13,7) metallic tube was reported by Kim et al[31]. Odom et al[32] showed the DOS of a semiconducting tube with indices (10,0). Resonances in Raman scattering experiments have provided evidence for such sharp peaks in the DOS of nanotubes[33].

24 Fig. 1-7: STM imaging and IV spectroscopy of SWNTs. (a) A comparison of the DOS obtained from an experiment (upper curve) and a π-only tight-binding calculation for a (13,7) SWNTs (lower curve). The broken vertical lines indicate the positions of van Hove singularity in the tunnelling spectra after consideration of thermal broadening convolution. The inset shows an atomic resolution image of the (13,7) tube. (b) A comparison of the DOS obtained from an experiment (upper curve) and a calculation for a (10,0) SWNTs (lower curve). The inset shows an atomically resolved image of the (10,0) SWNTs[34]. 11

25 Raman scattering from carbon nanotubes Resonant Raman scattering is an excellent probe for investigating the electronic and phonon structure in pristine and doped SWNTs materials due to the resonant coupling of the laser excitation energy to the transition energies between the van Hove singularities DOS [25, 33]. Since the gap between the vhss in the DOS differ from tube to tube based on the tube diameter and chiral angle, specific (n,m) tubes can be studied in this technique. Fig. 1-8 shows a Kataura plot (γ 0 = 2.90 ev), which is a plot of vhss transitions energies vs. the tube diameter[35]. This plot is a useful tool in determining whether semiconducting (red dots) or metallic SWNTs (black dots) will be in resonance in the Raman spectra for a given laser excitation wavelength. The notation, E s(m) jj, is used to identify a particular transition energy (j = 1,2,3,...) in a semiconducting (s) or metallic (m) SWNT. Two black vertical lines in Fig. 1-8 give the range of diameters (1.2~1.8nm) distribution of the tubes used in this work while the horizontal lines indicate the laser energies (2.54, 2.41, and 1.17 ev) used for exciting the Raman spectra (488nm, 514.5nm, and 1064nm). Using the known diameter distribution of the sample and the excitation energies of the sources, the 514.5nm/1064nm and the 488nm laser excitations will couple resonantly with semiconducting SWNTs. The Raman scattering experiments using 514.5nm and 488nm laser were performed at room temperature in air using the JOBIN YBON RAMANOR T The 1064nm data are acquired with a BOMEM FT-Raman system.

26 13 Fig. 1-8: Kataura plot of the interband transition energies Ejj Vs nanotube diameter. The two vertical lines represent our diameter distribution (1.2~1.8nm). the horizontal lines are the excitation laser lines. The process of Raman scattering in SWNTs can be described as follows: (1) The incident photon excites the electron from valence band to the conduction band. (2) Then the excited electron is scattered by emitting (stokes process) or absorbing (antistokes process) phonons [36]. The number of scattering events occurring before the electron relaxing back to its initial state defines the order of the Raman scattering (Fig. 1-9).

27 14 Fig. 1-9: Raman spectra, (a) first-order and (b) one-phonon second-order, (c) twophonon second order resonance Raman spectral processes. (top) incident photon resonance and (bottom) scattered photon resonance conditions. For one-phonon, second-order transitions, one of the two scattering events is an elastic scattering event (dashed lines). Resonance points are shown as solid circles[36] In SWNTs the strongest first-order Raman modes are the radial breathing mode (RBM) and the G-band (Fig. 1-10). The RBM band is related to the coherent vibration of the carbon atoms in the radial direction shown in Fig Fig. 1-10: Typical Raman spectra of bundles of SWNTs on quartz substrate.

28 15 Fig. 1-11: Atomic displacements of carbon related to RBM feature of SWNTs The RBN feature is most sensitive to tube diameter. The relation between RBM frequency and tube diameter is Eq. 1-7, where A and B are determined by experiment [37, 38]. For bundles of SWNTs with diameter of d t = nm, A=234 cm -1 nm and B=10 cm -1 nm was reported[38]. And A=248 cm -1 nm and B=0 has been found for isolated SWNTs on Si substrate [37, 39]. The RBM spectra for SWNT bundles contain an RBM contribution from different SWNTs in resonance with the excitation laser line[36]. This can be determined from the Kataura plot. ω RBM A d B 1-7 The G-band is the tangential mode (Fig. 1-13). This feature is generally related to the sp 2 carbon-carbon bond in-plane vibrations. In most Raman spectra for SWNTs we can observe that the G-band is composed of several peaks. Two Lorentzian peakfitting is the simplest approximation for the G-band. In this case, the G-band feature can be fitted by two Lorentzian peaks: one is around 1590 wavenumber (cm -1 ) and is called the G + peak. The other is around 1570 cm -1 which is called the G - peak. The G +

29 16 feature is known as associated with the vibration along the tube axis direction, while G- feature is associated with the vibration in the circumferential direction[36]. The G + band has no diameter dependence while G- band has different diameter coefficients for semiconducting and metallic tubes. Fig shows the lineshape of G-band for one semiconducting and one metallic nanotube. The G - lineshape is highly sensitive to whether the SWNT is metallic (Breit-Wigner-Fano (BWF) lineshape), or semiconducting (Lorentzian lineshape). Charge transfer to SWNTs can lead to an intensity increase or decrease of the BWF feature.[40, 41] Fig. 1-12: The G-band for one semiconducting SWNT and one metallic SWNT.[42]

30 17 Fig. 1-13: Atomic displacements of carbon related to the tangential mode (G-band) of SWNTs Two other features in our Raman spectra (Fig. 1-10) are second-order Raman spectra. They are the disorder-induced D-band and the G -band which is~2ω D [43]. These two bands are sensitive to the tube diameter and chirality since they depend on how the 2D electronic and phonon structure is folded into the 1D structure. SWNTs Raman measurements of D-band and G -band can be described by a double resonance process as in graphene [44]. In the double resonance (DR) process (Fig.1-9b,c), firstly, the electron absorbs a photon at a k state, then, scatters to k + q states, scatters back to a k state, and finally, emits a photon by recombining with a hole at a k state. Therefore, in a DR Raman process, two resonance conditions for three intermediate states should be satisfied, in which the intermediate k + q state is always a real electronic state and either the initial or the final k state is a real electronic state. In the case of the D-band, the two scattering processes consist of one elastic scattering event by defects of the crystal and one inelastic scattering event by emitting or absorbing a phonon, as shown in Fig. 1-9b. In the case of the G -band, both processes are inelastic scattering events involving two phonons.

31 18 Chapter 2 Structural, Optical and Electrical Characterization Techniques 2.1 Transmission electron microscopy (TEM) Transmission electron microscopy (TEM) is a powerful tool for characterization of nanomaterials. In order to image the material, a beam of electrons is transmitted through an ultra thin specimen. These electrons will interact with the specimen as they pass through. A magnified image is formed from the interaction of the electrons transmitted through the specimen onto an imaging device, such as a fluorescent screen or to be detected by a sensor such as a CCD camera. TEMs are capable of imaging at a significantly higher resolution than optical microscopes. The resolution of the microscope can be expressed as: 0.61, 2-1 where λ is the de Broglie wavelength of electrons given by: / 2-2 This enables the instrument to examine fine detail -even as small as a single column of atoms, which is tens of thousands times smaller than the smallest resolvable object in an optical microscope. At smaller magnifications, TEM image contrast is due to absorption of electrons in the material, due to the thickness and compositions of the material. At higher magnifications, complex wave interactions modulate the intensity of the image, requiring expertise to analyze the observed images. Alternate modes of use allow for

32 19 the TEM to observe modulations in chemical identity, crystal orientation, electronic structure and sample induced electron phase shift as well as the regular absorption based imaging. The first TEM was built by Max Knoll and Ernst Ruska in 1931, with this group developing the first TEM with resolving power greater than that of light in 1933 and the first commercial TEM was constructed in Electron energy loss spectroscopy (EELS) Electron energy-loss spectrometry (EELS) is the analysis of the energy distribution of electrons that have interacted inelastically with the specimen. A magnetic prism spectrometer is used to disperse the electrons so that the spectrum can be detected electronically and it offers sufficient energy resolution to distinguish all the elements in the periodic table. A typical EELS spectrum is shown in Fig. 2-1 Fig. 2-1: Idealised schematics of an EELS spectrum, indicating Zero-Loss peak, Plasomon resonance, and core-loss electron peak[45]

33 20 EELS works best at relatively low atomic numbers, where the excitation edges tend to be sharp, well-defined, and at experimentally accessible energy losses (the signal being very weak beyond about 3 kev energy loss). With EELS spectra it is possible to identify the different forms of the same element. For example there are clear differences between the spectra of diamond, graphite and amorphous carbon. When the electron beam traverses a thin specimen, it loses energy by a variety of processes. These inelastic collisions provide us tremendous information concerning the electronic structure of the specimen atoms and tell us the detailed nature of these atoms, e.g., dielectric response, bonding and nearest-neighbor distributions, etc. The reason why we do EELS is that we can separate these inelastically scattered electrons and quantify the information they contain. The derivation for getting the at.% of the element was first given by Egerton[46]. To quantify the EELS spectrum we have to extract intensity in the ionization edge by removing the plural-scattering background and integrating the intensity in the edge. When we are quantifying a K edge, the K-shell intensity above background I K is related to the probability of ionization P K and the total transmitted intensity I T is given by:. 2-3 In a thin specimen we can approximate I T to the incident intensity and we also assume that the electrons contribution to the edge have only undergone a single ionization (Eq. 2-4). exp. 2-4

34 21 N is the number of atoms per unit area of the specimen with thickness of t that contribute to the K edge. σ K is the cross section and λ K is the mean free path for ionization losses. Since λ K is large, Eq. 2-4 can be written as: and Thus, we can measure the absolute number of atoms per unit area of the specimen simply by measuring the intensity above background in the K edge and dividing it by the total intensity in the spectrum and the ionization cross section. For a spectrum containing two edges from elements A and B we have Details of background subtraction and determination of σ are described in David B. Williams and C.Barry Carter s book on Transmission Electron Microscopy.[47] 2.3 Fourier transform infrared spectroscopy (FTIR) Fourier transform infrared spectroscopy is a measurement technique for collecting infrared spectra. Instead of recording the amount of energy absorbed when the frequency of the infrared light is varied (using a monochromator), the IR light is guided through an interferometer, shown in Fig It contains two mutually perpendicular plane mirrors, one of which can move along an axis that is perpendicular to its plane. The movable mirror is moved at a constant velocity. Between the fixed mirror and the movable mirror is a beam splitter, where a beam of radiation from an external source can be partially reflected to the fixed mirror and partially transmitted to the movable mirror. After the beams return to the beam

35 22 splitter, they interfere and are again partially reflected and partially transmitted. Because of the interference, the intensity of each beam passing to the detector and returning to the source depends on the difference on path of the beams in the two arms of the interferometer. The variation in the intensity of the beams passing to the detector and returning to the source as a function of the path difference ultimately yields the spectral information[48]. This measured signal is the interferogram I(δ). Here δ=2vt v is the velocity of the moving mirror and t is the retardation time after zero retardation. After performing a Fourier transform on this signal, a spectrum identical to that from conventional (dispersive) infrared spectroscopy is obtained. 2, 2-7 where is the wavemumber in units of reciprocal centimeters. Fig. 2-2: Schematic representation of a Michelson interferometer

36 23 In this thesis work, a Bruker Vertex 80 V FTIR was used to take transmission spectra in the IR range. The schematics of the optical diagram are shown in Fig To cover the range of interest from 50cm -1 to 10000cm -1, different light sources (Quartz and Globar), various beam splitters( Bk7, CaF2, Kbr and Mylar film) and detectors (InGaAs, InSb and DLaTGS detector) are used. Fig. 2-3: Schematics of Bruker V80 FTIR spectrometer For the far-ir range normally the sample is deposited on a Si substrate. Except for the high refractive index (n=3.4) silicon is an excellent infrared window material, because of its many advantages, including being inexpensive, non-hygroscopic, chemically inactive, highly resistant to thermal and mechanical shocks, scratching and fogging. The transmission data of our Si substrate is shown in Fig In the mid-ir range ZnSe is preferred. The transmission is shown in Fig. 2-5.

37 24 Fig. 2-4: Transmission spectra of a silicon window Fig. 2-5: Transmission spectra of a slab of ZnSe windlow. The inset is a picture of the ZnSe substrate used in this work

38 λ 950 double beam spectrometer For transmission measurements in the range of 200nm to 3300nm a Perkin Elmer Lambda 950 double beam spectrometer is used. The UV/Vis region s resolution reaches 0.05nm with a deuterium lamp as the light source and a photomultiplier (PM) as the detector. For the NIR part the resolution reaches up to 0.20nm with a tungstenhalogen lamp light source and a lead sulfide (PbS) detector. The substrate widely used is quartz, and the transmission spectrum of the substrate is shown in Fig From 260 to 2000 nm the 1mm thick quartz has a transmission of 94% and there are no sharp absorption features. Fig. 2-6: Transmission of 1mm thick quartz substrate Fig. 2-7 shows the optical path for the instrument. The radiation I0(ω, t) split by the flip mirror M9 passing alternately through the sample and reference beams is

39 26 reflected by mirrors M11, M12, M13, and M11, M12, M13, respectively, of the optics in the detector assembly onto the appropriate detector. Mirror M14 is rotated to select the required detector. If T 0 (ω) is the optical coefficient of the first set of transfer optics, T R (ω) is the transmission of the reference sample, T S (ω) is the transmission of the sample, and T 1 (ω) is the optical coefficient of the second set of transfer optics. The radiation intensity at the detector for sample beam (I S ) and reference beam (I R ) is given by Eq. 2-8 and Eq. 2-9, respectively. Fig. 2-7: Schematic diagram of the λ950 optical system,, 2-8

40 27,, 2-9 The source radiation intensity is not a constant over time: there is a slow variation. This time variation of the source can be corrected in the double-beam configuration. By dividing Eq. 2-8 with Eq. 2-9, we get,,, From Eq. 2-10, the intensity ratio between the sample beam and the reference beam is no longer time-dependent. The conclusion is based on the assumption that the sample beam and reference beam are correlated and exhibit the same time dependence. This is a good assumption for our Lambda 950 spectrometer because the two beams are generated from the same source and are separated in time by 1/50 of a second; i.e., a 50 Hz separation is created by a chopper and is used to alternately direct the source energy to the sample and then the reference beam. Another important point to be noted here is the instrument purging with N 2 gas. Because O 2 absorbs radiation in the UV range below 190 nm, and H 2 O absorbs light in the wavelength range nm, nm, and nm, the best accuracy for these spectral ranges requires purging the system with nitrogen. To summarize, the Lambda 950 was found to have excellent stability and it can produce accurate and quantitative optical data.

41 Thermogravimetric ananlysis (TGA) Thermogravimetric Analysis (TGA) measures the weight changes in a material as a function of temperature (or time) under a controlled atmosphere. Its principal uses include measurement of a material's thermal stability and composition. In our experiment the TA instrument Q5000 is used (Fig. 2-8). The system uses an infrared furnace which offers a wide range of linear and ballistic heating rate from room temperature to 1200 o C. In the measurement two platinum pans are used, one for balancing and the other for sample holding. The equipment has a sensitivity of 0.1μg and can weigh as much as 100mg sample. Based on the fact that different carbon allotropes oxidize at different temperatures, TGA analysis can be used to determine the percentage of impurities such as amorphous carbon and graphitic shells in the SWNTs samples. The idea is simple: at a certain temperature a carbon allotrope reacts with the oxygen (the atmosphere used here is dry air) forming CO and CO 2 gases; this process produces a decrease in the total mass that is measured by the instrument. We can also find the optimal dry oxidation condition for removing amorphous carbon. Fig. 2-8: Picture of TA Q5000 system

42 Micro-Raman spectrometer: Jobin-Yvon T64000 A JY Horiba T6400 Raman spectrometer is used to collect Raman data from 488nm and 514.5nm laser excitation. Light is sorted by three monochromators with holographic gratings (1800 lines/mm) in which the first two monochromators (filter stage) can be coupled additively or subtractively and detected by a multichannel (CCD) detector. This instrument can be used both in micro as well as macro scattering geometry and allows measuring Raman modes at very small wave numbers close to the laser line. This feature is particularly useful to study RBM modes of large diameter SWNTs. In the micro-raman arrangement the laser is focused onto the sample using a microscope objective (100x, 50x or 10x) and the scattered light is collected by the same objective. A confocal aperture in the path of the scattered beam allows us to have a high spatial resolution of ~1μm. The microscope stage is equipped with a motorized XY Manipulator for precise positioning and selection of sample area. Raman spectra are acquires a laser source to irradiate the sample. Fig. 2-9 is a schematic diagram for the optical path in the T6400 instrument in the micro-raman geometry.

43 30 Fig. 2-9: Schematic diagram of T64000 emphasizing the optics of the confocal microscope used to collect Raman spectra of SWNTs. The objective spot size is around 1.0μm. Confocal aperture is used to define the scattering volume and remove stray light [49]. The excitation irradiation from the laser source is passed through a plasma filter to remove the associated plasma lines. The incident laser beam then goes through an iris and a spatial filter, which is imaged onto the sample via the objective. The laser light transmitted through the beam splitter is blocked while the reflected part is focused onto the sample by the microscope objective. The backscattered Raman light that is transmitted by the beam splitter proceeds through the triple monochromator to the detector. The confocal microscope is used to give a high lateral resolution. The objective serves two purposes: to focus the beam on the sample and to collect the backscattered light[49]. The collected back scattered light passes through the beam splitter and then through lens L3 to the adjustable confocal aperture. The

44 31 confocal aperture will reject stray light. The aperture size is chosen to be M 1μ, where M is the magnification given by M is the optical magnification of the ~1μ diameter beam spot on the sample produced by the objective in concert with the lens L3. The 1μ spot size is limited by the lateral resolution of the microscope. Passing through the confocal aperture, the Raman scattered light is then focused by lens L5 onto the entrance slit of the subtractive dispersion monochrometor (details are shown in Fig. 2-9 Fig is the schematic diagram of the internal optics in T6400. The polychromatic irradiation scattered from sample enters the first monochromator through entrance slit S1. This slit determines the instrument resolution. The grating G1 disperses the light and the exit slit S2 selects the bandpass. In the second monochromator, G2 recombines the dispersed irradiation into a polychromatic onto the slit S3. This configuration is used as a tunable filter in the spectral range defined by the scanning mechanism and the gratings. This polychromatic irradiation is then dispersed by G3 of the spectrograph. The spectrum is acquired with a liquid nitrogen cooled CCD. Fig. 2-10: Schematic representation of the T6400 double subtraction premonochromator with spectrograph[49].

45 Renishaw Invia Raman microscope Renishaw Invia Raman microscope is an automated micro-raman system. The optical ray diagram with all the optics is shown in Fig The incoming laser passes through a plasma filter, a neutral density filter (for attenuation), two lenses, four mirrors and finally focused to the sample stage by an objective lens. Then, the back scattered light is collected by the objective lens, passed through a Rayleigh filter (edge filter or notch filter) and is dispersed by a grating before being collected by the thermal CCD. The internal optics in this Renishaw Raman microscope is optimized for three laser lines (514.5, 647, 785 nm), allow detecting very small signals not in resonance with the laser excitation. A notch filter is used to pass light displaced about ~50-60 cm -1 on either side of the laser light. The instrument is able to detect vibrations with ω > 50 cm -1. Fig. 2-11: Schematic of the Renishaw Invia Raman microscope laser beam pass

46 Four-probe resistance measurements Instead of two-probe method we used the four-point probe method, shown in Fig When semiconductors are measured, the interface of the metallic probe and the semiconductor sample may form a Schottky barrier instead of an Ohmic contact. Such contact resistance may often outweigh the small resistance to be measured. This is the case for a network of SWNTs. Four probe measurement can effectively eliminate the problems of the contact resistance, as well as other series resistance errors. In the four probe method, a small current from a constant-current source is passed through the outer two probes and the voltage is measured between the inner two probes. For a thin film of thickness t much smaller than the film s width and length, the sheet resistance is given by Eq Fig. 2-12: Four probe inline measurement. Ω, 2-11

47 34 where CF is a correction factor. In the limit when the film (d x d) is far greater than the probe separation s in Fig. 2-12(d>>s), the correction factor becomes (π/ln2)=4.54. Some small current may flow through the set of leads used to measure the voltage. It is usually negligible (typically pa or less) and can generally be ignored for all practical purposes. Since the voltage drop across the leads originated from such small current is negligible, the voltage measured by the meter is essentially the same as the voltage across the sample. Consequently, the resistance can be determined much more accurately than with the two-probe method. In our measurement a Keithley 2400 SourceMeter provided the constant current. It strengths includes high precision, low-noise, highly stable DC current with capability of working as a highly repeatable and high-impedance multimeter. To offset its thermal emf, the current reversal algorithm is employed. That is, two measurements with currents of opposite polarity are programmed. Then, the two measurements are combined to eliminate unwanted offsets. Data are automatically acquired via an IEEE connection with a computer, as shown in Fig A LabVIEW based program is used to perform these electric measurements is shown Fig Fig. 2-13: Picture of the automated temperature dependent sheet resistance measurement setup.

48 35 Fig. 2-14: Program measuring the sheet resistance as a function of temperature for the SWNTs film. For the 3mmx3mm films of nanotube networks, van der Pauw method[50] is used to acquire the sheet resistance. A Keithley 2400 SourceMeter measures four combinations of current and voltage contacts. The sheet resistance R Ñ was calculated according to the relation [50] exp π R AB,CD R AB,DC R CD,AB R CD,BA 4R Ñ exp π R AD,BC R AD,CB R BC,DA R BC,AD 4R Ñ

49 36 Chapter 3 Synthesis and Purification Methods 3.1 Vapour-Liquid-Solid (VLS) model Vapour-Liquid-Solid (VLS) model is often used to describe the nucleation and growth of carbon nanofilaments. Although substantial progress has been made in the production of carbon nanotubes (CNTs) since its discovery, the growth mechanism is still poorly understood. It is widely believed that the Vapour-Liquid-Solid (VLS) model, first used to explain the growth of Si nanowires[51] and then to describe the catalyzed growth of carbon fibers[52], provides a valid description of CNT growth[53]. According to the VLS model, carbon (from feedstock such as carbon-rich gases or graphite) dissolves into the liquid catalyst particle, when the catalyst is supersaturated, carbon precipitates on the particle surface and nucleates the growth of CNTs. Recent thermodynamic calculation indicates that CNT growth is primarily driven by a concentration gradient [54]. In Fig. 3-1 the metal particle at the end of the SWNTs acts as a carbon sink during the growth. The reduction of carbon concentration near the cap edge leads to the diffusion of carbon atoms from regions of high carbon concentration to the cap, where they incorporated into the structure.

50 37 Fig. 3-1: Illustration of the concentration gradient across the catalytic metal particle during SWNT growth. Incorporation of atoms into the SWNT structure lowers the C concentration near the open SWNT end.[55] VLS might be a reasonable mechanism to explain the growth of carbon nanotubes using ARC and Pulsed Laser Vaporization (PLV), in which very high temperatures results in the liquid phases of both carbon and catalyst metal. However, this might not be true for CVD in growth because of the lower temperature. Nevertheless, the presence of a metallic nano-catalyst is necessary in all the techniques to promote the one dimensional growth. In the case of CVD, the catalyst is a must for the decomposition of the gas molecules (C 2 H 4 or C 2 H 6 ) to provide carbon for the nanotube growth. 3.2 Arc-discharge method The arc discharge was the first available method for the production of both MWNTs and SWNTs. This method is first developed by the research group at University

51 38 of Montpellier in France[56]. SWNTs can be produced in a carbon arc apparatus similar to the one depicted in Fig Fig. 3-2: Schematic diagram of the arc-discharge apparatus[57] The cathode is made of graphite and the anode is made of a mixture of graphite and metal catalysit particales(ni,y~30w%). An arc discharge, between the two electrodes in a gas atmosphere, vaporizes the anode electrode and produces the closed-ended SWNTs. The SWNTs formation is induced by the presence of the Ni-Y metal nanoparticles. A single nanoparticle can be the seed for the formation of a bundle of SWNTs. The high temperatures generated in the ARC discharge liquidize both the carbon and the metal and the nanotubes are formed by the VLS process. These carbon nanotubes are of average diameter 1.4nm and are found in ropes which are typically ~20nm in diameter or approximately 100 tubes per rope with lengths of 2 5 microns. These SWNTs are usually contaminated with other impurities, such as approximately 30 wt.% residual catalyst (NI, Y) which is usually encapsulated in carbon shells. Some amorphous carbon may also be found on the outer surfaces of the ropes. Fig. 3-3 shows a TEM image of the raw Arc SWNTs sample.

52 39 Fig. 3-3: TEM image of raw arc SWNTs 3.3 Pulse laser vaporization (PLV) Another efficient way to synthesize SWNTs is using a high power laser to vaporize a target made mainly of graphite and a small fraction fo metal catalyst[58]. The schematic diagram of our PLV setup is shown in Fig. 3-4.

53 40 Fig. 3-4: Schematic diagram of the pulsed laser vaporization (PLV) apparatus A pulsed Nd: YAG laser with a fixed pulse repetition rate of 10 Hz is used to vaporize the carbon target, which contains Ni/Co alloy. The laser beam is focused to a spot and scanned across the target via a computer controlled lens. The focused beam vaporizes both carbon and metal in the target. Just as in the case of the arc discharge, small droplets of metal-carbon alloy form and act as seeds to support the SWNTs growth by the VLS mechanism. Two coaxial quartz tubes are used in the PLV furnace, a stream of Ar/He gas flows between the two tubes and is preheated before entering the center tube containing the target. The gas flow is used to drive the SWNT product behind the target through the hot zone of the furnace before it sticks on the quartz tube walls in the colder zone. The PLV system can be set to operate at a particular pressure and gas flow rate. SWNTs are typically grown with a background furnace temperature of 1200 o C in our system. The diameter distribution of our PLV synthesized SWNTs is similar to that of the tubes produced by arc discharge, usually in the range nm.

54 Chemical vapor deposition (CVD) SWNTS can also be grown by CVD. CVD involves the decomposition of a gaseous compound of carbon, catalyzed by metallic nanopaticles, which also serves as nucleation sites for the initiation of carbon-nanotube growth. The advantage of CVD method is that it allows more control over the morphology and structure of the produced nanotubes[59]. With the CVD methos, one can produce well-separated individual nantube which is good enough to be directly used to fabricate nanoelectronics devices. Typically, the tubes are grown with Fe/Mo catalyst nanoparticles dispersed on a substrate. The growth temperature is around 1100K and a mixture of CH 4 and H 2 is used as the carbon feedstock[60]. 3.5 Sample purification The raw SWNT material obtained either by ARC discharge or PLV went through a three-step process to reduce the unwanted impurities (amorphous carbon, graphitic shells and metal catalyst particles). The first step is an oxidation to selectively remove amorphous carbon. Two of the most common methods for the removal of amorphous carbon attempt to take advantage of its higher reactivity with oxygen relative to SWNTs. One is a dry oxidation in air at moderate temperatures ~320 o C [61-63]. Another is refluxing raw soot in hydrogen peroxide H 2 O 2 [64]. During the reflux the solution is heated (60 o C) for 48 hours before going to the next step. If the metal catalyst in the nanotube soot is covered with a graphitic shell of carbon, the oxidation is also important to weaken this coating for subsequent acid digestion of the metal.

55 42 The second step of purification involves the removal of the metal with some acid, usually HNO 3 [63] or HCl [61, 62]. HNO 3 has been reported to aggressively attack the tube walls, while HCl is found to be passive with respect to attack on the carbons in the sample. This will remove most of the metal catalyst in the sample. The setup for reflux is shown in Fig. 3-5a. The third step is to remove the graphitic shells and other unwanted residues. This is usually done by filtration. For small amounts of sample (several mg), vacuum filtrating through a (polycarbonate membrane, 1 micron) and rinsing the material with de-ionized water (100 o C) works well (Fig. 3-5b). But during the filtration the retained SWNTs pack together to block the filter membrane pores. As this SWNTs filter cake thickens, the permeation rate of the solution slows dramatically. This problem can be easily solved with a cross flow filtration (CFF) (Fig. 3-5c). In CFF, the filtration membrane takes the form of a hollow-fiber, the wall of which is permeable to the solution (Fig. 3-5c). The filtrate is pumped down the bore of the fiber at some head pressure from a reservoir, and the major fraction of the fast flowing solution which does not permeate out the sides of the fiber is fed back into the same reservoir to be cycled through the fiber repeatedly. The fast hydrodynamic flow down the fiber bore (cross flow) sweeps the membrane surface, preventing the buildup of a filter cake. A second reservoir contains a buffer solution, which is used to make up the filtratereservoir solution volume lost to permeation through the fiber wall. [61]. The purified sample has a metal concentration of less than 1 wt.%, as demonstrated by the ash analysis (TA Instruments, model IR5000) and low magnification TEM images. The reaction conditions for the temperature programmed oxidation (TPO) were 5 o C/min from o C under a flow rate of dry air at 100 sccm. By taking the derivative of the TPO data (DTPO) we can obtain the maximum

56 43 oxidation temperature and the fractional percentages of the carbon species in the sample. The weight percentage of SWNTs can also be estimated from the area under the peak in the DTPO data, given in Fig In this DTPO data there are only two main peaks, one around 560 o C which is associated with the nanotubes, and the other is small, sitting at the foot of the first one at ~680 o C, which is related with the graphitic shells seen in the HRTEM image (Fig. 3-6 ). Fig. 3-5: Purification setup (a) reflux for wet oxidation with H 2 O 2 and acid treatment. (b) vacuum filtration setup (c) cross-flow filtration

57 Fig. 3-6: (a)tem image if raw PLV SWNTs and purified SWNTs (b)tpo analysis of purified SWNTs 44

58 Boron doping of SWNTs Chemical doping of various forms of sp 2 carbon have been the subject of considerable interest over the past 60 years beginning with graphite intercalation compounds. In the early 70 s, the chemical doping of graphite by a reaction with AsF 5 produced an intercalation compound with an electrical conductivity better than copper[10]. In chemical doping, the dopant chemisorbs on the carbon surface, and exchanges charge with the carbon, creating free carriers in the sp 2 network. SWNTs are p-doped in air due to oxygen covering (physisorbed) the tube wall. Both p-doping and n-doping can be achieved by covering the network with electron withdrawing and electron donating species. This has been demonstrated with NH 3 and NO 2. The disadvantage of physisorption doping is that stability is low, especially for applications with high operation temperature. The present work emphasizes the effects on the conductivity due to substitutional doping of SWNTs, where the dopant must substitute a carbon atom in the sp 2 lattice. Boron and nitrogen are the only two elements that can be incorporated into an sp 2 carbon network without significantly affecting the atomic arrangement in the hexagonal 2D lattice. Early studies have shown that boron can be substituted at the few % level into the graphene sheets in graphite[23]. The synthesis of B- and N-doped SWNTs may be separated into two categories: doping during nanotube growth and postsynthetic processing. Several synthetic routes have succeeded in direct synthesis of B-SWNTs and N-SWNTs including pulse laser ablation [15, 16, 22, 65, 66], Arc discharge [13, 14] and chemical vapor deposition [67-71]. In these routs boron or nitrogen containing reactants are mixed with the carbon and catalyst sources before the growth. As for the Arc (PLV) process electrodes (targets) containing boron concentrations are prepared by mixing a boron compound (B 4 C or B) with carbon paste and the Ni/Y catalysts. The advantage of direct doping is

59 46 that all the SWNTs are uniformly doped. The disadvantage is the high yield of unwanted side products (or phases), which are difficult to remove. In postsynthetic process, SWNTs are mixed with dopant molecules and react at a high temperature in order to promote the incorporation of dopant atoms into the carbon sp 2 lattice. For boron doping, the dopant source normally is B 2 O 3 [17-21]. These previous works showed that high levels of doping can be obtained but the doping level is not uniform, and that B-doping produces quite a few side products. Our goal here is to find a feasible way to uniformly dope SWNTs with boron to produce large quantity of clean B- SWNTs material suitable for applications, such as OLEDs and solar cells. Besides, the electrical and optical properties of these B-SWNTs are also of much basic and technological interest. 3.7 Oxidation of Carbon by B 2 O 3 In the presence of boron oxide, some carbon atoms in the tube wall are oxidized and form CO gas. The vacancies left by these carbon atoms will be filled with the remaining boron atoms, forming in this way the B x C (x<0.01) nanotubes at 900~1000 o C. Calculation of equilibrium in the carbothermic reduction of B 2 O 3 has been applied to the B-O-C system. The resulting five possible reaction paths are shown in Tab. 3-1[72], together with their Gibbs energy and the minimum temperature (equilibrium) required for these reactions to take place. These reactions are for pure graphitic (bulk) material of high crystalline quality in which no strain, defects or any other factor that could possibly lower the Gibbs energy and the equilibrium temperature as listed in the table.

60 47 The thermodynamic data listed the free energy of the formation of B 2 O 2, CO and boron. Eq. (1) is considering a reaction of liquid B 2 O 3 which still contains 20~30% boroxal rings. For the case of reaction (2) the B 2 O 3 is in a gas phase presented as discrete v-shaped O=B-O-B=O molecules. After boron is produced by reaction (1) or (2) it could be consumed by the excess B 2 O 3 and form B 2 O 2 as in Eq.4. The product of B 2 O 2 will react with carbon at even lower temperature leaving free boron to bind with the graphitic lattice. Tab. 3-1: Standard Gibbs energy, standard entropy ( S 0 298) and equilibrium temperature of B 2 O 3 consumption reactions[72] Reaction G o (J/mol) T eql ( o C) S 0 298(J/Kmol) B 2 O 3(l) +C (s) B 2 O 2(g) +CO (g) 1 545, T B 2 O 3(g) +C (s) B 2 O 2(g) +CO (g) 2 220, T B 2 O 3(l) +B (s) 1.5B 2 O 2(g) 3 451, T B 2 O 3(g) +B (s) 1.5B 2 O 2(g) 4 125, T B 2 O 2(g) +C (s) B (s) +CO (g) 5 94, T In the case of SWNTs, their unique hollow nanostructure and curvature induce some strain and provide a large surface area with more accessible active sites. These conditions facilitate the formation of CO (by oxidation) at lower temperatures, and the subsequent doping of boron into the rolled hexagonal graphene lattice structure. In our experiment we found that temperatures in the range of o C are enough to successfully B-dope the SWNTs. Early study of the thermal chemistry of B 2 O 3 deposited on graphene showed indirect evidences that the reaction path B 2 O 3(l) +3C 2B (s) +3CO (g) dominates at temperatures in the range of o C;while B 2 O 3(l) +C B 2 O 2(g) +CO (g) was more important at higher temperatures [73, 74]. It is clear that several competing processes may be involved. Temperature programmed desorption (TPD) and X-ray photoelectron

61 48 spectroscopy (XPS) studies of oxidation of graphitic carbon by B 2 O 3 suggest that B 2 O 3 may be reduced by carbon at temperatures above 700 o C, according to B 2 O 3(l) +3C 2B (s) +3CO (g) [74]. After the oxidation process, B 2 O 2 was also observed and it could be explained if a parallel reaction (2B 2 O 3(g) +2B (s) 3B 2 O 2 ) is taking place simultaneously [75, 76]. After carbon atoms were removed by oxidation, boron atoms will occupy the vacancies left by the oxidized carbon atom in the tube wall. At 900 o C boron will be bound with the neighboring carbon atoms and become a constituent part of the tube wall. The oxidation of carbon by B 2 O 3 needs to be at a high (>900 o C) temperature, but high temperature will also help remove the defects on the tube wall. In order to improve the doping processed an optimal temperature in the range of 900~1100 o C is needed. 3.8 Electronic structure of boron-doped SWNTs Electronic structure of boron-doped SWNTs was studied via first-principles methods based on the density functional theory by koretsune et al[77]. Since boron atom has a larger atomic radius than carbon atom[78], boron atom in the optimized structure locates outside of the original atomic position in the pristine nanotube to stretch the B-C bond length. From the calculated formation energy E (Fig. 3-7), indicates that B doping into the narrow-gap tubes needs less energy than that of the moderate-gap tubes.[77] Koretsune et al. [77]also calculated the band structure and DOSs of the pristine and B-doped SWNTs, as shown in Fig The energy gap of B-doped SWNTs is given in

62 49 Tab The correlation of B-doped SWNTs band structures with the optical measurements will be discussed in the following chapters. Fig. 3-7: Formation energy E as a function of nanotube diameter. (m,0) zigzag tubes of m=4-12 (filled circles) and (m,m)armchair tubes of m=4,5, and 6 (open circles) are plotted.[77] Fig. 3-8: Band structures and DOSs of pristine and B-doped (10,0) SWNT: (a) C 40, (b) BC 39, (c)) BC 79,(d)) BC 119, and (e) ) B 2 C 78.[77]

63 50 Tab. 3-2: Electronc structure of a B-doped SWNTs[77] C 40 BC 39 BC 79 BC 119 B 2 C 78 Gap (ev)

64 51 Chapter 4 Simultaneous Growth and Doping of SWNTs by ARC Discharge 4.1 Sample growth In this chapter discussions center on the synthesis of boron doped carbon nanotubes by arc discharge. Boron carbide powder is mixed with carbon paste and Ni- Y catalysts to form the arc discharge electrodes. A systematic study was conducted by varying boron concentration from 1 to 10 at.%. The material created from each electrode is denoted by the boron at.% in the electrode. The advantage of doping SWNTs via Arc discharge is that all the tubes created are uniformly doped. In this section, we will show HRTEM images and TPO data of raw SWNTs soot and SWNTs materials produced with N at.% boron in the electrode (N%B-SWNTs). Sample produced from electrodes with 1-10 at.% boron are shown in Fig. 4-1, Fig.4-2 and Fig. 4-3.

65 Fig. 4-1: TEM images of the material produced from electrodes with boron concentration of (a)1at.%, (b)2 at.% and (c)3 at.% 52

66 53 The TEM images show that the SWNTs yield decreases with an increase of boron content in the electrode. Four components in the sample, including bundles of SWNTs, amorphous carbon, graphitic carbon and ~10nm catalyst particles were clearly observed. HRTEM images reveal that the graphitic shells cover the catalyst particles. Fig. 4-1 shows that the SWNTs ropes dominate for the material grown from electrodes with boron concentration less than 3 at.%. The SWNTs bundles are coated with amorphous carbon. For the material grown with more than 5 at.% boron in the electrode, few nanotube bundles but nanohorns were observed instead, as indicated by arrows in Fig Moreover, large bamboo-structured tubes, with diameter around 20nm, were clearly seen. The dominant forms of carbon were graphitic flakes, nanohorns and bamboo-structured tubes, as shown in Fig Such observation is also supported by the temperature programmed oxidation (TPO) studies. This trend is also seen in the PLV samples, possibly because of poisoning of the catalyst particle which seeds the growth of the nanotubes [16]. Such catalyst poisoning might be understood as a boron-induced change in the seed composition, which in turn results in a completely new phase diagram that might not present an eutectic point above the temperatures used in the process. Worse still, catalyst poisoning may ruin the VLS process. A possible solution for this problem would involve the search of a novel catalyst material and a growth condition more suitable for VLS process when boron is present in the reaction.

67 Fig. 4-2: TEM image of material produced from electrodes with boron concentration of 5 at.% 54

68 55 Fig. 4-3: TEM image of material produced from electrodes with boron concentration of 10 at.% After removal of the amorphous coating on the SWNTs bundles by H 2 O 2 reflux the local boron content and how boron is fitted into the sp 2 lattice of the SWNTs wall were characterized with EELS. In the EELS K-edge measurement, care was taken to locate the regions of the material where the TEM beam probed only clean bundles of SWNTs, such that the contributions of B-doped amorphous carbon were minimized. For example, the EELS spectrum in Fig. 4-4(left) results from the entire image shown in Fig. 4-4(right). EELS spectra of a few bundles were obtained by using a selective area aperture, about 140nm; this condition was preferred instead of a focused nano probe because the larger probe area increases the signal-to-noise ratio for the K-edge and minimizes the tube wall damage. Fig. 4-4 displays TEM images of the isolated SWNTs bundles and the associated EELS spectra for high-quality samples produced with 1 at.% boron in the electrode. EELS spectrum shows the characteristic core-level excitations at 188 and 284 ev for boron and carbon, respectively. We don t have a well defined structure corresponding to the 1s π * and 1s σ * pre-ionization edges

69 56 characteristic of sp 2 hybridization [15, 69, 79]. The boron to carbon ratio of this sample is estimated to be less than 1 at.%. We found several examples of amorphous carbon with boron content higher than those in the tube walls, i.e., boron content as high as 10 at.% was observed in some amorphous carbon-rich regions. The boron detected here can be attributed to the amorphous material coated on the SWNTs bundle, and to boron substituting the carbon in the nanotube lattice. As will be discussed later, the Raman D-band study confirms the boron substitution in the lattice[80]. Fig. 4-4: HRTEM image and EELS spectrum of B-SWNTs. (left) HRTEM image of B-SWNTs bundles treated with H 2 O 2 for amorphous carbon removal. (right) EELS spectrum of the bundle shown in the left image 4.2 Temperature programmed oxidation (TPO) study The material produced by Arc discharge was chemically analyzed with TPO using a TA Instruments (TA Q5000). The reaction conditions included a 5 o C/min heating rate ( o C) and a dry air flow rate of 25 sccm. The DTPO curve is obtained by

70 57 calculating the first derivative of the TPO curve with respect to temperature. The preferential oxidation temperatures and the fractional percentages of the different phases of carbon in the soot can be derived from the DTPO data. The purity of SWNTs can also be estimated from the area under the peak in DTPO data. The data in Fig.4-5 are typical for the undoped (top) and 1%B-SWNTs (bottom) soot. The temperature was ramped linearly in time up to 1000 o C at a dry air flow rate of 25 sccm.the solid green line is the weight loss curve (m(t)) and the blue one is the derivative of the solid line (dm(t)/dt). The DTPO data were fitted by multiple Gaussians to determine the combustion temperatures of various carbon phases, including amorphous carbon (ac), SWNTs and graphitic carbon in each sample. The composite fit is shown in the right corner, represented by the red solid line. The individual Gaussians were displaced and displayed under the DTPO curve for clarity. The sudden drop of TPO data in 300 o C to 800 o C range is due to oxidation of carbon into gaseous CO 2 and CO. The leftover mass above 800 o C can usually be identified as a metal oxide or carbide.

71 Fig. 4-5: TPO and DTPO data of (a)swnts and (b)1%b-swnts 58

72 59 The fraction of each carbon phase can be evaluated by calculating the area under each Gaussian peak, and the metal content can be calculated from the weight above 800 o C. In SWNTs soot sample, we determined that the oxidation temperatures for amorphous carbon peaks at 358 o C, SWNTs at 426 o C and graphitic carbon at ~628 o C. We found that the fractional weight of the SWNT in all carbon materials was 58%, amorphous carbon 37% and 5% graphitic carbon. 35% of the total weight is catalyst metal. This percentage is in good agreement with the amount of metal catalyst we have in the electrode. As for the 1% boron doped material the oxidation temperature for the amorphous carbon peaks at 373 o C, SWNTs at 433 o C and graphitic carbon at 681 o C, respectively. The fractional weight of the SWNT in all carbon materials dropped to 40%, and that of amorphous carbon reduced to 14%. The fractional weight of graphitic carbon dramatically increased from 5% to 47%. The total weight of the metal was still 35%. Comparing the DTPO data of the doped and undoped materials it is easy to see an upshift of the SWNTs oxidation temperature, possibly originating from the coating of the catalyst particles by the graphitic shells and the decrease of amorphous carbon in the mixture. The DTPO data of 1%B-SWNTs does not fit properly with just three Gaussians. We noticed that the feature at around 680 o C in DTPO data is broad and the intensity is far greater than that of the undoped. The reason might be due to the creation of different sp 2 carbons (shells and horns and ribbons). 4.3 Raman study Fig. 4-6 gives the Raman spectra taken at room temperature for the semiconducting SWNTs excited with nm laser. Fig.4-6a focuses on the spectral

73 60 region from 1200~1850cm -1, which contains the D-band and the G-band of the SWNTs Raman spectrum, while Fig. 4-6b depicts the second-order G -band Raman spectra for the same set of samples. Fig. 4-6: Room temperature Raman spectra showing (a) D-band(~1340 cm -1 ),G-band (~1590 cm -1 ) and (b) second-order G band. The excitation laser is 514.5nm and spectra were normalized to the tangential G+ band intensity. The clearly observed radial breathing mode (RBM) in the Raman spectra (Fig. 4-7) is a sound evidence of the existence of nanotubes. This is also shown in the HRTEM images (Fig. 4-1 to Fig. 4-3). Based on a Lorentzian line fit to the data Fig. 4-6, Fig.4-7, Fig. 4-8 and Fig. 4-10, the RBM and the tangential mode (G-band) at~1590cm -1 do not show significant shifts or line broadening due to the low boron doping. This is a clear indication that the nanotube structure, as a whole, remains intact in products obtained from electrodes containing up to 3 at.% boron. However, the disorderinduced D-band at ~1350cm -1 does show a systematic increase in intensity, which

74 61 suggests that the increasing presence of boron in the electrode may cause a noticeable change in the degree of ordering in the sp 2 lattice. The Raman D-band is activated in sp 2 carbons by various forms of disorder in the carbon network and the scattering intensity depends on the disorder-induced relaxation of the 1 st order q=0 Raman selection rule for crystalline material. This increase in the D-band intensity is attributed to the activation of off-zone-center phonons due to relaxation of the strict selection rules for Raman scattering due to a double resonance process[81]. The increasing intensity of the D-band is also consistent with the earlier work in the B- doped graphite[82] and B-doped SWNTs produced by PLV technique[16], where a systematic increase in D-band intensity with an increase of B-doping was reported. Fig. 4-7: Room temperature Raman spectra showing Radio breathing mode (RBM) using laser with excitation energy of (a) 2.54eV (488nm) and (b) 2.41eV (514.5nm)

75 62 Fig. 4-8: Room temperature Raman spectra showing (a) D-band (~1350 cm -1 ), G-band (~1590 cm -1 ) and (b) second-order G -band(~2680cm -1 ). The excitation laser is at 488nm. Graphite films with 0.4 and 2.2 at.% boron were characterized in the B-doped graphite study using the 514.5nm excitation wavelength[83]. In the 0.4 at.% borondoped films, the D- and G -band frequencies downshifted by ~5 cm -1 while the G-band frequency downshifted ~1cm -1 [83]. When the boron concentration further increased from 0.4 at.% and reached 2.2 at.%, the D-band frequency upshifted by ~2 cm -1 while the G-band frequency upshifted by ~6 cm -1 and the G -band further downshifted by ~12 cm -1. The net result was an overall downshift of 3 and 17 cm -1, respectively, for D and G -band frequencies, and an upshift of 6 cm -1 for the G-band frequency as the boron concentration increased from 0 to 2.2 at.% [83].

76 63 For B-doped graphite the Hall coefficient measured at 3K for each of their boron- doped films was positive, implying p-type doping (presence of hole carriers) in the sample, which results in an upshift in the G-band frequency for boron concentration exceeding 2 at.%. Such upshift in the G-band frequency is not seen in the arc samples, because HRTEM measurement indicated that the presence of graphitic flakes and other forms of sp 2 carbons were mixed with the boron doped SWNTs bundles. These materials have a sharp first-order line at 1582~1585 cm -1 (2) (Fig. 4-9) corresponding to the Raman-active E 2g mode observed in single crystal graphite at the same frequency[8,9]. This explains why we are seeing slight downshift in the G-band as boron concentration increases in the electrodes. In chapter 6 we will show that for Raman spectra on pure B-SWNTs there is indeed an upshift of the G-band associated with boron doping, in agreement with the previously reported results in B- doped graphite.[83]

77 Fig. 4-9: Raman spectra (T=300K) from various sp2 carbon using Ar-ion laser excitation: (a)highly ordered pyrolytic graphite (HOPG), (b) boron-doped HOPG (BHOPG), (c) carbon nano particles derived from the pyrolysis of benzene and graphitized at 2820 o C, (d) as-synthesized carbon nanoparticles (~850 o C), (e) glassy carbon[84]. 64

78 65 Fig. 4-6b shows that the G -band at 2673 cm -1 in the undoped SWNTs bundles systematically downshifts with the increasing B concentrations. This is in agreement with the work of McGuire et al. on boron doped SWNTs from PLV technique. For their samples the G -band for SWNTs bundles systematically downshifted from 2677 cm -1 to 2670 cm -1 with increasing B concentrations in the targets. Previous work on boron doped HOPG by Hishiyama et al. also showed the G -band downshifts from 2725 cm -1 in a HOPG film to 2720 cm -1 in boron-doped HOPG films [83]. The G -band downshift is a result of the weaker C-B bond compared to the C-C bond energies. Although the G -band has a diameter dependence [36], the fact that the RBM frequency remains the same for all boron-doped levels suggests that the diameter dependence is not the cause of the downshift. For the Raman spectra using 488nm laser excitation on the same set of samples, we didn t observe the downshift in the G -band frequency. But for the Raman spectra of the same set of samples using 1064nm laser excitation this down shift is observed (Fig. 4-10). The RBM mode appears at 160 cm -1 for the 1064nm excitation. Comparing the 1%B-SWNTs sample with the undoped SWNTs we can see that the D-band intensity increases and G -band downshifts 9 cm -1. The huge intensity of G -band in 2%B-SWNTs (Fig. 4-10) can be attributed to the boron-induced changes in the electronic properties of SWNTs [16].

79 66 Fig. 4-10: Room temperature Raman spectra showing R-band(~160cm-1), D-band (~1270 cm-1), G-band (~1590 cm-1) and (b) second-order G -band(~2530 cm-1). The excitation laser is at 1064nm and spectra were normalized to the tangential G+ band intensity. 4.4 Resistance measurement Four-probe technique was used to measure the line resistance of the films with thickness greater than 2μm, and results are shown in Tab. 4-1 Current of 100μA is passed between the two outer probes while the voltage across the other two inner probes is measured. The 4 probes are separated at equal distance about 1.58mm apart. Since this separation is far smaller than the film width and length, a correction factor of 4.54 is used here. The resistance of the 1%B-SWNTs and 2%B-SWNTs were found to be about ~40% of that of the undoped sample. At this stage we can t

80 67 distinguish explicitly the effects of doping of bundle contacts and the doping of nanotube bundle itself. In the following chapters we will have a more systematic study of the resistance of these films. For the 3%B-SWNTs even when the doping concentration might be higher on each nanotube bundle, the resistance is increasingly higher than that of the undoped, probably because of the decrease of nanotube concentration in the material. Tab. 4-1: Sheet Resistance of undoped, 1%B-SWNTs, 2%B-SWNTs and 3%B-SWNTs Sample Undoped 1%B-SWNTs 2%B-SWNTs 3%B-SWNTs Sheet Resistance 5326Ω/ 2218Ω/ 2340Ω/ 8706Ω/ 4.5 Summary Boron has been successfully substituted into the carbon sp 2 framework of SWNTs by the Arc discharge method. We have found that boron doping significantly reduces its sheet resistance. The boron impurity in the nanotube wall changes the electronic properties of the nanotube. The Raman spectra and TEM study reveal that the electrodes with large boron content fail to grow SWNTs.

81 68 Chapter 5 Post-Growth B-Doping in Reactive Environments (NH 3 ) 5.1 Sample preparation After growth of a defect-free SWNTs, the high stability of C=C bond should make it difficult to substitute a boron atom. Post growth B-doping methods probably involve, as a first step, the removal of some C-atoms, e.g. as CO or CO 2 to create or increase the number of wall defects. In a second step, a reaction with a suitable boron precursor might be necessary to heal the carbon vacancies in the sp 2 lattice of the tube wall with boron atoms. Since the boron atom has a larger atomic radius than the carbon atom[78], the boron atom in the optimized structure locates outside the original lattice site in the pristine nanotube to stretch the B-C bond length. This B-C bond stretch will give rise to change in the lattice structures, including outward movement of boron atom. Boron doping into the narrow-gap tubes needs less energy than that into the bigger-gap tubes[77]. This can be understood as follows: if a boron is doped into a graphene layer, which is the infinite diameter limit of the SWNTs, to stretch the B-C bonds, carbon atoms should be pushed away or the boron atom should move higher or lower from the graphene layer accompanied by symmetry breaking. On the other hand, if a boron atom is doped into SWNTs to stretch the B-C bonds, the boron atom can move outward from the tube surface without symmetry breaking. When the curvature is large, a slight movement is sufficient to stretch the B-C bonds. Thus, it needs a relatively small energy to substitutionally dope a boron atom into a SWNT with a large curvature.

82 69 To decrease unwanted impurities in the material, we obtained B-doped SWNTs by a post-growth high-temperature chemical reaction of ARC-derived SWNTs in direct contact with B 2 O 3 while flowing NH 3 [17]. The processed ARC tubes used in this experiment still contain a high concentration of amorphous carbon and other impurities that affect the electrical properties of the sample. The following discussion focuses on how boron doping improves the optical transmission and sheet resistance of these thin films of SWNTs networks. The SWNTs material investigated here was obtained from CarboLex, Inc., and is specified to consist of ~50 70 vol% carbon as bundled SWNTs. The tubes were produced by an arc plasma discharge using a Ni/Y catalyst. The Raman spectrum (514.5 nm excitation) was found to be typical of high quality arc-derived tubes, i.e., weak D- band scattering, a Raman radial band centered at ~160 cm -1 and a stronger tangential band at 1592 cm -1 [36]. Typical high resolution transmission electron microscope (HRTEM) images on similar CarboLex material have shown that the arc-tubes are present in bundles ~3-5 μm long, with bundle diameters in the range of nm, i.e., typically containing ~ tubes[85]. The SWNTs themselves have diameters in the range of 1.2~1.6 nm[85]. In this study, the as-prepared (AP) SWNT material from CarboLex went through a mild three-step post-synthesis process to reduce the amorphous carbon (ac) and metal catalyst content: (step 1) reflux in 3 M HNO 3 for 48 h, (step 2) neutralization with a NaOH solution, followed by (step 3) filtration and rinsing with de-ionized water. After the treatment, the material was denoted as processed, and the processed arc material as p-swnts. The metal content in the acid-treated material was determined by ash analysis (combustion in dry air) in a thermo-gravimetric apparatus (TA Instruments, Inc., model IR5000) to be ~15 wt% catalyst residue. The fairly high metal

83 70 content should not impact the sheet resistance which is dominated by percolating SWNTs bundle networks. The p-swnts were then mixed with powdered B 2 O 3 (grain size ~10µm) and placed in a quartz boat for the high temperature reaction. The weight fraction of B 2 O 3 to SWNTs was ~ 5. The mixed SWNTs-B 2 O 3 powders were heated at 900 C for 4 h in a quartz tube reactor with flowing NH 3 (110 sccm at 200 mbar) as the reactive atmosphere[86]. The product was then washed and filtered 3 times (polycarbonate membrane, 1 micron) using hot de-ionized water (100 C) to remove the residual B 2 O 3. The ac coating of the bundles is expected to increase the bundle-bundle contact resistance. The carbon-encased metal nanoparticles, on the other hand, probably make a relatively unimportant contribution to the sheet resistance, but can produce an undesirable reduction in the optical transparency. In the ideal material, these metal particles need to be stripped off by a more effective purification. For example, a selective oxidation of the arc SWNTs, prior to acid treatment, in flowing dry air (~400 o C for 15 min) has been shown[63] to weaken the carbon coating on the catalyst particles to such a degree that HNO 3 or HCl reflux can be used to leach out the metal, leaving hollow carbon cores behind. This selective oxidation was not done here. 5.2 TEM and EELS studies The impacts of the B 2 O 3 reaction on the morphology of the SWNTs as well as the at.% boron-doped SWNTs were studied with TEM and EELS. The measurements were made in a JEOL-2010F microscope with a Gatan Enfina TM 1000 EELS system. EELS information on the hybridization state of both boron and carbon were obtained by studying the K-edge absorption of each element [15, 69, 79]. The energy dispersion

84 71 was 0.2 ev/channel and the absolute energy-loss scale was calibrated using the graphitic π* peak at 285 ev. Fig. 5-1 is the typical low-magnification TEM micrographs of bundles of the p- SWNTs (Fig. 5-1a) and B-SWNTs material (Fig. 5-1b). The TEM images show a web of entangled SWNTs bundles, and numerous carbon-encased metal nanoparticles can be seen attached to the bundle walls. The carbon encasement unfortunately protected quite a few metal cores of the catalyst nanoparticles from acid attack during the processing. The p-swnts bundle walls were observed at higher resolution (HR) to be reasonably free from amorphous carbon coating. That is, the HRTEM images of the p- SWNTs bundle surfaces are similar to those shown in Fig. 5-1c for the B-SWNTs material. The HRTEM images prove that the B 2 O 3 -NH 3 process does not appear to introduce significant damage to the nanotube walls. Raman scattering, however, is a more sensitive probe in this regard. Below we discuss the results of Raman scattering of these B-SWNTs samples. Fig. 5-1: TEM images of B-doped and undoped SWNTs bundles, (a) TEM image of undoped SWNTs bundles. (b) TEM image of B-SWNTs. Dark dots are residual NiY catalyst (c) HRTEM image of one B-SWNTs bundle (d) EELS spectrum of the bundle shown in (c)

85 72 The bundle size distribution, before and after the B-doping, was measured by TEM. We found that a typical bundle diameter was ~15 nm, and that 80% of the bundles had diameters in the range of 10 nm<d<20 nm. Most importantly, the B- doping process was not observed to create any significant change in the bundle size distribution. So, we don t expect that additional parallel percolation paths formed as a result of the B-doping splitting bundles into smaller diameter bundles. Had this occurred, even at fixed optical density we would expect a reduction in sheet resistance on decreasing bundle diameter. In the EELS K-edge measurement, we were careful to locate regions of the material where the TEM beam probed only clean bundles of SWNTs, such that the contributions of the B-doped amorphous carbon were minimized. For example, the EELS spectrum in Fig. 5-1d comes from the entire image shown in Fig. 5-1c. EELS spectra of a few bundles were obtained by using a selected area aperture of diameters~140nm. The EELS spectrum in Fig. 5-1 captures both the K-edge of boron and carbon. The shape of the boron K-edge at 188 ev, shows the structure corresponding to the 1s π * and 1s σ * pre-ionization edges characteristic of sp 2 hybridization [15, 69, 79]. We therefore can conclude that the majority of the boron in the bundle has been incorporated into the sp 2 nanotube carbon network. It should be noted that we found several examples of amorphous carbon with B content higher than that in the tube walls, i.e., B content as high as 10 at.% was observed in some amorphous carbon regions, whereas typical values of 1-3 at.% boron were found for clean bundles of B-SWNTs. 1-2 at.% variation in boron was observed from bundle to bundle and may be associated with inhomogeneous chemical contact between the p- SWNTs and the B 2 O 3. It should also be mentioned that we did not observe the EELS

86 73 peak at 194 ev corresponding to B 2 O 3 impurities[86]. This indicates that the rinsing with hot DI-water efficiently removes most of the B 2 O 3 residuals. 5.3 Raman measurement When it comes to characterization of the defects in the nanotube walls, the D- band scattering of Raman spectroscopy outperforms TEM[36]. The Raman D-band is activated in sp 2 carbons by various forms of disorder in the carbon network, and the scattering intensity depends on the disorder-induced relaxation of the 1 st order q=0 Raman selection rule for crystalline material. Recent theories for the D-band attribute the scattering to a double resonance scattering process[81]. Low power (P<2mW) micro-raman experiments were performed on the SWNTs in a Renishaw INVIA micro- Raman spectrometer with cooled CCD using an Ar-ion gas laser (514.5 nm) focused onto ~1 μm spot of the sample. In Fig. 5-2, we display Raman spectra in the range of ~ cm -1 for a series of SWNTs films. This region includes the contributions of radial and tangential first-order allowed Raman scattering and D-band scattering. The Raman spectra are labeled (a)-(e): (a) as-prepared (AP) CarboLex (AP-SWNTs); (b) AP- SWNTs after HNO 3 reflux and neutralization to ph=7 (i.e., spectrum for p-swnts); (c) p-swnts, but after an 1100 o C vacuum anneal (2 h, 10-5 Torr) ; (d) B-SWNTs (no vacuum anneal); and Fig. 5-2e B-SWNTs after 1100 o C annealing in vacuum (12 h, 10-5 Torr).

87 74 Fig. 5-2: Room-temperature Raman spectra showing radial R-band (~160cm -1 ), D-ban (~1350 cm -1 ) and G-band (~1590 cm -1 ) of (a) as-prepared (b) after HNO 3 reflux (c) after HNO3 reflux NaOH wash to ph=7 and 1100 o C vacuum anneal (d) after B 2 O 3 -NH 3 treatment (e) afte B 2 O 3 -NH 3 treatment and subsequent 1100 o C anneal. Starting with the bottom Raman spectrum (a) in Fig. 5-2, the AP-SWNTs material can be seen to exhibit a RBM band at ~160cm -1. The frequency of this band is linearly related to the inverse of the individual nanotube diameter[87]. The G-band is located at ~1592 cm -1, and is also in good agreement with the literature[36]. Weak D- band scattering can be observed at ~1350 cm -1 in the AP-SWNTs spectrum, either from ac or imperfect tube walls. Our experience with arc-tubes subjected to various purification treatments[63] indicates that HNO 3 reflux can induce a substantial increase in the wall disorder (defects plus functionalization, i.e. COOH), as observed by an increase in D-band scattering. The effect of the increase in wall disorder due to HNO 3 reflux can be seen in spectrum (b) in the figure, i.e., increased D-band intensity

88 75 along with a line broadening for the RBM bands and G-bands. We have identified these changes in the Raman spectrum with the removal of C-atoms, followed by the addition of functional groups (e.g., -COOH, -OH)[63]. The wall disorder-induced broadening of the Raman bands can be seen with naked eye in (b). However, it was also quantified using a multi-lorentzian line shape analysis. More importantly, we found that a short term vacuum annealing of the p-swnts, i.e., ~2 hr at 1100 o C in 10-5 Torr, significantly reduced the D-band scattering intensity induced by the purification process (by a factor of ~10), as shown in spectrum (c). After the vacuum annealing the D-band of the p-swnts is similar to that of the AP-SWNTs material provided by the manufacturer (c.f., (a)). The spectra (d) in Fig. 5-2 refers to the B-SWNTs before a high-t vacuum annealing. These tubes can be seen to exhibit a strong D-band. Recalling that our thermo-chemical B-doping takes place at 900~1000 o C, had there been no reaction of B 2 O 3 with the tube wall, we would have expected a passive high-t environment to reduce the D-band intensity in the p-swnts material. However, the D-band was found to increase in strength after the thermo-chemical reaction with B 2 O 3. Therefore, the D-band increase is interpreted as evidence that a B-reaction with the tube walls has introduced disorder. To further test the connection of the increased D-band scattering with possible B-doping, we then vacuum annealed (1100 o C for 12 hr) a B-SWNTs sample. The Raman spectrum is shown in (e). Unlike the response of the p-swnts material to such a vacuum anneal, only a small reduction in the D-band scattering is observed after the thermal anneal of the B-SWNTs. The results suggest that the D- band scattering that remains in the B-doped sample is associated with B-substitution but not with the remaining C-atom defects. Furthermore, we can also conclude that the boron atom is stable in the tube walls at the annealing temperature (T=1100 o C), at least over the annealing time of a few hours. Our proposal on the connection

89 76 between the D-band intensity and B-substitution is also consistent with the earlier work in B-doped graphite, where a systematic increase in D-band intensity with increasing B-doping was reported[82]. 5.4 Transmission spectrum Optical transmission spectra were collected with a double-beam UV-Vis-NIR spectrometer (Perkin-Elmer Inc., model Lambda 950). Absorption/reflection structure and loss derived from the substrate were eliminated by dividing the transmission of the film/substrate with the measured value of the clean substrate. Fig. 5-3 shows the optical density OD=(-log 10 (T)) vs photon energy for the p-swnts and B-SWNTs films on ZnSe substrates. ZnSe was chosen because it allows measuring the transmission spectrum throughout the visible and IR ranges. Fig. 5-3: Optical density of thin SWNTs films on ZnSe (a) p-swnts (b) after B 2 O 3 -NH 3 processing and a vacuum anneal at 1100 o C.

90 77 The OD for the p-swnts and B-SWNTs films are shown in Fig. 5-3a and b, respectively. The actual spectra are shown in the insets together with a linear optical background (solid line), identified with the low energy tail of carbon π-π* electronic interband absorption. After the background subtracting, the resulting data (thick black lines) exhibited three broad bands that were fitted to Lorentzians via the method of least squares. The thin red solid line is the result of the fit of the data (background removed) and the fit has been displaced slightly downward for clarity. The individual Lorentzian components are also shown in the figure. The three bands observed in the top spectrum (a) for the p-swnts are labeled according to the literature as E s(m) jj, where jj=11,22 refers to the transitions between the closest and next closest pair of van Hove singularities and s(m) denotes the semiconducting (metallic) tubes[88]. Strictly speaking, the absorption involves the production of the exciton associated with van Hove singularity[25]. The optical transition energies in SWNTs sensitive to excitonic effects have been studied in detail through fluorescence and Raman spectroscopy experiments [36, 89, 90]. Both theoretical calculations and experimental measurements show that the exciton binding energies are large in carbon nanotube, corresponding to a substantial fraction of the band gap[91]. Electron-electron interaction may play an important role in determining the optical transition energies. The position (photon energy) of these exciton bands for the p-swnts is in good agreement with the literature for bundled arc tubes [35, 92]. A single broad Lorentzian was used to fit each E jj s(m) band. In reality, each broad band actually contains many unresolved contributions of various (n,m) tubes with different (but very close) diameter. The bottom spectrum in Fig. 5-3 is the OD for a B-SWNTs film. Similar to the top spectrum for the p-swnts, three broad optical absorption peaks are observed and at similar energies. However, we found that B-doping downshifts each

91 78 of these broad Lorentzians by ~30meV relative to their position in the p-swnts material, while no significant change was observed in the widths of these bands. It is possible that the E jj bands observed in the B-SWNTs material may stem from the absorption of unreacted p-swnts or from both the B-doped and unreacted p-swnts. It is not possible for us to distinguish the contributions of these two types of tubes. We may reasonably expect that the outside tube wall in the large bundles of SWNTs thermo-chemically exposed to B 2 O 3 /NH 3 will be more heavily doped. However, we had no probe available to confirm that primary doping occurs to the outer tubes in the bundle. If the preferential doping of the external surface does occur, then the local doping per tube in the outer tubes would be significantly higher than the value of 1~3 at.% obtained by EELS, which represents an average over the entire bundle. A new optical absorption peak can be observed at low photon energy in the B-SWNTs film (Fig. 4b). The peak is located by a single Lorentzian band (least squares fitting) to be at 0.40 ± 0.02 ev. Borowiak-Palen et al.[86], who first used the B 2 O 3 /NH 3 reaction to prepare B-dope SWNTs, also observed this 0.4 ev band in their OD spectrum and identified it with BC 3 tubes. They made the identification based on the electronic density of states calculations for BC 3. Calculations have also been made regarding the electronic structure of borondoped SWNTs [17, 93]. They all indicate that B-doping at low at.% level will downshift the Fermi level (E F ) of semiconducting tubes. For example, one calculation shows that a few at.% boron substitution downshifts the Fermi level by ~0.7 ev into a hybridized acceptor band. Without calculated electric dipole matrix elements, it is difficult to identify the states involved in the 0.4eV absorption. However, if one requires vertical transitions (k-conserving), it seems most reasonable to identify the 0.4eV absorption band with transitions in the semiconducting or metallic tubes involving hybridized

92 79 boron-carbon bands split by 0.4 ev, with E F located in the highest partially occupied hybridized band. All the results considered (i.e., EELS, Raman scattering and Optical Absorption), we conclude that a thermo-chemical B 2 O 3 /NH 3 reaction with bundled SWNTs under our conditions does produce B-substituted SWNTs. 5.5 Summary We have successfully substituted boron into the carbon sp 2 framework of SWNTs via a post growth doping and have observed significant changes in the electronic properties of the spray-deposited SWNT thin films. We have found that boron doping significantly reduces its sheet resistance. We propose that the lower sheet resistance in B-SWNTs may result from the higher conductivity in the former semiconducting tubes. The preliminary results presented here show that the B-doped SWNTs films can be a promising new electronics material for organic electronic devices because of its many strengths, including low cost, transparency, high T/R Ñ and flexibility, easy deposition on plastic substrates, etc.

93 80 Chapter 6 Post-Growth B-Doping by Thermal Evaporation of B 2 O Post-growth boron doping As discussed in Chapter 5,boron was doped by annealing the mixture of SWNTs and B 2 O 3 powder in the argon and/or NH 3 reactive atmosphere. Several technical problems surfaced, including highly inhomogeneous B-doping, fairly low B 2 O 3 vapor pressure even at 1000 o C(<10-6 bar)[94], heavily doped SWNTs because of the direct contacts of B 2 O 3 melted grains, and formation of troublesome secondary phases, such as boron carbides and boron nitrate which could affect the final electric properties of the SWNTs films. Characterization of the optical and electrical properties of the boron doped SWNTs bundles should be done with highly clean materials. One better solution could be the use of SWNTs prepared by pulse laser vaporization (PLV). Fig. 6-1 shows the TEM micrograph of the purified, PLV produced SWNTs with some impurities, including amorphous carbon, a small amount of graphitic shells and other contaminants. A weak D-band scattering in its Raman spectrum (514.5nm excitation) shows the high quality of the nanotubes[36]. The TEM images in Fig. 6-1b revealed that the tubes form bundles, ~10µm long with a bundle diameter around 20nm. The SWNTs in each bundle have diameter in the range of 0.9~1.3 nm[1].

94 81 Fig. 6-1: (a) HRTEM image of purified undoped SWNTs bundles. (b) low resolution TEM image of purified SWNTs. A novel three-step cleaning was developed to remove the amorphous carbon coating on the bundles, and other unwanted impurities. The first step is the reflex in a heated (60 o C) H 2 O 2 solution for 48 hours to oxidize the amorphous carbon and to weaken the graphitic shells surrounding the metal catalyst particles. Second, 3M HCl is added to the solution, and the 24 hours reflex followed eliminates most of the metal catalyst of the SWNTs sample. Besides, the two refluxes create quite a few defects on the tube walls, providing favorable chance for boron to enter the sp 2 lattice in doping. Finally, the purified material is filtrated with polycarbonate membrane (with pore size of 1 micron), and rinsed with de-ionized water at 100 o C. The ash analysis with TA Instruments, model IR5000 shows a metal concentration of less than 1wt.%. The TEM images in Fig. 6-1 shows that the three-step cleaning removes most of the impurities. The purified SWNTs were divided into two parts, one part was spray-deposited on quartz substrates for boron doping, and the other was kept as the control sample. The idea is to demonstrate that boron doping reduces the sheet resistance of SWNTs

95 82 films without changing the optical transmittance in the visible range. The transmission of the spray-deposited films was evaluated with Perkin-Elmer lamda 950 from 200 to 2500nm. Then, the SWNTs films with the same thickness were divided into two groups; one was vacuum (10-6 Torr) annealed at 1000 o C for 1 hour, and the other was doped with boron oxide in gas phase. Fig. 6-2 is the setup used for boron doping of the predeposited SWNTs films. Since the vapor pressure of B 2 O 3 at ~1000 o C is very low (<10-6 bar) [94], the B 2 O 3 powder source was installed as close as possible to the films in the one end closed quartz tube to maximize the B 2 O 3 vapor concentration above the SWNTs films, and increase the contact time as shown in Fig After effectively degassing at 200 o C for one hour, the system was heated up to 1000 o C for 30 minutes. The interaction of B 2 O 3 vapor and the SWNTs films is expected to generate CO and CO 2, which further promotes boron atoms entering into the walls of the SWNTs. A cartoon describing the doping process is shown in Fig A flow tate of 100 sccm of argon with 10% hydrogen passes the tube so as to keep a high boron oxide vapor pressure, to blow away CO and CO 2, and to create a pressure gradient along the quartz tube.

96 83 Fig. 6-2: Schematics of the CVD setup for doping pre-deposited SWNTs. A half inch tube is placed in a 1 inch diameter quartz tube. Substrate with nantubes pre deposited on is placed closed to the open end of the tube. B 2 O 3 powder was located at the closed end of the quartz to have a higher gas pressure. A flow of 100 sccm of Argon with 10% Hydrogen was used to have a pressure gradient along the tube. Fig. 6-3: Schematics of a nantube with some vacancies on the tube wall in interacting with boron oxide. The final produce is a nanotube with boron substituting the carbon on the tube wall.

97 EELS measurement The local B-content and B-bonding geometry in the B-SWNTs bundles can be studied with EELS. The composition and bonding information of the B-SWNTs were evaluated by ensuring that the EELS probe only the clean bundles well-separated of other non-nanotube fractions in the sample. Fig. 6-4 displays the TEM images of the isolated SWNTs bundles and the associated EELS spectra for high-quality samples produced by doping boron in a gas phase. EELS spectrum shows the characteristic core-level excitations at 188 and 284 ev for boron and carbon, respectively. The boron signal is very weak, however, the π* and σ* features corresponding to the boron edge can still be distinguished. Being consistent with previous reports on B-doped graphite[83] and B-C-N nanotubes [14] the relative intensity of π* peak of boron is larger than that in the carbon border, owing to the larger number of the empty p states localized on boron due to its local deficient of valence electrons. The estimated boron concentration in the nanotubes is less than 1 at.% Fig. 6-4: (a) HRTEM image of B-SWNTs bundle (b) EELS spectrum of the bundle

98 Raman spectroscopy Raman spectroscopy is capable of characterizing the optical and electrical properties of the B-doped tubes. The G-band in graphite involves an optical phonon mode between the two dissimilar carbon atoms A and B in the unit cell. The corresponding mode in SWNTs bears the same name. In contrast to the graphite Raman G-band, which exhibits one single Lorentzian peak at 1582 cm -1 related to the tangential mode vibrations of the C atoms, the SWNTs G-band is composed of several peaks due to the phonon wave vector confinement along the SWNTs circumferential direction and due to the symmetry-breaking effect associated with SWNTs curvature [36]. The G-band feature for SWNTs consists of two main components, one peaked at ~1590 cm 1 (G + ) and the other peaked at about ~1570 cm 1 (G ). The G + feature is associated with carbon atom vibrations along the nanotube axis (LO phonon mode) and its frequency ω G+ is sensitive to charge transfer from the dopant added to the SWNTs (up-shifts) in ω G+ for acceptors, and downshifts for donors as in graphite intercalation compounds (GICs)[95, 96]). Low power (P<2mW) Micro-Raman experiments were performed at room temperature on the SWNTs at a spot of ~1μm in a Renishaw INVIA micro-raman spectrometer. Fig. 6-5 shows that the G + band for the B-SWNTs is up shifted by ~2cm -1. The upshift agrees well with our boron s p-doping. The D-band of the B-SWNTs is bigger than that of the annealed SWNTs, which indicates that boron atoms substitute the tube wall, causing some disorders. A detailed description and interpretation of the D-band were already presented in chapter 5.3.

99 86 Fig. 6-5: Room-temperature Raman spectra showing radial R-band (~184cm-1), D-band (~1350 cm-1) and G-band (~1590 cm-1) of annealed SWNTs and after B2O3 treatment B-SWNTs. 6.4 Resistance measurement The purified SWNTs films deposited on quartz substrate were divided into 4 pieces. We doped half of the sample and vacuum annealed the others. The in line resistance of the doped and undoped SWNT films were measured with four probe technique, as previously described in Fig. 2-12, and the results are listed in Tab Current of 100μA is passed between the outer two probes, while voltage is taken with the two inner probes. The 4 probes are separated at equal distance about 1.58mm

100 87 apart. Tab. 6-1 clearly shows that the inline resistance of the B-doped SWNTs films was reduced by 25-30% with respect to the undoped samples. Tab. 6-1: Sheet Resistance of B-SWNTs and SWNTs films Films T=52% (at 550nm ) Sheet Resistance SWNTs1 SWNTs2 B-SWNTs1 B-SWNTs2 169Ω/ 141Ω/ 117 Ω/ 106 Ω/

101 88 Chapter 7 Post-Growth B-Doping by High Temperature Treatment of Boron Oxide Decorated SWNTs 7.1 Sample preparation Introducing B 2 O 3 in a gas phase lowered the sheet resistance of the films. The Raman and EELS results show that the boron is fitted into the sp 2 lattice. But the high substrate temperature (1000 o C) in boron-doping is a limitation. Since one of the advantages of B-SWNTs over indium tin oxide (ITO) is that it is possible to deposit the material at a low temperature on any substrate. Another problem is that the gas phase B 2 O 3 can t get in connect with all the bundles in a thick film, so only the top layer can be doped. Distributing the boron-based ions and molecules over the SWNT surface via wet chemical route, as a possible solution to the problem, was developed. It was reported that after the B 2 O 3 is dissolved by alcohol, boric acid and boric acid esters form[97]. In dilute aqueous solution of B 2 O 3, the fully dissolved B 2 O 3 reacts with water to form a variety of ionic molecules (BnOm + and BnOmH + )[98]. The drying and degassing, by heating at 180 o C of the aqueous solution of B 2 O 3 in Ar + H 2 atmosphere, are capable of dehydrating all these boron compounds and restoring the boron oxide[99]. Such wet chemical route is expected to homogenously distributing the boron-based ions and molecules over the SWNT surface. In the wet chemical method, first, boron oxide (B 2 O 3 ) powder was dissolved in isopropanol, and the purified SWNTs were added into the isopropanol + B 2 O 3 solution. Second, the mixture was thoroughly dispersed in an ultrasonic bath. Third, the SWNT suspension (B 2 O 3 + SWNTs + isopropanol) was filtered through a polycarbonate

102 89 membrane with a pore size of one micron. The materials, collected from membrane and held in an alumina boat inside the quartz tube filled with Ar+ H 2 gas (10% H 2 ) at atmospheric pressure, were dried, dehydrated and degassed. The setup is shown in Fig The quartz tube was first heated for 2 hours at 180 o C to remove all the moisture with the Ar+H 2 gas flowing at a rate of 100 sccm. Finally, the B-doping was done by heating the system at 1000 o C for 30 minutes. After the system was cooled down the material was retrieved from the alumina boat and the possible residues of physisorbed boron/boron oxides were completely removed by repeatedly rinsing and filtering with polycarbonate membrane, (pore size of 1 micron) using IPA. Based on the result of the core level electron energy loss spectroscopy (EELS) shown in Fig. 7-2c, the at.% of boron was estimated to be around 1%. Fig. 7-1: Schematics of the CVD setup for doping SWNTs. A flow of 100 sccm of Argon with 10% Hydrogen was used.

103 90 Fig. 7-2: TEM and HRTEM images of the SWNTs bundles, (a) TEM image of bundled SWNTs. (b)hrtem image of several p-swnt bundles.(c) EELS spectra of B-SWNT bundle 7.2 Sheet resistance measurment The sheet resistance of the films measured with four-probe technique are shown in Tab Here, we use the van der paul geometry described in Chapter 2.8. The sheet resistance of the undoped sample is 3~6 times higher than that of the doped ones. These data shows that the bulk material s sheet resistance is reduced by doping, and further analysis is described in Chapter 8.

104 91 Tab. 7-1: Sheet Resistance of undoped and B-SWNTs of difference thickness Tansmission 550nm 2% 20% 50% 63% 85% 90% SWNTs 98Ω/ 131Ω/ 539Ω/ 742Ω/ Ω/ Ω/ B-SWNTs 32Ω/ 79Ω/ 133Ω/ 240Ω/ 666 Ω/ 1855 Ω/ 7.3 Summary A novel technique has been successfully developed to mass produce clean boron-doped SWNTs. Since the high temperature step is done before creating the film, the newly-developed approach is suitable for applications in fabrication of electronics devices. We have observed significant changes in the electronic properties of the boron-doped SWNTs. The sheet resistance measurement shows that the B-doping considerably improves the conductivity of the nanotube films. In later chapter, we will show that the B-doping increases the free carriers in the semiconducting tubes.

105 92 Chapter 8 Correlation Between Optical and Electrical Properties The figure of merit (FOM), defined as FOM=T/R, where T is the optical transmission (e.g. at 550 nm for visible light range) and R is the sheet resistance of the film, is an important factor in application of the percolating bundled SWNTs films as a transparent electrode. The previous reports on the subject include a 30Ω/ with more than 70% transmittance in the visible spectrum (2.3) [100], 120Ω/ with 80% (0.67) [101] and 160Ω/ at 87% transparency (0.54) [102], (where their calculated FOMs are listed in the parentheses. For our sample the FOM is around 0.37 with ~70% transmittance. Here, the discussion mainly focuses on lowering the sheet resistance by boron doping. Several factors may possibly affect the FOM. For example, the sheet resistance can be higher or lower than the two-probe resistance, depending on the electrode geometry. A true sheet resistance measurement is to be preferred as described in Chapter 3. Additionally, unintentional doping caused by adding surfactants in the dispersion process of the SWNTs is going to decrease the sheet resistance; this effect should not be permanent and it will partially disappear when the surfactants are evaporated away, but it will mask the actual effect of boron substitutional doping in our study. To eliminate the possibility of charge transfer from the surfactants to the nanotube wall, our nanotubes are just dispersed in IPA which will be totally removed when heated to 120 o C. The lack of surfactants in the suspension leads to the tube bundles agglomeration during the deposition process and the films created are less uniform. If the material were well debundled and dispersed, SWNTs can form more conducting paths with the same amount of nanotubes. This would lead to dramatic

106 93 improvement in lowering the sheet resistance without affecting the transmission. In order to minimize the effect of physisorbed molecules in our experiments, the sheet resistance is measured in vacuum after a 200 o C vacuum degassing. This will remove the oxygen and water molecules which decrease the sheet resistance in an uncoutrolled and unstable way. In some B-SWNTs sample sheet resistance after degasing is almost 100% bigger than that measured in air, this will be further discussed in section UV-Vis Transmission measurement To prepare uniform SWNTs films for optical and electrical measurements, the p-swnts and Boron-reacted SWNTs powders indentified here as (B-SWNTs) were first sonicated briefly in isopropanol (AE Spectral Grade) using a probe sonicator (Misonix Inc., model 2000) for 1 hour at 30W. It should be done carefully, as it may cut the nanotubes into shorter segments [103]. This cutting is undesirable simply because longer tubes percolate at lower density. The SWNTs thin films were prepared by spraying the SWNTs-suspended isopropanol onto the clean quartz substrate via short bursts of air with a small air brush, as shown in Fig. 8-1 (Paasche Inc. model VL#1), such as used to paint fabric. The suspension was continuously sonicated in a bath sonicater at low power during the spraying. The patterned films were made by spraying through a 3x3 mm 2 square hole in a stainless steel mask. The film thickness was controlled by counting the number of bursts. The substrates were maintained at 120 o C during the spraying to rapidly evaporate the isopropanol.

107 94 Fig. 8-1: Spray setup and films of nanotube deposited on quartz substrate The SWNTs films on quartz were vacuum-degassed at 1100 C at 7x10-7 Torr for 12 hours to remove solvent and functional groups attached to the tube walls before electrical contacts were fabricated[63]. Optical transmission spectra were collected with λ950 spectrometer. For each film the transmission at 550nm is recorded. The optical transmission data of the doped (in method discussed in Chapter 6) and undoped samples are presented in Fig The data clearly show that the transmission spectra in the UV-Vis region were not affected by boron doping. The optical transmission (%T) at 550nm almost stayed the same for the two different processes.

108 95 Fig. 8-2: optical transmittance of SWNT films with the same thickness. One went through a doping process the other went through a high temperature annealing(1000 o C). 8.2 Sheet resistance measurement To measure the D.C. sheet resistance, tiny gold and chromium electrode were first evaporated at the 4 corners of the rectangular films, then, the quartz piece was glued to a chip carrier, and finally the four electrodes were carefully wired, as shown in Fig. 8-3.

109 96 Fig. 8-3: Schematic diagram of resistance measurements probe for temperature range 4-500K Four combinations of current and voltage contacts were measured via an integrated current source-dvm system (Keithley; model 2400). The sheet resistance R Ñ was calculated with Eq (van der Pauw method [50]; c.f., Fig. 8-4).

110 97 Fig. 8-4: Schematic view of van der Pauw contacts (1,2,3,4)to SWNT film. Fig. 8-5: Sheet resistance VS degassing time (200 o C in 10-7 torr vacuum) for both B- SWNTs and undoped SWNsT samples.

111 98 Before measuring the sheet resistance by van der Pauw method[50], the sample was carefully degassed in vacuum. Fig. 8-5 shows that the degassing increased the sheet resistance in both samples (undoped and B-doped) by ~100 and 200 Ω/, respectively. Recent experimental studies[104] reported that the measured SWNTs electronic properties are extremely sensitive to the presence of molecular oxygen in the nanotube. The impact of oxygen adsorption on modification of the barrier at the metal-semiconductor contacts was also reported[105]. It has been also observed that some small-gap semiconducting nanotubes exhibit metallic behavior when they are exposed to oxygen[106]. As shown in Fig. 8-5 after a 24 hour vacuum degassing at 200 o C the room temperature resistance almost doubled for the B-SWNTs. 8.3 Correlation between optical and electrical properties Fig. 8-6, Fig.8-7 and Fig. 8-8 reveal the correlation between the optical transmission at 550nm and the sheet resistance of the films. Interesting finding in Fig. 8-6 is that the measured and fitted data of the sheet resistances of the p-swnts and B-SWNTs tend to approach each other in the limit of thick or poorly transmitting films. In this limit, the benefit of B-doping is minimized. In the high transmission region with T >70% in Fig. 8-6, and T>90% in Fig. 8-8, the sheet resistance shows a steep increase, possibly because the film is at the percolation threshold and because the number of conducting paths formed by bundles of nanotubes is rather small in the network. The sample used in Fig. 8-6 has more impurities, as shown in the TEM images in Chapter 5, which leads to the high sheet resistance and a lower percolation threshold starting at around T=70%. Clean material and long tubes are preferred for application in transparent electrodes because a high percolation threshold means we

112 99 can reach the desired sheet resistance with less material and less material means higher transmission. The continuous lines in this plot are the fittings by using the model described next. Fig. 8-6: Transmittance (at 550nm) and sheet resistance of B-doped SWNT (doped as in chapter 5 with the presence of NH 3 )and undoped p-swnt films (processed Arc SWNTs). Insert: Transmittance in the visible range for undoped SWNT p-films.

113 100 Fig. 8-7: Transmittance (at 550nm) and sheet resistance of B-doped (doped as in chapter 6) SWNsT and undoped p-swnts films (purified PLV SWNTs). Fig. 8-8: Transmittance (at 550nm) and sheet resistance of B-doped SWNT (doped as in chapter 7) and undoped p-swnt films (purified PLV SWNTs).

114 101 In order to obtain a mathematical relation between sheet resistance and optical transmittance, we start analyzing the transmission of light through a film with thickness d smaller than the wavelength λ ( Fig. 8-9 ). The transmission is given by: [107, 108] Fig. 8-9: Reflection and transmission through a slab with thickness d 1 4 exp 1 4Rexp. 8-1 Where R is the reflectivity, α is the absorption coefficient, φ r is the phase change upon reflection and β is the phase change on once passing through the medium. In the limit of very thin, highly conducting films in air, it can be shown that Eq. 8-1 can simplify to: /. 8-2 In the case of ε 1 2 << ε 2 2, the sheet resistance is given by:. 8-3

115 Sheet resistance and optical transmission can be related as Eq. 8-4 or Eq. 8-5 [109] Ñ and s s 8-5 where s dc is the DC conductivity and s opt is the optical conductivity at the wavelength at which T is measured, and c is the speed of light. Eq. 8-5 is therefore used to fit our experimental data shown in Fig. 8-6, Fig.8-7 and Fig. 8-8 by adjusting the single parameter β in a least squares fit. The fitted curves in Fig. 8-6 correspond to Ñ and Ñ, i.e., 7170Ω and to processed and B-doped material. 4853Ω, where the superscripts P and B refer, respectively, The boron-doping induced changes to s opt via free carrier absorption and new interband transitions are expected to contribute most significantly to s opt for photon energy<< 1eV. As demonstrated in Fig. 8-2 The B-doping of SWNTs didn t change the optical transmission in the UV-Visible region. We therefore make the approximation s /s 1, at the photon energy of interest ~2.5eV. The ratio of β s can then be written as: s s. 8-6 As a result, the ratio of the values for β yields the ratio of dc conductivities s /s 2/3. For the material acquired as discussed in Chapter 6 (Fig.8-7),

116 2350Ω and Ω, the ratio of the dc conductivities is s /s 1/3.7. For the material acquired as described in Chapter 7 (Fig. 8-8), the ratio of the D.C. conductivities is s /s 1/ Ωand 70Ω, The sheet resistance R Ñ and optical transmission of the p-swnts films have been modeled as a parallel network of the percolating SWNTs bundles[109]. Our approach to the modeling is similar to that of Gruener and co-workers[110]. In Fig. 8-10, we show a schematic diagram of a two-dimensional percolating nanotube bundle network in a plane. Four conducting paths have percolated between the electrodes shown in red. The nanotubes within a bundle are considered to be a set of parallel resistors. The distribution of the bundle diameter was ignored. Suppose the individual nanotube resistance is r, we expect the following hierarchy: pristine semiconducting tube (r s ) >> B-doped (formally semiconducting tubes) (r d ; the subscript d refers to doped) > metallic tube (r m ). We may consider the possibility that not all semiconducting tubes in the network are significantly B-doped. The resistance for each bundle is taken as the parallel combination of the various tubes in the bundle. Taking into account the resistance hierarchy, the bundle resistance for a pristine bundle (no B-doping) can be written as, where r m is the resistance of one metallic nanotube, and n m is the number of metallic tubes in the bundle. Similarly, for B-doped bundles, and using r d <<r s, we can write the bundle resistance as, where n d is the number of degenerately B-doped tubes per bundle. Assuming the electrically interconnected bundles form conducting paths as shown schematically in the Fig. 8-10, and that those paths are primarily in parallel to

117 each other, then the sheet resistance of the square-shaped p-swnts and B-SWNTs films can be written as Eq. 8-7 and Eq Ñ 8-7 Ñ 8-8 Fig. 8-10: Schematic percolation of a two-dimensional nanotube network made of SWNT bundles. where R C and Rc are, respectively, the total contact resistance between the p-swnts and B-SWNTs bundles in a typical percolating path. The summations in Eq. 8-7 and Eq. 8-8 are over the series of connections of bundles in a percolating path. The variables N and N are, respectively, the number of parallel equivalent conducting routes between the current contacts of the p-swnts and B-SWNTs films. Solid lines in Fig. 8-6, Fig.8-7 and Fig. 9-1 are fits to the data based on a model of parallel percolating paths, as shown schematically in Fig Therefore, using Eq. 8-7 and Eq. 8-8, the measured values for T and the assumption that the number of percolating parallel paths are approximately equal, i.e., N=N, we must have. To proceed further, one has to make some assumptions about the bundle contact resistances R C and R C. If we assume R C ºR C, i.e., the contact

118 resistances are not changed by thermo chemical B-doping, then., then Eq. 8-7 to Eq. 8-6 show that B-doping leads to at least a factor of 3 decrease in the bundle resistance. However, the assumption that R c =R c may be incorrect. The B- doping may possibly lower the contact resistance between the bundles. The problem of contact resistance reduction remains open and need further study. To proceed further, one has to know the relative contribution made by the contact resistances Rc and Rc. For example, it is commonly accepted that the contact resistance dominates the path resistance in nanotube films. If this is the case, then doping may lower the contact resistance between bundles by a factor of 2/3~1/3.7. One may argue that the barrier to charge transfer from bundle to bundle has been lowered as the bundle conductivity increases. But other explanations are also possible. To further understand the contact resistance, the temperature dependent sheet resistance is measured, as discussed in the following Chapter. 105

119 106 Chapter 9 Temperature Dependent Resistance of SWNTs and B-SWNTs 9.1 Introduction The temperature dependence of the electric resistance in homogeneous ordered materials (i.e. metals or semiconductors) can provide insite on the scattering mechanisms ruling the transport of the electric carriers (i.e. electron and/or holes) in the material. For example, near room temperature, the electrical resistance of a typical semiconductor decreases with rising temperature, while for a metal it increases. At temperatures lower than Debye temperature, with decreasing temperatures the resistance of a metal decreases as T 5, in this range the main mechanism is the scattering of electrons by phonons. At even lower temperatures, the dominant scattering mechanism is the electron-electron interaction; in this range the resistance decreases as T 2. At some temperature, the scattering by impurities will dominate the electric resistance of the metal, when this happens the resistance saturates to a constant value. The behavior of semiconductors depends on whether the material is intrinsic (no doping) or extrinsic (p- or n-doping). In these materials the valence band and the conduction band are separated by an energy gap (E g ). For intrinsic semiconductors, the electric resistance decreases exponentially with an increase of temperature. At absolute temperatures near zero degree, an intrinsic semiconductor should theoretically behave as a perfect insulator since no free carriers are available to contribute to the electric transport of charge. As the temperature increases more

120 107 electrons acquire enough energy to make a transition from the valence to the conduction band, leaving holes (positive charges) in the valence band. At this point both charge carriers (electrons and holes) can flow freely through the material. Depending on the band gap energy, these electronic transitions can start sooner or later; and the temperature dependence of the resistance changes from one compound to another. For extrinsic (p- or n-doped) semiconductors the temperature profile of the resistance is somehow more complicated. At very low temperatures, an increase in temperature will ionize the donor (or acceptor) impurities, i.e. electrons (or holes) are injected into the conduction (or valence) band. This produces an abrupt decrease of the electric resistance. After most of the donors or acceptors have been ionized, the resistance keeps increasing but slightly due to the reducing mobility of carriers similar to a metal. For higher temperatures these materials will behave like an intrinsic semiconductor because the carriers introduced by the impurities (donors or acceptors) become negligible compare to the thermally generated carriers (inter-band transitions). The SWNTs films is even more complex; not only because the complexity of the individual bundles (composed by metallic and semiconducting SWNTs) but also because it is a network in which the bundle-to-bundle contact resistance plays an important role in its resultant electric behavior. So, the SWNTs film should be treated as an inhomogeneous disordered material. Different models have been developed in order to understand the main scattering mechanisms of the SWNTs films. For instance, the fluctuation induced tunneling model and the Mott variable range hopping model that we will describe in details next in this chapter.

121 108 The resistance of the SWNTs networks in the films results from two parts: one is the bundle resistance, and the other is the bundle-bundle junction resistance (Fig. 9-1). Each nanotube bundle is made up of about 10 to 100 tubes, being either metallic or semiconducting. Since the resistivity of the metallic tubes is 4 to 5 orders of magnitude smaller than that of the semiconducting tubes[25], the metallic ones in the bundle will short the semiconducting ones, the bundle resistance is dominated by the metallic tubes. At low temperature the semiconducting tubes are frozen out (no carriers are thermally generated) the bundles should show more metallic behavior. A linear term of the temperature dependent resistance is expected to come from the metallic tubes in the bundle. There will also be resistance at the bundle-bundle junctions. Conduction of the films may possibly originate from the tunneling or hopping (i.e. phonon-assisted tunneling between electronic localized states centered at different positions.) between the metallic region of one bundle to the metallic region of the other (Fig. 9-1). As the B-doping turns the semiconducting tubes on each bundle into more metallic ones, the bundle resistance decreases. As for the contacts between bundles, the high conductivity area will increase due to the transformation of semiconducting tubes to the B-doped semiconducting tubes, and the contact resistance decreases. As a result, the overall conductivity of the network is significantly improved.

122 109 Fig. 9-1: Schematic diagram of carbon nanotube network and bundle-bundle junction. M stands for metallic tubes and S stands for semiconducting tubes. The impact of temperature on the sheet resistance was experimentally evaluated in a closed cycle refrigerator (CCR), with the temperature monitored by a silicon diode sensor (Fig. 9-2). The data were acquired in a temperature range from 8 ~ 310K. The sample is first degassed in vacuum (10-7 Torr) at 200 o C for 24 hours. After degassing and cooling down to room temperature, the sample is transferred into the CCR, and the base pressure is rapidly pumped down to 10-6 Torr via a diffusion pump to minimize the possible air contaminations.

123 110 Fig. 9-2: Closed cycle refrigerator (CCR) 9.2 Fluctuation induced tunneling (FIT) Kaiser et al. proposed that the temperature dependence of conductivity of SWNTs mats may be similar to that of organic conducting polymers [111, 112], typically showing a crossover from metallic to non-metallic behavior as temperature decreases. The resistivity can be modeled with Eq In this model the majority of the nanotube bundles have a linear metallic resistivity and there is a potential barrier between bundles[111]. exp 9-1