Temperature Effects on the Electronic Conductivity of
Single-Walled Carbon Nanotubes
by
Mark Daniel Mascaro
Submitted to the Department of Materials Science and Engineering
in partial fulfillment of the requirements for the degree of
Bachelor of Science in Materials Science and Engineering
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
June 2007
@ Massachusetts Institute of Technology 2007. All rights reserved.
//
................
. ..... . .. .
Author.............................
Department of Materials Science and Engineering
May 21, 2007
Certified by................
Francesco Stellacci
Assistant Professor
Thesis Supervisor
Accepted by.........
Caroline Ross
Chair of the Undergraduate Committee
MASSACHLSEMTTS INSTITUTE
OF TECHNOLOGY
JUL 17 2008
LIBRARIES
ARCHNES
Temperature Effects on the Electronic Conductivity of
Single-Walled Carbon Nanotubes
by
Mark Daniel Mascaro
Submitted to the Department of Materials Science and Engineering
on May 21, 2007, in partial fulfillment of the
requirements for the degree of
Bachelor of Science in Materials Science and Engineering
Abstract
The room-temperature electronic conductivity and temperature dependence of conductivity
were measured for samples of carbon nanotubes of three types: pristine; functionalized
with a nitrobenzene covalent functionalization, which are expected to display poor electronic
conductivity; and functionalized with a carbene covalent functionalization, which are expected to display pristine-like conductivity. Measurements were taken via four-point probe
measurement across palladium contacts on a silicon surface coated with a distribution of
nanotubes. Room-temperature measurements indicate that carbene-functionalized tubes
do exhibit significantly greater conductivity than nitrobenzene-functionalized tubes, but
also significantly less than pristine tubes. Statistically different distributions of resistances
observed in similarly prepared samples indicate that this measurement technique is strongly
affected by uncontrollable and so far uncharacterizable parameters of the employed sample
preparation technique. Measurements at varying temperature indicate the expected linear
relationship of resistance and temperature is dominated in pristine and carbene samples by
a effect, possibly contamination-related, which significantly and permanently increases the
resistance of samples after cycling to high temperatures, and which occurs repeatedly with
additional cycling. Carbene-functionalized samples were observed to exhibit similar temperature behavior to pristine samples, while nitrobenzene-functionalized samples displayed
erratic, unpredictable behavior.
Thesis Supervisor: Francesco Stellacci
Title: Assistant Professor
Contents
1 Introduction
1.1
Methods of Achieving Dispersion and Solvation of Nanotubes .
1.2
A Theoretical Conductivity-Preserving Covalent Functionalization .
1.3
Thesis Goals . . .
..
.
............................
2 Experimental Procedure
2.1
2.2
15
Sample Preparation ......................
. . .. . ..
16
2.1.1
Carbene Samples ...................
. . .. . ..
16
2.1.2
Nitrobenzene Samples
. . .. . ..
17
2.1.3
Pristine Samples
. . .. . ..
17
2.1.4
Electrode Fabrication .................
. . .. . ..
18
Measurement Procedures ...................
. . .. . ..
19
2.2.1
Multi-Pad Temperature Variation Measurements. .
. . . . . . .
19
2.2.2
Point-Dwell Equilibration Measurements . . . . . .
. . . . . . .
19
2.2.3
Single-Pad Temperature Cycle Measurements
. . . . . . . 20
................
...................
. . .
3 Results and Discussion
3.1
3.2
21
Room-Temperature Measurements . . . . . . . . . . . . . .
. . . . . . . 21
3.1.1
Measurement Data ..................
. .. . .. . 21
3.1.2
Statistical Analysis of Conductivity Measurements.
. . . . . . . 24
Multi-Pad Temperature-Variation Measurements . . . . . .
. . . . . . . 28
3.3
Point-Dwell Equilibration Measurements ...................
3.4 Single-Pad Temperature Cycle Measurements
4 Conclusion
A Aggregate Voltage and Resistance Curves
B Single-Pad Heat Ramp Data
. ........
.
........
30
32
List of Figures
1-1 Representations of the S, O closed, and O open configurations for functional
group attachment.
................................
2-1 Schematic of electrodes and contact pads.
13
. ...................
15
3-1 A typical well-formed I-V curve ..........................
22
3-2 Voltage and resistance logarithmic plots for carbene prior. . ..........
23
3-3 AFM image of interdigitated electrodes.
. ..................
.
24
3-4 Tabulated low-bias and full-curve resistance values. . ..............
25
3-5 Temperature-dependent resistance data for carbene BL303...........
29
3-6 Temperature-dependent resistance data for nitrobenzene tw2-172-3. ......
29
3-7 Temperature-dependent resistance data for pristine SP0300 20........
.
30
3-8 Resistance values for rapid sampling of a single pad with time........
.
31
3-9 Temperature cycle for carbene prior pad e7, shown as absolute resistance and
as normalized to 50 'C and baseline one. . ..................
.
33
3-10 Temperature ramp for carbene prior pad c4 in 10-degree intervals. .......
35
3-11 Temperature cycle of pristine prior, pad b2, and its baseline resistance. . . .
36
3-12 Temperature cycle of pristine prior, pad b4 and its baseline resistance.....
36
3-13 Temperature cycle of nitrobenzene tw2-172-3, pad c9, and its baseline resistance. 37
3-14 Temperature ramp for nitrobenzene tw2-172-3 pad e3 in 10-degree intervals.
38
A-1 Voltage and resistance logarithmic plots for carbene prior. . ..........
44
A-2 Voltage and resistance logarithmic plots
carbene BL303.....
A-3 Voltage and resistance logarithmic plots
carbene BL320.
A-4 Voltage and resistance logarithmic plots
nitrobenzene ethanol.
A-5 Voltage and resistance logarithmic plots
nitrobenzene tw2-172-3.
A-6 Voltage and resistance logarithmic plots
pristine SPO300 20.
A-7 Voltage and resistance logarithmic plots
pristine prior ......
B-i Resistance for temperature cycle of carbene prior, pad c4, and its baseline
resistance, shown as resistance values and normalized to baseline one.
B-2 Temperature cycle of carbene prior, pad c5, and its baseline resistance.
B-3 Temperature cycle
carbene prior, pad c6, and its baseline resistance.
B-4 Temperature cycle
carbene prior, pad c7, and its baseline resistance.
B-5 Temperature cycle
carbene BL303, pad d3, and its baseline resistance.
B-6 Temperature cycle
carbene BL303, pad d5, and its baseline resistance.
B-7 Temperature cycle
pristine prior, pad bl, and its baseline resistance.
B-8 Temperature cycle
pristine prior, pad b2, and its baseline resistance.
B-9 Temperature cycle
pristine prior, pad b3, and its baseline resistance.
B-10 Temperature cycle
pristine prior, pad b4 and its baseline resistance..
B-11 Temperature cycle
pristine prior, pad b5, and its baseline resistance.
B-12 Temperature cycle of pristine prior, pad b6, and its baseline resistance.
B-13 Temperature cycle of nitrobenzene tw2-1l72-3, pad c3, and its baseline resistance.
B-14 Temperature cycle of nitrobenzene tw2-1i72-3, pad c4, and its baseline resistance.
B-15 Temperature cycle of nitrobenzene tw2-1 72-3, pad c8, and its baseline resistance.
B-16 Temperature cycle of nitrobenzene tw2-1 72-3, pad c9, and its baseline resistance.
B-17 Temperature cycle of nitrobenzene tw2-1 72-3, pad dl, and its baseline resistance.
B-18 Temperature cycle of nitrobenzene tw2-172-3, pad d5, and its baseline resistance. 55
Chapter 1
Introduction
Carbon nanotubes display unique electrical [7] and mechanical [14] properties of great interest
in materials applications. To briefly touch upon mechanical properties, nanotubes exhibit
a spectacularly high elastic modulus and are highly mechanically resilient, undergoing a
reversible buckling transition rather than a breakage when subjected to severe stress [9]. For
this reason, nanotubes are being evaluated for use in a number of mechanical applications,
such as well-defined, mechanically robust scanning probe microscope tips capable of imaging
fine recesses due to their large aspect ratio [8]. The electrical properties of nanotubes are even
more unique, in that they are tunable over a wide range; depending upon the chirality of the
nanotube, anything from metallic conductivity to a considerable band gap (approximately
0.5 eV at 1.5 nm diameter) may be observed [7, 9]. Nanotubes can sustain very high current
density [18] and show great potential as nanoscale, virtually one-dimensional electronic
elements or "quantum wires" due to their ballistic conduction capabilities [15, 17]. Beyond
straightforward applications of their mechanical and electrical properties, nanotubes are
being considered for a host of other applications based on alteration of their surface chemistry,
for example functionalization with biomolecules to form tiny, specific electronic biosensors,
an approach which also capitalizes on the aforementioned electronic versatility [5].
This thesis is concerned with analysis of the electrical conductivity of nanotubes which
have been chemically functionalized in two substantially different ways, compared with nanotubes in the pristine state, and also with the change in this conductivity with temperature.
An understanding of the temperature behavior of conductivity is a valuable tool in designing
electronic systems based on nanotubes. Before discussing the functionalization schemes in
question, however, a brief discussion of the behavior of nanotubes in solution and the general
need for nanotube functionalization techniques is in order.
1.1
Methods of Achieving Dispersion and Solvation of
Nanotubes
In order to effectively include nanotubes in any sort of system or device, there must be an
effective way to manipulate and control them. To create a composite material, at least a
good dispersion is required [10]; to control the assembly of nanotubes in any device requires
both good dispersion/dissolution and a manipulable chemical functionalization [6, 2].
With regards to achieving a good dispersion, nanotubes in their pristine state are exceptionally poorly-behaved:
nanotubes have significant van der Waals adhesion to each
other, as well as hydrophobic interactions with solvents for which this is a concern, and
accordingly tend to clump together in ropelike bundles of parallel tubes [16, 9]. As a result
of this strong adhesion, good dispersion is difficult to achieve. An alternative route is to
achieve a solution, rather than a dispersion, of nanotubes; unfortunately, nanotubes in their
pristine state are also insoluble in organic solvents [4, 16].
However, through the use of
covalent and non-covalent functionalization, surfactants, and various purification methods,
good dispersion/dissolution of nanotubes has been achieved.
It is possible to employ a surfactant to disrupt nanotube van der Waals adhesion as well
as hydrophobic interactions; sodium dodecyl sulfate (SDS) has been shown to produce a
good suspension with sonication [3]. This method avoids damaging the nanotubes, unlike
oxidation purification techniques, and may even be employed to size-separate nanotubes by
filtration. The surfactant can then be washed free by organic solvents. This achieves good
dispersion, and leaves the nanotubes in the pristine state, though this procedure does not
itself provide a chemical handle. Alternatively, nanotubes can be made soluble in water by
wrapping linear polymers around them, disrupting hydrophobic and inter-tube interactions
[2]. This method can provide a functional handle and can allow nanotubes to be solubilized
with other organic molecules.
Oxidative purification methods used to clean graphite and amorphous carbon from nanotube samples, such as sonication in acids or treatment with piranha, additionally have a
functionalizing effect [1]. These procedures cut treated nanotubes into a number of openended cylinders, the ends of which now display oxygenated functionalities such as carboxyl
groups [1]. Samples purified by this method form a stable suspension in water with the help
of surfactants [13]. While method is more stable than the use of a surfactant alone, it does
damage the nanotubes significantly, and it is inapplicable for nanowires or in other cases
where long, continuous nanotubes are desired.
Covalent chemical functionalization offers both a manipulable chemical handle and the
possibility of tailoring surface chemistry and electronic structure[6]. This makes it a powerful
option, so long as one is careful of dispersion; functionalizing tubes still in their natural
bundled state leads to problematic uneven functionalization [10]. In addition to providing
chemical control, functionalization can increase solubility in water, and so it seems an
ideal choice [10]. However, forming covalent bonds to the nanotube sidewalls disrupts the
electronic structure; the additional covalent bond converts a previously sp 2-hybridized carbon
into an sp3 -hybridized carbon, eliminating the symmetry of the pristine nanotubes [11]. The
sp 3-hybridized carbon serves as a scattering point, and covalent functionalization which
bonds to the nanotube sidewall in this fashion destroys the delocalized electronic structure
observed in pristine nanotubes [2]. A primary effect of this is a massive drop in conductivity
across the tube for even modest amounts of functionalization, both predicted theoretically
and observed experimentally [1, 12]. In the interest of taking advantage of nanotubes in
electronic systems, this loss of conductivity cannot be allowed. The ideal solution would
provide all the benefits of chemical functionalization without affecting conductivity, and
indeed, a covalent functionalization that in principle does not affect conductivity has been
theorized by Lee and Marzari [12].
1.2
A Theoretical Conductivity-Preserving Covalent Functionalization
Young-Su Lee and Nicola Marzari have theorized a form of covalent functionalization based
on carbene or nitrene chemistry which, according to simulations, will preserve conductivity
[12]. This functionalization bonds to two sidewall carbons, and may cleave the bond between
the two in the process, forming a three-membered ring and preserving the uniform sp 2
hybridization of a pristine nanotube. This attachment may occur in three ways, two without
sidewall bond cleavage and one with: the "S" configuration, in which the functionalizing group
binds between two carbons forming a line skew to the nanotube length axis, and in which
the sidewall bond remains intact; the "O closed" configuration, in which the functionalizing
group binds to a pair of carbons forming a line orthogonal to the nanotube length axis, and
in which the sidewall bond remains intact; and the "O open" configuration, in which the
cleavage is orthogonal to the nanotube length axis, and the sidewall bond is broken. These
three configurations are shown in figure 1-1.
Functionalizing an O site pair is always more energetically favorable than binding to an S
site pair. Additionally, the O closed configuration is unstable, and the O open configuration
is preferred. This is good news for electronic conductivity: it is the O open configuration
which preserves sp2 hybridization, and the functional group is a very weak scattering center.
Without the strong scattering centers associated with sp3 hybridization, a tube functionalized
with this chemistry in the O open configuration should approach the conductivity of a pristine
nanotube. Simulation of O open functionalization with CC12 showed that with a closely12
&
.'3
i
"
kt
(a,
rr\
·
S
rnl
x
'4
0 closed
0 open
Figure 1-1: Representations of the (a) S (b) O closed and (c) O open configurations
for functional group attachment. Image taken from [121.
spaced distribution of 30 functional groups on an otherwise infinite nanotube, conductivity is
reduced only 25%; by comparison, simulated functionalization with hydrogen to a comparable
degree showed reduction of conductivity to practically zero.
This functionalization scheme, is of great potential value. It offers all the benefits of other
covalent chemical functionalization schemes while retaining most of the electronic properties
of pristine nanotubes. However, its theoretically low impact on conductivity must be verified
experimentally.
1.3
Thesis Goals
This thesis is concerned with analysis of electrical conductivity measurements taken by
a standard 4-point probe technique across a pair of interdigitated electrodes placed on
a surface with deposited single-walled carbon nanotubes. This method was employed for
samples of three types: pristine nanotubes, nanotubes functionalized with nitrobenzene in
a traditional conductivity-destroying covalent scheme, and nanotubes functionalized with
the theoretically conductivity-preserving carbene chemistry of Lee and Marzari.
Firstly,
this thesis is concerned with determining whether or not this measurement technique is
valid for determining the effects of carbene chemistry on conductivity, that is, whether the
conductivity of nanotubes so functionalized is comparable to that of pristine nanotubes.
Secondly, this thesis is concerned with establishing the effect of temperature, as well as
repeated temperature cycling, on samples of all three aforementioned sample types.
The synthesis procedures for the various samples are discussed in the first part of chapter
2, followed by the conductivity measurement setup and parameters.
Chapter 3 presents
and discusses the results of room-temperature resistance measurements, temperature-variant
measurements, and measurements of changes in the room-temperature resistance following
repeated temperature cycling, in that order. Conclusions are discussed in chapter 4.
Chapter 2
Experimental Procedure
To evaluate conductivity, four-point probe resistance measurements were performed on electrodes formed on silicon wafers containing a distribution of nanotubes on their surfaces. One
measurement point consists of a pair of palladium electrodes attached to larger palladium
contact pads, deposited by electron beam and optical lithography, respectively; since the
probe measurement is destructive of the underlying surface, a large contact area is necessary
for repeated measurements. The palladium electrode fingers form an interdigitated pair of
two-prong electrodes as shown in figure 2-1, which are predicted to be spanned by some
number of nanotubes on the surface.
200 pm
I :i
-L-------
Ic
r . ...
.
.
I
2.5 pm
6.5 pm
Figure 2-1: Schematic of electrodes and contact pads.
Each sample wafer was prepared with five columns of ten pairs of pads, totaling 50
measurement points per wafer. Several wafers of each functionalization type-five carbene
wafers, three nitrobenzene wafers, and three pristine wafers-were examined. These wafers
were produced by various protocols; each sample is named and discussed below.
2.1
Sample Preparation
All samples were prepared in advance by Tan Mau Wu of the Stellacci group. Carbenefunctionalized nanotubes were synthesized by Brenda Long of the Stellacci group, except
for carbene prior, synthesized by Benjamin Wunsch; nitrobenzene-functionalized nanotubes
were synthesized by Tan Mau Wu. Pristine single-walled carbon nanotubes of the HiPco
type were obtained from Carbon Nanotechnologies Inc.
2.1.1
Carbene Samples
Nanotubes were functionalized with dichlorocarbene by the procedure described in [6]. BL303,
BL3337 and BL320 were prepared from nanotube batch P0304. BL319 was prepared from
nanotube batch SP0300. Carbene prior was prepared from batch P0276.
The five carbene-functionalized samples, labeled BL303, BL319, BL320, and BL337, were
prepared by identical procedures.
Carbon nanotubes were added to dimethylformamide
(DMF) and sonicated for 30-120 minutes, until they were well-dispersed. Wafer surfaces
were cleaned first with piranha, a 3:1 solution of sulfuric acid and hydrogen peroxide,
for 15 minutes. Next, samples were exposed to SC1, a 5:1:1 solution of water, hydrogen
peroxide, and ammonium hydroxide for 15 minutes at 80 'C. Thirdly, samples were exposed
to oxygen plasma for five minutes, and then to HCl vapor for 60 seconds. Finally, samples
underwent vapor deposition of aminopropyltriethoxysilane (APTES) for 120 minutes to form
a monolayer. This four-step cleaning procedure was common to all samples.
Cleaned samples were immersed in the nanotube solution overnight for approximately 36
hours. Samples were then dipped briefly into DMF and blow-dried with nitrogen. Finally,
samples were sonicated in DMF for 1 minute.
The sample BL320 was found to be less functionalized, as x-ray photoelectron spectroscopy (XPS) indicated lower levels of chlorine.
2.1.2
Nitrobenzene Samples
Nanotubes were functionalized by the procedure described in [10]. All nitrobenzene samples
were prepared from batch P0304.
The nitrobenzene-functionalized sample labeled tw2-172-3 was prepared as follows. Carbon nanotubes were added to DMF and sonicated for 30-120 minutes, until well-dispersed.
The substrate was subjected to the common four-step cleaning procedure. A drop of the
nanotube solution was cast onto the sample and allowed to sit for 60 seconds, following which
the sample was blow-dried with nitrogen. No post-deposition cleaning was performed.
The nitrobenzene-functionalized sample labeled prior ethanol was prepared by first sonicating nanotubes in ethanol for 30-120 minutes, until well-dispersed. The substrate was
subjected to the common four-step cleaning procedure and allowed to soak in the ethanol
nanotube solution overnight, approximately 15 hours. The sample was then sonicated in
ethanol for 1 minute.
The nitrobenzene-functionalized sample labeled prior IPA was prepared by an identical
procedure to prior ethanol, using isopropyl alcohol (IPA) in place of ethanol.
2.1.3
Pristine Samples
The pristine sample labeled sp0300 12 was prepared by first sonicating nanotubes in dichlorobenzene (DCB) for 30-120 minutes, until well-dispersed.
The sample was subjected to the
common cleaning procedure. A drop of the DCB nanotube solution was cast onto the sample
and allowed to evaporate in an atmosphere of DCB vapor. The sample was then sonicated
in DCB for 15 minutes.
The pristine sample labeled sp0300 20 was prepared similarly, except that the drop of
the DCB nanotube solution cast onto the sample was allowed to sit for 1 minute before
being blow-dried with nitrogen, and that no post-deposition cleaning was performed. The
pristine sample labeled pristine prior was also prepared similarly, except that the substrate
was immersed in the nanotube solution overnight, approximately 15 hours, and that no
post-deposition cleaning was performed. Both sp0300 12 and sp0300 20 were prepared from
nanotube batch SP0300; pristine prior was prepared from batch P0304.
2.1.4
Electrode Fabrication
The interdigitated electrode fingers were fabricated on the sample surfaces via electron beam
lithography. A drop of poly(methyl methacrylate) (PMMA) was employed as a resist, spun
on each sample at 3000 rpm for 90 seconds.
Samples were then baked at 130 'C for
90 minutes. Electron beam writing was performed by the staff of the Scanning-ElectronBeam-Lithography Facility at MIT's Research Laboratory of Electronics. The PMMA was
developed for 60 seconds in 2:1 IPA:methylisobutylketone and a 25 nm layer of palladium
was deposited by electron beam evaporation. Samples were then immersed in acetone for
approximately 1 minute and then sonicated in acetone for 10 seconds.
The contact pads were fabricated by optical lithography. Samples were baked at 130
oC
for 5 minutes. A drop of NR-7 negative photoresist was spun on each sample at 3000
RPM for 60 seconds.
Samples were then baked on a hotplate at 150 'C for 60 seconds.
The resist was exposed with an EML high-res tool with ultraviolet filter for 15 seconds, and
samples were again baked at 100 'C for 60 seconds. The photoresist was developed with
RD-6 developer and water for 15 seconds. A 25-50 nm layer of palladium was deposited by
electron beam evaporation, and the samples were immersed in acetone for 2-3 minutes for
metal liftoff.
2.2
Measurement Procedures
Measurements were carried out using an Agilent 4155C semiconductor parameter analyzer
and a Cascade Microtech probe station with temperature-controlled chuck. Temperaturecontrolled measurements employed a Temptronic temperature controller. Four-point probe
measurements were taken, placing two probes on each contact pad, and the data were read
indirectly from the semiconductor parameter analyzer over a GPIB interface. Samples were
scanned from -1 V to 1 V, typically at 50 mV intervals. This simple procedure was employed
for all room-temperature measurements; temperature-variant measurements are described
below.
2.2.1
Multi-Pad Temperature Variation Measurements
In this set of measurements, the sample was brought to a particular temperature and a series
of pads was measured before changing the temperature. In this way, the entire sample is
slowly cycled from low to high temperature and back, taking a measurement of all pads
at each temperature plateau. Probes were lifted away from the sample during temperature
adjustment, as it was observed that probes would slide randomly and scratch away the
contact pads if left in contact during temperature adjustment. One carbene (BL303) and one
nitrobenzene sample (tw2-172-3) were measured at 0, 25, 80, and 120 "C, with an additional
post-ramp 25 "C measurement. One pristine sample (SP0300 20) was measured at 25, 80,
100, and 120 "C, with an additional post-ramp 25 "C measurement. 50 mV scanning intervals
were used.
2.2.2
Point-Dwell Equilibration Measurements
To determine whether the sample required time to equilibrate after the thermal controller
indicated that a certain temperature had been achieved, a test was made in which carbene
BL303 pad e8 was subjected to several rapid, repeated measurements at a low bias of
19
+0.3 V immediately after the temperature controller indicated the target temperature.
Measurements were taken as rapidly as the machine could complete them, approximately
every four seconds, and measurements were made from 50 to 200 'C at 10-degree intervals.
Probes were left in contact with the pad during the heating process to ensure immediate
readiness upon reaching the target temperature.
2.2.3
Single-Pad Temperature Cycle Measurements
Two types of single-pad heat ramp measurements were taken. In the first type, intended
primarily to show the behavior of a single pad with changing temperature, a single pad was
chosen as the target. Probes were placed in contact with the sample and a measurement was
taken with 20 mV scanning intervals. The temperature was ramped up in 10 degree increments from 50 to 200 'C, leaving the probes in contact with the sample during temperature
changes and repositioning them only if contact was broken. An additional measurement was
taken at 50 'C post-ramp.
In the second type, intended to show the behavior of a single pad with changing temperature as well as the behavior of pads after repeated temperature cycling, a wafer was selected,
and a selection of pads from that wafer was chosen as the target. First, baseline resistance
measurements were taken for each pad. Probes were then brought into contact with a chosen
pad while measurements were taken at 50, 100, 150, and 200 'C on both the ramp up and the
ramp down. Probes were left in contact during temperature changes and were repositioned
only if contact was broken. In this way, all pads on the wafer were subjected to a temperature
cycle while one was measured. Following this temperature cycle, baseline measurements of
each pad were repeated, another pad was subjected to measurement during an additional
temperature cycle, and so forth, until all pads had been measured during a temperature
cycle.
Chapter 3
Results and Discussion
3.1
3.1.1
Room-Temperature Measurements
Measurement Data
An ideal I-V curve for this setup has a predominantly linear shape, as shown in figure 3.1.1.
The wider blue curve represents the absolute value of the measured current; the thinner
green curve marks the instantaneously-calculated resistance, that is, the value of V/I at that
point. Samples displayed wide variation across their pads, often over 3 orders of magnitude,
with a few notable exceptions. Figure 3-2, showing voltage and resistance curves for carbene
prior, is shown here as an example; figures A-2 A-3 show the set of voltage and resistance
curves for BL303 and BL320, respectively, and are found in appendix A. BL319 had only
two functional pads, and BL337 had zero; accordingly there are no charts of this type for
those samples. For purposes of this analysis, a nonfunctional pad is one which displays a
noisy curve in the femtoamp range and/or which displays a nonmonotonic I-V curve, both
of which indicate that the observed data is not an accurate representation of any number of
nanotubes with good contact. Carbene prior is the most consistently-behaved of the samples,
displaying a tightly-packed distribution of curves and spanning approximately two orders of
magnitude in current distribution. Carbene BL303 is more typical, spanning five orders.
7
.
t I
•o
.4
4,,
4
Voltage (V)
Figure 3-1: A typical well-formed I-V curve, taken from carbene prior, pad e7.
Nitrobenzene samples ethanol and tw2-172-3 are shown in figures A-4 and A-5, respectively; IPA displayed only one good test site and is accordingly omitted.
In this case,
both samples showed a spread over at least four orders of magnitude. The I-V curves for
nitrobenzene ethanol indicating high resistance also show a significant level of noisiness, a
trend which is observed in almost all samples for pads indicating very high resistances. These
curves may be due to poor metal-nanotube contact.
Pristine samples SP0300 20 and prior are shown in figures A-6 and A-7, respectively;
SP0300 12 contained no functional test sites and is omitted. Pristine prior varies over only
one order of magnitude with a few significant outliers, most likely indicating relatively even
distribution of nanotubes on the surface. In addition, the I-V curves in the outliers display
a spectacular level of noise, absent in the remainder of the pristine prior pads, that can only
be indicative of poor contact. Pristine SP0300 20, by comparison, varies evenly over 8 orders
of magnitude, and though samples displaying higher resistance are indeed noisier, there is
no sudden transition indicative of error as there is in pristine prior.
To determine the relative conductivity of the three sample types, it is necessary to
22
IL5
,--3
..--.-..
.".....
:.._..._ ___..
ci
. ............................
............
a,
Cr
cI
3
ci)"
cii
~
----- L_~~_;-:`-
.-.
.....
.. .~.~
.......
....
.
. ........
.. ..........
.
-
...
-i
-- - - -- --
ic-c
1
Voltage (V)
-1
-0.5
0
0.5
1
Voltage (V)
Figure 3-2: Voltage and resistance logarithmic plots for carbene prior.
control for the density and distribution of nanotubes on the surface. From the variation
in resistance within samples of all types, it is clear that the distribution of nanotubes on
the surface of these samples is spectacularly uneven; pristine prior and carbene prior are
the only potential exceptions. It was thought that perhaps the count of nanotubes on a
given pad could be obtained by direct analysis with atomic force microscopy; unfortunately,
this proved impractical, as evidenced by figure 3-3. The tubes are simply not countable.
In place of this method, statistical analysis techniques were considered, employing the ttest and analysis of variations (ANOVA); if indeed the carbene-functionalized samples have
similar electronic properties to pristine samples, as theorized, then the data sets should show
statistical similarity. However, there is a stumbling block to this analysis: it cannot be
guaranteed that the density of nanotubes is comparable between different, samples prepared
by the same method. This can be shown by consideration of the differences between the
carbenes BL303, BL319, BL320, and BL337, which were identically prepared.
The lack of any meaningful data from BL337 may be indicative of a complete failure of
the deposition process, which would prove only that the deposition process can fail, and is not
sufficient to show that the deposition process is necessarily particularly uneven. BL319, with
two functional pads, might be similarly dismissed. BL320, however, shows several functional
Figure 3-3: AFM image of interdigitated electrodes.
test pads, though not nearly so many as BL303, and with a higher average resistance; BL303
displays significant clustering around 105 Q, while BL320 is centered between 106 and 108 Q.
However, this observational analysis is inexact at best; a more rigorous statistical analysis is
required.
3.1.2
Statistical Analysis of Conductivity Measurements
Firstly, for ease of comparison, it is valuable to reduce each test pad to a single resistance
value. This was accomplished by performing linear regression on the I-V curve; to reduce
the effect of nonlinearities in the curve, this calculation was performed for low-bias portions
of the curve, 0.1 to 0.3 V and -0.3 to -0.1 V, and the resistance was taken to be the average
of these two calculations. For comparison, a secondary set of resistance values was taken
by performing linear regression across the entire I-V curve, excluding the zero-voltage point.
Since most of the samples displayed significant nonlinearity at larger voltages, the low-bias
calculated resistances were used for purposes of statistics. The tabulated resistances can be
seen in figure 3-4.
From these tabulated resistances a few things are visually obvious. First and foremost is
24
Resistance (Q)
102
104
106
108
·
·
1012
1010
Mi
0rZQE·
Carbene BL303
(moo =M1P?~S
:/0
Carbene BL319
Carbene BL320
Carbene Prior
I
-Vm
Nitrobenzene Ethanol
0
Nitrobenzene IPA
Nitrobenzene tw2-172-3
1014
0
0
OoG
S
0
CD aamro~~ca
TV cd
0
0
Pristine Prior
0X&XZ
Pristine SP0300 20
MI
0
I
or-01(M cc
I
V
00
I
I
Figure 3-4: Tabulated low-bias (red, top) and full-curve (blue, bottom) resistance
values.
that the carbene data are indeed widely spread. While it is tempting to say that the carbene
prior sample approaches the conductivity of the pristine prior sample, the distribution
for carbene BL303 is not dissimilar to nitrobenzene tw2-172-3-nor, for that matter, is it
dissimilar to pristine SP0300 20. At a first glance, then, the data seem inconclusive.
The statistical first step is to perform simple single-factor ANOVA on the set of carbene
samples. One-way ANOVA is a test which evaluates a number of groups, which are assumed
to be roughly normally distributed with equal variance, and determines whether the null
hypothesis that all groups were samples from the same population can be refuted. The
ultimate test statistic is the p-value; for p-values less than 0.05, the null hypothesis can be
refuted at the 95% significance level. That is, a low p-value indicates significant variation
between the tested groups.
ANOVA over the set of resistance values for BL303, BL319, and BL320 has a p-value of
0.00432. Excluding the smallest population, BL319, ANOVA returns a p-value of 0.001187.
While ANOVA is known to be of limited use on such small populations, it is still important
that BL303, BL319, and BL320 show such significant differences. As a secondary consideration, ANOVA of the resistance values for BL303, BL319, and carbene prior returns a p-value
of 9.67.10-6, suggesting that the carbene prior sample differs significantly from the BLn
samples. This is, of course, intuitively clear from the tabulated resistance chart. It should
be noted that ANOVA in this fashion assumes a normal distribution of resistance values,
which may not be the case; assuming that resistance goes roughly as the inverse number
of nanotubes bridging a pair of electrodes, we can instead assume a normal distribution of
nanotubes on the surface, which is more realistic. In this case, single-factor ANOVA must
be performed over the sets of reciprocal resistance values. In this case, BL303-BL319-BL320
returns a p-value of 0.2720, and BL303-BL319-prior returns a p-value of 3.62 .10-16. Under
this assumption, it seems that there is less certainty that the BLn samples differ statistically,
but BL320 has an extremely small sample size, and omitting it, the p-value falls to 0.1136.
This is significant only under the assumption that the number of working pad pairs has no
26
implications for the distribution on the surface, which is hardly a safe assumption.
It is the case that these samples have been prepared from different nanotube batches,
which may differ significantly from each other. It is difficult to account for this potential
effect, but it should be noted that BL303 and BL319 do not display great similarity despite
being made by the same procedure and from the same batch; it is therefore very unlikely
that different batch properties are solely responsible for the observed differences.
Without clear statistical agreement within a sample type, there is no basis for comparison
with other sample types, as there is no way to take into account the unknown density
and distribution of nanotubes. Particularly, with significant differences between samples
which were identically prepared, it is clear that there is an uncontrollable aspect of sample
preparation that a significant impact on sample behavior for which this analysis technique
cannot compensate. It must be concluded that this combination of measurement and statistical methods is not sufficient to determine the relative conductivities of the three studied
nanotube types; this measurement technique is too sensitive to uncontrollable irregularities
in sample preparation. There are several potential sources of error in this data: first, poor
metal-nanotube contact, though this is expected to be similar across all samples prepared by
this method; second, contamination of the nanotube itself, at the nanotube-metal interface,
or of the probe tips due to other machine users; third, and most likely, inhomogeneity in
the dispersion of nanotubes on the surface. Other methods must be employed to verify the
conductivity effect of the carbene functionalization, or a way to analyze or control for the
distribution and density of nanotubes on the sample surface must be devised.
However, without forgetting the statistical caveat, there are some conclusions that may
be drawn. Consider the very close spacing of the highest-conductivity results in, for example,
pristine prior and carbene prior. With this very close spacing of points, it can be inferred
that there is a dense distribution of tubes, such that the impact of very small changes
in density is minor; it follows that differences in the spacing of these highest-conductivity
clusters are likely to be significant. Therefore, the difference in resistance of approximately
27
one order of magnitude between carbene prior and pristine prior may in fact be a good
approximation of the true difference in conductivities; the same might be said, with more
reservation, about carbene BL303 and nitrobenzene tw2-172-2. Without statistical backing,
though, it is necessary to take these observations with a grain of salt.
Nitrobenzene requires an additional note. tw2-172-2, despite being the only nitrobenzene
to show a significant number of functional pads, has an incredibly wide distribution of
resistance values; it additionally gave generally noisy results through testing. This breadth of
distribution indicates that either the dispersion of nanotubes across the surface is extremely
uneven, or that the functionalization is asymmetric in some way, though this is a more remote
and more complex possibility.
3.2
Multi-Pad Temperature-Variation Measurements
Figures 3-5, 3-6, and 3-7 show the temperature-dependent resistance for carbene BL303 (alal0), nitrobenzene tw2-172-3 (al, a2, a5-10, bl-b5, b7-b9, el-e9), and pristine SP0300 20
(a2-a7, b6, b8, c2, c4, c8, d6, el, e2). Pads for each sample were selected arbitrarily from
those pads which had previously exhibited normal behavior and relatively low resistance,
which in turn typically indicated relatively low levels of noise in the I-V curve. It is expected
that increasing temperature will increase resistance approximately linearly.
The data are wholly inconclusive: carbene BL303 shows increasing and decreasing resistances with temperature in roughly equal measure, and nitrobenzene shows no behavior
of any predictable sort; pristine SP0300 20 at best indicates there may be some tendency
towards increasing resistance with temperature. This method of sweeping across pads and
then adjusting temperature, then, seems unreliable; it is possible that sweeping repeatedly
between pads introduces some amount of contamination. It is also possible that samples
do not equilibrate immediately at the target temperature, and therefore the pads measured
long after reaching the measurement temperature will display different behavior than those
.. ^9
0
20
40
60
80
Temperature (C)
100
120
Figure 3-5: Temperature-dependent resistance data for carbene BL303.
^13
1012
1011
1010
10i
107
106
5
1n
,v
0
20
40
60
Temperature (C)
80
100
120
Figure 3-6: Temperature-dependent resistance data for nitrobenzene tw2-172-3.
1012
1010
o
10 8
106
4
10n
20
40
60
80
Temperature (C)
100
120
Figure 3-7: Temperature-dependent resistance data for pristine SP0300 20.
measured briefly after. To eliminate this possibility, two additional tests were performed:
firstly, a test to examine the possibility of a significant equilibration effect, and second, an
alternative form of temperature-variation test such that the probes would remain on one pad
for the duration of the cycle, eliminating whatever noise-introducing effect might stem from
repeatedly switching pads.
3.3
Point-Dwell Equilibration Measurements
Figure 3-8 shows a series of measurements taken rapidly on carbene BL303 e8 at ±0.3 V for
up to a few minutes starting immediately when the chuck indicated the target temperature.
The instantaneous resistance for each point was calculated, and then the pair of values was
averaged to give the indicated resistance. Ideally, full measurements would be taken to yield
more accurate resistance values, but the need for rapid sampling prevents this.
It is not expected that equilibration behavior will vary between sample types; the question
is whether the sample reaches thermal equilibrium at the same time as the chuck, and
Sx
104
_
5.2
!0
5
00
4.8
*0
Ci
(D4.6
S4.4
4.2
4
-
3.8
QA
40
50
I
I
I
I
60
70
80
90
I
100
110
120
130
140
I
I
I
150
160
170
180
Temperature (close-packed dots indicate repeated sampling)
Figure 3-8: Resistance values for rapid sampling of a single pad with time.
therefore, measuring one sample of known behavior is sufficient. If the sample does not
equilibrate immediately, this will be reflected in a clear change in resistance with repeated
rapid sampling.
From the wide scattering of points at each temperature, it is clear that equilibration
is either effectively immediate, or that the effect is so minor as to not be visible through
the noise in the measurements.
There are apparently significant and permanent leaps
visible at 70 TC and 100 TC, but these seem to be unique, and do not correlate with
any known phenomenon; indeed, a similar leap downwards can be seen at 90 'C, and it
is assumed that these are simply measurement artefacts of some sort. For purposes of these
experiments, then, it is assumed that there is no need to wait any significant period of time
for equilibration when changing temperatures. This additionally suggests that the rate of
change in temperature will have little if any effect on these measurements; the sample is
31
always effectively at the chuck temperature at the limit of the heating rate of the chuck.
These data lend validity to the rapid-equilibration assumption employed in taking singlepad temperature cycle measurements, below. Measurements were generally taken very soon
after equilibration, though occasionally with some delay; it is assumed that variation in the
sampling delay time will have no measurable impact on the results.
3.4
Single-Pad Temperature Cycle Measurements
For these measurements, bear in mind that a set of pads from a given wafer was selected
for measurement, and while only one pad was probed per temperature cycle, the entire
wafer was subjected to that temperature cycle; it was observed in section 3.2 that measuring
multiple pads as part of one temperature cycle was ineffective. At the completion of each
temperature cycle, all pads were again measured. Figure 3-9 shows an example of the data
taken by this method, and is representative temperature cycle graph for carbene prior, pad
e7. The curves on the left half of each graph show the resistance of the pad as a function
of temperature during the temperature cycle in which that pad was measured, solid for the
ramp up, and dashed for the ramp down; the curves on the right show the change in the
50 TC baseline resistance as a function of the number of temperature cycles experienced by
the sample. In the normalized graphs, the temperature ramp resistance is normalized to its
initial 50 TC value, and the baseline resistance is normalized to the first baseline resistance.
For example, a pad ramped at cycle two has its baseline resistance normalized to baseline 1,
and its temperature ramp normalized to its initial value which should be extremely close to
baseline 2.
Full data for carbene samples may be found in Appendix A; figures B-1-B-4 show the
temperature ramps for carbene prior and B-5 and B-6 show the ramps for carbene BL303.
For carbene prior, at least, there is a clear overall effect of cycling: resistance increases
with increasing temperature, sometimes with outlying points, and upon cooling returns to
S10n4
Ramped at Cycle 4
Ramp Up
--- Ramp Down
Baseline Resistance
1. 45
--- ---
1.6
Ramped at Cycle 4
Ramp Up
--- Ramp Down
Baseline Resistance
1.4
1.35
1.5
1.1.3
a:1 25
1•.4
1.1.2
S1.3
15
a: l1.1.1
1.2
1.05
''
'
50
100
150
Temperature (C)
200
1
2
3
Cycle Number
4
1
50
100
150
Temperature (C)
200
1
2
3
4
Cycle Number
Figure 3-9: Temperature cycle for carbene prior pad e7, shown as absolute resistance
and as normalized to 50 'C and baseline one.
a value significantly higher than its pre-cycling resistance, though the intermediate points
exhibit some noise. At this temperature resolution, and with this degree of noise, it is not
clear whether the increase with temperature is linear or not. Examination of the normalized
resistances does not show a pattern in the fraction by which resistance increased between 50
and 200 'C. Several pads show an increase of approximately 10%, but there are insufficient
data to draw a conclusion from this. A second effect is also clear in the carbene prior
data: repeated cycles permanently increase the room-temperature baseline resistance. It is
not believed that this baseline effect is the sole reason for the increase in resistance with
temperature, but rather that it is observed in conjunction with the ideal linear increase in
resistance with temperature. The ideal graph, then, is a pair of lines, both monotonically
increasing with temperature, with the maximum observed at 200 OC and the ramp down
decreasing from this to a value above the initial resistance; this ideal behavior is not observed,
and it is difficult to separate the direct effect of temperature from the baseline effect in
interpreting the temperature ramp curves, which in many cases show continually increasing
resistance throughout the cycle. This indicates dominance of the baseline effect over the
predicted linear-in-temperature effect, at least in this sample.
Carbene BL303 displays significantly noisier curves; though the endpoints in the case
of carbene BL303 d3 fit the pattern, BL303 d5 does not. It is believed that this is simply
noise or a severely damaged pad; the response of carbene BL303 pads was too poor to
gather a significant number of additional data points, and so carbene prior is taken as the
standard. It is notable that despite indeterminate temperature variations, carbene BL303
displays a significant increase in baseline resistance after being cycled, similar to carbene
prior; this independent effect is in fact common to all sample types. There are several
possible explanations for this, as there are three contact points which could be affected.
Firstly, there is probe-to-surface contact; it is possible that surface damage could result in
an increasingly poor connection that increases resistance, but it is expected that such an effect
would also cause significant similar error in other tests, which was not observed. Secondly,
there is electrode-to-nanotube contact, but the opposite of the observed result is expect for
this contact: in an ideal vacuum environment, the contact resistance between the nanotube
and the metal electrode is expected to decrease at elevated temperature due to an annealing
effect. This effect is permanent, with no predicted change upon repeated cycling. Therefore,
barring some unknown effect, the increase in baseline resistance is almost certainly not at
the electrode-nanotube contact. The third possibility is within the nanotubes themselves,
bridging the electrodes. A maximum temperature of 200 'C is not expected to have any
effect on the functionalization itself, but it is possible that the nanotubes, which are quite
sensitive to their environment, are somehow permanently but gradually altered by exposure
to the atmosphere at an elevated temperature. There is the possibility of introducing general
contamination via the probe tips. Finally, there is the possibility that there is in fact a
small or long-timescale equilibration time which is unaccounted for. When heating at the
maximum rate over a 10 degree interval, perhaps the lag of the sample temperature behind
the chuck temperature is minimal, but becomes significant when heating over a 50 degree
interval. This is pure conjecture, however, and the assumption that the sample equilibrates
near-instantaneously is believed to be a valid one.
Any of these could cause the observed changes in baseline resistance; fortunately, despite
being a significant effect in its own right, it does not seem to impact the form of the
4
x10
2.4Z 2.2
2
oc 1.8
1.6
1.4
19
50
100
150
200
Temperature (C)
Figure 3-10: Temperature ramp for carbene prior pad c4 in 10-degree intervals.
temperature ramp resistance curve, only its initial point. It is therefore not a concern in
drawing conclusions about the general behavior of these samples with direct temperature
variation. Some combination of these effects, however, is clearly responsible for the tendency
for resistance returning to a higher value after undergoing one full cycle. It is an interesting
question, and perhaps a good choice for future study, if attempts to control contamination can
lead to the curve becoming symmetric for the ramp up and ramp down. A higher-resolution
temperature ramp for carbene prior pad c4, shown in figure 3-10, displays approximately
the ideal behavior, and lends evidence to the expectation that there is a reversible linear-intemperature effect as well as an overall baseline increase.
Figures B-7-B-12 show the temperature ramps for pristine prior; figures 3-11 and 3-12 are
representative samples. The behavior is quite similar to carbene prior, though a significant
number of pads display a decreasing resistance with temperature before returning to a higherthan-initial value. This is difficult to explain: it was at first thought that these samples were
somehow less contaminated or susceptible to contamination, and this was an observation of
the annealing effect mentioned previously, but that effect is permanent after a single cycling,
and should not be observed on samples undergoing a 5th or 6th temperature cycle. The
cycling effect on the baseline is very similar to what was observed in the carbenes; it can
35
Ramped at Cycle 2
Ramp Up
--- Ramp Down
Baseline Resistance
Ramp Up
--- Ramp Down
Baseline Resistance
Ramped at Cycle 2
65UU
I.
Fr
'
'
'
7
1. 7
6000
1. 6
1.5
a 5500
4
5000
t1 . 3
"'
m 4500
T_
4000
---
I
0
50
100
150
Temperature (C)
200 1
2 3 4 5
Cycle Number
6
Al
j~7
_______________
50
100
150
Temperature (C)
~
I
I
__________
200 1 2 3 4 5
Cycle Number
Figure 3-11: Temperature cycle of pristine prior, pad b2, and its baseline resistance.
Ramped at Cycle 4
-
Ramp Up
---
Ramp Down
-Baseline
Resistance
Ramped at Cycle 4
Temperature (C)
---
Ramp Up
Ramp Down
Baseline Resistance
Cycle Number
Figure 3-12: Temperature cycle of pristine prior, pad b4 and its baseline resistance.
Ramped at Cycle 3
,x10
.2·
0. I
Ramp Up
- - - Ramp Down
Baseline Resistance
-
Ramp Up
--- Ramp Down
Baseline Resistance
Ramped at Cycle 3
2.5
1.1
2.4
.2
62.3e
E 2.2
4)
0.9
e 2.1
2
4
0.8
tR
50
100
150
Temperature (C)
200
1
2
3
Cycle Number
50
100
150
Temperature (C)
200 1
2
3
Cycle Number
Figure 3-13: Temperature cycle of nitrobenzene tw2-172-3, pad c9, and its baseline
resistance.
be concluded that the pristine sample exhibits the same type of behavior overall, with an
unknown resistance-decreasing effect dominating on some pads. It can also be observed that,
as in the case of pristine b4 (see figure 3-12), the resistance does indeed decrease during the
ramp down, while still increasing monotonically during the ramp up and ending the ramp
down at a higher value than pre-ramping. This is a case in which the baseline-increasing
effect is not overwhelming and the effects of the reversible linear temperature dependence
can still be observed; it may be considered the ideally demonstrative result which displays
both the baseline-increasing effect and the predicted linear dependence.
Nitrobenzene, in contrast to the other two sample types, displays no predictable behavior
whatsoever, and it is difficult to choose a representative example. Nitrobenzene tw2-172-3
pad c9 is shown in figure 3-13; full data is available in Appendix B, figures B-13-B-18.
The baseline variation is similar to all other samples, but the temperature cycles show
enormous, random spikes and unpredictable curves. A temperature ramp was also taken
for nitrobenzene tw2-172-3 pad e3 in 10 degree intervals to observe the behavior at improved
temperature resolution, for the ramp up only, plus one post-ramp sampling; figure 3-14
confirms that the behavior for this nitrobenzene sample is essentially noisy, with some
tendency to decrease with increasing temperatures, and particularly at high temperatures.
Despite the unpredictable behavior of the nitrobenzene samples, there is a clear conclusion
37
,,x 105
6
o0
-Ramp
0o Post-Ramp
5.5
5
4.5
24
3.5
3
9A
50
Figure 3-14:
intervals.
100
150
Temperature (C)
200
Temperature ramp for nitrobenzene tw2-172-3 pad e3 in 10-degree
to be drawn. Carbene and pristine samples are relatively similar in temperature behavior in
light of the nitrobenzene data, a point which suggests that carbene, despite being a covalent
functionalization like nitrobenzene, behaves in at least one aspect more like a pristine than a
nitrobenzene tube. Combined with the observation in section 3.1, this suggests that carbenefunctionalized nanotubes may indeed have similar electronic properties to pristine nanotubes,
or at the very least that further investigation is warranted, though this combination of sample
preparation and measurement methods is insufficient to quantify this behavior.
Chapter 4
Conclusion
Silicon wafers coated with distributions of nanotubes both pristine and covalently functionalized with nitrobenzene and carbene chemistries were analyzed for electronic conductivity
through direct measurement of resistance and the variation in resistance with temperature.
Measurements of room-temperature conductivity showed distributions of resistance values
over several orders of magnitude within single samples, and statistically significant differences
in resistance distributions between samples of the same type prepared by similar methods;
this combination of sample preparation and measurement methods is extremely sensitive to
nanotube density and distribution, which are difficult to control or measure, and therefore,
it is not sufficient to determine absolutely the differences in conductivity between the three
sample types. However, observation of closely-spaced resistance values indicates that some
carbene samples display conductivity very similar to pristine samples, within one order
of magnitude, while similarly closely-spaced nitrobenzene samples are several orders of
magnitude higher. It is quite possible that carbene-functionalized nanotubes can be shown
to display electronic properties quite similar to those of pristine nanotubes, in agreement
with theory, if a measurement method is devised which compensates for the distribution
problem.
Cycling the samples through a temperature ramp revealed a pronounced permanent
39
change to the room-temperature resistance of each sample; after each cycle, baseline resistance tended to increase considerably and permanently.
This is believed to occur in
conjunction with a reversible, linear change in resistance with temperature, but in many
cases the change in room-temperature resistance dominated this effect.
In general, and
despite an unexpected trend of decreasing resistance with increasing temperature shown
by many carbene prior pads, carbene and pristine samples showed similar behavior with
temperature variation. In turn, carbene and pristine samples show nothing in common with
the noise-dominated, unpredictable behavior of the nitrobenzene samples aside from the
universal increase in baseline resistance with temperature. Though carbene and nitrobenzene
are both covalent functionalizations, carbene-functionalized samples display many properties
common to pristine samples, in agreement with the theory [12].
There are many potential sources of the observed broad distribution of resistances, and
of the observed baseline-increasing temperature effect. Firstly, there is the possibility that
metal-nanotube contact is poor. It is expected that contact will be similar for all samples
prepared by this method, so this effect is unlikely to affect room-temperature data, but
high-temperature behavior (in the absence of the known annealing effect) is uncertain. More
likely, there is the possibility of contamination of the nanotubes due to exposure to the
atmosphere at both room and high temperatures. There may also be contamination of the
metal-nanotube contact and of the probe tips, but the likelihood of these factors having a
significant effect is uncertain. It is possible that there are contaminants in the nanotube
solution used for deposition, but this is extremely unlikely to affect these conductivity
measurements. Finally, there is the question of homogeneity in the dispersion. Dispersion
in the solvent used for deposition may be a problem in the case of pristine tubes, which
are known to be difficult to disperse, as well as some carbene tubes. Casting and surface
conditions may also come into play in controlling the distribution of nanotubes on the surface.
It is believed that this experiment can be repeated, with additional controls placed
over some of these factors, to yield more conclusive data. Samples can be prepared in a
more uniform manner, employing only a single batch of nanotubes to eliminate any possible
differences in the initial material. With significant additional data, it may be possible to
evaluate the influence of inhomogeneous dispersion by statistical methods. The significant
challenge is contamination; characterizing the effects of atmosphere at room and elevated
temperatures may be instrumental in evaluating the temperature behavior of these samples.
Appendix A
Aggregate Voltage and Resistance
C urves
____
___r____
__._ _________________
________
__C__________C
__~_~~___
C_·_____·_____
-
...
..
....
.-.-
·~__~__~__
_____
_.____.___
----------.....
--~------~F~-~i~,~:
-- - ------------·--. .._~
,---··---··-··----·;------------·::~--·-::---------=---4.~-.~r~:::.....·-.
---- ~~-----
I_· __
·I·-.__~~_~~~__~~~~
....
~_·····
0
Voltage (V)
Voltage (V)
Figure A-1: Voltage and resistance logarithmic plots for carbene prior.
S^10
-^-4
10 [
10
-5
9
10
10
10-6
-----,
---.-.
,--`-----~~~~~ -- -~~``-v------~_~--
c..c·-·---------
108
---- 1`2~-,
-----
··-----~,~..~
_.,--....i
-7
10
~r~---
a 10
-8
10
---
-----------
c-------
.------ ~
--~L
106
-
S10
i
- o
105
10
~ft~s;2~igi
_......
-····---- ···-- ·--- ~~~
----------··--------···----
-1
-0.5
0
Voltage (V)
0.5
1
-1
0
Voltage (V)
Figure A-2: Voltage and resistance logarithmic plots for carbene BL303.
.,12
1U
-4
I I
I
-6
10
I
--·-·--·
· ·-· --.
_
;-_,_____ -2.`"---,
/~~·
\·,
~~_~__~~___..__~_*----'`~,~~__~~_~
C -8
---
'A--
10
/I
_----~"--~"~L-~
·-
/---~
0
- __~_
--~
... .....
-.- ..
a)
- 10
~~--.-----------i~:I_
......
__-.~-------~c-~--
~.....
•-·--
·--
·
_
\
1
10 - 2
14
I
1 -
-1
-0.5
Voltage (V)
1
I
I
-
Voltage (V)
Figure A-3: Voltage and resistance logarithmic plots for carbene BL320.
.^-8
oC
a"
0,
-1
-0.5
0
Voltage (V)
1
-1
-0.5
0
Volage (V)
0.5
Figure A-4: Voltage and resistance logarithmic plots for nitrobenzene ethanol.
I
-4
IU
Np-
r:
-
10 6
1012
__.---,,-----~/~_~-,_/
\
j
v
i
/---
810
1010
10
-·---.,r-·---
`'sx,
.----'
-~~P~CY~~Z
r"~~-/
:o
``
)10
i0-10
~2-,
r~i~---~
'-~---=-CL-L-~.I-~Bs~~L·~
-
10 12
106
\.,,
",•_
\
/ \/j
- 14
10
-1
0
Voltage (V)
1
-1
0
Voltage (V)
Figure A-5: Voltage and resistance logarithmic plots for nitrobenzene tw2-172-3.
,•14
1U
1012
/7~
N,
f-7
1010
--
0
:3
c
)I
108
.~1(
a,
106
SI
-1
-U.b
1
Voltage (V)
I
-1
Voltage (V)
Figure A-6: Voltage and resistance logarithmic plots for pristine SP0300 20.
I
z
a
iS
a
1?
S
a-
It
0
=
!
I
IV
-1
-0.5
0
Voltage (V)
0.5
1
-1
-0.5
0
Voltage (V)
0.5
Figure A-7: Voltage and resistance logarithmic plots for pristine prior.
48
Appendix B
Single-Pad Heat Ramp Data
-
Ramp Up
-
a pda
10
Baseline Resistance
--
Ramped at Cycle 1
xln4
yl
Ramp Up
---Ramp Down
Baseline Resistance
--- Ramp Down
Ramped at Cycle 1
1i
----~-_~c=-ii
I ---------
54
4+
I
I
[
*1
_
+
_
_
_
100
150
Temperature (C)
50
_
200
1
_
2
3
4
Cycle Number
50
100
150
Temperature (C)
200 1
2
3
4
Cycle Number
Figure B-1: Resistance for temperature cycle of carbene prior, pad c4, and its baseline
resistance, shown as resistance values and normalized to baseline one.
Ramped at Cycle 2
.x 10
Ramp Up
--- Ramp Down
-Baseline
Resistance
-Ramp
Up
--- Ramp Down
Baseline Resistance
Ramped at Cycle 2
[
3. 8
I
i
3.6
S3. 4
a( 3
1
2. 8
T
Ii
i
-,
2
1.2
1. 1
2. 6
I
2. Y
50
100
150
Temperature (C)
200 1
I
2
3
4
Cycle Number
I
n
1l
I
_______
200 1
2
3
Cycle Number
100
150
Temperature (C)
Figure B-2: Temperature cycle of carbene prior, pad c5, and its baseline resistance.
-
I
x 10
".1"
4
Ramped at Cycle 3
Ramp Up
--- Ramp Down
Baseline Resistance
Ramped at Cycle 3
-
Ramp Up
---
Ramp Down
-
Baseline Resistance
1.8
1.75
1.7
1.65
S1.6
1.55
..
1.5
1.45
1.4
1.35
1
50
100
150
Temperature (C)
200 1
2
3
4
Cycle Number
n
i1
50
.;;;'
--- /
100
150
Temperature (C)
'
200 1
2
3
Cycle Number
Figure B-3: Temperature cycle of carbene prior, pad c6, and its baseline resistance.
Ramp Up
- - - Ramp Down
Baseline Resistance
Ramped at Cycle 4
xr 104
Ramped at Cycle 4
- Ramp Up
- -- Ramp Down
- Baseline Resistance
----------
1.6
[
1
1.5
o 1.4
• 1.3
50
1.2
-
10-
10-20-1----
-
100
150
Temperature (C)
50
200 1
2
3
4
Cycle Number
50
100
150
Temperature (C)
200 1
2
3
4
Cycle Number
Figure B-4: Temperature cycle of carbene prior, pad c7, and its baseline resistance.
4
-- Ramp Up
--- Ramp Down
Baseline Resistance
Ramped at Cycle 2
14
x5
9. 5rIl
-
I
Ramped at Cycle 2
i
I
I
p
I
50
T
50
Ramp Up
--- Ramp Down
-Baseline
Resistance
10
Ic
10
100
150
Temperature (C)
I
0
200
1
n
Cycle Number
2
OF
. 50
100
150
Temperature (C)
j
200
1
Cycle Number
2
Figure B-5: Temperature cycle of carbene BL303, pad d3, and its baseline resistance.
Ramped at Cycle 1
r
Ramp Up
--- Ramp Down
Baseline Resistance
I-
r
n
r
I
r
50
100
150
Temperature (C)
200
1
Cycle Number
Figure B-6: Temperature cycle of carbene BL303, pad d5, and its baseline resistance.
50
100
150
Temperature (C)
200 1 2 3 4 5
Cycle Number
6
Ou
IOU
Iu
I
uuV
I
Temperature (C)
4
!L
3
0
Cycle Number
Figure B-7: Temperature cycle of pristine prior, pad bl, and its baseline resistance.
-
Ramp Up
Ramp Up
--- Ramp Down
Baseline Resistance
- - - Ramp Down
-
Ramped at Cycle 2
Baseline Resistance
Ramped at Cycle 2
nnl
6500
I
1.
6000
I
I I ·
1
·
1
1
1 I
1
7
1.6
5500
a
5
cc
a) .4
c
5000
.3
co
4500
.2
1.1
4000
*
--
-
___________________________
50
100
150
Temperature (C)
200 1
2 3 4 5
Cycle Number
0
50
100
150
Temperature (C)
________________.
200 1 2 3 4 5
Cycle Number
Figure B-8: Temperature cycle of pristine prior, pad b2, and its baseline resistance.
-Ramp
---
Up
Ramp Down
50
100
150
Temperature (C)
200 1
2 3 4 5 6
Cycle Number
Figure B-9: Temperature cycle of pristine prior, pad b3, and its baseline resistance.
-Ramp
Up
--- Ramp Down
Baseline Resistance
Ramped at Cycle 4
·
6U00
I ---.------
5500
·
·
·
·
----
5000
S4500
4000
'
3500
""N
100
150
Temperature (C)
50
^
200 1 2 3 4 5 6
Cycle Number
Ou
luu
b1u
D
Temperature (C)
zu
1 2e
4 b
Cycle Number
U
Figure B-10: Temperature cycle of pristine prior, pad b4 and its baseline resistance.
4
x 10
Ramped at Cycle 5
-Ramp
Up
- - - Ramp Down
Baseline Resistance
4
3..53
5
S2. .5
a,
ir
C
+
II.
I
-I
I
I
+
j
_______
50
100
150
Temperature (C)
____.
200 1 2 3 4 5
Cycle Number
Temperature (C)
Cycle Number
Figure B-11: Temperature cycle of pristine prior, pad b5, and its baseline resistance.
Ramped at Cycle 6
3000
-... -
Ramp Up
--- Ramp Down
-Baseline
Resistance
1.7
.
p,
1.6
2800
1.5
Ramped at Cycle 6
Ramp Up
- -- Ramp Down
-Baseline
Resistance
K
I
A
Co
S 2600
c 1.4
CO
2400
cc
rr
a
-a
a
1.3
2200
a
2000
ool001
0
50
100
150
Temperature (C)
,
200 1 2 3 4 5 6
Cycle Number
I
----- ---------------'
-.
50
100
150
Temperature (C)
. . .
200 1 2 3 4 5
Cycle Number
Figure B-12: Temperature cycle of pristine prior, pad b6, and its baseline resistance.
-
x 1n5
Ramp Up
--- Ramp Down
Baseline Resistance
Ramped at Cycle 1
Ramp Up
-- -Ramp Down
Baseline Resistance
Ramped at Cycle 1
z
\r
( C
r ~ 5Il
r
·
50
100
150
Temperature (C)
200
1
·
2
Cycle Number
Figure B-13: Temperature cycle of nitrobenzene tw2-172-3, pad c3, and its baseline
resistance.
-Ramp
Up
--- Ramp Down
I
bU
1UU
1bU
Temperature (C)
ZUU
1
ou
Z
Cycle Number
IUU
bU
Temperature (C)
ZUU
1
Z
j
Cycle Number
Figure B-14: Temperature cycle of nitrobenzene tw2-172-3, pad c4, and its baseline
resistance.
Ramp Up
-- -Ramp Down
Baseline Resistance
Ramped at Cycle 2
L
ovv
700
r
r
600
0o
I
16500
a:
S400
D 300
r
r
r
r
,
r
r
r
r
r
I
ar:
200
r
100
0
Temperature (C)
Cycle Number
t
i mý =ýýM
ý= ý =ý
50
100
150
Temperature (C)
200
J•
1
2
Cycle Number
Figure B-15: Temperature cycle of nitrobenzene tw2-172-3, pad c8, and its baseline
resistance.
Ramp Up
-
- - - Ramp Down
Ramped at Cycle 3
w
__x 105
·
-
Baseline Resistance
·
·
Ramped at Cycle 3
Ramp Up
--- Ramp Down
Baseline Resistance
·
1.
2.4
0
. 2.3
a
S2.2
S2.1
I
0.
9
I
0.
8
I
2
0.
r
1.9
*
--I
'I
I
I
.
I
50
100
150
Temperature (C)
200
1
2
Cycle Number
50
100
150
Temperature (C)
200
1
2
3
Cycle Number
Figure B-16: Temperature cycle of nitrobenzene tw2-172-3, pad c9, and its baseline
resistance.
---
x
41
107
Cycle 2
Ramped at
.
.
-
_.
I
I
1
j
I
Ramp Up
Ramp Down
Baseline Resistance
.
._
I
2
I
I
s-i
V
i
· I ___________
______________
50
100
150
Temperature (C)
200
1
2
Cycle Number
Temperature (C)
Cycle Number
Figure B-17: Temperature cycle of nitrobenzene tw2-172-3, pad dl, and its baseline
resistance.
Sx 105
Ramped at Cycle 3
Ramp Up
- - - Ramp Down
Baseline Resistance
ii[7,
{y/
K~
1*.~
4
F~----_
50
I
_
_
_
100
150
Temperature (C)
_
200
I
I
_
1
_
2
Cycle Number
_
Ou
1Uu
lou
Temperature (C)
ZUU 1
2
Cycle Number
Figure B-18: Temperature cycle of nitrobenzene tw2-172-3, pad d5, and its baseline
resistance.
Bibliography
[1] Jeffrey Bahr and James Tour. Covalent chemistry of single-wall carbon nanotubes.
Journal of Materials Chemistry, 12:1952-1958, 2002.
[2] Sarbajit Banerjee, Tirandai Hemraj-Benny, and Stanislaus S Wong. Covalent surface
chemistry of single-walled carbon nanotubes. Advanced Materials, 17:17-29, 2005.
[3] Jean-Marc Bonard, Theirry Stora, Jean-Paul Salvetat, Frederic Maier, Th6mas Stbckli,
Claus Duschl, Laszl6 Forr6, Walt A de Heer, and Andre Chitelain. Purification and
size-selection of carbon nanotubes. Advanced Materials,9(10):827-831, 1997.
[4] Jian Chen, Mark A Hamon, Hui Hu, Yongsheng Chen, Apparao M. Rao, Peter C.
Eklund, and Robert C. Haddon. Solution properties of single-walled carbon nanotubes.
Science, 282(5386):95-98, 1998.
[5] Robert Chen, Sarunya Bangsaruntip, Katerina Drouvalakis, Nadine Wong Shi Kam,
Moonsub Shim, Yiming Li, Woong Kim, Paul Utz, and Hongjie Dai. Noncovalent
functionalization of carbon nanotubes for highly specific electronic biosensors. Proc.
Nat. Acad. Sci. USA, 100:4894, 2003.
[6] Y. Chen, R. C. Haddon, S. Fang, A. M. Rao, P. C. Eklund, W. H. Lee, E. C. Dickey,
E. A. Grulke, J. C. Pendergrass, A. Chavan, B. E. Haley, and R. E. Smalley. Chemical
attachment of organic functional groups to single-walled carbon nanotube material.
Journal of Materials Research, 13(9):2423-2431, 1998.
[7] Hongjie Dai. Carbon nanotubes: Synthesis, integration, and properties. Accounts of
Chemical Research, 35:1035-1044, 2002.
[8] Hongjie Dai,
Jason H. Hafner,
Andrew G. Rinzler,
Daniel T. Colbert, and
Richard E. Smalley. Nanotubes as nanoprobes in scanning probe microscopy. Nature,
384(6605):147-150, 1996.
[9] Cees Dekker. Carbon nanotubes as molecular quantum wires. Physics Today, 52:22-28,
1999.
[10] Christopher A. Dyke and James M. Tour. Unbundled and highly functionalized carbon
nanotubes from aqueous reactions. Nano Letters, 3(9):1215-1218, 2003.
[11] K. Kamaras, M. E. Itkis, H. Hu, B. Zhao, and R. C. Haddon. Covalent bond formation
to a carbon nanotube metal. Science, 301(5639):1501, 2003.
[12] Young-Su Lee and Nicola Marzari.
Cycloaddition functionalizations to preserve or
control the conductance of carbon nanotubes.
Physical Review Letters, 97:116-801,
2006.
[13] Jie Liu, Andrew G. Rinzler, Hongjie Dai, Jason H. Hafner, R. Kelley Bradley, Peter J.
Boul, Adrian Lu, Terry Iverson, Konstantin Shelimon, Chad B. Huffman, Fernando
Rodriguez-Macias, Young-Seok Shon, T. Randall Lee, Daniel T. Colbert, and Richard E.
Smalley. Fullerene pipes. Science, 280(5367):1253-1256, 1998.
[14] R. Saito, G. Dresselhaus, and M. S. Dresselhaus.
Physical Properties of Carbon
Nanotubes. Imperial College Press: London, 1998.
[15] Sander J. Tans, Michel H. Devoret, Hongjie Dai, Andreas Thess, Richard E. Smalley,
L. J. Geerligs, and Cees Dekker. Individual single-wall carbon naotubes as quantum
wires. Nature, 386:474-477, 1997.
[16] Andreas Thess, Roland Lee, Pavel Nikolaev, Hongjie Dai, Pierre Petit, Jerome Robert,
Chunhui Xi, Young Hee Lee, Seong Gon Kim, Andrew G Rinzler, Daniel T Colbert,
Gustavo E Scuseria, David Tominek, John E. Fischer, and Richard E. Smalley.
Crystalline ropes of metallic carbon nanotubes. Science, 273:483-587, 1996.
[17] C. T. White and T. N. Todorov. Carbon nanotubes as long ballistic conductors. Nature,
393:240-242, 1998.
[18] Zhen Yao, Charles L. Kane, and Cees Dekker. High-field electrical transport in singlewall carbon nanotubes. Physical Review Letters, 84(13):2941-2944, 2000.