Unraveling the 13C NMR chemical shifts in single-walled carbon nanotubes: Dependence on diameter and electronic structure

Article abstract of DOI:10.1021/ja211181q

The atomic specificity afforded by nuclear magnetic resonance (NMR) spectroscopy could enable detailed mechanistic information about single-walled carbon nanotube (SWCNT) functionalization as well as the noncovalent molecular interactions that dictate ground-state charge transfer and separation by electronic structure and diameter. However, to date, the polydispersity present in as-synthesized SWCNT populations has obscured the dependence of the SWCNT ¹³C chemical shift on intrinsic parameters such as diameter and electronic structure, meaning that no information is gleaned for specific SWCNTs with unique chiral indices. In this article, we utilize a combination of ¹³C labeling and density gradient ultracentrifugation (DGU) to produce an array of ¹³C-labeled SWCNT populations with varying diameter, electronic structure, and chiral angle. We find that the SWCNT isotropic ¹³C chemical shift decreases systematically with increasing diameter for semiconducting SWCNTs, in agreement with recent theoretical predictions that have heretofore gone unaddressed. Furthermore, we find that the ¹³C chemical shifts for small diameter metallic and semiconducting SWCNTs differ significantly, and that the full-width of the isotropic peak for metallic SWCNTs is much larger than that of semiconducting nanotubes, irrespective of diameter.

Full text of DOI:10.1021/ja211181q

Article
pubs.acs.org/JACS
13
Unraveling the C NMR Chemical Shifts in Single-Walled Carbon
Nanotubes: Dependence on Diameter and Electronic Structure
Chaiwat Engtrakul,*,# Veronica M. Irurzun,‡ Erica L. Gjersing,# Josh M. Holt,# Brian A. Larsen,#
Daniel E. Resasco,‡ and Jeffrey L. Blackburn*,#
#National Renewable Energy Laboratory, 1617 Cole Boulevard, Golden, Colorado 80401, United States
‡School of Chemical, Biological, and Materials Engineering, University of Oklahoma, Sarkeys Energy Center, Norman, Oklahoma
73019, United States
S
* Supporting Information
ABSTRACT: The atomic specificity afforded by nuclear
magnetic resonance (NMR) spectroscopy could enable
detailed mechanistic information about single-walled carbon
nanotube (SWCNT) functionalization as well as the non-
covalent molecular interactions that dictate ground-state
charge transfer and separation by electronic structure and
diameter. However, to date, the polydispersity present in as-
synthesized SWCNT populations has obscured the depend-
ence of the SWCNT 13C chemical shift on intrinsic parameters
such as diameter and electronic structure, meaning that no
information is gleaned for specific SWCNTs with unique chiral indices. In this article, we utilize a combination of 13C labeling
and density gradient ultracentrifugation (DGU) to produce an array of 13C-labeled SWCNT populations with varying diameter,
electronic structure, and chiral angle. We find that the SWCNT isotropic 13C chemical shift decreases systematically with
increasing diameter for semiconducting SWCNTs, in agreement with recent theoretical predictions that have heretofore gone
unaddressed. Furthermore, we find that the 13C chemical shifts for small diameter metallic and semiconducting SWCNTs differ
significantly, and that the full-width of the isotropic peak for metallic SWCNTs is much larger than that of semiconducting
nanotubes, irrespective of diameter.
varies as a function of diameter (d) and chiral angle, properties
captured in the unique chiral indices (n,m) of each SWCNT.
Nanotube properties follow patterns based on the nanotube
family (λ), determined by the relationship
1. INTRODUCTION
For the π aromatic allotropes of carbon, e.g., fullerenes, single-
walled carbon nanotubes (SWCNTs), and graphene/graphite,
the 13C nuclear magnetic resonance (NMR) spectrum is a
unique chemical signature that can relay critical information on
bond hybridization, density of states, curvature, and coupling to
nuclear spins of adjacent molecules. 13C NMR has been a
critical tool for identifying unique fullerenes (e.g., Cx, where x =
56, 60, 70, 78), confirming the symmetry of particular fullerene
isomers,1 tracking functionalization reactions of the fullerene
sidewall,2,3 and analyzing endohedral fullerenes.4 13C NMR has
also been used to characterize both covalent SWCNT sidewall
functionalization5−7 and noncovalent interactions such as the
interaction with encapsulated water.8 However, the same level
of chemical specificity achieved for fullerenes has not been
attained for SWCNTs because of the polydispersity present in
as-produced SWCNT samples.
Recent advances in the separations of SWCNTs by electronic
structure, diameter, and even chirality (left- versus right-
handed) have been achieved by the clever utilization of
surfactants to affect SWCNT properties such as density9,10 or
mobility through a fractionating column.11−13 Such separations
have initiated a new era of SWCNT research, in which highly
enriched samples can shed light on the molecular-level
properties of unique SWCNTs. SWCNT electronic structure
2n + m = 3p + λ
(1)
where n and m are the chiral indices of a particular SWCNT
and p is an integer. Two thirds of SWCNTs are in the λ = 1 and
λ = 2 families and are semiconducting (s-SWCNTs), while the
other one third, in the λ = 0 family, are metallic or semimetallic
(m-SWCNTs).14 For over a decade, theoretical calculations
have predicted that the 13C chemical shift of a SWCNT should
depend sensitively on diameter and electronic structure,15−20
but the polydispersity of as-produced SWCNT samples has
precluded validating these computations. Whereas fullerene 13C
NMR can be said to yield truly molecular-level information,
SWCNT 13C NMR has remained at the materials level, with no
unequivocal information on the relation of the observed 13C
NMR spectrum to SWCNTs with specific chiral indices,
diameters, or electronic structures. Only one recent report has
examined the dependence of the 13C chemical shift on SWCNT
electronic structure, within a narrow diameter range.21 Another
Received: November 29, 2011
Published: February 14, 2012
© 2012 American Chemical Society
4850
dx.doi.org/10.1021/ja211181q | J. Am. Chem. Soc. 2012, 134, 4850−4856

Journal of the American Chemical Society
Article
prereduced under H2 (300 sccm) for 30 min at 545 °C. Subsequently,
the temperature was raised to 680 °C under He (300 sccm). Carbon
monoxide (300 sccm) was finally fed at 680 °C for 30 min. 13C-
Labeled SWCNTs were produced by using a mixed gas feed of
ultrahigh purity 12CO and 13CO (Cambridge Isotope Laboratories,
Inc.) with a 13C partial pressure of 0.2. Pure 12C and 13C SWCNT
were also synthesized for comparison. The as-produced samples were
purified by acidic treatments and a series of oxidation steps.
Large diameter SWCNTs were prepared by laser vaporization (LV)
of a graphite target containing 3% (by weight) each of nickel and
cobalt catalysts.32 The 12C graphite utilized in the target was 7−10 μm
graphite purchased from Aldrich. For 13C syntheses, 20% of the
graphite was replaced with amorphous 13C (Cambridge Isotope
Laboratories, Inc.), as described previously.21,33,34 No further
purification was performed before dispersion of the LV SWCNTs
for density gradient ultracentrifugation (DGU) separation.
Density Gradient Ultracentrifugation. Several different DGU
experiments were performed to achieve the range of samples analyzed
in this study. Small-diameter CoMoCAT SWCNTs with three
different 13C labeling levels (0%, partially labeled, and 100%) were
subjected to a single-surfactant separation to achieve separation by
both diameter and electronic structure.9 For this separation, purified
CoMoCAT SWCNTs were dispersed in 2% aqueous sodium cholate
by sonication at 30% with a 1/4 in. ultrasonic tip (Cole Parmer, 750
W) for 1 h. This dispersion was not centrifuged prior to separation by
DGU. The SWCNT dispersion was then blended with a 60% iodixanol
solution (Optiprep density gradient medium) to produce a dispersion
with ∼24% iodixanol. This dispersion was injected into a linear density
gradient that varied from 7.5% and 22.5% iodixanol with 2% sodium
cholate. DGU separation was achieved by centrifuging at 41 000 rpm
(207 570g average centripetal force, 288 244g maximum centripetal
force) in an SW-41 rotor for 12 h at 20 °C. This DGU separation leads
to a thick, dark purple band that lies near the middle of the centrifuge
tube above a much larger dark gray band that extends almost to the
bottom of the centrifuge tube. The bottom layer of the separation, ∼1/
2 in. above the bottom of the tube, consists of a dense mat of either
bundled SWCNTs and/or SWCNTs to which residual catalyst metals
are attached. Four fractions (F1−F4) were removed from the
centrifuge tube, with the bottom fraction (F4) being removed from
approximately 1 in. above the lowest-lying dense mat. This was done
to ensure that all fractions did not contain appreciable amounts of
residual catalyst metals that would broaden NMR peaks.
recent report examined the effect of diameter and wall number
on the 13C chemical shift.22
Beyond the fundamental importance of understanding the
intrinsic magnetic properties of SWCNTs, 13C-labeled
SWCNTs with well-characterized electronic structure could
enable detailed investigations into important covalent and
noncovalent SWCNT interactions. For example, the simple
concept of utilizing 13C-labeled SWCNTs in separation
processes outlined in this study should allow for a detailed
probe of the interfacial coupling of unique SWCNT species
with the molecules used for separation, thus enabling predictive
models for the molecular archetypes that will yield highly
selective and efficient separations. Such samples may also
enable unequivocal identification of sidewall functional groups,
such as endoperoxides, that have been theoretically predicted to
control SWCNT optical and electrical properties.23,24 Finally,
as a technological example, SWCNTs are being explored as
potential absorbers and electron acceptors in organic photo-
voltaic composites with semiconducting polymers.25,26 It has
been demonstrated experimentally that the excited-state charge
carrier dynamics in such composites depend sensitively on the
SWCNT electronic structure.26 However, very little is known
experimentally about the atomic structure at the polymer−
SWCNT interface and how this interfacial structure varies with
SWCNT electronic structure or diameter. Several recent
investigations have used two-dimensional (2D) NMR to
probe molecular ordering, domain size, and composite
morphology in organic photovoltaic (OPV) composites of
semiconducting polymers and fullerenes.27,28 Because excited-
state interactions such as electron/hole transfer are ultimately
dictated by this ground-state interfacial structure, using 1D- and
2D-NMR to gain a detailed knowledge of how polymers
interface with SWCNTs of known electronic structure could
aid in the rational design of better SWCNT OPV active layers.
In this article, we combine the advantages of highly enriched
samples separated by electronic structure and diameter with
13C labeling to enhance the sensitivity and spectral resolution in
solid-state NMR spectroscopy of SWCNTs. The unprece-
dented range of highly enriched, well-characterized 13C-labeled
SWCNT samples produced here allows us to rigorously address
the theoretically predicted dependence of 13C chemical shift on
SWCNT diameter and electronic structure. We utilize
absorbance, Raman, and photoluminescence spectroscopies to
thoroughly characterize the diameter and electronic structure
variations for 13C-labeled SWCNTs separated by density
gradient ultracentrifugation. Subsequent characterization by
solid-state magic angle spinning (MAS) NMR demonstrates
that the isotropic 13C chemical shift for semiconducting
SWCNTs systematically decreases with increasing diameter,
with a functional form that is well characterized by recent first
principle calculations.17 The 13C chemical shifts for small
diameter metallic and semiconducting SWCNTs are found to
differ significantly, while for larger diameter SWCNTs they are
nearly equivalent. Finally, the full-width of the isotropic peak
for metallic SWCNTs is much larger than that of semi-
conducting nanotubes, irrespective of diameter, a trend that to
our knowledge has not been captured by theoretical
calculations.
Both 13C-labeled LV SWCNTs and CoMoCAT were subjected to a
cosurfactant DGU separation that enables enrichment primarily by
electronic structure.9,21 Small diameter CoMoCAT SWCNTs were
only subjected to a DGU separation tailored for extraction of enriched
m-SWCNTs as the most buoyant fraction. For this separation, the
SWCNTs are first sonicated in a 1% sodium cholate (SC) solution
(30% power, 1 h). Then, sodium dodecyl sulfate (SDS) was added to
bring the total concentration of surfactant up to 2% while achieving a
3:2 ratio of SDS:SC by weight. The density gradient for m-SWCNT
separation ranged from 20% to 35% iodixanol, and m-SWCNT
separation was achieved by 16 h of centrifugation at 41 000 rpm and
12 °C. After centrifugation, the CoMoCAT m-SWCNTs were
extracted as a thick yellow band near the middle of the centrifuge
tube. LV SWCNTs were subjected to the same conditions for m-
SWCNT separation. The LV m-SWCNTs were extracted as a thick
blue band near the middle of the centrifuge tube. LV SWCNTs were
also subjected to a DGU separation tailored for s-SWCNT enrichment
as the most buoyant fraction. The cosurfactant mixture for s-SWCNT
enrichment was 4:1 SC:SDS (2% total surfactant concentration).
DGU enrichment was performed in an SW-41 rotor (LX-100P
Beckman centrifuge) at a speed of 41 000 rpm. S-SWCNT enrichment
was achieved by 16 h of centrifugation at 41 000 rpm and 25 °C.
Preparation of Solid-State NMR Samples. Solid-state samples
were prepared for 13C NMR by filtering a particular enriched SWNT
dispersion through a 0.05 μm mixed cellulose ester (MCE) filter.
Following filtration, the enriched SWNT buckeypaper was rinsed with
∼50 mL of DI water to remove excess surfactant and iodixanol. The
filter and buckeypaper were then removed and placed into a beaker of
2. EXPERIMENTAL DETAILS
SWCNT Synthesis. Small diameter SWCNTs were produced by
following the CoMoCAT process.29−31 Briefly, 1 g of calcined catalyst
was set into a vertical quartz reactor of 1/2 in. diameter and
4851
dx.doi.org/10.1021/ja211181q | J. Am. Chem. Soc. 2012, 134, 4850−4856

Journal of the American Chemical Society
Article
presented in our previous communication21 and is not
recounted here for the sake of brevity. For the small diameter
SWCNTs synthesized by the CoMoCAT CVD method,30 we
undertook a detailed analysis by absorbance, Raman, and
photoluminescence excitation (PLE) spectroscopies. Initially,
we performed single-surfactant DGU separations on three
CoMoCAT samples containing the naturally occurring 12C:13C
ratio (99:1) or labeled with a low level (<20%) or 100% 13C.
The top fraction collected from each of these separations was
highly enriched with (6,5) SWCNTs, as evidenced by the
absorbance spectra in Figure 1a.36 The absorbance spectrum of
acetone to dissolve the MCE filter. After ∼16 h, the buckeypaper was
removed and placed in another beaker of acetone to ensure full
dissolution and removal of the MCE. The enriched buckeypaper was
then subjected to mild heating in vacuum to remove surface-bound
molecules. This degas step was performed by folding the buckeypaper
into a preweighed platinum packet that was then placed into a quartz
tube (fitted with a valve) that is custom-made for temperature-
programmed desorption (TPD) experiments. This sample holder was
then attached to the TPD apparatus and evacuated to a base pressure
of ∼3 × 10−8 Torr. A tube furnace placed around the quartz tube
heated the sample to 200 °C over 1 h while molecular desorption was
monitored with a mass spectrometer. Following this degas step, the
platinum packet and sample were reweighed to determine the mass of
the buckeypaper to be measured in the NMR experiment. The Mettler
Toledo microbalance used to measure the platinum and samples
masses has an accuracy of 0.1 μg. The masses of samples measured in
these experiments ranged from ∼100 to 600 μg.
1H−13C CP-MAS experiments were performed on various SWCNT
samples that were soaked in 4 M NaOH for seven days at room
temperature. After being soaked in base, SWCNT samples were dried
briefly under ambient conditions.
MAS Solid-State NMR. High-resolution, solid-state 13C NMR
spectra were collected on a Bruker Avance 200 MHz spectrometer (4.7
T) operating at 50.13 MHz for 13C at room temperature, with MAS
rotation rates of 7 kHz under a nitrogen atmosphere. The spectra were
acquired under high-power proton decoupling using a π/2 pulse
length of 5.2 μs and a recycle delay of 10 s and were referenced to
adamantane (spectra acquired over ∼98 h). NMR samples were
loaded into zirconium oxide rotors (7 mm outer diameter) and packed
between two plugs of either polyethylene or Teflon under ambient
conditions. Polyethylene plugs were used in single pulse experiments
and Teflon plugs in 1H−13C CP-MAS experiments in order to
minimize background signals, balance the rotor, center the small
amount of SWCNT sample in the NMR coil, and provide a barrier
against exposure to ambient conditions. The feature at 139 ppm in the
13C NMR spectra (e.g., Figure 3a) is due to unsaturated impurities in
the polyethylene plugs. CP experiments on the F1-CoMoCAT sample
were carried out at room temperature using a 5 ms contact time,
spinning rate of 7 kHz, and a pulse repetition rate of 1.0 s (spectra
acquired over ∼43 h). CP experiments on the MF1-CoMoCAT
SWCNTs and LV SWCNTs (94% metallic and 92% semiconducting)
were carried out on a Bruker Avance III 600 MHz spectrometer (14.09
T) operating at 150.9 MHz for 13C at room temperature (zirconium
oxide rotors, 3.2 mm outer diameter). The spectra were obtained
under high-power proton decoupling using a π/2 pulse length of 3.4
μs, 5 ms contact time, spinning rate of 15 kHz, and pulse repetition
rate of 1.0 s (spectra acquired over ∼43 h). A total of ∼300K scans
were collected and averaged (spectra acquired over ∼86 h) for the
MF1-CoMoCAT SWCNT sample because of the low sample mass of
∼100 μg.
Raman and Absorption Spectroscopies. Raman spectroscopy
was performed on the raw samples (before separation) and the
enriched samples after separation. Raman spectra were obtained on
unlabeled (99% 12C), 100% 13C-labeled, and partially 13C-labeled
liquid-phase dispersions using 20 mW of 2.33 eV (532 nm) laser
excitation. Solid-state Raman spectra were obtained using 7 mW of
2.54 eV (488 nm) laser excitation. The slit width was 0.1 mm, allowing
for a resolution of 2−4 cm−1 across the measured spectrum.
Absorbance spectra were performed on a Cary 500 UV−vis-IR
spectrometer. Samples were measured in solution phase in 1 cm
cuvettes. Photoluminescence excitation (PLE) maps were obtained
using a home-built Fourier transform spectrometer described
previously.35
Figure 1. (a) Absorbance spectra of (6,5) enriched CoMoCAT
SWCNTs labeled with different levels of 13C, taken after DGU. (b)
Solution-phase G band Raman spectra (λexc = 532 nm) of (6,5)
enriched CoMoCAT SWCNTs labeled with different levels of 13C,
taken after DGU. (c) Absorbance spectra of the five CoMoCAT 13C-
labeled (14% 13C) fractions studied by MAS NMR. (d) Solid-state
RBM and G band Raman spectra for samples F1, F3, and MF1 (λexc
=
488 nm).
each sample is dominated by a sharp peak at 980 nm and a
smaller peak at 571 nm, representing the first and second
excitonic transitions (S11 and S22) of (6,5) SWCNTs,
respectively.
The isotope-induced softening of the Raman-active E1g mode
(G band), shown in Figure 1b, was used to deduce the level of
13C incorporation into the partially 13C-labeled sample that was
used for subsequent NMR studies. Replacement of 12C atoms
with 13C atoms results in a predictable red shift of the SWNT
Raman features because the heavier 13C mass softens the
characteristic vibrations. These red shifts are clearly visible in
the spectra shown in Figure 1b. The theoretical shift (ω13C) for
a sample with a 13C labeling ratio (m13C/[m13C + m12C]) of η13C
can be calculated by the following relation:
ω
= ω
12C
− [(ω
− (ω
×
m
/m ))
13C
12C 12C
12C 13C
× η
13C
]
(2)
3. RESULTS AND DISCUSSION
A rigorous analysis of the SWCNT 13C chemical shift first
requires a detailed understanding of the polydispersity in both
electronic structure and diameter. A detailed analysis of the
large diameter SWCNTs synthesized by laser vaporization is
where ω12C is the Raman shift for an unlabeled sample and
m12C and m13C are the atomic masses of 12C and 13C,
respectively. Equation 2 can be rearranged to yield eq 3,
which allows for the labeling level to be calculated based on the
4852
dx.doi.org/10.1021/ja211181q | J. Am. Chem. Soc. 2012, 134, 4850−4856

Journal of the American Chemical Society
Article
observed Raman shifts for an unlabeled sample and a sample
(7,4), and (11,2) species,38 although no quantitative
information can be determined for the relative concentration
of each species. A full comparison of the RBM spectra for F1−
F4 in the Supporting Information demonstrates that the
weighting of m-SWCNT features systematically increases with
increasing fraction number.36 Collectively, the Raman spectra
illustrate that the five 13C-labeled samples produced via DGU
span the range of a highly m-SWCNT enriched sample (MF1),
to samples of mixed electronic structure (F2−F4), and finally
to a sample highly enriched in (6,5) s-SWCNTs (F1).
labeled with any level (η13C) of 13C:
ω
− ω
13C
12C
− (ω
η
=
13C
ω
× 0.9608)
(3)
12C
12C
Figure 1b illustrates a Raman shift of 1590.5 cm−1 for the G
band of the unlabeled (6,5) sample. Equation 3 predicts that
the 100% 13C-labeled sample (η13C = 1) should have a G band
shift of 1528 cm−1, in good agreement with the observed shift
of 1529.4 cm−1. The partially labeled sample has a G band shift
of 1581.5 cm−1, which translates to a 13C level, η13C, of 14%
when plugged into eq 3.
PLE spectroscopy allows for easier visualization of the
individual s-SWCNTs that comprise each sample and affords a
rigorous analysis of diameter distributions. Figure 2a and 2b
Figure 1c compares the absorbance of the five fractions
collected from DGU separations of the 14% 13C-labeled
CoMoCAT SWCNTs. We note three important regions for
the excitonic transitions of these SWCNTs: (1) the S11
transitions between ∼825 and 1550 nm, (2) the S22 transitions
between ∼515 and 775 nm, and (3) the M11 transitions
between ∼415 and 515 nm. The most buoyant single-surfactant
fraction, F1, is highly enriched with small diameter (6,5)
SWCNTs, as evidenced by the dominance of one large S11 peak
at ∼980 nm and S22 peak at ∼571 nm. Each subsequent, less
buoyant fraction contained successively broader diameter
distributions of s-SWCNTs centered at progressively larger
diameters. Because of the overlap of M11 and S22 peaks for this
particular diameter range, it is difficult to precisely discern the
contribution of m-SWCNTs to the spectra of fractions F1−F4.
However, the cosurfactant separation yields a buoyant yellow
band highly enriched with m-SWCNTs, shown in Figure 1c as
MF1. The absorbance spectrum of MF1 is dominated by three
peaks corresponding to m-SWCNTs in the range of 400−500
nm, in addition to a series of s-SWCNT peaks nearly identical
to the peaks found in F1.36 The similar distribution of s-
SWCNT peaks (although with much lower intensity relative to
the strong m-SWCNT peaks) suggests that the diameter
distribution of MF1 is roughly equivalent to that of F1.
To more precisely examine the distribution of electronic
structures (s-SWCNTs versus m-SWCNTs), resonance Raman
was performed on the solid-state samples prepared from the
dispersions shown in Figure 1c. Figure 1d compares the G band
and radial breathing mode (RBM) spectra of samples F1, F3,
and MF1. For these samples, the 488 nm excitation wavelength
is resonant with the S22 transitions of s-SWCNTs with relatively
small diameters and for the M11 transitions of m-SWCNTs with
relatively large diameters. The RBM spectrum of the F1 sample
is dominated by one large peak at 304 cm−1, corresponding to
the (6,5) SWCNT, along with small shoulders at 290 and 370
cm−1 from minority s-SWCNT species.37 The G band spectrum
of the F1 sample displays two sharp peaks at 1592 and 1520
cm−1, corresponding to the G+ and G− branches of the G band
for s-SWCNTs, respectively.14 Keeping in mind that the s-
SWCNT diameter distribution of MF1 is roughly equivalent to
that of F1, significant differences exist in the Raman spectra of
MF1 due to the enrichment in m-SWCNTs. A series of peaks
in the range of 220 to 290 cm−1 correspond to the RBMs of m-
SWCNTs,38 while the 304 cm−1 (6,5) RBM peak constitutes
only a small shoulder. The G band consists of a sharp peak at
1592 cm−1 and a very broad shoulder peaking at ∼1520 cm−1,
corresponding to the LO and TO branches of the G band of m-
SWCNTs, respectively. The RBM peaks for MF1 are analyzed
in detail in the Supporting Information to determine precisely
which m-SWCNT species are present in this sample.36 This
analysis suggests that the MF1 sample contains the (6,6), (8,5),
Figure 2. (a) PLE map of sample F4. (b) PLE map of sample F1. (c)
PL intensity histogram used to approximate the diameter distribution
of samples F1 through F4. The average diameter, ⟨d⟩, and standard
deviation (Δd, in parentheses) are given for each sample. (d) Section
of the graphene lattice with red hexagons highlighting the relative PL
intensities of semiconducting SWCNTs present in sample F1. The
thickness of the lines are 20, 4.2, 3, 2.5, 1, and 1 for the (6,5), (8,3),
(7,5), (6,4), (8,4), and (10,2) SWCNTs, respectively. Blue hexagons
highlight the m-SWCNTs identified by resonance Raman spectrosco-
py for MF1 (see Supporting Information for details). No quantitative
information on relative amounts of each species present can be
gleaned from Raman or absorbance spectroscopies.
shows photoluminescence excitation (PLE) maps of F4 and F1;
the full suite of PLE maps is given in the Supporting
Information.36 The PLE map for F4 shows significant
contributions for ten s-SWCNT species, while this distribution
is reduced to only six species for F1, with the (6,5) SWCNT
dominating the PLE map. Diameter distributions for each
fraction were extracted from the PLE maps for correlation with
13C NMR chemical shifts and peak widths. The extracted
diameter distributions, shown in Figure 2c, clearly confirm the
diameter selection of the DGU separation, with increasing
fraction numbers demonstrating progressively broader distri-
butions (Δd) centered at larger average diameters (⟨d⟩). A
section of the graphene lattice is given in Figure 2d for easier
visualization of the distribution of chiral species in samples F1
and MF1. The thickness of the red hexagons gives the relative
weighting for s-SWCNT species in the PL spectrum of F1,
while the blue hexagons denote m-SWCNTs identified by the
analysis of MF1 RBM peaks.36 While we cannot definitively
determine quantitative weighting for the amounts of m-
4853
dx.doi.org/10.1021/ja211181q | J. Am. Chem. Soc. 2012, 134, 4850−4856

Journal of the American Chemical Society
Article
SWCNT species in MF1, it is likely that the sample is
dominated by the (6,6) and (7,4) m-SWCNTs because of their
proximity to the dominant (6,5) s-SWCNT in the sample.
With the detailed knowledge of electronic structure and
diameter for the 14% 13C-labeled SWCNTs in hand, we now
turn to the nuclear magnetic resonance spectra. Figure 3a
spectra obtained for bulk samples,22 and attests to the high
degree of enrichment and purity achieved through the DGU
separation.
To study the NMR spectra in more detail, each spectrum in
Figure 3 was deconvoluted into either multiple or single
Lorentzian peak(s) (Figure 4a−c). The 13C resonance for F1,
Figure 3. (a) 13C NMR spectra of F1−F4. The dashed black line is a
guide for the eye, demonstrating the progressive shift to lower
frequency as the diameter distribution broadens to include larger
diameter SWCNTs and metallic SWCNTs. (b) 1H−13C CP-MAS
spectra of 13C-labeled CoMoCAT SWCNT samples: MF1 (red) and
F1 (blue); 13C-labeled LV SWCNT samples: 94% m-SWCNTs
(brown) and 92% s-SWCNTs (purple). The asterisk indicates the
unsaturated impurities present in the polyethylene plugs.
Figure 4. (a−c) Single- or multi-Lorentzian fits to samples F1, MF1,
and F4, respectively. Fits for F2 and F3 are shown in the Supporting
Information. (d) Plot of NMR fwhm (as obtained through the
Lorentzian fitting routine) to the standard deviation of the diameter
distribution (Δd, as obtained from PLE plots) for the peak assigned to
s-SWCNTs.
displays the solid-state MAS 13C NMR spectra for the 14% 13C-
labeled CoMoCAT samples obtained from DGU separations.
1
The H−13C cross-polarization (CP) MAS method has been
used previously to obtain significant enhancements in carbon
magnetization for SWCNTs under various pH conditions and
decrease NMR acquisition time.33,34 In this study, the CP-MAS
method was useful for obtaining high signal-to-noise for
samples with relatively small masses (MF1 and LV SWCNTs).
shown in Figure 4a, is fit well with one Lorentzian centered at
129 ppm (fwhm =5.4 ppm). In contrast, the 13C resonance for
the F4 sample, shown in Figure 4c, requires two Lorentzians:
one centered at 127.3 ppm (fwhm =7.8 ppm) and one centered
at 123.6 ppm (fwhm =10.5 ppm). The spectrum for MF1 is
successfully fit with a single Lorentzian, centered at 125.1 ppm
(fwhm =13.7 ppm). We interpret the spectral fits using the
following model. The narrow 129.0 ppm peak for F1 is easily
assigned to s-SWCNTs on the basis of the very narrow
diameter and chirality distribution, dominated by the (6,5) s-
SWCNT (Figure 2). Similarly, the metal-enriched MF1 sample,
dominated by m-SWCNTs, displays a single broad peak shifted
to lower frequency relative to the semiconducting (6,5)
SWCNTs in F1. As shown in Figure 4c, the 13C resonance
for the F4 sample contains two populations of SWCNTs,
represented by a narrow and broad component. The F1 and
MF1 samples allow us to assign the narrow peak at 127.3 ppm
to s-SWCNTs and the broader peak at 123.6 ppm to m-
SWCNTs. The relative areas of the peaks (2:1) suggests a
semi−metal ratio of ∼2:1, matching the expected statistical
ratio of a random distribution. The larger fwhm for the s-
SWCNT component of F4 (relative to F1) makes intuitive
sense if we assume that the chemical shift varies as a function of
diameter, as theoretically predicted,16,17 meaning a distribution
of diameters will have a distribution of chemical shifts. Indeed,
Figure 4d demonstrates that the width of the isotropic 13C peak
assigned to s-SWCNTs scales directly with the width of the
diameter distribution obtained by PLE analysis (Figure 2c).
Initial semiempirical calculations predicted a large diameter-
independent chemical shift difference of ∼11 ppm between m-
1
Figure 3b shows the H−13C CP-MAS spectra for selected
CoMoCAT samples and 20% 13C-labeled laser vaporization
(LV) synthesized SWCNT samples (94% metallic and 92%
semiconducting) after soaking the SWCNT samples in sodium
hydroxide and serves to illustrate the effects of electronic
structure and diameter. Several interesting trends are
immediately obvious from Figure 3. First, as the CoMoCAT
fraction number increases from F1 to F4, the 13C chemical shift
progressively shifts to lower frequency and the peak envelope
broadens (Figure 3a). Second, the spectrum for the metal-
enriched MF1 sample is shifted significantly to lower frequency
and is very broad relative to fractions F1−F4 (Figure 3b).
Finally, the chemical shift of ∼129 ppm for the ⟨d⟩ = 0.77 nm
(6,5) enriched F1 fraction is shifted to higher frequency relative
to the chemical shift typically observed for the most commonly
studied large diameter SWCNTs (1.2 < ⟨d⟩ < 1.4 nm).21,33
This difference is clearly illustrated by Figure 3b, which displays
the 1H−13C CP-MAS spectra for the DGU-separated LV
SWCNTs with ⟨d⟩ = 1.25 nm. The LV s-SWCNT (92%)
sample exhibits a 13C resonance at ∼121 ppm (fwhm = 5.0
ppm) that is shifted to lower frequency by 8 ppm relative to the
smaller diameter (6,5) enriched sample (F1), demonstrating a
strong diameter dependence for the 13C chemical shift. The
full-width-at-half-maximum (fwhm) for the LV s-SWCNT and
F1 samples are ∼5 ppm. To the best of our knowledge, these
are the narrowest 13C resonance spectra ever observed for
SWCNTs, on the order of 4−5 times narrower than typical
4854
dx.doi.org/10.1021/ja211181q | J. Am. Chem. Soc. 2012, 134, 4850−4856

Journal of the American Chemical Society
Article
and s-SWCNTs.20 However, more recent first-principles
computations have converged on a diameter dependence of
the SWCNT 13C chemical shift that follows the generalized
functional form
diameter of ∼1.4 nm, but at small diameters such as ∼0.8 nm,
the chemical shift for m-SWCNTs would be shifted to lower
frequency relative to the small diameter s-SWCNTs. Our
experimental data qualitatively confirm this trend. The 13C
chemical shifts for both s- and m-SWCNTs with ⟨d⟩ ∼1.25 nm
are nearly equivalent, with the large diameter m-SWCNTs
shifted slightly to higher frequency relative to the equivalent
diameter s-SWCNTs. In contrast, when ⟨d⟩ is reduced to
between 0.77 and 0.82 nm, the observed 13C chemical shift for
the m-SWCNTs is shifted to lower frequency by 3−4 ppm
relative to the s-SWCNTs with similar ⟨d⟩.
We note that although our experimental data confirm the
generally predicted trend of increasing chemical shift differ-
ences for m- versus s-SWCNTs with decreasing diameter, the
magnitude observed experimentally is much less than that
predicted by first-principles computations. For the diameter
range explored here experimentally, both Zurek et al.16 and Lai
et al.17 predict that the λ = 0 family should be shifted to lower
frequency by ∼8 ppm relative to the λ = 2 family, while the
experimentally observed difference is only ∼4 ppm. It is
important to appreciate that the m-SWCNTs expected in the
population distributions in our study are proximal to the (6,5)
SWCNT and thus encompass chiral and armchair m-SWCNTs
(Figure 2d), not the zigzag m-SWCNTs computed by first-
principles computations. Additionally, we note that there is a
high degree of uncertainty associated with the computed
chemical shifts of m-SWCNTs due to the large unit cell that
must be addressed for convergence. These considerations
prohibit us from a rigorous quantitative comparison of our data
to first principles calculations. However, we hope that the
qualitative comparisons made here, using samples with well-
characterized electronic structures, serve as a starting point for
future refinement of the quantitative correlation between
theory and experiment.
α
δ(ppm) = A(λ)/D + B
(4)
where D is the tube diameter, B is the chemical shift limit for
infinite diameter, and A(λ) is a constant that depends on the
nanotube family.16 The values for A, B, and α in eq 4 vary
significantly, depending on the particular calculation, and are
summarized in Table S1 and Figure S5 of Supporting
Information for the three most detailed theoretical reports
available.15−17
To determine the appropriate scaling relationship for the
SWCNT 13C chemical shift, we plotted our experimental data
against the theoretically predicted chemical shift values from
references 15−17. A full comparison to references 15−17 is
found in the Supporting Information.36 Figure 5 compares our
Finally, for both the small diameter CoMoCAT SWCNTs
and the large diameter LV SWCNTs, the isotropic peak for the
m-SWCNTs has a fwhm roughly double that of s-SWCNTs
with similar ⟨d⟩. The reason for this broadening is unclear but is
likely related to coupling of the 13C nuclear spins to electron
spins of free carriers, as was seen recently for heavily doped
SWCNTs.34 To our knowledge, such differences have not been
predicted by first-principles calculations.
Figure 5. Plot of 13C chemical shift as a function of SWCNT diameter
for s- and m-SWCNTs. Experimental points, obtained in this study, are
given as green diamonds (s-SWCNTs) and blue triangles (m-
SWCNTs). Lines show calculated diameter dependence of chemical
shift from Lai et al. for λ = 2 s-SWCNTs (black), λ = 1 s-SWCNTs
(gray), and λ = 0 m-SWCNTs (black triangles). Detailed figure in
Supporting Information denotes sample number for each sample.
4. CONCLUSION
experimental data with the theoretical predictions from Lai et
al.,17 for which we found the best agreement. Specifically, our
data for s-SWCNTs follow the λ = 2 family dependent curve
(solid black line) predicted by Lai et al.17 Because of inherent
uncertainties associated with the theoretical predictions and the
limited number of data points, we hesitate in emphasizing the
quantitative agreement of our data to these calculations.
However, our data are the first to demonstrate that the
generalized functional form of eq 4 predicts the qualitative
dependence of the 13C chemical shift on s-SWCNT diameter.
Additional samples highly enriched in other single species, as
well as further refinements in theoretical calculations, are
needed to fully quantify this diameter dependence.
The data in Figure 5 also allow for the evaluation of how the
SWCNT electronic structure (i.e., metallic versus semi-
conducting) affects the 13C chemical shift and how this effect
varies with diameter. The black triangles plot the computed
chemical shifts for five (n,0) λ = 0 metallic SWCNTs.17 The
calculations shown in Figure 5 predict that the chemical shift
for s- and m-SWCNTs would be nearly overlapping at a
In conclusion, we have presented the first rigorous study of the
effects of both diameter and electronic structure on the 13C
chemical shift in SWCNTs, using SWCNTs highly enriched by
DGU and labeled with 13C. We hope that the results displayed
here will aid in the fundamental understanding of the physical
principles governing the 13C chemical shift in SWCNTs and
can help to refine future computational efforts. Beyond the
fundamental knowledge of the SWCNT 13C chemical shift
accessible through such samples, a number of applied studies
could benefit from the DGU-separated 13C-labeled SWCNTs
demonstrated here. In particular, we suggest the incorporation
of 13C-labeled SWCNTs into a variety of different separation
protocols, such that the resulting SWCNTs can be subjected to
detailed multidimensional NMR analyses to understand the
interfacial interactions between SWCNTs with known (n,m)
indices and molecules such as, DNA,12,13 diazonium salts, and
anionic/cationic surfactants9−11 typically utilized for separa-
tions. Furthermore, type-separated 13C-labeled SWCNTs could
provide detailed information regarding the molecular ordering
4855
dx.doi.org/10.1021/ja211181q | J. Am. Chem. Soc. 2012, 134, 4850−4856

Journal of the American Chemical Society
Article
(20) Latil, S.; Henrard, L.; Goze Bac, C.; Bernier, P.; Rubio, A. Phys.
Rev. Lett. 2001, 86, 3160.
and noncovalent interactions occurring at important interfaces,
such as those between SWCNTs and conducting polymers,
fullerenes, or inorganic semiconductor surfaces.
(21) Engtrakul, C.; Davis, M. F.; Mistry, K.; Larsen, B. A.; Dillon, A.
C.; Heben, M. J.; Blackburn, J. L. J. Am. Chem. Soc. 2010, 132, 9956.
(22) Abou-Hamad, E.; Babaa, M. R.; Bouhrara, M.; Kim, Y.; Saih, Y.;
Dennler, S.; Mauri, F.; Basset, J. M.; Goze-Bac, C.; Wagberg, T. Phys.
Rev. B 2011, 84, 165417.
(23) Ghosh, S.; Bachilo, S. M.; Simonette, R. A.; Beckingham, K. M.;
Weisman, R. B. Science 2010, 330, 1656.
(24) Ju, S.-Y.; Kopcha, W. P.; Papadimitrakopoulos, F. Science 2009,
323, 1319.
(25) Ferguson, A. J.; Blackburn, J. L.; Holt, J. M.; Kopidakis, N.;
Tenent, R. C.; Barnes, T. M.; Heben, M. J.; Rumbles, G. J. Phys. Chem.
Lett. 2010, 1, 2406.
ASSOCIATED CONTENT
* Supporting Information
■
S
Determination of the MF1 population distribution, solution-
phase Raman spectra of samples F1−F4, PLE maps of samples
F1−F4, multi-Lorentzian fitting of F2 and F3, and a detailed
comparison of experimental 13C chemical shifts to available
theoretical predictions. This material is available free of charge
via the Internet at http://pubs.acs.org.
(26) Holt, J. M.; Ferguson, A. J.; Kopidakis, N.; Larsen, B. A.; Bult, J.;
Rumbles, G.; Blackburn, J. L. Nano Lett. 2010, 10, 4627.
(27) Yang, C.; Hu, J. G.; Heeger, A. J. J. Am. Chem. Soc. 2006, 128,
12007.
(28) Nieuwendaal, R. C.; Snyder, C. R.; Kline, R. J.; Lin, E. K.;
VanderHart, D. L.; DeLongchamp, D. M. Chem. Mater. 2010, 22,
2930.
AUTHOR INFORMATION
Corresponding Author
chaiwat.engtrakul@nrel.gov; jeffrey.blackburn@nrel.gov
■
Notes
The authors declare no competing financial interest.
(29) Alvarez, W. E.; Pompeo, F.; Herrera, J. E.; Balzano, L.; Resasco,
D. E. Chem. Mater. 2002, 14, 1853.
(30) Resasco, D. E.; Alvarez, W. E.; Pompeo, F.; Balzano, L.; Herrera,
J. E.; Kitiyanan, B.; Borgna, A. J. Nanopart. Res. 2002, 4, 131.
(31) Resasco, D. E.; Herrera, J. E.; Balzano, L. J. Nanosci. Nanotechnol.
2004, 4, 398.
ACKNOWLEDGMENTS
■
Raw CoMoCAT SWCNTs were supplied by the University of
Oklahoma. SWCNT separation, spectroscopy, and NMR
characterization was performed at NREL and was supported
by the Solar Photochemistry program of the U.S. Department
of Energy, Office of Science, Basic Energy Sciences, Division of
Chemical Sciences, Geosciences and Biosciences, under
Contract No. DE-AC36-08GO28308 to NREL.
(32) Dillon, A. C.; Gennett, T.; Jones, K. M.; Alleman, J. L.; Parilla, P.
A.; Heben, M. J. Adv. Mater. 1999, 11, 1354.
(33) Engtrakul, C.; Davis, M. F.; Gennett, T.; Dillon, A. C.; Jones, K.
M.; Heben, M. J. J. Am. Chem. Soc. 2005, 127, 17548.
(34) Mistry, K. S.; Larsen, B. A.; Bergeson, J. D.; Barnes, T. M.;
Teeter, G.; Engtrakul, C.; Blackburn, J. L. ACS Nano 2011, 5, 3714.
(35) McDonald, T. J.; Jones, M.; Engtrakul, C.; Ellingson, R. J.;
Rumbles, G.; Heben, M. J. Rev. Sci. Instrum. 2006, 77, 053104.
(36) See Supporting Information for additional details.
(37) Doorn, S. K.; Heller, D. A.; Barone, P. W.; Usrey, M. L.; Strano,
M. S. Appl. Phys. A: Mater. Sci. Process. 2004, 78, 1147.
(38) Strano, M. S.; Doorn, S. K.; Haroz, E. H.; Kittrell, C.; Hauge, R.
H.; Smalley, R. E. Nano Lett. 2003, 3, 1091.
REFERENCES
■
(1) Diederich, F.; Whetten, R. L.; Thilgen, C.; Ettl, R.; Chao, I.;
Alvarez, M. M. Science 1991, 254, 1768.
(2) Reed, C. A.; Kim, K.-C.; Bolskar, R. D.; Mueller, L. J. Science
2000, 289, 101.
(3) Troshin, P. A.; Avent, A. G.; Darwish, A. D.; Martsinovich, N.;
Abdul-Sada, A. a. K.; Street, J. M.; Taylor, R. Science 2005, 309, 278.
(4) Kurotobi, K.; Murata, Y. Science 2011, 333, 613.
(5) Zhang, L.; Yang, J.; Edwards, C. L.; Alemany, L. B.; Khabashesku,
V. N.; Barron, A. R. Chem. Commun. 2005, 3265.
(6) Alemany, L. B.; Zhang, L.; Zeng, L.; Edwards, C. L.; Barron, A. R.
Chem. Mater. 2007, 19, 735.
NOTE ADDED AFTER ASAP PUBLICATION
■
Equation 4 contained errors in the version published ASAP on
February 28, 2012; the correct version reposted February 29,
2012.
(7) Zeng, L.; Alemany, L.; Edwards, C.; Barron, A. Nano Res. 2008, 1,
72.
(8) Wang, H.-J.; Xi, X.-K.; Kleinhammes, A.; Wu, Y. Science 2008,
322, 80.
(9) Arnold, M. S.; Green, A. A.; Hulvat, J. F.; Stupp, S. I.; Hersam, M.
C. Nat. Nanotechnol. 2006, 1, 60.
(10) Green, A.; Duch, M.; Hersam, M. Nano Res. 2009, 2, 69.
(11) Liu, H.; Nishide, D.; Tanaka, T.; Kataura, H. Nat. Commun.
2011, 2, 309.
(12) Tu, X.; Manohar, S.; Jagota, A.; Zheng, M. Nature 2009, 460,
250.
(13) Tu, X.; Hight Walker, A. R.; Khripin, C. Y.; Zheng, M. J. Am.
Chem. Soc. 2011, 133, 12998.
(14) Saito, R.; Dresselhaus, G.; Dresselhaus, M. S. Physical Properties
of Carbon Nanotubes; Imperial College Press: London, 1998.
(15) Marques, M. A. L.; d’Avezac, M.; Mauri, F. Phys. Rev. B 2006,
73, 125433.
(16) Zurek, E.; Autschbach, J. Int. J. Quantum Chem. 2009, 109, 3343.
(17) Lai, L.; Lu, J.; Song, W.; Ni, M.; Wang, L.; Luo, G.; Zhou, J.;
Mei, W. N.; Gao, Z.; Yu, D. J. Phys. Chem. C 2008, 112, 16417.
(18) Zurek, E.; Autschbach, J. In Computational Nanoscience;
Bichoutskaia, E., Ed.; Royal Society of Chemistry: Cambridge, 2011;
p 279.
(19) Zurek, E.; Pickard, C. J.; Walczak, B.; Autschbach, J. J. Phys.
Chem. A 2006, 110, 11995.
4856
dx.doi.org/10.1021/ja211181q | J. Am. Chem. Soc. 2012, 134, 4850−4856