Anal. Chem.

2008, 80, 2672-2679

Separation of Single-Walled Carbon Nanotubes by

Use of Ionic Liquid-Aided Capillary Electrophoresis
M. López-Pastor, A. Domı́nguez-Vidal,† M. J. Ayora-Cañada,† B. M. Simonet, B. Lendl,‡ and
M. Valcárcel*

Department of Analytical Chemistry, University of Córdoba, E-14071 Córdoba, Spain

This paper presents a simple, highly efficient method for In addition to difficulties in their synthesis, the poor dispers-
analyzing single-walled carbon nanotube (SWNT) bundles ibility of CNTs makes their characterization and processing
based on (1) ultrasound-assisted solubilization/disper- difficult. Specifically, single-walled carbon nanotubes (SWNTs)
sion of SWNTs in the ionic liquid 1-butyl-3-methylimida- exist as heavily entangled aggregates of ropes or bundles4
zolium tetrafluoroborate, (2) encapsulation of the nano- containing hundreds of closely packed CNTs. Interactions between
tubes in sodium dodecyl sulfate micelles, and (3) analysis CNTs result mainly from van der Waals attractions with intertube
by capillary electrophoresis. The process by which SWNTs contact energies of 0.5 eV/nm. For many industrial applications,
disperse in the ionic liquid was studied by Raman using uniformly dispersed CNTs is crucial with a view to ensuring
spectroscopy. No degradation of SWNTs was observed an appropriate final dispersion state of particles in the end product
under mild sonication conditions. The shape and position and, together with the particular characteristics of the CNTs, the
changes observed in the Raman spectral bands for the desired properties.
nanotubes are ascribed to debundling and interaction with Successfully integrating CNTs into nanostructured materials
the ionic liquid. Separation of solubilized SWNTs was
typically meets with major challenges including the mass produc-
accomplished by using a 50 mM formic acid solution at
tion and purification of high-purity CNTs and ensuring uniform
pH 2.0 as background electrolyte and a potential of -10
dispersion and preferential alignment. This has raised the need
kV. Under these conditions, separation was completed
for fast, powerful methods for CNT characterization. Available
within only 4 min. Eighteen peaks for SWNTs were
methods for purity assessment and characterization of CNTs rely
identified in the analysis of commercial SWNT bundles.
on thermal gravimetric analysis,5,6 Raman spectroscopy,7-9 X-ray
The two types of bundles studied exhibited distinct, highly
characteristic electrophoretic profiles which could be used diffraction (XRD) spectroscopy,10,11 transmission electron micros-
to control SWNTs purity. copy (TEM),12 scanning electron microscopy (SEM),13 atomic
force microscopy (AFM),14 UV-vis-near-IR spectroscopy,15,16 and
Ever since their discovery, carbon nanotubes (CNTs) have
(4) Thess, A.; Lee, R.; Nikolaev, P.; Dai, H.; Petit, P.; Robert, J.; Xu, C.; Lee, Y.
aroused enormous interest among researchers on account of their
H.; Kim, S. G.; Rinzler, A. G.; Colbert, D. T.; Scuseria, G. E.; Tománek, D.;
peculiar electronic and mechanical properties and also their high Fischer, J. E.; Smalley, R. E. Science 1996, 273, 483-487.
chemical stability.1 In fact, CNTs possess a unique body of (5) Dillon, A. C.; Gennet, T.; Jones, K. M.; Alleman, J. L.; Parilla, P. A.; Heben,
M. J. Adv. Mater. 1999, 11, 1354-1358.
properties that make them ideal candidates for a wide range of
(6) Chiang, I. W.; Brinson, B. E.; Smalley, R. E.; Margrave, J. L.; Hauge, R. H.
potential nanotechnology and material science applications.2 The J. Phys. Chem. B 2001, 105, 1157-1161.
exact properties of CNTs are highly sensitive to their degree of (7) Gibson, R. F.; Ayorinde, E. O.; Wen, Y. F. Compos. Sci. Technol. 2007, 67,
1-28.
graphitization and diameter (or chirality) and also to whether they
(8) Sato-Berru, R. Y.; Basiuk, E. V.; Saniger, J. M. J. Raman Spectrosc. 2006,
are prepared in single-wall or multiwall form.1 In fact, the synthetic 37, 1302-1306.
method used dictates CNT properties such as the distribution of (9) Arepalli, S.; Nikolaev, P.; Gorelik, O.; Hadjiev, V.; Holmes, W.; Files, B.;
Yowell, L. Carbon 2004, 42, 1783-1791.
diameters and lengths, degree of entanglement, defects, chirality,
(10) Zhu, W.; Miser, D.; Chan, W.; Hajaligol, M. Mater. Chem. Phys. 2003, 82,
and crystallinity,3 as well as the overall quality of the product. 638-647.
However, careful analytical characterization is required in order (11) Rinzler, A. G.; Liu, J.; Dai, H.; Nikolaev, P.; Huffman, C. B.; Rodrı́guez-Macı́as,
F. J.; Boul, P. J.; Lu, A. H.; Heymann, D.; Colbert, D. T.; Lee, R. S.; Fischer,
to ensure efficient use of this new type of material.
J. E.; Rao, A. M.; Eklund, P. C.; Smalley, R. E. Appl. Phys. A 1998, 67,
29-37.
* To whom correspondence should be addressed. Phone and Fax: +34 957 (12) Liu, P.; Liu, L.; Zhang, Y. Phys. Lett. 2003, 313, 302-306.
218616. E-mail: [email protected]. (13) Chiang, I. W.; Brinson, B. E.; Huang, A. Y.; Willis, P. A.; Bronikowski, M.
†
Present address: Department of Physical and Analytical Chemistry, Uni- J.; Margrave, J. L.; Smalley, R. E.; Hauge, R. H. J. Phys. Chem. B 2001,
versity of Jaén, Paraje Las Lagunillas s/n, E-23071, Jaén, Spain. 105, 8297-8301.
‡
Present address: Institute of Chemical Technologies and Analytics, Vienna (14) Bhushan, B.; Kasai, T.; Nguyen, C. V.; Meyyappan, M. Microsyst. Technol.
University of Technology, Getreidemarkt 9/164AC, A-1060, Wien, Austria. 2004, 8-9, 633-639.
(1) Dai, H. Surf. Sci. 2002, 500, 218-241. (15) Kuhlmann, U.; Jantoljak, H.; Pfander, M.; Bernier, P.; Journet, C.; Thomsen,
(2) Baughman, R. H.; Zazhidov, A. A.; de Heer, W. A. Science 2002, 297, 787- C. Chem. Phys. Lett. 1998, 294, 237-240.
792. (16) Kuzmany, H.; Burger, B.; Thess, A.; Smalley, R. Carbon 1998, 36, 709-
(3) Ajayan, P. M. Chem. Rev. 1999, 19, 1787-1799. 712.

2672 Analytical Chemistry, Vol. 80, No. 8, April 15, 2008 10.1021/ac701788r CCC: $40.75 © 2008 American Chemical Society
Published on Web 03/15/2008
scanning tunneling spectroscopy (STM).17 Although these tech- of ionic liquids.26 Also, ionic liquids containing CNTs at concentra-
niques are effective for the rapid detection of impurities, they can tions above 0.5-1 wt % are known to form gels that can be used
only provide global information for CNTs. Common technology to produce new soft materials.27 Such gels have already been used
can therefore hardly allow mixtures consisting of very similar in analytical chemistry to develop electrochemical sensors and
CNTs such as newly synthesized SWNT bundles to be discrimi- biosensors.28-31 Also, ionic liquids have been used as solvents in
nated. In fact, spectroscopic techniques provide inadequate the functionalization of SWNTs.32
resolution for this purpose and microscopic techniques have In the present work, we used Raman spectroscopy to charac-
limited precision (less than 1 nm) in measurements smaller than terize SWNT dispersions in an ionic liquid. Although some Raman
100 nm.18 spectra for SWNTs in ionic liquids have previously been reported,
The tendency of CNTs to form aggregates or bundles poses a they were obtained from SWNT gels of room-temperature ionic
major problem in obtaining discriminative information with a view liquids (RTILs) where the nanotubes were at much higher
to their characterization. Nanotubes can be dispersed mechanically concentrations than in our dispersions; also, the focus was placed
or by altering their surface energy with a physical (noncovalent) on first-order Raman spectra alone.26
or chemical (covalent) treatment. Analytically, mechanical disper- In addition, we propose a simple capillary electrophoresis
sion methods and noncovalent treatments are the most attractive method for characterizing SWNTs in bundles. The method is
as they preserve the initial structure of the CNTs. Mechanical based on the dispersion/solubilization of SWNTs in the ionic liquid
methods such as ultrasonication and high shear mixing are 1-butyl-3-methylimidazolium tetrafluoroborate, encapsulation of the
effective in separating CNT bundles and aggregates but have the SWNTs within a surfactant micelle, and subsequent analysis by
serious drawback that they can fragment nanotubes as well.19 capillary electrophoresis. To the best of our knowledge, this is
Noncovalent surface treatments based on surfactants and polymers the first time that CNTs have been solubilized and dispersed in
have been widely used in the preparation of both aqueous and an ionic liquid for their characterization.
organic solutions.20,21 Nanotube dispersions have to date been
analyzed by size exclusion chromatography,22 field flow fraction- EXPERIMENTAL SECTION
ation,23 and capillary electrophoresis.24,25
Chemicals. The ionic liquid used was 1-butyl-3-methylimida-
The ionic liquids have been widely used in chemistry for
zolium tetrafluoroborate (bmimBF4), which was purchased from
synthesis, but they have also been used in separations, spectros-
Solvent Innovation GmbH (Köln, Germany) and used as received.
copy, electrochemistry, and mass spectrometry.33 Water-im-
The SWNTs were of two different types. Both were supplied by
miscible ionic liquids have been even used for the extraction of
Sigma-Aldrich, in diameters of 0.7-1.2 nm (SWNT-1) or 1.2-1.5
gold nanoparticles from aqueous matrixes. Concretely, 1-butyl-3-
nm (SWNT-2) and lengths from 2 to 20 µm (purity 50-70% by
methylimidazolium hexafluorophosphate has been used to quan-
volume). Formic acid, which was used to prepare the electro-
titatively extract gold nanoparticles and gold nanorods from
phoretic solutions, was also purchased from Sigma-Aldrich.
aqueous solutions.34 The sizes and shapes of the nanoparticles
Sodium hydroxide was obtained from Prolabo (France); HPLC-
and nanorods were not affected by the phase transfer. In this work,
grade methanol and hydrochloric acid were supplied by Panreac
we used an ionic liquid to disperse CNTs more efficiently than
(Barcelona, Spain); sodium dodecyl sulfate (SDS) was purchased
do surfactant and polymers. Carbon nanotubes have been shown
from Sigma-Aldrich (Barcelona, Spain). Finally, purified water (18
to interact with imidazolium ions, which are major building blocks
mΩ) was obtained from a Millipore Milli-Q water purification
(17) Hassanien, A.; Tokumoto, M. Carbon 2004, 42, 2649-2653. system.
(18) Graham, D. Analyst 2007, 132, 95-96. Equipment. Carbon nanotubes were dispersed and solubilized
(19) Lu, K. L.; Lago, R. M.; Chen, Y. K.; Green, M. L. H.; Harris, P. J. F.; Tsang,
in the ionic liquid by using a UP400S ultrasound processor from
S. C. Carbon 1996, 34, 814-816.
(20) Jiang, L.; Gao, L.; Sun, J. J. Colloid Interface Sci. 2003, 260, 89-94. Hielscher Ultrasonic GmbH (Teltow, Germany). The maximum
(21) Gong, X.; Liu, J.; Baskaran, S.; Voise, R. D.; Young, J. S. Chem. Mater. 2000, acoustic power density was 460 W/cm2, and the maximum
12, 1049-1052.
amplitude was 210 µm.
(22) Duesberg, G. S.; Muster, J.; Krstic, V.; Burghard, M.; Roth, S. Appl. Phys. A
1998, A67, 117-119. Dispersed nanotubes were analyzed by using an Agilent HP3D
(23) Chen, B.; Selegue, J. P. Anal. Chem. 2002, 74, 4774-4780. capillary electrophoresis system equipped with a diode array
(24) Suárez, B.; Simonet, B. M.; Cárdenas, S.; Valcárcel, M. J. Chromatogr., A
2006, 1128, 282-289.
detector (Waldbronn, Germany). A fused-silica capillary 30 cm
(25) Doorn, S. K.; Fields, R. E.; Hu, H.; Hamon, M. A.; Haddon, R. C.; Selegue, long × 75 µm i.d. was used for separation. New capillaries were
J. P.; Majidi, V. J. Am. Chem. Soc. 2002, 124, 3169-3174. sequentially conditioned by flushing at 2 bar with 1 M HCl, 0.1
(26) Fukushima, T.; Kosaka, A.; Ishimura, Y.; Yamamoto, T.; Takigawa, T.; Ishii,
N.; Aida, T. Science 2003, 300, 2072-2074.
M NaOH, and Milli-Q water for 5 min each. Before each run, the
(27) Fukushima, T.; Aida, T. Chem. Eur. J. 2007, 13, 5048-5058. capillary in use was flushed at 2 bar with 0.1 M NaOH for 3 min,
(28) Tao, W.; Pan, D.; Liu, Q.; Yao, S.; Nie, Z.; Han, B. Electroanalysis 2006, 17, then with Milli-Q water for 1 min, and finally, running buffer for
1681-1686.
(29) Maleki, N.; Safari, A.; Tajabadi, F. Anal. Chem. 2006, 78, 3820-3826.
6 min. After each run, the capillary was flushed at 2 bar with
(30) Yu, B.; Zhou, F.; Liu, G.; Liang, Y.; Huck, W. T. S.; Liu, W. Chem. Commun. Milli-Q water for 2 min, methanol for 5 min, and Milli-Q water for
2006, 2356-2358. 2 min.
(31) Du, P.; Liu, S.; Wu, P.; Cai, C. Electrochim. Acta 2007, 52, 6534-6547.
(32) Price, B. K.; Hudson, J. L.; Tour, J. M. J. Am. Chem. Soc. 2005, 127, 14867- The separation background electrolyte (BGE) consisted of 50
14870. mM formic acid at pH 2.0. The applied voltage used for electro-
(33) Anderson, J. L.; Armstrong, D. W.; Wei, G. T. Anal. Chem. 2006, 78, 2893- phoretic separation was -10 kV, and the temperature 20 °C.
2902.
(34) Wei, G. T.; Yang, Z.; Lee, C. Y.; Yang, H. Y.; Wang, C. R. C. J. Am. Chem. Samples were introduced by sequential hydrodynamic injection:
Soc. 2004, 126, 5036-5037. 5-fold diluted BGE at 50 mbar for 30 s first and then the sample
Analytical Chemistry, Vol. 80, No. 8, April 15, 2008 2673
at the same pressure but for 15 s. Electropherograms were
recorded at a detection wavelength of 240 nm.
Finally, Raman spectra for solid and solubilized/dispersed
SWNTs were recorded on a Renishaw (in Via Reflex) spectrometer
equipped with a Peltier-cooled CCD detector coupled to a Leika
microscope. The 520.7 cm-1 peak for a silicon standard was
employed for calibration. A 785 nm diode laser with a maximum
power of 300 mW was used as excitation source. Spectra were
recorded by using a ×20 lens and gathered for 60 s (three
accumulations, 20 s each).
Nanotube Solubilization/Dispersion Procedure. Single-
walled nanotube dispersions were prepared by placing 1 mg of
each type of nanotube in a glass beaker containing 2.7 g of
bmimBF4 and inserting an ultrasound probe with a power density
of 63 W/cm2 for 5 min. The concentrated dispersion thus obtained Figure 1. Raman spectra for commercial solid SWNT-1 (black line)
was diluted at convenience with ionic liquid and resonicated in and solid SWNT-2 (red line) as obtained with a 785 nm laser at 1.5
order to prepare more dilute samples by 10-fold dilution with a mW.
100 mM SDS aqueous solution and sonication as described above.
The final solution was directly introduced into the electrophoretic
distinguishable owing to the low transmittance of the available
capillary for analysis.
experimental setup below 200 cm-1.
In order to identify potential effects of sonication, both types
RESULTS AND DISCUSSION
of SWNTs were dispersed in bmimBF4, using variable power
The principal aim of this work was to separate SWNTs of very
densities, numbers of cycles, and times. Thus, the power density
similar shape and dimensions. To this end, we used commercially
was varied from 46 to 460 W/cm2, the number of cycles from 0.2
available SWNT bundles of a high quality as target analytes in
to 1 (a cycle of 0.2 means power discharge for 0.2 s and a pause
order to demonstrate the high efficiency of ionic liquid-aided
of 0.4 s), and times from 2 to 40 min. Following sonication, SWNTs
capillary electrophoresis. This section describes and discusses the
in the resulting dispersions were separated by centrifugation at
solubility-dispersibility tests conducted on the CNTs in the ionic
4000 rpm, washed with methanol three times to remove the ionic
liquid and also the optimization of the capillary electrophoresis
liquid, and dried under a nitrogen stream. The remaining SWNT
(CE) separation procedure.
solids were analyzed by Raman spectroscopy.
Dispersion/Solubilization of SWNTs in the Ionic Liquid.
The Raman spectra for the sonicated SWNTs always exhibited
Preliminary experiments conducted in our laboratory exposed a
an increased G/D ratio. The D-band is associated with disordered
high capacity of ionic liquids (ILs) to disperse and solubilize
sp3-hybridized carbon present as impurities and SWNT defects.37
SWNTs. As stated above, SWNTs exist as ropes and bundles
The G/D ratio, which increases with increasing graphite crystallite
which are heavily entangled with one other and form aggregates
size,9,36,38 is a measure of purity. We thus ascribed the increased
or agglomerates. We chose to use an imidazolium-based ionic
G/D ratio obtained to the sample treatment removing carbon-
liquid to disperse SWNTs with a view to their chemical charac-
aceous impurities from the CNT surface.
terization. However, inspection of the resulting IL-SWNTs
With the use of harsh sonication conditions, a slight upper
samples under an optical microscope revealed the presence of
shift in the bands up to 7 cm-1 was observed (see Figure 2). These
micrometer-sized aggregates. Therefore, effective dispersion
shifts, which could be attributed to degradation of SWNTs, were
required the assistance of ultrasound. Also, because ultrasound
higher in SWNT-1 than in SWNT-2. This effect was not observed
can damage CNTs,19 the SWNTs were dispersed in an ionic liquid
under milder sonication conditions, so in order to avoid degrada-
(bmimBF4) in order to preserve their integrity and prevent
tion 63 W/cm2 was used to prepare the IL-SWNT dispersions.
aggregation.
As confirmed by microscopic inspection, 5 min was long enough
Raman spectroscopy has been used to assess the integrity of
to obtain appropriate dispersions.
SWNTs in bmimBF4 and to provide information about the
A typical Raman spectrum of the so-obtained dispersions is
interaction between the CNTs and the ionic liquid.
shown in Figure 3 together with the pure ionic liquid spectrum.
Figure 1 depicts the Raman spectra obtained for the bundled
Despite the contribution of the ionic liquid to the spectrum, it is
SWNTs studied. The spectra exhibit the typical bands of SWNTs,
still possible to distinguish the characteristic features of the
namely,8,9,36 the G-band (1592 cm-1), which comprises several
SWNTs. The Raman bands of the SWNT can be better observed
tangential modes due to stretching vibrations in SWNT sidewall
in the difference spectrum, which was calculated following
C-C bonds, the D-band (ca. 1290 cm-1), which is assigned to
baseline correction and normalization to the 2974 cm-1 band for
disordered carbon, and its second-order harmonic (the G′-band,
the ionic liquid.
ca. 2575 cm-1). The low-frequency radial breathing mode (RBM)
typically observed in the region of 100-300 cm-1 was barely (37) Dresselhaus, M. S.; Dresselhaus, G.; Jorio, A.; Souza Filho, A. G.; Pimenta,
M. A.; Saito, R. Acc. Chem. Res. 2002, 35, 1070-1078.
(35) O’Connell, M. J.; Bachilo, S. M.; Huffman, C. B.; Moore, V. C.; Strano, M. (38) Laspade, P.; Marchand, A.; Couzi, M.; Cruege, F. Carbon 1984, 22, 375-
S.; Haroz, E. H. Science 2002, 297, 593-596. 385.
(36) Dresselhaus, M. S.; Dresselhaus, G.; Joria, A.; Souza Filho, A. G.; Saito, R. (39) Kahn, D.; Lu, J. Phys. Rev. B 1999, 60, 6535-6540.
Carbon 2002, 40, 2043-2061. (40) Henrard, L.; Popov, V.; Rubio, A. Phys. Rev. B 2001, 64, 205403/1-10.

2674 Analytical Chemistry, Vol. 80, No. 8, April 15, 2008

Figure 2. Raman spectra for commercial bundled (black line) and
sonicated (red line) SWNT-1. Treatment conditions: dispersion in IL
by sonication at continuous power density of 460 W/cm2 for 20 min
and subsequent centrifugation, washing, and drying. Raman spectra
were recorded with 785 nm laser at 1.5 mW.

To further investigate the possible interactions between

bmimBF4 and the SWNTs, Raman spectra of dispersions contain-
ing variable amounts of nanotubes were recorded. Figure 4 shows
the Raman spectra for dispersions containing SWNT-1 at concen-
trations from 4 × 10-4 to 184 × 10-4 % (w/w). To minimize ionic
liquid contribution to the spectra, we focused on the G- and G′-
bands.
A blue-shift of the G′-band was observed in the dispersions.
The magnitude of this shift ranged from 6 to 10 cm-1 and
depended on the concentration of SWNTs. The lower the SWNT
concentration was, the stronger the shift with respect to the solid
SWNT. The fact that the ionic liquid strongly affects the Raman
spectra of SWNTs (especially at low concentrations) can be
ascribed to specific interactions between the imidazolium ion
component and the π-complex of the SWNT. Such interactions
lead to the formation of a new system, IL-coated SWNTs, which
consists of isolated or weakly interacting SWNTs. The extent of
formation of the π-complex depends on the degree of overlap
between pz orbitals in the carbon atoms or, in others words, on Figure 3. Raman spectra recorded using 785 nm laser at 30 mW
the cylindrical shape of the SWNTs. Since the studied bundle for (bottom) pure bmimBF4, (middle) dispersion containing 184 ×
types, SWNT-1 and SWNT-2, differed in their dimensions, they 10-4 % SWNT-1 in bmimBF4, and (top) difference spectrum.
also differed in their π-complexes and hence in the way they
interacted with the imidazolium ion, thereby leading to disparate
shifts in the G′-band. The increasing similarity between the spectra their efficient debundling and makes the resulting dispersions
of the more concentrated dispersions and the SWNT bundles can suitable for subsequent analysis of SWNTs by CZE.
be ascribed to increasing intertube interactions and even to Capillary Electrophoresis Analysis. Nonfunctionalized CNTs
formation of microscopic aggregates. In fact, SWNT concentra- have so far been analyzed as surfactant-coated species by capillary
tions higher than 184 × 10-4 % (w/w) were not considered zone electrophoresis (CZE).24,25 On the basis of our own experi-
because the presence of such aggregates was observed. ence, however, SWNTs are more efficiently solubilized and
Concerning the G-band, although no significant changes in its dispersed in ionic liquids. We thus conducted preliminary tests
position were observed, a decreased bandwidth was apparent with a view to separating IL-coated SWNTs by CZE. The high
(Figure 4). This is seemingly the result of the high-frequency side conductivity of the ionic liquid and its high viscosity precluded
of the band being lost and can be ascribed to a decrease in its use in pure form as BGE. Instead, we used bmimBF4 solutions
intertube interactions in the dispersions. prepared in water, methanol, or aqueous 10 mM ammonium
Our Raman results reveal that the use of SWNTs at low acetate adjusted to pH 8.0 or 5.0. In all cases, the concentration
concentrations and their sonication under mild conditions ensures of ionic liquid was adjusted in order to obtain a current intensity
Analytical Chemistry, Vol. 80, No. 8, April 15, 2008 2675
 cation, which, as expected, migrated before the electroosmotic
flow (EOF). In any case, the high concentration of SWNTs used,
and their strong tendency to interact with the capillary inner walls,
gave rise to a strong, broad peak which resulted in an unstable
baseline. In addition to peak overlap, this resulted in irreproducible
migration times by effect of the bmim cation interacting with the
capillary walls. In order to avoid these problems, the polarity of
the CE system was switched and an acid BGE was used to
eliminate the EOF. This method of separating the analytes was
made possible by surfactant-coated SWNTs possessing residual
negative charge.
The BGE studied included acetic acid at pH 3.5, citric acid at
pH 2.5, and formic acid at pH 2.5. Acetic acid was discarded
because pH 3.5 was not low enough to eliminate the EOF. Also,
citric acid detracted from baseline stability, probably by effect of
its forming citrate complexes with the capillary walls. Therefore,
Figure 4. Raman spectra for various dispersions of SWNT-1 at
concentrations from 4 × 10-4 to 184 × 10-4 % in bmimBF4 by using only formic acid, at pH 1.5-3.5, was further tested as BGE. A pH
a 785 nm laser at 30 mW. value of 2 was finally adopted as a compromise between analysis
time and stability. The formic acid concentration was set at 50
mM in order to ensure adequate buffering capacity. These
conditions facilitated the resolution of analyte peaks within a few
below 120 µA during electrophoretic separation since higher
minutes; however, the peaks were broad and irreproducible. These
currents rendered the system unstable.
shortcomings were ascribed largely to the high conductivity of
Because SWNTs dispersed in the ionic liquid exhibited a high
sample. Thus, although the ionic liquid was 10-fold diluted during
conductivity, the samples were introduced hydrodynamically into
SWNT micellization, the sample conductivity exceeded that of the
the capillary. Direct analysis and the use of stacking and flow
BGE. We therefore assayed inserting a plug of highly diluted BGE
reversal techniques to remove excess ionic liquid all proved
immediately before the sample. This had a strongly favorable effect
inefficient. The complete absence of peaks in the spectra was
on peak shapes, which became much sharper. The 5-fold dilution
ascribed to the high conductivity of the samples and the low
stability of IL-coated SWNTs in the electrophoretic system. of the BGE was adopted since higher dilutions resulted in
Because direct analysis of SWNTs dispersed in the ionic liquid irreproducible migration times.
provided poor results, encapsulating IL-dispersed SWNTs in SDS Once the concentration of the plug was selected, the influence
micelles was assayed. To this end, SWNTs solubilized and of its volume was examined. Applying a pressure of 50 mbar for
dispersed in the ionic liquid were mixed with an aqueous solution 30 s was found to be the optimum choice since lower volumes
containing the surfactant SDS. Mixing caused the ionic liquid to had a negligible effect, probably as a result of the plug mixing
be dissolved in the water and the SWNTs to be coated with the with the sample during hydrodynamic sample introduction.
surfactant during the micelle forming process. In order to avoid Figure 5 shows the typical electropherograms obtained from
the formation of aggregates, the process was also assisted by the analysis of the two commercial SWNT samples. As can be
ultrasonication (see the Experimental Section). A concentration seen, it is possible to obtain 18 distinct peaks for the SWNTs
of 0.5% SDS has been proposed as optimum for the dispersion of within in only 3 min. The separation window determined by the
CNTs in aqueous solution.25 In our previous work, we find that peak of the ionic liquid was about 6 min, but this time can be
the presence of a polymer such as hydoxypropyl methyl cellulose increased by using longer capillaries. Resolution was higher than
(HPMS) has a favorable influence on the dispersion being possible that obtained in previous work.24,25 In fact, rather than the single
to use a SDS concentration of 17.3 mM. In this work the peak obtained under the conditions used before,24 using IL-aided
concentration of SDS used to coat the SWNTs was studied over electrophoretic separation provides 18 well-defined peaks. The
the range of 10-200 mM. The use of a concentration of 200 mM short separation window in which the analytes were resolved must
resulted in a high instability of the baseline. Finally a 100 mM be remarked. As can be seen the peaks were separated in a few
concentration was recommended. This concentration is higher minutes. On the other hand, very narrow peaks were observed.
than the concentration used in the previous work. However, the The peaks obtained using SDS and HPMC were clearly broad
absence of polymer as well as the high ionic strength due to the which was due to the overlapping of the peaks. It should be noted
high concentration of the ionic liquid must be taken into account. that, as inferred from the Raman spectra, such peaks are not due
As expected from previous experience,24 the samples thus to SWNT degradation during ultrasonication. The commercial
obtained (SDS surfactant coated SWNTs) were stable in the CNTs as well as the dispersions of SWNTs in ionic liquid were
electrophoretic system. However, they provided long, irreproduc- also analyzed by TEM. For that, commercial samples were
ible migration times typical of anionic species separated by CZE dispersed in acetone and located into the grid for TEM analysis.
in the normal polarity mode. In addition, the electropherograms The dispersions of SWNT in IL were directly located into the grid
were severely distorted by effect of interferences from the ionic and washed with methanol to remove the ionic liquid. Figure 6
liquid. Thus, the peak of the anion was very close to those for shows a characteristic TEM image of the commercial analyzed
the SWNTs; however, the strongest interference was due to the samples before dispersion. As can be seen, CNTs are aggregated
2676 Analytical Chemistry, Vol. 80, No. 8, April 15, 2008
Figure 5. Electropherograms used to analyze commercial SWNT bundles for quality and purity by using the recommended operating conditions.
The size distribution was as follows: (A) diameters of 0.7-1.2 nm and lengths of 2-20 µm (SWNT-1) and (B) diameters of 1.2-1.5 nm and
lengths of 2-20 µm (SWNT-2). (C) Electropherogram corresponding to the separation of SWNT-2 dispersed in SDS and without prior dispersion
in the ionic liquid.

in characteristic strands. This aggregation made the identification aggregation of CNTs during the preparation of the sample in the
of the CNT size distribution not possible. The electron microscopic grid. The microscopic examination of the CNT surface did not
analysis of the dispersions showed the presence of bigger strands reveal the presence of detectable superficial defects produced by
than those depicted in Figure 6. This was attributed to the the ultrasonic radiation.
Analytical Chemistry, Vol. 80, No. 8, April 15, 2008 2677
 such impurities were not observed in the electron microscopic
evaluation of the dispersion. It can thus be affirmed that the peaks
observed were due to SWNTs and not to the presence of other
impurities. The individual identification of the peaks was discarded
because the CE-Raman coupling does not provide sufficient
sensitivity due to the small sample volume introduced into the
capillary and the short time available for measurement (note the
small width of the peaks).
By using the recommended procedure, the reproducibility of
the method was tested by dispersing/solubilizing six times the
same solid sample. Then, each solution was analyzed three times
by CE. The RSD of the migration times ranged between 4.8% and
6.2%, a result which points out the usefulness of the proposed
dispersion procedure.
The separation of SWNT samples directly dispersed in SDS
Figure 6. TEM image of a strand of SWNTs. The sample was first under the above-described electrophoretic conditions (i.e., with
dispersed in the ionic liquid, then filtered and washed with methanol, no prior dispersion in the ionic liquid) produced an electrophero-
and finally examined by TEM. gram with a poor resolution of the peaks (see Figure 5C). Such
a poor separation efficiency can be ascribed to the unavoidable
presence of aggregates when only SDS is used to disperse the
Commercially available SWNTs have some impurities (see SWNT bundles. In comparison with previous work,24 these
Figure 6), but the high dilution of the sample resulted in a conditions are not the best with a view to ensuring appropriate
negligible concentration of impurities in the dispersions. In fact, dispersion of CNTs with SDS. However, even under the best

Figure 7. Electropherograms used to analyze (A) SWNT-2 and (B) bundles of the same material following ultrasonic degradation.

2678 Analytical Chemistry, Vol. 80, No. 8, April 15, 2008

conditions, it is impossible to achieve the resolution obtained CONCLUSIONS
with IL-aided electrophoretic separation. Ionic liquids such as bmimBF4 provide a powerful medium for
Although the proposed system is quite robust, obtaining solubilizing SWNT bundles with the aid of mild ultrasonication.
reproducible migration times entails thoroughly flushing the Mixing the bundles with an aqueous solution of SDS ensures
capillary as described in the Experimental Section. The main efficient encapsulation of well-dispersed SWNTs in micelles. This
limitation of this technique is the absence of appropriate standards sample preparation method facilitates highly efficient electro-
for correlating migration times with SWNT structure. However, phoretic separation and characterization of SWNT bundles. The
the proposed methodology can be easily used to control the proposed analytical method is highly suitable for quality control
of SWNTs as it affords the highest CE resolution reported to date.
synthesis and purification of CNTs. As noted in the Experimental
Section, SWNT-1 and SWNT-2 exhibited a similar size distribution
ACKNOWLEDGMENT
and yet their electropherograms were quite different. Therefore,
The authors are grateful to the Spanish Ministry of Education
the proposed methodology clearly affords classification of
and Science for funding this work as a part of Project CTQ2007-
SWNT samples.
60426.
Figure 7 compares the electropherograms for SWNT-2 prior
to and after degradation with ultrasound. As can be seen, the
degraded SWNTs exhibited shorter migration times. This is Received for review August 23, 2007. Accepted February
consistent with a decreased SWNTs length since small SWNTs 11, 2008.
possess a higher electrophoretic mobility. AC701788R

Analytical Chemistry, Vol. 80, No. 8, April 15, 2008 2679