Raman spectroscopic characterization of single walled carbon nanotubes: influence of the sample aggregation state

Abstract

Raman spectroscopy has been employed in analytical sciences for purity determination of carbon nanotube samples based on the consideration of G-/D-band intensity ratios. This work demonstrates the role of aggregation in these feature bands, which, in the case of single walled carbon nanotubes (SWNTs), has proved to be crucial for G-/D-band intensity ratio measurements. We have found variation in the relative intensities of G- and D-bands across a sample of SWNTs without any other treatment, discarding the possible influence of the laser beam or sample focusing. In the case of multiwalled carbon nanotubes (MWNTs), this effect is less notorious. Thus, to achieve a good representativeness of Raman measurements, it is important to consider the sample preparation procedure in order to avoid aggregation, which has an effect over the signals, making difficult the subsequent interpretation of results.

---

Introduction

Carbon nanotubes have attracted much attention due to their huge potential in many fields. Potential applications of these nanostructures include use in nanoelectronics,¹ energy storage,² battery and field-emitting-display technologies.³ They also show great potential for sensors,⁴ composites,⁵ catalytic supports,⁶ and as membrane materials for analytical separations.⁷ This increased use of these nanostructures opens up new challenges in analytical chemistry since new methods for their characterization are needed. Microscopy-based techniques such as scanning and transmission electron microscopy (SEM and TEM) and atomic force microscopy (AFM) provide exact information about shape and size, however, some of their most important shortcomings are the low representativeness of the results due to the small sample volumes that are analyzed, as well as sample preparation, which involves drying, and might result in their aggregation.

Ultraviolet-visible-near infrared (UV-vis-NIR) spectroscopy has been employed for distinguishing individualized dispersions of single walled carbon nanotubes from bundles of them by visualization of the broadening and the red-shift of the absorption peaks.⁸ The UV-vis absorption value at 500 nm has been used to quantify different SWNT dispersions,⁹ although it is difficult to distinguish SWNT contributions from other carbonaceous impurities or the dispersing agents. Near infrared luminescence studies performed on individual SWNTs revealed spectral variations within a given (n,m) type of nanotubes,¹⁰ which allows the determination of chirality/diameter distribution present in individually dispersed nanotubes. However, metallic nanotubes produced by quenching and surfactants might influence the peak.

Raman spectroscopy has been widely employed for the characterization of carbon nanotubes. The resonant Raman scattering from SWCNTs generates an intense and easily measurable signal. The main strengths of Raman are its simplicity,¹¹ versatility of measurement conditions, rapidity and non-destructive and non-invasive nature. The information provided about vibrational properties can be correlated with the structure and electronic properties of the nanotubes. Carbon nanotube dispersions have been quantified by Raman spectroscopy normalizing the spectra with respect to the area of a solvent peak,¹² analyzing carboxylated single walled carbon nanotubes (c-SWNTs) previously preconcentrated in an ionic liquid and retained on a membrane¹³ or measuring c-SWNTs preconcentrated on MWNT-modified membranes.¹⁴ Raman spectroscopy has been also used for purity and defect characterization of SWNTs¹⁵ since its mode around 1350 cm⁻¹ is sensitive to structural defects in the graphitic sp² network typical of carbonaceous impurities, such as amorphous carbon particles.¹⁶

The intensity ratio of the tangential mode of SWCNTs (G-band) to the D-band has been used to discuss the purity.^17–19 As has been reported, the evaluation based on the G-/D-band intensity ratio has uncertainty as to whether the D-band intensity reflects the amount of impurity particles or the defect density on the sidewalls. For example, when the D-band mainly reflects carbonaceous impurities in a sample, the G-/D-band intensity ratio becomes a good index of SWCNT purity. Meanwhile, when there are fewer carbon impurities in the sample, the G-/D-band intensity ratio can be used to discuss SWCNT defects.²⁰

In this work, we aim to point out the need for carefully considering operational conditions when measuring Raman spectra of SWNTs. Usually, Raman measurements are acquired at a specific position of the carbon nanotube sample, and this study indicates the importance of sample treatment for debundling of aggregates of nanotubes. In fact, the aggregation state of the sample takes an important role in the featured-bands, as in the case of the G-/D-band intensity ratio, which may lead to an ambiguity in the consideration of different purities of samples when actually the differences are just in aggregation. If it is not taken into account mistakes may be committed. Differences in RBM bands of SWNTs spectra^21–24 or a G-band broadening²⁵ have been described to occur when samples with different aggregation states are measured.^21,22 Herein we have proved that a sample of SWNTs produces spectra with different G-/D-band intensity ratios just by variation of their aggregation distribution across the sample itself, discarding the possible influence of impurities in the sample, laser beam or sample focusing. Thus, in order to achieve a good representativeness in Raman measurements for quality assessment it is necessary to control the aggregation state. As far as we are concerned the relationship of the G-/D-band intensity ratio with the different aggregation states along a carbon nanotube sample has not been previously reported. Dispersion of the carbon nanotube sample with the aid of surfactants is proposed as an effective methodology to control aggregation of bundles of carbon nanotubes. This effect has been observed in the case of single walled carbon nanotubes while for the multiwalled carbon nanotubes it has been found that the effect of aggregation is less significant.

Materials and methods

Chemicals

Carbon nanotubes of different nature were studied in this work, namely single walled carbon nanotubes (SWNTs) purchased from Shenzhen Nanotech Port Co. Ltd (NTP) (Shenzhen, China), with a purity over 90%, an outer diameter of <2 nm, a length of 5–15 μm and a special surface area of 500–700 m² g⁻¹. Multiwalled carbon nanotubes (MWNTs) were provided by CheapTubes (Brattleboro, USA) with an outer diameter of 10–30 nm, a length of 10–30 μm, >90% purity in weight, and ashes content <1.5% in weight. Ethanol and Triton X-100 were provided by Fluka (Buchs, Switzerland).

Equipment

For Raman measurements confocal Raman equipment (alpha500, manufactured by WITec GmbH) was employed. For excitation a frequency double Nd:YAG laser at 532 nm (second harmonic generation) was used, which resulted in a penetration depth in silicon of about 0.5 μm. Raman spectra were collected using a 600 g mm⁻¹ diffraction grating. The laser beam was focused on the sample surface onto a spot of 1 or 3.45 mm in diameter using a 100×/0.95 Nikon or 20×/0.4 Zeiss objectives, respectively. Laser powers were measured directly on the sample stage and were typically between 0.5 and 1.5 mW. Measurement time was about 20 ms for each point of mapping, which typically consisted of 150 × 150 points.

An ultrasound bath without heating (Ultrason model, Selecta, 50 W, 60 Hz) was employed for sample preparation. A Vibracell™ 75041 ultrasonic probe (750 W, 20 KHz, Bioblock Scientific, Illkirch, France) equipped with a 3 mm probe was also employed to prepare the dispersions. Transmission electron microscopy (TEM) images were acquired with a JEOL JEM-1400 transmission electron microscope. The AFM image was acquired with an alpha500 WITec AFM microscope using tapping mode.

Sample treatment and preparation

As received single walled carbon nanotubes were dispersed in ethanol and ultrasonicated to remove impurities of the sample. Ethanol is a pure solvent which does not leave residues on evaporation. Ultrasonication conditions were soft (50 W, 10 min) in order not to introduce defects in carbon nanotubes that will affect the intensity of the Raman D-band. It could be checked by comparison of the Raman spectra of solid samples with the dispersed one that there is no functionalization or degradation of the latter carbon nanotubes. Immediately after the ultrasound step, the dispersion was drop-dried on a microscope glass slice. Nanotubes aggregated as ethanol evaporated, although the size of aggregates was smaller than in an untreated solid sample. Thus, aggregates can be found in the slice, and also free nanotubes at the edges of the aggregates and in the inter-aggregate regions above the crystal surface of the specimen.

Single walled carbon nanotubes were also dispersed in 5 w/v% Triton X-100 solution by using an ultrasonic probe for 10 minutes (750 W, 20 kHz) equipped with a 3 mm probe set at 20% amplitude. Pulses of energy of 20 s on and 20 s off were employed in order to avoid sample heating. No degradation or shortening of carbon nanotubes was observed after sonication treatment.

Raman measurements

Both acquisition time and accumulated spectra for measuring single Raman spectra were optimized. The Raman spectrum at a certain point of the sample was acquired with an integration time of 1 s, the spectrum being the sum of 20 acquisitions.

Raman images in the case of the Raman study to evaluate the influence of the aggregation state shown in Fig. 4 were obtained by measuring a total of 150 points per line and 150 lines per image over a scan width and height of 110 μm × 110 μm with an integration time of 0.05 s. The Raman image of single walled carbon nanotubes dispersed in Triton X-100 (Fig. 6a) was acquired by measuring 250 points per line and 250 lines per image at a scan area of 50 × 50 μm with an integration time of 0.1 s. The image shown in Fig. 7, in which more isolated nanotubes are shown, was acquired by measuring 100 points per line and 100 lines per image at a scan area of 8 × 8 μm with an integration time of 0.1 s. In the case of multiwalled carbon nanotubes the Raman image (Fig. 8a) was obtained measuring a total of 90 points per line and 90 lines per image at a surface of 72 × 72 μm with an integration time of 0.05 s. Finally, the depth scan shown in Fig. 3a were acquired measuring 50 points per line and 100 lines per image for a scan width of 10 μm and a scan depth of 50 μm, with an integration time of 0.05 s.

Results and discussion

Raman spectra of solid as received pristine single walled carbon nanotubes

Raman spectra of a solid sample of pristine single walled carbon nanotubes without any modification or purification step have been recorded at different points of the sample. Differences in the G-/D-band intensity ratio were found as can be seen in Fig. 1a. The sensitivity of the D-band to structural defects in the graphitic sp² network typical of carbonaceous impurities, such as amorphous carbon particles, has been described.^15,16 The sample has not been purified, and, although the supplier states that the purity of carbon nanotubes is over 90%, it could be thought that differences are produced by carbonaceous impurities present in the solid sample. These impurities could have been closely entangled with SWNTs or stuck on the outer surface of carbon nanotubes. The presence of impurities in those carbon nanotubes employed in this study was discarded by microscopy measurements.

Fig. 1 Raman spectra of (a) as received solid pristine single walled carbon nanotubes (spectra normalized to G-band, changes in the relative intensity of D-band are observed) and (b) SWNTs after dispersion in ethanol and subsequently deposited on a glass support (normalized to D-band, changes are observed for the G-band). Each spectrum (a1–a4 and b1–b3) was taken at a different position of the same sample. Spectra were taken with an integration time of 1 s and accumulating 20 spectra with an EMCCD Gain output amplifier.

Raman spectra of single walled carbon nanotubes after dispersion in ethanol

In order to eliminate a hypothetical contribution of impurities, carbon nanotubes were previously suspended in ethanol and ultrasonicated as previously explained for sample treatment. Fig. 2 shows a TEM photograph of the nanotube sample after ethanol treatment once solvent evaporated. As can be seen, there are bundles of aggregated carbon nanotubes but also more isolated nanotubes at the edges of such aggregates. We acquired spectra at different locations within this sample, in which aggregated and non-aggregated carbon nanotubes are present, and the previously observed differences in the G-/D-band intensity ratio for untreated samples persisted (Fig. 1b).

Fig. 2 Transmission electron microscopy (TEM) photograph carried out with a JEOL JEM-1400 transmission electron microscope. The sample was prepared by placing one drop of a solution of SWNTs in ethanol onto a copper TEM grid with a Carbowax forward.

For purity evaluation using Raman spectroscopy, the G-band peak around 1593 cm⁻¹, which is derived from the longitudinal optical (LO) phonons of semiconducting SWCNTs,¹⁵ has been used²⁶ due to the fact that it is less sensitive to the excitation laser energy than the RBM intensity, and there is no significant diameter dependence of the G-band intensity of the LO phonon of semiconducting SWCNTs, while the RBM intensity is more sensitive to their diameter and chirality. Itkis et al.¹⁸ have described that the G-band area is proportional to the relative purity. In addition to the G-band, the G-/D-band intensity ratio has been also employed as an index of purity for characterization of carbon nanotubes in terms of defects density. The intensity ratio of G-band to D-band reflects both the purity and defect density of a SWNT sample. Pure SWNTs can also have a considerable D-band intensity due to structural defects, thus, evaluation of purity through the G-/D-band intensity ratio has uncertainty as to whether the D-band reflects the amount of impurity particles or the defect density on SWNT sidewalls.¹⁵

As can be observed in Fig. 1b, RBM bands also change in the spectra. Their dependence on the aggregation state of carbon nanotubes has been previously described²⁰ since some modes which are initially off resonance in the individually dispersed material, with aggregation, as transitions shift to lower energy, are brought into resonance while others move away. The intensity of the peak at 267 cm⁻¹ has been shown to be proportional to the degree of aggregation.²² It has also been reported that the influence of aggregation on the apparent metallic to semiconductor ratio of dielectrophoretically deposited SWNTs can compromise conclusions based on Raman alone.²⁷ Moreover, the linewidths of the G′-band for solubilised SWNTs have been observed to monotonously decrease with sonication energy density and can be used to monitor the gradual solubilisation of SWNTs.²⁵ Liu et al.²⁸ employed simultaneously Raman measurements and photoluminescence spectroscopy to evaluate the bundling states of SWNTs by using again the 267 cm⁻¹ band and the sum of photoluminescence G-band intensity ratios.

We have focused on changes of G-/D-band intensity ratios and their dependence on aggregation state, a parameter which, as far as we are concerned, had not been considered until date. As will be explained below, changes in G-/D-band intensity ratios observed in this work are not a consequence of changes in purity or sample focusing, nor are problems of laser damage of the sample. It can be concluded that observed changes are due to different aggregation states across the sample.

Influence of the laser beam

The influence of laser power on the spectra of the nanotubes has been studied. The optimum power selected for recording the spectra was 1.0 mW for a laser wavelength of 532 nm using a 20× lens. The laser power is low in order not to damage the sample when acquiring the spectra. It has been described that the use of higher laser power when measuring dispersions of carbon nanotubes in a solution allows better heat dissipation.²⁹ In addition, the laser power to measure the Raman spectrum of bundles of carbon nanotubes must be lower than for isolated nanotube samples due to the poor inter-tube thermal conductivity in CNT bundles.³⁰

Since it had been previously described in the literature that the laser beam may affect the bands,^15,31,32 the influence of laser irradiation during a period of time over the sample was studied. In addition, laser irradiation is known to cause shifts in Raman bands due to thermal effects which lead to an increase in D peaks.³¹ We have found that there is no change in the profile of the considered bands during the exposure windows usually employed for measurements. Spectra of untreated solid SWNTs with different times of laser exposure (0, 3, 6 and 9 min, respectively), showed a G-/D-band intensity ratio that is almost constant. An upshift of baseline was observed, having no influence when normalizing spectra. This observation agrees with that of almost a constant G-/D-band intensity ratio for subsequent measurement after the first power sweep.³³ In any case, measurements have been carried out using the same laser power for all the studies presented in this work.

Influence of focusing

We are measuring a relatively large area of the sample, which implies the presence of bundles of carbon nanotubes, which may have different heights. Sample focusing affects the intensity of the spectra, usually by increasing or decreasing the whole bands of it. The possible influence on the considered parameter (G-/D-band intensity ratio) measured at different heights across a sample was investigated. The small diameter of the multi-mode fiber (50 μm) of our Raman spectrometer acts as a pinhole allowing for confocal microscopy and hence increasing spatial resolution. This allows performing Raman spectra capture of a depth profile of the sample. When acquiring the spectrum at different heights, no significant differences were found in the G-/D-band intensity ratio.

In addition, Fig. 3a shows the Raman image of a SWNT sample across its whole diameter. In this case, instead of acquiring discrete spectra at a series of points, a Raman image was acquired by measuring at different points in an x- and z-axis area (50 points per line and 100 lines per image for a scan width of 10 μm and a scan depth of 50 μm, with an integration time of 0.05 s). The image depicted corresponds to the intensity of the G-band. In order to corroborate our observations, the collection of spectra was subjected to cluster analysis by K-Nearest Neighbors (KNN) under the same conditions explained as follows and, as can be observed in Fig. 3b, the spectrum of each cluster is very similar, so we can conclude that focusing is not the source of variations in G-/D-band intensity ratios in the spectra. Fig. 3c depicts the regions of the image which correspond to the spectrum in Fig. 3b with the same colour.

Fig. 3 (a) Raman image which represent areas with a high intensity of the G-band (yellow) where nanotubes are present. The scan was taken in depth, by measuring in a XZ plane. (b) Raman spectra result of cluster analysis of the spectra collection. There is a little influence of focusing on the D/G relationship. (c) Distribution of the cluster in the sample. The color code of the cluster and Raman spectrum is the same.

Influence of the aggregation state

In order to investigate the source of the observed variations, a Raman image which is composed of spectra acquired at every point on a surface was recorded. Fig. 4a shows an optical microscopy photograph of the sample analyzed. Nanotubes were treated with ethanol and ultrasonicated as previously explained. Specifically, a total of 150 points per line and 150 lines per image with a scan width and height of 110 μm × 110 μm with an integration time of 0.05 s were measured. From the set of spectra we can see differences in the collected spectra.

Fig. 4 (a) Photography acquired with an optical microscope employing a 20× objective of a SWNTs sample ultrasonicated in ethanol and drop-dried on a glass support. (b) Distribution across the sample of the different clusters classified when the set of spectra corresponding to a Raman image was submitted to K-NN analysis. (c) Corresponding spectrum of each cluster.

To prove if it is possible to acquire cluster spectra with similar G-/D-band intensity ratios and then relate them to different properties of the sample, the whole set of spectra was submitted to K-NN cluster analysis. K-Nearest Neighbors is a discriminant, non-parametric technique, based on the distance between objects in a space of dimension equal to the number of variables explored. The sample is assigned to a class where the samples of the training set closest to it have been classified. Only the K closest objects are used to make the assignment. The distance criterion used was the Euclidean distance. Initially, a classification model was constructed in which all spectra were used as the training set. In the second stage, to validate the classification model thus obtained and its stability in predicting, a cross-validation step was performed with five cancellation groups (the spectra are randomly divided into five groups, each of them containing 20% of the total), four of which were used as the training set and the fifth as the prediction set. To perform this cross-validation procedure, the same process was repeated five times with five different training and prediction sets, ensuring that all the samples were included at least once in the prediction set.

K-NN cluster analysis differentiated three regions in the image (see Fig. 4b) which possessed spectra with different G-/D-band intensities. As depicted in Fig. 4b, the regions correspond to the nuclei of the aggregates and concentric outer regions, so it seems logical to think that changes may be produced by the different aggregations inside bundles and around them, which may affect the resonance of carbon nanotubes. In fact, regions which have a low density of carbon nanotubes between aggregates, whose dimensions oscillate between 7 and 25 μm width, have a spectrum similar to bundle edges, in which nanotubes are less aggregated.

The G- and D-band intensities were obtained from their maximum peak counts. The G-/D-band intensity ratio averages at three aggregation levels are 5 ± 0.6 for regions with low aggregation, 9.2 ± 0.8 for regions of medium aggregation and 25 ± 1 for regions of high aggregation, as shown in Fig. 5. It can be seen that there are three different values of relation, related to the aggregation state of the sample. The color code presented in Fig. 5 correlates with the colors assigned to each cluster associated with a different value of the G-/D-band intensity.

Fig. 5 Representation of the G/D ratio obtained from their maximum peak counts for three different levels of aggregation. The color code corresponds to the different clusters classified and depicted in Fig. 4.

Spectra of debundled single walled carbon nanotubes

Exfoliation of carbon nanotubes has been attempted by ultrasonication in the presence of organic solvents³¹ or aqueous solutions of surfactants,³⁴ copolymers,³⁵ polyelectrolytes³⁶ and DNA molecules.³⁷ In previous work,²⁹ the suitability of surfactants to disperse carbon nanotubes without interfering with the Raman signal has been demonstrated. The interaction of linear ionic surfactants such as SDS and CTAB with the nanotube surface is weaker than in the case of Triton X-100 by virtue of its aromaticity. Molecules having a benzene ring structure adsorb more strongly to the graphitic surface due to π–π stacking interactions.³⁸

SWNTs were dispersed in a 5 w/v% Triton X-100 solution with the aid of an ultrasonic probe as explained in sample treatment. Again, a Raman image of the sample was acquired and subsequently submitted to K-NN cluster analysis similarly to that previously described. Fig. 6 shows the results obtained. Fig. 6a depicts the Raman image which reflects the intensity of the G-band, while Fig. 6b classified the different points of the image in two clusters, whose corresponding spectra are shown in Fig. 6c. TEM images corroborated the more debundled state of this sample, as can be seen in Fig. 6d. In this case, the differences in G-/D-band intensity ratios are less significant, the values obtained being 4.59 and 4.34 for the two different classified clusters. Thus, it can be concluded that a suitable control of the aggregation state of the sample is crucial for a representative value of the G-/D-band intensity ratio. In fact, this parameter has proved to be effective for the characterization of mixtures of single and multiwalled carbon nanotubes in a reproducible way when carbon nanotube samples are dispersed in a surfactant solution.²⁹

Fig. 6 (a) Raman image which reflects the intensity of the G-band in a sample of well-dispersed SWNTs in 5 w/v% Triton X-100 drop-dried on a glass support, (b) assignment of regions to the different clusters when the set of spectra was submitted to K-NN analysis, (c) Raman spectra corresponding to each cluster and I_G/I_D calculated for each spectrum, (d) TEM photograph captured from the JEOL JEM-1400 transmission electron microscope of a SWNT sample dispersed and ultrasonicated in 5 w/v% Triton X-100 prepared placing one drop of the dispersion onto a copper TEM grid with a Carbowax forward.

The Raman image was also acquired in a region containing just some SWNTs. Fig. 7e shows the optical microscopy photograph of the scanned surface and Fig. 7d shows an AFM image of the sample. The Raman image (shown in Fig. 7a) was composed of 100 points per line and 100 lines per image at a scan area of 8 × 8 μm measured with an integration time of 0.1 s. Similarly to that described above, the set of spectra was submitted to K-NN cluster analysis, which classified them into three different regions (see in Fig. 7b cluster classification and the corresponding spectra in Fig. 7c). In this case it should be pointed out that the more significant differences found that lead to the classification of the spectra in clusters were differences in G- and G′-band wavenumbers. As can be seen in the insets of Fig. 7c, the G-band positions of the spectra are 1568, 1571 and 1575 cm⁻¹, respectively for the blue, green and red spectra, while the G′-band appears at 2681, 2687 and 2690 cm⁻¹. It seems that the spectrum corresponding to the interior of the carbon nanotubes shows the features at lower wavenumbers.

Fig. 7 (a) Raman image which reflects the intensity of the G-band in a sample of well-dispersed SWNTs in 5 w/v% Triton X-100 drop-dried on a glass support, (b) assignment of regions to the different clusters when the set of spectra was submitted to K-NN analysis, (c) Raman spectra corresponding to each cluster, the insets show the wavenumbers at which the G- and G′-bands appear in each cluster, (d) AFM image of the sample, (e) photography acquired with an optical microscope employing a 20× objective of the SWNTs sample.

Influence of the aggregation state of multiwalled carbon nanotubes

The Raman spectrum of MWNTs has been less studied than that of SWNTs. In order to elucidate if these changes due to aggregation also take place when measuring the spectrum of a sample of multiwalled ones, similar measurements were performed. When acquiring the spectrum of solid (without any preparation treatment) as received multiwalled carbon nanotubes, differences in the G-/D-band intensity ratios were also observed although less significant than in previous studies (Fig. 8).

Fig. 8 Raman spectra of solid multiwalled carbon nanotubes acquired from the same sample at different points (a–d). Spectra were taken with an integration time of 1 s and accumulating 20 spectra with an EMCCD Gain output amplifier.

In order to corroborate it, we measured a Raman image of MWNTs dispersed in methanol and drop-dried on a glass support (Fig. 9a), measuring a total of 90 points per line and 90 lines per image at a surface of 72 × 72 μm with an integration time of 0.05 s. The set of spectra was submitted to analogous K-NN analysis than that described for SWNTs. In this case it can be concluded that aggregation affects the G-/D-band intensity ratio to a less extent. Fig. 9b shows the different regions of the clusters corresponding to the Raman spectra in Fig. 9c with the same colour code.

Fig. 9 (a) Raman image which reflects the intensity of the G-band in a sample of MWNTs dispersed in ethanol and drop-dried on a glass support, (b) assignment of regions to the different clusters when the set of spectra was submitted to K-NN analysis, and (c) Raman spectra corresponding to each cluster and I_G/I_D calculated for each spectrum.

In the case of MWNTs, most of the characteristic bands of SWNTs such as the RBM Raman feature are not usually observed owing to the larger diameter of the outer tubes and the ensemble average of inner tube diameter broadens the signal. The less influence of the bundle state of the sample on MWNT features may be attributed to the presence of several layers. In the case of SWNTs the bands associated with isolated nanotubes are associated with a single-layer, while in the case of bundles of SWNTs the signal is influenced by the nanotubes present around them.

Conclusions

The need for careful consideration of the aggregation state of a sample of single walled carbon nanotubes when measuring the Raman spectrum has been demonstrated because of its influence on G-/D-band intensity ratios, which have been usually employed to discuss sample purity. The possible influence of the laser beam or sample focusing has been discarded. K-Nearest Neighbors classified a set of spectra of a region of carbon nanotubes with different aggregation states into three different clusters, according to their G-/D-band intensity ratios. Thus, regarding the variability introduced by aggregation, a sample pretreatment is proposed which consists of dispersion of carbon nanotubes with the aid of surfactants, such as Triton X-100, and the subsequent measurement of the free carbon nanotubes placed on a glass substrate. It has been proved that when analyzing debundled carbon nanotubes the reproducibility of the measurements is consistent. The influence of aggregation on MWNTs has also been studied and it has been found that this sample is less sensitive to changes produced by this variable than in the case of SWNTs.

Acknowledgements

The authors wish to thank Spain's Ministry of Innovation and Science for funding Project CTQ2007-60426 and Junta de Andalucia for Project FQM02300. A.I. López-Lorente also wishes to thank the Ministry for the award of a Research Training Fellowship (Grant AP2008-02939).

References

1. H. Wang, J. Luo, F. Schäffel, M. H. Rümmeli, G. A. D. Briggs and J. H. Warner, Nanotechnology, 2011, 22, 245305.
2. R. P. Raffaelle, B. J. Landi, J. D. Harris, S. G. Bailey and A. F. Hepp, J. Mater. Sci. Eng. B, 2005, 116, 233–243.
3. S. A. Miller, V. Y. Young and C. R. Martin, J. Am. Chem. Soc., 2001, 123, 12335–12342.
4. N. Sinha, J. Ma and J. T. W. Yeow, J. Nanosci. Nanotechnol., 2006, 6, 573–590.
5. P. J. F. Harris, Int. Mater. Rev., 2004, 49, 31–43.
6. J. M. Planeix, N. Coustel, B. Coq, V. Brotons, P. S. Kumbhar, R. Dutartre, P. Geneste, P. Bernier and P. M. Ajayan, J. Am. Chem. Soc., 1994, 116, 7935–7936.
7. A. I. López-Lorente, B. M. Simonet and M. Valcárcel, Anal. Chem., 2010, 82, 5399–5407.
8. A. Hagen and T. Hertel, Nano Lett., 2003, 3, 383–388.
9. V. A. Sinani, M. K. Gheith, A. A. Yaroslavov, A. A. Rakhnyanskaya, K. Sun, A. A. Mamedov, J. P. Wicksted and N. A. Kotov, J. Am. Chem. Soc., 2005, 127, 3463–3472.
10. A. Hartschuh, H. N. Pedrosa, L. Novotny and T. D. Krauss, Science, 2003, 301, 1354–1356.
11. A. Jorio, M. A. Pimenta, A. G. Souza Filho, R. Saito, G. Dresselhaus and M. S. Dresselhaus, New J. Phys., 2003, 5, 139.
12. C. G. Salzmann, B. T. T. Chu, G. Tobias, S. A. Llewellyn and M. L. H. Green, Carbon, 2007, 45, 907–912.
13. A. I. López-Lorente, B. M. Simonet and M. Valcárcel, Talanta, 2013, 105, 75–79.
14. A. I. López-Lorente, M. L. Polo-Luque and M. Valcárcel, Anal. Chem., 2013, 85, 10338–10343.
15. Y. Miyata, K. Mizuno and H. Kataura, J. Nanomater., 2011, 2011, 786763–786770.
16. S. Reich, C. Thomsen and J. Maultzsch, Carbon Nanotubes: Basic Concepts and Physical Properties, John Wiley & Sons, Berlin, Germany, 2004.
17. A. C. Dillon, T. Gennett, K. M. Jones, J. L. Alleman, P. A. Parilla and M. J. Heben, Adv. Mater., 1999, 11, 1354–1358.
18. M. E. Itkis, D. E. Perea, R. Jung, S. Niyogi and R. C. Haddon, J. Am. Chem. Soc., 2005, 127, 3439–3448.
19. H. Kataura, Y. Kumazawa, Y. Maniwa, Y. Ohtsuka, R. Sen, S. Suzuki and Y. Achiba, Carbon, 2000, 38, 1691–1697.
20. D. A. Heller, P. W. Barone, J. P. Swanson, R. M. Mayrhofer and M. S. Strano, J. Phys. Chem. B, 2004, 108, 6905–6909.
21. V. N. Popov and P. Lambin, Phys. Rev. B, 2006, 73, 165425.
22. D. Yoon, J. B. Choi, C. S. Han, Y. J. Kim and S. Baik, Carbon, 2008, 46, 1530–1534.
23. N. Izard, D. Riehl and E. Anglaret, Phys. Rev. B, 2005, 71, 195417.
24. A. Kumatani and P. A. Warburton, Appl. Phys. Lett., 2008, 92, 243123.
25. J. F. Cárdenas, Carbon, 2008, 46, 1327–1330.
26. K. K. Kim, J. S. Park, S. J. Kim, H. Z. Geng, K. H. An, C. M. Yang, K. Sato, R. Saito and Y. H. Lee, Phys. Rev. B., 2007, 76, 205426.
27. L. M. Ericson and P. E. Pehrsson, J. Phys. Chem. B, 2005, 109, 20276–20280.
28. T. Liu, Z. Xiao and B. Wang, Carbon, 2009, 47, 3529–3537.
29. A. I. López-Lorente, B. M. Simonet and M. Valcárcel, Analyst, 2013, 138, 2378–2385.
30. M. S. Dresselhaus, G. Dresselhaus, R. Saito and A. Jorio, Phys. Rep., 2005, 409, 47–99.
31. K. R. Moonoosawmy and P. Kruse, J. Phys. Chem. C, 2009, 113, 5133–5140.
32. P. Corio, P. S. Santos, M. A. Pimenta and M. S. Desselhaus, Chem. Phys. Lett., 2002, 360, 557–564.
33. J. Judek, C. Jastrzebski, A. Malopeszy, M. Mazurkiewicz, L. Stobinski and M. Zdrojek, Phys. Status Solidi A, 2012, 2, 313–316.
34. V. C. Moore, M. S. Strano, E. H. Haroz, R. H. Hauge and R. E. Smalley, Nano Lett., 2003, 3, 1379–1382.
35. I. Cotiuga, F. Picchioni, U. S. Agarwal, D. Wouters, J. Loos and P. J. Lemstra, Macromol. Rapid Commun., 2006, 27, 1073–1078.
36. Y. Liu, L. Gao, S. Zheng, Y. Wang, J. Sun, H. Kajiura, Y. Li and K. Noda, Nanotechnology, 2007, 18, 365702.
37. M. Zheng, A. Jagota, E. D. Semke, B. A. Diner, R. S. Mclean, S. R. Lustig, R. E. Richardson and N. G. Tasii, Nat. Mater., 2003, 2, 338–342.
38. R. Rastogi, R. Kaushal, S. K. Tripathi, A. L. Sharma, I. Kaur and L. M. Bharadwaj, J. Colloid Interface Sci., 2008, 328, 421–428.

---