1888
J. Phys. Chem. C 2007, 111, 1888-1894
Temperature Dependence of Electronic Transitions of Single-Wall Carbon Nanotubes: Observation of an Abrupt Blueshift in Near-Infrared Absorption Anni Siitonen, Henrik Kunttu, and Mika Pettersson* Nanoscience Center, Department of Chemistry, P.O. Box 35, FIN-40014, UniVersity of JyVa¨skyla¨, Finland ReceiVed: June 26, 2006; In Final Form: October 27, 2006
Near-infrared (NIR) absorption spectra of single-wall carbon nanotube (SWNT) films are studied between 10 and 293 K. The most prominent effect is the shift of bands with temperature. Some nanotubes show a redshift of transition upon increasing temperature while some show blueshift and others show no shift. The shift is interpreted to originate mainly from the effect of strain induced in the tubes because of interaction with the environment. In particular, at temperatures T ) 175-225 K, for some bands, there is an abrupt large blueshift, which is interpreted to originate from interaction of the nanotubes with water. Two models could be considered to explain the effect: (1) strain induced by a phase transition of water confined inside the nanotubes or (2) adsorption-desorption of water on the surfaces of nanotubes, but the current experimental data does not allow to distinguish unambiguously between the two possibilities. Some evidence of the presence of adsorbed water is obtained from coupled thermogravimetric and Fourier transform infrared (FTIR) measurements which indicate that water is released from purified nanotubes upon heating in synthetic air atmosphere.
Introduction Optical spectroscopy in its different varieties has proven to be very useful in studying single-wall carbon nanotubes (SWNT).1-7 Optical absorption spectroscopy probes directly transitions between the van Hove singularities in the density of states (DOS) giving direct information on the electronic structure of SWNTs. Tunable optical excitation and emission spectroscopy has been succesfully used to assign the chiral indices of more than 30 different nanotubes.4 Raman spectroscopy is perhaps the most widely used technique in the characterization of SWNTs since it gives direct information on the tube diameters.1,8 Combining electronic absorption/emission spectroscopy with Raman spectroscopy can provide assignment of the chiral indices of different nanotubes, which is one of the major goals of current research of SWNTs. There has been a relatively few systematic experimental investigations of temperature dependence of spectroscopic observables of SWNTs although such studies can give valuable information. For example, they can provide information on the coupling between electrons and phonons or between optical and acoustic phonons. Moreover, since measurement of band positions by electronic and vibrational spectroscopy has become a general tool to identify the chiralities of nanotubes, it is important to investigate how sensitively the band positions depend on external factors, such as temperature. This is especially important when theoretical calculations are used to aid the assignments, since they usually refer to 0 K temperature. When analyzing the temperature dependence, it is important to keep in mind that intrinsic and extrinsic effects can lead to very different behavior. Capaz et al. studied theoretically temperature dependence of the band gap of isolated semiconducting nanotubes because of electron-phonon coupling.9 They found that most gap shifts at 300 K were negative and small (less than 12 meV), although some tubes with small chiral angles showed * To whom correspondence should be addressed. E-mail: mijopett@ cc.jyu.fi.
nonmonotonic behavior. In general, the shifts were diameterand chirality-dependent. Anharmonic effects were found to be negligible compared to harmonic contribution. Cronin et al. investigated, by tunable Raman spectroscopy, temperature dependence of the optical transition energies of individual suspended nanotubes, thus minimizing the environmental effects.10 They observed small downshifts for all the tubes studied and interpreted this as being due to electron-phonon coupling, in agreement with the predictions of Capaz et al.9 Similar observations were done by Karaiskaj et al. in their photoluminescence studies of semiconducting nanotubes.11 On the contrary, nanotubes in various environments, such as in bundles, wrapped in polymers or in surfactants, show larger shifts and the sign of the shift can be negative or positive depending on the tube family.10-12 This effect is caused by strain induced in nanotubes by difference in thermal expansion coefficients of nanotubes and their environment. Theoretical calculations of the strain-induced band gap shift are in agreement with the experimental observations.12-14 Several investigations have concentrated on temperature effects on frequencies and bandwidths of the Raman lines of SWNTs.15-18 Studies of the temperature dependence of the Breit-Wigner-Fano Raman line in metallic SWNT bundles yielded information on the coupling of phonons to the electronic continuum.19 Quite recently, the temperature dependence of luminescence spectra in SWNTs in micelles was studied.20 The temperature dependence of transition energy and line width indicated that the excitonic states are coupled to the radial breathing mode (RBM) of the nanotubes. According to our knowledge, no corresponding data in absorption below room temperature has been reported so far. In another study, photoluminescence of suspended SWNTs was investigated between 300 and 5 K.21 Fantini et al. reported in their resonance Raman study that the E22S transition energies (E22S ) transition between the second van Hove singularities of semiconducting nanotubes) are redshifted for S1 [(2n + m) mod 3 ) 1] (n, m ) chiral indices) nanotubes and are blueshifted for S2 [(2n + m) mod 3
10.1021/jp0639851 CCC: $37.00 © 2007 American Chemical Society Published on Web 01/12/2007
Electronic Transitions of Carbon Nanotubes ) 2] nanotubes suggesting that the direction of the shift could be used for identifying the type of a nanotube.22 This type of behavior has been attributed to the effect of strain on the band gap which depends on the type of nanotube. This phenomenon allows to divide nanotubes into different categories on the basis of the sign of the band gap shift. Very recently, two separate studies appeared, where a sudden shift of the band gap transition was induced either thermally or by laser irradiation.23,24 It was suggested that this effect was due to adsorption/desorption of some gases on the nanotubes but the identity of the gas was unclear. These observations indicate yet another type of environment-induced mechanism for the shift of electronic transitions in carbon nanotubes and point out the complexity of phenomena connected with the environmental effects on the electronic properties of SWNTs. On the other hand, strong environmental dependence of the band gap shift could be advantageous from the point of view of potential applications, such as gas sensors. In this article, we report systematic studies on temperature dependence of the electronic transitions in SWNT films at temperatures between 10 and 293 K. We probe the transitions directly in absorption in contrast to the previous investigations which were carried out by monitoring emission from the nanotubes. In addition, thermogravimetry coupled with Fourier tranform infrared (FTIR) measurements is used to investigate water content of purified nanotubes. Experimental Methods Experiments were performed by using SWNT material prepared by chemical vapor deposition (CVD) (Thomas Swan Carbon Materials’ Elicarb SW nanotubes (purity 70-90%)). According to specifications, SWNTs were 0.9-1.7 nm in diameter. We measured the Raman spectrum of the material with 532-nm excitation. The RBM mode frequencies span the range 160-330 cm-1, which indicates, by using the relation between the RBM frequency and nanotube diameter from ref 4, the diameter distribution of 0.7-1.4 nm. The purity of the sample was further tested by thermogravimetry which indicated less than 4% of metallic impurities. The solute sample was prepared by weighing 2 mg of nanotube material and then sonicating it in 20 mL of ortho-dichlorobenzene (ODCB) for 5 min with a tip-sonicator (Teopal UP 200s, 24 kHz, 150 W). The sample was then centrifuged (Eppendorf mini Spin plus) for about 2.5 h with maximum speed (corresponding to 14 000g). A thin film of SWNT material was prepared by drying a few drops of supernatant on a MgF2- substrate while gently warming it on a heat plate. To check that the results were not sensitive to the solvent or to the substrate material, the experiments were repeated by using toluene as a solvent or CaF2 as a substrate. In both cases, the results were similar. The sample substrate was placed in a closed-cycle helium cryostat (APD Cryogenics) which was pumped overnight down to ∼10-7 mbar vacuum with a turbo pump. The cryostat was equipped with a resistive heater and a controller (Lakeshore) for accurate adjustment of temperature. The temperature was measured with a silicon diode attached to the cold head near the MgF2 window frame. The near-infrared (NIR) absorption spectra were measured with a Nicolet Magna-IR 760 spectrometer by using a Quartz beamsplitter and an MCT detector. Five hundred spectra were averaged and the resolution used was 4 cm-1. The cryostat was cooled from room temperature to 10 K and the spectra were measured at 15 different temperatures. During measurements, the temperature was stabilized within (0.5 degrees. After cooling, the sample was warmed up back
J. Phys. Chem. C, Vol. 111, No. 5, 2007 1889
Figure 1. NIR-absorption spectrum of an SWNT film at room temperature (293 K) (-----) and at 10 K (s).
to the room temperature and spectra were recorded again at the same temperatures as in the cooling cycle. Two separate samples were prepared and similar measurements were performed for both samples. The results were comparable in both cases. The thermogravimetric measurements were carried out with a thermogravimetric analyzer (Perkin-Elmer TGA 7), which was coupled to an FTIR spectrometer (Perkin-Elmer 2000 FTIR) by a heated gas connection kept at 190 °C. Approximately 6 mg of nanotube powder was placed in a Pt cup and the temperature was ramped between 25-850 °C with a ramping rate of 5 °C/min. FTIR spectra were taken every 1.6 min averaging 20 scans and using the resolution of 2 cm-1. Measurements were conducted in synthetic air (AGA, 80% N2, 20% O2, purity 99.999%, water content max 3 ppm) and nitrogen (AGA, purity 99.999%, water content max 3 ppm) atmospheres. The gas flow-rate was 80 mL/min. To reliably measure small amounts of water released from SWNTs, several measurements were made in identical conditions. Before starting the series of measurements, the empty furnace was heated to 850 °C and was held at that temperature over several hours under flushing with synthetic air to remove any adsorbed water from the system walls. In addition, prior to measurements, the sample compartment and the gas lines were flushed with synthetic air overnight to further reduce the amount of background water. During the measurements, the cyuvette compartment of the spectrometer was flushed with nitrogen gas. Despite these precautions, the amount of water vapor in the measurements fluctuated to some extent during the measurement which lasted up to 8 h. The actual measurement was repeated five times: twice with a sample and three times without a sample (0-measurement) to judge the reliability of the results. Different approaches were tested when attempting to wet nanotubes in a controlled way. In the first approach, the nanotube sample was placed in a sealed container together with a cup of water at room temperature (20 °C) to expose nanotubes to saturated vapor pressure of water at this temperature. In another approach, the sample was heated at 110 °C in a closed container together with a water container to expose nanotubes to higher concentrations of water vapor. Results and Discussion Spectral Observations. Figure 1 shows the NIR absorption spectrum of a thin film sample at room temperature and at 10
1890 J. Phys. Chem. C, Vol. 111, No. 5, 2007
Figure 2. Baseline-corrected NIR-absorption spectra of an SWNT film at (A) 10 K, (B) 175 K, and (C) 293 K. The main bands are numbered (1-5) in A. The dashed line shows the multi-Gaussian fit to the data.
K. The spectrum consists mainly of a continuum type of absorption with some oscillatory structure on top of the continuum. This type of a spectrum is typical for SWNT material prepared by the CVD method.25 The oscillatory structure originates from individual electronic resonances of SWNTs with different chiralities. In the region from 5000 cm-1 to ∼10 000 cm-1, the structures are mainly due to transitions between the first (E11) and second (E22) pairs of singularities in the DOS of the semiconducting SWNTs.26 The excited states can also be considered as excitonic in nature.27 It can be seen in Figure 1 that the oscillatory structure becomes more visible when going from room temperature to 10 K. The continuum contribution in the spectra was removed by making a smooth baseline correction by fitting the spectrum with a function of the type y ) y0 + ax + bx2 + c[1 - exp(-k(x - x0))]. Further analysis was concentrated on the region ∼5600-8800 cm-1 where the individual transitions were most prominent. According to tight-binding calculations, in this region, the transitions should originate either from E11 transitions for nanotubes of diameter about 0.7-1.2 nm or from the E22 transitions for nanotubes of diameter >1.4 nm.26 However, recent calculations by density functional theory (DFT) and experimental findings would suggest that for the E22 transitions in this energy range the diameter should be at least 1.6 nm.4,28 Since the SWNT material used in this study should contain nanotubes with diameter less than 1.6 nm, we believe that the observed bands originate only from the E11 transitions. Accordingly, the diameter range of these nanotubes should be ∼0.91.6 nm.4 The data was smoothed by five-point adjacent averaging and the resulting spectra were fitted with nonlinear least-squares fitting with a sum of seven Gaussian functions. The results of this procedure for temperatures 293, 175, and 10 K are shown in Figure 2. The choice of Gaussian functions is justified since the spectra are inhomogeneously broadened. The
Siitonen et al.
Figure 3. Temperature dependence of transition energies of the five main bands marked in Figure 2. The data are obtained from a multiGaussian fit to the measured spectra. The lines are only guides to the eye.
quality of the fit also supports this choice. On the other hand, the obtained bandwidths should be taken with caution. Since the baseline is arbitrarily generated, it is not possible to obtain the absolute bandwidth (fwhm) unambiguously. However, the relative change in the bandwidth can still be determined reliably. The widths obtained via fitting varied between 140-320 cm-1 at 10 K and between 220-500 cm-1 at room temperature showing clear increase of the width with temperature. It is interesting to compare these values with the widths obtained in emission measurements where values of approximately 250 cm-1 (30 meV) were obtained.20 The similarity of the values suggests that our baseline is properly chosen. However, because of the problems with the baseline and inhomogeneity, the main source of information is in the band positions which can be more accurately determined. Temperature Dependence. Figure 3 shows the temperaturedependent shifts for five different bands numbered according to Figure 2. The temperature effect on the band positions is reversible and, within the accuracy of ∼10 K, there is no hysteresis as can be seen in Figure 4, where the temperature effect for band no. 1 is shown for both cooling and heating cycle. By inspecting Figure 3, three types of behavior are evident: blueshift, redshift, and no shift within the error limits upon increasing the temperature. For bands 1 and 2, there is a smooth blueshift at temperatures below 175 K and a sudden, very strong blueshift at ∼200 K. The low-temperature data for bands 1, 2, and 4 are shown in detail in Figure 5. Low-Temperature Shift. We first consider the interpretation of the low-temperature (
