- Email: [email protected]
Raman spectroscopy of pristine and doped single wall carbon nanotubes
Thin Solid Films 331 (1998) 141±147
Raman spectroscopy of pristine and doped single wall carbon nanotubes A.M. Rao a, b, S. Bandow c, E. Richter a, P.C. Eklund a, b,* a
Department of Physics and Astronomy, University of Kentucky, Lexington, KY 40506, USA b Center for Applied Energy Research, University of Kentucky, Lexington, KY 40506, USA c Research Center for Molecular Materials, Institute for Molecular Science, Myodaiji, Okazaki 444, Japan
Abstract Pulsed laser vaporization (PLV) of a heated, Fe/Ni or Co/Ni catalyzed carbon target in Argon gas has been used to synthesize single-wall carbon nanotubes (SWNTs). The Raman spectrum of SWNTs is very different from that observed for graphite (the nanotube's ¯at parent) and these differences can be understood as a result of the cyclic boundary condition imposed on a graphene sheet rolled up to form a seamless nanotube. Optical resonances are observed which are associated with the one-dimensional character of the electronic states of these novel quantum wires. The high frequency tangential mode frequency is signi®cantly upshifted (downshifted) when the SWNT bundles are exposed to acceptor (donor) dopants. q 1998 Elsevier Science S.A. All rights reserved. Keywords: Single wall carbon nanotubes; Raman scattering; Quantum con®nement; Doped single wall carbon nanotubes
1. Introduction Single wall nanotubes (SWNTs) were discovered in 1993 in the carbonaceous by-products from an arc discharge between carbon electrodes in an inert atmosphere [1,2]. They have been observed directly in electron microscopes and in scanning tunneling microscopes. In Fig. 1 the sp 2 based structure of a seamless carbon nanotube is shown schematically. The hexagonal arrangement of C-atoms, identical to that in ¯at carbon sheets of graphite, is apparent. Three different subclasses of seamless nanotubes can be formed: armchair (n,n), zigzag (n,0) and chiral (n,m±n) where the integers n and m are used to de®ne the symmetry of the nanotube [3]. The notation (n,m) de®nes the atomic coordinates for the 1D unit cell of the nanotube. For more details on carbon nanotubes see Ref. [3]. It is of interest to note that `armchair' and `zigzag' tubes have rows of hexagons aligned parallel to the tube axis and `chiral' tubes can be formed such that the hexagon rows spiral up along the tube axis. The tube shown in Fig. 1 is a (10,10) armchair tube. The diameter of an armchair tube is given by d A 1:357n [3], so that a (10,10) nanotube has a diameter of Ê. 13.57 A Until recently, research on the physical properties of these quantum carbon wires have been slowed by the fact that previous arc discharge methods produced mostly carbon soot and only a few percent carbon nanotubes * Corresponding author.
[1,2]. Thus experimental signals in these samples were often dominated by the response of the soot. Recently, a pulsed laser vaporization (PLV) technique was discovered at Rice University which produced over 70% tubes [4]. The tubules are organized into regular hexagonal arrays to form `ropes' or `bundles' and exhibit high aspect ratios (length/ diameter) .10 3. The diameter distribution of the tubules was reported to be sharply peaked at t1.36 nm, close to the diameter predicted for a (10,10) `armchair' tube [4]. Details of the laser-assisted process developed at Rice University are available elsewhere [4]. Brie¯y, a carbon target containing 1±2% Ni/Co catalyst is maintained in ¯owing Ar in an oven (12008C). The carbon target is vaporized by a two-pulse sequence from a YAG and frequencydoubled YAG laser at a 10-Hz repetition rate. Nanotubes (.70%) and carbon nanospheres (t30%) and fullerenes (t1±2%) form in the hot carbon plasma and drift downstream under ¯owing Ar where they are collected on a water-cooled cold ®nger. The carbon material was ®rst soaked in CS2 to remove solubles (i.e. fullerenes) and the insolubles were then subjected to micro®ltration to physically separate the tubes and soot. This was accomplished by dispersing the material in benzalkonium chloride using an ultrasonic bath [5]. In this paper, we wish to explore the nature of these onedimensional (1D) carbon quantum wires using Raman spectroscopy. We will also discuss the doping-induced shifts in the SWNT Raman-active mode frequencies. Finally, we discuss the effect of the catalyst on the tube yield, and the
0040-6090/98/$ - see front matter q 1998 Elsevier Science S.A. All rights reserved. PII S0 040-6090(98)009 10-9
142
A.M. Rao et al. / Thin Solid Films 331 (1998) 141±147
Fig. 1. Schematic representation of an `armchair' (10,10) SWNT. The solid circles represent the carbon atoms.
evolution of the tube diameter distribution with increasing growth environment temperature T. 2. Pristine SWNTs In Fig. 2, we display the Raman spectrum (300 K) for puri®ed single wall nanotubes obtained in the backscattering geometry using 514.5-nm Ar laser radiation [6]. The SWNT samples were synthesized at the Rice University using the PLV method and a growth environment temperature T of 12008C [4]. For comparison, the calculated Raman
Fig. 2. Raman spectrum (top) of SWNT sample taken with 514.5 nm excitation at t2 W/cm 2. The asterisks (*) in the spectrum indicates features that are tentatively assigned to second-order Raman scattering. The four bottom panels are the calculated Raman spectra for armchair (n,n) nanotubes, n 8±11, as obtained within a bond-polarizability model [6]. The downward pointing arrows in the lower panels indicate the positions of the remaining weak, Raman-active modes [6].
Fig. 3. Room temperature Raman spectra for puri®ed SWNTs excited at ®ve different laser frequencies. The laser frequency and power density for each spectrum is indicated, as are the vibrational frequencies [6].
spectrum is shown below for n 8±11 armchair tubes. The frequencies were calculated in a phenomenological force constant model using the same CZC force constants used to ®t vibrational data for a ¯at graphene sheet [7]. The theoretical Raman intensities were calculated using a bond polarizability model by Richter and Subbaswamy [6]. It may be noticed that some of the experimental bands are narrow and some are broad. This difference is attributed to an inhomogeneous line-broadening mechanism based on the theoretical observation that particular vibrational modes exhibit a strong tube diameter dependence, while others exhibit a rather weak dependence. Thus, considering that our sample contains a distribution of tube diameters, the Raman lines can be sharp (similar to a linewidth in graphite t6 cm 21) or broad, depending on whether or not the mode frequency is strongly diameter-dependent. The intense line seen at 186 cm 21 is identi®ed with the radial breathing mode in which all C-atoms are displaced radially outward in phase. The strong lines observed near 1600 cm 21 are related to the intralayer vibrations in graphite which are observed at
A.M. Rao et al. / Thin Solid Films 331 (1998) 141±147
143
tubes and the fact that these different diameter tubes have different optical resonances. In 1D systems it is well known that E 21/2 singularities exist in the electronic density of states (DOS). These singularities manifest themselves as spikes in the DOS calculated for n 8±11 armchair tubes [6,8]. The allowed optical transitions for these tubes can be shown to be between ®lled valence states in spikes (v1,v2) below the Fermi energy level EF to empty conduction band states (c1,c2) above EF [6,8]. Raman scattering from an (n,n) tube will dominate the spectrum when the laser photon energy matches the energy difference between these E 21/2 spikes for that (n,n) DOS. The center of gravity of each Raman band then shifts to the frequency of the (n,n) vibrational mode being resonantly driven by the laser ®eld [6]. Thus, the unusual laser frequency dependence of the Raman spectra of carbon nanotubes is a direct consequence of 1D quantum con®nement effects in carbon nanotubes. Furthermore, calculations within a simple tight binding model for a family of armchair tubes ((8,8) to (11,11)) show that, for a ®xed excitation frequency, the resonance curve-derived from the (v2,c2) electronic transition for smaller diameter armchair tube may overlap with the resonance curve derived from the (v2,c2) electronic transition for relatively larger diameter armchair tubes [8]. Therefore, one should expect a complicated dependence of the peak positions and intensities on the incident excitation frequencies. Fig. 4. Raman scattering spectra for pristine SWNT bundles reacted with various donor and acceptor reagents. From top to bottom: I2, Br2, pristine SWNT, Rb, K. The asterisks (*) indicate the positions of peaks associated with halogen reactant. The peak frequencies indicated in parentheses for the K, Rb doped SWNTs are the renormalized phonon frequencies [10].
1582 cm 21. In the nanotube, the cyclic boundary conditions around the tube waist activate new Raman and IR modes that are not observable in a well-ordered ¯at graphene sheet or in graphite ± this is the ®rst consequence of the onedimensional (1D) nature of the nanotube. The room temperature Raman scattering data on the same sample (Fig. 3) were also collected with four different laser excitation frequencies (647.1, 780, 1064 and 1320 nm) at low power densities [6]. Both the Raman peak intensities and frequencies were observed to change dramatically with laser excitation frequency. Resonantly enhanced Raman scattering occurs when the energy of the incident photon matches the transition energy of a strong optical absorption band [9]. Normally in a resonant Raman scattering process, with variable laser frequency and ®xed incident ¯ux the Raman line intensity can vary by several orders of magnitude, but the Raman line frequency is ®xed. The large shifts in frequency that were also observed is quite unusual [6]. We have identi®ed these large shifts in Raman line frequency with a diameter-dependent optical absorption which promotes resonant scattering from particular diameter tubes [8]. The shifting frequencies are both the result of the sample being a collection of different diameter
3. Doped SWNTs In Fig. 4, we display the T 300 K Raman spectra of pristine SWNTs and those doped with the anticipated acceptors (I2,Br2) and donors (K,Rb). As discussed above for the pristine SWNTs, the peak at 186 cm 21 has been identi®ed with the Ag symmetry radial breathing mode, and the peak at 1593 cm 21has been assigned to an unresolved Raman triplet identi®ed with tangential C-atom displacement modes [6]. For the SWNT bundles reacted with I2 or Br2 (K or Rb) dopants, the Raman active mode frequencies are expected to upshift (downshift) due to the transfer of electrons from (to) the carbon p (p *) states in the tubules to (from) the dopant molecules [10]. Signi®cant shifts in the frequencies of the radial and tangential modes were observed for all dopants indicated in Fig. 4, except for the I2-doped SWNT bundles. This is perhaps not surprising, as I2 does not intercalate into graphite. The low frequency radial breathing mode in K or Rb doped sample is not evident in the spectrum ± it may have downshifted below our lower frequency limit (t100 cm 21), or, alternatively, it may have broadened to the point it cannot be detected above the background. The asymmetric peak at t1565 cm 21 observed in both alkali metal-doped SWNT bundles is identi®ed with a BreitWigner-Fano (BWF) resonance [10] which usually involves an interference between Raman scattering from continuum excitations and that from a discrete phonon. The degree of coupling (1/q) determines the departure of the lineshape
144
A.M. Rao et al. / Thin Solid Films 331 (1998) 141±147
breakdown in the pristine nanotube selection rules associated with the doping. The linewidths are also very large, suggesting broadening through the electron phonon interaction or through inhomogeneous doping. It should also be mentioned that BWF lineshapes have been observed previously in doped C60 (K3C60, Rb3C60 [12]). The value of 1/q t20.35 we observe here for the M-doped SWNT bundles is about a factor of three smaller than those observed for the broad BWF resonance in C8M compounds [13] and similar values reported for the low frequency Hg symmetry, intramolecular vibrations in M3C60 [12]. Using the charge-transfer- induced shifts in the E2g2 high frequency intralayer mode in the H2SO4 doped GICs, we ®nd f t 1=19 (f is the transferred charge per C-atom), or one free hole per 19 C-atoms in the bromine saturation doped SWNTs [10]. For the alkali metal dopants (K,Rb), the two contributions (i.e. charge transfer and interference (1/q)) to the BWF frequency v 0 cannot be separated without a detailed analysis of the continuum excitations. Therefore, for the alkali metal doped SWNT bundles we cannot estimate f at this time.
Fig. 5. High frequency Raman spectrum for K and Rb doped SWNT bundles (T 300 K; excitation 514.5 nm). The spectra are ®t with a superposition of Lorentzians and an asymmetric Breit-Wigner-Fano lineshape on a linear continuum and the dashed lines represent the individual contributions to the spectrum. In both cases, the calculated spectrum nearly superimposes on the data. Peak frequencies are indicated; the values in parentheses are the renormalized phonon frequencies.
from a symmetric Lorentzian. It should be noted that when the coupling is signi®cant, the BWF peak position can be considerably displaced from the value of the uncoupled discrete phonon [11]. This is referred to as a renormalization of the `bare' phonon frequency. If the continuum scattering in the SWNTs is electronic in origin, then the degree of frequency renormalization can be an important measure of the electron±phonon interaction. The BWF lineshape is given by: I v I0 {1 1 v 2 v0 =qG}2 ={1 1 v 2 v0 =G2 }
(1)
where I0, v 0, and G are, respectively, the intensity, renormalized frequency and broadening parameter. In Fig. 5, we show a ®t to the high frequency experimental spectra for Rband K- doped SWNTs. The dashed lines represent the individual Lorentzians used for the peaks at t949, 1145, 1222, 1263 and 1329 cm 21 and the BWF lineshape with its peak at 1567 cm 21. The Lorentzian peaks are tentatively assigned to charge-transfer-softened SWNT modes predicted in theoretical calculations [8]. However, the doping-induced increase in the Raman activity of the mid frequency (650±1350 cm 21) phonons is not understood and may be related to a
4. Effect of the growth temperature on the diameter distribution and chirality of single wall carbon nanotubes Finally, we discuss the in¯uence of the laser target and tube growth environment temperature (T) on the diameter and symmetry of SWNTs produced by the PLV method at the Institute for Molecular Science, Japan. A two-pulse sequence from a Nd/YAG laser (532-nm radiation followed by 1064-nm radiation at 20 Hz and coaxial with the 532-nm radiation) were focused in a 3-mm diameter spot on carbon targets containing either Fe/Ni (0.6:0.6 at.%) or Co/Ni (2.6:2.6 at.%). Each target was placed in a quartz tube near the center of a 30-cm tube furnace. Flowing Ar gas (100 sccm, 500 Torr) was introduced at the front of the furnace that carried the SWNTs and other carbon products produced in the carbon plasma to the rear of the furnace where they were collected just outside on a water-cooled Cu cold ®nger. The average temperature near the carbon target and growth environment (T) was controlled to within ^18C. From electron microscopy, X-ray diffraction and Raman experiments, we observe a systematic increase in the average tube diameter with increasing growth temperature above a threshold temperature T0 t 8508C for tube production. The same general trend was observed for the target containing the two different bi-metallic catalysts (Fe/Ni and Co/Ni), i.e. lower T values promote the growth of smaller diameter tubules in smaller diameter bundles. As discussed below, the diameter dependence of the frequency of the Ramanactive breathing mode are consistent with the production of bundles of tubules with chiral angles u with less than t108
A.M. Rao et al. / Thin Solid Films 331 (1998) 141±147
145
frequency (v r) vs. the inverse tube diameter (1/d) for all symmetry types, i.e. (n,n), (n,0), and (n,m). The calculated values can be found in Ref. [15] for comparison to experimental frequencies. As can be seen from Fig. 6, the calculated frequencies all fall on a straight `universal' line well ®t by the function:
vr 223:75 cm21 ´nm=d nm
Fig. 6. Calculated radial frequency v r vs. inverse tube diameter (1/d) for armchair (W), zigzag (K) and chiral (L) symmetry tubes (Eq. (2)).
of the (n,n) or `armchair' direction, similar to the range of angles reported earlier by Cowley et al. [14] for PLV grown tubes synthesized at T 12008C by the Rice University group. In Fig. 6, we plot the calculated radial breathing mode
Fig. 7. Low frequency, room temperature Raman spectra for nanotubes produced using Fe/Ni catalyst at the temperatures and excitations indicated in the ®gure. In (a±d), the shift of intensity towards lower frequency radial modes with increasing T signals the increased production of larger diameter tubes (see text).
(2)
As anticipated, for d 1, vr 0, as this frequency corresponds to the q 0 transverse acoustic (TA) phonon of a ¯at graphene sheet and this mode thereby corresponds to a rigid displacement of the sheet normal to its surface. Our calculations show clearly that v r is only sensitive to inverse diameter and is not sensitive to the helicity (or symmetry) of the particular tube. Thus, this Raman-active radial mode, which can exhibit quite a large resonant Raman cross section [6,8], is ideal for determining the diameter of SWNTs. In Fig. 7, we display the low frequency region of the Raman spectrum for SWNTs where the resonantly enhanced lines associated with the radial breathing modes of various diameter SWNTs are expected. The data were taken at 300 K on tubes grown at T 1000, 920, 860, and 7808C using an Fe/Ni catalyzed target. The spectra are unpolarized and were collected in the backscattering con®guration. The bottom four spectra (Fig. 7a±d) were excited using Nd/ YAG laser radiation (1064 nm) and represent the change in the radial mode scattering with increasing temperature (bottom to top). The upper four spectra (Fig. 7d±g) were all taken on the SWNT sample grown at 10008C but at different excitation wavelengths (1064, 647, 514.5, and 488 nm). These spectra were chosen to show the strong resonance effects (i.e. peak shifting and intensity change) as discussed in Ref. [6]. As the data in Fig. 7 indicate, to ®nd all the radial mode frequencies associated with a given sample, it is necessary to use several excitation wavelengths. All the peak positions were obtained from a Lorentzian lineshape analysis (for example, see Fig. 7d) were recorded for tubes produced both for the Fe/Ni and the Co/Ni catalyzed targets [15]. They have all been identi®ed with the resonantly enhanced radial breathing mode for tubes with differing diameter (see Table 1) [15]. Other lower symmetry, low frequency modes exhibit a signi®cantly weaker Raman scattering cross-section [6]. From the calculated relation vr t 1/d, it is evident that smaller (larger) diameter tubes should exhibit higher (lower) radial mode frequencies. Thus the evolution of the spectra in Fig. 7a±d, i.e. the shift of intensity towards lower frequency radial modes with increasing T, signals the increased production of larger diameter tubes, consistent with the Xray diffraction (XRD) and transmission electron microscopy (TEM) data (see Ref. [15]). A similar general trend is observed with other excitation wavelengths and also on SWNTs prepared from Co/Ni catalyzed targets. In Fig. 8, we display a map of chiral vectors C which can
146
A.M. Rao et al. / Thin Solid Films 331 (1998) 141±147
Table 1 Experimentally determined radial mode frequencies for SWNTs produced using the Fe/Ni (indicated by (*)) and the Co/Ni catalyzed targets a Experiment 1064 nm
647.1 nm
514 nm
230 213 204 190
177
258* 232* 230
222, 221* 215, 214*
224*, 222
222, 221*
206* 206
203 203* 193 192*
186* 185
182
179* 174, 174*
180*, 178 178* 174, 173*
167, 167*
167*, 166
167, 166*
163, 163*
163*
163, 163*
180 179* 171, 170*
160
257 255* 242, 240*
231* 232
192, 192* 189
157 150
145
Notation
Diameter (nm)
275.0 273.7 253.6 251.6 242.4 235.7 234.9 232.5 219.8 218.5 206.2 205.7 204.1 194.0 193.1 191.3 183.3 182.9 181.8 178.0 177.6 173.6 172.9 171.7 165.0 164.7 163.9 162.6 157.1 156.6 155.7 150.0 149.8 149.2 148.2 143.4 142.9 142.2
(6,6) (7,5) (7,6) (8,5) (10,3) ³ (7,7) (8,6) (9,5) (8,7) (9,6) (8,8) (9,7) (10,6) (9,8) (10,7) (11,6) (9,9) (10,8) (11,7) (12,6) ³ (13,5) ³ (10,9) (11,8) (12,7) (10,10) (11,9) (12,8) (13,7) (11,10) (12,9) (13,8) (11,11) (12,10) (13,9) (14,8) (12,11) (13,10) (14,9)
0.81 0.82 0.88 0.89 0.92 0.95 0.95 0.96 1.02 1.02 1.08 1.09 1.10 1.15 1.16 1.17 1.22 1.22 1.23 1.24 1.26 1.29 1.29 1.30 1.36 1.36 1.37 1.38 1.42 1.43 1.44 1.49 1.49 1.50 1.51 1.56 1.56 1.57
488 nm
276 259 252
Theory
152
146 142
a The calculated radial mode frequencies were obtained as vr , 223.75/d. The mode frequencies correspond to (n,m) values which lie within the shaded area marked in the 108 wedge (Fig. 8) with the exception of those indicated by ³.
be used to identify all possible SWNTs with diameter d # 1:65 nm using the standard carbon nanotube notation involving the pair of integers (n,m). The unit cell, tube axis and diameter of an (n,m) tube are all uniquely speci®ed by C. By de®nition, C n; m na 1 mb, where vectors a and b are the primitive lattice vectors of a graphene sheet (see Fig. 8). The solid dots (´) locate the tips of C (n,m) for all `zigzag' (n,0), `armchair' (n,n) and `chiral' (n, m ±n) tubes with d , 1:65 nm. As discussed below, most of the vectors relevant to this work can be assigned to the t108 wedge closest to the so-called `armchair' tube direction (n,n). This 108 wedge angle is consistent with the range of helix or chiral angles reported by Cowley et al. [14] for SWNTs prepared by PLV at T 12008C. The tube diameter d is related to the length of the chiral vector C by the simple relationship d uCu=2p
[3]. Thus, the thin circular arcs in Fig. 8 represent constant tube diameters. Since vr t 1=d, these arcs also represent constant radial frequency arcs. It is interesting to see that, within this 108 wedge, the tubes are organized very nearly on `constant frequency' arcs. Thus, we expect all tubes within this wedge with diameters between that of a (5,5) and a (12,12) tube to exhibit only 15 different radial mode frequencies (see Fig. 8). Furthermore, since the electron± phonon broadening in graphite is t6 cm 21 [16], we do not anticipate resolving the four frequencies, for example, on the (10,10) arc inside the 108 wedge. It should be noted that seven of the ®fteen anticipated frequencies stem from arcs which do not contain an armchair tube, i.e. they must be chiral symmetry tubes. A few experimental frequencies we observe cannot be explained by tubes which fall outside this
A.M. Rao et al. / Thin Solid Films 331 (1998) 141±147
147
Fig. 8. Chiral vector map for carbon nanotubes with diameter d , 1:65 nm. The arcs locate tubes of the same diameter d and theory predicts vr t 1/d. The experimentally observed frequencies are attributed to the eleven resolvable breathing mode frequencies wr for tubes occupying the shaded 108 wedge closest to the armchair (n,n) direction.
108 degree wedge closest to the armchair direction. They are highlighted in the chiral map as open circles. Taken collectively, the data presented in this section show that the mean diameter of nanotubes increases with increasing temperature of the growth environment in the PLV process, and the tubes collect into bundles at all growth temperatures used in this study. A broad range of carbon nanotube diameters with uniform diameter distribution have been observed and the Raman data are interpreted to indicate that they primarily fall in a wedge of the chiral vector map within 108 of the `armchair' direction. Acknowledgements The authors are grateful for the ®nancial support from NSF (OSR-94-52895) and DOE (DE-F22-90PC90029). This paper is based on the results published in Refs. [6] and [10], and those in Ref. [15]. This work represents the fruits of a dynamic, multi-university collaboration between experimentalist and theoreticians.
References [1] S. Iijima, T. Ichihashi, Nature 363 (1993) 603. [2] D.S. Bethune, et al., Nature 363 (1993) 605. [3] M.S. Dresselhaus, G. Dresselhaus, P.C. Eklund, Science of Fullerenes and Carbon Nanotubes, Academic Press, New York, 1996. [4] A. Thess, et al., Science 273 (1996) 483. [5] S. Bandow, A.M. Rao, K.A. Williams, A. Thess, R.E. Smalley, P.C. Eklund, J. Phys. Chem. B. 101 (1997) 8839. [6] A.M. Rao, E. Richter, S. Bandow, et al., Science 275 (1997) 187. [7] R.A. Jishi, et al., Chem. Phys. Lett., 209 (1993) 77. [8] E. Richter, K.R. Subbaswamy, et al., Phys.Rev. Lett., 79 (1993) 2738. [9] W. Hayes, R.A. Loudon, in: Lighting Scattering by Crystals, Wiley, New York, 1978. [10] A.M. Rao, et al., Nature 388 (1997) 257. [11] P.C. Eklund, K.R. Subbaswamy, Phys Rev. B 20 (1979) 5157. [12] P. Zhou, et al., Phys. Rev. B 48 (1979) 5157. [13] P.C. Eklund, et al., Phys. Rev. B 16 (1977) 3330. [14] J.M. Cowley, P. Nikolaev, A. Thess, R.E. Smalley, Chem. Phys. Lett. 265 (1997) 379. [15] S. Bandow, et al., Phys. Rev. Lett. 80 (1998) 3779. [16] M.S. Dresselhaus, G. Dresselhaus, Adv. Phys. 30 (2) (1981) 139.
Raman spectroscopy of pristine and doped single wall carbon nanotubes A.M. Rao a, b, S. Bandow c, E. Richter a, P.C. Eklund a, b,* a
Department of Physics and Astronomy, University of Kentucky, Lexington, KY 40506, USA b Center for Applied Energy Research, University of Kentucky, Lexington, KY 40506, USA c Research Center for Molecular Materials, Institute for Molecular Science, Myodaiji, Okazaki 444, Japan
Abstract Pulsed laser vaporization (PLV) of a heated, Fe/Ni or Co/Ni catalyzed carbon target in Argon gas has been used to synthesize single-wall carbon nanotubes (SWNTs). The Raman spectrum of SWNTs is very different from that observed for graphite (the nanotube's ¯at parent) and these differences can be understood as a result of the cyclic boundary condition imposed on a graphene sheet rolled up to form a seamless nanotube. Optical resonances are observed which are associated with the one-dimensional character of the electronic states of these novel quantum wires. The high frequency tangential mode frequency is signi®cantly upshifted (downshifted) when the SWNT bundles are exposed to acceptor (donor) dopants. q 1998 Elsevier Science S.A. All rights reserved. Keywords: Single wall carbon nanotubes; Raman scattering; Quantum con®nement; Doped single wall carbon nanotubes
1. Introduction Single wall nanotubes (SWNTs) were discovered in 1993 in the carbonaceous by-products from an arc discharge between carbon electrodes in an inert atmosphere [1,2]. They have been observed directly in electron microscopes and in scanning tunneling microscopes. In Fig. 1 the sp 2 based structure of a seamless carbon nanotube is shown schematically. The hexagonal arrangement of C-atoms, identical to that in ¯at carbon sheets of graphite, is apparent. Three different subclasses of seamless nanotubes can be formed: armchair (n,n), zigzag (n,0) and chiral (n,m±n) where the integers n and m are used to de®ne the symmetry of the nanotube [3]. The notation (n,m) de®nes the atomic coordinates for the 1D unit cell of the nanotube. For more details on carbon nanotubes see Ref. [3]. It is of interest to note that `armchair' and `zigzag' tubes have rows of hexagons aligned parallel to the tube axis and `chiral' tubes can be formed such that the hexagon rows spiral up along the tube axis. The tube shown in Fig. 1 is a (10,10) armchair tube. The diameter of an armchair tube is given by d A 1:357n [3], so that a (10,10) nanotube has a diameter of Ê. 13.57 A Until recently, research on the physical properties of these quantum carbon wires have been slowed by the fact that previous arc discharge methods produced mostly carbon soot and only a few percent carbon nanotubes * Corresponding author.
[1,2]. Thus experimental signals in these samples were often dominated by the response of the soot. Recently, a pulsed laser vaporization (PLV) technique was discovered at Rice University which produced over 70% tubes [4]. The tubules are organized into regular hexagonal arrays to form `ropes' or `bundles' and exhibit high aspect ratios (length/ diameter) .10 3. The diameter distribution of the tubules was reported to be sharply peaked at t1.36 nm, close to the diameter predicted for a (10,10) `armchair' tube [4]. Details of the laser-assisted process developed at Rice University are available elsewhere [4]. Brie¯y, a carbon target containing 1±2% Ni/Co catalyst is maintained in ¯owing Ar in an oven (12008C). The carbon target is vaporized by a two-pulse sequence from a YAG and frequencydoubled YAG laser at a 10-Hz repetition rate. Nanotubes (.70%) and carbon nanospheres (t30%) and fullerenes (t1±2%) form in the hot carbon plasma and drift downstream under ¯owing Ar where they are collected on a water-cooled cold ®nger. The carbon material was ®rst soaked in CS2 to remove solubles (i.e. fullerenes) and the insolubles were then subjected to micro®ltration to physically separate the tubes and soot. This was accomplished by dispersing the material in benzalkonium chloride using an ultrasonic bath [5]. In this paper, we wish to explore the nature of these onedimensional (1D) carbon quantum wires using Raman spectroscopy. We will also discuss the doping-induced shifts in the SWNT Raman-active mode frequencies. Finally, we discuss the effect of the catalyst on the tube yield, and the
0040-6090/98/$ - see front matter q 1998 Elsevier Science S.A. All rights reserved. PII S0 040-6090(98)009 10-9
142
A.M. Rao et al. / Thin Solid Films 331 (1998) 141±147
Fig. 1. Schematic representation of an `armchair' (10,10) SWNT. The solid circles represent the carbon atoms.
evolution of the tube diameter distribution with increasing growth environment temperature T. 2. Pristine SWNTs In Fig. 2, we display the Raman spectrum (300 K) for puri®ed single wall nanotubes obtained in the backscattering geometry using 514.5-nm Ar laser radiation [6]. The SWNT samples were synthesized at the Rice University using the PLV method and a growth environment temperature T of 12008C [4]. For comparison, the calculated Raman
Fig. 2. Raman spectrum (top) of SWNT sample taken with 514.5 nm excitation at t2 W/cm 2. The asterisks (*) in the spectrum indicates features that are tentatively assigned to second-order Raman scattering. The four bottom panels are the calculated Raman spectra for armchair (n,n) nanotubes, n 8±11, as obtained within a bond-polarizability model [6]. The downward pointing arrows in the lower panels indicate the positions of the remaining weak, Raman-active modes [6].
Fig. 3. Room temperature Raman spectra for puri®ed SWNTs excited at ®ve different laser frequencies. The laser frequency and power density for each spectrum is indicated, as are the vibrational frequencies [6].
spectrum is shown below for n 8±11 armchair tubes. The frequencies were calculated in a phenomenological force constant model using the same CZC force constants used to ®t vibrational data for a ¯at graphene sheet [7]. The theoretical Raman intensities were calculated using a bond polarizability model by Richter and Subbaswamy [6]. It may be noticed that some of the experimental bands are narrow and some are broad. This difference is attributed to an inhomogeneous line-broadening mechanism based on the theoretical observation that particular vibrational modes exhibit a strong tube diameter dependence, while others exhibit a rather weak dependence. Thus, considering that our sample contains a distribution of tube diameters, the Raman lines can be sharp (similar to a linewidth in graphite t6 cm 21) or broad, depending on whether or not the mode frequency is strongly diameter-dependent. The intense line seen at 186 cm 21 is identi®ed with the radial breathing mode in which all C-atoms are displaced radially outward in phase. The strong lines observed near 1600 cm 21 are related to the intralayer vibrations in graphite which are observed at
A.M. Rao et al. / Thin Solid Films 331 (1998) 141±147
143
tubes and the fact that these different diameter tubes have different optical resonances. In 1D systems it is well known that E 21/2 singularities exist in the electronic density of states (DOS). These singularities manifest themselves as spikes in the DOS calculated for n 8±11 armchair tubes [6,8]. The allowed optical transitions for these tubes can be shown to be between ®lled valence states in spikes (v1,v2) below the Fermi energy level EF to empty conduction band states (c1,c2) above EF [6,8]. Raman scattering from an (n,n) tube will dominate the spectrum when the laser photon energy matches the energy difference between these E 21/2 spikes for that (n,n) DOS. The center of gravity of each Raman band then shifts to the frequency of the (n,n) vibrational mode being resonantly driven by the laser ®eld [6]. Thus, the unusual laser frequency dependence of the Raman spectra of carbon nanotubes is a direct consequence of 1D quantum con®nement effects in carbon nanotubes. Furthermore, calculations within a simple tight binding model for a family of armchair tubes ((8,8) to (11,11)) show that, for a ®xed excitation frequency, the resonance curve-derived from the (v2,c2) electronic transition for smaller diameter armchair tube may overlap with the resonance curve derived from the (v2,c2) electronic transition for relatively larger diameter armchair tubes [8]. Therefore, one should expect a complicated dependence of the peak positions and intensities on the incident excitation frequencies. Fig. 4. Raman scattering spectra for pristine SWNT bundles reacted with various donor and acceptor reagents. From top to bottom: I2, Br2, pristine SWNT, Rb, K. The asterisks (*) indicate the positions of peaks associated with halogen reactant. The peak frequencies indicated in parentheses for the K, Rb doped SWNTs are the renormalized phonon frequencies [10].
1582 cm 21. In the nanotube, the cyclic boundary conditions around the tube waist activate new Raman and IR modes that are not observable in a well-ordered ¯at graphene sheet or in graphite ± this is the ®rst consequence of the onedimensional (1D) nature of the nanotube. The room temperature Raman scattering data on the same sample (Fig. 3) were also collected with four different laser excitation frequencies (647.1, 780, 1064 and 1320 nm) at low power densities [6]. Both the Raman peak intensities and frequencies were observed to change dramatically with laser excitation frequency. Resonantly enhanced Raman scattering occurs when the energy of the incident photon matches the transition energy of a strong optical absorption band [9]. Normally in a resonant Raman scattering process, with variable laser frequency and ®xed incident ¯ux the Raman line intensity can vary by several orders of magnitude, but the Raman line frequency is ®xed. The large shifts in frequency that were also observed is quite unusual [6]. We have identi®ed these large shifts in Raman line frequency with a diameter-dependent optical absorption which promotes resonant scattering from particular diameter tubes [8]. The shifting frequencies are both the result of the sample being a collection of different diameter
3. Doped SWNTs In Fig. 4, we display the T 300 K Raman spectra of pristine SWNTs and those doped with the anticipated acceptors (I2,Br2) and donors (K,Rb). As discussed above for the pristine SWNTs, the peak at 186 cm 21 has been identi®ed with the Ag symmetry radial breathing mode, and the peak at 1593 cm 21has been assigned to an unresolved Raman triplet identi®ed with tangential C-atom displacement modes [6]. For the SWNT bundles reacted with I2 or Br2 (K or Rb) dopants, the Raman active mode frequencies are expected to upshift (downshift) due to the transfer of electrons from (to) the carbon p (p *) states in the tubules to (from) the dopant molecules [10]. Signi®cant shifts in the frequencies of the radial and tangential modes were observed for all dopants indicated in Fig. 4, except for the I2-doped SWNT bundles. This is perhaps not surprising, as I2 does not intercalate into graphite. The low frequency radial breathing mode in K or Rb doped sample is not evident in the spectrum ± it may have downshifted below our lower frequency limit (t100 cm 21), or, alternatively, it may have broadened to the point it cannot be detected above the background. The asymmetric peak at t1565 cm 21 observed in both alkali metal-doped SWNT bundles is identi®ed with a BreitWigner-Fano (BWF) resonance [10] which usually involves an interference between Raman scattering from continuum excitations and that from a discrete phonon. The degree of coupling (1/q) determines the departure of the lineshape
144
A.M. Rao et al. / Thin Solid Films 331 (1998) 141±147
breakdown in the pristine nanotube selection rules associated with the doping. The linewidths are also very large, suggesting broadening through the electron phonon interaction or through inhomogeneous doping. It should also be mentioned that BWF lineshapes have been observed previously in doped C60 (K3C60, Rb3C60 [12]). The value of 1/q t20.35 we observe here for the M-doped SWNT bundles is about a factor of three smaller than those observed for the broad BWF resonance in C8M compounds [13] and similar values reported for the low frequency Hg symmetry, intramolecular vibrations in M3C60 [12]. Using the charge-transfer- induced shifts in the E2g2 high frequency intralayer mode in the H2SO4 doped GICs, we ®nd f t 1=19 (f is the transferred charge per C-atom), or one free hole per 19 C-atoms in the bromine saturation doped SWNTs [10]. For the alkali metal dopants (K,Rb), the two contributions (i.e. charge transfer and interference (1/q)) to the BWF frequency v 0 cannot be separated without a detailed analysis of the continuum excitations. Therefore, for the alkali metal doped SWNT bundles we cannot estimate f at this time.
Fig. 5. High frequency Raman spectrum for K and Rb doped SWNT bundles (T 300 K; excitation 514.5 nm). The spectra are ®t with a superposition of Lorentzians and an asymmetric Breit-Wigner-Fano lineshape on a linear continuum and the dashed lines represent the individual contributions to the spectrum. In both cases, the calculated spectrum nearly superimposes on the data. Peak frequencies are indicated; the values in parentheses are the renormalized phonon frequencies.
from a symmetric Lorentzian. It should be noted that when the coupling is signi®cant, the BWF peak position can be considerably displaced from the value of the uncoupled discrete phonon [11]. This is referred to as a renormalization of the `bare' phonon frequency. If the continuum scattering in the SWNTs is electronic in origin, then the degree of frequency renormalization can be an important measure of the electron±phonon interaction. The BWF lineshape is given by: I v I0 {1 1 v 2 v0 =qG}2 ={1 1 v 2 v0 =G2 }
(1)
where I0, v 0, and G are, respectively, the intensity, renormalized frequency and broadening parameter. In Fig. 5, we show a ®t to the high frequency experimental spectra for Rband K- doped SWNTs. The dashed lines represent the individual Lorentzians used for the peaks at t949, 1145, 1222, 1263 and 1329 cm 21 and the BWF lineshape with its peak at 1567 cm 21. The Lorentzian peaks are tentatively assigned to charge-transfer-softened SWNT modes predicted in theoretical calculations [8]. However, the doping-induced increase in the Raman activity of the mid frequency (650±1350 cm 21) phonons is not understood and may be related to a
4. Effect of the growth temperature on the diameter distribution and chirality of single wall carbon nanotubes Finally, we discuss the in¯uence of the laser target and tube growth environment temperature (T) on the diameter and symmetry of SWNTs produced by the PLV method at the Institute for Molecular Science, Japan. A two-pulse sequence from a Nd/YAG laser (532-nm radiation followed by 1064-nm radiation at 20 Hz and coaxial with the 532-nm radiation) were focused in a 3-mm diameter spot on carbon targets containing either Fe/Ni (0.6:0.6 at.%) or Co/Ni (2.6:2.6 at.%). Each target was placed in a quartz tube near the center of a 30-cm tube furnace. Flowing Ar gas (100 sccm, 500 Torr) was introduced at the front of the furnace that carried the SWNTs and other carbon products produced in the carbon plasma to the rear of the furnace where they were collected just outside on a water-cooled Cu cold ®nger. The average temperature near the carbon target and growth environment (T) was controlled to within ^18C. From electron microscopy, X-ray diffraction and Raman experiments, we observe a systematic increase in the average tube diameter with increasing growth temperature above a threshold temperature T0 t 8508C for tube production. The same general trend was observed for the target containing the two different bi-metallic catalysts (Fe/Ni and Co/Ni), i.e. lower T values promote the growth of smaller diameter tubules in smaller diameter bundles. As discussed below, the diameter dependence of the frequency of the Ramanactive breathing mode are consistent with the production of bundles of tubules with chiral angles u with less than t108
A.M. Rao et al. / Thin Solid Films 331 (1998) 141±147
145
frequency (v r) vs. the inverse tube diameter (1/d) for all symmetry types, i.e. (n,n), (n,0), and (n,m). The calculated values can be found in Ref. [15] for comparison to experimental frequencies. As can be seen from Fig. 6, the calculated frequencies all fall on a straight `universal' line well ®t by the function:
vr 223:75 cm21 ´nm=d nm
Fig. 6. Calculated radial frequency v r vs. inverse tube diameter (1/d) for armchair (W), zigzag (K) and chiral (L) symmetry tubes (Eq. (2)).
of the (n,n) or `armchair' direction, similar to the range of angles reported earlier by Cowley et al. [14] for PLV grown tubes synthesized at T 12008C by the Rice University group. In Fig. 6, we plot the calculated radial breathing mode
Fig. 7. Low frequency, room temperature Raman spectra for nanotubes produced using Fe/Ni catalyst at the temperatures and excitations indicated in the ®gure. In (a±d), the shift of intensity towards lower frequency radial modes with increasing T signals the increased production of larger diameter tubes (see text).
(2)
As anticipated, for d 1, vr 0, as this frequency corresponds to the q 0 transverse acoustic (TA) phonon of a ¯at graphene sheet and this mode thereby corresponds to a rigid displacement of the sheet normal to its surface. Our calculations show clearly that v r is only sensitive to inverse diameter and is not sensitive to the helicity (or symmetry) of the particular tube. Thus, this Raman-active radial mode, which can exhibit quite a large resonant Raman cross section [6,8], is ideal for determining the diameter of SWNTs. In Fig. 7, we display the low frequency region of the Raman spectrum for SWNTs where the resonantly enhanced lines associated with the radial breathing modes of various diameter SWNTs are expected. The data were taken at 300 K on tubes grown at T 1000, 920, 860, and 7808C using an Fe/Ni catalyzed target. The spectra are unpolarized and were collected in the backscattering con®guration. The bottom four spectra (Fig. 7a±d) were excited using Nd/ YAG laser radiation (1064 nm) and represent the change in the radial mode scattering with increasing temperature (bottom to top). The upper four spectra (Fig. 7d±g) were all taken on the SWNT sample grown at 10008C but at different excitation wavelengths (1064, 647, 514.5, and 488 nm). These spectra were chosen to show the strong resonance effects (i.e. peak shifting and intensity change) as discussed in Ref. [6]. As the data in Fig. 7 indicate, to ®nd all the radial mode frequencies associated with a given sample, it is necessary to use several excitation wavelengths. All the peak positions were obtained from a Lorentzian lineshape analysis (for example, see Fig. 7d) were recorded for tubes produced both for the Fe/Ni and the Co/Ni catalyzed targets [15]. They have all been identi®ed with the resonantly enhanced radial breathing mode for tubes with differing diameter (see Table 1) [15]. Other lower symmetry, low frequency modes exhibit a signi®cantly weaker Raman scattering cross-section [6]. From the calculated relation vr t 1/d, it is evident that smaller (larger) diameter tubes should exhibit higher (lower) radial mode frequencies. Thus the evolution of the spectra in Fig. 7a±d, i.e. the shift of intensity towards lower frequency radial modes with increasing T, signals the increased production of larger diameter tubes, consistent with the Xray diffraction (XRD) and transmission electron microscopy (TEM) data (see Ref. [15]). A similar general trend is observed with other excitation wavelengths and also on SWNTs prepared from Co/Ni catalyzed targets. In Fig. 8, we display a map of chiral vectors C which can
146
A.M. Rao et al. / Thin Solid Films 331 (1998) 141±147
Table 1 Experimentally determined radial mode frequencies for SWNTs produced using the Fe/Ni (indicated by (*)) and the Co/Ni catalyzed targets a Experiment 1064 nm
647.1 nm
514 nm
230 213 204 190
177
258* 232* 230
222, 221* 215, 214*
224*, 222
222, 221*
206* 206
203 203* 193 192*
186* 185
182
179* 174, 174*
180*, 178 178* 174, 173*
167, 167*
167*, 166
167, 166*
163, 163*
163*
163, 163*
180 179* 171, 170*
160
257 255* 242, 240*
231* 232
192, 192* 189
157 150
145
Notation
Diameter (nm)
275.0 273.7 253.6 251.6 242.4 235.7 234.9 232.5 219.8 218.5 206.2 205.7 204.1 194.0 193.1 191.3 183.3 182.9 181.8 178.0 177.6 173.6 172.9 171.7 165.0 164.7 163.9 162.6 157.1 156.6 155.7 150.0 149.8 149.2 148.2 143.4 142.9 142.2
(6,6) (7,5) (7,6) (8,5) (10,3) ³ (7,7) (8,6) (9,5) (8,7) (9,6) (8,8) (9,7) (10,6) (9,8) (10,7) (11,6) (9,9) (10,8) (11,7) (12,6) ³ (13,5) ³ (10,9) (11,8) (12,7) (10,10) (11,9) (12,8) (13,7) (11,10) (12,9) (13,8) (11,11) (12,10) (13,9) (14,8) (12,11) (13,10) (14,9)
0.81 0.82 0.88 0.89 0.92 0.95 0.95 0.96 1.02 1.02 1.08 1.09 1.10 1.15 1.16 1.17 1.22 1.22 1.23 1.24 1.26 1.29 1.29 1.30 1.36 1.36 1.37 1.38 1.42 1.43 1.44 1.49 1.49 1.50 1.51 1.56 1.56 1.57
488 nm
276 259 252
Theory
152
146 142
a The calculated radial mode frequencies were obtained as vr , 223.75/d. The mode frequencies correspond to (n,m) values which lie within the shaded area marked in the 108 wedge (Fig. 8) with the exception of those indicated by ³.
be used to identify all possible SWNTs with diameter d # 1:65 nm using the standard carbon nanotube notation involving the pair of integers (n,m). The unit cell, tube axis and diameter of an (n,m) tube are all uniquely speci®ed by C. By de®nition, C n; m na 1 mb, where vectors a and b are the primitive lattice vectors of a graphene sheet (see Fig. 8). The solid dots (´) locate the tips of C (n,m) for all `zigzag' (n,0), `armchair' (n,n) and `chiral' (n, m ±n) tubes with d , 1:65 nm. As discussed below, most of the vectors relevant to this work can be assigned to the t108 wedge closest to the so-called `armchair' tube direction (n,n). This 108 wedge angle is consistent with the range of helix or chiral angles reported by Cowley et al. [14] for SWNTs prepared by PLV at T 12008C. The tube diameter d is related to the length of the chiral vector C by the simple relationship d uCu=2p
[3]. Thus, the thin circular arcs in Fig. 8 represent constant tube diameters. Since vr t 1=d, these arcs also represent constant radial frequency arcs. It is interesting to see that, within this 108 wedge, the tubes are organized very nearly on `constant frequency' arcs. Thus, we expect all tubes within this wedge with diameters between that of a (5,5) and a (12,12) tube to exhibit only 15 different radial mode frequencies (see Fig. 8). Furthermore, since the electron± phonon broadening in graphite is t6 cm 21 [16], we do not anticipate resolving the four frequencies, for example, on the (10,10) arc inside the 108 wedge. It should be noted that seven of the ®fteen anticipated frequencies stem from arcs which do not contain an armchair tube, i.e. they must be chiral symmetry tubes. A few experimental frequencies we observe cannot be explained by tubes which fall outside this
A.M. Rao et al. / Thin Solid Films 331 (1998) 141±147
147
Fig. 8. Chiral vector map for carbon nanotubes with diameter d , 1:65 nm. The arcs locate tubes of the same diameter d and theory predicts vr t 1/d. The experimentally observed frequencies are attributed to the eleven resolvable breathing mode frequencies wr for tubes occupying the shaded 108 wedge closest to the armchair (n,n) direction.
108 degree wedge closest to the armchair direction. They are highlighted in the chiral map as open circles. Taken collectively, the data presented in this section show that the mean diameter of nanotubes increases with increasing temperature of the growth environment in the PLV process, and the tubes collect into bundles at all growth temperatures used in this study. A broad range of carbon nanotube diameters with uniform diameter distribution have been observed and the Raman data are interpreted to indicate that they primarily fall in a wedge of the chiral vector map within 108 of the `armchair' direction. Acknowledgements The authors are grateful for the ®nancial support from NSF (OSR-94-52895) and DOE (DE-F22-90PC90029). This paper is based on the results published in Refs. [6] and [10], and those in Ref. [15]. This work represents the fruits of a dynamic, multi-university collaboration between experimentalist and theoreticians.
References [1] S. Iijima, T. Ichihashi, Nature 363 (1993) 603. [2] D.S. Bethune, et al., Nature 363 (1993) 605. [3] M.S. Dresselhaus, G. Dresselhaus, P.C. Eklund, Science of Fullerenes and Carbon Nanotubes, Academic Press, New York, 1996. [4] A. Thess, et al., Science 273 (1996) 483. [5] S. Bandow, A.M. Rao, K.A. Williams, A. Thess, R.E. Smalley, P.C. Eklund, J. Phys. Chem. B. 101 (1997) 8839. [6] A.M. Rao, E. Richter, S. Bandow, et al., Science 275 (1997) 187. [7] R.A. Jishi, et al., Chem. Phys. Lett., 209 (1993) 77. [8] E. Richter, K.R. Subbaswamy, et al., Phys.Rev. Lett., 79 (1993) 2738. [9] W. Hayes, R.A. Loudon, in: Lighting Scattering by Crystals, Wiley, New York, 1978. [10] A.M. Rao, et al., Nature 388 (1997) 257. [11] P.C. Eklund, K.R. Subbaswamy, Phys Rev. B 20 (1979) 5157. [12] P. Zhou, et al., Phys. Rev. B 48 (1979) 5157. [13] P.C. Eklund, et al., Phys. Rev. B 16 (1977) 3330. [14] J.M. Cowley, P. Nikolaev, A. Thess, R.E. Smalley, Chem. Phys. Lett. 265 (1997) 379. [15] S. Bandow, et al., Phys. Rev. Lett. 80 (1998) 3779. [16] M.S. Dresselhaus, G. Dresselhaus, Adv. Phys. 30 (2) (1981) 139.