Electronic properties of Cs-intercalated single-walled carbon nanotubes derived from nuclear magnetic resonance

We report on the electronic properties of Cs-intercalated single-walled carbon nanotubes (SWNTs). A detailed analysis of the 13C and 133Cs nuclear magnetic resonance (NMR) spectra reveals an increased metallization of the pristine SWNTs under Cs intercalation. The ‘metallization’ of CsxC materials where x=0–0.144 is evidenced from the increased local electronic density of states (DOS) n(EF) at the Fermi level of the SWNTs as determined from spin–lattice relaxation measurements. In particular, there are two distinct electronic phases called α and β and the transition between these occurs around x=0.05. The electronic DOS at the Fermi level increases monotonically at low intercalation levels x<0.05 (α-phase), whereas it reaches a plateau in the range 0.05⩽x⩽0.143 at high intercalation levels (β-phase). The new β-phase is accompanied by a hybridization of Cs(6s) orbitals with C(sp2) orbitals of the SWNTs. In both phases, two types of metallic nanotubes are found with a low and a high local n(EF), corresponding to different local electronic band structures of the SWNTs.


2
The electronic properties of single-walled carbon nanotubes (SWNTs) have attracted a great deal of interest in the last decade. Non-intercalated SWNTs have been shown both theoretically [1,2] and experimentally [3] to exhibit either metallic or semi-conducting properties depending on their chirality and diameter. Tuning the electronic density of states (DOS) is of paramount scientific and technological interest. One of the ways to proceed is to intercalate nanotubes with alkali metals, which is known to change the electronic properties of the nanotubes due to electron transfer from the alkali atoms to the nanotubes. Despite the progress made in the last decade, there are still many open questions concerning this transfer. A specific question concerning the charge transfer is if under some conditions the alkali metal does not fully transfer its valence electrons to the nanotube carbons [4,5]. In this systematic intercalation study, we have applied 13 C and 133 Cs nuclear magnetic resonance (NMR) to investigate seven samples of 13 C-enriched (10%) Cs-doped SWNTs (Cs x C, x = 0-0.143). We show that doped SWNTs exhibit two distinct electronic phases called α and β and that the transition between these phases occurs around x = 0.05. The electronic DOS, n(E F ), at the Fermi level increases monotonically at low intercalation level x < 0.05 (α-phase), whereas it reaches a plateau in the range 0.05 x 0.143 at high intercalation level (β-phase). We present strong evidence that this new β-phase is accompanied by a hybridization of Cs(6s) orbitals with the C(sp 2 ) orbitals of the SWNTs.
Raw SWNT samples were synthesized by arc-discharge using Pt and Rh as catalysts as described in [6]. The SWNTs had an average diameter of 1.4 ± 0.2 nm and a bundle length of several micrometers [7,8]. All SWNTs were 13 C isotope enriched to 10% in order to increase the signal-to-noise ratio in the same way as previously reported by Ruemelli et al [9]. The preparation of Cs-intercalated SWNTs proceeded by saturating SWNTs by Cs vaportransport onto degassed SWNT at 200 • C in a cycled process for 10 days. Theoretical and experimental studies have shown that the intercalation of SWNT saturates at a stoichiometry of approximately Cs 0.14 C [10][11][12]. The enriched saturated sample was then mixed in different ratios with the pristine SWNT sample by a Schlenk-like method. All samples were treated under secondary vacuum. Before the samples were mixed and annealed at 300 • C for 2 weeks the mixtures were analyzed with static NMR. The different line shapes of the pristine and saturated intercalated SWNTs reflecting the difference in chemical and Knight shift anisotropy is clear from figure 1(a). By extracting the contribution of the two different line shapes (pristine and saturated) in the different mixtures before annealing, the ratio of non-intercalated SWNTs and saturated intercalated SWNTs could be estimated. After the heat treatment the samples were again controlled by static NMR. The change in line shape of the samples ( figure 1(b)) confirms the homogeneity of the samples with stoichiometries Cs x C (x = 0.028-0.14) (errors ≈ 4-8%). 13 C NMR experiments were performed on Bruker and Tecmag Apollo NMR solid state spectrometers at a magnetic field of 4.7 T and a Larmor frequency of 50.3 MHz. Static NMR spectra were obtained in the temperature range of 5-300 K. 13 C NMR line shifts are referred to tetramethylsilane (TMS). 133 Cs NMR measurement were carried out on a home-built pulsed NMR spectrometer equipped with a sweep magnet working at 9.1 T and a Larmor frequency of 50.7 MHz. 133 Cs NMR line shifts are referred to an aqueous solution of 1 M CsNO 3 . 13 C spin-lattice relaxations were obtained using a saturation recovery pulse technique and a Hahn echo as the detection sequence. Figure 1(b) displays the static 13 C-NMR spectra of pristine and Cs-intercalated SWNTs. The spectrum of pristine SWNTs shows a characteristic sp 2 powder line shape in agreement with earlier studies [8]. However, already at low intercalation levels, x = 0.028-0.038, the 3 (a) (b) Figure 1. (a) Static 13 C NMR spectra at room temperature before annealing. 'Pristine': pristine SWNTs; 'I)' and 'II)': mixture of saturated Cs 0.14 C and pristine SWNTs; and 'Saturated': saturated Cs 0.14 C. Fitted tensors are shown to explain how the doping ratio was established. Spectra I and II were fitted with both types of tensors (giving a doping ratio of 0.033 ± 0.003 and 0.091 ± 0.005, respectively). The fits were performed by following the procedure described in [8,18]. (b) Static 13 C NMR spectra at room temperature after annealing the samples at 300 • C for 2 weeks. Spectra show pristine SWNTs and Cs x C intercalated SWNTs for different intercalation levels x increasing from the bottom to the top as indicated in the figure.
NMR spectra exhibit a clear anisotropy reduction of about 200 ppm due to a change into a more metallic-like shift anisotropy as reported by Latil et al [13]. All samples show a similar and strong further narrowing of the line shape when the intercalation level is increased from x = 0.05 to 0.14, indicating that the samples exhibit an increasing trend in metallic behavior. There are two known contributions to the total 13 C NMR shift of doped SWNTs, namely the Knight shift K and the chemical shift σ , where the Knight shift arises from the hyperfine coupling of the nuclei to conduction electron spins and the chemical shift summarizes all contributions from orbital currents due to local magnetic fields. Both Knight shift K and T 1 relaxation are affected by the DOS at the Fermi surface n(E F ) if shift and relaxation are dominated by the magnetic interaction with the conduction electrons of the metal [14]. In carbon nanotubes as well as in fullerenes and other anisotropic and complex conducting materials the simple classical relations applied to classical metals do not hold. Instead, the anisotropy of the hyperfine interaction as well as non-classical electrodynamics need to be considered. It is beyond the scope of this paper to review these theoretical aspects. We therefore just refer to the results of earlier publications [15][16][17]. A review of these topics can be found in [18,19].
The relaxation rate can be expressed as , is the anisotropy parameter with dipolar contribution d corresponding to the anisotropic but axially symmetric hyperfine interaction and the other symbols have their usual meaning. The values of A iso = 5.8 × 10 7 s −1 and A dip = 1.1 × 10 7 s −1 leading to the value of ε = 3.9 are derived from the work of Pennington and Stenger [18] for sp 2 -bonded carbon materials. There is a compact notation of the relation between the isotropic Knight shift and the relaxation time as [17] with and S K = 1 for uncorrelated electrons as considered here. In this communication we evaluate only the relaxation rate 1/T 1 and its proportionality to n(E F ) 2 , a technique that has been shown to be appropriate to estimate n(E F ) [20][21][22][23].
The observed relaxation curves signal a two-component model with one slow relaxing and one fast relaxing component, which can be rationalized by the two types of SWNTs to be discussed in the following. The magnetization recovery, M(t), was correspondingly fitted with a double exponential function of the type (inset of figure 2(a) at room temperature) where M s and M m are the equilibrium magnetizations for the two components. T 1s and T 1m are the two different T 1 relaxation times for the two components. The relative magnitudes, M s /M 0 and M m /M 0 , with M 0 = M s + M m do not differ in a systematic way for the different stoichiometries, and are determined to be 60% ± 7% and 40% ± 7%. The errors of the T 1s and T 1m values vary around 10% for different temperatures. For the pristine SWNTs, the slow and the fast relaxing components derived from the fit at room temperature result in T 1s = 75 s and T 1m = 2 s and were assigned to semiconducting and metallic tubes, respectively [8,22]. The two components of SWNT samples are also in agreement with recent high-resolution NMR experiments [24]. On doping we observe a decrease of both relaxation times, T 1s and T 1m , as expected due to the increase of the DOS of both the metallic and semiconducting tubes. They reach comparable values to the one for the pristine metallic SWNT (e.g. Cs 0.028 C T 1s = 9 s and T 1m = 1.3 s). With increased intercalation, both relaxation rates 1/T 1s and 1/T 1m increase until they 'saturate' (T 1s = 2.2 s and T 1m = 0.25 s) for samples with x = 0.05-0.14. In our case, we assign the two components to two kinds of metallic tubes (α l -and α h -type) obtained by doping the originally semiconducting and metallic SWNTs. Recent NMR studies on pristine SWNTs [25] and double-walled carbon nanotubes (DWNTs) [23,26] have used partly different models based on a stretched distribution of T 1 values around one average. In studies of the DWNTs this is rationalized by a purely metallic state of the inner tubes in the DWNTs according to the authors. For all our samples with different doping levels two components of T 1 values are present. It should, however, be noted that the twocomponent model adopted here is a simplified picture that enables us to estimate the DOS of the SWNTs. In a more complicated view, the heterogeneity of the SWNTs would require one to use a distribution of T 1 values around T 1s and T 1m . The previous use of the two-component model in similar studies [8,22] and regarding T 1s and T 1m as representing the average values of the two distributions in our opinion justifies the use of the simplified model. 6 Using equation (1) that takes into account both isotropic and anisotropic hyperfine interactions in contrast to earlier studies of non-intercalated SWNTs [8,22] and doped C 60 [27,28], we can determine the DOS at the Fermi level, n(E F ), for our samples ( figure 2(b)). As a direct consequence of the Korringa law ( figure 2(a)) the two types of tubes with long and short relaxation times are calculated to have a 'low' DOS (α l -type tubes, 60%) or a high DOS (α h -type tubes, 40%), respectively. The values of n(E F ) were derived using the relaxation times determined at room temperature (figure 2). Low-temperature data, however, support well the derived values shown in figure 2(b). Our derived value for n(E F ) is significantly higher than the corresponding value for non-intercalated tubes in the paper by Tang et al [22] but with a very similar DOS ratio between the high-and low-DOS tubes. Referring to the discussion above about the necessity of taking also anisotropic hyperfine interactions into account for carbon nanotubes, we do believe that our derived values are more realistic but that the absolute values of n(E F ) largely depend on the parameters a and d.
Note that n(E F ) does not follow a simple linear dependence but increases steadily to an intercalation level corresponding to x = 0.05 where it starts to flatten out. This is especially clear for tubes with a low DOS. Our observations are in excellent agreement with Kazaoui et al [29] and Grigorian et al [30] who observed a maximum in the conductivity of Csintercalated SWNTs for similar intercalation levels. The behavior of the DOS observed here as well as the conduction plateau observed by Kazaoui et al [29] and Grigorian et al [30] could indicate a different charge transfer behavior of Cs atoms added in the high intercalation regime. To further explore this charge transfer behavior, we have performed 133 Cs NMR experiments as shown in figure 3. At low intercalation levels, a single quadrupolar broadened line located at 0 ppm is observed which we assign to Cs(α). At higher intercalation levels above x = 0.038, a second line located around 2700 ppm appears and is assigned to a different type of Cs ion denoted Cs(β) which increases progressively with Cs concentration.
The Cs(α) line position around 0 ppm is a signature of almost fully ionized Cs +1 cations (Cs 0 metal is found at 14600 ppm), whereas the strong paramagnetic shift of the Cs(β) line indicates the presence of Cs +δ cations with approximately δ = 1 − (2700/14600) = 0.82, which signals some hybridization with the carbon orbitals [31]. The two lines at high Cs concentration may be explained by the existence of two non-equivalent Cs adsorption sites in the SWNT bundle as proposed by Bendiab et al [32] and Bantignies et al [33]. Here we favor, however, two different intercalation phases, namely the α-phase corresponding to Cs(α) and the β-phase corresponding to Cs(β) in comparison to the stage I Cs-GIC compound CsC 8 [30]. We point out, however, that both these phases contain two types of tubes as revealed by the relaxation experiments.
The paramagnetically shifted Cs(β) line implies a strong hyperfine coupling between Cs nuclei and the conduction electrons of the carbon nanotubes arising from the strong Fermi contact-like electrons-nuclear interaction with the hybridized Cs(6s)-C(sp 2 ) orbitals. This is supported by a strong Korringa-like relaxation of the Cs(β) line (not shown here). Moreover, a well-developed quadrupolar interaction of Cs(β) that does not vanish at higher temperatures suggests a rather rigid bonding of Cs(β) in contrast to Cs(α) that show large mobility at higher temperatures.
A limited charge transfer was also proposed by Lu et al [4] who found a charge transfer limitation in K-intercalated nanotubes theoretically. Above a stoichiometry of K 0.037 C, the authors claim a preferential hybridization between K (4s) orbital and nearly free electron states of the nanotube, leading to a limit in charge transfer. Such hybridization was also proposed theoretically by Miyake and Saito [5]. Their calculated effect is expected to be even more significant for Cs(β) due to the higher hybridization affinity of cesium. This might in fact explain our observation of the plateau in the DOS n(F F ) of carbon nanotubes in the region from Cs 0.05 C to Cs 0.14 C as well as the conductivity plateau observed by Kazaoui et al [29].
We have shown that intercalated SWNTs have a complex metallic ground state, which is tunable by the Cs content. The T 1 -rates of all samples follow a linear relation in agreement with Korringa's law. At low intercalation levels the local electronic DOS at the Fermi level increases consistently up to x = 0.05 but flattens out for higher intercalation levels. At high intercalation levels the spectra indicate a transition into a two-phase intercalation model, which we label as αand β-phase. Static 133 Cs NMR measurements affirm this model and show that at high intercalation levels a certain fraction of the Cs atoms does not take part in the full charge transfer to the SWNT carbons but instead leads to a hybridization between the valence electrons of the Cs atoms and the valence electrons of the nanotubes.
Our study confirms and explains several recent theoretical and experimental studies on alkali-metal-intercalated carbon nanotubes. This study proves the versatility of NMR to extract significant information on the metallic and local properties of intercalated carbon nanotubes. Moreover, this study provides important additional knowledge about the variety of electronic states of carbon nanotubes, which is of significant importance for future applications of SWNTs.