1: \documentclass{elsart}
2: \usepackage{amssymb}
3: \usepackage{epsfig}
4: \newcommand{\lapx}{\mbox{\raisebox{-4pt}{$\,\buildrel<\over\sim\,$}}}
5: \newcommand{\gapx}{\mbox{\raisebox{-4pt}{$\,\buildrel>\over\sim\,$}}}
6:
7: \begin{document}
8:
9: \begin{frontmatter}
10: \title{Intrinsic Coulomb blockade in multi-wall carbon nanotubes}
11: \author[l1]{R. Egger}
12: \author[l2]{, A.O. Gogolin}
13: \address[l1]{Institut f\"ur Theoretische Physik IV,
14: Heinrich-Heine-Universit\"at, \\ D-40225 D\"usseldorf, Germany}
15: \address[l2]{Department of Mathematics, Imperial College, 180 Queen's Gate,\\
16: London SW7 2BZ, United Kingdom}
17: \begin{abstract}
18: Carbon nanotubes provide a new class of molecular wires that
19: display new and exciting mesoscopic transport properties.
20: We provide a detailed theoretical description for transport in
21: multi-wall nanotubes, where both disorder and strong interactions
22: are important. The interplay of both aspects leads to a particularly effective
23: intrinsic Coulomb blockade for tunneling. The relation to recent
24: experiments is discussed.
25: \end{abstract}
26:
27: \begin{keyword}
28: Carbon nanotubes \sep Coulomb blockade \sep Quantum wires
29: \PACS 71.10.-w \sep 71.20.Tx \sep 72.80.Rj
30: \end{keyword}
31:
32: \end{frontmatter}
33:
34: \section{Introduction}
35: \label{Intro}
36:
37: Transport in molecular wires has received a lot of
38: attention during the past decade. Besides
39: fundamental interest, much of the relevance of this field comes
40: from potential applications in the realm of
41: molecular electronics \cite{joachim}.
42: In this article, we will focus on one specific class
43: of molecular wires, namely {\sl carbon nanotubes} \cite{dekker99}.
44: Nanotubes provide a remarkable and exciting arena for
45: mesoscopic transport phenomena involving strong
46: electron correlations.
47: The primary quantity theoretically analyzed below
48: is the energy-dependent tunneling density of states (TDOS)
49: for tunneling into the nanotube,
50: $\nu(E) \sim \,{\rm Re}\,\int_0^\infty dt \, e^{iEt}
51: \langle \psi^{}(t)\psi^\dagger(0)\rangle$,
52: where energies are measured relative to the Fermi level $E_F$
53: (we put $\hbar=1$).
54: The energy dependence
55: of the TDOS directly governs the (nonlinear) conductance
56: of a nanotube connected to an STM tip or to metallic leads
57: via bad contacts (tunnel junctions). It also determines
58: the intrinsic conductance in the presence of strong
59: impurity backscattering or weak links.
60: From a more fundamental point of view, the TDOS
61: provides precious information about the importance
62: of Coulomb interactions and electronic correlations
63: in such a molecular wire.
64: Recent experiments on individually contacted single-wall nanotubes (SWNTs),
65: which are composed of a single wrapped graphite sheet, have
66: convincingly established the {\sl ballistic} (essentially defect-free)
67: nature of electronic transport in SWNTs over distances of the
68: order 1 $\mu$m and beyond \cite{nt1,nt2}.
69: Due to their small radius $R\approx 1$ to 2 nm, SWNTs are
70: characterized by strong transverse momentum quantization,
71: and under normal circumstances only two
72: spin-degenerate transport bands
73: are present at the Fermi level (assuming a metallic SWNT).
74: In such a one-dimensional (1D) conductor, the electron-electron
75: interactions are expected to be of crucial importance.
76: In fact, the field-theoretical analysis \cite{egger97,kane97}
77: predicted the breakdown of conventional Fermi-liquid theory.
78: On not too low energy scales, in practice meaning that the temperature
79: should be above the sub-mK regime, a Luttinger liquid
80: (LL) phase should emerge in SWNTs. The LL is the generic phase of interacting
81: 1D electrons, and is characterized by the absence
82: of Landau quasiparticles, implying a smeared Fermi surface.
83: In addition, a LL exhibits
84: spin-charge separation, electron fractionalization, and
85: anomalous transport properties \cite{ll2}.
86: Recent SWNT experiments \cite{ll1,ll1a,ll1b,ll1c,ll1d} reported
87: clear evidence for the elusive LL behavior
88: of 1D interacting fermions.
89: These charge transport experiments have measured the
90: TDOS for tunneling into the SWNT. According to LL theory,
91: the energy dependence of the TDOS is a power law,
92: with an exponent that explicitly depends on whether one tunnels into the end
93: or into the middle of the SWNT. The predicted exponents
94: have by now been observed experimentally with good
95: precision \cite{ll1a,ll1b,ll1c,ll1d}.
96: For a review on the status of theory and experiment regarding electronic
97: correlation effects in SWNTs, see Ref.~\cite{review}.
98:
99: The situation is more complex and controversial for
100: multi-wall nanotubes (MWNTs),
101: which are the focus of this article.
102: Existing experimental observations for MWNTs
103: do not seem to easily fit into the framework of well-established theories
104: for disordered electrons. The reason appears
105: to be linked to the presence of strong electron-electron interactions
106: as will be exemplified below for
107: the case of the TDOS.
108: The structure of this article is as follows.
109: In Sec.~\ref{sec2}, we summarize basic transport
110: properties of MWNTs and the experimental situation.
111: In Sec.~\ref{sec3}, phenomenological Coulomb blockade
112: theory is reviewed, which is then given a microscopic
113: justification based on
114: a nonlinear $\sigma$ model calculation.
115: The theoretical predictions
116: for the TDOS, in particular the numerical solution of the Coulomb blockade
117: equations is discussed in Sec.~\ref{sec4}.
118: Finally, some conclusions can be found in Sec.~\ref{sec5}.
119:
120: \section{Electronic transport in MWNTs}
121: \label{sec2}
122:
123: Multi-wall nanotubes
124: are composed of several (typically ten) concentrically arranged
125: graphite shells, with outermost radius of the order $R\approx 5$ to 10~nm
126: and length $L$ up to several 100~$\mu$m. It is rather obvious that
127: the main difference between MWNTs and SWNTs, apart from
128: the larger radii of MWNTs, should come from the presence of inner shells.
129: Because of the large radius,
130: the transverse quantization energy is only $v_F/R\approx 0.2$~eV,
131: where the Fermi velocity is $v_F=8\times 10^5$ m/sec,
132: and therefore one needs to be careful about the position of the
133: Fermi level. While in an undoped MWNT, electron-hole
134: symmetry enforces $E_F=0$, in basically all tubes studied so
135: far a rather strong doping effect was present,
136: $|E_F|\approx 0.3$ to $0.5$~eV \cite{sc4}.
137: The physical origin of the doping is largely open,
138: but may be a result of charge transfer from oxygen or from the substrate
139: or the attached leads. This means in practice that
140: typically 20 spin-degenerate subbands are present
141: (instead of only two as in SWNTs),
142: and therefore doped MWNTs correspond to {\sl multi-channel
143: molecular quantum wires}. Looking for the moment only at the outermost shell,
144: the bandstructure of a perfect (clean) tube corresponds to a
145: Dirac ``light cone'', $E(\vec{k})=v_F |\vec{k}|$, around each of
146: the two gapless K points \cite{dekker99},
147: with $\vec{k}=(k,k_\perp)$ and quantized transverse momentum, $k_\perp=n/R$,
148: where $n=-N,\ldots,N$ and $N=[k_F R]$.
149: The number $M=2N+1 \approx 10$ of spin-degenerate 1D subbands at each
150: K point arising from periodic
151: boundary conditions around the circumference is
152: determined by the doping (Fermi) level via $k_F=|E_F|/v_F$.
153: The $n$th subband is then characterized by a separate Fermi velocity,
154: $v_n=v_F \sqrt{1-(n/k_F R)^2}$,
155: and Fermi momentum, $k_n = k_F v_n/v_F$.
156: For clarity, we focus on doped MWNTs throughout this article,
157: where $M\gg 1$.
158:
159: With one exception \cite{frank}, available experiments agree that
160: electronic transport in MWNTs is not ballistic (as in SWNTs) but
161: {\sl diffusive} \cite{langer,sc1,sc2,jap,liu,graug,kim,sc3,finn,sc5,Lu}.
162: Estimates for the mean free paths differ substantially in
163: different studies, and seem to depend on many aspects, e.g.~MWNT fabrication,
164: purification and preparation, as well as the energy regime probed
165: experimentally.
166: To mention some of the experimental evidence for
167: diffusive behaviors, there are typical
168: weak localization features, universal conductance fluctuations
169: and $h/2e$ oscillations in the magnetoconductance.
170: These experiments also show that electronic transport
171: proceeds only through the outermost shell (which is contacted
172: by external leads), unless that shell has been intentionally
173: damaged. There are a number of theoretical arguments
174: supporting this observation \cite{kane,egger99,prl}, and we shall
175: assume an effective single-shell model in what follows.
176: Inner shells then cause a screening of the electron-electron
177: interaction potential.
178: For a computation of the TDOS, the latter effect, as well as the
179: spin and K point degeneracy,
180: can be absorbed by a suitable renormalization of the
181: electron-electron interaction potential $U_0(\vec{q})$ \cite{egger99}.
182: Therefore we may simplify the computation and
183: consider only spinless electrons at
184: one K point. In addition,
185: since different shells always have incommensurate lattices
186: due to different curvature or helicity,
187: a quasiperiodic ionic potential from inner shells acts on
188: outermost-shell electrons. The effect of such a potential
189: is expected to be similar to a random potential described by a
190: mean free path $\ell=v_F\tau$.
191: Moreover, true disorder imposed by imperfections, substrate
192: inhomogeneities, tube processing or defects is very likely present.
193: Together with the inner-shell potential, this may explain the
194: reported mean free paths, $\ell \leq 100$~nm.
195: Estimating the localization length as $\xi \approx 4 M \ell$
196: using the standard Thouless argument,
197: electronic transport at low energy scales, $E\tau \ll 1$,
198: is therefore diffusive (but not localized)
199: for not exceedingly long doped MWNTs. However, for
200: $E\tau \gg 1$, there is a ballistic regime, where the
201: Luttinger liquid picture \cite{egger99} is appropriate
202: again, with minor modifications.
203: In this article, we will mainly focus on the more complex
204: energy regime $E\tau \ll 1$, i.e.~assume a sufficiently
205: dirty MWNT. In fact, we impose the condition $\ell \lapx 2\pi R$ such that
206: transport around the circumference can be considered as
207: diffusive. The opposite limit $\ell \gapx 2\pi R$ has recently been addressed
208: in Ref.~\cite{gl1}.
209:
210: At low energy scales, $E<v_F/R$, many groups have by now
211: experimentally observed pronounced {\sl zero-bias anomalies} in the TDOS
212: of an individual MWNT \cite{sc2,liu,graug,kim,sc3,finn,Lu}.
213: Most of these experimental results are described by a power-law
214: TDOS, $\nu(E)\propto E^\alpha$,
215: just as in a LL, with exponents clustering around $\alpha\approx
216: 0.3 \pm 0.1$ \cite{sc2,graug,sc3,finn}.
217: Remarkably, this value is of the same order of magnitude as the
218: exponents in a SWNT and hence the interpretation in terms of a LL may seem
219: obvious. One exception to this
220: result has been reported in Ref.~\cite{liu},
221: where $\alpha\approx 0.04\pm 0.02$.
222: Occasionally, also logarithmic dependencies have been observed \cite{Lu}
223: in MWNT bundles, where probably the electron-electron interaction
224: is externally screened. Such logarithmic dependencies could
225: be explained by tunneling into an effectively 2D diffusive system
226: with weak Coulomb interactions \cite{altshuler}.
227: Furthermore, the TDOS at the end of the MWNT, while still
228: of power-law form, is characterized by a doubling of the
229: boundary exponent, $\alpha_{\rm end}=2 \alpha$.
230: Unfortunately, it appears to be difficult to explain these findings
231: by Luttinger liquid theory, at least for the majority
232: of the quoted experimental studies. Although
233: LL theory can be extended
234: to ballistic multi-mode wires with inner-shell screening
235: \cite{egger99}, the presence
236: of many subbands in a doped MWNT inevitably implies rather
237: small exponents. Even optimistic estimates yield exponents that
238: are at least one order of magnitude smaller than observed.
239: For that reason, below we address the role of disorder for
240: the zero-bias anomaly.
241:
242: Building upon our original paper \cite{prl}, we provide a theoretical
243: description for the TDOS of MWNTs within the general
244: framework of {\sl Coulomb blockade theory} \cite{sct}.
245: We focus on sufficiently low energy scales $E < v_F/R$, where
246: it is sufficient to take a fixed number $M$ of subbands,
247: and thereby ignore
248: van Hove singularities associated with the opening
249: of new 1D subbands as energy is varied.
250: As is shown below, on intermediate energy scales, an apparent
251: power law suppression of the TDOS is found, which is distinct
252: from the LL power laws of a ballistic system.
253: This {\sl intrinsic Coulomb blockade} phenomenon
254: arises because of the suppression of
255: tunneling into an {\sl strongly interacting disordered metal}.
256: Because of strong interactions, a perturbative Altshuler-Aronov-Lee
257: (AAL) approach \cite{altshuler} is not possible.
258: The corresponding nonperturbative problem for 2D systems
259: has been studied in Refs.~\cite{finkel,belitz,levitov,kopietz,ka}.
260: Very recently, besides our own paper \cite{prl},
261: the 1D case has attracted considerable interest in the
262: theory community \cite{gl1,roll,kopietz2}.
263:
264:
265: \section{Intrinsic Coulomb Blockade}
266: \label{sec3}
267:
268: The key ingredient in Coulomb blockade theory is the probability $P(E)$ that
269: a tunneling electron excites electromagnetic modes
270: with energy $E$ in the system \cite{sct}. The theory
271: is meaningful if these modes are harmonic, and then $P(E)$
272: directly determines the TDOS according to a relation
273: first explicitly given in Ref.~\cite{roll},
274: \begin{equation} \label{tdossct}
275: \frac{\nu(E)}{\nu_0} = \int_{-\infty}^\infty dE' \frac{1+\exp(-E/k_B T)}{1+
276: \exp(-E'/k_B T)} P(E-E') \;,
277: \end{equation}
278: where $\nu_0$ is the non-interacting DOS.
279: The probability $P(E)$ is expressed as
280: \begin{equation}\label{pe}
281: P(E) = \frac{1}{\pi} {\rm Re} \int_0^\infty dt \exp[iEt+J(t)] \;,
282: \end{equation}
283: with the phase correlation function
284: \begin{equation} \label{jf}
285: J(t) = \int_0^\infty d\omega \frac{I(\omega)}{\omega}
286: \Bigl \{\coth(\omega/2 k_B T) [\cos(\omega t)-1]
287: - i \sin(\omega t) \Bigr \}
288: \end{equation}
289: for a given spectral density $I(\omega)$ of electromagnetic modes.
290: As a probability, $P(E)$ is normalized,
291: \begin{equation} \label{normp}
292: \int_{-\infty}^\infty P(E) = 1
293: \end{equation}
294: and fulfills the detailed balance relation, $P(-E)=\exp(-E/k_B T) P(E)$.
295:
296: For clarity, we now focus on the most interesting zero-temperature case.
297: Provided $I(\omega)$ remains finite for low frequencies,
298: Eqs.~(\ref{tdossct}) and (\ref{jf}) straightforwardly lead to
299: a power law for the TDOS with exponent $\alpha = I(\omega\to 0)$.
300: We then should establish the harmonic nature
301: of the electromagnetic modes and compute the low-frequency
302: spectrum $I(\omega)$. If $I(\omega\to 0)$ is finite,
303: a power law would directly follow.
304: Notice that a perturbative treatment of
305: interactions is apparently not sufficient, as
306: a power-law TDOS is inconsistent
307: with conventional perturbative (1D or 2D) AAL
308: predictions \cite{altshuler}.
309: In some studies \cite{sc3,finn}, $I(\omega)$ is phenomenologically
310: parameterized in terms of the total impedance $Z(\omega)$, i.e.~the
311: MWNT is modelled as a transmission line. Under such an
312: approach, one obtains $\alpha= Z(0)/(h/2e^2)$.
313: This simple transmission line model can directly explain
314: the doubling of the end exponent, because in the bulk
315: case one has effectively two resistances in parallel as
316: compared to the end case. However, since this purely
317: phenomenological approach can hardly represent a satisfactory
318: theory, we pursue a {\sl microscopic}
319: approach. It should also be
320: stressed at this point that the 1D pseudo-gap TDOS
321: found for small $E$ \cite{prl,gl1} is apparently
322: outside the reach of transmission line modelling.
323: Furthermore, the doubling of the end exponent
324: can be verified from microscopic theory as well.
325: The derivation is discussed at length in Ref.~\cite{prl}, and here
326: we focus on the bulk TDOS alone.
327:
328: Recent field-theoretical developments
329: allow to incorporate the Coulomb interactions
330: in a nonperturbative way \cite{finkel,kopietz,ka}.
331: Adapting the Keldysh nonlinear $\sigma$ model approach for
332: interacting disordered systems worked out in Ref.~\cite{ka},
333: the TDOS can be computed in analytical form
334: for arbitrary interaction strength.
335: This calculation is certainly on sound footing for long-ranged interactions
336: (which is the case for MWNTs) in 2D. For
337: truly 1D systems, however, the asymptotic low-energy
338: behavior of the TDOS resulting from this approach, see Ref.\cite{gl1},
339: has been questioned recently \cite{kopietz}.
340: In any event, the final result
341: yielded by this theory indeed reproduces phenomenological Coulomb blockade
342: theory, since the electromagnetic modes are found to be Gaussian
343: with spectral density \cite{prl,ka}
344: \begin{equation} \label{iw}
345: I(\omega) = \frac{\omega}{\pi} {\rm Im}
346: \sum_{\vec{q}} \frac{1}{(Dq^2 - i\omega)^2}
347: \left( U_0^{-1}(\vec{q}) +
348: \frac{\nu_0 Dq^2}{Dq^2-i\omega}\right)^{-1} \;.
349: \end{equation}
350: Here the diffusion coefficient for
351: charge diffusion on the tube surface is $D=v_F^2 \tau/2$.
352: In Eq.~(\ref{iw}), the $\vec{q}$ summation includes
353: an integral over the momentum
354: parallel to the tube direction (a very long MWNT is assumed),
355: and a summation over the discrete transverse momenta $q_\perp = n/R$
356: for integer $n$. For consistency,
357: the $n$ summation is restricted to $|n|\leq N$,
358: albeit the detailed value for the cut-off is not essential.
359: The Fourier-transformed Coulomb
360: interaction potential $U_0(\vec{q})$ includes the effect
361: of external screening by nearby gates or the substrate,
362: but {\sl not}\ of internal screening which is fully accounted for by
363: Eq.~(\ref{iw}). In what follows, to keep the discussion simple, we
364: consider an effectively short-range
365: interaction potential characterized by a constant $U_0$; for the
366: case of a $1/r$ potential, see, e.g.~Ref.~\cite{ka}.
367: Since the dominant contributions to $I(\omega)$ come from small
368: $q$, it is justified to integrate
369: over the longitudinal momentum in Eq.~(\ref{iw})
370: directly (there is no UV divergence), leading to
371: \begin{eqnarray} \label{iw1}
372: I(\omega) &=& \frac{U_0 }{2\pi (D^\ast-D)} \, {\rm Re} \,
373: \sum_{n=-N}^N \Bigl [
374: (-i\omega/D^\ast + n^2/R^2)^{-1/2} \\ \nonumber && \quad -
375: (-i\omega/D + n^2/R^2 )^{-1/2} \Bigr ]
376: \end{eqnarray}
377: with the field diffusion constant $D^*=D (1+\nu_0 U_0)$.
378: Although we do not present the derivation of Eq.~(\ref{iw}) here,
379: we feel it is important to summarize the main approximations
380: entering this result:
381: \begin{enumerate}
382: \item
383: The regime $\ell\lapx 2\pi R$ is considered, but
384: Eq.~(\ref{iw1}) should also yield
385: useful results for somewhat larger $\ell$, since then
386: the $n=0$ mode dominates the Coulomb blockade completely. The
387: $n=0$ mode is unaffected by assumptions concerning transversal motion.
388: \item
389: As a consequence of the assumed diffusive
390: behavior, the spectral density (\ref{iw1})
391: should only be used for $\omega \tau \ll 1$.
392: The $I(\omega\to 0) \sim \omega^{-1/2}$
393: behavior in Eq.~(\ref{iw}) due to the $n=0$ mode
394: implies that the dominant contribution
395: to the Coulomb blockade indeed results from these
396: low-energy collective modes. In Eq.~(\ref{jf}) we therefore truncate the
397: integration at the upper limit $1/\tau$. Since the higher energy
398: modes are equivalent to a Luttinger liquid, which in turn
399: only leads to comparatively weak Coulomb blockade effects
400: in this regime, their omission
401: is not expected to create serious problems.
402: \item
403: The Coulomb interaction potential should be
404: sufficiently long-ranged and smooth to
405: allow for semi-classical (WKB-type) treatments. The main
406: effect of the interaction is then to change the phase of
407: electron wavefunctions but not the amplitude.
408: Since Coulomb blockade is a low-energy
409: collective phenomenon and the interaction
410: potential is rather long-ranged in MWNTs,
411: this approximation should be justified.
412: \item
413: Only the intrinsic electrodynamic modes
414: of the MWNT are considered to be responsible
415: for the Coulomb blockade, but not
416: the attached external circuit. For sufficiently long MWNTs
417: such as the ones in Ref.~\cite{sc3}, the intrinsic
418: resistance is large and
419: environmental Coulomb blockade can safely be neglected.
420: \end{enumerate}
421:
422: Above the Thouless energy for diffusion around the
423: circumference,
424: \begin{equation} \label{thouless}
425: E_T \approx \frac{D}{(2\pi R)^2} \;,
426: \end{equation}
427: where charge diffusion is essentially two-dimensional,
428: we then expect basically an exponentiated 2D AAL law.
429: In contrast, for sufficiently low energy scales,
430: $E < E_T$, 1D behavior takes over, where
431: the $n=0$ mode in Eq.~(\ref{iw1}) becomes more and more
432: important as the energy scale is decreased.
433: In these two limits one can obtain the TDOS analytically.
434: In the {\sl 2D diffusive energy regime} $E > E_T$, a power
435: law emerges by converting the $n$-summation into an integral.
436: The bulk exponent follows directly as $\alpha=I(\omega\to 0)$, and reads
437: \cite{prl}
438: \begin{equation} \label{alph}
439: \alpha = \frac{R}{2\pi \nu_0 D} \ln(D^*/D) \;.
440: \end{equation}
441: In effect, the AAL logarithmic correction therefore {\sl exponentiates}
442: into a {\sl power law} \cite{prl}. We wish to stress that the
443: derivation of Eq.~(\ref{alph}) works only in the true 2D limit,
444: characterized by a large number of bands $M$ or by $\ell \ll R$.
445: On the other hand, for sufficiently low energy scales, only the $n=0$
446: mode responsible for 1D perturbative AAL corrections is important.
447: Keeping only the $n=0$ term in Eq.~(\ref{iw1})
448: results in a divergent (``sub-Ohmic'') behavior of
449: the low-frequency spectral density, $I(\omega)\sim \omega^{-1/2}$,
450: and hence to the appearance of a pseudo-gap as $E\to 0$ \cite{weiss},
451: \begin{equation} \label{1dlimit}
452: \nu(E) \sim \exp(-E_0/E) \;,
453: \end{equation}
454: where we neglect a prefactor exhibiting
455: power-law energy dependence. Alternatively, this leads to
456: logarithmic corrections of the scale $E_0$ in Eq.~(\ref{sce0}) below.
457: The result (\ref{1dlimit}) agrees with the findings of Refs.~\cite{gl1,roll}.
458: Using a stationary-phase evaluation
459: of $P(E)$ in Eq.~(\ref{pe}),
460: the energy scale $E_0$ in Eq.~(\ref{1dlimit}) follows as \cite{weiss}
461: \begin{equation}\label{sce0}
462: E_0 = \frac{U^2_0}{8\pi D \left(1+\sqrt{D^\ast/D}\right)^2 } \;,
463: \end{equation}
464: For strong interactions, $D^\ast\gg D$, this can be simplified
465: to $E_0= U_0/8\pi \nu_0 D$.
466: Remarkably, Eq.~(\ref{1dlimit}) representing the exponentiated 1D AAL law
467: does {\sl not}\ reproduce the perturbative 1D AAL $1/\sqrt{E}$ behavior under
468: a naive direct expansion of the exponential \cite{gl1,roll}.
469:
470: \section{Numerical solution of Coulomb blockade equations}
471: \label{sec4}
472:
473: To analyze the full energy dependence of the TDOS,
474: numerical methods are mandatory. Using the spectral density (\ref{iw1}),
475: the phase correlation function $J(t)$ in Eq.~(\ref{jf}) can
476: be evaluated, which then allows for the computation of $P(E)$
477: according to Eq.~(\ref{pe}). Finally, Eq.~(\ref{tdossct}) yields
478: the TDOS. For simplicity, we again focus on the $T=0$
479: limit, although the finite-temperature case is also directly
480: accessible. A convenient check that the numerical procedure
481: has converged is provided by
482: the normalization condition for $P(E)$, see Eq.~(\ref{normp}),
483: which is accurately fulfilled for the results reported below.
484: For convenience,
485: the energy scale is set by $v_F/R$ throughout
486: this section. We consider a situation
487: with $k_F R=5.5$ so that $N=5$ and the number of bands is $M=11$, and
488: use the value
489: $U_0/2\pi v_F = 1$ to parameterize the interaction strength.
490: Unfortunately, it appears to be rather difficult to compute a realistic value
491: for $U_0$ (except possibly by ab-initio methods).
492: The above choice corresponds to rather strong interactions,
493: but for comparison
494: we have also carried out calculations
495: for $U_0/v_F=1$ (not shown, but see below).
496:
497: \begin{figure}
498: \epsfysize=10cm
499: \epsffile{f1.eps}
500: \caption{ \label{fig1}
501: Numerical result for the energy-dependence of the TDOS at $\ell=R$
502: on a double-logarithmic scale (for other parameters, see text).
503: The dotted line represents a power-law fit with exponent $\alpha=1.97$.
504: Inset: Arrhenius plot of the same data. The dotted line
505: has slope $E_0=0.015$.}
506: \end{figure}
507:
508:
509: \begin{figure}
510: \epsfysize=10cm
511: \epsffile{f2.eps}
512: \caption{ \label{fig2}
513: Same as Fig.~\ref{fig1} but for $\ell=10 R$.
514: The dotted line is a power-law fit with exponent $\alpha=1.1$.
515: Inset: Arrhenius plot of the same data. The dotted line
516: has slope $E_0=0.00025$.}
517: \end{figure}
518:
519:
520: Let us start with a rather dirty MWNT, $\ell = R$.
521: In the mentioned units,
522: the Thouless scale is $E_T= 0.013$, and the
523: plot of the TDOS in Fig.~\ref{fig1} indeed shows an apparent {\sl power law}
524: even for energies well below $E_T$, extending up to $E\approx 0.1$
525: over approximately one decade.
526: The inset shows that for $E\to 0$, the predicted pseudo-gap behavior
527: emerges. For this parameter set,
528: the power-law exponent $\alpha=0.23$ predicted
529: from the 2D limit, Eq.~(\ref{alph}), is much smaller than the
530: numerically observed exponent $\alpha=1.97$.
531: The estimate (\ref{alph}) is therefore too crude and really restricted
532: to the true 2D regime, since it
533: ignores the special role played by the $n=0$ contribution in
534: the spectral density.
535: Nevertheless, we find a clear power-law behavior at intermediate
536: energy scales. Importantly, the
537: regime of validity for the power law is not set by the Thouless scale,
538: but by a smaller energy scale.
539: Similarly, the value $E_0=0.015$ extracted from the slope of the
540: Arrhenius plot in the inset
541: of Fig.~\ref{fig1} is significantly smaller than the
542: value $E_0=0.07$ predicted by Eq.~(\ref{sce0}).
543: This deviation is probably linked to strong logarithmic renormalizations
544: of the scale (\ref{sce0}) by the power-law prefactor
545: not written out in Eq.~(\ref{1dlimit}), and is always observed in
546: our calculations.
547: Since this renormalization of $E_0$ makes this scale quite
548: small, the pseudo-gap TDOS is
549: in practice difficult to distinguish from the power law
550: except at very low energies. This may offer a (somewhat trivial)
551: explanation for the experimental difficulties encountered in finding
552: pseudo-gap behavior. Finally, for high energies
553: close to $v_F/R$, the non-interacting DOS is approached.
554:
555: Next we discuss the {\sl quasi-ballistic limit}, taking $\ell=10 R$.
556: The features shown in Fig.~\ref{fig2}
557: are qualitatively similar as in Fig.~\ref{fig1},
558: namely a pseudo-gap at very low energies turns into a power law at intermediate
559: energies. The power law exponents become systematically smaller by increasing
560: $\ell$, in this case $\alpha=1.1$.
561: The power law crosses over into the pseudo-gap as $E\to 0$, with $E_0=0.00025$
562: again much smaller than the expected value $E_0=0.007$.
563: As this power law feature is definitely not linked to the
564: Thouless scale, it is not related to the exponent (\ref{alph})
565: for tunneling into a 2D interacting disordered metal.
566: We have checked for this value of $\ell$ that the
567: power law persists for smaller $U_0$. In fact, the
568: exponent $\alpha$ then systematically decreases, and
569: for $U_0/v_F=1$ is $\alpha\approx 0.3$, which would be in
570: good agreement with experiment.
571: Since the interaction strength in the experiments conducted in Ref.~\cite{sc3}
572: did probably not vary much from tube to tube, the robustness of
573: the observed exponents with respect to changes in $\ell/R$
574: is encouraging. We mention in passing that
575: for suspended MWNTs or smaller doping levels, one may
576: reach a regime of stronger interactions, where again
577: power-law behavior at
578: intermediate energies is predicted, but with larger exponents.
579:
580: \section{Conclusions}
581: \label{sec5}
582:
583: MWNTs represent a unique laboratory for exploring mesoscopic
584: physics in the presence of electron-electron interactions.
585: Here we have addressed one aspect, namely the zero-bias
586: anomaly of the tunneling density of states due to
587: Coulomb interactions among the electrons.
588: Assuming a sufficiently dirty MWNT with mean free path
589: less than the circumference,
590: the spectral density $I(\omega)$ of Coulomb blockade theory
591: has been computed, and the numerical
592: solution of the resulting equations for the TDOS was presented. The results
593: show power-law behavior at intermediate energy scales,
594: extending down to quite low energies over typically one to
595: two decades. Remarkably, the power law is seen at
596: energies less than the Thouless
597: scale for diffusion around the circumference, i.e.~it does
598: not appear to reflect tunneling into a 2D diffusive electron
599: liquid. At very low energies,
600: the power-law behavior $\nu(E) \sim E^\alpha$
601: crosses over into a pseudo-gap of the
602: form $\nu(E) \sim \exp(-E_0/E)$, although it
603: should be stressed that in practice both are
604: sometimes hard to distinguish
605: because of the rather small scale $E_0$.
606: According to our numerical study,
607: the power-law exponent $\alpha$ would be consistent with
608: typical exponents around $\alpha\approx 0.3$
609: for interaction strength $U_0/v_F\approx 1$,
610: but is significantly larger for stronger interactions.
611: We note that the power law is very robust, and has always been
612: observed in our calculations, regardless of the chosen values
613: for $\ell/R$ or $U_0$.
614:
615: Support by the DFG under the Gerhard-Hess program and by the
616: EPSRC under Grant No.~GR/N19359 is gratefully acknowledged.
617:
618: \begin{thebibliography}{00}
619:
620: \bibitem{joachim}
621: For a review, see C. Joachim, J.K. Gimzewski and A. Aviram,
622: Nature 408 (2000) 541, and references therein.
623:
624: \bibitem{dekker99} C. Dekker, Physics Today 52(5) (1999)
625: 22, and:
626: Special Issue on carbon nanotubes, Physics World 13(6) (2000).
627:
628: \bibitem{nt1}
629: S.J. Tans, M.H. Devoret, H. Dai, A. Thess, R.E. Smalley, L.J. Geerligs
630: and C. Dekker, Nature 386 (1997) 474.
631:
632: \bibitem{nt2}
633: S.J. Tans, M.H. Devoret, R.J.A. Groeneveld and
634: C. Dekker, Nature 394 (1998) 761.
635:
636: \bibitem{egger97}
637: R. Egger and A.O. Gogolin, Phys. Rev. Lett. 79 (1997) 5082;
638: Eur. Phys. J. B 3 (1998) 281.
639:
640: \bibitem{kane97}
641: C.L. Kane, L. Balents and M.P.A. Fisher,
642: Phys. Rev. Lett. 79 (1997) 5086.
643:
644: \bibitem{ll2}
645: For a recent textbook, see
646: A.O. Gogolin, A.A. Nersesyan and A.M. Tsvelik,
647: Bosonization and Strongly Correlated Systems
648: (Cambridge University Press, Cambridge, 1998),
649: and references therein.
650:
651: \bibitem{ll1} M. Bockrath, D.H. Cobden, J. Lu, A.G. Rinzler,
652: R.E. Smalley, L. Balents and P.L. McEuen,
653: Nature 397 (1999) 598.
654:
655: \bibitem{ll1a}
656: J. Nygard, D.H. Cobden, M. Bockrath, P.L. McEuen and P.E. Lindelof,
657: Appl. Phys. A 69 (1999) 297.
658:
659: \bibitem{ll1b}
660: Z. Yao, H. Postma, L. Balents and C. Dekker,
661: Nature 402 (1999) 273.
662:
663: \bibitem{ll1c}
664: H. Postma, M. de Jonge, Z. Yao and C. Dekker,
665: Phys. Rev. B 62 (2000) 10653.
666:
667: \bibitem{ll1d}
668: H. Postma, T. Teepen, Z. Yao, M. Grifoni and C. Dekker,
669: Science 293 (2001) 76.
670:
671: \bibitem{review}
672: R. Egger, A. Bachtold, M. Fuhrer, M. Bockrath, D. Cobden and P. McEuen,
673: in: Interacting Electrons in Nanostructures, edited by
674: R. Haug and H. Schoeller (Springer Lecture Notes in Physics, 2001).
675:
676: \bibitem{sc4}
677: M. Kr{\"u}ger, M.R. Buitelaar, T. Nussbaumer,
678: C. Sch{\"o}nen\-berger and L. Forr{\'o},
679: Appl. Phys. Lett. 78 (2001) 1291.
680:
681: \bibitem{frank}
682: S. Frank, P. Poncharal, Z.L. Wang and W.A. de Heer,
683: Science 280 (1998) 1744.
684:
685: \bibitem{langer}
686: L. Langer, V. Bayot, E. Grivei, J.P. Issi, J.P. Heremans,
687: C.H. Olk, L. Stockman, C. Van Hasendonck and Y. Bruynseraede,
688: Phys. Rev. Lett. 76 (1996) 479.
689:
690: \bibitem{sc1} A. Bachtold, C. Strunk, J.P. Salvetat, J.M. Bonard,
691: L. Forr{\'o}, T. Nussbaumer and C. Sch{\"o}nenberger, Nature
692: 397 (1999) 673.
693:
694: \bibitem{sc2} C. Sch{\"o}nenberger, A. Bachtold, C. Strunk, J.P. Salvetat
695: and L. Forr{\'o}, Appl. Phys. A 69 (1999) 283.
696:
697: \bibitem{jap}
698: A. Fujiwara, K. Tomiyama, H. Suematsu, M. Yumura and K. Uchida,
699: Phys. Rev. B 60 (1999) 13492.
700:
701: \bibitem{liu}
702: K. Liu, Ph. Avouris, R. Martel and W.K. Hsu,
703: Phys. Rev. B 63 (2001) 161404.
704:
705: \bibitem{graug}
706: E. Graugnard, P.J. de Pablo, B. Walsh, A.W. Ghosh,
707: S. Datta, and R. Reifenberger,
708: Phys. Rev. B 64 (2001) 125407.
709:
710: \bibitem{kim}
711: J. Kim, K. Kang, J. Lee, K. Yoo, J. Kim,
712: J.W. Park, H.M. So, and J.J. Kim,
713: J. Phys. Soc. Jpn. {\bf 70}, 1464 (2001).
714:
715:
716: \bibitem{sc3} A. Bachtold, M. de Jonge, K. Grove-Rasmussen,
717: P.L. McEuen, M. Buitelaar and C. Sch{\"o}nenberger,
718: cond-mat/0012262.
719:
720: \bibitem{finn}
721: R. Tarkiainen, M. Ahlskog, J. Penttil\"a, L. Roschier,
722: P. Hakonen, M. Paalanen and E. Sonin,
723: cond-mat/0104019.
724:
725:
726: \bibitem{sc5}
727: C. Sch\"onenberger, M. Buitelaar, M. Kr\"uger, I. Widmer,
728: T. Nussbaumer and M. Iqbal, cond-mat/0106501.
729:
730: \bibitem{Lu} Li Lu, unpublished data.
731:
732: \bibitem{kane} A.A. Maarouf, C.L. Kane and E.J. Mele,
733: Phys. Rev. B 61 (2000) 11156.
734:
735: \bibitem{egger99} R. Egger, Phys. Rev. Lett. 83 (1999) 5547.
736:
737: \bibitem{prl} R. Egger and A.O. Gogolin,
738: Phys. Rev. Lett. 87 (2001) 066401.
739:
740: \bibitem{gl1} E.G. Mishchenko, A.V. Andreev and L.I. Glazman,
741: cond-mat/0106448.
742:
743: \bibitem{altshuler}
744: B.L. Altshuler and A.G. Aronov, in: Electron-electron interactions
745: in disordered systems, edited by A.L. Efros and M. Pollak
746: (Elsevier, Amsterdam, 1985).
747:
748: \bibitem{sct} Single Charge Tunneling, edited by
749: H. Grabert and M.H. Devoret, NATO-ASI Series B: Vol.~294
750: (Plenum Press, 1992).
751:
752:
753: \bibitem{finkel}
754: A.M. Finkelstein, Zh. Eksp. Teor. Fiz. 84 (1983) 168
755: [Sov. Phys. JETP 57 (1983) 97].
756:
757: \bibitem{belitz}
758: D. Belitz and T.R. Kirkpatrick,
759: Phys. Rev. B 48 (1993) 14072.
760:
761: \bibitem{levitov}
762: L.S. Levitov and A.V. Shytov,
763: Pis'ma Zh. Eksp. Teor. Fiz. 66 (1997) (200)
764: [JETP Lett. 66 (1997) 214].
765:
766: \bibitem{kopietz}
767: P. Kopietz, Phys. Rev. Lett. 81 (1998) 2120.
768:
769: \bibitem{ka}
770: A. Kamenev and A.V. Andreev, Phys. Rev. B 60
771: (1999) 2218.
772:
773: \bibitem{roll} J. Rollb\"uhler and H. Grabert,
774: Phys. Rev. Lett. 87 (2001) 126804.
775:
776: \bibitem{kopietz2}
777: L. Bartosch and P. Kopietz, cond-mat/0108463.
778:
779: \bibitem{weiss}
780: U. Weiss, Quantum dissipative system, 2nd enlarged edition,
781: Chapter 20.2.3 (World Scientific, Singapore, 1999)
782:
783: \end{thebibliography}
784:
785: \end{document}
786:
787: