1: \documentclass[12pt]{iopart}
2: \usepackage[dvips]{graphicx}
3: \usepackage{amssymb}
4: \usepackage{amsfonts}
5:
6: \begin{document}
7: \title[Beyond the Linearity of Current-Voltage Characteristics in Multiwalled Carbon Nanotubes]{Beyond the Linearity of Current-Voltage Characteristics in Multiwalled Carbon Nanotubes}
8:
9: \author{B. Bourlon$^{1,2}$, C. Miko$^3$, L. Forr\'o$^3$, D.C. Glattli$^{1,4}$ and A. Bachtold$^{1,5}$}
10: \address{$^1$ Laboratoire Pierre Aigrain, Ecole Normale Sup\'{e}rieure, 24
11: rue Lhomond, 75231 Paris 05, France}
12: \address{$^2$ Department of Applied Physics, California Institute of Technology, Pasadena, California 91125, USA}
13: \address{$^3$ EPFL, CH-1015, Lausanne, Switzerland} \address{$^4$ SPEC, CEA Saclay,
14: F-91191 Gif-sur-Yvette, France}
15: \address{$^5$ ICN and CNM-CSIC, Campus Universitat Autonoma de Barcelona, E-08193 Bellaterra, Spain}
16: %\ead{}
17:
18: \begin{abstract}
19: We present local and non-local electron transport measurements on individual multi-wall nanotubes for bias voltage between 0 and about 4 V. Local current-voltage characteristics are quite linear. In contrast, non-local measurements are highly non-linear; the differential non-local conductance can even become negative in the high-bias regime. We discuss the relationship between these results and transport parameters such as the elastic length, the number of current carrying shells, and the number of conducting modes.
20:
21: \end{abstract}
22: \pacs{73.63.Fg, 73.50.Fq, 72.10.Di}
23: \submitto{\SST} \maketitle
24:
25: \section{Introduction}
26: Multiwalled carbon nanotubes (MWNT) have recently attracted a lot
27: of attention as possible interconnects in future integrated
28: circuits $\cite{weiapl2001,kreume2002,liapl2003}$. MWNTs can carry
29: several hundreds of $\mu$A in a section of about 100 nm$^2$,
30: which corresponds to a current-carrying capacity much larger than
31: the one of today's interconnects. MWNTs consist of several nested
32: cylindrical graphene shells. Their electronic transport properties
33: have been studied mainly with devices where the electrodes are in
34: direct contact with the outer shell. Understanding the electronic
35: transport properties of a MWNT requires knowledge of
36: the transport properties of each shell. More precisely, the
37: conductance of a MWNT depends on parameters such as $l_e$ the mean
38: free path, $N_{shell}$ the number of current-carrying shells, and
39: $N_{mode}$ the number of conducting modes. It has been possible to
40: determine these parameters mainly in two distinct regimes. The two
41: regimes, which are set by the bias voltage applied to the
42: conductor, are the low-bias regime (voltage below $kT=25$ meV) and
43: the high-bias regime (around a few volts).
44:
45: Results from previous studies on how
46: the mean free path, the number of shells, and the number of modes
47: per shell contributing to conduction vary between the low- and
48: the high-bias regime are reviewed in this paragraph. We note that these results have been
49: obtained on MWNT devices that have been fabricated following a
50: specific fabrication process (see below). For samples prepared using
51: a different method the results can be different
52: $\cite{a1,a2,a3,a3b,a4,a5,a6,a7}$. For example, the doping and the
53: mean-free path can be different due to a different device
54: preparation. Also, in some of these papers, the MWNT is suspended
55: so that the heat produced by Joule heating is less evacuated.
56:
57: - $l_e$ the mean free path. In the low-bias regime, it has been
58: found that the transport is diffusive with a mean free path
59: $l_{e}$ $\lesssim$ $100\ nm$
60: $\cite{bachtoldnature1999,schAPA1999,bourlonprl2004n2}$. The
61: diffusion arises from disorder but the microscopic nature of these
62: scattering centers is not yet known. At high-bias, the mean-free
63: path becomes shorter with a mean free path $l_{e} \sim 5-10\ nm$
64: $\cite{yaoprl2000,javeyprl2004,bourlonprl2004n1}$. This regime has
65: been intensively studied these last few years for SWNTs. The reduction
66: of $l_{e}$ is attributed to the strong coupling between electrons
67: and optical phonons
68: $\cite{yaoprl2000,maulprl2004,dubayprb2003,ferrari}$.
69:
70: -$N_{mode}$ the number of modes per shell participating
71: significantly in conduction. In the low-bias regime, it has been
72: shown that MWNTs are doped with a Fermi level that is hundreds
73: millivolts away from the charge neutrality point
74: $\cite{krugerAPL2001}$. The corresponding number of modes is about
75: $N_{mode}=10-30$. In the high-bias regime, the number of modes
76: that significantly contribute to conduction is suppressed to a
77: few modes $\cite{bourlonprl2004n1,collins2001,collinsprl2001}$.
78: This has been attributed to the Zener backscattering that occurs
79: between the lower and upper non crossing subands
80: $\cite{bourlonprl2004n1,ananprb2000}$.
81:
82: -$N_{shell}$ the number of shells participating significantly in
83: conduction. In the low-bias regime, non-local resistance
84: measurements showed that essentially two shells contribute to
85: conduction when the distance between electrodes is below around $1
86: \mu m$ $\cite{bourlonprl2004n2}$. This changes at high bias where
87: most of the shells carry current. This result has been obtained
88: from the electrical breakdown experiments, which consist of
89: removing the shells one by one by injecting a large current in the
90: MWNT $\cite{a4,bourlonprl2004n1,collins2001,collinsprl2001}$.
91:
92: Between the low- and high-bias regimes there are no experimental
93: methods to probe $l_e$, $N_{mode}$, and $N_{shell}$. The purpose
94: of this article is to discuss the relationship between the
95: current-voltage (I-V) characteristics of MWNTs and the variations
96: of $l_e$, $N_{mode}$, and $N_{shell}$ with bias voltage.
97: Surprisingly, the local I-V characteristics appear mostly linear
98: up to several volts. At first sight, this may suggest that $l_e$,
99: $N_{mode}$, and $N_{shell}$ do not change as the bias voltage is
100: increased, in opposition to the above summary. However, we will
101: see that the variation of $l_e$, $N_{mode}$, and $N_{shell}$ can
102: be observed in non-local I-V characteristics that are highly
103: non-linear.
104:
105:
106:
107: \section{Experimental results}
108:
109: The MWNTs are synthesized by arc-discharge evaporation and
110: carefully purified $\cite{bonaram1997}$. Nanotubes are sonicated in dichloroethane and
111: then dispersed on an oxidized Si substrate. Finally, they are
112: contacted by top Cr/Au electrodes using electron-beam lithography.
113: Typical two-point resistances range from 5 to 30 $k\Omega$.
114:
115: We carry out local and non-local measurements. In non-local
116: measurements, the voltage drop is measured outside the region
117: between the current biased electrodes, while for local
118: measurements the current that flows through the voltage biased
119: electrodes is measured (see schematic in Fig. 1(b)) Local
120: measurements in Fig. 1(a) show that the current is roughly linear
121: with the bias voltage up to 3 volts. This could suggest that
122: parameters such as $l_e$, $N_{mode}$, and $N_{shell}$ remain
123: constant with the bias voltage. However, Fig. 1(b) shows that the
124: non-local measurements are highly non-linear. The non-local
125: voltage $\Delta V$ first increases linearly up to 1.5 V and then
126: decreases. Similar results are obtained for several other devices.
127: This is the main result of the paper. It shows that the transport
128: characteristics of MWNTs change between the low- and high-bias
129: regimes.
130:
131: It was shown that the non-local voltage at low-bias mainly results
132: from the current pathway through the two outermost shells
133: $\cite{bourlonprl2004n2}$. The current pathway at low bias was
134: shown in Ref. $\cite{bourlonprl2004n2}$ to be described by a
135: resistive transmission line with the intrashell resistance about
136: $\sim$ 10 k$\Omega$/$\mu$m and the intershell conductance about
137: $\sim$(10 k$\Omega$)$^{-1}$/$\mu$m, see Fig. 4(a). In Fig. 1(b)
138: the reduction of $\Delta V$ at high bias suggests that the current
139: pathway in the MWNT changes. Less current flows through the
140: outermost shell in regions that are not between the biased
141: electrodes. A change in the current pathway between the low- and
142: the high-bias regimes is not surprising when considering the
143: variations of $l_e$, $N_{mode}$, and $N_{shell}$ discussed in the
144: introduction.
145:
146: When looking at the high non-linearity of non-local measurements,
147: it is however surprising that the local $I-V$ curves appear so
148: linear. To try to clarify this issue, we have studied more
149: than 50 different nanotubes. Three representative measurements of
150: local $dI/dV-V$ characteristics are presented in Fig. 2. Between 0
151: and about 2 V, we obtain all the possible slopes for $dI/dV-V$;
152: $dI/dV-V$ can increase, remain constant, or decrease. At higher
153: voltages, however, $dI/dV$ systematically decreases. The different
154: variations of $dI/dV-V$ suggest that transport characteristics,
155: such as $l_e$, $N_{mode}$, and $N_{shell}$, vary differently for
156: different tubes as the voltage is increased.
157:
158: Figure 3 summarises the local measurement over all the tubes. The figure shows that there is a correlation between the slope of $dI/dV-V$ and
159: geometrical parameters of MWNTs such as the length and diameter. A reduction of $dI/dV-V$ is more probable to occur for
160: long and thin MWNTs, while an enhancement is more probable for short and thick tubes.
161:
162: It is interesting to compare these results to measurements on
163: SWNTs. The enhancement of $dI/dV-V$ can be observed at low bias
164: voltage up to about 100 mV, which is usually attributed to Coulomb
165: interaction $\cite{bockrath1999}$. It is however very rare to
166: observe in SWNTs an enhancement of $dI/dV-V$ up to 2V as it is the
167: case for short and thick MWNTs. This difference can be attributed
168: to the large diameter shells in MWNTs that have shorter separations
169: in energy between subbands. High energy subbands are then easier
170: to populate with electrons and thus can easily carry current at high bias
171: voltage. In addition, the different behaviour between SWNTs and
172: short thick MWNTs can also be attributed to the fact that MWNTs
173: consist of several SWNTs. The enhancement of $dI/dV-V$ can reflect
174: the enhancement of the number of shells that carry current.
175: Further discussion will be found in the section below.
176:
177: \section{Discussion}
178:
179: We now discuss the relationship between these results and transport parameters such as $l_e$, $N_{mode}$, and $N_{shell}$. Unfortunately, we are not at the stage where we are able to quantify local and non-local $I-V$ characteristics in MWNTs. We can neither account for the length nor the diameter dependences of the local $I-V$ characteristics reported in Fig. 3. However, in the following, we will explain our understanding of how $l_e$, $N_{mode}$, and $N_{shell}$ change from the low- to the high-bias regime.
180:
181: In Fig. 4(a), we describe MWNTs by multiple resistive lines. Each shell is modeled by a
182: series of resistances that depend on both the number of
183: modes and the transmissions of these modes. This model has successfully described measurements in the low-bias regime in which the current is shown to flow through the two outermost shells $\cite{bourlonprl2004n2}$. As the bias voltage is increased, more shells are expected to participate in the conduction. Moreover, the number of modes per shell, and
184: the transmission of each of these modes are also expected to change. The main problem is to
185: know how these parameters change with the bias voltage.
186:
187: We first look at the transmission $T(V)$. Yao et al.
188: $\cite{yaoprl2000}$ have shown that the current in SWNTs saturates
189: at high voltage. This has been attributed to the suppression of
190: the transmission due to scattering processes between electrons and
191: optical phonons with an energy of $\hbar \Omega = 160$ $meV$. At
192: first sight we can assume that an electron is scattered as soon as
193: its energy acquired from the electric field reaches $\hbar \Omega$
194: so that the mean free path $l$ is given by $e(V/L)l=\hbar \Omega$.
195: The transmission, which is obtained by dividing $l$ by $L$, is
196: then $T(V)=\hbar\Omega/eV$. The conductance is obtained using the
197: Landauer formula (Fig. 4(b)). The figure shows how the
198: differential conductance goes to zero as the bias voltage is
199: increased.
200:
201: We now discuss the variation in the number of modes that
202: participate in conduction as the bias voltage is increased. For
203: simplicity, we start the discussion for undoped tubes. As a first
204: approximation, one may simply count the number of subbands that
205: lie between the energy range $eV$. In the high-bias regime there
206: would be several tens of such subbands for a 10 nm diameter shell.
207: This however overestimates the number of efficient modes that
208: contribute significantly to the current. It has been found that
209: all shells carry a very similar current at high-bias and that this
210: current is close to the saturation current of about 20 $\mu A$
211: which is flowing through metal SWNTs. This suggests that the
212: number of efficient subbands per shell in MWNTs is close to the
213: number of conducting modes in SWNTs, which is 2.
214:
215: More precisely, this reduction of efficient modes has been
216: attributed to intersubband Zener tunnelling. This process involves
217: electrons tunnelling from the top of valence subbands to the
218: bottom of conduction subbands (see Fig. 5(a)) $\cite{ananprb2005}$. The transmission of
219: this process is dramatically lower for electrons that lie in
220: subbands far in energy from the charge neutrality point, so that
221: such subbands do not contribute much to the conduction.
222:
223: Zener tunnelling is expected to be voltage dependent. The
224: transmission is expected to become higher as the bias voltage is
225: enhanced, since the electric field reduces the length of the
226: barrier. In addition, the high current in the MWNT enhances the
227: temperature through Joule heating $\cite{chen2005}$ $\footnote{In
228: Ref. $\cite{chen2005}$ the temperature has been shown to reach
229: about 2000 C in the high-bias regime. These measurements have been
230: carried on suspended MWNTs. In our case, the MWNTs are lying
231: directly on the silicon oxide layer, so that the produced heat can
232: be better evacuated. It makes difficult to estimate the
233: temperature in our experiments; it can be between 300 and 2000
234: K.}$, so that electrons can easier pass the Zener barrier. This
235: results in an increase of the differential conductance as shown
236: in Fig. 4(c).
237:
238: We now take into account the fact that our tubes are significantly
239: p-doped $\footnote{This can be shown by measuring the conductance
240: as a function of the voltage applied on the backgate. The
241: conductance is observed to be reduced by around a factor 2 as the
242: gate voltage is swept from -100V to +100V. We never observed an
243: enhancement of the conductance as the gate voltage is increased
244: even at very large positive gate voltages.}$. Kruger et al. have
245: shown using an electrochemical gate that the Fermi level is
246: shifted by around 0.3 eV from the charge neutrality point, which
247: corresponds to $N_{mode} =10-30$ $\cite{krugerAPL2001}$. Figure
248: 5(b) shows the schematic of a semiconducting shell with a few
249: subbands since it is difficult to represent much more subbands.
250: The number of conducting modes is here $N_{mode} =4$ and the
251: states contributing to transport are represented by the dark gray region
252: set by $eV$. Figure 5(c) shows the case at high-bias. We assume
253: that electrons travel at constant energy between the electrodes.
254: Most electrons have then to cross the gap via Zener tunneling.
255: Only electrons in subbands close to the charge neutrality point can then
256: carry current since the others are backscattered when they try to
257: cross the gap. Thus, many of the subbands that conduct at low-bias
258: become inefficient at high-voltage. We notice however that these
259: subbands generate a low but non-negligible current since some
260: electrons do not have to cross the gap, represented in the dark gray region in Fig.
261: 5(c). Overall, as the bias voltage is increased from zero, the
262: number of modes $N_{mode}$ that carry a significant amount of current first
263: decreases due to the Zener backscattering at the gap. At higher
264: voltage, however, the differential conductance increases again
265: since the Zener transmission is enhanced.
266:
267:
268:
269: We here turn our attention to the number of shells that contribute
270: significantly to electron transport. Adding conducting shells
271: obviously increases the differential conductance. It has been
272: shown that in the linear regime only one or two shells participate
273: in the conduction $\cite{bourlonprl2004n2}$. This is different
274: from the high-bias regime where almost all shells have been shown
275: to participate
276: $\cite{a4,bourlonprl2004n1,collins2001,collinsprl2001}$. However
277: little is known about how $N_{shell}$ increases between the linear
278: and high-bias regime (Fig. 4(d)). Indeed, the probability for an
279: electron to go into the adjacent shell depends on the shell
280: resistances that are given by $l_e$ and $N_{mode}$
281: $\cite{trioprb2004,c1,c2}$, as illustrated in the resistive line
282: model in Fig. 4(a). The probability is enhanced (reduced) as the
283: resistances are larger (smaller). Because the behaviour of
284: $N_{mode}$ is not well understood, it is difficult to predict how
285: electrons penetrate in inner shells as the voltage increases.
286:
287:
288: The variation of $dI/dV-V$ can also be a consequence of the
289: contact between the nanotube and the electrode. Indeed, it has
290: been shown that $dI/dV-V$ increases with $V$ due to Coulomb
291: interaction. When the resistance of the contact is larger than $
292: \sim h/2e^2$, the functional form is expected to follow a power law
293: $\cite{b1,b2,b3,b4}$. The exponent has been measured to be about 0.3.
294: For lower contact resistance, the functional form of $dI/dV-V$
295: remains to be determined. However, the enhancement of $dI/dV-V$ is
296: expected to be much less pronounced. It is thus suitable to work
297: with low-ohmic contacted MWNTs in order to disregard the
298: contribution from the contacts in the analysis of $I-V$
299: characteristics.
300:
301: \section{Conclusion}
302:
303: We have presented local and nonlocal measurements in MWNTs. The
304: main result is that nonlocal $I-V$ characteristics are highly non
305: linear. The nonlocal voltage can even decrease at high bias
306: voltage. This suggests that the current pathway changes as the
307: bias voltage increases. This is attributed to the variation of
308: transport parameters such as $l_e$, $N_{mode}$, and $N_{shell}$.
309: As the bias voltage increases, the mean-free path is expected to
310: decrease, while the number of shells is expected to increase. The
311: behaviour of the number of modes that significantly contributes to
312: transport is expected to be more complicated when working with
313: doped MWNTs; $N_{mode}$ first decreases and then increases.
314: Further studies are needed to understand transport properties of
315: MWNTs between the low and high-bias regimes. In particular, it
316: would be interesting to account for the length and diameter
317: dependences of the $I-V$ characteristics that are reported in Fig.
318: 3. For this, it would be very interesting to study DWNTs, which are
319: simpler systems.
320:
321: \ack
322: We thank B. Pla\c{c}ais for discussions, and P. Morfin, F. R. Ladan and
323: C. Delalande for support. LPA is CNRS-UMR8551 associated to University Paris 6 and 7.
324: The research has been supported by the DGA, ACN, sesame, the Swiss National Science
325: Foundation, and its NCCR "Nanoscale Science".
326:
327: \section*{References}
328: \bibliographystyle{prsty}
329: \bibliography{./biblio}
330:
331: \clearpage
332: \begin{figure}[htp]
333: \centering
334: \includegraphics[width=10cm,keepaspectratio,clip]{./figure1.eps}
335: \begin{center}
336: \caption{\label{fig1.fig} \footnotesize{(a) Local and (b)
337: Non-local measurements on a same MWNT at room temperature in air.
338: The schematic shows the measurement setup. The non-local
339: measurements are obtained with four electrodes that contact the
340: MWNT, while the local measurements are obtained with two
341: electrodes. The voltage $V$ is applied, while the current $I$ is
342: recorded for the local measurement and $\Delta V$ is recorded for
343: the non-local measurement}}
344: \end{center}
345: \end{figure}
346: \clearpage
347:
348: \clearpage
349: \begin{figure}[htp]
350: \centering
351: \includegraphics[width=12cm,keepaspectratio,clip]{./figure2.eps}
352: \caption{\label{fig2.fig}
353: \footnotesize{(a) Local current-voltage
354: characteristics for different MWNT devices. (b) Corresponding
355: differential conductances. All the possible slopes for $dI/dV-V$
356: can be obtained below about 2 V; $dI/dV-V$ can increase, remain
357: constant, or decrease.}}
358: \end{figure}
359: \clearpage
360:
361: \clearpage
362: \begin{figure}[htp]
363: \centering
364: \includegraphics[width=12cm,keepaspectratio,clip]{./figure3.eps}
365: \caption{\label{fig3.fig} \footnotesize{Slope of the local
366: differential conductance (V below about 2V) for more than 50 MWNT
367: devices as a function of diameter and length. The blue discs
368: correspond to positive slopes, the orange triangles pointing
369: downwards to constant slopes, and the red triangles pointing
370: upwards to negative slopes.}}
371: \end{figure}
372: \clearpage
373:
374: \clearpage
375: \begin{figure}[htp]
376: \centering
377: \includegraphics[width=14cm,keepaspectratio,clip]{./figure4.eps}
378: \caption{\label{fig4.fig} \footnotesize{(a) Schematic of a
379: multiple transmission line that describes transport of MWNTs. (b)
380: The calculated contribution of $dI/dV-V$ that results from the
381: scattering between electrons and optical phonons. (c) The
382: calculated contribution of $dI/dV-V$ that results from the Zener
383: tunneling. The tube is undoped and semiconducting. The enhancement
384: of $dI/dV-V$ corresponds to an enhancement of modes that
385: contribute significantly to transport. The differential
386: conductance is obtained from the Zener transmission
387: $T_{Z}=\exp(-\frac{4\sqrt{2m^*}LE^{3/2}}{3e\hbar V})$ with a shell
388: length $L=1$ $\mu m$, an effective mass $m^{*}=6.5 \times
389: 10^{-33}kg$ and a gap $E=53$ $meV$ $\cite{bourlonprl2004n1}$. The
390: values of $m^{*}$ and $E$ correspond to a diameter of $15$ $nm$.
391: (d) The number of current-carrying shells (and so $dI/dV$)
392: increases with the bias voltage. The behavior of $N_{shell}$ as a
393: function of $V$ is not known. The curves are drawn by hand.}}
394: \end{figure}
395: \clearpage
396:
397: \clearpage
398: \begin{figure}[htp]
399: \centering
400: \includegraphics[width=8cm,keepaspectratio,clip]{./figure5.eps}
401: \caption{\label{fig5.fig} \footnotesize{Schematic of the potential
402: variation in space. The boxes show the band diagram of a
403: semiconducting shell. (a) Zener tunneling from the top of the
404: valence subband to the bottom of the conduction subband. The shell
405: is undoped. The linear voltage drop along the tube results from
406: the short mean-free path that arises from the scattering between
407: electrons and optical phonons. (b) Low-bias regime for a doped
408: shell. (c) High-bias regime for a doped shell.}}
409: \end{figure}
410: \clearpage
411:
412:
413: \end{document}
414: