1: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
2: \documentclass[prb,twocolumn,showpacs,amsmath,amssymb,floatfix,noshowpacs]{revtex4}
3:
4: \usepackage[dvips]{graphicx}% Include figure files
5: \usepackage{dcolumn}% Align table columns on decimal point
6: \usepackage{bm}% bold math
7: \usepackage{times}
8:
9: %%
10: %% Float stuff
11: %%
12: \setlength{\floatsep}{4pt plus 4pt minus 4pt}
13: \setlength{\textfloatsep}{8pt plus 4pt minus 4pt}
14: \setlength{\dbltextfloatsep}{8pt plus 4pt minus 4pt}
15: %% \setlength{\intextsep}{0pt plus 2pt minus 2pt}
16: \setcounter{totalnumber}{4}
17: \setcounter{topnumber}{4}
18: \setcounter{bottomnumber}{4}
19: \renewcommand{\textfraction}{0.0}
20: \renewcommand{\topfraction}{1.0}
21: \renewcommand{\bottomfraction}{1.0}
22:
23: %\nofiles
24: %\addtolength{\voffset}{1.2cm} %shke: for letter paper
25: %\linespread{1.00}
26:
27: \begin{document}
28:
29: %\preprint{APS/123-QED}
30:
31: \title{
32: Nanotube-Metal Junctions: 2- and 3- Terminal Electrical Transport
33: }
34:
35: \author{ San-Huang Ke,$^{1,2}$ Weitao Yang,$^{1}$ and Harold U. Baranger$^{2}$}
36:
37: \affiliation{
38: $^{\rm 1}$Department of Chemistry, Duke University, Durham, NC 27708-0354
39: \\
40: $^{\rm 2}$Department of Physics, Duke University, Durham, NC 27708-0305
41: }
42:
43: \date{April 4, 2006; \textbf{J. Chem. Phys. 124, 181102 (2006)}; DOI: 10.1063/1.2200356}% It is always \today, today,
44: % but any date may be explicitly specified
45:
46: \begin{abstract}
47: We address the quality of electrical contact between carbon nanotubes and
48: metallic electrodes by performing first-principles calculations for the
49: electron transmission through ideal 2- and
50: 3-terminal junctions, thus revealing the physical limit of tube-metal
51: conduction.
52: The structural model constructed involves surrounding the tube by the metal
53: atoms of
54: the electrode as in most experiments;
55: we consider metallic (5,5) and \textit{n}-doped semiconducting
56: (10,0) tubes surrounded by Au or Pd. In the case of metallic tubes, the contact
57: conductance is shown to approach the ideal $4e^2/h$ in the limit of large
58: contact area.
59: For three-terminals, the division of flux among the different transmission
60: channels depends strongly on the metal material. A Pd electrode has nearly
61: perfect tube-electrode transmission and therefore turns off the straight
62: transport along the tube. Our results are in good agreement with some recent
63: experimental reports and clarify a fundamental discrepancy between theory and
64: experiment.
65: \end{abstract}
66:
67: %\pacs{73.40.Cg, 72.10.-d, 85.65.+h}
68: \maketitle
69:
70: %\setstretch{1}
71: %\section{Introduction}
72:
73: Carbon nanotubes (CNTs) have received extensive experimental attention for more
74: than a decade \cite{Avouris04403,Mceuen04272,Heer04281}, and are considered a
75: possible basis for nanoelectronic technology independent of silicon. A major
76: issue is the quality of CNT/metal contacts: obtaining the minimum contact
77: resistance is critical to access the intrinsic electric properties of CNTs.
78: Despite extensive experimental effort to improve the contact transparency and
79: reveal the relevant factors behind it -- metal material, contact structure, and
80: type of tube, for instance -- a clear picture is still not available.
81:
82: On the theoretical side, it is highly desirable to be able to simulate from the
83: first principles the
84: electron transport through CNT/metal junctions and thus to improve our
85: understanding of this important issue. So far, first-principles
86: studies of the contact transparency between metallic
87: CNTs and metals have been carried out [(3,3), (4,4), or (5,5) tubes with Al,
88: Au, or Ti, for instance]
89: \cite{Taylor01245407,Nardelli01245423,Palacios03106801,Liu03193409}. The
90: results are, however, quite scattered, and agreement between theory and
91: experiment has not yet been achieved: For Al electrodes, an equilibrium
92: conductance of $\sim\!1\,G_0$ ($=2e^2/h$, the conductance quanta) was found
93: by some calculations \cite{Taylor01245407,Palacios03106801} while
94: $\sim\!2\,G_0$ was found by another \cite{Nardelli01245423}.
95: For Ti electrodes, $1.7\,G_0$ was obtained by one calculation
96: \cite{Liu03193409} and $\sim\!1.2\,G_0$ by another \cite{Palacios03106801},
97: while experiment \cite{Kong01106801} found $\sim\!2\,G_0$.
98: For Au electrodes, a recent calculation \cite{Palacios03106801} showed
99: $\sim\!1\,G_0$ while a value of $1.5\,G_0$ was found experimentally
100: \cite{Nygard00342}.
101: On the other hand, a model calculation \cite{Choi99R14009} using a jellium model for the electrode
102: even showed that the ideal CNT/metal conductance will not be larger than
103: $1\,G_0$.
104: With regard to the metal used, it was found theoretically that Ti is better
105: than either Au or Al for contact transparency \cite{Palacios03106801}. On the
106: other hand, experimentally, Pd, for which no theoretical calculation is
107: available, was found superior to Ti \cite{Mann031541}.
108:
109: It has been unclear how to explain these discrepancies.
110: %, made especially puzzling because the ideal contacts used theoretically
111: %suggest that the calculations should yield upper bounds.
112: One possible reason, as is supported by this work,
113: is the difference in contact structure between theory and
114: experiment. In all these calculations, simplified contact models were adopted
115: due to computational cost or methodology: On the carbon side, the tube-metal
116: connection is made by either straight $\sigma$-bonds from the tube end or
117: $\pi$-bonds from the side of the tube, while the metal is modeled by a thin
118: nanowire or small cluster. However, in most experimental situations, a tube is
119: surrounded by metal atoms \cite{Soh99627,Martel01159,McEuen0278,
120: Wind03058301,Yaish04046401,Biercuk041}.
121: As the contact structure and quality changes from case
122: to case, experimental results are also scattered. In such a situation, the key
123: contribution to be made by theory is to reveal the physical limit to which
124: experimental measurement may approach by improving the contact quality.
125:
126: Furthermore, compared to metallic CNTs, semiconducting CNTs are much more
127: important for potential electronic applications. This naturally raises the
128: issue of contact transparency between doped semiconducting tubes and metal
129: junctions. In addition, electron transport through multi-terminal structures is
130: a key property in moving toward applications. Neither of these fundamental
131: issues has been tackled previously, to the best of our knowledge, using
132: first-principles calculations.
133:
134: \begin{figure}[b]
135: \includegraphics[angle= 0,width=8.0cm]{str.eps}
136: \caption{Side and top views of the device region of a (5,5)-tube-Au junction,
137: where the Au electrode consists of 7 atomic layers (7L).
138: The length of the tube in the device region is 32.6{\AA} and
139: the dimensions in the lateral directions are 20.4 and 22.44{\AA},
140: respectively.
141: The three terminals are denoted by t$_1$, t$_2$, and t$_3$.
142: }
143: \label{fig_str}
144: \end{figure}
145:
146: \begin{figure*}[t]
147: %\includegraphics[angle= 0,width=8.5cm]{t_2t5x5.eps}
148: \includegraphics[angle= 0,width=11cm]{t_2t5x5.eps}
149: \caption{Two-terminal transmission $T_{\rm 2term.}(E)$ (tube to metal) for (a) (5,5)-Au
150: and (b) (5,5)-Pd with different widths of the metal electrode, as indicated in
151: the legends. Also plotted is the transmission for a pure (5,5) tube, as a
152: comparison.}
153: \label{fig_t2t5x5}
154: \end{figure*}
155:
156: In this paper, we first construct a better
157: \cite{Soh99627,Martel01159,McEuen0278,Wind03058301,Yaish04046401,Biercuk041}
158: structural model of an ideal contact, as shown in Fig.~\ref{fig_str} for the
159: device region, in which
160: an infinitely long tube is surrounded by metal atoms of an electrode.
161: %In this model, the tube is infinitely long, eliminating the quantization in
162: %the transport direction which bedevils calculations using a short CNT segment
163: %\cite{Palacios03106801}. The large active region surrounding the contact
164: %ensures that the tube-metal interaction is fully included.
165: We use this structural model, first, to investigate the two-terminal tube-metal
166: transparency, $T_{\rm 2term.}$ (from terminal 1 to 3), obtaining the physical limit of
167: tube-metal conductance for an ideal contact. Second, we then study
168: three-terminal electron transport by including explicitly the self-energies of
169: all the leads shown in Fig.~\ref{fig_str}. Specifically, we investigate the
170: division of current among transmission straight-through the junction,
171: probability to turn the corner, and reflection. In both 2- and 3- terminal
172: cases, we consider the dependence on the contact quality (simulated by the
173: width of the metal electrode), the metal material, and the type of tube
174: (metallic or semiconducting).
175:
176: %Our calculations show that the conductance of 2t metallic tube-metal junctions
177: %will be approaching 2$G_0$ with gradually improved contact quality, while that
178: %of doped semiconducting tube systems is around $G_0$. The convergence,
179: %however, depends strongly on the metal material and Pd is better than Au. For
180: %the 3t systems, the division of the different component transports depends
181: %strongly on the metal material, and the `better' Pd electrode will have near
182: %perfect side transport and therefore turns off the straight transport through
183: %the wraped tube. The phsical limit of 2$G_0$ for the 2t conductance is
184: %consistent with recent experimental findings \cite{Kong01106801,Mann031541}.
185: %The dominant side transport in the 3t-Pd system is in good agreement with a
186: %recent experimental report which showed that the transport through the tube
187: %will be turned off by a Pd electrode \cite{Mann031541}.
188:
189: The structural model (Fig.~\ref{fig_str}) can be regarded as an ideal CNT-metal
190: junction because of three features: (1)~The tube is perfect and infinitely
191: long, thus eliminating the quantization in the transport direction which
192: bedevils calculations \cite{Palacios03106801} using a short CNT segment.
193: %in which several different tube lengths must be studied in order to approximately show the true behavior.
194: (2)~The tube is surrounded by metal atoms, as is usual in experiments
195: \cite{Soh99627,Martel01159,McEuen0278,Wind03058301,Yaish04046401,Biercuk041}.
196: (3)~The large active region around the countact
197: ensures that the tube-metal interaction is fully included.
198:
199: Contact quality will depend experimentally on the number of good carbon-metal
200: connections, which is essentially determined by whether the metal wets the CNT
201: surface. If the metal (like Au) does not wet the CNT surface, there will be few
202: connections even though the electrode is large. In our mod\-els, because of
203: the ideal structure, the number of good car\-bon-metal connections is
204: substantial even for a small electrode (Fig.~\ref{fig_str}, top view). As a
205: proxy, we change the number of good carbon-metal connections by adjusting the
206: contact width, in this way simulating the changing contact
207: quality.
208:
209: We consider three widths for the metal electrodes -- 3, 5, and 7 atomic layers
210: (denoted by 3L, 5L, and 7L) -- and two kinds of metals -- Au and Pd.
211: The electronic states of the two electrodes are very different: Au has active \textit{s}
212: states while Pd has only \textit{d} states. We consider three kinds of CNTs:
213: (5,5), (10,0), and Na-doped (10,0) -- denoted Na@(10,0) hereafter -- in which
214: dopant Na atoms are adsorbed periodically on the inner surface of the tube.
215: [The adsorption position is determined by minimizing the atomic force on the
216: Na atom and is found to be slightly off the tube axis and
217: approximately above one of the C atoms.]
218: The concentration is one Na atom per 40 C atoms. Because of the large charge
219: transfer from Na to the tube ($\sim\!0.7$ electrons per Na atom by Mulliken
220: population analysis), the (10,0) tube is therefore heavily \textit{n}-doped.
221:
222: The electronic structure of the junctions is calculated by density functional
223: theory (DFT) using periodic boundary conditions and a localized basis set with
224: a finite range \cite{Soler022745}.
225: In the device region, as shown in Fig.~\ref{fig_str}, the number
226: of atoms included is $\sim$ 350 -- 500, depending on the
227: metal width and material.
228: The large lateral dimensions (see Fig.~\ref{fig_str} (top view)) ensure
229: that the separation among the tube and its
230: images is larger than 12{\AA} which is much larger than the range of the basis
231: functions used. As a result, a good convergence with respect to the lateral
232: dimensions can be expected. The convergence with respect to the length of the
233: tube in the device region is also found good, as will be discussed later.
234:
235: We adopt a DFT supercell which is larger than the
236: device region shown in Fig.~\ref{fig_str} for all the different systems: $\sim$
237: 500 -- 700 atoms are included, depending on the electrode width and material.
238: To avoid too expensive computational cost,
239: a single-zeta plus polarization basis set (SZP) is adopted
240: for all atomic species. Our test calculation for a small system shows that the
241: use of SZP results in only minor differences from results using a higher-level
242: double-zeta plus polarization basis set. We use optimized Troullier-Martins
243: pseudopotentials \cite{Troullier911993} for the atomic cores and the PBE
244: version of the generalized gradient approximation \cite{Perdew963865} for the
245: electron exchange and correlation. The contact atomic structure is optimized by
246: eliminating the atomic forces on the atoms around the contact region, i.e.,
247: the carbon atoms underneath the metal and the metal atoms contacting the tube.
248: The relaxation is found quite small (see Fig.~\ref{fig_str} (top view) for
249: the (5,5)-Au system) because of the choices of the tubes, (5,5) and (10,0), and
250: the orientation of the metals, (001).
251:
252: Electron transport through the CNT-metal junction is calculated using a
253: Green function method \cite{Datta95,Ke04085410}, in which the device region and
254: the leads (not shown in Fig.~\ref{fig_str}) are treated exactly on the same
255: footing \cite{Ke04085410}. The retarded Green function of the
256: device region, $\mathbf{G}_{D}(E)$, is determined by the Hamiltonian given by
257: DFT [$\mathbf{H}_{D}$] combined with the self-energies for the semi-infinite
258: leads [${\Sigma}_{i}(E)$]:
259: \begin{equation}
260: \mathbf{G}_{D}(E) = \Big[ E\mathbf{S}_{D} - \mathbf{H}_{D}
261: -\sum_{i=1}^{N_t} {\Sigma}_{i}(E) \Big]^{-1}, \label{equ_g}
262: \end{equation}
263: where $N_t$ is the number of leads or terminals. The transmission at any
264: energy, $T(E)$, is calculated from the Green function, and the conductance,
265: $G$, then follows from a Landauer-type relation. We adopt the following
266: notation:
267: $T_{\rm 2term.}$ is the two-terminal transmission between tube and metal (terminals 1
268: and 3) \textit{without} electrode 2 attached, which means that
269: the tube on the right side of Fig.1 (side view) is somehow terminated and all the
270: electron wave entering that part of tube will be totally reflected.
271: $T_{\rm straight}$ is the transmission
272: straight through the junction from tube to tube, $T_{\rm turn}$ is the tube-to-metal
273: transmission \textit{with} the presence of electrode 2, and $R$ is reflection
274: from the tube back to itself.
275:
276: %$T_{13}$ stands for the transmission coefficient from t$_1$ to t$_3$ (tube to
277: %metal) {\it without} the presence of t$_2$ (i.e., $\Sigma_{2}$ in
278: %Eq.~\ref{equ_g}); $T_{ij+k}$ stands for the transmission coefficient from
279: %t$_i$ to t$_j$ {\it with} the presence of t$_k$ (i.e., $N_t$=3 in
280: %Eq.~\ref{equ_g}); $R$ is the reflection coefficient in the 3t systems.
281:
282: \begin{figure*}[t]
283: \includegraphics[angle= 0,width=11.6cm]{t_3t5x5_357L.eps}
284: \caption{Three-terminal transmission functions $T_{\rm straight}(E)$,
285: $T_{\rm turn}(E)$, and $R(E)$
286: for (5,5)-Au [first column, (a)-(c)] and (5,5)-Pd [second column, (d)-(f)]. The
287: width of the metal electrode is indicated in the legends. Also plotted is the
288: transmission function of a pure (5,5) tube as a comparison (dashed). The Pd
289: traces converge much more quickly than those for Au.}
290: \label{fig_t3t5x5}
291: \end{figure*}
292:
293: In Fig.~\ref{fig_t2t5x5} we show the transmission $T_{\rm 2term.}(E)$ for the (5,5)-Au
294: and (5,5)-Pd junctions with different widths of the metal electrodes. As a
295: comparison, $T(E)$ for the pure (5,5) tube is also plotted; it shows, as
296: expected, perfect steps and two channels around the Fermi energy. Note that
297: throughout this work the position of the Fermi energy is set to be exactly the
298: same as in the pure tube. This is required since very far from the device
299: region the tube (lead) is completely charge neutral.
300:
301: In Fig.~\ref{fig_t2t5x5}, the dependence of $T_{\rm 2term.}(E)$ on the electrode width
302: for the two metals is clear: For Au, the dependence is strong. In contrast,
303: for Pd the dependence is much weaker: the overall shape of $T_{\rm 2term.}(E)$ for the
304: three widths is similar, and the result is nearly the same for 5L and 7L.
305: This result for Pd also indicates that in our calculation the convergence with
306: respect to the length of the tube included in the device region is good.
307: Despite the difference in width dependence, the equilibrium conductance, $G$,
308: of both systems increases with increasing electrode width: for the thinnest,
309: hence poor quality, electrode $G \!\sim\! G_0$, while it approaches the
310: physical limit of $2\,G_0$ for the widest electrode.
311:
312: Even the widest electrode (7L) considered here is, of course, still much
313: thinner than those used in experiments. However, our calculation bears on real
314: experimental situations because the number of good carbon-metal connections may
315: be similar in that the contact structure considered here is perfect while in
316: real experimental situations it is usually far from perfect. For example, Au is
317: thought not to wet the tube surface but rather will form nanoparticles near the
318: surface \cite{JieLiu}. As a result, although the Au electrode can be very wide,
319: the tube will pass through the space between these Au nanoparticles with very
320: few good C-Au connections formed. In this case increasing the diameter of the
321: tube will improve the contact quality as more C-Au connections will be formed
322: and, therefore, increase the conductance. In fact, a recent experiment using
323: tubes with different diameters \cite{Javey03654} shows that a larger diameter
324: yields a larger conductance. Obviously, increasing the electrode width here is
325: similar to increasing the tube diameter: both increase the number of C-Au
326: connections and so increase the conductance. \textit{A key result here is that
327: for good contact quality, the equilibrium conductance approaches the physical
328: limit of 2 $G_0$ for both Au and Pd.} This finding differs from a previous
329: first-principles calculation \cite{Palacios03106801} which used a short segment of tube
330: contacted by two Au electrodes through several carbon-$\pi$ bonds, and also
331: differs from a previous model calculation \cite{Choi99R14009} which shows that CNT/metal
332: conductance will not be larger than 1 $G_0$, but is
333: consistent with recent experiments \cite{Kong01106801,Mann031541}.
334:
335: %\begin{figure}[t]
336: %\includegraphics[angle= 0,width=6.0cm]{t_3t5x5_7L.eps}
337: %\caption{Three-terminal transmission functions $T_{\rm straight}(E)$,
338: %$T_{\rm turn}(E)$, and $R(E)$
339: %for (a) (5,5)-Au and (b) (5,5)-Pd for the best quality electrodes (7L). The
340: %transmission for a pure (5,5) tube is plotted for comparison (dashed). At the
341: %Fermi energy, the transmission is apportioned among all possibilities in the
342: %case of Au. In contrast, note the striking dominance of $T_{\rm turn}$ for a Pd
343: %electrode.}
344: %\label{fig_t3t5x5a}
345: %\end{figure}
346:
347: \begin{figure*}[t]
348: \includegraphics[angle= 0,width=12cm]{t_2t10x0.eps}
349: \caption{Two-terminal transmission, $T_{\rm 2term.}(E)$, with semiconducting tubes: (a)
350: Na@(10,0)-Au and (b) Na@(10,0)-Pd systems with an electrode width of 5L.
351: The transmission function of the bare Na@(10,0) tube is shown in each panel for
352: comparison; in addition, (a) includes $T_{\rm 2term.}(E)$ for the undoped (10,0)-Au
353: system.}
354: \label{fig_t2t10x0}
355: \end{figure*}
356:
357: The three-terminal transmission in the (5,5)-metal case is shown in
358: Fig.~\ref{fig_t3t5x5} for the different metal widths. [Note that reflection
359: $R(E) \!=\! T_{\rm tube}(E) \!-\! T_{\rm straight}(E) \!-\! T_{\rm turn}(E)$.]
360: For the widest case (7L) (see Figs.~\ref{fig_t3t5x5} (a) and (d)) a striking feature
361: of the result for Pd is that the straight-through transmission is almost turned
362: off while transmission into the
363: metal electrode dominates. This feature is in very good agreement with a recent
364: experiment \cite{Mann031541} showing that a Pd electrode suppresses inner
365: electron transport through the tube. For Au, $T_{\rm turn}$ is also larger
366: than $T_{\rm straight}$ but they are still comparable and the latter is not turned
367: off.
368:
369: %Turning to the dependence of the three-terminal transmission on the electrode
370: %width, Figs.~\ref{fig_t3t5x5a} and
371: In Figs.~\ref{fig_t3t5x5}, we see that for Au the
372: apportioning of transmission among the different components depends
373: significantly on the electrode width: For small width, $T_{\rm straight}(E)$
374: and $T_{\rm turn}(E)$
375: are comparable around the Fermi energy, while for the widest electrode (7L),
376: $T_{\rm turn}(E)$ begins to dominate due to the improved contact quality. In the Pd
377: system, the convergence as a function of width is much faster (as for the
378: two-terminal case Fig.~\ref{fig_t2t5x5}); indeed, one sees only minor changes
379: between the curves.
380: This quick convergence indicates that the carrier injection takes place mainly
381: at the very edge of the junction, being consistent with the experimental
382: observation \cite{Mann031541}.
383: Thus for Pd, the contact quality is already good (i.e., the width is large
384: enough) even
385: though the junction is very thin. Since the contact structure is the same, this
386: difference between Au and Pd electrodes is due to their different electronic
387: states: The Pd \textit{d} states have stronger interaction with C \textit{p}
388: states than Au \textit{s-d} states. As a result, Pd is a better electrode
389: material than Au.
390:
391: %In Fig.~\ref{fig_t3t5x5} it turns out that for both junctions the reflection
392: %is small in the 2-channel gap but always jumps up at the edges. --
393: %explanation?
394:
395: Two additional features of Figs.~\ref{fig_t3t5x5} are
396: worth comments. First, $T_{\rm turn}(E)$ is always smaller than the two-terminal result
397: $T_{\rm 2term.}(E)$ of Fig.~\ref{fig_t2t5x5}. This is reasonable since in the
398: three-terminal case some of the initial flux escapes into electrode 2 while for
399: two terminals that flux will be totally reflected. Second, the reflection
400: increases in all cases near the energies where more modes start to propagate in
401: the pure tube. These are examples of threshold singularities
402: \cite{LandauLifshitzQM}.
403:
404: %\begin{figure}[t]
405: %\includegraphics[angle= 0,width=8.5cm]{t_3t5x5_357L.eps}
406: %\caption{Three-terminal transmission functions $T_{\rm straight}(E)$,
407: %$T_{\rm turn}(E)$, and $R(E)$
408: %for (5,5)-Au [first column, (a)-(c)] and (5,5)-Pd [second column, (d)-(f)]. The
409: %width of the metal electrode is indicated in the legends. Also plotted is the
410: %transmission function of a pure (5,5) tube as a comparison (dashed). The Pd traces
411: %converge much more quickly than those for Au.}
412: %\label{fig_t3t5x5}
413: %\end{figure}
414:
415: We now turn from metallic tubes to discussing the doped (10,0) semiconducting
416: tube. The doping is achieved by adding Na atoms periodically on the inner
417: surface of the (10,0) tube with a high concentration as described earlier.
418: %To save the computational cost, here we consider only one width (5L) of the metal electrodes.
419:
420: The two terminal transmission is shown in Fig.~\ref{fig_t2t10x0} for both Au
421: and Pd electrodes. For comparison, the transmission is shown for two additional
422: cases: $T_{\rm 2term.}(E)$ for the Na@(10,0) tube alone and for the undoped (10,0)-Au
423: system. For the undoped (10,0)-Au system, the Fermi energy is at the middle of
424: the gap because one of the leads is the semiconducting (10,0) tube. After the
425: tube is \textit{n}-doped, its Fermi energy enters the conduction band, and the
426: transmission through the Na@(10,0) tube in the absence of any metal has a value
427: of 2 around the Fermi energy. As in the metallic case, here the Pd electrode
428: results in a larger two-terminal conductance ($0.79\,G_0$) than the Au
429: electrode ($0.56\,G_0$). Both are smaller than the two-terminal conductance of
430: the (5,5) systems.
431:
432: An interesting feature of the results in Fig.~\ref{fig_t2t10x0} is that
433: $T_{\rm 2term.}(E)$ for the doped Na@(10,0)-Au case is approximately a constant shift of
434: that for the undoped (10,0)-Au system. Thus the role of the dopants is mainly
435: to give electrons to the tube, and therefore to shift its Fermi energy, without
436: noticeably altering the electronic structure of the tube. Our calculations
437: indicate this is a general rule of thumb. Following this idea, we can
438: approximately get the two-terminal conductance of a $p$-doped (10,0)-metal
439: junction by simply shifting the Fermi energy slightly into the valence band (to
440: around $-1$\,eV in Fig.~\ref{fig_t2t10x0}. For a Pd electrode, this $p$-type
441: conductance is about $1\,G_0$ which is in good agreement with a recent
442: experimental report where the Fermi energy of a semiconducting CNT contacted by
443: Pd electrodes was shifted into the valence band by applying a back gate [see
444: Fig. 1 (c) in Ref.~\cite{Javey03654}].
445:
446: \begin{figure*}[t]
447: \includegraphics[angle= 0,width=12cm]{t_3t10x0.eps}
448: \caption{Three-terminal case with semiconducting tubes: $T_{\rm straight}(E)$,
449: $T_{\rm turn}(E)$,
450: and $R(E)$ for (a) Na@(10,0)-Au and (b) Na@(10,0)-Pd with an electrode width of
451: 5L. For comparison, the transmission function of the Na@(10,0) tube is also
452: shown.}
453: \label{fig_t3t10x0}
454: \end{figure*}
455:
456:
457: The three-terminal transmission functions for the semiconducting tubes cases
458: are shown in Fig.~\ref{fig_t3t10x0}. For Na@(10,0)-Au, the straight-through and
459: tube-metal transmission around the Fermi energy are comparable, just like for
460: the corresponding (5,5) system [Fig.~\ref{fig_t3t5x5} (b)]. Both of them are,
461: however, much smaller than the reflection because the Fermi energy now is at
462: the edge of the gap. For a Pd electrode, on the other hand, $T_{\rm turn}$ is
463: comparable to the reflection, near the Fermi energy, and both of them are much
464: larger then the straight-through transport. Therefore, for an $n$-doped
465: semiconducting (10,0) tube, the Pd electrode turns off conduction through the
466: tube, just like for the metallic (5,5) tube.
467:
468: \newpage
469: In summary, we have calculated transmission through both 2- and 3- terminal
470: CNT-metal junctions using an ideal structure in which the tube is surrounded by
471: a Au or Pd electrode. In this way we have established the physical limit to
472: which experiments will approach by improving the contact quality.
473:
474: The main quantities studied are the contact transparency in the case of
475: two-terminal systems and the division among the different transmission
476: coefficients for three-terminals.
477: %; their dependence on the contact quality, metal, and type of tube have been systematically investigated.
478: The two main findings are:
479:
480: (1) For two-terminal junctions, when the tube is metallic the conductance will
481: approach $2\,G_0$ as the contact quality improves, while for the doped
482: semiconducting tube systems, the conductance is about $1\,G_0$. In both cases,
483: Pd is better than Au for contact transparency.
484:
485: (2) For three terminal junctions, the relative magnitude of the different types
486: of transmission depends significantly on the metal material. The ``better'' Pd
487: electrode yields near perfect transmission between the tube and electrode and,
488: therefore, turns off the straight transport through the tube for both the (5,5)
489: and Na@(10,0) tubes.
490:
491: %Our results for the two-terminal Au and three-terminal Pd systems are in good
492: %agreement with recent experiments.
493: Our results are in good agreement with recent experiments
494: and clarify a fundamental discrepancy between theory and
495: experiment.
496:
497: {\bf Acknowledgments.}
498: This work was supported in part by the NSF (DMR-0506953).
499:
500: \begin{thebibliography}{25}
501: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
502: \expandafter\ifx\csname bibnamefont\endcsname\relax
503: \def\bibnamefont#1{#1}\fi
504: \expandafter\ifx\csname bibfnamefont\endcsname\relax
505: \def\bibfnamefont#1{#1}\fi
506: \expandafter\ifx\csname citenamefont\endcsname\relax
507: \def\citenamefont#1{#1}\fi
508: \expandafter\ifx\csname url\endcsname\relax
509: \def\url#1{\texttt{#1}}\fi
510: \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi
511: \providecommand{\bibinfo}[2]{#2}
512: \providecommand{\eprint}[2][]{\url{#2}}
513:
514: \bibitem[{\citenamefont{Avouris}(2004/June)}]{Avouris04403}
515: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Avouris}},
516: \bibinfo{journal}{MRS Bulletin} \textbf{\bibinfo{volume}{29 (6)}},
517: \bibinfo{pages}{403 } (\bibinfo{year}{2004/June}).
518:
519: \bibitem[{\citenamefont{McEuen and Park}(2004/April)}]{Mceuen04272}
520: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{McEuen}} \bibnamefont{and}
521: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Park}}, \bibinfo{journal}{MRS
522: Bulletin} \textbf{\bibinfo{volume}{29 (4)}}, \bibinfo{pages}{272}
523: (\bibinfo{year}{2004/April}).
524:
525: \bibitem[{\citenamefont{de~Heer}(2004/April)}]{Heer04281}
526: \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{de~Heer}},
527: \bibinfo{journal}{MRS Bulletin} \textbf{\bibinfo{volume}{29 (4)}},
528: \bibinfo{pages}{281} (\bibinfo{year}{2004/April}).
529:
530: \bibitem[{\citenamefont{Taylor et~al.}(2001)\citenamefont{Taylor, Guo, and
531: Wang}}]{Taylor01245407}
532: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Taylor}},
533: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Guo}}, \bibnamefont{and}
534: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Wang}},
535: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{63}},
536: \bibinfo{pages}{245407 (13 pages)} (\bibinfo{year}{2001}).
537:
538: \bibitem[{\citenamefont{Nardelli et~al.}(2001)\citenamefont{Nardelli,
539: Fattebert, and Bernholc}}]{Nardelli01245423}
540: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Nardelli}},
541: \bibinfo{author}{\bibfnamefont{J.-L.} \bibnamefont{Fattebert}},
542: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Bernholc}},
543: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{64}},
544: \bibinfo{pages}{245423 (5 pages)} (\bibinfo{year}{2001}).
545:
546: \bibitem[{\citenamefont{Palacios et~al.}(2003)\citenamefont{Palacios,
547: P\'{e}rez-Jim\'{e}nez, Louis, SanFabi\'{a}n, and
548: Verg\'{e}s}}]{Palacios03106801}
549: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Palacios}},
550: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{P\'{e}rez-Jim\'{e}nez}},
551: \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Louis}},
552: \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{SanFabi\'{a}n}},
553: \bibnamefont{and}
554: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Verg\'{e}s}},
555: \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{90}},
556: \bibinfo{pages}{106801 (4 pages)} (\bibinfo{year}{2003}).
557:
558: \bibitem[{\citenamefont{Liu}(2003)}]{Liu03193409}
559: \bibinfo{author}{\bibfnamefont{Y.}~\bibnamefont{Liu}}, \bibinfo{journal}{Phys.
560: Rev. B} \textbf{\bibinfo{volume}{68}}, \bibinfo{pages}{193409 (4 pages)}
561: (\bibinfo{year}{2003}).
562:
563: \bibitem[{\citenamefont{Kong et~al.}(2001)\citenamefont{Kong, Yenilmez,
564: Tombler, Kim, Dai, Laughlin, L.~Liu, and Wu}}]{Kong01106801}
565: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Kong}},
566: \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Yenilmez}},
567: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Tombler}},
568: \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{Kim}},
569: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Dai}},
570: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Laughlin}},
571: \bibinfo{author}{\bibfnamefont{C.~J.} \bibnamefont{L.~Liu}},
572: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Wu}},
573: \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{87}},
574: \bibinfo{pages}{106801 (4 pages)} (\bibinfo{year}{2001}).
575:
576: \bibitem[{\citenamefont{Nyg{\aa}rd et~al.}(2000)\citenamefont{Nyg{\aa}rd,
577: Cobden, and Lindelof}}]{Nygard00342}
578: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Nyg{\aa}rd}},
579: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Cobden}}, \bibnamefont{and}
580: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Lindelof}},
581: \bibinfo{journal}{Nature} \textbf{\bibinfo{volume}{408}}, \bibinfo{pages}{342
582: } (\bibinfo{year}{2000}).
583:
584: \bibitem[{\citenamefont{Choi et~al.}(1999)\citenamefont{Choi, Ihm, Yoon, and
585: Louie}}]{Choi99R14009}
586: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Choi}},
587: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Ihm}},
588: \bibinfo{author}{\bibfnamefont{Y.-G.} \bibnamefont{Yoon}}, \bibnamefont{and}
589: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Louie}},
590: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{60}},
591: \bibinfo{pages}{R14009 } (\bibinfo{year}{1999}).
592:
593: \bibitem[{\citenamefont{Mann et~al.}(2003)\citenamefont{Mann, Javey, Kong,
594: Wang, and H.~Dai}}]{Mann031541}
595: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Mann}},
596: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Javey}},
597: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Kong}},
598: \bibinfo{author}{\bibfnamefont{Q.}~\bibnamefont{Wang}}, \bibnamefont{and}
599: \bibinfo{author}{\bibfnamefont{N.}~\bibnamefont{H.~Dai}},
600: \bibinfo{journal}{Nano Letters} \textbf{\bibinfo{volume}{3}},
601: \bibinfo{pages}{1541 } (\bibinfo{year}{2003}).
602:
603: \bibitem[{\citenamefont{Soh et~al.}(1999)\citenamefont{Soh, Quate, Morpurgo,
604: Marcus, Kong, and Dai}}]{Soh99627}
605: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Soh}},
606: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Quate}},
607: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Morpurgo}},
608: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Marcus}},
609: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Kong}}, \bibnamefont{and}
610: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Dai}},
611: \bibinfo{journal}{Appl. Phys. Lett.} \textbf{\bibinfo{volume}{75}},
612: \bibinfo{pages}{627 } (\bibinfo{year}{1999}).
613:
614: \bibitem[{\citenamefont{Martel et~al.}(2001)\citenamefont{Martel, Wong, Chan,
615: and Avouris}}]{Martel01159}
616: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Martel}},
617: \bibinfo{author}{\bibfnamefont{H.-S.} \bibnamefont{Wong}},
618: \bibinfo{author}{\bibfnamefont{K.}~\bibnamefont{Chan}}, \bibnamefont{and}
619: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Avouris}}, in
620: \emph{\bibinfo{booktitle}{Proc. IEDM 2001}} (\bibinfo{publisher}{Institute of
621: Electrical and Electronics Engineers}, \bibinfo{address}{Piscataway, NJ},
622: \bibinfo{year}{2001}), p. \bibinfo{pages}{159}.
623:
624: \bibitem[{\citenamefont{McEuen et~al.}(2002)\citenamefont{McEuen, Fuhrer, and
625: Park}}]{McEuen0278}
626: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{McEuen}},
627: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Fuhrer}}, \bibnamefont{and}
628: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Park}},
629: \bibinfo{journal}{IEEE Trans. Nanotechnol.} \textbf{\bibinfo{volume}{1}},
630: \bibinfo{pages}{78 } (\bibinfo{year}{2002}).
631:
632: \bibitem[{\citenamefont{Wind et~al.}(2003)\citenamefont{Wind, Appenzeller, and
633: Avouris}}]{Wind03058301}
634: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Wind}},
635: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Appenzeller}},
636: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Avouris}},
637: \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{91}},
638: \bibinfo{pages}{058301 (4 pages)} (\bibinfo{year}{2003}).
639:
640: \bibitem[{\citenamefont{Yaish et~al.}(2004)\citenamefont{Yaish, Park,
641: Rosenblatt, Sazonova, Brink, and McEuen}}]{Yaish04046401}
642: \bibinfo{author}{\bibfnamefont{Y.}~\bibnamefont{Yaish}},
643: \bibinfo{author}{\bibfnamefont{J.-Y.} \bibnamefont{Park}},
644: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Rosenblatt}},
645: \bibinfo{author}{\bibfnamefont{V.}~\bibnamefont{Sazonova}},
646: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Brink}}, \bibnamefont{and}
647: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{McEuen}},
648: \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{92}},
649: \bibinfo{pages}{046401 (4 pages)} (\bibinfo{year}{2004}).
650:
651: \bibitem[{\citenamefont{Biercuk et~al.}(2004)\citenamefont{Biercuk, Mason, and
652: Marcus}}]{Biercuk041}
653: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Biercuk}},
654: \bibinfo{author}{\bibfnamefont{N.}~\bibnamefont{Mason}}, \bibnamefont{and}
655: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Marcus}},
656: \bibinfo{journal}{Nano Letters} \textbf{\bibinfo{volume}{4}},
657: \bibinfo{pages}{1 } (\bibinfo{year}{2004}).
658:
659: \bibitem[{\citenamefont{Soler et~al.}(2002)\citenamefont{Soler, Artacho, Gale,
660: Garc\'{i}a, Junquera, Ordej\'{o}n, and S\'{a}nchez-Portal}}]{Soler022745}
661: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Soler}},
662: \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Artacho}},
663: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Gale}},
664: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Garc\'{i}a}},
665: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Junquera}},
666: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Ordej\'{o}n}},
667: \bibnamefont{and}
668: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{S\'{a}nchez-Portal}},
669: \bibinfo{journal}{J. Phys.:Condens. Matter} \textbf{\bibinfo{volume}{14}},
670: \bibinfo{pages}{2745} (\bibinfo{year}{2002}).
671:
672: \bibitem[{\citenamefont{Troullier and Martins}(1991)}]{Troullier911993}
673: \bibinfo{author}{\bibfnamefont{N.}~\bibnamefont{Troullier}} \bibnamefont{and}
674: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Martins}},
675: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{43}},
676: \bibinfo{pages}{1993} (\bibinfo{year}{1991}).
677:
678: \bibitem[{\citenamefont{Perdew et~al.}(1996)\citenamefont{Perdew, Burke, and
679: Ernzerhof}}]{Perdew963865}
680: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Perdew}},
681: \bibinfo{author}{\bibfnamefont{K.}~\bibnamefont{Burke}}, \bibnamefont{and}
682: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Ernzerhof}},
683: \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{77}},
684: \bibinfo{pages}{3865} (\bibinfo{year}{1996}).
685:
686: \bibitem[{\citenamefont{Datta}(1995)}]{Datta95}
687: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Datta}},
688: \emph{\bibinfo{title}{Electronic Transport in Mesoscopic Systems}}
689: (\bibinfo{publisher}{Cambridge University Press},
690: \bibinfo{address}{Cambridge, England}, \bibinfo{year}{1995}).
691:
692: \bibitem[{\citenamefont{Ke et~al.}(2004)\citenamefont{Ke, Baranger, and
693: Yang}}]{Ke04085410}
694: \bibinfo{author}{\bibfnamefont{S.-H.} \bibnamefont{Ke}},
695: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Baranger}}, \bibnamefont{and}
696: \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{Yang}},
697: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{70}},
698: \bibinfo{pages}{085410 (12 pages)} (\bibinfo{year}{2004}).
699:
700: \bibitem[{\citenamefont{Liu}(2005)}]{JieLiu}
701: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Liu}},
702: \bibinfo{journal}{private communication} (\bibinfo{year}{2005}).
703:
704: \bibitem[{\citenamefont{Javey et~al.}(2003)\citenamefont{Javey, Guo, Wang,
705: Lundstrom, and Dai}}]{Javey03654}
706: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Javey}},
707: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Guo}},
708: \bibinfo{author}{\bibfnamefont{Q.}~\bibnamefont{Wang}},
709: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Lundstrom}},
710: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Dai}},
711: \bibinfo{journal}{Nature} \textbf{\bibinfo{volume}{424}}, \bibinfo{pages}{654
712: } (\bibinfo{year}{2003}).
713:
714: \bibitem[{\citenamefont{Landau and Lifshitz}(1977)}]{LandauLifshitzQM}
715: \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{Landau}} \bibnamefont{and}
716: \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Lifshitz}},
717: \emph{\bibinfo{title}{Quantum Mechanics}} (\bibinfo{publisher}{Pergamon
718: Press}, \bibinfo{address}{Oxford}, \bibinfo{year}{1977}).
719:
720: \end{thebibliography}
721:
722: \end{document}
723: