1: \documentclass[twocolumn,prb,showpacs,preprintnumbers,amsmath,amssymb]{revtex4} %PRL dos columnas
2: \usepackage{graphicx}% Include figure files
3: \usepackage{dcolumn}% Align table columns on decimal point
4: \tolerance = 10000
5: \input epsf
6:
7: \begin{document}
8:
9:
10: \title{Single-channel transmission in gold one-atom contacts and chains}
11: \author{G. Rubio-Bollinger$^1$, C. de las Heras$^1$, E. Bascones$^{2,3,*}$, N. Agra\" it$^1$, F. Guinea$^2$, S. Vieira$^1$ }
12:
13: \affiliation{ $^1$ Instituto Universitario de Ciencia de
14: Materiales "Nicol\' as Cabrera",
15: Universidad Aut\'onoma de Madrid, E-28049 Madrid, Spain.\\
16: $^2$Instituto de Ciencia de Materiales de Madrid, CSIC,
17: Cantoblanco, E-28049
18: Madrid, Spain \\
19: $^3$Departament of Physics, University of Texas at Austin, Austin,
20: TX 78712\\}
21:
22: %\date{\today}
23:
24: \begin{abstract}
25: We induce superconductivity by proximity effect in thin layers of
26: gold and study the number of conduction channels which contribute
27: to the current in one-atom contacts and atomic wires. The atomic
28: contacts and wires are fabricated with a Scanning Tunneling
29: Microscope. The set of transmission probabilities of the
30: conduction channels is obtained from the analysis of the $I(V)$
31: characteristic curve which is highly non-linear due to multiple
32: Andreev reflections. In agreement with theoretical calculations we
33: find that there is only one channel which is almost completely
34: open.
35:
36: PACS numbers: 73.40.-c, 73.20.-r, 74.50.-h
37: \end{abstract}
38:
39:
40:
41: \maketitle
42:
43: Much effort has been devoted in the last decade to the
44: understanding of electron transport processes and mechanical
45: properties of atomic-sized point contacts and chains between
46: metallic electrodes \cite{review}. Coherent electron transport in
47: these nanostructures can be understood in the frame of the
48: scattering formalism. The conductance $G$ of these nanocontacts is
49: given by the Landauer formula \mbox{$G=G_0\sum_{i=1}^{N}\tau_{i}
50: $}, where $\tau_i$ are the transmission probabilities for each of
51: $N$ conductance channels and $G_0=2e^2/h$ is the conductance
52: quantum. For a given contact realization the conductance channels
53: are in general neither completely open nor completely closed and
54: the transparency $\tau_i$ of each channel depends on the material
55: forming the contact \cite{Nature,Sirvent}, detailed atomic
56: arrangement and applied stress \cite{PRLplateau}. In this report
57: we study the channel transparency set $\{\tau_{i}\}$ both in gold
58: one-atom contacts and chains of single gold atoms \cite{Chains},
59: for which theoretical models
60: \cite{PRLplateau,Cuevas98,Lopez-Ciudad99,abinitio,Palacios02}
61: predict an almost completely open single channel, and therefore a
62: conductance close to $G_0$.
63:
64: It is well established that conductance histograms (CH) show that
65: the conductance of one-atom contacts of gold is close to
66: $G_0$\cite{review} but the values of $\tau_i$ cannot be obtained
67: from the only measurement of the total conductance. However, the
68: marked non-linearity of the current-voltage characteristic
69: (\textit{IV}) of superconducting contacts has been exploited to
70: obtain the set of transmission coefficients $\{\tau_i\}$ of
71: atomic-sized aluminum contacts \cite{PRLset}. The channel
72: transmission probabilities are extracted by fitting the measured
73: \textit{IV} curve to a sum of $N$ independent \textit{IV}s
74: calculated for individual channels with a given transmission
75: probability \cite{Averin95,Cuevas96}. This method was afterwards
76: extended to other superconducting materials, showing that the
77: number of conducting channels contributing to the current in
78: one-atom contacts is limited by the valence of the atom at the
79: contact \cite{Nature}.
80:
81: Gold is not superconducting. However we take advantage of the
82: superconducting correlations induced in a thin layer of normal
83: metal in contact with a superconductor. The energy spectrum is
84: modified and a gap is opened at the Fermi energy. At low voltages
85: transport is dominated by Andreev reflection processes which
86: results in non-linear \textit{IV} curves. Here we use the
87: aforementioned method \cite{PRLset,Nature} to analyze the
88: transmission coefficients of proximity induced superconducting
89: one-atom contacts and atomic chains of gold.
90:
91: The proximity effect has been previously exploited
92: \cite{Nature,Scheerproxy} to get information of the $\{\tau_i\}$
93: in gold atomic contacts. In the first experiment \cite{Nature} the
94: \textit{IV} curves were fitted to a sum of theoretical
95: \textit{IV}s corresponding to BCS superconductors and the
96: contribution of a single channel in one-atom contacts is reported.
97: However, the energy dependence of the density of states and the
98: probability of Andreev reflection at a proximity induced
99: superconducting contact are smeared with respect to the BCS model
100: \cite{Belzig96}. The modifications due to the proximity character
101: of the superconducting correlations were taken into account in the
102: later experiment \cite{Scheerproxy}. Most of the experimental
103: \textit{IV}s recorded at the last conductance plateau before
104: contact breaking could be fitted with a single channel. Several
105: channels were necessary to fit the \textit{IV} curves of some
106: contacts with conductance smaller than $G_0$. However, in these
107: experiments the conductance of the smallest contacts is usually
108: much smaller than $G_0$, being the first peak of the histogram
109: placed at $0.6$ $G_0$ and plateaux not nearly flat. These results
110: are not consistent with what has been usually reported in gold
111: atomic contacts in the normal state \cite{review}. Here we report
112: experiments in which the conductance histograms for proximity
113: superconducting gold and normal gold are in remarkable agreement
114: (Fig 1). We also study the channel content of chains of gold
115: atoms.
116:
117:
118: Nanocontacts are produced by pressing two wires crosswise against
119: each other. The wires are used as electrodes in place of tip and
120: sample of an STM. The separation and contact size between the
121: wires can be controlled with the piezoelectric positioning system
122: the microscope. The advantage of using two crossed wires is the
123: possibility of selecting the position of the point contact along
124: both wires using the coarse lateral displacement capability of the
125: microscope, allowing for the exploration of point contacts at
126: different spots along the wires. The wires (0.25 mm diameter) are
127: made of bulk lead and are in a first preparation step covered by
128: thermal evaporation with a thick layer of lead (900 nm at a rate
129: of 0.8 nm/s). This thick layer of lead provides a clean surface. A
130: thin layer of gold (28 nm, rate of 0.1 nm/s) is then evaporated on
131: top of the Pb layer. The lead and gold deposition sequence is
132: performed without breaking the vacuum in the chamber ($<10^{-6}$
133: mbar), thus preventing the presence of oxide at the Pb-Au
134: interface and ensuring good electrical contact between both
135: layers. The substrate is at a temperature of 80 K during film
136: deposition. One-atom contacts are fabricated by slight indentation
137: ($<3$ nm) and subsequent retraction of the electrodes. Further
138: retraction results in contact breaking and a jump to the
139: tunnelling regime. Experiments are done at 1.8 K. During
140: nanocontact pull-off the conductance trace is step-like (see Fig.
141: 1). This characteristic behavior has been shown to be due to the
142: mechanical processes that take place during contact breaking
143: \cite{PRLforce}; conductance plateaus corresponding to elastic
144: deformation stages and sharp conductance changes related to sudden
145: rearrangements of the atoms in the narrowest part of the
146: nanocontact.
147:
148:
149: \begin{figure}
150: %\vspace{0mm}
151: \begin{center}
152: \leavevmode
153: \epsfxsize=85mm % it stands for PRL columns width
154: \epsfbox{Fig1.eps}
155: \end{center}
156: \vspace{0mm}
157:
158: \caption{Differential conductance at a bias voltage of 10 mV of proximity induced superconducting gold contact during pull-off.
159: One-atom contact formation (left) and atomic chain (right). The length of the conductance plateau indicates that
160: a chain longer than five atoms is formed. Inset: conductance histograms of gold in the normal state (bottom)
161: and proximity superconducting (shifted upwards 10000 counts).}
162: \label{FIG. 1.}
163: \end{figure}
164:
165:
166: For gold nanocontacts in the normal state the last conductance
167: plateau before contact breaking is close to $G_0$ and corresponds
168: to the smallest possible contact: a one-atom contact. Despite the
169: inherent variability of the exact conductance trace during contact
170: pull-off, it has been shown in several experiments that the
171: conductance histogram (CH) over a large number of contact breaking
172: realizations displays peaks at conductance values that are
173: characteristic of the chemical nature of electrodes\cite{review}.
174: CH in gold nanocontacts have a characteristic first peak at a
175: conductance close to $G_0$. We show in the inset of Fig. 1 a
176: comparison between CH obtained for bulk gold tips in the normal
177: state and in proximity induced superconducting gold nanocontacts.
178: Due to the presence of excess current in the \textit{IV}s of
179: superconducting nanocontacts it is necessary to measure the
180: differential conductance at a fixed bias voltage well above
181: $\Delta/e$, where $\Delta$ is the energy of the superconducting
182: gap. The remarkable agreement between both CH supports the
183: validity of our analysis also for atomic-sized contacts of gold in
184: the normal state. The low value of the conductance of the first
185: peak in the conductance histogram in Ref. \onlinecite{Scheerproxy}
186: was probably due to enhanced elastic scattering related to sample
187: preparation method.
188:
189: We show in Fig. 2 representative \textit{IV} curves recorded at a
190: one-atom contact, in an atomic chain and in the tunneling regime,
191: up to a voltage $2\Delta_0/e$, where $\Delta_0=1.40$ meV is the
192: bulk superconducting gap for lead. We show both the measured
193: \textit{IV} curves (symbols) and theoretical fitting (lines). The
194: theoretical \textit{IV}s in the contact regime are calculated by
195: solving the time-dependent Bogoliubov de Gennes equations within
196: the scattering formalism \cite{Averin95,Zaitsev97}. This method
197: requires the knowledge of the Andreev reflection amplitude of
198: probability ($a(E)$) at the contact, where the voltage drops.
199: $a(E)$ can be computed if the normal and anomalous Green functions
200: are known\cite{Zaitsev97}. In the tunneling regime the
201: \textit{IV}s are calculated by the usual convolution of the
202: densities of states at both sides of the barrier\cite{Tinkham}.
203:
204:
205: As both Au-Pb electrodes are only very weakly coupled at the weak
206: link, to calculate the density of states and the Andreev
207: reflection amplitude of probability, we model our system as two
208: independent normal-superconducting (NS) structures and solve
209: self-consistently the Usadel equations\cite{Usadel70,Belzig96}.
210: Usadel equations provide a quasiclassical description of the Green
211: functions of a superconductor in the dirty limit, in which
212: electronic transport is diffussive. Elastic impurity scattering is
213: included in the Born approximation and is characterized by a mean
214: free path $l$ (or a diffusion coefficient $D=v_Fl/3$). A
215: description based on Usadel equations was recently used to explain
216: the \textit{IV} curves of lead nanostructures under the influence
217: of a magnetic field and proximity effect, providing excellent
218: quantitative agreement with experiment both in the contact and
219: tunneling regime\cite{Suderow00,Suderow02}.
220:
221:
222: The two NS are assumed to be equal and consist of a dirty normal
223: layer with thickness $d_N$ which is bounded at one end by vacuum
224: and joined to a dirty semi-infinite superconductor at the other.
225: Note that the diffussive description is expected to be valid in
226: the case $d_N\ll l$. The Green functions which are relevant to
227: calculate the IV curves are the ones at the vacuum-bounded edge of
228: the normal layer. The proximity effect is also, affected by the
229: existence of a barrier at the interface, as quasiparticles
230: normally reflected do not contribute to it. In a diffusive system,
231: the mismatch between the characteristic parameters (conductivity
232: and diffusion coefficients) of the normal and superconducting
233: metals, leads to an effective barrier for the quasiparticles. It
234: can be described through the parameter
235: $\Gamma=(D_S/D_N)^{1/2}\sigma_N/\sigma_S$. In our model we assume
236: $\Gamma=1$ and a vanishing resistance of the
237: interface\cite{note1}. With these assumptions the superconducting
238: correlations in the normal metal are described by $\Delta_0$ and
239: the value of $d_N/\xi_S$, where $\xi_S$ is the coherence length of
240: the superconductor. In the atomic chain and one-atom contact the
241: transmission channels set enters also as fitting parameters. A
242: similar model was used by Scheer {\it et al}\cite{Scheerproxy} in
243: the description of Au contacts with superconductiviy induced by
244: proximity effect. In their case, however, the mismatch parameter
245: $\Gamma$ was also used as a fitting parameter.
246:
247:
248:
249:
250:
251: \begin{figure}
252: \vspace{0mm}
253: \begin{center}
254: \leavevmode
255: \epsfxsize=85mm
256: \epsfbox{Fig2.eps}
257: \end{center}
258: \vspace{0mm} \caption{Main figures show the \textit{IV} curves
259: corresponding to the tunneling regime (b), a single-atom contact
260: (bottom curve in (a)) and an atomic chain (top curve in (b)). The
261: curve corresponding to the atomic chain has been displaced for
262: clarity. Experimental results are shown by symbols and theoretical
263: fittings by solid lines. The transmissions obtained in the contact
264: regime are $T_1 = 0.995$ (single atom), and $T = 0.96$ (chain).
265: Inset in (a) shows the density of states used to calculate the
266: theoretical curves -see text. Inset in (b) show the derivative of
267: the experimental (the one with noise) and calculated tunneling
268: curves.} \label{FIG. 2.}
269: \end{figure}
270:
271: Solid lines in Fig. 2 show the calculated \textit{IV} which for
272: the same sample fitting parameter, $d_N/\xi_S=0.81$, best fit both
273: the curves in the tunneling and the contact regime. Assuming the
274: nominal thickness $d_N=28$ nm it corresponds to a superconducting
275: coherence length $\xi_S=34$ nm, in good agreement with the values
276: obtained in Ref. \onlinecite{Suderow00} and
277: \onlinecite{Suderow02}.
278:
279: The tunneling regime \textit{IV} curve, plotted in Fig. 2b, shows
280: a gap smaller than $\Delta_0$ and a bump characteristic of NS
281: structures \cite{Wolf82}. It is however poorly fitted
282: \cite{notefit}, which means that the density of states used, shown
283: in the inset of Fig. 2(a), does not agree completely with the
284: experimental one. The source of disagreement in the fitting can be
285: better understood by looking at the derivative of the experimental
286: and theoretical tunnel \textit{IV} curves, shown in inset in Fig.
287: 2 (b). The theoretical curve shows an asymmetric peak at the gap
288: which decays slowly as the energy increases and results from the
289: one in the density of states. This asymmetric peak structure is
290: characteristic of the diffusive regime for parameters which give
291: gaps in the normal metal much smaller than $\Delta_0$. The peak in
292: the experimental curve is more symmetric. More rounded peaks (and
293: tunneling curves with a bump more similar to the one obtained
294: experimentally) can be obtained for much smaller values of
295: $d_N/\xi_S$. In the diffusive regime the gap induced in the
296: spectrum of the normal metal decreases with $d_N/\xi_S$. Lower
297: values of the fitting parameter induce gaps with values much
298: closer to $\Delta_0$ and could not explain the strong reduction of
299: the gap found experimentally and would give worse fits.
300:
301:
302:
303:
304: Scheer {\it et al.} \cite{Scheerproxy} also found disagreement in
305: the fit of the tunneling conductance and related it to the
306: non-diffusive character of the samples, being the condition
307: $d_N\ll l$ not fulfilled. The proximity effect is a consequence of
308: the coherent superposition of Andreev reflection and is strongly
309: influenced by the length of the path which the electron travels
310: between Andreev reflection processes. Thus, the energy dependence
311: of the induced pair correlations is very sensitive to the degree
312: of disorder in the normal metal and the shape of the spectrum
313: differs considerably in the clean and dirty
314: limits\cite{Arnold78,Wolf82,Belzig96}. We have also tried, without
315: success, to fit the experimental curves assuming that there is no
316: disorder being the transport in the electrodes ballistic, instead
317: of diffusive\cite{Arnold78,Wolf82,Kieselmann87,Vecino}. As
318: concluded by Scheer {\it et al.} \cite{Scheerproxy}, we think
319: that the lack of good fittings in the tunneling regime is due to
320: transport in the gold layer being in the weakly disordered regime,
321: instead of in the diffusive or clean limits.
322:
323:
324: Given the uncertainty in the density of states, the accuracy with
325: which the channel transmission set is obtained in the contact
326: regime is slightly reduced, compared to the ones done in BCS
327: superconductors. The single-atom curve can be reasonably well
328: fitted with a very open channel with transmission $T_1=0.995$. The
329: chain is reasonably well fitted by a single channel with
330: transmission $T = 0.96$. This result is representative of the
331: general behavior that we observe thus confirming that one widely
332: open channel is responsible for the conduction in single atom gold
333: contacts, in agreement with theoretical
334: predictions\cite{Cuevas98,Lopez-Ciudad99,abinitio,Palacios02}.
335:
336:
337:
338:
339: This work has been supported by the Spanish DGI through grants
340: MAT2001-1281 and PB0875/96. Additional funding comes from the ESF
341: Vortex programm, the Welch Foundation, and the NSF under grants
342: DMR-0210383 and DMR0115947. E.B. acknowledges financial support
343: from a fellowship of the Comunidad Aut\'onoma de Madrid during the
344: early stages of this work. We acknowledge W. Belzig for helpful
345: support.
346:
347: $^*$ Present address: Theoretische Physik, ETH-H\"onggerberg,
348: CH-8093 Z\"urich, Switzerland.
349: \begin{thebibliography}{99}
350:
351: \bibitem{review} N. Agra\"it,
352: A. Levy Yeyati and J.M. van Ruitenbeek, submitted to Phys. Rep.
353: (cond-mat/0208239).
354:
355: \bibitem{Nature}E. Scheer, N. Agra\"it, J.C. Cuevas, A. Levy Yeyati, B. Ludoph, A. Mart\'{\i}n-Rodero, G. Rubio-Bollinger, J.M. van Ruitenbeek, C. Urbina,
356: Nature {\bf 394}, 154 (1998).
357:
358: \bibitem{Sirvent} C. Sirvent, J. G. Rodrigo, S. Vieira, L. Jurczyszyn, N. Mingo, F.
359: Flores, Phys. Rev. B {\bf 53}, 16086 (1996).
360:
361: \bibitem{PRLplateau}J.C. Cuevas, A. Levy-Yeyati, A. Mart\'{\i}n-Rodero,
362: G. Rubio-Bollinger, C. Untiedt and N. Agra\"it, Phys. Rev. Lett. {\bf 81},
363: 2990 (1998).
364: \bibitem{Chains} A.I.Yanson, G. Rubio-Boliinger, H.E. van den Brom, N. Agra\"it and J.M. van Ruitenbeek, Nature {\bf 395}, 780 (1998). H. Ohnishi, Y. Kondo and K. Takayanagi, Nature {\bf 395}, 783 (1998).
365: \bibitem{Cuevas98}
366: J.C. Cuevas, A. Levy-Yeyati and A. Mart\'{\i}n-Rodero. Phys. Rev.
367: Lett. {\bf 80}, 1066 (1998)
368: \bibitem{Lopez-Ciudad99}
369: T. L\'opez-Ciudad, A. Garc\'{\i}a-Mart\'{\i}n, A.J. Caama\~no and
370: J.J. S\'aenz. Surf. Sci. Lett., {\bf 440}, L887 (1999).
371: \bibitem{abinitio}
372: M. Brandbyge, J.L. Mozos, P. Ordej\'on, J. Taylor and K. Stokbro,
373: Phys. Rev. B
374: {\bf 65}, 165401 (2002).
375: S.K. Nielsen, M. Brandbyge, K. Hansen, K. Stokbro, J.M. van
376: Ruitenbeek and F. Besenbacher, Phys. Rev. Lett. {\bf 89}, 066804
377: (2002).
378: \bibitem{Palacios02}
379: J.J. Palacios, A.J. P\'erez-Jim\'enez, E. Louis,
380: E. SanFabian and J.A. Verg\'es, Phys. Rev. B {\bf 66}, 035322 (2002).
381: Note that while in this work a conductance larger than one quantum
382: and the contribution of three channels is predicted, one of these
383: channels is almost completely transmitting. The single-channel
384: conductance is recovered for atomic chains.
385:
386:
387: \bibitem{PRLset}E. Scheer, P. Joyez, D. Esteve, C. Urbina and M.H. Devoret, Phys. Rev. Lett. {\bf 78}, 3535 (1997)
388:
389: \bibitem{Averin95}
390: D. Averin and A. Bardas, Phys. Rev. Lett. {\bf 75}, 1831 (1995).
391: \bibitem{Cuevas96}
392: J. C. Cuevas, A. Martin-Rodero, A. Levy-Yeyati, Phys. Rev. B. {\bf
393: 54}, 7366 (1996).
394:
395: \bibitem{Scheerproxy} E. Scheer \textit{et al.}, Phys. Rev. Lett \textbf{86},
396: 284 (2001).
397: \bibitem{Usadel70}
398: K.D. Usadel, Phys. Rev. Lett. {\bf 25}, 507 (1970).
399: \bibitem{Belzig96}
400: W. Belzig, C. Bruder and G. Sch\"on, Phys. Rev. B, \textbf {54},
401: 9443 (1996).
402: \bibitem{PRLforce} G. Rubio, N. Agra\"it and S. Vieira, \textit{Phys. Rev. Lett.} {\bf 76}, 2302 (1996).
403: \bibitem{Tinkham}{\it Introduction to Superconductivity}, M. Tinkham, Mc Graw-Hill,1996.
404: \bibitem{Zaitsev97}
405: A.V. Zaitsev and D.V. Averin, Phys. Rev. Lett. {\bf 80}, 3602
406: (1998).
407: \bibitem{Suderow00} H. Suderow, E. Bascones, W. Belzig, F. Guinea and S. Vieira,
408: Eur. Phys. Lett. {\bf 50}, 740 (2000).
409: \bibitem{Suderow02} H. Suderow, E. Bascones, A. Izquierdo, F. Guinea and
410: S. Vieira, Phys. Rev. B {\bf 65}, 100519 (2002).
411: \bibitem{note1}
412: We have found that to introduce the influence of the mismatch of
413: the Fermi velocities and mean free path in gold and lead or of an
414: interface resistance does not improve significantly the quality of
415: the fittings and would introduce more fitting parameters.
416: \bibitem{Arnold78}
417: G.B. Arnold, Phys. Rev. B, {\bf 18},1076 (1978)
418: \bibitem{Wolf82}
419: G.B. Arnold, E.L. Wolf, Phys. Rev. B {\bf 25}, 1541 (1982).
420: \bibitem{Vecino} E. Vecino, A. Martin-Rodero, and A. LevyYeyati Phys. Rev. B
421: {\bf 64}, 184502 (2001)
422:
423: \bibitem{Kieselmann87}
424: G. Kieselmann, Phys. Rev. B {\bf 35},6762 (1987)
425: \bibitem{notefit}
426: Slightly better agreement can be obtained with a larger $d_N/\xi$.
427: Larger values of $d_N/\xi$ do not fit properly the \textit{IV}
428: curve in the contact regime for any value of the channels
429: transmission. .
430:
431: \end{thebibliography}
432:
433:
434:
435: \end{document}
436: