1: \documentclass[aps,twocolumn]{revtex4}
2: \usepackage{graphicx}
3: %\documentstyle[multicol,aps,epsfig]{revtex}
4:
5: \begin{document}
6:
7: \title{Transition from tunneling to direct contact in tungsten nanojunctions}
8:
9: \author{ A.~Halbritter, Sz.~Csonka and G.~Mih\'aly}
10: \affiliation{ Department of Physics, Institute of Physics,
11: Budapest University of Technology and Economics, 1111 Budapest,
12: Hungary}
13: \author{E.~Jurdik, O.Yu.~Kolesnychenko, O.I.~Shklyarevskii$^\dag$,
14: S.~Speller and H.~van~Kempen}
15: \affiliation{NSRIM, University of Nijmegen, Toernooiveld 1, 6525
16: ED Nijmegen, the Netherlands}
17:
18: %\date{\today}
19:
20: \begin{abstract}
21: We apply the mechanically controllable break junctions technique
22: to investigate the transition from tunneling to direct contact in
23: tungsten. This transition is quite different from that of other
24: metals and is determined by the local electronic properties of the
25: tungsten surface and the relief of the electrodes at the point of
26: their closest proximity. The conductance traces show a rich
27: variety of patterns from the avalanche-like jump to a mesoscopic
28: contact to the completely smooth transition between direct contact
29: and tunneling. Due to the occasional absence of an adhesive jump
30: the conductance of the contact can be continuously monitored at
31: ultra-small electrode separations. The conductance histograms of
32: tungsten are either featureless or show two distinct peaks related
33: to the sequential opening of spatially separated groups of
34: conductance channels. The role of surface states of tungsten and
35: their contribution to the junction conductance at sub-{\AA}ngstrom
36: electrode separations are discussed.
37: \end{abstract}
38:
39: \pacs{PACS number(s): 73.40.Jn, 72.15.-v}
40:
41:
42: \maketitle
43:
44:
45: \section{Introduction}
46:
47: %1
48: The study of transition from tunneling to direct contact and
49: electrical transport through atomic-sized metallic conductors has
50: been the object of great attention during the last decade.
51: Different types of phenomena, related to both the quantum
52: character of transport and the atomic discreteness of the contact,
53: were observed in 3D nanoconstrictions produced by scanning
54: tunneling microscopes (STM) or mechanically controllable break
55: junctions (MCBJ) \cite{agrait_review}. In particular, conductance
56: measurements of breaking nanowires demonstrated a step-like
57: structure of conductance versus electrode separation traces
58: $G(z)$. For single-valence $s$-metals conductance plateaus are
59: close to the integer multiples of the quantum conductance unit
60: $G_0=2e^2/h$. However, simultaneous measurements of both the force
61: and the conductance for breaking gold nanowires demonstrated that
62: jumps between plateaus in the conductance staircase are always
63: correlated with relaxations of the mechanical force and,
64: therefore, with atomic rearrangements in the nanoconstriction
65: \cite{rubio}. Individual conductance curves $G(z)$ are inherently
66: irreproducible due to the different dynamical evolution of the
67: connective necks during the break. Therefore, analysis of
68: experimental data includes construction of conductance histograms
69: based on a large number of conductance traces. Peaks in the
70: conductance histograms are related to the statistically more
71: probable atomic configurations in the connective neck between the
72: electrodes \cite{krans}. For polyvalent metals the main (and
73: sometimes the only) feature in the conductance histograms is a
74: peak corresponding to one-atom point contact \cite{thesis}. The
75: conductance through such a contact is determined by a few
76: conductance channels intimately related to atomic orbitals
77: \cite{cuevas}. Transition from tunneling to single-atom contact
78: occurs in an avalanche-like way at an electrode separation of
79: $\sim 1.5$~\AA~due to the metallic adhesion forces \cite{sutton}.
80: This sudden jump in conductance precluded measurements of $G(z)$
81: at sub-{\AA}ngstrom distances between the electrodes for all
82: metals studied to date.
83:
84: %2
85: While measuring thermal expansion of MCBJ electrodes for different
86: materials, we took notice of the unusually high stability of
87: tungsten tunnel junctions at very close electrode separations
88: \cite{thermo}. In another study, transmission and scanning
89: electron microscopy images showed no evidence of connective neck
90: formation between tungsten wires \cite{correia}. Moreover,
91: measurements of adhesive forces between an atomically-defined
92: W(111) trimer tip and a Au(111) sample revealed no spontaneous
93: jump to contact \cite{cross}. These unusual properties of tungsten
94: nanocontacts motivated us to carry out further extensive
95: investigations.
96:
97: %3
98: In this paper we present our experiments with tungsten MCBJ. Our
99: aim was to examine the behavior of tungsten nanojunctions during
100: the transition between tunneling and direct contact and to what
101: extent this behavior is influenced by the unique mechanical and
102: electronic properties of tungsten. We show that very often the
103: adhesive jumps of conductance are absent and that the transition
104: to single atom contact is smooth. This permitted us to investigate
105: the junction at ultra-small electrode separations. By employing
106: the conductance histogram technique we determined the preferential
107: values of conductance during the fracture of the junction to be
108: about $1$~$G_0$ and $2$~$G_0$. Also, we measured the conductance
109: histograms of tantalum and molybdenum (as two nearest neighbors of
110: tungsten in the periodic table of elements) to provide additional
111: support for our interpretation.
112:
113: \section{Experimental setup}
114:
115: \begin{figure}
116: \includegraphics[width=0.5\columnwidth]{fig1.eps}
117: \caption{Modified sample mounting. The tungsten wire is glued on
118: top of shear piezo-ceramic attached to a phosphor bronze bending
119: beam. The distance between the electrodes is controlled by
120: vertically pushing the bending beam at the middle or by applying
121: voltage on the left-hand side shear piezo. All distances are
122: exaggerated for clarity.} \label{fig1}
123: \end{figure}
124:
125: %4
126: We employed the traditional MCBJ technique, described elsewhere
127: \cite{agrait_review}, with a modified sample mounting (Fig.~1).
128: This mounting includes two $6 \times 2.5 \times 1$~mm pieces of
129: shear piezo-ceramic (which gives horizontal displacement of its
130: surface when voltage is applied) glued at the center of a phosphor
131: bronze bending beam. A polycrystalline tungsten wire with a
132: diameter of 100~$\mu$m was attached to the top of the piezo's with
133: two small drops of hard epoxy (Stycast 850FT). The central section
134: of the wire was then electrochemically etched in a 25\% solution
135: of KOH down to $5$--$10$~$\mu$m at its thinnest part. The wire was
136: broken at $4.2$~K in an ultra-high vacuum environment by bending
137: the beam. The distance between the electrodes was fine tuned by
138: changing the deflection of the bending beam or applying voltage to
139: the left-hand side shear piezo. The relative displacement of the
140: electrodes was calibrated using the exponential part of $G(z)$
141: traces in the tunneling regime (assuming a work function for
142: tungsten $\phi \simeq 4.5$~eV) as well as by measuring field
143: emission resonance spectra \cite{rsi}. The contact conductance was
144: measured using a current-to-voltage converter with a gain of
145: 0.1~V/$\mu$A. Conductance versus electrode separation traces
146: $G(z)$ were measured in a slow mode ($2$--$30$~points/s) using
147: Keithley 2182 nanovoltmeters in order to cover over 7 orders of
148: magnitude in the conductance during transition from tunneling to
149: direct contact. Conductance traces for building conductance
150: histograms were recorded with an AT-MIO-16XE-50 National
151: Instruments data acquisition board (sampling rate of
152: $20000$~points/s and resolution of $16$~bits). During this
153: acquisition a ramp voltage with a frequency of $5$--$50$~Hz was
154: applied to the shear piezo to establish a repeated fracture of the
155: junction.
156:
157: %5
158: The sample surfaces were characterized by scanning the electrodes
159: with the two shear piezo's in the constant current mode. We found
160: atomically flat parts up to $5$--$10$~nm in length alternating
161: with irregularities of $5$--$15$~\AA~ in height. It should be
162: noted that these STM-like measurements with two ``blunt''
163: electrodes are rather qualitative and the numbers presented are
164: only rough estimates.
165:
166: %6
167: We analyzed thoroughly more than $200$ conductance histograms
168: measured for $12$ different samples.
169:
170: \section{Results}
171:
172: \begin{figure}
173: \includegraphics[width=0.6\columnwidth]{fig2.eps}
174: \caption{Two extreme cases of $G(z)$ traces for W MCBJ. 1 --
175: avalanche-like transition from tunneling to a low-ohmic contact.
176: The upper panel shows the contact conductance in the process of
177: retraction. 2 -- transition from tunneling to direct contact
178: without spontaneous formation of an adhesive neck.} \label{fig2}
179: \end{figure}
180:
181: %7
182: We investigated the transition from the tunneling regime to direct
183: contact for tungsten MCBJ by performing conductance measurements
184: during slow, continuous approach or retraction of the electrodes.
185: These measurements revealed a rich variety of $G(z)$ traces. All
186: of them fall between two extreme cases shown in Fig.~\ref{fig2}.
187: Curve 1 demonstrates an avalanche-like transition from $G \approx
188: 0.01$--$0.1$~$G_0$ to $100$--$1000$~$G_0$. This indicates the
189: formation of a mesoscopic contact with a cross-section of
190: $10$--$100$~nm$^2$ between two flat portions of the electrodes.
191: Subsequent retraction of the electrodes produced only small
192: changes in the contact conductance over a retreat distance of
193: $\gtrsim 4$~nm, which were followed by sudden disconnection (upper
194: panel in Fig.~\ref{fig2}). During $5$--$10$~minutes after
195: disconnection, the distance between the electrodes increased as
196: deduced from the transient decrease of the measured tunnel
197: current. After the full relaxation of the sample, the original
198: curve could be reproduced accurately, indicating that no
199: irreversible changes of the electrode relief occurred. The other
200: extreme case -- curve 2 in Fig.~\ref{fig2} -- demonstrates a
201: smooth transition from tunneling to direct contact with no sign of
202: spontaneous formation of an adhesive neck. The majority of $G(z)$
203: curves exhibited combination of smooth changes of the conductance
204: and sudden jumps with amplitudes ranging from $0.1G_0$ up to few
205: $G_0$.
206:
207: \begin{figure}
208: \includegraphics[width=.6\columnwidth]{fig3.eps}
209: \caption{Conductance histograms for W MCBJ based on $10000~G(z)$
210: traces. (a) -- featureless histogram. (b) -- histogram with peaks
211: close to integers of $2e^2/h$. Insets in (a) and (b): histograms
212: for the corresponding data sets after filtering out the
213: featureless curves.} \label{fig3}
214: \end{figure}
215:
216: %8
217: We also performed statistical analysis of conductance traces by
218: employing the conductance histogram technique. We found that
219: tungsten nanojunctions have a particular tendency to show
220: reproducible, almost identical conductance traces when indentation
221: of the electrodes is not sufficient. If proper care is taken and
222: in every cycle contacts of $G \gtrsim 30$--$40$~$G_0$ are formed,
223: the break occurs in a large variety of ways. In this latter case a
224: statistical approach to data analysis is justified. In sharp
225: contrast to other metals, the conductance histograms of tungsten
226: do not necessarily exhibit the same pattern when the contact site
227: is shifted by the right-hand side shear piezo or when the sample
228: is replaced. A major part of histograms demonstrated no
229: distinctive features [Fig.~\ref{fig3}(a)]. However, on many
230: occasions the histograms showed two peaks at respectively
231: $1$--$1.2$ and $2$--$2.2$~$G_0$ [Fig.~\ref{fig3}(b)], or only one
232: of those two.
233:
234: \begin{figure}
235: \includegraphics[width=.6\columnwidth]{fig4.eps}
236: \caption{Typical conductance traces for W. Curve $1$ --
237: featureless curve. Curve $2$ -- smooth transition to tunneling
238: with distinct plateaus. Curves $3$a and $3$b -- traces with sudden
239: jumps of conductance.} \label{fig4}
240: \end{figure}
241:
242: %9
243: In Fig.~\ref{fig4} four characteristic conductance traces are
244: presented. Curve 1 shows a smooth and completely featureless
245: transition from direct contact to tunneling. For a tungsten
246: junction this kind of transition is frequently present and gives
247: rise to the featureless background in the conductance histograms.
248: This background is suppressed when these featureless curves are
249: rejected from the data set by employing a computer filtering
250: algorithm \cite{note_filter}. Then, either peaks invisible in the
251: original histogram appear [inset in Fig.~\ref{fig3}(a)] or the
252: contrast of peaks is significantly improved [inset in
253: Fig.~\ref{fig3}(b)]. Peaks in the histogram naturally arise from
254: conductance traces with distinct features -- plateaus -- at about
255: $1G_0$ and/or $2G_0$ as demonstrated by curves 2, 3a and 3b in
256: Fig.~\ref{fig4}. While curve 2 shows a smooth transition with no
257: jumps, in the case of curves 3a and 3b clear jumps occur near
258: $1G_0$ and $2G_0$, respectively. We note here that conductance
259: traces like those represented by curve 2 are unique to tungsten
260: and were not observed in other metallic contacts.
261:
262: %10
263: In our previous analysis we used an algorithm to select the traces
264: with long plateaus \cite{note_filter}. However, it is also
265: important to differentiate between the smooth curves and those
266: with sudden jumps. For this reason every trace was characterized
267: by the largest jump between neighboring data points in the
268: conductance interval $G/G_0 \in [0.3, 3.3]$. By combining these
269: two approaches the traces were classified into the three
270: categories presented in Fig.~\ref{fig4}. Using the data set for
271: histogram in Fig~\ref{fig3}(b) we found that $\sim 30\%$ of traces
272: exhibited a smooth and featureless transition and roughly $10$\%
273: of curves show plateaus without sudden jumps. In the rest of the
274: traces clear conductance jumps were found. In order to quantify at
275: which conductance values conductance jumps with $\Delta G_{max}
276: \geq 0.7 G_0$ are most likely to occur, we also built up
277: histograms for the position of these sudden transitions. The peaks
278: in the so-built histogram presented in Fig.~\ref{fig5} clearly
279: demonstrate that large conductance jumps are most probable at $G
280: \simeq 1G_0$ and $2G_0$. We emphasize that the relative
281: occurrence of the three different types of conductance traces is
282: sample and site-dependent, and in certain data sets only traces
283: with smooth transitions were present.
284:
285: \begin{figure}
286: \includegraphics[width=.6\columnwidth]{fig5.eps}
287: \caption{Histogram for conductance values at which jumps with
288: amplitude $\Delta G > 0.7G_0$ occur.} \label{fig5}
289: \end{figure}
290:
291: \section{discussion}
292:
293: %MECHANICAL PROPERTIES
294:
295: %11
296: Our experimental results for tungsten nanojunctions strikingly
297: differ from those obtained for other metals studied so far, that
298: exhibit conductance traces always falling into the same
299: (material-specific) pattern. Especially, the adhesive jump to
300: contact or the tensile overstretching followed by sudden
301: disruption have respectively been found to be an unavoidable
302: attendant of contact formation or breakage. This has precluded
303: measurements of conductance at small electrode separations. The
304: occasional absence of these processes is unique for tungsten
305: junctions and we attribute it to the specific mechanical
306: properties of this metal. In particular, this odd junction
307: stability at ultra-small electrode separations permitted us to
308: study the previously unaccessible regime of nanocontact
309: conductance.
310:
311: %12
312: The adhesive jump is governed by the the competition between the
313: adhesive and tensile forces \cite{taylor}. The tensile forces are
314: determined by the stiffness of the electrodes, which is composed
315: from the stiffness of the nanoscale junction itself and the
316: stiffness of the whole setup. The latter can have significant
317: importance in an atomic force microscope with a soft cantilever,
318: however it is not expected to play any role in an MCBJ due to the
319: rigid sample mounting. The stiffness of the junction depends on
320: the strength of the bonds in the material, but also the junction
321: geometry and the crystallographic orientation of the electrodes
322: have a crucial influence on it. Contacts with large opening angle
323: have large stiffness, while in contacts with small opening angle a
324: large amount of layers are involved in the elastic deformation and
325: thus the stiffness is reduced. The stiffness is also reduced, if
326: the contact surfaces are not perpendicular to the contact axis.
327: Then, the electrodes can bend like a cantilever and the elastic
328: behavior is defined by the smaller shear forces.
329:
330: %13
331: The elastic behavior of the material is basically determined by
332: the strength of the bonds between the atoms, therefore it is
333: similar at any length-scale and atomic scale systems are well
334: described by bulk elastic constants \cite{rubio}. The bulk elastic
335: modulus of tungsten ($E_{\mathrm{W}}=411$\,GPa) is outstandingly
336: large among all metals. (As a comparison the Youngs modulus of Au,
337: Pt, and Nb is $E_{\mathrm{Au}}=78$\,GPa,
338: $E_{\mathrm{Pt}}=168$\,GPa, and $E_{\mathrm{Nb}}=105$\,GPa,
339: respectively.) On the other hand, the junction geometry is not
340: controllable in MCBJ experiments. In previously studied metals the
341: conductance traces imply that the junction breaks through the
342: process of rupture. In this case the contact geometry is
343: ``self-organized'' during the neck pulling, so the conditions for
344: the adhesive jumps are similar during each disconnection. Our
345: measurements imply that tungsten breaks in a more brittle way and
346: a large variety of contact geometries are established during the
347: fractures. This, together with the outstanding elastic modulus of
348: tungsten can explain our observations. In ``stiff'' contact
349: geometries the elastic forces can overcome the adhesion and
350: thereby enable a smooth transition between tunneling and direct
351: contact, while in softer contact geometries the adhesive jumps are
352: still present.
353:
354: %14
355: Our interpretation can be tested by comparing the mechanical
356: behavior during the break of further metals with different elastic
357: modulus. For this reason we performed measurements on two
358: neighbors of tungsten in the periodic system, molybdenum and
359: tantalum. These metals have similar electronic properties to
360: tungsten. The elastic modulus of molybdenum is relatively large
361: ($E_{\mathrm{Mo}}=329$\,GPa), while tantalum is much softer
362: ($E_{\mathrm{Ta}}=186$\,GPa). In agreement with that we found that
363: tantalum always breaks in a jump-like way, while in molybdenum a
364: part of the conductance traces showed a smooth transition.
365:
366: %15
367: To explain the avalanche-like motion of hundreds to thousands of
368: atoms in the extreme case presented in Fig.~\ref{fig2} as curve 1,
369: we consider the following three possibilities. The first one
370: corresponds to the motion of a number of atomic layers in the
371: direction normal to the electrode surfaces as assumed in the
372: calculations by Taylor \textit{et al.} \cite{taylor}. This
373: possibility would require an avalanche of a macroscopic amount of
374: metal. Such a situation can take place only for contacts that act
375: as a very soft spring and thus we do not consider it as the most
376: likely explanation. The second possibility, that would result in
377: the same behavior of the contact conductance, is the transition to
378: direct contact due to bending of the electrodes. This arrangement
379: is likely to occur when flat parts of the electrode surfaces are
380: not perpendicular to the electrode axis. Then, the component of
381: the adhesive force that is normal to the electrode axis causes
382: bending. During our experiments, we observed anomalously high
383: sensitivity of the tungsten MCBJ to acoustic vibrations, such as
384: human voice, that suggests a transverse, spring-like motion of the
385: electrodes. This observation is in favor of the ``bending model.''
386: As the last possibility, the contact between the electrodes may
387: emerge as a result of dislocation glide or homogeneous shear
388: motion of one of the electrodes \cite{agrait_review}. However, in
389: this case it is not possible to explain the reversible behavior of
390: contact conductance through many formation-breaking cycles.
391:
392: %ELECTRONIC PROPERTIES
393:
394: %16
395: The smooth conductance traces with plateaus at the first two
396: conductance quanta (e.g., curve 2 in Fig.~4) are very close to
397: those expected for conductance quantization in short constrictions
398: \cite{torres}. In spite of its attractiveness, we discarded this
399: effect as the possible origin of the peaks in our histograms.
400: Extremely small distances between the plateaus in conductance
401: traces of tungsten, $\Delta z \simeq 30$~pm, suggest a
402: constriction with an opening angle close to 90 degrees (an
403: orifice). In this case the conductance quantization must be
404: completely suppressed \cite{torres}. The shape of these traces is
405: better explained by the sequential closing of conduction channels
406: due to different spatial distribution of the involved electronic
407: states.
408:
409: %17
410: For transition $d$-metals five conductance channels are expected
411: to contribute to the conductance of a single-atom contact
412: according to the number of valence orbitals. The transmission
413: values of these conductance channels were extensively studied for
414: single-atom niobium junctions by tight binding calculations
415: \cite{cuevas} and by measuring subgap structure in the
416: superconducting state \cite{niobium}. The calculations performed
417: for a simplified one-atom contact geometry revealed a single
418: dominant channel (as a result of the hybridization between the $s$
419: and $d_{z^2}$ orbitals), two medium-sized channels and two smaller
420: channels that yield together a net conductance of
421: $G=2$--$3$~$G_0$. Both the conductance and the transmission
422: coefficients showed a good agreement with the values determined
423: from the experiment. Such a detailed investigation of the
424: transmission eigenvalues is not available for the rest of the
425: $d$-metals, and therefore the conduction properties of monoatomic
426: contacts can only be deduced from the conductance histograms. For
427: all non-magnetic transition metals studied so far (Nb, V, Rh, Pd,
428: Ir, Pt) the histograms show a single well-defined peak centered in
429: the region of $G=1.5$--$2.5~G_0$, that is attributed to the
430: conductance through a single-atom contact
431: \cite{agrait_review,thesis}. To date, theoretical calculations
432: have not been performed for the conductance of a single-atom
433: tungsten contact, but its value is presumably in the same range of
434: $G \simeq 1.5$--$2.5~G_0$.
435:
436: %18
437: In tungsten the first peak in the histogram is situated at
438: $1$~$G_0$, which is too low to be attributed to a single atom
439: contact with all the channels open. Therefore, rather the second
440: peak at $2$~$G_0$ is interpreted as the conductance of a
441: monoatomic contact geometry, where the center atom is closely
442: bound to both electrodes. The peak at $1$~$G_0$ is assumably
443: connected to an arrangement where the electrodes are so much
444: separated that only a part of the electronic states have
445: noticeable overlap.
446:
447: %19
448: It is well known from STM studies that tungsten surfaces have long
449: protruding surface modes, which have a crucial role in achieving
450: atomic resolution. First principle calculations for the W(001)
451: surface demonstrated that the tunnel current is primarily
452: generated by the $5d_{z^2}$ dangling-bond surface states
453: \cite{ohnishi,chen}. At distances $>2$~\AA~from the nuclei the
454: charge density of the surface state is much higher than that of
455: the atomic $6s$-state \cite{post} and so the surface states play a
456: key role in the electron transport at small separations. The
457: conductance through a single channel formed by such surface states
458: can be a good candidate for explaining the conductance plateaus at
459: $1$~$G_0$.
460:
461: %20
462: The occasional absence of adhesive jump enables us to monitor the
463: transition between the two situations. However, the sequential
464: closing of conductance channels -- like curve 2 in Fig.~\ref{fig4}
465: -- may only occur for specific electrode configurations, which
466: explains the relatively low percentage of this type of conductance
467: traces.
468:
469: \begin{figure}
470: \includegraphics[width=0.6\columnwidth]{fig6.eps}
471: \caption{Conductance histograms for tantalum (a) and molybdenum
472: (b) based on 20000 conductance traces. Inset in (b) shows a
473: histogram for selected 10\% of the curves, exhibiting plateaus
474: around 1$G_0$.} \label{fig6}
475: \end{figure}
476:
477: %21
478: The conductance histograms of tantalum and molybdenum were also
479: studied. These two neighbors of tungsten in the periodic system
480: have very similar electronic structure to W. Both for Ta and Mo
481: the histograms show a single, well-defined peak situated around
482: $2$~$G_0$ and $2.5$~$G_0$, respectively (see Fig.~\ref{fig6}).
483: This gives further support to the assumption that the conductance
484: through a single atom contact is in the range of $2$~$G_0$ for
485: tungsten as well. Though the electronic properties of Mo surfaces
486: are similar to tungsten \cite{molybdenum} the histograms show no
487: indication for a peak at $1$~$G_0$. In some occasions the computer
488: filtering of the original data set reveals conductance traces with
489: plateaus close to the conductance quantum, and accordingly a small
490: peak around 1~$G_0$ is recovered in the histogram (see inset in
491: Fig.~\ref{fig6}(b)).
492:
493: \section{Conclusions}
494:
495: %22
496: In conclusion, our measurements have shown that the evolution of
497: conductance during the break of tungsten nanojunctions essentially
498: differs from the ususal behavior of metallic contacts. Due to the
499: large elastic modulus of W and the the non-ductile contact
500: fracture, the adhesive jumps in the conductance traces are
501: occasionally absent. Thank to these completely smooth transitions
502: from direct contact to tunneling, the closing of conductance
503: channels can be continuously monitored at ultra small electrode
504: separations. The study of conductance histograms has shown, that
505: the preferential values of conductance in tungsten junctions are
506: $1$\,$G_0$ and $2$\,$G_0$. The smooth conductance traces with
507: plateaus at these two values are explained by the sequential
508: closing of conductance channels due to their different spatial
509: distribution. The conduction through the long-protruding surface
510: states of tungsten might be responsible for the plateaus at
511: $1$\,$G_0$.
512:
513: %23
514: The limitations of MCBJ technique precluded us from drawing a more
515: exact and persuasive conclusion. The current data can greatly be
516: improved by simultaneous measurements of the conductance and the
517: forces in the course of electrode approach for contacts with a
518: well-defined tip/sample shape and orientation. To our opinion, the
519: details of the process of avalanche-like transition to direct
520: contact and subsequent break can be visualized with a high
521: resolution transmission electron microscope operating in UHV.
522:
523: \section*{Acknowledgments}
524:
525: %23
526: The authors are grateful to A.J. Toonen, J. Hermsen and J.
527: Gerritsen for invaluable technical assistance. Part of this work
528: was supported by the Stichting voor Fundamenteel Onderzoek der
529: Materie (FOM) which is financially supported by the Nederlandse
530: Organisatie voor Wetenschappelijk Onderzoek (NWO). The Hungarian
531: Research Funds OTKA TO37451, TS040878, N31769 and a NWO grant for
532: Dutch-Hungarian cooperation are also acknowledged. O.I.S. wishes
533: to thank the NWO for a visitor's grant.
534:
535: %REFERENCES
536:
537: \begin{references}
538: \item[$^\dag$] Also at: B. Verkin Institute for Low Temperature
539: Physics \& Engineering, National Academy of Science of Ukraine, 47
540: Lenin Av., 61164, Kharkov, Ukraine.
541:
542: \bibitem{agrait_review} J.M. van Ruitenbeek, in "Metal Cluster on
543: Surfaces: Structure, Quantum Properties, Physical Chemistry"' K.H.
544: Meiwes-Broer, ed. (Springer-Verlag, Heidelberg, 2000) 175-210; N.
545: Agrait, A. Levy Yeyati, and J.M. van Ruitenbeek, to appear in
546: Phys. Reports.
547:
548: \bibitem{rubio} G. Rubio, N. Agra\"{\i}t, and S. Viera,
549: Phys. Rev. Lett. {\bf 76}, 2302 (1996).
550:
551: \bibitem{krans} J.M. Krans, J.M. van Ruitenbeek, V.V. Fisun, I.K. Yanson,
552: and L.J. de Jongh, Nature (London) {\bf 375}, 767 (1995).
553:
554: \bibitem{thesis} A.I. Yanson, Ph.D. thesis, University of Leiden, 2001.
555:
556: \bibitem{cuevas} J.C. Cuevas, A.Levy Yeyati, and A. Mart\'{\i}n-Rodero,
557: Phys. Rev. Lett. {\bf 80}, 1066 (1998).
558:
559: \bibitem{sutton} A.P. Sutton and J.B. Pethica,
560: J.Phys.: Cond. Mat. {\bf 2}, 5317 (1990); U. Landman, W.D.
561: Luedtke, N.A. Burnham, and R. Colton, Science {\bf 248}, 454
562: (1990).
563:
564: \bibitem{thermo} O.Yu. Kolesnychenko, A.J. Toonen, O.I.
565: Shklyarevskii,
566: and H. van Kempen, Appl. Phys. Lett. {\bf 79}, 2707 (2001).
567:
568: \bibitem{correia} A. Correia, M.I. Marqu\'es, and N. Garc\'{\i}a,
569: J. Vac. Sci. Techn. B {\bf 15}, 548 (1997).
570:
571: \bibitem{cross} G. Cross, A. Schirmeisen, A. Stalder,
572: P. Gr\"utter, M. Tschudy, and U. D\"urig, Phys. Rev. Lett. {\bf
573: 80} 4685 (1998).
574:
575: \bibitem{rsi} O.Yu. Kolesnychenko, O. I. Shklyarevskii, and H. van Kempen,
576: Rev. Sci. Instrum. {\bf 70}, 1442 (1999).
577:
578: \bibitem{note_filter} The integral
579: $\int(| \partial_z G(z)| + \varepsilon)^{-1} dz$, where
580: $\varepsilon > 0$ is a small number used to prevent singularities,
581: was calculated for each conductance trace within an interval
582: $G(z)/G_0 \in [0.3,3.3]$ and its value was used to select the
583: curves with distinct plateaus.
584:
585: \bibitem{taylor} P.A. Taylor, J.S. Nelson, and B.W. Dodson,
586: Phys. Rev. B {\bf 44}, 5834 (1991).
587:
588: \bibitem{riedle} J. Riedle, P. Gumbsch, and H.F. Fischmeister,
589: Phys. Rev. Lett. {\bf 76} 3594 (1996).
590:
591: \bibitem{torres} J.A. Torres, J.I. Pascual, and J.J. Saenz,
592: Phys. Rev. B {\bf 49}, 16581 (1994).
593:
594: \bibitem{niobium} B. Ludoph, N. van der Post, E.N. Bratus, E.V.
595: Bezuglyi, V.S. Shumeiko, G. Wendin, and J.M. van Ruitenbeek, Phys.
596: Rev. B {\bf 61}, 8561 (2000).
597:
598: \bibitem{ohnishi} S. Ohnishi and M. Tsukuda,
599: Solid State Commun. {\bf 71}, 391 (1989).
600:
601: \bibitem{chen} C.J. Chen,
602: Phys. Rev. B {\bf 42} 8841 (1990).
603:
604: \bibitem{post} M. Posternak, H. Krakauer, A.J. Freeman, and D.D.
605: Koeling, Phys. Rev. B {\bf 21}, 5601 (1980).
606:
607: \bibitem{thesisk} O.Yu. Kolesnychenko,
608: Ph.D. thesis, University of Nijmegen, 2002.
609:
610: \bibitem{molybdenum} S.B. Legoas, A.A. Araujo, B. Laks, A.B.
611: Klautau, and S. Frota-Pess\^{o}a, Phys. Rev. B {\bf 61}, 10417
612: (2000).
613:
614: \end{references}
615:
616: \end{document}
617: