cond-mat0703589/Al.tex
1: %
2: %  Manuscript prepared in IOP-Style for submission to Nanotechnology
3: %
4: 
5: \documentclass{iopart}
6: %\documentclass[12pt]{iopart}
7: \usepackage{graphicx}% Include figure files
8: \usepackage{dcolumn}% Align table columns on decimal point
9: \usepackage{bm}% bold math
10: %\usepackage{rotating}
11: %\usepackage{graphpap}
12: 
13: \newcommand{\gsim}{{\textstyle{\lower 2pt \hbox{$>$} \atop \raise 1pt \hbox{$ \sim$}}}}
14: % \newcommand{\lsim}{{\textstyle{\lower 2pt \hbox{$<$} \atop \raise 1pt \hbox{$ \sim$}}}}
15: 
16: \newcommand\onlinecite[1]{\cite{#1}}
17: 
18: \begin{document}
19: 
20: % Suggested title revision:
21: \title{Electronic and atomic shell structure in aluminum nanowires}
22: %\title{Observation of shell structure in aluminum nanowires}
23: 
24: \author{A.~I. Mares$^{1}$, D.~F. Urban$^{2}$, J. B\"{u}rki$^{3}$, H. Grabert$^{2}$, \\
25:         C. A. Stafford$^{3}$ and J.~M. van Ruitenbeek$^{1}$}
26: 
27: % \affiliation{$^{1}$Kamerlingh Onnes Laboratorium, Universiteit Leiden,
28: %     P.O. Box 9504, 2300 RA Leiden, The Netherlands \\
29: %     $^{2}$Physikalisches Institut, Albert-Ludwigs-Universit\"{a}t, D-79104 Freiburg, Germany \\
30: %     $^{3}$Department of Physics, University of Arizona, Tucson, Arizona 85721, USA}
31: 
32: \address{$^{1}$Kamerlingh Onnes Laboratorium, Universiteit Leiden,
33:     P.O. Box 9504, 2300 RA Leiden, The Netherlands}
34: \address{$^{2}$Physikalisches Institut, Albert-Ludwigs-Universit\"{a}t, D-79104 Freiburg, Germany}
35: \address{$^{3}$Department of Physics, University of Arizona, Tucson, Arizona 85721, USA}
36: 
37: \ead{urban@physik.uni-freiburg.de}
38: 
39: \submitto{\NT,\hspace{1cm}\today}
40: 
41: \begin{abstract}
42: 
43: We report experiments on aluminum nanowires in ultra-high vacuum
44: at room temperature that reveal a periodic spectrum of
45: exceptionally stable structures. Two ``magic'' series of stable
46: structures are observed: At low conductance, the formation of
47: stable nanowires is governed by electronic shell effects whereas
48: for larger contacts atomic packing dominates. The crossover
49: between the two regimes is found to be smooth. A detailed
50: comparison of the experimental results to a theoretical stability
51: analysis indicates that while the main features of the observed
52: electron-shell structure are similar to those of alkali and noble
53: metals, a sequence of extremely stable wires plays a unique role
54: in Aluminum. This series appears isolated in conductance
55: histograms and can be attributed to ``superdeformed''
56: non-axisymmetric nanowires.
57: 
58: \end{abstract}
59: \pacs{73.40.Jn, 61.46.+w, 68.65.La}
60: 
61: 
62: \maketitle
63: 
64: 
65: 
66: 
67: \section{Introduction}
68: 
69: Metal nanowires have attracted considerable interest in the past
70: decade, experimental research has burgeoned, and various
71: techniques have been employed to produce them for many different
72: materials (see\ \onlinecite{Agrait03} for a recent review).
73: Conductance  histograms, built out of thousands of conductance
74: traces recorded during the breaking of a nanowire in a
75: mechanically controllable break junction (MCBJ), reveal the
76: existence of exceptionally stable wires: Due to their stability,
77: these wires are formed more often than others, and show up as
78: peaks in a conductance histogram taken at a temperature large
79: enough to allow the contact to explore the available
80: configuration space.
81: 
82: 
83: Such experiments on alkali metals
84: \cite{yanson99,Yanson01,Yanson04} have revealed a sequence of
85: stable ``magic'' wires that is periodic in $\sqrt{G}$, where $G$
86: is the electrical conductance. A theoretical analysis of nanowire
87: stability \cite{kassubek01,urban04,burki05a} within the nanoscale
88: free-electron model (NFEM)
89: \cite{stafford97,ruitenbeek97,Buerki05b} has connected this magic
90: sequence of wires to the electron shell structure, similar to
91: what is known for metal clusters
92: \cite{deheer93,brack93,martin96}. Groups of discrete transverse
93: electronic states are bunched together, with gaps between them.
94: When the electronic Fermi energy is within one of these gaps, the
95: wire is exceptionally stable. The periodicity in $\sqrt{G}$ can
96: be understood semi-classically: writing the mesoscopic
97: contribution to the wire energy as a series in classical periodic
98: orbits in a circular billiard
99: \cite{yanson99,kassubek01,Buerki05b}, the main contribution is
100: found to come from the diameter orbit, and to be periodic in the
101: wire radius, and thus proportional to $\sqrt{G}$ via the Sharvin
102: formula \cite{torres94,hoppler98}.
103: A supershell structure was also observed \cite{Yanson00}
104: in the form of a periodic modulation of the peak heights, and it
105: was argued that the stable wires at the nodes of the supershell
106: structure are non-axisymmetric \cite{urban04,urban04b,Urban06}.
107: 
108: 
109: At larger wire diameters, and therefore larger conductance, a
110: crossover to an atomic shell structure was found \cite{yanson01a}.
111: In this case, shell closings correspond to crystalline wires with the
112: completion of additional atomic layers.
113: Recently, electron-shell structure has also been observed for the
114: noble metals gold \cite{Diaz03,mares04}, silver \cite{mares05}, and
115: copper \cite{mares05}. Except for the latter, the crossover to
116: atomic shell effects at larger diameter could be
117: confirmed \cite{mares04,mares05}.
118: 
119: Inspired by these results and by the research done on metal
120: clusters, we extend our investigation to multivalent metals,
121: choosing as a first step aluminum.
122: 
123: 
124: 
125: Conductance histograms for aluminum atomic size contacts recorded
126: using the MCBJ technique at 4.2 K were first reported by Yanson
127: \etal \cite{yanson97}. Four peaks were identified situated close
128: to 1, 2, 3 and 4 times the conductance quantum $G_0=2e^2/h$.
129: Scheer \etal \cite{scheer97} analyzed the transmission modes for
130: Al point contacts by recording current-voltage characteristics in
131: the superconducting state. For a single atom, three modes were
132: found to contribute to the conductance, with individual
133: transmissions lower than one summing together to a total
134: conductance of about one quantum unit.
135: 
136: 
137: A previous experiment on aluminum shell structure \cite{medina03}
138: reports both experimental results and molecular dynamics
139: simulations of the evolution of Al nanowires while breaking. The
140: experiment was performed using a scanning tunnelling microscope
141: (STM) at room temperature (RT) in ultra high vacuum (UHV). The
142: authors attribute the oscillation period of the peaks to filling
143: of octagonal facets, claiming that the electronic shell effect is
144: absent. Our results reported below are significantly different
145: since we present evidence of electronic shell effects. The
146: difference with our experiments appear in several aspects:
147: Firstly, in the histogram from \onlinecite{medina03} there is a
148: large contribution in the low conductance regime, including below
149: 1 $G_0$. This is usually an indication of the presence of
150: contaminants, especially expected in the case of STM soft
151: indentations \cite{hansen00}. Secondly, in \onlinecite{medina03}
152: the peaks have low weight and a derivative is performed in order
153: to identify their position. Moreover, from one single histogram
154: one cannot exclude the occurrence of electronic shell effects,
155: since electronic and atomic shell effects compete in the
156: stabilization of the nanowire and in different measurements one
157: or the other effect can dominate, as it will be shown below.
158: Although the experimental results in \onlinecite{medina03} agree
159: with the theoretical predictions based on molecular dynamics
160: simulations, the latter calculation does not take into account
161: the electronic contribution to the total energy of the nanowire,
162: excluding the electronic shell effects from consideration.
163: 
164: This paper is organized as follows. The experimental results are
165: presented in \sref{sec:ExperimentalResults}, followed by a
166: detailed analysis of the data in \sref{sec:DataAnalysis}. A
167: theory explaining the observed electronic shell structure and
168: allowing the prediction of the shapes of the wires is presented in
169: \sref{sec:ElectronicShellEffects}. Atomic shells and the
170: crossover between electronic and  atomic shell effects are
171: analyzed in \sref{sec:AtomicShellEffects}.
172: \Sref{sec:Superdeformed} addresses the issue of superdeformed
173: nanowires. Finally, a short summary is given in \sref{sec:Summary}.
174: 
175: 
176: 
177: \section{Experimental results}
178: \label{sec:ExperimentalResults}
179: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
180: 
181: 
182: \begin{figure}[b]     %%%  Figure 1
183:   \begin{center}
184:     \includegraphics[width=8cm]{figure1__Al4010b+inset.eps}
185:   \caption{(a) Conductance histogram for aluminum at RT-UHV constructed
186:     from 7000 scans at a bias voltage of 50 mV. The inset shows the
187:     presumed atomic positions in the normal cross-section of the wire. In (b) we plot the
188:     conductance values at the peak position vs.\ their index while in
189:     (c) $ \sqrt{G/G_0}$ for the peak positions vs.\ peak index. For (c)
190:     a good linear dependence is obtained with a period $\Delta
191:     \sqrt{G/G_0}$=0.29 $\pm$0.002.} \label{fig:Al4010}
192:   \end{center}
193: \end{figure}
194: 
195: 
196: 
197: 
198: We report conductance histograms recorded for Al nanowires in UHV
199: at room temperature using the MCBJ set-up presented in
200: \onlinecite{mares04} and \onlinecite{mares05}. A metallic nanowire
201: is bent in a controlled way and its diameter at the weakest point
202: is reduced down to atomic size. The wire is initially broken under
203: UHV, ensuring in this way clean surfaces. In order to perform a
204: reliable statistical analysis we have investigated six different
205: samples. The behavior was reproduced on every sample and typical
206: histograms are shown in figures \ref{fig:Al4010}, \ref{fig:G22},
207: \ref{fig:histSuperDeformed}. In the histogram from
208: \fref{fig:Al4010} we can see a remarkably long series of peaks up
209: to a conductance value of 100 $G_{0}$.
210: 
211: 
212: The first step in finding the origin of the peaks in the
213: conductance histogram is to verify whether the peaks show a
214: certain periodicity. We can see in \fref{fig:Al4010} (b) and (c)
215: that a linear dependence is found when plotting the square root
216: of the conductance as a function of the peak index, while when
217: plotting directly the conductance the dependence is far from
218: linear. Note that the square root of the conductance is
219: proportional to the root-mean-square (rms) radius $\rho$ of the
220: wire, as can be inferred from the corrected Sharvin formula
221: \cite{torres94,hoppler98},
222: \begin{eqnarray}
223: \label{eq:SharvinConductance}
224:     G_S&=&G_0\left(\frac{k_F^2\rho^2}{4}-\frac{k_F{\cal
225:     P}}{4\pi}+\frac{1}{6}\right),
226: \end{eqnarray}
227: which gives a semi-classical approximation to the conductance.
228: Here $\cal P$ is the perimeter of the wire's cross section and $k_F$ is the Fermi
229: wavevector. In the case of the histogram of \fref{fig:Al4010}
230: the period found is $\Delta \sqrt{G/G_0}$=0.29 $\pm$0.002.
231: 
232: 
233: \begin{figure}[b]     %%% Figure 2
234:   \begin{center}
235:      \includegraphics[width=8cm]{figure2__G22.eps}
236:   \end{center}
237:   \caption{Conductance histogram for Al taken under RT-UHV conditions, recorded at a
238:      bias voltage of 20 mV constructed from 2000 scans. Inset: the peak
239:      oscillation has two periods: in the low conductance regime up to
240:      about 40 $G_{0}$ the period is $\Delta \sqrt{G/G_0}$= 0.47 $\pm$0.03 whereas for
241:      larger diameters we find $\Delta \sqrt{G/G_0}$= 0.25 $\pm$0.01.}
242:      \label{fig:G22}
243: \end{figure}
244: 
245: As for the other metals, we have studied how the histograms vary
246: from one experiment to the next, and a characteristically
247: different histogram can be seen in \fref{fig:G22}. There a
248: crossover to a different period occurs. Two periods of oscillation
249: can be identified around $\Delta \sqrt{G/G_0}$=0.47 and 0.25, with
250: a crossover at a conductance value of about 40 $G_{0}$. The first
251: period is similar to the period of the electronic shell effect in
252: noble metals \cite{mares04,mares05}. The second one, similar to
253: the period of the histogram of \fref{fig:Al4010}, is somewhat
254: larger than what was obtained for the atomic shell effect in noble
255: metals, for which typically $\Delta \sqrt{G/G_0}$ is found to be
256: around 0.21.
257: 
258: 
259: \begin{figure}     %%% Figure 3
260:   \begin{center}
261:      \includegraphics[width=8cm]{figure3__G51422.eps}
262:   \end{center}
263:      \caption{Special type of histograms recorded for Al
264:      at RT-UHV showing two dominant peaks near 14 and 22
265:      $G_{0}$, while the lower curve shows a third peak near 5 $G_{0}$.
266:      The two curves are vertically offset for clarity.
267:      The histograms are recorded on two different samples,
268:      both at a bias voltage of 20 mV containing 1000 traces each.
269:      In the individual conductance traces long
270:      plateaus can be seen at these values, shown in the inset.}
271:      \label{fig:histSuperDeformed}
272: \end{figure}
273: 
274: Sometimes a new type of histogram appears having only two or
275: three peaks as one can see in \fref{fig:histSuperDeformed}. This
276: type of histogram appears at the beginning of the measurements on
277: a fresh sample just after the initial breaking, and we see it in 7
278: out of 30 histograms. The position of the two major peaks is
279: always close to 14 and 22 $G_{0}$. In two of these histograms a
280: peak near 5 $G_{0}$ also appears. All three peaks are related to
281: very long plateaus at these values in the individual conductance
282: traces [see inset of \fref{fig:histSuperDeformed}].
283: 
284: 
285: Finally, in \fref{fig:GAlair} we show a conductance histogram
286: recorded for aluminum in air at room temperature. We see distinct
287: peaks that have a period close to the electronic shell effect
288: found in noble metals and for Al in UHV. It is known that bulk Al
289: is a very reactive metal that forms an oxide layer on the surface.
290: Possibly this layer prevents the immediate contamination inside
291: the bulk of the nanowire.
292: 
293: \begin{figure}[t]       %%% Figure 4
294:   \begin{center}
295:     \includegraphics[width=8cm]{figure4__GAlair.eps}
296:   \end{center}
297:   \caption{Conductance
298:     histogram for Al at room temperature in air constructed from 1000
299:     traces at a bias voltage of 20 mV. Inset: the peak oscillation
300:     has a period  $\Delta \sqrt{G/G_{0}}$=0.48 $\pm 0.04$.}
301:   \label{fig:GAlair}
302: \end{figure}
303: 
304: 
305: 
306: \section{Theoretical analysis of Al histograms}\label{sec:DataAnalysis}
307: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
308: 
309: In order to obtain more information about the stable wire
310: configurations realized in experiments, we have analyzed in
311: detail the conductance histogram of \fref{fig:G22}. We fit the
312: data set with a function consisting of a sum of peaks of
313: lorentzian shape with height, width and peak position as fitting
314: parameters, and the result is shown in \fref{fig:LorentzFit} and
315: \tref{tab:LorentzFit}. Note that the number of peaks used for the
316: fit is not unique, as the fit can always be improved by including
317: additional broad peaks of lower intensities. However, the
318: parameters of additional peaks can not be determined with high
319: confidence, especially at larger conductance where the curve is
320: more noisy.
321: 
322: \begin{figure}[t]       %%%  Figure 5
323:   \begin{center}
324: %     \includegraphics[width=8cm, angle=-90]{fig.ps}
325:      \includegraphics[width=8cm, angle=-90]{figure5__CompareExpTheoD.eps}
326:   \end{center}
327:   \caption[]{Experimental conductance histogram data from \fref{fig:G22}
328:     fitted with lorentzian peaks.  The dotted lines show the individual peaks.
329:     The positions, weights, and widths of the peaks are given in \tref{tab:LorentzFit}.}
330:   \label{fig:LorentzFit}
331: \end{figure}
332: 
333: 
334: 
335: 
336: 
337: Besides the 18 dominant conductance peaks, marked with numbered
338: black arrows in \fref{fig:LorentzFit}, we can identify a
339: number of very shallow peaks in the conductance histogram. They
340: originate from less stable wire configurations, that are less
341: frequently realized. These ``peaks'' are marked with capital
342: letters A--D in \fref{fig:LorentzFit}.
343: 
344: \Fref{fig:Periodicity} shows $\sqrt{G}$ versus peak number for
345: three different peak-selection criteria: Peaks 1--18 of
346: \tref{tab:LorentzFit} are plotted as red squares, reproducing the
347: inset of \fref{fig:G22}, with a transition from electron-shell to
348: atomic-shell structure at a conductance $G\sim 40 G_0$.  When the
349: shallow peaks A--D are included (green stars), the transition to
350: atomic-shell structure occurs already at $G\sim 20 G_0$.
351: Conversely, when only the largest-weight peaks ($>5\%$) at high
352: conductance are included (blue $+$ symbols), the slope
353: corresponding to electron-shell structure is seen to extend all
354: the way up to the highest peak fit, $G\sim 60 G_0$.  Note that at
355: a temperature sufficient to display electron-shell effects, the
356: weights of the peaks with $G< 20 G_0$ may be suppressed due to
357: thermally-induced breaking of the contact, as previously observed
358: in alkali metals \cite{yanson99}. \Fref{fig:Periodicity} suggests
359: that electronic- and atomic-shell effects compete in the regime
360: $20 G_0 < G < 60 G_0$, with atomic-shell effects becoming
361: increasingly important with increasing $G$.
362: 
363: 
364: 
365: \Table{\label{tab:LorentzFit} Conductance $G_{\rm exp}$, relative
366: weight, and width $\Delta G$ of the peaks in the histogram shown
367: in \fref{fig:LorentzFit}. The values were obtained by fitting the
368: data set to a sum of lorentzian peaks, with height, width, and
369: peak position as fitting parameters. The next two columns give the
370: parameters of the stable Al nanowires determined by a linear
371: stability analysis within the NFEM: $G_{th}$ is the quantized
372: conductance and $\lambda$ is the quadrupole deformation [cf.\
373: \eref{eq:quadrupole}]. The last two columns give the mean channel
374: transmission $\langle T_n \rangle\equiv G_{exp}/G_{th}$ or series
375: resistance $R_s$ needed to shift the theoretical peaks down to the
376: corresponding experimental values. For completeness, the NFEM
377: predictions are shown for peaks 1--5, although the model is not
378: expected to be quantitatively accurate for few-atom contacts in
379: multivalent metals, as evidenced by the large values of the
380: effective series resistance, or the small channel transmissions
381: \cite{scheer97}. The last three lines of the Table give the
382: positions of the three main peaks of \fref{fig:histSuperDeformed},
383: and the corresponding theoretical parameters.}
384: % {cccccccc}
385: % \begin{indented}
386: % \item[]\begin{tabular}
387: % {|c|c|c|c|c|c|c|c|}
388: %     \colrule
389:     \br
390:     peak&
391:      $\begin{array}{cc} {G}_{\rm{exp}} \\ \left[G_0\right]\end{array}$ &
392:      $\begin{array}{cc} {\rm{weight}} \\ \left[\%\right]\end{array}$ &
393:      $\begin{array}{cc} \Delta{G} \\ \left[G_0\right]\end{array}$ &
394:      $\begin{array}{cc} {G_{th}} \\ \left[G_0\right]\end{array}$ &
395:      $\lambda$ &
396:      $\langle T_n \rangle$ &
397:      $\begin{array}{cc} {R_s} \\ \left[\Omega\right]\end{array}$
398:     \\
399: %     \colrule
400:     \mr
401:   1  &  1.4 & 0.2 & 0.5  & 3  & 0    & 0.45 & 5185 \\
402:   2  &  3.6 & 0.8 & 0.7  & 5  & 0.17 & 0.72 & 993 \\
403:   3  &  5.3 & 1.1 & 1.0  & 6--7 & 0--0.3 & 0.76--0.88 & 289--596 \\
404:   4  &  7.5 & 0.2 & 0.7  & 9   & 0.11 & 0.83 & 291 \\
405:   5  &  9.5 & 1.3 & 1.4  & 12  & 0   & 0.79 & 283 \\
406: %   \colrule
407:     \mr
408:   6  & 13.5 & 1.3 & 1.3  & 14 & 0.3 & 0.96 & 36 \\
409:   7  & 16.5 & 9.2 & 1.5  & 17 & 0    & 0.97 & 24 \\
410:   A  & 18.6 & 0.5 & 1.3  & -- & --   & -- & -- \\
411:   8  & 20.8 & 6.7 & 2.2  & 23 & 0    & 0.90 & 59 \\
412:   B  & 23.7 & 1.6 & 1.3  & -- & --   & -- & -- \\
413:   9  & 25.8 & 9.5 & 2.8  & 29 & 0.06 & 0.89 & 55 \\
414:   C  & 28.3 & 0.5 & 1.0  & -- & --   & -- & -- \\
415:   10 & 30.8 & 10.5 & 2.5 & 34 & 0    & 0.90 & 40 \\
416:   D  & 33.7 & 0.6 & 1.4  & -- & --   & -- & -- \\
417:   11 & 36.7 & 9.0 & 3.4  & 42 & 0    & 0.87 & 45 \\
418:   12 & 40.0 & 3.7 & 2.2  & -- & --   & -- & -- \\
419:   13 & 43.4 & 8.7 & 3.2  & 51 & 0    & 0.85 & 44 \\
420:   14 & 46.7 & 3.5 & 2.6  & -- & --   & -- & -- \\
421:   15 & 50.4 & 8.9 & 3.6  & 59 & 0.05 & 0.85 & 37 \\
422:   16 & 53.9 & 4.7 & 2.7  & -- & --   & -- & -- \\
423:   17 & 57.7 & 8.6 & 4.5  & 67 & 0    & 0.86 & 31 \\
424:   18 & 61.8 & 8.9 & 3.5  & 72 & 0.04 & 0.86 & 30 \\
425: %     \colrule
426:   \mr
427:   I  &  5.3 & --  & --   & 6--7 & 0.3 &  0.76--0.88 & 289--596 \\
428:   II  & 14.1 & -- & -- & 14 & 0.3 & 1.0 & 0 \\
429:   III  & 21.9 & -- & -- & 23 & 0.3 & 0.95 & 28 \\
430: %   E  &  5.3 & ?  & ?   & 6--7 & 0.3 &  0.76--0.88 & 289--596 \\
431: %   F  & 14.1 & 20 & 1.8 & 14 & 0.3 & 1.0 & 0 \\
432: %   G  & 21.9 & 51 & 2.9 & 23 & 0.3 & 0.95 & 28 \\
433: %   \colrule
434:   \br
435: % \end{tabular}
436: % \end{indented}
437: %  \label{tab:LorentzFit}
438:  \endTable
439: % \end{table}
440: 
441: 
442: 
443: \subsection{Electronic Shell Effects}
444: \label{sec:ElectronicShellEffects}
445: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
446: 
447: Recently, the stability of alkali and noble metal nanowires was
448: extensively studied theoretically
449: \cite{kassubek01,urban04,burki05a,urban04b,Urban06,Urban03,Buerki03}
450: using the Nanoscale Free-Electron Model (NFEM)
451: \cite{stafford97,Buerki05b}. It was found that nanowire stability
452: is determined by a competition between surface and shell
453: contributions. A linear stability analysis considering
454: cylindrical geometries as well as wires with broken axial
455: symmetry revealed a ``magic'' sequence of stable nanowires, that
456: allows for a consistent interpretation of experimental
457: conductance histograms for alkali metals, including both the
458: electronic shell and supershell structures
459: \cite{urban04,urban04b}. The same sequence of stable wires is
460: observed in histograms for Au \cite{Diaz03,mares04}, suggesting a
461: possible universality.
462: 
463: 
464: \begin{figure}[t]      %%%  Figure 6
465:   \begin{center}
466:     \includegraphics[angle=270, width=0.6\textwidth]{figure6__AtomicShell-transition.eps}
467:   \end{center}
468:   \caption{(colour online) Linear dependence of the peak positions in $\sqrt{G/G_{0}}$ using
469:     various selection criteria:
470:     (i) Main peaks, as in the inset of \fref{fig:G22} (red squares);
471:     (ii) Keeping only the peaks with weight larger than 5\% at large
472:     conductance (blue $+$ symbols);
473:     (iii) Keeping all peaks in \tref{tab:LorentzFit} (green stars).
474:     }
475:   \label{fig:Periodicity}
476: \end{figure}
477: 
478: Guided by the importance of conduction electrons in the cohesion
479: of metals, and by the success of the jellium model in describing
480: metal clusters \cite{deheer93,brack93}, the NFEM replaces the
481: metal ions by a uniform, positively charged background that
482: provides a confining potential for the electrons. The electron
483: motion is free along the wire, and confined in the transverse
484: directions. The surface properties of various metals can be fit
485: by using appropriate surface boundary conditions
486: \cite{Garcia-martin96,urban04}.
487: 
488: 
489: Aluminum is a trivalent metal, but the Fermi surface of bulk Al
490: resembles a free-electron Fermi sphere in the extended-zone
491: scheme. This suggests that the NFEM might be applicable to Al
492: nanowires, although the continuum approximation is more severe
493: than for monovalent metals. A first-principles theoretical
494: calculation of the electronic structure of Al nanowires indeed
495: reveals that, although the atomic structure has to be taken into
496: account, the band structure has a free-electron like dispersion
497: relation \cite{ditolla00}. Free-electron jellium
498: models \cite{brack93} have already successfully been used for
499: modeling Al clusters \cite{Heer89,Milani91,Jarrold93} and
500: explaining the occurrence of magic
501: numbers \cite{Schriver90,Persson91,Pellarin93}. De Heer \etal
502: \cite{Heer89,Milani91} have specifically addressed the
503: question of the applicability of the jellium model by comparing
504: measured electronic properties of Al$_n$ clusters to the values
505: calculated with the jellium model. For the polarizability they
506: find good agreement \cite{Heer89} for $n>40$, while when
507: considering ionization potentials and electron affinities, it is
508: found that clusters with $n>13$ are already accurately described
509: by the jellium model \cite{Milani91}.
510: % Using the relation between the Fermi wavevector and
511: Using a charge neutrality argument, the atomic volume is given by
512: $v_a = 9\pi^2/k_F^3$ for a trivalent metal, the radius $R$ of an
513: Al$_{13}$ cluster can be estimated as $k_F R = 6.5$, while that of
514: Al$_{40}$ is $k_F R = 9.5$. Aluminum nanowires with these radii
515: would have electrical conductances of $7~G_0$ and $18~G_0$,
516: respectively, according to \eref{eq:SharvinConductance}. We
517: therefore expect the NFEM to become quantitatively accurate for
518: nanowires with $G \gsim 12 G_0$.
519: 
520: Ogando \etal \cite{ogando02} have self-consistently calculated
521: the electron shell potential for Aluminum within a local density
522: approximation using the stabilized jellium model. The free energy
523: shows oscillations characteristic of the shell effect. Despite
524: using a more elaborate self-consistent jellium model, their
525: results predict the same sequence of energetically preferred
526: cylindrical wires and the same positions for the supershell beat
527: minima as the NFEM, which \emph{a posteriori} underlines the
528: applicability of our approach.
529: 
530: Guided by the points given above, we use the NFEM to interpret
531: the electron-shell structure observed in \fref{fig:G22}. A
532: general method for performing linear stability analysis of metal
533: nanowires, including the modelling of material-specific surface
534: properties, was recently presented \cite{Urban06}. Here, we apply
535: the method of \onlinecite{Urban06} to Aluminum and concentrate on
536: presenting the results.
537: 
538: We examine straight wires aligned along the $z$-axis and, in order
539: to account for an energetically favorable symmetry breaking
540: (Jahn-Teller effect), both cylindrical and quadrupolar cross
541: sections are considered. The wire geometry is given by the radius
542: function
543: %
544: \begin{equation}
545:     r(\varphi)=\rho\left(\sqrt{1-\lambda^2/2}+\lambda\cos(2\varphi)\right)
546: \label{eq:quadrupole}
547: \end{equation}
548: %
549: in cylindrical coordinates $r,\,\varphi,\,z$, characterized by
550: the rms radius $\rho$ and the quadrupole deformation parameter
551: $\lambda$. For a given geometry $(\rho,\lambda)$ we calculate the
552: grand canonical potential $\Omega$ of the electron gas and
553: examine its change with respect to small perturbations. Only if
554: $\Omega$ increases for every possible small deformation of the
555: initial geometry is the wire (linearly) stable. Note that this
556: rigorous thermodynamic definition of stability is not the same as
557: identifying stable wires as minima of the shell potential, as is
558: done for example in \onlinecite{ogando02}. Although the
559: most-stable wires do correspond to minima of the shell potential,
560: not every minimum of the shell potential corresponds to a stable
561: wire, as has been shown by considering non-axisymmetric
562: deformations \cite{Urban06}.
563: 
564: \begin{figure}[t]       %%% Figure 7
565: \newlength\figonewidth\setlength\figonewidth{12.8cm}
566: %     \begin{minipage}[c]{\figonewidth}\hspace*{-6mm}
567:   \begin{center}
568:     \includegraphics[width=\figonewidth,clip=,draft=false]{figure7__Stabdia-Al-m2.eps}
569:   \end{center}
570: %     \end{minipage}\
571: %     \begin{minipage}[c]{4.5cm}
572:   \caption[]{(colour online) Energetically stable cylindrical and
573:       quadrupolar Al wires as function of temperature (in units of
574:       $T_\rho=T_F/k_F\rho$, see text). The deviation from axial
575:       symmetry is coded via the gray (colour) scale shown on the
576:       right. The numbered arrows label the configurations that
577:       enter the analysis of the shell structure and give the
578:       number of the corresponding peaks in the histogram.
579:      % The dashed area at $G=6\,G_0$ correspond to an unphysical
580:      % axisymmetric wire, discussed in the text.
581:       }
582:   \label{fig:StabdiaAl}
583: %     \end{minipage}\vspace*{-3mm}
584: \end{figure}
585: 
586: 
587: 
588: 
589: \Fref{fig:StabdiaAl} displays the results of the stability
590: analysis as a function of temperature: shaded (coloured) areas
591: represent stable geometries, where the deviation from axial
592: symmetry $\lambda$ is coded via a gray (colour) scale. The
593: temperature is displayed in units of $T_\rho\equiv T_F/k_F\rho$,
594: where $T_F$ is the Fermi temperature, which provides a good
595: measure of the total stability of a metal nanowire \cite{Urban06}.
596: 
597: 
598: A series of very stable cylindrical wires can be identified, with
599: conductance values $G/G_0=3,6,12,17,23,34,42,51,...$ Their
600: absolute stability is periodically modulated, and remarkably
601: stable quadrupolar wires are found at the nodes of this supershell
602: structure at $G/G_0=9,29,59,...$ These two series combined form
603: the ``universal'' series of stable wires explaining the observed
604: shell and supershell structures in alkali and noble metals. Two
605: additional stable highly-deformed wires that are well separated in
606: conductance from their neighbors, and are thus expected to give
607: pronounced peaks in the conductance histogram, are labeled peak
608: Nos.\ 2 and 6 in \fref{fig:StabdiaAl}.  The theoretical quantized
609: conductance values and quadrupole deformations of this sequence
610: of stable nanowires  are given in \tref{tab:LorentzFit}.
611: 
612: \begin{figure}[t]     %%%  Figure 8
613:   \begin{center}
614:     \includegraphics[angle=270,width=0.7\textwidth]{figure8__ElectronShells2.eps}
615:   \end{center}
616:   \caption{(colour online) Comparison of the stable Al nanowires
617:     within the NFEM (solid circles) with the dominant experimental
618:     peaks (blue symbols with error bars). Here the experimental error
619:     bars indicate the peak widths, and the theoretical conductance
620:     values were corrected by a mean series resistance of 36$\Omega$.
621:     The NFEM predictions for the low-conductance peaks are shown in
622:     green, for completeness.
623:   \label{fig:NFEM_vs_EXP} }
624: \end{figure}
625: 
626: 
627: 
628: In \fref{fig:NFEM_vs_EXP}, this theoretical series of most stable
629: wires is compared to the dominant (electron-shell) peaks extracted
630: from the histogram of \fref{fig:G22}, plotted as (blue) $+$
631: symbols in \fref{fig:Periodicity}. The theoretical conductance
632: values have been corrected by subtracting a series resistance of
633: $36\,\Omega$, which was fitted to phenomenologically describe the
634: backscattering of electron waves by defects, surface
635: irregularities, and phonons \cite{Ludoph99b, Buerki99}. Aside from
636: a systematic shift in the theoretical peaks with $G\leq 12 G_0$,
637: where the NFEM is expected to be inaccurate, there is a perfect
638: one-to-one correspondence of the stable Al nanowires determined
639: within the NFEM and the experimental electron-shell structure.
640: This correspondence is also shown in \tref{tab:LorentzFit}, along
641: with the mean channel transmission $\langle T_n \rangle$ (or
642: series resistance $R_s$) necessary to shift the theoretical
643: quantized conductance values down to the experimental values. As
644: expected \cite{scheer97}, $\langle T_n \rangle$ is significantly
645: less than one for contacts with 3--6 channels due to strong mode
646: mixing of the $s$- and $p$-orbitals, but it approaches unity for
647: $G>12G_0$.  Eventually, $\langle T_n \rangle$ is expected to
648: decrease again as the system crosses over from ballistic to
649: diffusive transport.
650: 
651: %JB and CAS: Proposed modified new paragraph
652: The NFEM predicts a rather universal sequence of dominant peaks in
653: conductance histograms, stemming from electron-shell effects. This
654: sequence is expected to be present for free-electron-like metals,
655: and is consistent with previous experiments on alkali
656: \cite{yanson99,urban04} and noble metals \cite{Diaz03,mares04},
657: and is confirmed by the data on Aluminum reported here. For non
658: free-electron-like materials, other models must be used to
659: determine the stability. For example, classical molecular dynamics
660: simulations such as \cite{Pauly06}, predict very different,
661: material specific (i.e. nonuniversal) low conductance histograms
662: for Ag, Pt and Ni nanowires.
663: 
664: 
665: 
666: 
667: \subsection{Atomic Shell Effects}
668: \label{sec:AtomicShellEffects}
669: 
670: Similar to what is seen in aluminum clusters \cite{li98b}, the
671: conductance histogram in \fref{fig:G22} shows a crossover to a
672: different peak period of about $\Delta \sqrt{G/G_0}\simeq$ 0.25 at
673: larger conductance values. This shift in periodicity is associated
674: with the appearance of a series of peaks of increasing intensity
675: (cf.\ peaks A, B, C, D, 12, 14, and 16 in \fref{fig:LorentzFit})
676: that cannot be explained in terms of electron shell effects. This
677: new periodicity can be understood as an atomic shell effect: at
678: large diameters the nanowire will be crystalline and stable wires
679: are expected to form for densely-packed structures with closed
680: facets, that minimize the surface energy. In the case of
681: face-centered cubic lattice, the [110] orientation has been shown
682: to be most favorable to form nicely faceted long nanowires
683: \cite{Rodrigues00,jagla01}. We assume that the bulk fcc packing of
684: aluminum is preserved in nanowires similar to what has been
685: observed for gold and silver nanowires in transmission electron
686: microscopy by Rodrigues {\it et al.} \cite{Rodrigues00,
687: rodrigues02}. The lowest energy surfaces for Al are the $(111)$
688: planes \cite{vitos98} and we can construct a densely packed wire
689: along a $[110]$ axis with four $(111)$ facets and completing the
690: hexagon with two larger $(100)$ facets. One obtains a nearly
691: hexagonal cross-section of the wire illustrated in the inset of
692: \fref{fig:Al4010}. The faceting structure of stable wires is
693: reproduced in a Monte Carlo simulation of the thinning-down of a
694: fcc wire by Jagla and Tosatti \cite{jagla01}. Filling a whole
695: atomic shell results in an increase of the nanowire radius with
696: the period of $\Delta \sqrt{G/G_0}=1.98$. Assuming that stable
697: wires are found when completing a single facet of the hexagon, one
698: obtains a six times smaller period $\Delta \sqrt{G/G_0}=0.33$. An
699: average experimental period obtained from 11 separate measurements
700: would give an value of about $\Delta \sqrt{G/G_0}\sim $0.3.
701:  As in the case of the electronic
702: shell effect period, the agreement can be improved by correcting
703: the theoretical value by subtracting a mean series resistance of
704: $36\,\Omega$.
705: 
706: On top of this correction, a lower period is expected due to the
707: interplay between electronic and atomic shell effects. The
708: relatively higher slopes observed for samples where the atomic
709: shell effect is dominant, like in the histogram of
710: \fref{fig:Al4010}, as compared to the histogram of \fref{fig:G22}
711: comes in support to this assumption.
712: 
713: The crossover between electronic and atomic shell filling varies
714: between different experiments, similar to what was found for
715: alkali and noble metal nanowires \cite{yanson01a, mares04,
716: mares05} which is likely due to differences in the local crystal
717: orientation of the leads connecting the nanowire. On top of this,
718: for a given histogram the cross-over position is somewhat
719: arbitrary, depending on the peak-selection criteria (see
720: \fref{fig:Periodicity}). When considering only the dominant peaks
721: in the conductance histogram, mostly a value close to 40 $G_{0}$
722: is observed. Still, the crossover from electronic to atomic shell
723: effects is found to be a smooth transition rather than a sharp
724: feature, if we include the number of shallow ``peaks''(cf.\ peaks
725: A, B, C, and D in \fref{fig:LorentzFit}) which start to show up at
726: $G\simeq19G_0$ (cf.\ \fref{fig:Periodicity} and discussion).
727: 
728: These findings are in striking analogy to what was previously
729: found for Al clusters. The early photoionization spectrometry
730: measurements \cite{Schriver90,Persson91,Pellarin93} on Al$_n$
731: report electronic shell structure up to $n\sim100$ while
732: subsequent experiments \cite{Martin92,Naher93} give evidence for
733: atomic shell effects for $n\gsim200$. The crossover from
734: electronic to atomic shell effects was studied in
735: \onlinecite{li98b,Baguenard94} and a crossover in the range of
736: $n\sim75$ -- 100 atoms was reported \onlinecite{li98b}. We can
737: convert the number of atoms to a radius as in
738: \sref{sec:ElectronicShellEffects} using $v_{a}=9\pi^2/k_F^3$.
739: Then the crossover occurs in the range of $k_F\rho\simeq11.7$ --
740: 12.8, which corresponds to $G/G_0\simeq28$ -- 35 via
741: \eref{eq:SharvinConductance}, which agrees well with our
742: observations.
743: 
744: 
745: 
746: 
747: 
748: 
749: \subsection{Superdeformed wires}
750:  \label{sec:Superdeformed}
751: 
752: 
753: \begin{figure}      %%%  Figure 9
754:   \begin{center}
755:     \includegraphics[angle=270, width=0.7\textwidth]{figure9__ConfigurationSpace.eps}
756:   \end{center}
757:   \caption{(colour online) Stability diagram for Al wires at fixed temperature $T=0.4\,T_\rho$.
758:     Thick (red) lines mark stable wires in the configuration space of conductance $G$ and
759:     deformation parameter $\lambda$. The box emphasizes a series of very stable
760:     superdeformed wires, whose peanut-shaped cross section is shown as an inset.
761:     This sequence is thought to
762:     be responsible for the type of histogram shown in \fref{fig:histSuperDeformed}.}
763:   \label{fig:StabdiaAlphasespace}
764: \end{figure}
765: 
766: The third type of histogram seen in our experiment (cf.\
767: \fref{fig:histSuperDeformed}) is qualitatively different. It
768: shows up only at the beginning of the measurement on a fresh
769: sample, just after the initial breaking.  Two dominant peaks near
770: 14 and 22 $G_0$ are observed, along with a third peak near 5
771: $G_0$ in two out of seven cases. An intriguing possibility is
772: that this third type of histogram is due to an isolated sequence
773: of ``superdeformed'' nanowires, that is, nanowires whose cross
774: section has an aspect ratio near 2:1.  Superdeformation is a
775: well-known feature of nuclear shell structure \cite{Nolan88}, but
776: has not been previously observed in nanowires.
777: 
778: Our stability analysis shows that the series of wires with
779: $\lambda\sim0.3$ ($\mbox{aspect ratio} \approx 1.9$), with
780: conductances of 6, 14, and 23 $G_0$, although less stable than the
781: magic cylinders, are quite stable compared to other
782: highly-deformed structures (cf.\ \fref{fig:StabdiaAl}). More
783: importantly, they are very isolated in configuration space, as
784: illustrated in \fref{fig:StabdiaAlphasespace}. Thus, if the
785: initial structure of the nanocontact formed in the break junction
786: is rather planar, with a large aspect ratio, then as the junction
787: is pulled apart, it will maintain a large aspect ratio as it necks
788: down elastically. Along such a trajectory in configuration space
789: (schematically indicated by a blue arrow in
790: \fref{fig:StabdiaAlphasespace}), the most stable structures
791: available are those of the superdeformed family (indicated by the
792: blue box).  Moreover, it is plausible that under elastic
793: deformation during repeated elongation/compression cycles, the
794: system would simply jump back and forth between these
795: superdeformed structures, which are geometrically similar to each
796: other. Eventually, however, a large structural fluctuation could
797: drive the nanocontact into the axisymmetric lower energy minimum,
798: where it would remain during further breaking cycles. This
799: scenario is analogous to the decay-out process of highly-excited
800: superdeformed nuclei, which remain in the superdeformed energy
801: well thoughout a long sequence of decays, before finally
802: tunnelling into the (lower minimum) normal-deformed energy well
803: \cite{Nolan88}.
804: 
805: 
806: Evidence of superdeformation has not been reported in any of the
807: previous experiments on alkali and noble metals, presumably
808: because highly-deformed structures are intrinsically less stable
809: than nearly axisymmetric structures, due to their larger surface
810: energy. Aluminum is unique in this respect, since its surface
811: tension is very small in natural units (i.e., in units of
812: $E_Fk_F^2$), some five times smaller than the value for
813: gold \cite{Tyson77}, thus favoring the possibility of large
814: deformations \cite{Urban06}.
815: 
816: 
817: 
818: \section{Summary and Conclusions}\label{sec:Summary}
819: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
820: 
821: 
822: We report experimental and theoretical evidence that the stability
823: of aluminum nanowires is governed by shell filling effects. The
824: experimental investigation is done by recording conductance
825: histograms at room temperature in UHV. Two ``magic'' series of
826: stable structures are observed and the corresponding periodicities
827: of the histogram peaks are very similar to the ones obtained
828: recently for noble and alkali metal nanowires. Therefore we
829: suggest, that also for Al the exceptionally stable structures are
830: related to electronic and atomic shell effects. Concerning the
831: electronic shell effect, the NFEM can explain the conductance and
832: geometry (cylindrical or quadrupolar) of the stable structures. On
833: the other hand, the atomic shell effect peaks can be understood
834: considering close-packed hexagonal structures. In addition, new
835: stable structures are found, that sometimes appear isolated in the
836: conductance histogram. Following our stability analysis, we can
837: attribute them to superdeformed nanowires. Remarkably, some of the
838: peaks having the period expected for electronic shell effect
839: survive under ambient conditions.
840: 
841: 
842: % \section*{Acknowledgments}
843: \ack We would like to thank Gijs van Dorp for assistance with the
844: measurements. This work is part of the research program of the
845: "Stichting FOM" (A.I.M. and J.M.v.R) and it was supported by the
846: EU Training Network DIENOW (A.I.M., D.F.U., H.G. and J.M.v.R), the
847: DFG (D.F.U. and H.G.) and by NSF Grant Nos.\ 0312028 (J.B.\ and
848: C.A.S.) and 0351964 (J.B.).
849: 
850: \section*{References}
851: \bibliographystyle{unsrt}
852: \bibliography{Al}
853: % \References
854: 
855: % \bibliographystyle{apsrev}
856: 
857: \end{document}
858: