1: \documentstyle[multicol,aps,prl,epsf]{revtex}
2:
3: \pagestyle{empty}
4: \begin{document}
5:
6: \draft
7:
8: \title{Effect of bonding of a CO molecule on the conductance of atomic metal wires}
9:
10: \author{M. Kiguchi, D. Djukic, and J.M. van Ruitenbeek}
11:
12: \address{Kamerlingh Onnes Laboratorium, Universiteit Leiden, Postbus 9504, NL - 2300 RA Leiden, The Netherlands}
13:
14:
15: \date{\today}
16:
17: \maketitle
18:
19: \begin{abstract}
20:
21: We have measured the effect of bonding of a CO molecule on the conductance of Au, Cu, Pt, and Ni atomic contacts
22: at 4.2 K. When CO gas is admitted to the metal nano contacts, a conductance feature appears in the conductance
23: histogram near 0.5 of the quantum unit of conductance, for all metals. For Au, the intensity of this fractional
24: conductance feature can be tuned with the bias voltage, and it disappears at high bias voltage (above $\sim$ 200 mV). The
25: bonding of CO to Au appears to be weakest, and associated with monotomic Au wire formation.
26:
27: \end{abstract}
28:
29: \medskip
30:
31:
32:
33:
34: \begin{multicols}{2}
35: \narrowtext
36:
37: \section{INTRODUCTION}
38: \label{sec1}
39:
40: Fabrication and characterization of nano structures have attracted wide attention because of its potential
41: application in atomic scale electronic devices. The development of fabrication methods such as mechanically
42: controllable break junctions (MCBJ) and a scanning tunnelling microscopy, has made it possible to fabricate
43: even monatomic metal wires \cite{1,2}. The properties of the nano wires have been studied in detail by various
44: methods \cite{3}. On the basis of the understanding of the metal nano wire, the influence of molecules binding to
45: the metal nano contacts has been discussed in recent years \cite{4,5,6,7,8,9,10,11,12}. Li et al. reported a change in the
46: conductance of metal nano wires due to adsorbates. The change was attributed to adsorbate scattering and
47: rearrangement in the atomic configuration of the nano wire induced by adsorption \cite{4}. Under specific
48: circumstances a single molecule can be made to form a bridge between metal electrodes just after breaking
49: the nano contact. The first conductance measurement of a single molecule was performed by Reed et al. who
50: investigated 1,4 benzendithiol \cite{5}. Conductance measurements of single molecules have been studied by
51: several experimental techniques, including MCBJ and electromigration break junctions \cite{9}, and various
52: interesting phenomena have been observed \cite{6,9}. Although these results are promising there are large
53: discrepancies between experimental results obtained by different groups, and between experiments and
54: theory. One of the reasons for these discrepancies is uncertainty regarding the configuration of the molecule
55: and the nature of the molecule-metal interface. The molecules selected for these studies are usually
56: composed of several carbo-hydride rings and are anchored to gold metal leads by sulphur groups. In view of
57: the difficulties, it seems natural to step back and focus on even simpler system, such as H$_{2}$ and CO.
58:
59: For H$_{2}$ on Pt nano contacts detailed studies have been performed by conductance measurement and
60: point contact spectroscopy. It was found that the conductance of a single H$_{2}$ molecule anchored between Pt
61: leads is slightly below 1 $G_{0}$, where $G_{0}$ = $2e^{2}/h$ is the quantum conductance unit. The configuration of the H$_{2}$
62: molecule in the junction and the number of conductance channels were determined by combined
63: experimental and theoretical studies \cite{7,8}. The effect of H$_{2}$ on metal nano contacts was also studied for Au,
64: Fe, Co, Ni, and Pd \cite{10,11,12}. Going beyond the simplest molecule, H$_{2}$, we concentrate here on CO for the
65: follows reasons. First, CO is one of the simplest molecules for which atomic and electronic structures for
66: bonding to metal surfaces are well determined. CO on the metal surfaces can be studied by conventional
67: surface analysis methods, including Auger electron spectroscopy (AES) \cite{13} and low energy electron
68: diffraction (LEED) \cite{14}, while H$_{2}$ is hard to detect by these methods \cite{15}. We can discuss the results for the
69: metal nano contacts based on the understanding obtained for flat metal surfaces. Second, both electron
70: donation and back donation from the CO molecule occur at the same time, which are important charge
71: transfer mechanisms for the understanding of chemical interactions between a molecule and a metal. In case
72: of CO, the 5$\sigma$ orbital (which is the highest occupied molecular orbital: HOMO) and 2$\pi^{*}$ orbital (lowest
73: unoccupied molecular orbital: LUMO) stay near the Fermi level of the metal, involving electron donation
74: from the 5$\sigma$ orbital to the metal, and back donation from the metal d band to the 2$\pi^{*}$ orbital. Third, CO is a
75: poisonous gas. If the conductance of the metal nano contact is changed in a specific way by CO, it may be
76: applicable as CO gas sensors.
77:
78: The effect of CO on the Pt nano contacts is discussed in more detail elsewhere \cite{11,16}. The CO single
79: molecule anchored between Pt leads was characterized by conductance measurement and point contact
80: spectroscopy. In the present study, we report on the effect of CO on Ni, Cu, Pt, and Au nano contacts. While
81: Au and Cu are noble metals, Ni and Pt are transition metals, for which the Fermi level lays inside the metal d
82: band and back donation from the metal to the CO 2$\pi^{*}$ orbital should play an important role. Reflecting the
83: change in the chemical interaction the adsorption energy of CO on Ni, Pt, Cu, and Au flat surfaces decreases
84: in that order \cite{15}. We discuss the relationship between this adsorption character on macroscopic metal
85: surfaces and the conductance behaviour of metal nano contacts. For Au nano contacts we present a more
86: detailed analysis of the conductance histograms and conductance traces.
87:
88:
89: \section{EXPERIMENTAL}
90: \label{sec2}
91:
92: The measurements have been performed using the mechanically controllable break junction technique
93: \cite{3}. A small notch was cut at the middle of Ni, Cu, Pt, and Au wires in order to fix the breaking point. The
94: wires used were 0.1 mm in diameter, about 1 cm long. The wire was glued on top of a bending beam and
95: mounted in a three-point bending configuration inside a vacuum chamber. Once under vacuum and cooled to
96: 4.2 K the wire was broken by mechanical bending of the substrate. Clean fracture surfaces are exposed by
97: breaking and these remain clean for days in the cryogenic vacuum. The bending can be relaxed to form
98: atomic-sized contacts between the wire ends using a piezo element for fine adjustment. CO was admitted via
99: a home made capillary equipped with a heater wire running along the capillary that prevents premature
100: condensation of the CO gas. After admitting a small amount of CO gas in the sample chamber we waited
101: some time for the gas to diffuse to the end of the insert. About 3000 digitized conductance traces were used
102: to build each conductance histogram in the present study.
103:
104:
105: \section{RESULTS}
106: \label{sec3}
107:
108:
109: Figure~\ref{fig1} shows the conductance histograms for Ni, Cu, Pt, and Au nano contacts before and after
110: admitting CO. The conductance histograms are observed to change by admitting CO, most dramatically for
111: Pt and Ni. For Pt, two new peaks appear in the conductance histogram, one near 0.5 $G_{0}$ and one near 1 $G_{0}$, in
112: agreement with our previous studies \cite{11,16}. A peak appears in the conductance histogram near 0.5 $G_{0}$ for Ni.
113: In both cases the peak for the clean metal that represents a single metal atom contact (at 1.6 $G_{0}$ for Pt and 1.4
114: $G_{0}$ for Ni) is suppressed after admission of CO. Weak features near 0.5 $G_{0}$ also appear in the conductance
115: histograms for Au and Cu, although the change is much smaller than for Pt and Ni.
116:
117: For Pt, the 0.5 $G_{0}$ feature is likely due to the conductance of a single CO molecule anchored between Pt
118: electrodes \cite{16,17}. For Ni, Cu, and Au, it is not clear whether the new feature near 0.5 $G_{0}$ is due to the
119: conductance of a single CO molecule, or whether it is due to the conductance of the metal nano wire whose
120: surface is covered by CO. In both cases the conductance of the contact should vary with the type of metal
121: because the electronic structure of the CO-metal bond depends strongly on the metal. Therefore, it is
122: interesting to observe that a conductance feature near 0.5 $G_{0}$ is found for all metals.
123:
124: The importance of the changes in the conductance histograms for Ni, Pt, Cu, and Au decreases in that
125: order. This order agrees with the adsorption properties of CO molecules on metal surfaces \cite{15}. On transition
126: metal (Ni, Pt) surfaces, the electron donation from the 5$\sigma$ orbital to the metal and the back donation from the
127: metal to the 2$\pi^{*}$ orbital occur simultaneously, and a strong chemical bond is formed between the CO
128: molecule and the metals. The present results show that the adsorption properties on macroscopic surfaces
129: reflect the adsorption properties on atomic-scale metal nano structures. On the other hand, there is a
130: difference in the adsorption properties between the macroscopic surface and the atomic scale nano structure.
131: While a CO molecule does not adsorb chemically on macroscopic Au surfaces, CO molecules appear to
132: coordinate to Au nano contacts judging from the change in the conductance histogram (Fig.~\ref{fig1}(a)). Indeed,
133: novel catalytic properties have been reported for Au nano particles \cite{18}. Theoretical calculations support the
134: idea that the Au nano contact is chemically active, with strong chemisorption of O$_{2}$ and CO \cite{19}.
135:
136: For Au nano contacts we observed systematic changes in the conductance behaviour and reproducible
137: conductance traces, characteristic of Au nano contacts in the presence of CO. For this reason we concentrate
138: on Au nano contacts in the following. The conductance histogram changes with bias voltage, and the change
139: is continuous and reversible (Fig.~\ref{fig2}). The fractional conductance feature is clearly observed at low bias
140: voltage, and its intensity decreases with bias voltage. Figure~\ref{fig3} shows the intensity of the fractional feature
141: normalized to the intensity of the peak at 1 $G_{0}$, plotted as a function of the bias voltage. The fractional
142: conductance feature gradually decreases and disappears at a bias voltage of about 200 mV, indicating that the
143: structure having a fractional conductance is unstable at high bias voltage, suggesting that the interaction
144: between the CO molecule and the Au nano contact is weak. It is noteworthy that the fractional conductance
145: feature appears at low bias voltage, even after it first disappeared at high bias voltage. This indicates that the
146: CO molecule is not blown off the surface but remains bound to one of the two electrodes at high bias
147: voltage.
148:
149: In order to find evidence as to how the CO molecule is adsorbed on, or incorporated into the Au nano
150: wire, we have analysed the conductance traces. Figure~\ref{fig4} shows typical breaking and return traces for Au after
151: admitting CO. The stretching length was calibrated with the length histogram of the last conductance plateau
152: for clean Au \cite{20}. There are three characteristics in the conductance traces. The last plateau, by which we
153: mean the part having a conductance near 1 $G_{0}$, extends to quite long length; the conductance drops abruptly
154: just before breaking to about 0.5 $G_{0}$; and the conductance smoothly increases with further stretching on this
155: fractional conductance plateau. These characteristics are observed reproducibly and they are typical for Au
156: nano contact in the presence of CO. In the following these characters are discussed.
157:
158: First, we discuss the length of the Au nano contact. Figure~\ref{fig5} shows the length histogram of the last
159: conductance plateau for Au after admitting CO. The length of the last plateau is taken here as the distance
160: between the points at which the conductance drops below 1.2 $G_{0}$ and 0.2 $G_{0}$, respectively. A conductance at 1
161: $G_{0}$ corresponds to a clean Au monatomic contact or chain of Au atoms. The 1$G_{0}$-plateau was observed to
162: stretch up to 1 nm in length. Figure~\ref{fig6} shows the average of the return distance as a function of the length of
163: the last plateau. Apart from an offset of 0.5nm due to the elastic response of the banks \cite{2}, the relation is
164: approximately 1:1, suggesting that a fragile structure is formed with a length corresponding to that of the last
165: plateau, which is unable to support itself when it breaks and collapses onto the banks on either side. The long
166: plateaux of over 1 nm length and the 1:1 relation were also observed for clean Au monatomic wires \cite{2},
167: supporting the formation of the monatomic wire in the present case. While clean Au forms a monatomic wire
168: under UHV at low temperature the results show that Au monatomic wires also form in the presence of CO.
169:
170: Second, we discuss the change in conductance during the breaking process. The conductance drops
171: abruptly down to about 0.5 $G_{0}$ just before breaking the contact, which produces a fractional feature in the
172: conductance histogram. This conductance drop should be caused by the adsorption of CO on the Au nano
173: wire or incorporation of CO into the Au wire because it is not observed for clean Au. It should be noticed
174: that the wire breaks once the conductance has dropped to about 0.5 $G_{0}$. For Pt in the presence of H$_{2}$ and Ag
175: in the presence of O$_{2}$, the conductance decreases stepwise after the conductance drops from 1 $G_{0}$, and the
176: contact can be stretched further \cite{21,22}. This indicates that CO adsorption or incorporation destabilized the
177: Au nano wire.
178:
179: The destabilization mechanism of the Au nano wire by CO we will now discuss in terms of the charge
180: transfer between the CO molecule and Au, and in relation to the stabilization mechanism of the Au
181: monatomic wire. The Au d band is located below the Fermi level, and thus, the electron back donation from
182: the Au d band to 2$\pi^{*}$ orbital of CO does not play an important role. The donation from the 5$\sigma$ orbital of CO
183: to the metal is dominant, and the electron would be transferred from CO to Au. The charge transfer between
184: CO and Au was studied for Au nano particles deposited on TiO$_{2}$, which showed novel catalytic properties
185: \cite{23}. Adsorption of CO on Au particles induces charge transfer from Au to TiO$_{2}$, suggesting that the electron
186: is transferred from CO to Au. A monatomic wire is formed only for 5d metal as a result of the following
187: mechanism. For 5 d metals, the s occupation increases at the expense of the d electrons due to relativistic
188: effects in electronic structure. Since the top of the d band consists of states with anti-bonding character that
189: are now partially depleted, the d bond becomes stronger. While the d electrons tend to compress the lattice,
190: the s electrons exert an opposing Fermi pressure. At the surface, the spill out of the s electron cloud into the
191: vacuum relieves some of the s electron pressure, and allows a strengthening of the bonds at the surface. The
192: monatomic wire geometry allows for an even larger s pressure relief. Therefore, the bond in the wire is much
193: stronger than the bond in the bulk, and monatomic wires are formed for 5d metals, Ir, Pt and Au \cite{3,24}.
194: When CO adsorbs on the surface of the Au monatomic wire, spill out of the s electron cloud into the vacuum
195: will be inhibited. Furthermore, the density of states (DOS) near the Fermi level with anti-bonding character
196: will increase due to the charge transfer from CO to Au. These mechanisms will decrease the bond strength of
197: the Au in the wire, and the Au monatomic wire will be destabilized by CO adsorption or incorporation.
198:
199: Third, we discuss the increase in conductance with the stretch length after the conductance drops down
200: to near 0.5 $G_{0}$. A similar increase in the conductance with elongation is observed for Al nano contacts \cite{25},
201: and for Au nano contacts in the presence of H$_{2}$ \cite{26}. The former is explained by the change in DOS at Fermi
202: level. The elongation of the Al contact induces a narrowing the s$_{pz}$ band, leading to an increase in the DOS at
203: the Fermi level, and a concomitant increase in the conductance. The latter (the effect for Au+H$_{2}$) was
204: explained as due to a gradual change in the molecular orientation. Barnett et al. used density functional
205: theory (DFT) to calculate the conductance of the various atomic configurations, including a system having a
206: H$_{2}$ molecule incorporated into a Au monatomic wire. The conductance of the perpendicular configuration,
207: for which the molecular axis is perpendicular to the contact axis, was calculated to be about 0.1 $G_{0}$, and the
208: conductance of the parallel configuration increased up to 0.25 $G_{0}$ \cite{27}. Based on these calculation results the
209: increase in conductance was explained by the change in configuration of hydrogen molecule. The H$_{2}$
210: molecule incorporates into the Au wire in the perpendicular configuration and the molecule turns toward an
211: orientation parallel to the contact axis, which shows up as an increasing conductance as the electrodes are
212: pulled apart. The present results may be explained by both a change in the DOS or a changing orientation of
213: the CO molecule. Recent DFT calculation results for Pt nano contacts in the presence of CO show a slight
214: increase in the conductance with the elongation \cite{17}, and this mechanism may also apply to Au in the
215: presence of CO.
216:
217: \section{CONCLUSIONS}
218: \label{sec4}
219:
220: We have studied the conductance of Ni, Cu, Pt, and Au nano contacts in the presence of CO. In all cases
221: a fractional conductance feature appears in the conductance histogram near 0.5 $G_{0}$ after admitting CO. The
222: change in the conductance histogram for Ni, Pt, Cu, and Au decreases in that order, in agreement with the
223: adsorption properties on macroscopic metal surfaces. For Au, the intensity of the fractional conductance
224: feature continuously and reversibly changes with the bias voltage. The feature disappears at a bias voltage of
225: 200 mV, indicating that the CO-Au bond is weak. The results give evidence that Au atomic chain formation
226: persists in the presence of CO, but bonding of a CO molecule weakens the chain and gives rise to a drop of
227: the conductance to about 0.5 $G_{0}$.
228:
229: \acknowledgments{
230: This work is part of the research program of the Stichting voor Fundamenteel Onderzoek der Materie (FOM),
231: which is financially supported by NOW..MK has been supported by the Yamada Science Foundation.}
232:
233: \begin{references}
234: \bibitem{1} Ohnishi H, Kondo Y and Takayanagi K 1998 Nature {\bf 395} 780
235: \bibitem{2} Yanson A I, Bollinger G R, van den Brom H E, Agrait N and van Ruitenbeek J M 1998 Nature {\bf 395} 783
236: \bibitem{3} Agrait N, Levy Yeyati A and van Ruitenbeek J M 2003 Phys. Rep. {\bf 377} 81
237: \bibitem{4} Li C Z, He H X, Bogozi A, Bunch J S and Tao N J 2000 Appl. Phys. Lett. {\bf 76} 1333
238: \bibitem{5} Reed M A, Zhou C, Muller C J, Burgin T P and Tour J M 1997 Science {\bf 278} 252
239: \bibitem{6} For example, van Ruitenbeek J M, Scheer E and Weber H 2005 Introducing Molecular Electronics ed G
240: Cuniberti, G Fagas and K Richter (Heidelberg: Springer) p253
241: \bibitem{7} Smit R H M, Noat Y, Untiedt C, Lang N D, van Hemert M C and van Ruitenbeek J M 2002 Nature {\bf 419}
242: 906
243: \bibitem{8} Djukic D, Thygesen K S, Untiedt C, Smit R H M, Jacobsen K W and van Ruitenbeek J M 2005 Phys.
244: Rev. B {\bf 71} 161402; Djukic D and van Ruitenbeek J M 2006 Nano Lett. {\bf 6} 789
245: \bibitem{9} Park, J.; Pasupathy, A. N.; Goldsmith, J. I.; Chang, C.; Yaish, Y.; Petta, J. R.; Rinkoski, M.; Sethna, J.
246: P.; Abruna, H. D.; McEuen, D. Nature 2002, {\bf 417}, 722?725; Liang, W.; Shores, M.P.; Bockrath, M.;
247: Long, J. R.; Park, H. Nature 2002, {\bf 417}, 725?729.
248: \bibitem{10} Csonka Sz, Halbritter A, Mihaly G, Jurdik E, Shklyarevskii O I, Speller S and van Kempen H 2003 Phys.
249: Rev. Lett. {\bf 90} 116803
250: \bibitem{11} Untiedt C, Dekker D M T, Djukic D and van Ruitenbeek J M 2004 Phys. Rev. B {\bf 69} 081401
251: \bibitem{12} Csonka Sz, Halbritter A, Milhaly G, Shklyarevskii O I, Speller S, and van Kempen H 2004 Phys. Rev.
252: Lett. {\bf 93} 16802
253: \bibitem{13} Carlson T A 1975 Photoelectron and Auger spectroscopy (New York: Plenum)
254: \bibitem{14} van Hove M A and Tong S Y 1979 Surface Crystallography by LEED (Heidelberg: Springer)
255: \bibitem{15} Zangwill A 1988 Physics at Surfaces (Cambridge: Cambridge University Press)
256: \bibitem{16} Djukic D et al. in preparation.
257: \bibitem{17} Strange M, Thygesen K S and Jacobsen K W 2006 Phys. Rev. B {\bf 73} 125424
258: \bibitem{18} Haruta M 2003 Chem. Record {\bf 3} 75
259: \bibitem{19} Bahn S R, Lopez N, Norskov J K and Jacobsen K W 2002 Phys. Rev. B {\bf 66} 81405
260: \bibitem{20}. The conductance traces are obtained as a function of piezo voltage. It is reported in \cite{2} that the length
261: histogram of the last conductance plateau for clean Au, which is defined as having a conductance
262: between 1.3 $G_{0}$ and 0.7 $G_{0}$, shows a number of equidistant maxima, at multiples of 0.257 nm \cite{3}. We use
263: this to determine the ratio between the relative change in piezo voltage with respect to the stretch length,
264: assuming that the inter-peak distance is 0.26 nm in the length histogram for clean Au. The stretch length
265: was calibrated in this way with clean Au before each measurement of the Au in the presence of CO.
266: \bibitem{21} Thijssen W H A, Marjenburgh D, Bremmer R H, and van Ruitenbeek J M 2006 Phys. Rev. Lett. {\bf 96}
267: 026806
268: \bibitem{22} Kiguchi M et al. unpublished results
269: \bibitem{23} Toyoshima I and Somorjai G A 1979 Cat. Rev. Sci. Eng. {\bf 19} 105
270: \bibitem{24} Smit R H M, Untiedt C, Yanson A I and van Ruitenbeek J M 2001 Phys. Rev. Lett. {\bf 87} 266102
271: \bibitem{25} Cuevas J C, Levy Yeyati A, Martin-Rodero A, Rubio Bollinger G, Untiedt C and Agrait N 1998 Phys.
272: Rev. Lett. {\bf 81} 2990
273: \bibitem{26} Csonka Sz, Halbritter A and Mihaly G 2006 Phys. Rev. B {\bf 73} 075405
274: \bibitem{27} Barnett R N, Hakkinen H, Scherbakov A G and Landman U 2004 Nano Lett. {\bf 4} 1845
275:
276:
277: \end{references}
278:
279: \begin{figure}
280: \begin{center}
281: \leavevmode\epsfxsize=65mm \epsfbox{fig1.eps}
282: \caption{
283: Conductance histograms for (a) Au, (b) Cu, (c) Pt and (d) Ni atomic-size contacts before (dotted-line)
284: and after (thick line) admitting CO. The bias voltage is (a) 0.03 V, (b) 0.05 V, (c) 0.15 V and (d) 0.2 V.}
285: \label{fig1}
286: \end{center}
287: \end{figure}
288:
289: \begin{figure}
290: \begin{center}
291: \leavevmode\epsfxsize=50mm \epsfbox{fig2.eps}
292: \caption{
293: Bias voltage dependence of conductance histograms for Au atomic-size contacts after admitting CO
294: for various values of the bias voltage.}
295: \label{fig2}
296: \end{center}
297: \end{figure}
298:
299: \begin{figure}
300: \begin{center}
301: \leavevmode\epsfxsize=50mm \epsfbox{fig3.eps}
302: \caption{Intensity of the fractional feature in the histograms normalized to the intensity of the peak at 1 $G_{0}$,
303: plotted as a function of bias voltage for Au atomic-size contacts after admitting CO.}
304: \label{fig3}
305: \end{center}
306: \end{figure}
307:
308: \begin{figure}
309: \begin{center}
310: \leavevmode\epsfxsize=50mm \epsfbox{fig4.eps}
311: \caption{Tycal breaking and return traces for Au after admitting CO recorded at a bias voltage of 100 mV.
312: The length of the last plateau is defined as the distance between the points at which the conductance drops
313: below 1.2 $G_{0}$ and 0.2 $G_{0}$, respectively.}
314: \label{fig4}
315: \end{center}
316: \end{figure}
317:
318: \begin{figure}
319: \begin{center}
320: \leavevmode\epsfxsize=50mm \epsfbox{fig5.eps}
321: \caption{The length histogram of the last conductance plateau as defined in Fig.~\ref{fig4}. The histogram is obtained
322: from 3000 experiments recorded at a bias voltage of 100 mV.}
323: \label{fig5}
324: \end{center}
325: \end{figure}
326:
327: \begin{figure}
328: \begin{center}
329: \leavevmode\epsfxsize=50mm \epsfbox{fig6.eps}
330: \caption{The average of the return distance as a function of the length of the last plateau as defined in Fig.~\ref{fig4}
331: recorded at a bias voltage of 100 mV. }
332: \label{fig6}
333: \end{center}
334: \end{figure}
335:
336: \end{multicols}
337: \end{document}
338:
339: