Random Telegraphic Conductance Fluctuation at AuPentaceneAu

Mar 13, 2009 - junctions. The distribution of the plateau lifetime obeyed an exponential distribution, but the decay constants were independent of the...
0 downloads 0 Views 829KB Size
NANO LETTERS

Random Telegraphic Conductance Fluctuation at Au-Pentacene-Au Nanojunctions

2009 Vol. 9, No. 4 1442-1446

Yuki Kihira,† Toshihiro Shimada,*,† Yutaka Matsuo,‡ Eiichi Nakamura,†,‡ and Tetsuya Hasegawa† Department of Chemistry, The UniVersity of Tokyo, Bunkyo-ku, Tokyo 113-0033, Japan, and ERATO Nakamura Functional Carbon Cluster Project, Bunkyo-ku, Tokyo 113-0033, Japan Received October 31, 2008; Revised Manuscript Received January 31, 2009

ABSTRACT Random telegraphic noises with a preferential jump height around 0.018G0 were observed in the conductance traces of the Au-pentacene-Au junctions. The distribution of the plateau lifetime obeyed an exponential distribution, but the decay constants were independent of the height of the adjacent conductance jumps or the measurement temperature. This result, along with the buckyferrocene (Fe(C60(CH3)5)C5H5) result, suggests that the conductance fluctuation originates from the current-induced geometrical fluctuation around a single molecule.

When the miniaturization of organic semiconductor devices occurs,1 the device characteristics become more affected by the properties of the nanostructures at the metal-molecule interface. Structural fluctuation at the electrode interfaces, for example, might act as a noise source in organic devices. However, their structures and properties on a microscopic scale have not been investigated in detail. On the basis of this aspect, the metal-molecule interface is closely related to the measurement of a single molecular conductance, including switching,2 between metal electrodes. Observation of the electron transport through single molecules has been actively studied since “break junction (BJ)” techniques came into common use.3-5 Nevertheless, most of the molecules used in the BJ experiments have contact with at least a pair of electrodes,6-8 which is the major difference with the interfaces in ordinary organic semiconductor devices. An exception to the molecular conductance measurement without an anchoring moiety is performed by scanning tunneling microscopy (STM), which requires very low temperatures (e4.2 K) for the precise positioning and handling of the molecule.9 Since the shape of the tip apex cannot be easily defined in the direct contact regime, it is still necessary to statistically accumulate the data of many junctions even by STM imaging. This is probably the reason why that kind of experiment is rarely performed. At higher temperatures, the molecular motion is thermally or electronically activated during the measurements.10-13 * Corresponding author. E-mail: [email protected]. † Department of Chemistry, The University of Tokyo. ‡ ERATO Nakamura Functional Carbon Cluster Project. 10.1021/nl803284t CCC: $40.75 Published on Web 03/13/2009

 2009 American Chemical Society

Figure 1. Experimental setup.

In this study, we intentionally used the motion of the atoms/ molecules as the noise source and measured the conductance fluctuation at the Au-molecule-Au nanocontact using lowtemperature STM and found that the fluctuation shows random telegraph signals. The molecule measured in detail was pentacene, because the junction between pentacene and gold is one of the most important examples of an electrode interface in the organic field effect transistors (FET).14 The parameters related to the conductance values of a single pentacene molecule have not been measured, and they are of particular importance in nanoscaled organic electronics. We also present the result using buckyferrocene (Fe(C60(CH3)5)C5H5),15 which exhibits different fluctuation values even at room temperature, in order to show that the conductance fluctuation probably originates from the molecular nanojunction. The principle of the method is described in Figure 1. The experiments were performed in a cryogenic ultrahigh vacuum

Figure 2. (a) A random telegraphic signal at an Au-pentacene-Au nanojunction (125 K). (b) Explanation of the analysis. (c) Histogram of the conductance jump at 125 K.

scanning tunneling microscope (UHV-STM) connected to a UHV preparation chamber. The Au(111) substrate was flameannealed in air, and then it was heated to 800 °C in UHV. A 15nm thick pentacene film was thermally deposited onto the substrate in the preparation chamber at room temperature. We found that this rather thick film is necessary for stable measurements, which probably embeds the nanojunction with an ample supply of the molecules. The Au tip was prepared by a standard electrochemical process for an STM tip.16 Liquid nitrogen was pumped into the inside tank to cool the UHV-STM, and then it was pumped out after the temperature inside was well below the measurement temperature. The measurements were carried out after all the liquid nitrogen had evaporated in order to avoid any vibration from its boiling. The temperature gradually increased from 91 to 225 K during the measurements, but this was slow enough for good temperature stability during a single measurement. The temperature was measured by a CERNOX thermometer thermally connected to the sample. The bias voltage applied between the tip and substrate was maintained at 129 mV unless otherwise stated. The Au tip was placed in contact with the substrate surface to make the conductance value (G) greater than 6G0 (G0 ≡ 2e2/h). We found that a dull Au tip does not penetrate the pentacene layer to make this condition. The tip was completely taken out of contact and put in again to make G > 6G0. This process was repeated several times in order to form a nanojunction composed of a mixture of Au and pentacene molecules on the Au substrate. The tip-surface distance was then controlled to make the junction conductance (1.0-5.0)G0. Next, the feedback loop of the tip height control was cutoff and the tip-sample current was monitored. After the junction current had stabilized, the conductance was continuously measured. The conductance traces were recorded at 100 µs intervals using a 14-bit digital oscilloscope. To achieve a good balance between the high-speed performance and low noise measurement, the input range of the oscilloscope was magnified in the following manner. Nano Lett., Vol. 9, No. 4, 2009

(1) The conductance values were measured for 20 ms, and the average was calculated. This was evaluated in order to determine whether the average was between 1.0G0 and 5.0G0. If not, the tip height was adjusted using the feedback control until the conductance average satisfied the above condition. The tip feedback control was cutoff again. This occasional tip height control was used to create a tolerance to the thermal drift. (2) The DA converter was set to the voltage corresponding to the final average conductance value in (1) as a pedestal. (3) The pedestal DA converter output was subtracted from the preamplifier output voltage using a low noise precision op-amp circuit, and the resulting signal was input to the 14-bit high-speed digital oscilloscope. The digital oscilloscope recorded 2000 points at 100 µs intervals (total 200 ms), and the recorded values were transferred to a computer. (4) The pedestal DA converter output was digitally added to the oscilloscope output by the measurement software, and true conductance values were obtained. Steps 1-4 were repeated for several minutes as the measurement at one temperature. It is expected that there are time gaps and tip height differences at every 2000 points corresponding to the feedback process (1). The conductance jumps at these points were omitted by the analysis program. The accuracy and the low noise amplitude operation, both better than (5 × 10-4)G0, were confirmed by measuring a dummy fluctuating resistor circuit made of standard resistors and relays. Figure 2a shows a typical conductance trace at 125 K. The trace had the shape of random telegraphic signals. It is easily noticed that the conductance jumps of random telegraphic signals seem to have several specific values. The random telegraphic fluctuation was not observed above ∼225 K. We made a histogram of the jump heights from each trace to obtain exact values of the fluctuation (see Figures 2c and 3). The histograms were obtained using the following algorithm. The analysis program seeks the “flat” part where the conductance maintains its value within a certain conductance range for ∆ t1 g 500 µs as shown in Figure 2b. The program searches the next flat part, which also lasts 1443

Figure 3. A histogram of the conductance jump height at 225 K.

Figure 4. A lifetime-conductance jump plot corresponding to Figure 3.

Figure 5. A histogram of the lifetime plateau corresponding to Figure 3.

∆t2 g 500 µs, and examines whether the time between them (∆ t3) is less than 10 ms. If the answer is yes, the conductance difference, ∆G, is added to the histogram. The conductance peaks of ∆G at 225 K found in Figure 3 are (i) 0.008G0 and (ii) 0.016G0. To investigate the origin of these peaks, we measured the lifetime of the plateaus after the fluctuation. The lifetime of the conductance plateau, ∆L, is defined as the interval time from jump to jump as shown in Figure 2 (∆L ) ∆t2 in this case). We used the interval time after the jump occurred in this analysis, but the lifetime before the jump actually showed the same result. In Figure 4, the logarithm of the jump occurrence is plotted in a twodimensional chart. The abscissa is the conductance difference, ∆G, and the ordinate is the plateau lifetime after the jump, ∆L. The ∆G peaks of (i) and (ii) clearly appear on the twodimensional (2D) plot in Figure 4. The distribution of the lifetime at each peak is shown in Figure 5. Both lifetimes corresponding to (i) and (ii) were excellently fitted by the exponential distributions. However, the log-linear plots show different slopes, indicating that the lifetimes were not the same in the measured single nanocontact. The decay time 1444

constants k can be obtained from the slopes of the plot. It should be noted that similar plots from the measurements with liquid nitrogen remaining in the STM cryostat were not fitted by the exponential distribution. This shows that the lifetime was strongly influenced by the mechanical vibration due to the liquid nitrogen bubbles. Figure 6a shows the comparison of ∆G at various temperatures ranging from 91 to 225 K. It is noticed that the main grouping of ∆G is around 0.018G0 regardless of the temperature as shown in Figure 6b. Figure 6c is the Arrhenius plot of all the measurements, where k stands for the decay time constant of the plateau after the jump. The Arrhenius plot, however, did not seem to show any correlation. This indicates that the peaks of ∆G at various temperatures were influenced more by other factors, e.g., the geometry of the individual junctions, than by the temperature. The conductance jumps of the random telegraphic signals have some specific ∆G values; the principal values are around 0.018G0 (see Figure 6b). It must be clarified whether these ∆G values in our measurement originate from the fluctuation at the Au-pentacene-Au junction, because the random telegraphic noises were observed even at pristine Au nanojunctions.17 However, it is reported that the values of the conductance jumps at an Au nanojunction are between 0.05G0 and 0.15G0. Because the conductance jumps in the present experiment were mostly below 0.05G0, the noises must be different from those at a pristine Au nanocontact. It is established that the random telegraphic noises appear due to the atomic motion in the case of the pristine Al nanocontacts;18 therefore, it is strongly suggested that the conductance fluctuations in this experiment come from the motion of the Au atoms or the pentacene molecules in the Au-pentacene-Au contacts. The principal values around 0.018G0 seem to correspond to the making and breaking of the junction based on the geometry which frequently appears. As for the value of 0.06G0, it might be due to a pristine Au nanocontact formed in parallel with the Au-pentacene-Au junctions. In order to show that the fluctuation is indeed related to the molecular conductance, we present the result of a similar experiment on buckyferrocene (Figure 7 inset) using the Au(111) substrate and the Au tip. Figure 7 shows a typical conductance trace and corresponding histogram taken at room temperature. We found that the random telegraph signal was observed even at room temperature, and the frequently observed ∆G value (around 0.035G0) is substantially greater than the case of pentacene. These differences are reasonably explained by the fact that buckyferrocene is heavier, and its conjugated π-electron system is more extended than that of pentacene, leading to less frequent molecular motion and greater channel transmittance in the nanojunctions. From the comparison between pentacene and buckyferrocene, we can strongly suggest that the stepwise conductance fluctuation is related to the molecular conductance even though the sample molecules do not have any anchoring. The broad peak in Figure 6b suggests that there must be stable geometries of the Au-pentacene (or buckyferrocene)Au junction and that their conductance differences distribute around the above-mentioned values. Although we cannot rule Nano Lett., Vol. 9, No. 4, 2009

Figure 6. (a) Conductance jump values plotted vs the measurement temperature. (b) The histogram of all the conductance jump data values. (c) Lifetime of the plateau plotted vs the measurement temperature.

Figure 7. Random telegraphic noise and corresponding histogram of conductance jump height at an Au-buckyferrocene-Au nanojunction.

out the possibility of the modification of the conductance of fluctuating Au-Au atomic contact by the surrounding molecules, the present experiment suggests that the intrinsic conductance values of the molecules without anchor terminals can be addressed as well as that of the molecules with anchoring parts. The 2D plot used here contains much information about the molecular junction. As we can see in Figure 5, the distribution of the lifetimes of the plateaus is excellently fitted by the exponential distribution. This indicates that the conductance jumps do not have a memory or their occurrence is governed by the Markovian process. Note that the lifetime did not show an exponential distribution when the liquid nitrogen remained during the measurement. This strongly suggests that the external mechanical noise without liquid nitrogen was not the direct source of the conductance fluctuation. This hypothesis is reinforced by noting that the decay constants k corresponding to certain jump values (∆G0) differ even within the same conductance trace (see Figure 5). This observation is naturally explained by assuming that there were several individual small junctions that fluctuate with different k values in the measured nanocontact. Measurements at various temperatures illustrate that neither the conductance jump values nor the lifetime have any correlation with temperature (Figure 6a,c). We consider that this result comes from two mechanisms. One is the variation in the stability of the individual junctions due to the absence of any anchoring atoms. Because the Au atom is much smaller than the fused π electron plane of pentacene, it is easily assumed that similar conductance values are obtained from the different arrangement of Au atoms at the junction, and yet the stabilities of these arrangements are quite different. Nano Lett., Vol. 9, No. 4, 2009

The other mechanism is the local heating by the current through the junctions which is much greater than the temperature effect. We can estimate the current through a single junction which produces the conductance fluctuation. The bias voltage of 129 mV multiplied by G corresponds to a current on the order of 10-50 µA. This current is much higher than that used in typical STM imaging of adsorbed molecules by orders of magnitude and is huge enough to induce the molecular motion which can be seen in the STM scan with an excessive current. It is naturally understood that the lack of anchoring atoms contributes to the mechanical instability against the current flow. We note that the results at lower bias voltages (30-129 mV) were within the deviation in Figure 6, which is consistent with the above discussion. In conclusion, random telegraphic noises were observed in the traces of the conductance at Au-pentacene-Au and Au-buckyferrocene-Au junctions. The telegraphic noises exhibited conductance jumps with particular values around 0.018G0 for pentacene and 0.035G0 for buckyferrocene. We believe that these values resulted from the Au-molecule-Au junctions, because the conductance fluctuation range of our junction was much smaller than that of the Au nanocontact and dependent on the molecules. The lifetime of the junctions obeyed exponential distributions and had different decay constants. This indicates that the conductance fluctuation observed at the Au-molecule-Au junction was not induced by external factors but by the structural fluctuation around a single molecule. The temperature dependence did not appear in this experiment probably because the current-induced molecular motion was dominant. It is suggested that the conductance differences due to the existence of a molecule at nanojunctions can be estimated by measuring the conductance fluctuations even if the molecule does not have an anchoring moiety. Acknowledgment. This work was supported by the Casio Science and Technology Foundation and GCOE program for Chemistry Innovations through Cooperation of Science and Engineering, MEXT, Japan. Discussions with Professor M. Tsukada and Professor T. Hitosugi (Tohoku University) are gratefully acknowledged. References (1) Hamers, R. J. Nature 2001, 412, 489. (2) Haiss, W.; Zalinge, H. van; Higgins, S. J.; Bethell, D.; Hbenreich, H.; Schiffrin, D. J.; Nichols, R. J. J. Am. Chem. Soc. 2003, 125, 15294. (3) Xu, B.; Tao, N. J. Science 2003, 301, 1221. 1445

(4) Ruitenbeek, J. van; Scheer, E.; Weber, H. B. Lect. Notes Phys. 2005, 680, 253. (5) Sumit, R. H. M.; Untiedt, C.; Ruitenbeek, J. van. Nanotechnology 2004, 15, S472. (6) Venkataraman, L.; Klare, J. E.; Nuckolls, C.; Hybertsen, M. S.; Steigerwald, M. L. Nature 2006, 442, 24. (7) Danilov, A.; Kubatkin, S.; Kafanov, S.; Hedegard, P; Stuhr-Hansen, N.; Moth-Poulsen, K.; Bjornholm, T. Nano Lett. 2008, 8, 1. (8) Xiao, X.; Xu, B.; Tao, N. J. Nano Lett. 2004, 4, 2, 267. (9) Temirov, R.; Lassise, A.; Anders, F. B.; Tautz, F. S. Nanotechnology 2008, 19, 065401. (10) Galperin, M.; Ratner, M. A.; Nitzan, A. J. Phys.: Condens. Matter 2007, 19, 103201. (11) Nitzan, A.; Ratner, M. A. Science 2003, 300, 1384.

1446

(12) Huang, Z.; Chen, F.; D’agosta, R.; Bennett, P. A.; Ventra, M. D.; Tao, N. Nat. Nanotechnol. 2007, 2, 2698. (13) Hihath, J.; Arroyo, C. R.; Rubio-Bollinger, G.; Tao, N.; Agrait, N. Nano Lett. 2008, 8, 6, 1673. (14) Service, R. F. Science 2000, 287, 5452, 415. (15) Sawamura, M.; Kuninobu, Y.; Toganoh, M.; Matsuo, Y.; Yamanaka, M.; Nakamura, E. J. Am. Chem. Soc. 2002, 124, 9354. (16) Libioulle, L.; Houbion, Y.; Gilles, J. M. J. Vac. Sci. Technol., B 1995, 13, 3. (17) Huntington, M. D.; Armstrong, J. N.; Sullivan, M. R.; Hua, A. Z.; Chopra, H. D. Phys. ReV. B 2008, 78, 035442. (18) Halbritter, A.; Csonka, Sz.; Kolesnychenko, O. Yu.; Mihaly, G.; Shklyarevskii, O. I.; Kempen, H. van. Phys. ReV. B 2002, 65, 045413.

NL803284T

Nano Lett., Vol. 9, No. 4, 2009