Tip-State Control of Rates and Branching Ratios in Atomic

NANO LETTERS

Tip-State Control of Rates and Branching Ratios in Atomic Manipulation

2005 Vol. 5, No. 5 835-839

Peter A. Sloan*,† and Richard E. Palmer Nanoscale Physics Research Laboratory, School of Physics and Astronomy, The UniVersity of Birmingham, Birmingham, B15 2TT, UK Received January 23, 2005; Revised Manuscript Received March 19, 2005

ABSTRACT We report the atomic manipulation properties of two distinct, stable, and reproducible states of a scanning tunneling microscope tip applied to chlorobenzene/Si(111)−(7×7). We show that the tip state influences the rates of (current-driven) molecular desorption and C−Cl dissociation as well as the branching ratio between these processes, but does not change the mediating electronic channel or the required number of electrons. These manipulation properties combined with the imaging properties of the two tip-states suggest the major difference between tip-states is their coupling efficiency to the π-states of the chlorobenzene molecule.

Single atom and molecule manipulation with the tip of a scanning tunneling microscope (STM) has become increasingly sophisticated over the past decade,1 fuelled partly by the interest in single-molecule devices.2,3 Crucial to both STM and molecular electronics is the interaction of the electrical contact with the molecule. In single-molecule electronics, it is already recognized that the electrical contact to the molecule4 cannot simply be regarded as a structureless point contact.5,6 During mechanical manipulation7-9 and electric field induced manipulation10,11 with the STM, the main role of the contact (the tip) is clear. In current-induced manipulation,12,13 however, the tip is usually regarded, in the absence of better information, as a point source of electrons,14,15 yet most studies report results varying from tip to tip.16,17 Here we demonstrate the different imaging and atomic manipulation properties of two distinct, stable and reproducible states of a tungsten STM tip, i.e., two different electrical contacts. We show, for the chlorobenzene/Si(111)-(7×7) system, that the state of the tip controls not only the absolute rates of desorption and dissociation but also the branching ratio between these two manipulation channels. Figures 1A and 1B show two different STM images of a Si(111)-(7×7) surface partly covered with chlorobenzene molecules, obtained at a surface bias voltage of +2 V in each case but with two different tip states; the chlorobenzene molecules image as (Figure 1A) dark “missing-adatom” features (as previously reported18) and (Figure 1B) as bright features; we term the two different tip states responsible for these two images as “dark-tip” and “bright-tip”.19 The image processing and analysis was carried out using Image SXM.20 * Corresponding author: E-mail: [email protected]. † Present address: Department of Chemistry, University of Toronto, 80 St. George Street, Toronto, Ontario, M5S 3H6, Canada. 10.1021/nl050142x CCC: $30.25 Published on Web 04/13/2005

© 2005 American Chemical Society

The schematic diagrams of Figure 1C show the two possible chlorobenzene bonding sites in the Si(111)-(7×7) unit cell (corner adatom and center adatom sites, 2,5 di-σ bonding in both cases); panels (a) and (c) were obtained with a dark-tip and (b) and (d) with a bright-tip. The height of the chlorobenzene feature (at the bonding silicon restatom site) relative to an unreacted and unperturbed neighboring silicon adatom is -0.40 ( 0.05 Å for the dark-tip and +0.04 ( 0.02 Å for the bright-tip (all errors and error bars are ( one standard deviation from the mean). The particular atomic orbital at the tip apex (at the Fermi level energy) is crucial in determining the degree of atomic resolution (surface corrugation) in the STM image.21,22 Generally, the more symmetric the tip state (s > p > d) the smaller the measured surface corrugation. For example, an STM image produced by modeling with a low symmetry tip (pz or dZ2) reproduces the Si(111)-(7×7) surface corrugation, but modeling with a spherically symmetric tip apex (an s-state tip) does not.23 The same is true for the Al(100) surface.22 The dark-tip image (Figure 1A) clearly shows a larger corrugation of the unreacted Si(111)-(7×7) than the bright-tip image (Figure 1B). This difference is more striking in the line profiles of Figure 1D. The imaged depth of a corner hole (marked X in Figures 1A, B, and D) gives a measure of the surface corrugation; it is nearly twice as deep in Figure 1A (dark-tip, 1.72 ( 0.03 Å, similar to ref 23) than in Figure 1B (bright-tip, 0.90 ( 0.02 Å). The brighttip, therefore, has the higher symmetry tip state. Voltage pulses (∼20 ms, -4 V, feedback loop off) applied to the tip while scanning the chlorobenzene/Si(111)-(7×7) surface initially generate the bright-tip or dark-tip state and induce switching between the states. Switching between

Figure 1. Constant current STM images of a Si(111)-(7×7) surface partly covered by chlorobenzene, obtained with two tip states, where chlorobenzene images as (A) a dark feature and (B) a bright feature (A and B are images of different areas). Images: 100×100 Å; V (sample) ) +2 V; I ) 50 pA; images filtered to remove high-frequency noise. (C) STM images (+2 V and 50 pA) showing the two possible types of bonding site within the unit cell for the chlorobenzene/Si(111)-(7×7) system (as depicted in the schematic cartoon) for (a and c) a dark-tip and (c and d) a bright-tip. (D) Cross-sections through an uncovered unit cell from (A) and (B) as marked. Also marked in panels A, B, and D are the location of the corner hole (X). (E) STM image (150×150 Å, +2 V, 50 pA, scanning bottom to top) containing a spontaneous tip change. (F) 3D view of the effect of the tip change in E (z scale has been magnified by a factor 10).

states can also occur spontaneously during an image scan as shown in Figure 1E (scanning bottom to top, tip change from dark-tip to bright-tip). From the 3D picture, Figure 1F, the tip clearly retreats away from the surface upon this state switch (on average 0.4 ( 0.1 Å, with an absolute tip to surface height of ∼ 7 Å24). Switching the other way results in the tip approaching the surface by a similar distance. Either the tip apex atoms rearrange themselves, or the tip picks up or deposits an atom (or molecule).25 While the change in tip height might suggest the pick up (dark-tip to bright-tip) or deposition (bright-tip to dark-tip) of an adsorbate, we observed no change of the surface composition after a spontaneous switch and thus rule out this possibility. Instead, we conclude that rearrangement of the tip apex itself must be responsible for the switching. Once produced, either spontaneously or by pulsing, both tip states were stable for typically 30 min, enough time to perform several manipulation experiments. For the chlorobenzene/Si(111)-(7×7) system, the STM can induce both molecular desorption and C-Cl bond dissociation.26-28 Figure 2 reveals how both molecular desorption (A and B) and C-Cl dissociation (C and D) is characterized by STM imaging. Figure 2A shows a surface imaged with a bright-tip at +2 V with many chlorobenzene adsorbate molecules clearly present. The tunneling parameters chosen allowed identification of the adsorbed molecules without disturbing them. The same area was then scanned with the desired manipulation parameters and then rescanned, 836

Figure 2B, with the passive set of imaging parameters. Figure 2B shows the surface with fewer chlorobenzene molecules than it had prior (Figure 2A) to the manipulation scan (not shown). The white circles in Figure 2A mark locations of two chlorobenzene molecules that were desorbed. Thus, the manipulation scan induced molecular desorption. Dissociation is characterized in a similar fashion. Figure 2C shows a surface, this time imaged with a dark-tip, before a manipulation scan and in Figure 2D the same area after a manipulation scan. Here, the evidence for dissociation is the creation of a chemisorbed chlorine atom (a bright feature at +2 V) on the surface.29 Two such chlorine atoms are marked by white squares in Figure 2D. Scanning before and after the manipulation scan also allows the state of the tip to be determined and to ensure the same tip state was used throughout a manipulation experiment. Comparison of pairs of these images, typically 350×350 Å, allowed the rates of the processes to be determined.30 Tunneling current population of the π-states of the chemisorbed molecule drives both desorption27,30 and dissociation31 and results in a sharp bias voltage onset for manipulation near +3 V.27,32 Figure 2E shows the normalized rates of dark-tip desorption and dissociation and bright-tip desorption as a function of voltage. Figure 2F shows, for both tip states, the absolute rates of desorption. The +3.0 V onset for both dark-tip and bright-tip desorption and for darktip dissociation,33 shown in Figure 2E, indicates that the π-states of the molecule mediate all these manipulation Nano Lett., Vol. 5, No. 5, 2005

Figure 2. (A, B, C, D) STM images of a Si(111)-(7×7) surface partly covered by chlorobenzene: (A) and (B) (both +2 V, 100 pA, 100×100 Å) taken with a bright-tip (A) before and (B) after a scan at -1.5 V (50 pA) which induced desorption. The white circles in (A) mark the locations of two chlorobenzene molecules that desorbed. (C) and (D) (both +2 V, 50 pA, 100×100 Å) taken with a dark-tip (C) before and (D) after a scan at +3.5 V (500 pA) that induced both dissociation and desorption. The white squares in (D) mark the location of two chlorine atoms generated by C-Cl bond dissociation. (E) Normalized bias voltage dependence of the rates (at 100 pA) of STM desorption induced by (2) a dark-tip and (4) a bright-tip, and also of (b) dark-tip dissociation. (F) Bias voltage dependence of the rates (logarithmic scale) of STM induced chlorobenzene desorption (at 100 pA) induced by (2) a dark-tip; each of the other symbols represents a bright-tip desorption experiment on a different day. Desorption results were taken with a variety of tunneling currents (typically < 50 pA) and scaled appropriately to 100 pA. All dissociation experiments used 100 pA tunneling current.

processes. Changing the tip-state, therefore, leaves the molecular electronic pathway unchanged. Figure 2F shows, however, that a change in the tip-state does change the absolute rate of desorption. There is more than an order of magnitude difference (a factor 50 ( 10) between bright-tip and dark-tip induced desorption, clearly evident at the higher bias voltages. The tip-state controls the probability that a tunneling electron induces desorption. Though we find that the STM tips always image the chlorobenzene/Si(111)-(7×7) system in one of two ways, i.e., they behave either as a bright-tip or as a dark-tip, there are subtle variations between the tips used on different days. This is not surprising as most STM manipulation experiments report a small degree of variation from tip to tip.7,8,10,16,17 The desorption rate data presented in Figure 2F are, in fact, the same as Figure 2E, but with the bright-tip results separated into sets of points corresponding to individual experiments, i.e., different days. The day-to-day spread of results for bright-tip desorption, especially evident at the higher voltages, is clearly smaller than the difference in the rates produced between the two different tip states. Hence, the difference in STM induced molecular manipulation between the two tip states is not attributable to minor dayto-day variations, but instead to a major change of the tip state, resulting in the large differences in the measured manipulation rates. Presented in Figure 3 are the tunneling current dependence of desorption and dissociation events for both tip states. Figure 3 also shows power law fits, Rate ∝ In, where the exponent n is equal to the number of electrons required to Nano Lett., Vol. 5, No. 5, 2005

Figure 3. Tunneling current dependence of the rate of STMinduced (2) dark-tip desorption of chlorobenzene and (b) dark-tip C-Cl dissociation (both at +3.5 V), as well as (4) bright-tip desorption and (O) bright-tip C-Cl dissociation (both at +3.0 V). The lines are least-squares fits and correspond to power laws, with rate ∝ In, where for dark-tip induced desorption n ) 0.8 ( 0.2 and dissociation n ) 1.8 ( 0.2, and for bright-tip induced desorption n ) 0.9 ( 0.1 and dissociation n ) 1.8 ( 0.3.

drive each process.34 Both the bright-tip (n ) 0.9 ( 0.1) and dark-tip (n ) 0.8 ( 0.2) rates of desorption are almost linearly dependent on the tunneling current (desorption probability per electron approximately constant). For dissociation, by contrast, the exponents for the bright-tip (n ) 1.8 ( 0.3) and dark-tip (n ) 1.8 ( 0.2) show that the tunneling current dependence is approximately quadratic in each case (dissociation is a two electron process31). Thus, the number of electrons that drive a particular process for both tip states is the same; desorption requires one electron 837

and dissociation two. The state of the tip does not influence the number of electrons required to drive a process; it is a molecular property. Figure 3 also allows us to compare the rates of desorption and dissociation for the two different tip states. In both cases the bright-tip is more efficient. There is a factor of 24 ( 2 difference between the desorption rates and a factor 43 ( 7 between the dissociation rates for the two tip states.35 What is the cause of this difference? Since the dark-tip has the lower desorption rate and is 0.4 Å closer to the adsorbate (Figures 1E and 1F), one obvious candidate is a mechanical tip/adsorbate interaction that suppresses desorption. However, changing the tunneling current in Figure 3 is accomplished by changing the tip height. From the linearity of the desorption rates in Figure 3, the probability per electron of inducing desorption is constant for tip heights differing by 1.74 ( 0.05 Å for the bright-tip27 and 0.93 ( 0.02 Å for the dark-tip. Both of these ranges are well above the measured height change upon a switch of tip-state (0.4 ( 0.1 Å), so we rule out a mechanical suppression of desorption. The same argument also rules out the electric field as the possible difference between tip states, since the voltage is constant while the tip-surface height is varying. It is well-known that the π-states of aromatic molecules control their electrical properties.3,36 Moreover, the interface between the molecule and any current carrying contact plays a crucial role in the conductivity because of the directionality of the p-orbitals of the carbon atoms in the benzene ring and hence the directionality of the π-orbitals that they form.5,6 Calculations have shown that, for s-state contacts (such as a gold atom), the crucial aspect is whether the ring is contacted at the side or on top.6 If the ring is parallel to a line bisecting two gold atom contacts (i.e., two side contacts) then due to a symmetry mismatch the π-states of the ring are unavailable to the electrons flowing from the gold s-state contact. Tilting the ring away from this bisecting line toward the perpendicular breaks the symmetry of the molecule-contact system (each contact now interacts with one face of the ring) and the conductance through the π-states rapidly increases,5,6 perhaps by 2 orders of magnitude.37 Similar considerations can be applied to the present STM experimental situation. The chlorobenzene molecule lies on the surface in a butterfly configuration;38 the ring is slightly buckled up either side of a central line. However, to first approximation, we can regard the ring as lying flat on the surface and thus perpendicular to the line bisecting the STM tip and the surface, i.e., perpendicular to the electronic contacts. In this picture an s-state tip (s-tip) should couple more efficiently to the π-states of the chlorobenzene molecule (since the tip is above the ring) than a p-state tip (p-tip), since any px and py contributions will not couple to the molecular π-states. Therefore, in constant current STM images (e.g., Figures 1A, B, and C), an s-tip will image a chlorobenzene molecule as a brighter (higher) object than a p-tip. Moreover, since both desorption and dissociation are mediated by resonant tunneling through the same π-states, the increased coupling for the s-tip should lead to an increase in the manipulation rate in this example over the p-tip. Thus 838

Figure 4. Rate of STM-induced C-Cl dissociation in chlorobenzene as a function of the rate of STM-induced chlorobenzene desorption for (b) a dark-tip and (O) a bright-tip. Data are taken from Figure 3; thus the bias voltages used were +3.5 V for the dark-tip and +3 V for the bright-tip.

a tentative assignment of the bright-tip to an s-tip state and of the dark-tip to a p-tip state would explain both the imaging characteristics (molecular brightness, surface corrugation) and manipulation characteristics (the difference in absolute rates) for the two tip states. For both tip states we now examine the rate of dissociation relative to the rate of desorption - the branching ratio. To do this Figure 4 re-plots the data of Figure 3, but shows the dissociation rate as a function of the desorption rate. As desorption is a one electron π-state mediated process, the desorption rate should be proportional to the fraction of the tunneling current resonantly tunneling though the π-states. The rate of desorption can therefore be viewed as a pseudo π-state tunneling current. Figure 4 shows that for any particular rate of desorption, the rate of dark-tip dissociation is more than an order of magnitude (a factor of 14 ( 2) larger than the corresponding rate of bright-tip dissociation. The key to understanding the change in the branching ratio may well lie in the mechanism of dissociation (presented in ref 31). In brief, we have proposed a two electron process, where the first electron forms a transient negative-ion π-state of the molecule but, because of a symmetry barrier, cannot induce C-Cl dissociation. Upon neutralization the molecule is left vibrationally excited.27,30 Some molecules subsequently desorb from the surface while others, though vibrationally excited, remain on the surface. If a second electron interacts with such a molecule before it has fully relaxed vibrationally, then the barrier to C-Cl breaking is relaxed and the C-Cl bond can break. The crucial aspect here is the need for the second electron, the one that induces dissociation, to interact with a vibrationally excited molecule, which has a perturbed π-state (see ref 31 for details). We therefore suggest that the coupling of the tip states to the vibrationally excited πstate is different from the coupling to the ground π-state. Thus, from Figure 4, the dark-tip appears to couple more efficiently to the vibrationally perturbed π-state (relative to the coupling to the ground π-state) than does the bright-tip. This work shows that the STM tip-state, when viewed as an electrical contact, is a tool that can control the rate of a chemical process as well as the branching ratio between Nano Lett., Vol. 5, No. 5, 2005

competing processes. It therefore opens up new possibilities for probing and controlling single-molecule physics and chemistry. Moreover, this control is achievable at room temperature and on a substrate of considerable technological importance. Acknowledgment. We thank the EPSRC and the European Research Training Networks “Manipulation of individual atoms and molecules” and AMMIST for financial support of this work. P.A.S. acknowledges studentship support from the School of Physics and Astronomy and EPSRC. References (1) Ho, W. J. Chem. Phys. 2002, 117, 11033. (2) Gimzewski, J. K.; Joachim, C. Science 1999, 283, 1683. (3) Joachim, C.; Gimzewski, J. K.; Aviram, A. Nature (London) 2000, 408, 541. (4) Hipps, K. W. Science 2001, 294, 536. (5) Kornilovitch, P. E.; Bratkovsky, A. M. Phys. ReV. B 2001, 64, 195413. (6) Emberly, E. G.; Kirczenow, G. Phys. ReV. Lett. 2003, 91, 188301. (7) Eigler, D. M.; Schweizer, E. K. Nature (London) 1990, 344, 524. (8) Bartels, L.; Meyer, G.; Rieder, K.-H. Phys. ReV. Lett. 1997, 79, 697. (9) Rosei, F.; Schunack, M.; Jiang, P.; Gourdon, A.; Lægsgaard, E.; Stensgaard, I.; Joachim, C.; Besenbacher, F. Science 2002, 296, 328. (10) Kobayashi, A.; Grey, F.; Williams, R. S.; Aono, M. Science 1993, 259, 1724. (11) Whitman, L. J.; Stroscio, J. A.; Dragoset, R. A.; Celotta, R. J. Science 1991, 251, 1206. (12) Salam, G. P.; Persson, M.; Palmer, R. E. Phys. ReV. B 1994, 49, 10655. (13) Ueba, H.; Persson, B. N. J. Surf. Sci. 2004, 566-568, 1. (14) Komeda, T.; Kim, Y.; Kawai, M.; Persson, B. N. J.; Ueba, H. Science 2002, 295, 2055. (15) Pascual, J. I.; Lorente, N.; Song, Z.; Conrad, H.; Rust, H.-P. Nature (London) 2003, 423, 525. (16) Quaade, U. J.; Stokbro, K.; Lin, R.; Grey, F. Nanotechnology 2001, 12, 265. (17) Soukiassian, L.; Mayne, A. J.; Carbone, M.; Dujardin, G. Phys. ReV. B 2003, 68, 035303.

Nano Lett., Vol. 5, No. 5, 2005

(18) Chen, X. H.; Kong, Q.; Polanyi, J. C.; Rogers, D.; So, S. Surf. Sci. 1995, 340, 224. (19) A DC electrochemical etch of polycrystalline tungsten wire in 2 M NaOH was used to make the STM tip, followed by in-vacuum electron bombardment to remove any residual tungsten oxide layer. (20) Barrett. S. D. Image SXM, 2004, http://www.ImageSXM.org.uk. (21) Hansma, P. K.; Tersoff, J. J. Appl. Phys. 1987, 61, R1. (22) Chen, C. J. Phys. ReV. Lett. 1990, 65, 448. (23) Demuth, J. E.; Koehler, U.; Hamers, R. J. J. Microsc. 1988, 152, 299. (24) Dujardin, G.; Mayne, A.; Robert, O.; Rose, F.; Joachim, C.; Tang, H. Phys. ReV. Lett. 1998, 80, 3085. (25) Bartels, L.; Meyer, G.; Rieder, K.-H. Appl. Phys. Lett. 1997, 71, 213. (26) Lu, P. H.; Polanyi, J. C.; Rogers, D. J. Chem. Phys. 1999, 111, 9905. (27) Sloan, P. A.; Hedouin, M. F. G.; Palmer, R. E.; Persson, M. Phys. ReV. Lett. 2003, 91, 118301. (28) Palmer, R. E.; Sloan, P. A.; Xirouchaki, C. Philos. Trans. R. Soc. London Ser. A 2004, 362, 1195. (29) Boland, J. J.; Villarrubia, J. S. Phys. ReV. B 1990, 41, 9865. (30) Alavi, S.; Rousseau, R.; Patitsas, S. N.; Lopinski, G. P.; Wolkow, R. A.; Seideman, T. Phys. ReV. Lett. 2000, 85, 5372. (31) Sloan, P. A.; Palmer, R. E. Nature (London) 2005, 434, 367. (32) Komeda, T.; Kim, Y.; Fujita, Y.; Sainoo, Y.; Kawai, M. J. Chem. Phys. 2004, 120, 5347. (33) No bright-tip dissociation was observed in these particular experiments as the low tunnelling currents used, typically < 50 pA (Figures 2E and F results are scaled to 100 pA), coupled with the low molecular coverage precluded any chance of inducing a dissociation event during the manipulation experiments. (34) Stipe, B. C.; Rezaei, M. A.; Ho, W.; Gao, S.; Persson, M.; Lundqvist, B. I. Phys. ReV. Lett. 1997, 78, 4410. (35) The bright-tip and dark-tip experiments of Figure 4 were performed with slightly different bias voltages, applying a correction factor determined from Figure 3A (3.3 ( 0.9) would lead to an even larger difference between the rates, 80 ( 20 for desorption and 140 ( 40 for dissociation. (36) Aviram, A.; Ratner, M. A. Chem. Phys. Lett. 1974, 29, 277. (37) Di Ventra, M.; Pantelides, S. T.; Lang, N. D. Phys. ReV. Lett. 2000, 84, 979. (38) Cao, Y.; Deng, J. F.; Xu, G. Q. J. Chem. Phys. 2000, 112, 4759.

NL050142X

839