Role of tip apex in inelastic electron tunneling spectroscopy of CO/Cu(111) with an

Transcription

1 Role of tip apex in inelastic electron tunneling spectroscopy of CO/Cu(111) with an STM/AFM Norio Okabayashi 1,2,*, Alexander Gustafsson 3, Angelo Peronio 1, Magnus Paulsson 3, Toyoko Arai 2, and Franz J. Giessibl 1 1 Institute of Experimental and Applied Physics, University of Regensburg, D Regensburg, Germany 2 Graduate School of Natural Science and Technology, Kanazawa University, Ishikawa, Japan 3 Department of Physics and Electrical engineering, Linnaeus University, Kalmar, Sweden * okabayashi@staff.kanazawa-u.ac.jp PACS numbers: Ef, Ps, b

2 Abstract By combining scanning tunneling microscopy and atomic force microscopy, we have investigated the role of the microscopic tip apex in the magnitude of inelastic electron tunneling spectroscopy (IETS) of CO single molecules on a Cu(111) surface, We found that the measured intensity strongly depends on the tip termination, in particular a tip with a single atom at its apex yields a stronger IETS signal over a tip with several atoms at its apex. This is due to a larger fraction of electrons tunneling through the CO molecule, rather than to a different efficiency of the inelastic tunneling. In addition, we experimentally observed a -5% shift in the energy of the frustrated rotational mode of a CO molecule adsorbed on a Cu adatom on Cu(111) compared to a CO molecule directly adsorbed on Cu(111). This shift is confirmed by density functional theory calculations. 1

3 Inelastic electron tunneling spectroscopy (IETS) with scanning tunneling microscopy (STM) is an effective method to analyze vibrational modes in a molecule with sub-nanometer lateral resolution [1,2]. The vibrational energy of a molecule on a substrate strongly depends on the surrounding environment, e.g., substrate structure and its chemical identity [3]. By studying these subtle changes of the vibrational energy using STM-IETS with a molecular functionalized tip, it has been demonstrated that STM-IETS can provide information on the inner structure of a molecule and hydrogen bonding between molecules [4, 5] similar to atomic force microscopy (AFM) [6]. These advantages of the STM-IETS have accelerated research in related fields [7-16]. Owing to recent progress in the theoretical description of IETS [17-22], the qualitative understanding has been improved considerably, however, a quantitative understanding has yet to emerge. For this purpose, the following two factors need to be distinguished: (1) the ratio of the tunneling current passing through a molecule to the total tunneling current (I molecule /I total ) and (2) the efficiency of the inelastic process for the tunneling current involving the molecule ( inel ). By using these two factors, the probability of the inelastic process can be expressed as inel I molecule /I total, where the geometrical structure of the substrate and the tip are expected to play an essential role. The geometrical structure of a metal tip apex can be determined using the carbon monoxide (CO) front atom identification (COFI) method in AFM [23,24], where the tip apex of a force sensor is probed by a CO molecule that stands upright on a metal surface (Fig. 1 top). The metallic tip apex atom has a dipole moment induced by the Smoluchowski effect [25], whose direction is opposite to that of the CO molecule [26]. Thus in the distance regime where the electrostatic interaction between the tip and the molecule dominates, the force between them is attractive. When the tip is scanned over the CO molecule, this attractive force appears as a dip (smaller value) in the frequency shift image for each atom at its apex, i.e., the number of the attractive force minima provides the number of atoms on the tip apex [27]. In this paper, we have investigated the tip-structure dependent IETS for CO molecules on a Cu(111) surface by combining STM and AFM. We have found that a tip with a single atom on its apex (single-atom tip) gives a stronger IET signal over a tip with three atoms on its apex (three-atom tip). 2

4 By comparing the IETS signal with a CO attached tip for a Cu adatom and the Cu substrate, we have found that the higher IETS intensity offered by the single-atom tip mainly originates from a more focused beam of tunneling electrons (higher value of I CO /I total ) rather than an alteration in the efficiency of the inelastic process ( inel ). The dominating influence of I CO /I total on the IETS intensity has been confirmed by calculations. The experiments are carried out in an ultra-high vacuum low-temperature (4.4 K) STM and AFM combined machine (LT-STM/AFM, Omicron Nanotechnology, Taunusstein, Germany). A Cu(111) surface was cleaned by repeated sputtering and annealing before CO molecules are adsorbed on the Cu(111) surface. The force acting between a CO molecule and the apex of the metallic tip is measured by a qplus sensor [28]. The sensor whose stiffness is k=1800 N/m oscillates at f 0 =47,375 Hz with a constant amplitude of 50 pm during all STM/AFM measurements. When an average force gradient <k ts > acts between the tip and the CO molecule, the sensor frequency is shifted by f=f 0 <k ts >/2k. The current <I t > is averaged for many cycles of the sensor oscillation, as the bandwidth of the current amplifier is small compared to f 0. In order to measure the conductance (di/dv) and IET signal (d 2 I/dV 2 ), a modulation voltage (2.3 khz and 3.5 mv rms ) is added to the sample bias and the first- and secondharmonics in the current are detected by lock-in amplifiers (HF2, Zurich Instruments, Zurich, Switzerland). Throughout the whole text, the IETS is normalized with the differential conductance, i.e., (d 2 I/dV 2 )/(di/dv) [10,14,18,20]. The tip is formed from an etched tungsten wire, cleaned by field evaporation and repeatedly poked into the Cu substrate to prepare various tip apexes [23,24,29]. The repeated poking processes likely cover the tip apex with Cu atoms. The poking processes also scatter Cu adatoms on the Cu(111) surface which are employed as a counter electrode for a CO attached tip [30]. Calculations are performed with the density functional theory program Siesta [31]. From the relaxed geometries obtained by Siesta, elastic transport properties are calculated by attaching electrodes using the software package TranSiesta [32]. Vibrational frequencies and STM images are obtained from the TranSiesta calculations using the post-processing package Inelastica [33]. The Siesta (TranSiesta) calculations utilize a supercell consisting of a 7 (17) layer thick 4 4 Cu slab together with 3

5 the CO-molecule and a pyramidal tip modeled by 4 Cu atoms on the reverse side of the slab. The computations were performed using the following parameters: Perdew-Burke-Ernzerhof functional, 200 Ry real space cutoff, 4 3 k-points and a DZP (SZP) basis set for CO (Cu). To simulate the STM images, we extended the Bardeen method of Ref. [34]. The real space solutions of the Kohn-Sham orbitals are found in the vacuum region using a finite difference approximation with the full potential from TranSiesta and boundary conditions set by the TranSiesta solutions close to the surface atoms. This method overcomes the limitations of the finite range basis set used in Siesta and allows us to correctly describe the dip in the current over the CO molecule which is not feasible with a standard TranSiesta calculation, see Ref. [35] for further details. Fig. 1(a)[(c)] shows a constant-height current image of a CO molecule on the Cu(111) surface by a single-atom tip [a three-atom tip], as confirmed by the simultaneously acquired f image (Fig. 1(b)[(d)]). A sample bias V t = 35 mv and an average current <I t >=10 na have been used. For both tips we see the dip in the current image at the position of the CO molecule [36], where the current on top of the CO molecule is larger for three-atom tip than for the single-atom tip owing to the larger tip area from which electrons can tunnel. Fig. 1(e) shows the IETS for CO molecules by the single-atom tip and the three-atom tip, where identical set-points are used for both tips (V t = 50 mv and <I t >=10 na on the CO molecule). As described in the supplementary information [37], a background IETS measured on the copper surface is subtracted from that on the CO molecule. In the case of the single-atom tip, the frustrated translational (~4 mev) and frustrated rotational (~35 mev) modes of the CO molecule [7-9] are clearly seen in its IETS. However, the IETS of the rotational mode acquired with the three-atom tip is 73% smaller than the intensity acquired with the single-atom tip. The smaller IETS intensity is also observed for a tip with two atoms on its apex. The strong intensity of the IETS provided by single-atom tips is confirmed by preparing different tips, which by COFI measurements (Fig. 2(a)) are single-atom tips. Cross-sections of current images at the constant-height scan (V t = 1 mv and <I t >=1 na on the Cu) for those tips are shown in Fig. 2(b). Note that tip #5 is the one used in Fig 1(a) and (b). In the case of tips #1 through #5, the minimum current acquired on the CO molecules is almost identical (0.24±0.02 na) and the value is 24% of that 4

6 on the Cu surface, thus these single-atom tips are judged to be sharp. On the other hand, tip #6 shows a higher ratio of the minimum current on the CO molecule to the Cu surface (40 %) than that of the other tips. Taking the observation of a single minimum in the COFI image for tip #6 (Fig. 2(a)) into account, this observation suggests the existence of a secondary atom contributing to the tunneling which is not located on the vicinity of the tip apex (Fig. 2(c)). The red dotted line is a cross-section for the tip #1 multiplied by 0.79 with an offset of 0.21 na, which is consistent with the cross-section for tip #6. We therefore suggest that 79% of the current comes from the single-atom tip apex with the remaining 21% coming from a secondary atom when positioned away from the CO molecule. On the other hand, when the tip apex atom is located on the CO molecule, this ratio is changed to 47% (= /( )) for the tip apex atom and 53% for the secondary atom. Note that f for tip #6 is considerably lower than #1 through #5, which is consistent with the picture that tip #6 is blunt and thus enhances the van der Waals force between the tip and the substrate. The normalized IETS is also consistent for the different sharp single-atom tips (tip #1 to #5), especially the rotational mode shows a stable intensity of 13.6 ±1.3 [1/V] (Fig. 2(d)). The intensity of the translational mode shows more variability, probably owing to an uncertainty in the background signal. On the other hand, the IETS with the blunt tip (#6) is considerably smaller, e.g, the intensity of the rotational modes is 58 % of those with sharp tips, which is roughly consistent with the ratio of the tunneling from the tip apex atom to the total current (47%) when the tip apex locates on the CO molecule. This similarity suggests the hypothesis that the efficiency of the inelastic process ( inel ) for the tunneling process involving the CO molecule is identical for sharp and blunt tips. The decrease in IETS intensity for the three-atom tip can then be rationalized in the same way as a decreased ratio of tunneling current involving the CO molecule to the total current (I CO /I total ). To investigate how the inelastic efficiency ( inel ) depends on the geometry we now present IETS measurements using different configurations. The IETS dependence on the electrode geometry opposite to a CO molecule has been investigated with the same set-point (V t = 50 mv and <I t >=5 na) for four different situations: IETS by a sharp single-atom tip for (1) a CO on a Cu adatom and (2) a CO on the Cu(111) surface (Fig. 3(a)) and IETS 5

7 by using a CO attached sharp single-atom tip for (3) a Cu adatom and (4) the Cu(111) surface (Fig. 3(b)) [37]. In the case of the single-atom tip in Fig. 3(a) the back-ground contribution has been subtracted, however this is not possible for the CO attached tip in Fig. 3(b), direct comparison between the IETS in (a) and (b) is therefore not possible. However, comparison between the IETS with the same tip is meaningful, i.e., the two curves within Fig. 3(a) and 3(b). Note that before the CO attachment to the tip and after the CO detachment from the tip, COFI images confirmed that the tip apex structure is unchanged [37]. In the four situations investigated in Fig. 3, tunneling electrons are emitted from the single-atom opposite to a CO molecule or on which a CO molecule is adsorbed, thus the electron beam is focused and I CO /I total is expected to be large and similar between the four cases. The similarity of the IETS intensity between two cases in (a) indicates that the inel does not depend on the existence of the Cu adatom under the CO molecule, which is quite reasonable. On the other hand, the similarity of the IETS intensity in (b) indicates that the structure of a counter electrode does not strongly influence the inel, which is not trivial. The observation that the inelastic efficiency ( inel ) is independent of the counter-electrode structure suggests that the stronger IETS intensity of the single-atom tip over the three-atom tip originates from a higher value of I CO /I total rather than inel. Hereafter we will discuss the dependence of I CO /I total on the number of atoms at the tip apex basing on theoretical simulations. The solid lines in Fig. 4 show the calculated cross-sections of constant-height current images on a CO molecule from the Bardeen method for (a) the single-atom tip and (b) the three-atom tip. The distance between the tip apex and the O atom is 500 pm for the single-atom tip and 590 pm for the three-atom tip, which provides approximately the same current on the Cu(111) surface at the sample bias of 1 mv: 11.5 pa for the single-atom tip and 14.2 pa for the three-atom tip. The ratio of the minimum current to the maximum current is 16 % for the single-atom tip and 45 % for the three-atom tip, which is consistent with the experiment (see dotted lines in Fig. 4). This result suggests that the Bardeen method (I Bardeen ) well reproduces the total tunneling current between the tip and Cu surface. On the other hand, calculations using TranSiesta (I TranSiesta ), which uses a localized basis set, do not capture the increased current when 6

8 the tip is moved away from the molecule. This failure is caused by the rapid decay of the localized basis orbitals in the vacuum region where the true wave-function is more delocalized as captured by the Bardeen method. This indicates that the increased current and decreased magnitude in the current dip for the three-atom tip is caused by current going from the tip directly to the extended states of the Cu substrate and not through the CO-molecule. Assuming that the TranSiesta calculation mainly captures the current through the molecule, where the basis set of the CO molecule at least qualitatively describes the wave-function, the theoretical ratio of I CO /I total between the two systems can be compared by approximating I CO /I total I TranSiesta /I Bardeen. In the present case, I CO /I total of the three-atom tip is 22 % of that by the single-atom tip, which is consistent with ratio of the experimentally observed IETS in Fig. 1(e) (27%). With relation to the Fig. 3 (a) and (b) we note that: (1) The energy of the CO rotational mode is shifted from 34.9 mev to 33.0 mev when the position of the CO molecule is changed from the Cu(111) substrate to the Cu adatom, which originates from the lower binding energy of the CO molecule bound to the adatom than that to the Cu(111) surface [39] and is consistent with our DFT calculation (from 32.0 mev to 29.0 mev). (2) The vibrational energy of the CO molecule attached to the single-atom tip is 32.0 mev, which is close to that value for the CO molecule on the Cu adatom. The similarity suggests that the tip-apex is likely coated by Cu atoms. In summary, by combing STM and AFM, we have investigated the electrode structure dependent IETS for CO molecules on the Cu(111) surface. We have found that when at least one apex of the electrodes is sharp and consists of a single-atom, the IETS intensity is largely determined by the ratio of the current through the molecule to the total current, I co /I total, rather than inelastic efficiency ( inel ). On the other hand, when the both electrodes are planar structures, the IETS intensity considerably decreases owing to the decreased ratio of I co /I total. The fraction of the current involving a CO molecule is also studied by the theoretical calculations, which explains the experiment results. The data presented here was acquired for well-defined electrode structures and can thus be used to benchmark the theory, aiming to an accurate, quantitative description of both elastic and inelastic processes. This would in turn promote the use of IETS to characterize a single-molecule in a nanometer sized gap 7

9 formed by electro-migration or break junction techniques [40,41], where various electrode structures are expected to appear. We are deeply indebted to Thomas Frederiksen, Aran Garcia-Lekue and Alfred. J. Weymouth for stimulating discussions, and to Daniel Meuer and Florian Pielmeier for the sample preparation and sensor construction. This study was partially supported by a funding (SFB 69) from Deutsche Forschungsgemeinschaft (F.J.G); by JSPS "Strategic Young Researcher Overseas Visits Program for Accelerating Brain Circulation" (T.A. and N.O.); by a Grant-in-Aid for Young Scientists (B) ( ) from MEXT (N.O.); by a grant from the Swedish Research Council ( ) (A.G. and M. P.). 8

10 References [1] W. Ho, J. Chem. Phys. 117, (2002). [2] B. C. Stipe, M. A. Rezaei, and W. Ho, Science 280, 1732 (1998). [3] H. J. Lee and W. Ho, Phys. Rev. B 61, R16347 (2000). [4] C. L. Chiang, C. Xu, Z. Han, and W. Ho, Science 344, 885 (2014). [5] P. Hapala, R. Temirov, F. S. Stefan Tautz, and P. Jelínek, Phys. Rev. Lett. 113, (2014). [6] L. Gross, F. Mohn, N. Moll, P. Liljeroth, and G. Meyer, Science 325, 1110 (2009). [7] L. J. Lauhon and W. Ho, Phys. Rev. B 60, R8525 (1999). [8] A. J. Heinrich, C. P. Lutz, J. A. Gupta, and D. M. Eigler, Science 298, 1381 (2002). [9] L. Vitali et al., Nano Lett. 10, 657 (2010). [10] N. Okabayashi, M. Paulsson, H. Ueba, Y. Konda, and T. Komeda, Phys. Rev. Lett. 104, (2010). [11] N. Okabayashi, M. Paulsson, H. Ueba, Y. Konda, and T. Komeda, Nano Lett. 10, 2950 (2010). [12] N. Okabayashi, M. Paulsson, and T. Komeda, Prog. Surf. Sci. 88, 1 (2013). [13] H. Gawronski and K. Morgenstern, Phys. Rev. B 89, (2014). [14] K. J. Franke, G. Schulze, and J. I. Pascual, J Phys Chem Lett 1, 500 (2010). [15] M. Grobis et al., Phys. Rev. Lett. 94, (2005). [16] K. Motobayashi, Y. Kim, H. Ueba, and M. Kawai, Phys. Rev. Lett. 105, (2010). [17] N. Lorente and M. Persson, Phys. Rev. Lett. 85, 2997 (2000). [18] M. Paulsson, T. Frederiksen, H. Ueba, N. Lorente, and M. Brandbyge, Phys. Rev. Lett. 100, (2008). [19] A. Garcia-Lekue, D. Sanchez-Portal, A. Arnau, and T. Frederiksen, Phys. Rev. B 83, (2011). [20] E. T. R. Rossen, C. F. J. Flipse, and J. I. Cerda, Phys. Rev. B 87, (2013). [21] G. Teobaldi, M. Penalba, A. Arnau, N. Lorente, and W. A. Hofer, Phys. Rev. B 76, (2007). [22] A. Troisi and M. A. Ratner, Phys. Rev. B 72, (2005). [23] J. Welker and F. J. Giessibl, Science 336, 444 (2012). [24] T. Hofmann, F. Pielmeier, and F. J. Giessibl, Phys. Rev. Lett. 112, (2014). 9

11 [25] R. Smoluchowski, Phys. Rev. 60, 661 (1941). [26] M. Schneiderbauer, M. Emmrich, A. J. Weymouth, and F. J. Giessibl, Phys. Rev. Lett. 112, (2014). [27] M. Emmrich et al., Science (DOI: /science.aaa5329). [28] F. J. Giessibl, Appl. Phys. Lett. 76, 1470 (2000). [29] M. Emmrich et al., to be published in Phys. Rev. Lett.. [30] L. Bartels, G. Meyer, and K. H. Rieder, Appl. Phys. Lett. 71, 213 (1997). [31] J. M. Soler et al., J. Phys.: Condens. Matter 14, 2745 (2002), [32] M. Brandbyge, J. L. Mozos, P. Ordejo n, J. Taylor, and K. Stokbro, Phys. Rev. B 65, (2002). [33] T. Frederiksen, M. Paulsson, M. Brandbyge, and A. P. Jauho, Phys. Rev. B 75, (2007). [34] O. Paz and J.M. Soler, Phys. Stat. Sol. (b) 243, 1080 (2006). [35] A. Gustafsson et al., in preparation. [36] R. K. Tiwari, D. M. Otalvaro, C. Joachim, and M. Saeys, Surf. Sci. 603, 3286 (2009). [37] See Supplemental Material. [38] For the tip#6, V t = 1 mv and <I t > = 1.0 na is used exceptionally. [39] A. J. Weymouth, T. Hofmann, and F. J. Giessibl, Science 343, 1120 (2014). [40] N. J. Tao, Nat. Nanotechnol. 1, 173 (2006). [41] R. H. M. Smit et al., Nature 419, 906 (2002). 10

12 Fig. 1. (color online) Constant-height, (a)[(c)] current and (b)[(d)] frequency shift images (1.5nm 1.5nm) for a CO molecule adsorbed on Cu(111) by a single-atom tip [three-atom tip]. The tip height is set on the Cu(111) substrate at V t = 35 mv and <I t >=10 na. (e) Normalized IETS for a CO molecule at a set-point of V t = 50 mv and <I t >=10 na, where the IETS on the Cu(111) surface is subtracted [37]. In the left upper panel, the light blue and dark blue area between the tip and the surface schematically represent tunneling processes involving/not involving the CO molecule, respectively. 11

13 Fig. 2. (color online) (a) COFI images (1.5nm 1.5nm) by the single-atom tips at the set-point of V t = 1 mv and <I t >=1.5 na [38] on the Cu(111) surface. Tip #5 is the one used in Fig 1(a) and (b). (b) Cross-sections of the constant-height current images for CO molecules with various single-atom tips (tip #1 to #6), where the tip heights are set on the Cu substrate at V t = 1 mv and <I t >=1 na. (c) Schematic images of the single-atom sharp and blunt tips. (d) IETS for CO molecules with the same tips shown in (a) and (b) where the set-point on a CO molecule is V t = 50 mv and <I t >=10 na. 12

14 Fig. 3. (color online) (a) IETS by a single-atom tip for a CO molecule on a Cu adatom (blue) and a CO molecule on the Cu(111) surface (red). (b) IETS by a CO attached tip for a Cu adatom (blue) and the Cu(111) surface (red). The back-ground signal is subtracted in (a) and not subtracted in (b). For all IETS measurements, the tip-height is set at V t = 50 mv and <I t >=5 na. 13

15 Fig. 4. (color online) Solid lines are constant-height current images for a CO molecule on Cu(111) surface at the bias of 1mV by (a) the single atom tip and (b) the three-atom tip. The distance between the tip apex atom and the O atom is 500 pm in (a) and 590 pm in (b). The solid squares are the results by the TranSiesta when the tip center is located on the CO molecule, where the current involving the CO molecule is dominantly estimated rather than the direct tunneling between the tip apex atom and Cu substrate. 14