STM studies

Fast-scanning STM studies

by Flemming Besenbacher*, Erik Lægsgaard, and Ivan Stensgaard

Dynamic processes on surfaces play a crucial role in important areas such as catalysis, thin-film growth, and sensor technology. With the advent of scanning tunneling microscopy (STM) in the 1980s, scientists could visualize these processes in real space. This has led to unprecedented new insight into, for example, diffusion, reaction, nucleation, and growth phenomena on surfaces at the atomic level, strongly improving the basis for surface engineering in materials science. An imperative for visualizing dynamic phenomena on surfaces is the ability to acquire adequate temporal resolution, i.e. to record STM images at a sufficiently high rate. In this article, we illustrate, primarily with examples from our own laboratory, how fastscanning STM can be applied successfully to studies of dynamic processes on surfaces, in particular (a) mass transportation induced by adsorption, (b) diffusion of adatoms on surfaces, (c) diffusion of vacancies on TiO2, and (d) the influence of molecular orientation on diffusion properties.

Interdisciplinary Nanoscience Center (iNANO), and Department of Physics and Astronomy, University of Aarhus, DK-8000 Aarhus C, Denmark *E-mail: [email protected]

26

May 2005

With the invention of STM in the 1980s by Binnig, Rohrer, and coworkers1, exploration of the atomicscale realm of surfaces in real space became feasible. What fascinates many scientists, and sets STM apart from most other surface techniques, is its ability to investigate the geometric, and in some cases also the electronic, structure of flat surfaces atom-by-atom. Up to now, STM has had its greatest impact in the field of surface science, although its impact in the fields of materials science, electrochemistry, and biology is growing steadily. Atomic, or at least molecular, real-space resolution can be achieved on a wide variety of surfaces in various environments, for example ultrahigh vacuum (UHV), well-controlled gas reaction cells, ambient conditions, and liquids (electrolytical cells). But, in all cases, the surface to be studied must be conductive or semiconductive2-5. Since its advent, STM’s development has been considerable. From being a rather complicated, home-built instrument that was sensitive to vibrations and noise in general, the STM has developed into a general, rather compact, stable, and variable-temperature instrument that is now commercially available at a fairly low cost; a fact that has diversified the application of the microscope tremendously such that new areas of applications are constantly being developed2-3,6. An important application of STM is the manipulation of single atoms or molecules on surfaces. By placing the tip of the microscope in close proximity to an atom or a molecule and exploiting attractive or repulsive forces between the tip

ISSN:1369 7021 © Elsevier Ltd 2005

REVIEW FEATURE

Adsorbate-induced mass transport

Fig. 1 The Aarhus STM11. The sample (1) is placed in a tantalum holder (2) held down on the STM top by springs (3). The top plate is thermally and electrically insulated from the STM body by three quartz balls (10) and mounted on a 0.6 kg Al block that can be cooled to -160°C or heated to 100°C. The tip (4) is held by a Macor® holder (5) glued to the top of the scanner tube (6). The scanner tube is 4 mm long with an outer/inner diameter of 3.2/2.1 mm and is glued to the rod (7) which, together with the piezo tube (9), form a small inchworm motor that is fixed to the STM body by the Macor ring (8) and used for coarse approach. The motor works in steps as small as 2 Å, but at full speed it moves at ~2 mm/min. The scan range is up to ±1 µm when antisymmetrical scan voltages of ±200 V are used. The Zener diode BZY93C75 (11) is used to counterheat the STM body during cooling. A compact and rigid design implies that the lateral and longitudinal resonance frequencies of the STM are as high as 8 kHz and 90 kHz, respectively. This reduces sensitivity to external vibrations and allows fast scanning.

and the atom/molecule, it is possible to push, pull, or slide the atom/molecule along a particular direction on the surface, or pick it up and drop it6. The most spectacular application of this is probably the formation of quantum corrals7, but manipulation has been very useful in, for example, revealing ‘frozen’ atomic structure under adsorbed, large molecules8,9 and changing the orientation of large molecules10 (see below). As far back as about 1990, we and several other research groups began to explore the possibility of using STM to visualize dynamic processes on surfaces, such as the mass transport associated with adsorbate-induced reconstructions and the diffusion of single adatoms and defects, by recording time-lapse series of images of the surface and replaying these in the form of a movie. Since the time required for recording a typical 256 × 256-pixel STM image in the constant-current mode for many STMs was of the order of minutes, such ‘real-time’ dynamic studies would only be possible if the mechanical construction of the STM and the STM electronics were optimized such that STM images could be recorded sequentially at approximately one (or even several) frames per second. Using our home-built Aarhus STM11 (see Fig. 1), which has a rigid and compact design and, thus, the required high mechanical resonance frequency, we obtained and presented the first STM movies at an STM conference in Baltimore in 199012. Below, we illustrate this fast-scanning STM capability by discussing a number of examples, primarily from our own research activities (see13 for movies).

One of the simplifying assumptions behind the descriptions of adsorption of atoms and molecules, and in particular chemical reactions between adsorbates on metal surfaces, has often been that the surface can be considered as a static ‘checkerboard’ that provides adsorption sites for, and bonds to, the adsorbing atoms and molecules, as well as pathways for the dissipation of energy in and out of the reaction coordinates. However, many studies have found that this is rather the exception. Instead, reactive adsorbates such as H, C, N, O, and S in many cases cause a restructuring of both metal and semiconductor surfaces4,5,14, i.e. atoms in surface layers tend to arrange themselves so that the total energy is a minimum; this will often involve a shift in the positions of the atoms with respect to each other. Dynamic STM studies have led to unprecedented new insight into this area of adsorbate-induced reconstructions of surfaces. We will briefly illustrate this by the now classical example of oxygen-induced restructuring on Cu(110). Around 1990, it had long been known that oxygen adsorbs dissociatively on Cu(110), resulting in a doubling of the unit cell to a (2×1) structure. After more than 20 years of research, the detailed atomic structure of this phase was still open to debate, although most studies favored a missing-row model in which every second [001] surface row was missing. By ‘filming’ the surface, i.e. recording sequential STM images of the same area of the surface, during the adsorption-mediated reconstruction process, the dynamics of the transformation were visualized revealing an unexpected mechanism15.

Fig. 2 Four snapshots from an STM movie of a Cu(110) surface recorded after progressively higher oxygen exposure; the oxygen pressure during recording was 10-8 torr. The imaged area (235 x 256 Å2) originally contained three step edges. It is seen that Cu atoms are removed from step edges during the oxygen exposure. Correlated with the Cu removal, added –Cu-O– rows nucleate and grow on the terrace along the [001] direction. The oxygen exposures are: (a) 0.3 L; (b) 3 L; (c) 19 L; and (d) 54 L (1 L = 10-6 torr × sec).

May 2005

27

REVIEW FEATURE

Fig. 2 shows four images from such an STM movie13. Several terraces of the Cu(110) surface are visible, separated by monoatomic steps. As oxygen exposure is increased, the removal of Cu atoms from the step edges is clearly observed and, simultaneously, anisotropic growth of added rows, consisting of –Cu-O– segments along the [001] direction, takes place on the terraces. When the surface is saturated by the (2×1) phase, the missing-row and the added-row models are indistinguishable, but the ways that they nucleate and grow clearly differ in terms of mass transport14-16: the missing-row model would lead to mass transport to the steps rather than away from the steps, as observed experimentally. The dynamics of the Cu/O system have been visualized in several other cases by fast-scanning STM. See14 for a review.

Dynamics of adatom surface diffusion A detailed understanding of the mechanisms and energetics of single atoms, molecules, and small clusters migrating across a surface is of utmost fundamental and technological interest, with key implications for the development of microscopic models, e.g. crystal and thin-film growth, heterogeneous catalysis, and sintering. It has been shown how fast-scanning, variable-temperature STM can visualize directly the vivid diffusion of atoms, molecules, and clusters on surfaces. We will illustrate this for the diffusion of Pt adatoms on the reconstructed Pt(110)-(1×2) surface17, in which every second close-packed row is missing (see Fig. 3). Insight into the atomistics of one-dimensional, random-walk migration of the deposited Pt adatoms was obtained by acquiring many consecutive STM images played back as an STM movie13. Our pattern-recognition software allows us to follow individual adatoms over extended periods, from minutes to hours, even for changing temperatures. Individual adatom positions can be tracked in time with atomic resolution, and a distribution of image-to-image displacements of the diffusing adatoms determined. From a thorough statistical analysis of such displacement distributions, the hopping rate h at variable temperatures T can be determined. By plotting rates versus 1/T and fitting the results on an Arrhenius diffusion equation, h = ν exp(-Ed/kBT), we obtain quantitative information for the diffusion parameters17, i.e. the activation barrier for diffusion, Ed = 0.81 ± 0.01 eV, and attempt frequency ν = 1010.7±0.2 s-1. The studies were performed under very clean, idealized ultrahigh-vacuum conditions. But, in most material processes of interest, diffusion happens on surfaces at higher pressure or

28

May 2005

Fig. 3 STM image showing the Pt(110)-(1x2) surface after a submonolayer amount of Pt has been deposited. Single adatoms and small islands are observed in the troughs of the missing-row reconstruction. STM movies reveal that the Pt atoms in the troughs perform one-dimensional random-walk diffusion. The imaged area is 140 × 140 Å2.

even at ambient conditions where the surface is covered by atoms and molecules. To simulate such effects, we have shown that adsorbates such as hydrogen and oxygen promote the self-diffusion of Pt on the Pt (110)-(1×2) surface18. From STM movies, we are able to show directly the formation of intermediate, activated Pt-H complexes, which have a significantly larger diffusivity (~500-fold increase) than other Pt adatoms at room temperature13. This corresponds to a lowering of the activation barrier by ~0.16 eV. From interplay with density functional theory calculations, we show that the Pt-H complex consists of an H atom trapped on top of a Pt adatom, and that this H atom reduces the diffusion barrier18. Several other examples have since been published that show that fast-scanning STM is also a very valuable technique for obtaining quantitative insight into the diffusion of gaseous and molecular adatoms. An example of this is our study of the dynamics of N atoms on the Fe(100) surface, resulting in the derivation of an activation barrier for diffusion of 0.92 ± 0.04 eV19, and the diffusion of large organic hexa-tert-butyl decacycline (HtBDC) molecules on Cu(110) surfaces20. While jumps to nearest-neighbor sites dominate in most diffusion processes, it was found that long jumps surprisingly play a dominant role in the diffusion of HtBDC on Cu(110). Again, the movies can be found in13. Wintterlin et al.21 have studied the random walk of individual oxygen atoms adsorbed on Ru(0001) at low coverage, and the equilibrium fluctuations of oxygen islands at a higher coverage at even greater imaging rates of up to 20 frames s-1. A hopping rate of 14 ± 3 s-1 is derived. But, since measurements were made only at room temperature, neither the activation energy nor the attempt frequency could be determined. The high imaging rates are obtained by scanning in the constant-height mode (rather than constant-

REVIEW FEATURE

current mode) and by applying sinusodial scan voltages instead of the usual triangular signals. See22 for movie. Recently, a rather surprising giant atomic slide-puzzle was revealed by Frenken and coworkers23 during their studies of the diffusion of a low density of In atoms embedded as tracer atoms in the uppermost layer of a Cu(001) crystal. STM movies24 (20 s per frame) show that the In atoms make long, concerted diffusion jumps at room temperature. The ultralow density of vacancies, diffusing rapidly within the surface, appears to be responsible for this motion, and a similar vacancy-assisted mobility of atoms may occur for many other surfaces, as also demonstrated for Pd/Cu(001)25. The fast-scanning STM studies discussed so far have all been performed in clean, well-controlled UHV environments. But electrochemical STM studies by Magnussen et al.26 have provided new insight into the electrochemical deposition and dissolution of metals, which are important processes in metal corrosion, technological etching, coating, and metal refinement. Deposition and dissolution are strongly localized at ‘active sites’ commonly attributed to kink sites at steps on the surface. Crystal growth and dissolution often proceed via either direct attachment/detachment of metal species in solution or direct attachment/detachment of metal adatoms at these sites26. At 5-10 images s-1, direct observations reveal equilibrium fluctuations caused by local removal/addition of atoms at atomic kinks in steps on the crystal surface27.

Oxygen vacancy diffusion on TiO2 Defects such as oxygen vacancies often dominate the electronic and chemical properties of transition metal oxides, the surfaces of which play an important role in a wide range of applications such as heterogeneous catalysis, photoelectrolysis, and biocompatibility. By controlling the nature and density of vacancies, new means of tailoring the adhesion and wetting of metal particles and thin films on oxide surfaces will become available. Very little is known about the diffusion of defects on transition metal oxide surfaces, but recent STM dynamics studies have resulted in unprecedented new insights into this important problem. The model studied is the rutile TiO2(110) surface, which has emerged as the prototypical system for fundamental surface science studies of transition metal oxides. The anisotropic TiO2(110) surface consists of alternating rows of five-fold-coordinated Ti atoms and two-fold-coordinated O atoms aligned along the [001] direction (see Fig. 4). The

Fig. 4 Ball model of the TiO2 (110) surface. A bridging O vacancy is marked by a circle. The arrow denotes the observed vacancy diffusion pathway. (A) and (B) are two consecutive images extracted from an STM movie. (C) The corresponding difference image shows that the vacancies jump perpendicular to the Ti/O rows.

contrast in the STM image of the TiO2(110) surface is dominated by electronic rather than geometric effects, such that protruding rows are assigned to Ti rows and troughs are accordingly identified as so-called ‘bridging’ O rows, although these protrude geometrically from the surface plane by ~1.2 Å. The distinct surface features seen in Figs. 4A and B, imaged as bright spots between Ti rows, are assigned to vacancy point defects (missing O atoms) in the bridging oxygen. Using high-resolution STM movies13, we have unraveled a surprising adsorbate-mediated diffusion mechanism of O vacancies on this rutile surface28. The movies reveal that diffusion is promoted by the presence of O molecules on the surface, and that the diffusion pathway is perpendicular to the bridging O rows on the TiO2(110) surface, as shown in Fig. 4C. From low-temperature STM movies, we find that O2 molecules diffuse along the Ti rows ([001] direction) on the pristine parts of the surface. When a diffusing O2 molecule encounters an O vacancy, a strong interaction occurs that leads to the diffusion of the O vacancy to a neighboring bridging O row. The microscopic steps leading to this surprising vacancy diffusion event can be understood on the basis of a simple oxygen exchange mechanism between lattice O atoms and adsorbed O2. When the O2 molecule reaches the O vacancy, an intermediate O2/vacancy complex is formed that subsequently dissociates; one O atom of the O2 molecule fills the initial vacancy, while the other is ejected laterally onto a neighboring Ti trough. This highly reactive O adatom subsequently extracts an O atom from a neighboring bridging O row, reverting to an adsorbed O2. The final configuration results from diffusion of the vacancy to a neighboring bridging O row with the O2 molecule staying in the same Ti trough as in the initial configuration.

May 2005

29

REVIEW FEATURE

Diffusion of large organic molecules Insight into the complex dynamics of large molecules on surfaces is essential for controlling the synthesis of molecular nanostructures with potential applications in emerging fields such as molecular electronics and nanomechanical sensors. We have investigated how the orientation and shape of large and complex organic molecules may affect their dynamics10. The test molecule, known as Violet Lander (VL, C108H104), was thermally evaporated onto a Cu(110) surface and the conformation and orientation of single, individual molecules were established using STM. The VL molecule consists of a central board with four spacer legs and adsorbs spontaneously on Cu(110) in a stable configuration with the board axis aligned to the [110] surface direction. At temperatures below 200 K, the molecule’s orientation can be manipulated into a new configuration by gently pushing it with the STM tip in a direction perpendicular to the board axis. In this manipulated state, the molecule is kinetically trapped in a metastable state with the molecular board rotated by 70° with respect to the [110] direction. Surprisingly, the rotated molecule shows a high diffusivity, as shown by dynamic STM movies13. Fig. 5 shows an image composed from three snapshots from a movie recorded at 180 K, confirming that only the rotated molecules diffuse. By carefully tracking the position of individual, diffusing molecules, the manipulated molecules in the rotated configuration are found to diffuse along the [110] direction with a diffusion constant of D = (4.8 ± 0.5) × 10-17 cm2 s-1. Since nonmanipulated molecules do not diffuse at all in the

Fig. 5 Color-coded STM image of Violet Lander molecules (VL, C108H104) on Cu(110). Three images from an STM movie were shown in a gray-scale height representation and transformed to red (R), green (G), and blue (B) images, respectively, by preserving only one of the (R,G,B) color components. A composite image was then formed by adding the color components pixel by pixel. Stationary features then show up again in the gray scale. Higher-lying molecules appear white and substrate terraces dark. Diffusing molecules, however, are displayed in color at the place they occupied at the time of recording. It is evident from the composite image that only rotated molecules diffuse.

time span of observation (several minutes), their diffusion constant must be smaller than ~5 × 10-19 cm2 s-1. Our results indicate that the diffusion constant for large complex molecules may change by at least two orders of magnitude when the molecular orientation is changed by rotating about an axis perpendicular to the substrate.

Concluding remarks The quest for ever-increasing acquisition rates is bound to continue. The challenge is to image processes on surfaces in real time, i.e. to depict the surface at rates so high that no diffusion jumps, reaction events, etc. are missed. But even with video-rate imaging, images will show the outcome of events rather than the events themselves. A diffusion event, e.g. the hopping of an adatom from one site to the next, may take just 10-13 s. So, mapping out the detailed transition by fast-scanning STM is completely out of reach. MT

REFERENCES 1. Binnig, G., et al., Phys. Rev. Lett. (1982) 49 (1), 57

15. Jensen, F., et al., Phys. Rev. B (1990) 41 (14), 10 233

2. Wiesendanger, R., Scanning Probe Microscopy, Cambridge University Press, (1994)

16. Coulman, D. J., et al., Phys Rev. Lett. (1990) 64 (15), 1761

3. Bonnell, D., Scanning Tunneling Microscopy and Spectroscopy, John Wiley and Sons, (2001)

18. Horch, S., et al., Nature (1999) 398, 134

4. Besenbacher, F., Rep. Prog. Phys. (1996) 59 (12), 1737

19. Pedersen, M. Ø., et al., Phys. Rev. Lett. (2000) 84 (21), 4898

5. Neddermeyer, H., Rep. Prog. Phys. (1996) 59 (6), 701

20. Schunack, M., et al., Phys. Rev. Lett. (2002) 88 (5), 56 102

6. Meyer, G., et al., Single Mol. (2000) 1 (1), 79

21. Wintterlin, J., et al.., Surf. Sci. (1997) 394 (1-3), 159

7. Crommie, M. F., et al., Science (1993) 262, 218

22. www.cup.uni-muenchen.de/pc/wintterlin/worke-frame.html

8. Schunack, M., et al., Phys. Rev. Lett. (2001) 86 (3), 456

23. van Gastel, R., et al., Nature (2000) 408, 665; Phys. Rev. Lett. (2001) 86 (8), 1562

9. Rosei, F., et al., Science (2002) 296, 328 10. Otero, R., et al., Nat. Mater. (2004) 3 (11), 779 11. Lægsgaard, E., et al., Rev. Sci. Instrum. (2001) 72 (9), 3537

30

17. Linderoth, T. R., et al., Phys. Rev. Lett. (1997) 78 (26), 4978

24. www.physics.leidenuniv.nl/sections/cm/ip/projects/dynamics/incopper/ incopper.htm

12. Besenbacher, F., et al., J. Vac. Sci. Technol. B (1991) 9 (2), 874

25. Grant, M. L., et al., Phys. Rev. Lett. (2001) 86 (20), 4588

13. www.phys.au.dk/camp/m-t

26. Magnussen, O. M., et al., Faraday Discuss. (2002) 121, 43

14. Besenbacher, F., and Stensgaard, I., In The Chemical Physics of Solid Surfaces and Heterogeneous Catalysis, King, D. A., and Woodruff, D. P., (eds.), Elsevier (1994), 7

27. www.ieap.uni-kiel.de/solid/ag-magnussen

May 2005

28. Schaub, R., et al., Science (2003) 299, 377