PHYSICAL REVIEW 8 VOLUME 39, NUMBER 8 15 MARCH 1989-I Atom-resolved surface chemistry studied by scanning tunneling microscopy and spectroscopy Ph. Avouris and R. Wolkow IBM Research Division, Thomas J. 6'atson Research Center, P. O. Box 218, Yorktown Heights, New York 10598 (Received 18 July 1988) We have used scanning tunneling microscopy and spectroscopy to study the reaction of Si(111)- (7X 7) with NH3. We have found that by use of topographs obtained at different energies, as well as atom-resolved spectra, reacted and unreacted surface sites can be imaged selectively. Thus we have been able to probe the spatial distribution of the surface reaction on an atom-by-atom basis. We find that there are significant differences in reactivity between the various dangling-bond sites on the Si(111)-(7 X 7) surface. Specifically, rest-atom sites are more reactive than adatom sites and, more- over, center-adatom sites are more reactive than corner-adatom sites. We ascribe the reduced reac- tivity at adatom sites to the delocalized nature of their dangling-bond state. We suggest that a bonding interaction between adatoms and the Si atoms directly below them is responsible for this behavior — a suggestion supported by electronic-structure calculations. Thus, while reaction at a rest-atom site can be considered a dangling-bond saturation process, reaction at an adatom site in- volves the formation of a hypervalent (fivefold-coordinated) adatom. We tentatively ascribe the reactivity differences between center and corner adatoms to differences in the strain they induce upon reaction on the dimer bonds. Atom-resolved spectroscopy allows us to probe interactions and charge transfer between surface sites, and for the first time, we can directly observe how chemisorp- tion affects the substrate electronic structure at neighboring unreacted sites. I. INTRODUCTION Scanning tunneling microscopy (STM) has been used to image the structure of clean surfaces aq. d has provided valuable insight into the nature of surface reconstruc- tions, particularly for semiconductor surfaces. ' Recently, it also has been used fo study the topography of ad- sorbed, primarily metallic, layers on semiconductors. The ability of STM to provide atomic-scale information on both geometric and electronic structure can lead to a completely new way of studying surface chemistry. Us- ing STM, one should be able to follow the extent and spa- tial distribution of a surface reaction on an atom-by-atom basis. By probing simultaneously the electronic structure of the different surface sites one then can relate electronic structure and reactivity. Moreover, in this way one could directly probe the interactions between surface sites. The existence of such interactions is inferred in a variety of experiments, such as in vibrational spectroscopy of adsor- bates and in the study of adsorption isotherms, where the adsorbate-substrate interaction is studied as a func- tion of coverage. The macroscopic and indirect nature of such measurements, however, does not allow one to discriminate between effects due to modification of the electronic structure of the substrate by the adsorption process and direct lateral interactions between the adsor- bates. In this paper, we present results of scanning tunneling microscopy (STM) and atom-resolved scanning tunneling spectroscopy (AR-STS) studies of the reaction of am- monia (NH3) with Si(111)-(7X7). This study indicates that the above goals can be achieved by combined STM and AR-STS studies. In previous work we have shown that the reactivity of Si(100)-(2X 1) towards a variety of molecular systems is strongly correlated with the pres- ence of surface dangling-bond states. However, all dan- gling bonds on the Si(100)-(2X1) surface are located on equivalent sites. A far more chemically interesting case and one which would take full advantage of the unique capability of STM to provide atom-resolved information, would be a study of the reactivity of a surface with several structurally distinct active sites. The Si(111)- (7X7) surface is deal in this respect. The nature of the 7 X 7 reconstruction itself has been the subject of intense study for about 30 years, and the role of STM (Ref. 7) was quite important in the resolution of this problem. Currently, the model proposed by Takayanagi et al. [the so-called dimer — adatom — stacking-fault (DAS) model] is generally accepted. The DAS model for the 7X7 unit cell is shown in Fig. 1. There are two triangular subunits each surrounded by nine Si dimers; in addition, there is a stacking fault in the left triangle. On the surface there are six triply coordinated Si atoms (labeled A and 8 in Fig. 1), so-called rest atoms. The top layer is composed of 12 Si adatoms (solid circles), and, finally, at the corners of the unit cell there are vacancies usually referred to as corner holes. For reasons that will become apparent later in the paper we separate the adatoms into two groups: the ones located next to a corner hole are termed corner adatoms, while the other six are called center adatoms. The most important chemical effect of the reconstruction is a severe reduction in the number of surface dangling bonds (DB's). While on the unreconstructed Si(111) sur- face there are 49 DB's, only 19 survive in the 7 X 7 unit cell. Of these, 12 are located on the two types of ada- toms, six on the rest atoms and one on the atom at the bottom of the corner hole. Thus, there are a variety of chemically active surface sites which would allow us to 39 5091 1989 The American Physical Society