Investigation of the structures of sulfur on Mo(100) by scanning tunneling microscopy

The structures of the four ordered overlayers of sulfur on the Mo(100) surface were investigated by scanning tunneling microscopy (STM). In order of increasing sulfur coverage, the overlayers have c(2×2), | 12 11 |, c(4×2), and p(2×1) low energy electron diffraction (LEED) patterns. Only the c(2×2) overlayer geometry has been determined from LEED I(V) analysis. An examination of point and line defects and domain boundaries in STM images provides information beyond that known from LEED on the required symmetries of the ordered overlayers. Several proposed structures were shown to be incompatible with these symmetries and therefore discarded. Only one model of the c(4×2) structure was found to be consistent with the symmetry of the STM images. This structure contains sulfur atoms occupying exclusively fourfold hollow sites. Unexpectedly, STM images of the surface with the p(2×1) LEED pattern did not have p(2×1) symmetry, but consisted of small domains of the c(4×2) structure.


I. INTRODUCTION
Molybdenum single crystals surfaces are of interest as a model for industrial molybdenum sulfi de catalysts. and p(2 X 1) symmetry. These coverages correspond to one, two, three, and two atoms per unit cell, respectively.
A LEED calculation performed on the c ( 2 >< 2) structure determined that the sulfur atoms reside at fourfold hollow sites.5 A later dynamical LEED11 and a tensor LEED12 calculation determined the precise position of the sulfur adatoms and the displacements of the substrate Mo surface

II. EXPERIMENT
Sample preparation and STM imaging were performed inside a standard surface science ultra high vacuum (UHV) chamber with a base pressure of 5 X 10-10 Torr.
Besides the STM, the chamber was equipped with Auger electron spectroscopy (AES), LEED, and Ar i-sputtering. Prior to imaging of the sulfur overlayer, AES was used to determine sulfur coverage on the surface. The sample was annealed to produce an ordered LEED pattern. STM images of the surface were obtained using several different tips and preparations of each structure. Several hundred STM images of the overlayer structures were obtained. As discussed in a separate paper, 10 sulfur adsorption was found to modify the structure of the atomic steps on the surface by causing the coalescence of steps and the conse quent enlargements of the ( 100) terraces.
The surface was imaged in both constant current (to pographic) and constant height mode. In most of the im ages the sample was biased negative relative to the tip.
Changing the polarity of the bias was found to have no effect on the images. The tunneling current was set at be Thermally induced drifts of the sample during imaging significantly distorted the images, particularly the topo graphic mode images that require longer acquisition time.
The correct shape and size of the images was determined from current mode images acquired quickly enough so that the drift had little effect on them. Using this information the images were replotted with the correct shape using a two-dimensional second-order fi t of the lattice. All the im ages shown in this paper are drift-corrected topographic mode images. To make interpretation easier, the images are strate. They were also Fourier fi ltered to remove spatial frequency components too high to represent real informa tion.

Ill. RESULTS
A. c(2X2) structure is likely that these areas are due to a different structure of sulfur and not to an impurity. They probably consist of sulfur which has not ordered into the c ( 2 X 2) structure due to insufficient annealing or too low a sulfur coverage. secondary maximum at the center of four primary maxima.
We believe that these images were produced by a different STM images of this structure could be explained by any of the three models. If we assume that, as in the c( 2 X 2) structure, the sulfur atoms have positive corrugation, the rows of sulfur atoms in Figs. 5 (A) and 5 (C) could corre spond to the rows in the image. However, the rows of bridge sites in model B could also produce the rows in the image. As the bridge atoms are higher above the surface than atoms at hollow sites one would expect the rows of bridge site atoms in model B to appear brighter in an STM image than the rows of hollow site atoms. The observed variations in the experimental shape of the rows can be explained in all cases as the effect of a particular tip struc ture and are not helpful in determining the correct unit cell structure.
Since the ! 3 I I structure coexisted on the surface with the c(2X2) structure it was possible to determine the rel ative position of the rows and the maxima of the c(2X2) overlayer in the STM images. Figure 6 shows an image of the boundary region between both structures. In this image the c(2X2) domain occupies the right-hand side and its There are two perpendicular mirror planes in each unit cell which cross at the maxima. The maximum corrugation between the rows is 0.45 A.
tice directions. However, since we have not yet ascertained that sulfur atoms have positive corrugation in this struc ture, the result is not conclusive.

C. c(4 X2) structure
At a coverage of approximately three quarters of a monolayer a sharp c( 4 X 2) LEED pattern was obtained.
STM images of the surface showed it was covered with ordered domains of approximately 100 A in size. An ex ample of an image of a single domain is shown in Fig. 7. The staggered vertical rows of maxima in the figure are separated by twice the Mo lattice vectors. The maxima along the rows are also separated by this distance. The symmetry of the maxima in this image is c ( 4 X 2), consis tent with the LEED pattern. There are two perpendicular mirror planes in each unit cell of the STM image which run through the maxima in the unit cells. Six possible model structures, shown in Fig. 8, have these two mirror planes and contain the three atoms per unit cell required to have 0.75 monolayers coverage. In models A-C the perpendic ular mirror planes cross at hollow sites while in models D-F they cross at bridge sites.
Half of these models can be discarded by observing the symmetry of domain boundaries in this structure. An im age of the boundary between two domains of c( 4 X 2) or dered sulfur rotated by 90° relative to each other [c(4X2) and c ( 2 X 2)] is shown in Fig. 9. A ( 1 X 1 ) lattice is plotted over this image and lined up with the maxima of one do main. In the rotated domain the lattice is also found to line up with the maxima. This indicates that the maxima must be located at fourfold (hollow or top) sites on the lattice. to "mix" into a (2 X 2) structureo The structure in this area can be explained as a periodic array of antiphase bound aries like that shown in Fig. lO(a). This region appears to have a fourfold symmetry, and extends away from both the c(4X2) and c(2X4) domains symmetrically. As two do- D. "p(2 X 1 )" structure Producing a sample with a sharp p(2 X 1) LEED pat tern proved to be difficult. Since it is not possible to pro duce the required saturation coverage of sulfur by exposure FIG. 11. A 70 X 70 A image of the surface after a p(2 X l) LEED pattern was obtained. The symmetry is not the expected p(2 X l). The surface is instead covered by small domains of one to three unit cells with c( 4 X 2) symmetry. Extra sulfur atoms at the domain boundaries increase the average sulfur coverage above that of the ordered c( 4 X 2) ovcrlayer. Bias= 100 mV, I= 1.0 nA.
to low pressures of H2S, the electrochemical sulfur source was used to deposit sulfur on the surface. Annealing of the surface was necessary to obtain an ordered LEED pattern.
Excessive annealing produced the c ( 4 X 2) pattern while insufficient annealing left the LEED pattern diffuse. Previ ous reports 4 • 5 of this structure have stated that the LEED pattern was often streaked, at least at some sulfur cover ages. In the present experiment the overlayer spots of the LEED pattern were streaked in the direction of the quarter order spots of the c( 4 X 2) LEED pattern at some electron energies. Adjusting the electron energy changed the inten sity of the streaking; at some energies it was nearly invisi ble.
Unlike the previous lower coverage structures, we were unable to obtain STM images of a surface structure con sistent with the symmetry of the LEED pattern. This result is different from the previous STM study of this surface in air in which images of a surface with apparent p( 2X 1) symmetry were obtained. 14 Because of the air environment the observed p(2X 1) ordering may be due to some impu rity which was absent in our UHV experiment.
A representative image of the surface with a p(2 X 1) LEED pattern is shown in Fig. 11. It consists of very small domains of c ( 4 X 2) ordered sulfur. This image is not ac tually inconsistent with the LEED pattern, but only with its apparent p(2X 1) symmetry. The c(4X2) LEED pat tern contains all the spots of the p(2X 1) pattern and ad ditional quarter order spots. The streaking we observed in the p(2 X 1) LEED pattern was in the direction of these spots. The lack of large domains on the surface would cause streaking of the higher-order spots as they corre spond to longer range order on the surface. The higher coverage of sulfur on this surface relative to the ordered c ( 4 X 2) overlayer may be explained by the large number of antiphase defects present which have a local coverage of one monolayer. As other authors have also reported streaking of thep(2Xl) overlayer LEED spots, it is pos sible that all reports of the p(2 X 1) LEED pattern were obtained from surfaces with small c( 4 X 2) domains. It may also be that some impurity is required to induce p(2 X 1 ) ordering of the sulfur.