Hot STM Labs

Hot STM Studies of MBE growth of Si on Si(001)

We studied the initial stages of Si MBE growth on Si(001) surfaces. In the initial stages of MBE growth at 530 K, many small one-dimensional (1D) islands are formed. The explanation of this curious shape anisotropy has been controversial. On the Si(001)-(2 × 1) surface, the mobility of Si monomers and dimers is high along the dimer rows and low across them. It has been observed that for certain growth conditions, deposited Si atoms form islands a single dimer row wide (1D islands). The long axes of these 1D islands are aligned perpendicular to the substrate dimer rows. It is puzzling that the islands grow perpendicular to the direction of fast diffusion. Because the mobility is high along the rows, one might expect that islands would capture atoms from a greater distance in this direction. This would lead to growth predominantly along the rows, contrary to what is observed.

Mo et al. proposed a kinetic model based on sticking anisotropy to explain the island shape anisotropy. If island ends are stickier than island sides, arriving adatoms stick preferentially to the ends, resulting in the growth of many 1D islands. Tsao and co-workers suggested that an opening or weakening of dimer bonds might explain this sticking anisotropy. Metiu et al. proposed an alternative model based on an exchange mechanism. Metiu suggested that an adatom arriving on the side of a Si island may displace an existing island atom to the top of the island. The displaced atom diffuses rapidly along the top of the island (the direction of fast diffusion) until reaching an end, where it can fall over the edge and stick. Adatoms arriving on island sides are transported to the ends via this exchange mechanism, resulting in enhanced 1D growth. Our ability to track individual islands microscopically with the STM allowed us to test these models directly.

FIG 1. 400 × 400 Å STM Image of the Si(001) surface at 536 K showing many small Si islands. The direction of fast monomer diffusion is parallel to the dimer rows, indicated by the black arrows.

In Fig. 1 we show an image of the Si(001) surface decorated with many Si islands. Islands have grown on the upper and lower terraces with their long axes aligned perpendicular to the substrate dimer rows. According to the simple sticking anisotropy model of 1D island growth, a long 1D island should be no more likely to capture material at its ends than a short one, for both islands have just two ends. The Metiu model, however, implies that the effective end capture probability should increase linearly with length. Because the exchange mechanism transports adatoms arriving on island sides to island ends, doubling an island's length should double the rate at which material is added to its ends. A portion of one movie is shown in Fig. 2. From detailed analysis of many movie images, we found that growth is independent of length, supporting the anisotropic sticking model. One advantage of our method is that we can select and follow just the 1D islands. The simple anisotropic sticking model is therefore confirmed as the cause the island shape anisotropy observed during growth. We then use our data to measure the anisotropic sticking ratio, yielding a sticking anisotropy ratio of 0.019 ± 0.003. Thus, an end site is roughly 50 times more likely to gain a block than a side site.

 

Figure 3. 800 × 800 Å STM images showing the growth of islands at 533 K. Time advances from left to right. Coverage increases from 0 to 0.1 ML at a rate of approximately 0.01 monolayers deposited per frame. The movie begins with the clean substrate before deposition. At this temperature 1-D islands form and coalesce. View Java Applet Movie or View Gif Animation

This result gives us insight into MBE growth of Si on Si(001). It is known that in step flow growth, type-B step edges grow faster than type-A edges, eventually causing double height steps to form. Although most material arriving at a step edge is incorporated at existing kink sites, the creation of new kinks (by addition of material to previously flat sections) is the rate-limiting step for the advance of the edge. Because the side of a 1D island is a type-A edge, adding a block there is like adding a block to a flat section of a type-A step edge. The end of a 1D island is a type-B edge; adding a block there is similar to adding a block to a flat section of type-B step edge. Thus the rapid growth of the type-B step edge and the highly anisotropic island shapes are both results of the sticking anisotropy.

Our measured sticking anisotropy, together with our previous measurements of edge fluctuations, also provides a detailed quantitative picture of the coarsening of small islands or features on the Si(001) surface. In this case, the type-B edges of each row fluctuate. The rows on the edge of an island fluctuate fastest, and when the ends cross, the entire row disappears (this process is very clear in Java Applet or Gif Animation movie images of island fluctuations) [Pearson 1995b]. Due to the sticking anisotropy, it is then difficult to nucleate a new row, and the island shrinks. This process will hasten the demise of smaller islands at the expense of larger islands and steps.

The actual microscopic mechanism underlying these step fluctuation and growth processes is not currently known. The current models include monomer, dimer, and dimer vacancy diffusion. The low prefactor that we observed for step fluctuations indicates that a multi-atom or collective process is involved. The activation energy of 0.97 eV that we measure is larger than the 0.67 eV estimated for monomer diffusion, and smaller than the dimer dissociation energy. On the other hand, 0.97 eV is very close to the roughly 1 eV activation energy for dimer diffusion. We suggest that for typical conditions, the monomers pair up into dimers.

Further investigation of epitaxial growth's early stages has therefore involved a detailed study of single and two dimer configurations and dynamics. This has lead to our recent observation of a novel dimer diffusion mechanism that provides for crossing substrate rows thereby extending the possible diffusion pathways into two dimensions. The Stealth, or C-type dimer configuration plays a key role in this mechanism. Surprisingly, the C dimer plays a similar role in the transformation between two four-atom configurations, a transformation which we have observed to lead to epitaxial row formation. Especially at elevated temperatures where process rates are increased, a conventional STM's image rate limits its applicability to surface dynamics studies. For this reason we have implemented a tracking technique which allows us to follow individual features along their diffusive paths. This has yielded dimer diffusion and dissociation activation energies.


Jump to Hot STM Labs

Copyright 1996 by the Regents of the University of Minnesota