Motivation—Exposure of the Si(001) surface to low pressures of molecular oxygen at temperatures exceeding 800C produces volatile SiO leading to a net removal of Si from the surface. Despite its importance as a potential procedure for cleaning Si, a comprehensive understanding of the kinetic processes underlying this etching phenomenon is still lacking. Current models are based on experiments in which the state of the surface during etching is unknown. We have addressed this problem using the unique imaging capabilities of the low energy electron microscope (LEEM).

Accomplishment—We have combined low energy electron microscopy with kinetic Monte Carlo  simulation and diffusion equation analysis to identify the mechanism and determine the energetics by which molecular oxygen etches Si(001) surfaces. Our results show that oxygen adsorption leads to vacancy creation on the terraces followed by diffusion of the vacancies to single-atom high steps. This mechanism allows us to model the etching process as a temperature-dependent flux of vacancies onto the surface and apply standard nucleation and growth theory to extract kinetic parameters.  On typical Si(001) terraces (~0.5mm wide or less), vacancy creation/diffusion leads to step flow similar in appearance to ordinary sublimation. However, on large terraces (~5mm or larger), vacancy concentration during etching can reach levels sufficient high to nucleate vacancy islands.  LEEM measurements of the growth rate of these islands provides quantitative information on the vacancy flux, and, in turn, the rate of etching.   Figure 1 shows a sequence of LEEM images illustrating the time evolution of the growth of 25 vacancy islands on an 11.1 mm wide Si(001) terrace.  The large terraces were prepared by a novel procedure involving the formation of square pits in the Si wafer and subsequent etching of the pit bottoms by oxygen exposure at elevated temperatures. The key observation from these measurements is that the growth rate of a given island depends on its location with respect to neighboring islands (e.g., island A grows faster than C in Fig. 1). Specifically, the growth rate is defined by a capture area within which vacancies are supplied to the island.  The observed growth rates are faithfully reproduced by kinetic Monte Carlo Simulations and a diffusion equation analysis (Fig 1). The validity of the diffusion equation analysis is established by the ability to fit the simultaneous growth of up to 25 islands with just one adjustable parameter (the vacancy flux).  A plot of the flux (effective etch rate) vs. inverse temperature for three different oxygen pressures is shown in Figure 2.  From these data we determine that the activation energy for Si etching is 2.0 ± 0.4 eV.  This result is significantly lower (by a factor of 2) than that measured in molecular beam experiments (for a review see: T. Engel, Surf. Sci. Rep. 18, 91 (1993)).

Significance—This large discrepancy in measured activation energies shows that etching of Si(001) surfaces by molecular oxygen is  much more complicated that can be inferred from measurements of desorption products alone. Real-space imaging of surfaces that are both well-defined and reproducible along with the ability to observe dynamic processes under steady state conditions avoids the uncertainties inherent in past investigations leading to a new understanding of the etching process. 
 
 

Fig. 1 (Top) LEEM images showing the growth of vacancy islands on an 11.1mm wide terrace on Si(001).  The sample temperature is 815C and the oxygen pressure is 7.8x10-8 Torr. (Bottom) Simulation of island growth with diffusion equation analysis.	Fig. 2 Arrhenius plots of the effective flux of vacancies for three oxygen pressures.  The activation energy of Si removal is 2.0± 0.4 eV.

Publication: 
J. B. Hannon, M. C. Bartelt, N. C. Bartelt and G. L. Kellogg, Phys Rev. Lett. 81, 4676 (1998).