Ultramicroscopy 42-44 (1992) 838 844 North-Holland STM studies of Si(100)-2 × 1 oxidation: defect chemistry and Si ejection Ph. Avouris and D. Cahill T.J. Watson Research Center, IBM Research Dil~ision, Yorktown Heights, NY 10598, USA Received 12 August 1991 The initial stages of oxidation of the Si(100)-2× 1 surface have been studied using STM and STS. In contrast to the Si(111)-7 x 7 surface, which has a metallic density of states (DOS), the dangling bonds of the Si dimers on the 2 × 1 surface are paired, leading to the formation of a surface gap and a vanishing DOS near E F. We observe a reduced reactivity of Si dimers towards 0 2 compared to that of Si adatoms on Si(111)-7 × 7, consistent with the reduced DOS near E F. Defects on the Si(100)-2× 1 surface which have a metallic DOS dominate the reactivity towards 0 2 in the early stages of the reaction. Among the new sites generated by the exposure to 0 2 are 1.4 A high bumps on top of the surface. Upon annealing of the O2-exposed surface or upon 0 2 exposure at an elevated temperature these bumps form elongated islands. Evidence is presented suggesting that the bumps and islands likely are due to silicon ejected to the surface by the oxidation reaction. Possible mechanisms and implications are discussed. 1. Introduction The oxidation of silicon is one of the most important processes in the chemistry of electronic materials, and as such it has received consider- able attention [1]. STM studies of the initial stages of Si(111)-7 x 7 oxidation have shown that the reaction is very site-selective, with adatom sites being particularly reactive [2-6]. The Si(111)-7 x 7 surface exemplifies a metallic character, i.e. it has a finite density-of-states (DOS) at the Fermi en- ergy (E v) with the Fermi level pinned by the adatom dangling-bond surface states. Among adatoms, the corner-sites on the faulted half of the unit-cell are the most reactive. We have inter- preted this reaction selectivity as reflecting the selectivity of the first step of the oxidation pro- cess which involves a substrate-to-O 2 charge- transfer (CT) process [6]. This CT process is optimized at sites with high DOS near E v. Cor- ner-adatoms on the faulted half of the unit-cell have indeed the highest DOS near E F [7]. Fur- ther support for the importance of the CT pro- cess came from experiments in which the occupa- tion of the dangling-bonds was altered by ~- dopant layers [8]. Here we focus on the Si(100)-2 × 1 surface which, unlike the Si(lll)-7 × 7 surface, has semi- conducting properties (surface band-gap ~ 0.6 eV) [9]. The 2 × 1 reconstruction arises when Si surface atoms pair to create Si dimers. These dimers can either be symmetric or asymmetric (buckled). Theoretical evidence supports the view that the buckled form has the lowest energy. However, at room temperature the two configu- rations of a buckled dimer interconvert very rapidly so that these interconverting dimers ap- pear symmetric in STM images. Buckling is also associated with an intra-dimer CT process in which charge is transferred from the "down" atom to the "up" atom [10]. Since the 2× 1 surface is semiconducting with a vanishing DOS near E v, we would expect a reduced reactivity towards O 2 compared to the Si(111)-7 x 7 sur- face. On the other hand, it is known that the E F of the 2 × 1 surface is pinned near the center of the gap, implying the presence of electronic states due to minority defect-sites [11]. Defect-induced 0304-3991/92/$05.00 © 1992 - Elsevier Science Publishers B.V. All rights reserved