Etching mechanism and surface structure

Hydrogenated silicon surfaces, simply prepared by dipping in HF, are the prototype of ideally passivated semiconductor surfaces [5]. They are electronically inert with the lowest recombination velocity ever reported [6] and are chemically resistant to air oxidation [5]. The pH of HF or NH4F solutions has a great influence on the microstructure of etched surfaces [7-9]. The images in Fig. 2 are typical large-scale views of n-Si(l 11) surfaces taken in 1 M NH4F solutions of pH 4 and 8 [9], Atomic resolution in fluoride solutions has been reported by others [10, 11]. As described in Section 1 the silicon substrate was cathodically biased for imaging.

By showing that flat terraces appear at elevated pH Fig. 2 illustrates the pH dependence of the etching mechanism [9, 12]. There are indeed two routes, one electrochemical and the other chemical (no free-charge carrier exchanged). At pH 4 the first route dominates and the surface is rough because the process is isotropic. As the pH increases, the rate of the chemical reaction increases leaving (111) terraces atomically flat because the process is highly anisotropic for steric reasons [9]. At alkaline pH, etching is even more anisotropic with (111) planes which etch more than 100 times slower than (110) and (100) planes [13]. This unique property has long been used for micromachining Si wafers. Most of our STM work was performed in these solutions where the dynamics of the process has been analyzed in great detail.

Figure 3 is a high-resolution image of a Si(l 11) in NaOH. It shows that the surface has a great chemical homogeneity and that the STM may have some chemical sensitivity. With the exception of a few =Si-OH [14] visible (white spots), which represent only 0.1 % of atomic sites, the surface is indeed (lxl) H-Si(lll). That =Si-OH appear as protrusions surrounded by a dark ring comes from an increased local density of states and a three-dimension (3D) potential distribution induced by the large

Fig. 2. 1000x1000 Â2 images of n-Si(lll) surfaces in ammonium fluoride solutions of pH 4 (a) and 8 (b). After [9].
Fig. 3. 103x 70 A2 STM image showing the surface of n-Si(l 11) inNaOH. The surface is (lxl)-H-terminated except for a few =Si-OH groups (after [14]).

electronegativity difference between Si and the OH group. The H-termination, though energetically less favorable when comparing the energies of Si-H and Si-O bonds, may be explained in terms of kinetics.

Fig. 4. Time sequences of STM images showing Si(l 11) etching in NaOH. (a): Cathodic bias (90s between images; frames are 1260x970 A2), (b): At the rest potential for 3 s using a special procedure is used (frames are 1400x1400 A2) (after [12]).

The substitution Si-H —» Si-OH, which is the very initial step of the dissolution, is indeed rate-determining in NH4F at all pH values [9] as well as in NaOH. The H-termination has been proved by in-situ FUR [15, 16].

The generation mechanism and the stability of Si-OH surface groups both depend on the surface site [9]. At step edges, hydroxyl groups may easily bind chemically on step di- and monohydrides, forming =Si-HOH or =Si-OH entities which are rapidly removed by chemical breaking of the Si-Si back bonds. This leads to the lateral migration of atomic steps observed at cathodic bias (Fig. 4a). Vertical =Si-OH group formation stems by contrast exclusively from the electrochemical process and therefore occurs only at potentials close to and positive of the rest bias. Vertical Si-OH groups are also more stable than step =Si-OH and not every group leads to the initiation of an etch pit. According to a simulation (Monte-Carlo method) it can indeed be inferred that the formation of a nano cluster of suboxide is necessary prior to the nucleation of an etch pit [17]. The triangular etch pits observed in Fig. 4(b) have been created at the rest bias but imaged at cathodic bias, using a special procedure [12].

The time sequences presented in Fig. 4 have been used to measure the local etch rates, i.e. on a 0.01 pm2 scale. Etch rates thus derived are 0.2 and 8 A/min"1, respectively, in sequences (a) and (b) of Fig. 4, values which are surprisingly in very good agreement with macroscopic determinations, performed on a mm2 scale, that is on a surface area 108 times larger [12]. More recently we have been able to determine the reaction rates on the atomic scale from the direct comparison of in-situ STM sequences and Monte-Carlo simulations.

0 0

Post a comment