On the Mechanism of Anisotropic Etching of Silicon

A new model is proposed that explains the anisotropy of the etch rate of single crystalline silicon in certain etchants. It is inspired from theories of crystal growth. We assume that the ( l l l ) face is flat on an atomic scale. Then the etch rate should be governed by a nucleation barrier of one atomic layer deep cavities. The origin of the nucleation barrier is that the formation of a too small cavity increases the free energy of the system due to the step-free energy. The step-free energy and the undersaturation governs the activation energy of the etch rate. Having the largest step-free energy, the (11 D-face etches the slowest. The model explains qualitatively why the etching is isotropic in certain etchants and anisotropic in others. In papers on anisotropic etching of silicon many authors are puzzled by the strong anisotropy of the etch rate (see the discussion in Ref. 1 and references therein). In some etchants [the most well-known are aqueous solution of KOH and ethylene diamine pyrocatechol (EDP)], the etch rate of the (111) planes is much smaller than in the other crystallographic directions. Depending on concentration of the etchant and the temperature the (111) direction etches slower than the other ones by a factor 100 and more. This fact is widely being used to micromachine t iny mechanical devices. 2 Etch rate and temperature dependence (described by an Arrhenius law) are anisotropic: slow etching goes with large activation energies. Recently, papers appeared which deal with the mechanism of anisotropic etching of silicon. 1'3'~ The basic idea is that in the (111)-plane of silicon there is only one dangling bond per silicon atom. Therefore there are three bonds to break for dissolution, while other planes [except the (110)] have more dangling bonds, accordingly a smaller number of bonds must be broken. a Present address: MESA-Research Institute, University of Twente, NL-7500 AE Enschede, The Netherlands. The anisotropy cannot be understood by this fact alone because in the dissolution process transferring a silicon atom from the solid to a molecule dissolved in the liquid the backbonds are not broken simultaneously. Seidel et al. 1 have proposed that in the rate-determining step the electronic state of the complexed silicon depends on the number of backbonds. The anisotropy of the activation energy and of the etch rate itself could be explained by such a model. A serious problem however arises if one realizes that on the (110) plane there are also three backbonds, but the etch rate is large [comparable to (100)] and the activation energy corresponds to a fast etching direction. The activation energy of the etch rate in anisotropic etching solutions depends on the etching system. This dependence is attributed to diffusion that plays a greater role in EDP than in KOH based solutions. However one should expect that at least the slow etch rates are not due to diffusion in the solution but to surface reactions, and diffusion should have a minor effect on the activation energy. There are other etchants in which silicon etches isotropically (aqueous solutions of HF:HNO~). Here, the chemical reaction is different; HNO~ acts as an oxidizer and the siltDownloaded 11 Nov 2008 to 130.89.19.116. Redistribution subject to ASCE license or copyright; see http://www.ecsdl.org/terms_use.jsp 2076 J. Electrochem. Soc., Vol. 140, No. 7, July 1993 9 The Electrochemical Society, Inc. con oxide is dissolved subsequently by HF. No reasons are given in the literature why this system etches isotropically. A simple explanation may be available that could clarify the points addressed above, and that opens a route for new research. I propose it is the physical state (being atomically flat or rough) of the various surfaces that are finally responsible for the anisotropy of etch rates and activa-