硅的湿化学蚀刻机理

ABSTRACT  

We review what can be said on wet chemical etching of  single crystals from the viewpoint of the science of  crystal growth. Starting point is that there are smooth  and rough crystal surfaces. The kinetics of smooth faces  is controlled by a nucleation barrier that is absent on  rough faces. The latter therefore etch faster by orders of  magnitude. The analysis of the diamond crystal structure  reveals that the { 11 1) face is the only smooth face in this  lattice - other faces might be smooth only because of  surface reconstruction. In this way we explain the  minimum of the etchrate in KOH:H20 in the <001>  direction. Two critical predictions concerning the shape  of the minimum of the etchrate close to <001> and the  transition from isotropic to anisotropic etching in  HFHN03 based solutions are tested experimentally. The  results are in agreement with the theory.  

I. INTRODUCTION  

Anisotropic wet chemical etching of single crystalline  silicon, gallium arsenide and quartz is one of the key  technologies for the fabrication of microsystems. Yet the  strong anisotropy of the etchrate in particular etching  solutions (e.g. KOH:H20, EDP, TMAH), and the  isotropy in others (e.g. HFHN03:HzO) is poorly  understood. Mostly the anisotropy of the etchrate is  related to chemical reactions on the crystal surface  oriented in different crystallographic directions. In this  respect maybe the most advanced picture has been  proposed by Seidel et al. 111. They assumed that the  complex formed by the attachment of an OH-ion to the  dangling bond, after the electron has been delivered to  the solid state, changes the back-bond energy of the  silicon atom with three back-bonds in a different way  than iin the situation when one has two OH attached to  the silicon atom with two back-bonds. The difficult point  however is that the silicon atoms have three backbonds  also in the flat (1101 face, not only in the flat { 1111 face;  therefore, etchrate and activation energy in these  crystallographic directions should be comparable in  contrast to experimental evidence.  

In this paper we add a number of new experimental  results which support the view given here. In particular,  we have looked with greater detail at the dependency of  the etchrate of silicon etched in KOH on the  crystallographic orientation close to the <loo> direction,  and we studied the transition to anisotropic silicon  etching in HFHN03:CH3COOH.  In the next section we briefly review the crystal growth  point of view of wet chemical etching. We then describe  our experiments and their results. A section of discussion  and conclusions follows.

11. THEORY  

In kinetics of crystal growth active sites for growth and  dissolution play a key role. These active sites are atoms  with as many bonds to the crystal as to the liquid (or  gaseous) environment. Such a site is called a kink site.  An atom in a kink site in a simple cubic lattice is shown in  fig. la. The heavy shaded atom has three bonds to the  crystal and three bonds to the liquid. In a dissolution situation it is commonly believed that this atom will  diffuse over the surface (fig. lb), until it either finds a  kink position again or it desorps and diffuses away from  the crystal in the liquid phase (indicated in fig. IC). In a  growth situation, an atom diffuses from the liquid to the  crystal (fig. IC), it diffuses over the crystal surface (fig.  1.b) until it is either desorped or it finds a kink site (fig.  la).

Fig1

This is very much diffeent from the {OOl) silicon face as  can be seen in fig. 3. The same operation - creating an  adatom-cavity pair - now costs no energy, because one  has to break two bonds in order to remove an atom from  the (001) face, but one gets them back by placing it back  to any position on this face.  

It is easily seen that the number of kink sites varies with  the crystallographic orientation in a very dramatic way.  The perfectly flat { 11 1) face in the diamond lattice has  no kink positions (three backbonds, one dangling bond  per atom), while on the (001) face of silicon every atom  has two back bonds and two dangling bonds - every  position is a kink position.