Site News

Become a guest author

guest postersMEMS Library is a forum for interested visitors to share ideas, discuss issues related to current work and obtained results. It is a place that provides...

Login

Banner
Anisotropic wet etching of silicon with alkaline etchants E-mail
Written by Colin   
Monday, 21 February 2011 00:00

anisotropic wet etching of silicon with alkaline solutions

Anisotropic wet etching of silicon is a commonly used bulk micromachining technique for fabrication of micromechanical structures because of its relatively simple implementation and low costs. In most applications silicon is etched chemically without using an electrical power source.

 

The reaction is given by [Allongue, 1993], [Bressers, 1995]:

 

Si+4H2O->Si(OH)4+2H2

 

The hydroxide ion does not appear explicitly in this reaction, but it is important since it catalyzes the reaction and ensures the solubility of the Si(OH)4 product. The etch rate depends on the temperature, etchant composition and the crystallographic orientation of the surface. Note that since this reaction involves a chemical reaction, its rate is generally not affected when the potential of the semiconductor is changed by an external power source.

 

Anisotropy

All alkaline etchants etch silicon anisotropically, which means that the etch rate is dependant on the crystal orientation. The fastest etching plane is usually the (100)-plane, while the (111)-plane has the lowest etch rate. Dissolution is limited by the chemical reaction kinetics. For isotropic etchants dissolution is limited by diffusion of reacting species to the semiconductor. The anisotropy can be clearly observed from surface profiles. If a (100)-oriented Si wafer is patterned with an etch mask, the etch front is bounded by the (111)-planes. The final shape of the etched pit is a V-groove or an inverted pyramid. The sides are aligned to the (110)-planes independently of the initial shape of the etch mask. 

 

final shape of etched pits in (100)-siliconFigure: A SEM photo and schematic of the final shape of etched pits in (100)-silicon. Note that the initial shape of the etch mask is not a square. After some time in a TMAH solution at 80oC the etched shape is bounded by (111)-facets.

 

Because of the anisotropy, structures in (100)-oriented silicon are always limited in their smallest dimensions. The angle between the <100> and <111> direction is 54.74o. Therefore, an etched pit with depth A has a minimum width of 2A/tan(54.74o). 

 

etch mask required for an etched pit with depth A

Figure: The minimum width of an etch mask required for an etched pit with depth A.

 

The low aspect ratio of the etched holes may be increased by using (110)-oriented silicon wafers. The limiting (111)-planes are perpendicular to the (110)-surface. Therefore, the angle between the sidewall of an etched pit and the surface is 90o [Kendall, 1979]. However, (110)-oriented wafers are not standard in todays IC processes and (100)-oriented wafers are usually used for micromachining purposes.

 

There are many theories which try to explain the anisotropic etching of silicon (for example [Elwenspoek, 1993-1996], [Kendall, 1990], [Palik, 1985], [Seide, 1990]) but the exact mechanism is still unclear. One theory explains the anisotropy in terms of the number of dangling bonds for different surface atoms. A (111) surface atom is bonded to three atoms in the crystal lattice and has only one dangling bond while a (100) surface atom is bonded to two atoms in the crystal and has two dangling bonds. Therefore, (100) surface are more reactive than (111) planes. This gives only a qualitative understanding of the anisotropic etching and cannot explain the anisotropy ratio of typically more than 100 [Bressers, 1995].

 

Convex structures

The anisotropic nature of the etching means that the formation of concave corners is quite straightforward. However, convex corners present complications. When a free-standing structures such as the seismic mass of a bulk-micromachined accelerometer are made, the exposure of a convex corner is inevitable. At a convex corners many different crystal planes are exposed to the etchant at the same time, resulting in considerable undercutting. For most micromachining applications, this is undesirable because it results in a badly defined shape (and mass). Different corner compensation masks have been realized to produce square corners (reported in for example: [Enoksson, 1997], [van Kampen, 1995], [Mayer, 1990]). The simplest structure consists of a beam oriented along the <100>-direction. The width of this beam should be twice the etched depth. Th disadvantage is that the beam has to be rather long since it is undercut at the end at a considerable rate. The (411)-plane is the dominant plane causing the undercutting. More area-efficient structures have been designed, consisting of (110)- and (100)- oriented complex patterns.

 

Etch mask

The part of the wafer which is exposed to the solution is defined by an etch mask. The etchant should have a high selectivity (etch-rate ratio) towards the etch-mask material. Commonly used materials are silicon oxide and silicon nitride. Inert metals, i.e. gold, platinum, have also been used for this purpose [Petersen, 1982].

 

Commonly used etchants

The most commonly used etchants are solutions of ethylenediamine pyrocatechol (EDP) [Finne, 1967], hydrazine [Declercq, 1975], [Mehregany, 1988], KOH [Bean, 1978], [Kovacs, 1998], [Lee, 1969] and tetramethylammonium hydroxide (TMAH) [Merlos, 1993], [Tabata, 1992]. The first two are highly toxic and require special care during handling. They are less popular than the last two. The KOH-based etchant is non-toxic and has a relatively high etch rate and a high anisotropy ratio (anisotropic ratio is the (100)/(111) etch-rate ratio). However, it is not compatible with CMOS-processes due to mobile K+ ion contamination. After etching, it is therefore impossible to perform thermal steps. In recent years, the TMAH etchant has received more interest since it is fully CMOS compatible The tetramethylammonium ion is rather large and does not diffuse into the silicon lattice. An interesting feature is the possibility of passivating aluminum metallization by adding silicon or silicic acid [Sarro, 1998]. Another advantage over KOH solutions is the high selectivity towards silicon oxide. This means that relatively thin layers of silicon oxide can be used as an etch mask. TMAH solutions have a lower etch rate than KOH solutions (typically a factor of 2) and they also have a lower anisotropy ratio (typically a factor of 4). As a consequence, it is virtually impossible to produce well-defined convex corners with existing corner compensation structures. An effective corner compensation structure is not possible because its dimensions tend to become too large for device applications. Furthermore, it is more expensive to dispose of organic solutions (such as TMAH) than non-organic solutions (such as KOH).

 

Etch rate of silicon and masking layers

In this article, TMAH solutions are mainly used. Hence, etch-rate data are only shown for this etchant. 

 

The etch rate of (100)-oriented Si as a function of temperature and TMAH solution concentrationFigure: The etch rate of (100)-oriented Si as a function of temperature and TMAH solution concentration (5% (a), 15% (b), 20% (c), and 25% (d)).

 

In this figure the measured etch rate for different TMAH solutions at temperatures between 65oC and 95oC is shown [Sarro, 1998]. Note that a higher concentration results in a lower etch rate. A higher temperature results in a higher etch rate. The etch rate of a KOH solution with a typical concentration of 33% at a typical etching temperature of 80oC is 120 µm/hr [Kwa, 1995]. This is 4 times higher than the etch rate of a typical TMAH solution of 25% at the same temperature.

 

The etch rate of different masking layers as a function of temperature in TMAH solutionFigure: The etch rate of different masking layers (PECVD oxide (a), PECVD silicon nitride (b), thermal oxide (c) and LPCVD silicon nitride (d)) as a function of temperature in a 5% and 25% TMAH solution. The solid curves are guide to the eye.

 

In this figure the etch rate of different etch-mask layers (PECVD silicon oxide, PECVD silicon nitride, thermally grown silicon oxide and LPCVD silicon nitride) is shown [Sarro, 1998]. The trends are the same as for silicon etching; the lower the concentration, the higher the etch rate and a higher temperature leads to a higher etch rate. Note that all layers have a very low etch rate and are therefore suitable as materials for etch masks. For comparison, thermally grown silicon oxide has an etch rate of 12 nm/hr in a 25% TMAH solution at 80oC and an etch rate of 1000 nm/hr in a 33% KOH solution at the same temperature [Kwa, 1995]. This demonstrates the high selectivity towards silicon oxide of TMAH solutions. The selectivity is 2500 for TMAH solutions and it is 120 for KOH solutions.

 

Surface morphology

It is important for many sensors applications that the etching of the semiconductor results in a smooth surface [Kwa, 1995]. The surface morphology is dependent on the etching conditions. 

 

etched surface in a 5% and a 25% TMAH solutionFigure: SEM pictures of an etched surface in a 5% (left image) and a 25% (right image) TMAH solution at 80oC.

 

In this figure two surfaces are shown, which were etched in different TMAH solutions. Solutions with a high concentration (>20%) tend to result in a smoother finish than solutions with low concentration. Since high concentration also results in a lower etch rate, a trade-off between etch rate and surface roughness has to be made. For TMAH solutions the temperature proved to be unimportant for the surface morphology. It can be seen that the surface roughness consists of pyramidal hillocks. The exact mechanism of hillock formation is not clear. One theory is that hillocks are due to the formation of hydrogen gas. The bubbles adhere to the surface and thereby create local etch masks [Campbell, 1995]. Addition of oxygen to the solution was reported to reduce hillock formation because hydrogen reacted with oxygen to form water. Another theory attributes hillock formation to limited silicon deposition from silane (SiH4) during the etching process. SiH4 is only stable at low OH- concentrations explaining the concentration dependence of the surface morphology [Bressers, 1996]. It was observed that etching under anodic polarization or with added oxidizing agents results in hillock suppression and a smooth surface. Since in this case the hydrogen generation rate is unchanged, Bressers concluded that hillock formation cannot be due to hydrogen bubbles.

 

If you liked this article, you may also like the book: Galvanic Etching of Silicon: For Fabrication of Micromechanical Structures by the same author.

 

You can also guest post on MEMS Library website or leave a comment if you want to share your thoughts. You may also want to visit our forum.

 
Joomla Templates: by JoomlaShack