硅碱性蚀刻中的绝对蚀刻速率

The absolute etch rate of silicon (1 1 1) during wet chemical etching in aqueous KOH solution has been investigated with optical interferometry, using masked samples. The etch rate is constant at 0.62 ± 0.07 m/h and independent of alkaline concentration for 1–5 M KOH solutions at 60 ◦C. Only at lower alkaline concentrations, the etch rate decreases. Adding isopropanol slightly lowers the absolute etch rate. The activation energy of the etching reaction is 0.61 ± 0.03 eV in standard KOH solutions and 0.62 ± 0.03 eV with 1 M isopropanol added to the solution. This indicates that the reaction is determined by reaction kinetics and not by transport limitations. In all cases the surfaces are covered by shallow etch pits, not related to defects in the crystal. This implies that the actual factor that determines the etch rate is the 2D nucleation of new vacancy islands at the bottom of these pits. This process is likely catalyzed by a local accumulation of reaction products, which preferentially occurs near the mask edges.

1. Introduction

Anisotropic wet chemical etching of silicon is an extensivelyused process and an essential step in micro-electromechanical sys-tems (MEMS) manufacturing [1-5]. The process is based on thesignificant difference in etch rate between the f1 0 01 and (1 1 0faces versus the (1 1 11 faces of the silicon crystal [6-9]. As the Si-(1 1 11 plane is the slowest etching plane in KOH solution [10-14).wet chemical anisotropic etching is normally applied to masked(1 1 0) and (1 0 0) wafers, resulting in complex three-dimensionalstructures bounded by f1 1 11 planes. The difference in etch ratebetween the slowest f1 1 11 faces and the other two crystallo-graphic orientations of silicon during etching is also known as theanisotropy ratio, which is a major factor in determining MEMSquality. In contrast to the fast etching (1 0 0) and (1 1 0) planes,knowledge of the absolute etch rates of "exactly” oriented (1 1 1)faces is still lacking. In addition to the processing of (1 00) and(1 1 0) Si wafers, knowledge of Si-f1 1 11 etching is also importantfor the processing of (1 1 1) oriented wafers as a route to obtainsmooth surfaces and different kinds of free-standing microstruc-tures[10,15-17].

A second aspect of the wet chemical etching of Si-(1 1 1) usingalkaline etchants is its mechanism, which totally differs from thatof the other silicon crystal orientations. Whereas the KOH etchingof the other, eitherexact orvicinal, Si-faces (h kl)involves step flowstep bunching, pyramid formation or kinetic roughening (18], the etching of the exact f1 1 11 plane proceeds by a repeated 2D nucle-ation of vacancy islands, which then expand by step flow, leadingto the development of shallow, point bottomed etch pits (19,20].Such a process of step generation can be induced by the stressfields around dislocations or stacking faults ending on the (1 1 1)face (21 ,22], but in our study the pits are not related to defects andthe 2D nucleation ofsteps is likely induced by a local accumulationofsilicate reaction products (19,20.However, quantitative data onthe kinetics of this etching process is not yet available.

Methods based on the etching of wagon wheel patterns[11,23-25] or of semiconductor (hemi)spheres [18,26] have theadvantage of yielding dissolution rates for a whole range of orientations in a single experiment. Unfortunately, these otherwisepowerful approaches do not provide information on the etch rateof the exact (1 1 11 face, as this is readily etched away and replacedby vicinal faces close to f1 1 11. This follows from the kinematicwave theory of crystal dissolution, according to which the slow-est etching face on a curved surface will disappear during etching(27]. Examination of cleaved silicon samples with etch structuresboundedby f1 1 11 faces, usingopticalmicroscopyorscanningelec-tron microscopy (SEM), can provide information on absolute etchrates of (1 1 1) surfaces. However,this approach is quite tedious andthe etch rate might be affected by the presence of the mask con-tacting the (1 1 1) faces. Another experimentally very challengingway of determining the etch rates is by using in situ STM obser-vation of the Si-(1 1 1) surfaces during etching [28]. Although thismethod provides detailed information on the etching process at ananometer scale, the measured etch rates are not representative forthe larger scales used in MEMS technology. Moreover, this methodcan only be used for relatively low alkaline concentrations. From the above it is clear that, although measuring absolute etch rates ofSi-(1 1 11 accurately is highly relevant in MEMS technology, it hasbeen proven not to be a trivial exercise.A large number of studies have been carried out on the etchrate and surface morphology of the (1 00), (1 1 01 and other fh klfaces of silicon etched in alkaline solutions, with and without addi-tives, e.g. see Refs. [29-31]. However, the (1 1 1) faces receivedless attention, as the commonly used methods for the measure-ment of etch rates are not suited for the slowly etching f1 1 1)faces.

In this study we determine accurate f1 1 11 etch rates by etch-ing partially masked Si-f1 1 11 wafers followed by measurementof the resulting height differences between the masked and freesurface areas using phase shifting interferometry (PSl). By thisstraightforward approach the problems encountered by the meth-ods mentioned above are avoided. The absolute etch rates aremeasured with and without isopropanol (IPA) additive and asa function of time, the diameter of the non-masked areas, KOHconcentration and temperature. The activation energy of the wetchemical etch reaction is determined from the temperature seriesComplementary to the etch rate measurements, the morphology ofthe etched surfaces is inspected using PSl and differential interfer-ence contrast microscopy (DICM).These morphologic observationsprovide important information on the influence of mask edges onthe surface profile and the measured etch rates.