Anisotropic Etching of Crystalline Silicon in Alkaline Solutions

The anisotropic etching behavior of single-crystal silicon and the behavior of SiO2 and Si3N4 in an ethylenediaminebased solution as well as in aqueous KOH, NaOH, and LiOH were studied. The crystal planes bounding the etch front and their etch rates were determined as a function of temperature, crystal orientation, and etchant composition. A correlation was found between the etch rates and their activation energies, with slowly etching crystal surfaces exhibiting higher activation energies and vice versa. For highly concentrated KOH solutions, a decrease of the etch rate with the fourth power of the water concentration was observed. Based on these results, an electrochemical model is proposed, describing the anisotropic etching behavior of silicon in all alkaline solutions. In an oxidation step, four hydroxide ions react with one surface silicon atom, leading to the injection of four electrons into the conduction band. These electrons stay localized near the crystal surface due to the presence of a space charge layer. The reaction is accompanied by the breaking of the backbonds, which requires the thermal excitation of the respective surface state electrons into the conduction band. This step is considered to be rate limiting. In a reduction step, the injected electrons react with water molecules to form new hydroxide ions and hydrogen. It is assumed that these hydroxide ions generated at the silicon surface are consumed in the oxidation reaction rather than those from the bulk electrolyte, since the latter are kept away from the crystal by the repellent force of the negative surface charge. According to this model, monosilicic acid Si(OH)4 is formed as the primary dissolution product in all anisotropic silicon etchants. The anisotropic behavior is due to small differences of the energy levels of the backbond surface states as a function of the crystal orientation. Anisotropic etchants for crystalline silicon have been known for a long time (1-3). Their first applications included the etching of V-grooves on <100> silicon or Ugrooves on <110> silicon, in order to fabricate MOS transistors for high power and high current densities (4). Increasing attention has been paid to this etching technology, after recognizing its unique capabilities for micromachining three-dimensional structures (5-9). Due to the strong dependence of the etch rate on crystal direction and on dopant concentration, a large variety of silicon structures can be fabricated in a highly controllable and reproducible manner. Typical structures include thin membranes, deep and narrow grooves, and cantilevers with single or double sided suspension. Important fields of application include the fabrication of passive mechanical elements, sensors, and actuators, as well as micro-optical components (8, 10). Among the best known examples are sensors for pressure (8), acceleration (11), and flow (12), as well as ink jet nozzles (13), connectors for optical waveguides (14), and major components of a gas chromatograph (15). All anisotropic etchants are aqueous alkaline solutions, where the main component can be either organic or inorganic. The first organic system was proposed in 1962 and consisted of hydrazine (N2H4) and water with the addition of pyrocatechol (C6H4(OH)2) (16). It was shown that pyrocatechol is not a necessary component, and might well be omitted (2). Experiments were made with iso-2-propyl alcohol as a third component, which was shown to act as a moderator (3). In a later work, hydrazine was substituted by ethylenediamine (NH2(CH2)~NH2), which is more stable and less toxic than the former (2). Purely inorganic aqueous solutions of KOH and NaOH have been known to etch silicon anisotropically for a long time (1). A different system with improved etching behavior was obtained by the addition of isopropyl alcohol (17). J. Electrochem. Soc., Vol. 137, No. 11, November 1990 9 The Electrochemical Society, Inc. 3613 In general, all aqueous solutions containing hydroxides of other alkali metals, like LiOH, and CsOH (18) perform in a similar manner. Aqueous solutions of ammonium hydroxide (NH4OH), were also reported to etch anisotropically (19). The same type of solution with the addition of H202 is frequently employed for the cleaning of silicon wafers (20). More complicated derivatives of ammonium hydroxide, e.g., so-called quaternary ammonium hydroxides like tetramethyl ammonium hydroxide (N(CH3)4OH) and choline ((CH3)3N(CH2CH2OH)OH) can also be used as anisotropic etchants (21). General Considerations Solutions consisting of ethylenediamine, water, and pyrocatechol (EDP) are among the most widely employed. Many essential results concerning the composition of this system and its crystal orientation dependence were reported in the work done by Finne and Klein (2). They found that pyrocatechol can be omitted, so that in its most primitive form, the etchant solely consists of ethylenediamine and water. With nonaqueous ethylenediamine no etching was achieved, indicating that water is an active and necessary component. The maximal etch rate occurred at an ethylenediamine to water molar ratio of approximately 1:2. By the addition of pyrocatechol, a strong increase of the etch rate by a factor of three was obtained, saturating at a concentration of approximately 4 mole percent (m/o). For the three main silicon crystal orientations <100>, <110>, and <111>, Finne and Klein found an etch rate ratio of 17:10:1. The effects of further additives were studied by Reisman et al. (22). They found that when exposing the EDP solution to oxygen, 1,4-benzoquinone and other products are formed, leading to an increase of the etch rate and a darkening of the solution. This effect can be avoided by continuously purging the etching apparatus with an inert gas (2, 22). They also found that trace quantities of pyrazine (C4H4N2) lead to an increase of the etch rate. However, similar to pyrocatechol, this effect nearly saturates at a concentration of 3g pyrazine per liter ethylenediamine. Since commercial ly available ethylenediamine usually contains an unknown trace amount of pyrazine, Reisman et al. (22) proposed to intentionally add enough pyrazine to the solution so that the saturation level is reached. For the ethylenediamine-water-pyrocatechol system several recipes were proposed. Reisman et al. developed two specific solutions optimized for use where either a high etch rate is required ("F"), or where slower etch rates and/or lower temperatures are desired ("S") (22). Their specific composit ions are listed in Table I, together with a recipe used by Finne and Klein ("T") (2) and another one proposed by Bassous ("B") (23). For both solutions suggested by Reisman et al. (22) an activation energy of 0.36 eV on <100> silicon was found, which increased to 0.47 eV when no pyrazine was added. Furthermore, they determined an anlsotropy ratio for the <100>/<111> silicon etch rates of 19 and 13.5 with and without pyrazine, respectively. Finne and Klein were the first authors to publish reaction equations for the etching process (2). Based on a chemical analysis of the reaction products and on the observation that hydrogen evolves during etching in a stoichiometric ratio of approximately 2 H2/Si, they proposed an oxidation-reduction step with hydroxide ions and water reacting with the silicon surface, followed by a chelation stage involving pyrocatechol Si + 2(OH)+ 4H~O -~ Si(OH)6-+ 2H2 [1] Si(OH)6+ 3C6H4(OH)2---> Si(O2C6H03-+ 6H20 [2] They assumed the chelation to be the slow step, unless pyrocatechol was added at a concentration exceeding 4 mole percent (m/o). In that case they considered the oxidation reaction to be rate limiting. Pyrocatechol was assumed to act mainly as an agent to increase the solubility of the silicon compound, thus increasing the reaction rate. The above ment ioned oxidation-reduction equation was used by several authors in later publications without Table I. Composition of different EDP solutions published by Finne and Klein (2), Reisman et al. (22), and Bassous (23) Type S (22) F (22) B (23) T (2) Water ml 133 320 320 470 ED 1 1.0 1.0 1.0 1.0 Pyrocatechol g 160 320 160 176 Pyrazine g 6 6 major modifications (22, 24). The gross reaction proposed in this equation does not provide an obvious explanation for the anisotropic behavior of the etchant. For this purpose it must be broken up into its fundamental reaction steps. It was noticed by several workers that residues might appear on the silicon surface. The occurrence of this phenomenon depends on the composit ion of the solution and on its saturation level with silicon. For EDP solutions, Reisman et al. (22) have found that this tendency increases with the water content of the solution. With respect to the aging of the solution, quantitative results were given by Wu et al. (25). They found that in a one liter solution (type F and type B) at a temperature of 100~ a max imum amount of 10g silicon can be etched without producing solid residues. This value was slightly higher for the F etchant which was attributed to its larger content of pyrocatechol. A chemical analysis of the residues showed that they consisted mainly of SiO~ with additional trace amounts of reaction products. The work of Abu-Zeid et al. (26) showed that, for ethylenediamine-based solutions, the silicon etch rate can be increased considerably by stirring. They also showed that the etch rate depends on the effective silicon area being exposed to the solution and its geometry. An increase can be observed, when the area of the active region gets smaller. These results indicate that diffusion processes influence the silicon dissolution rate considerably. For hydrazine water solutions a very similar behavior to the one observed in EDP solutions was found (3, 27, 28). At a temperature of 118~ an etch rate ratio of 16:9:1 for the {100}:{110}:{111} planes was determined, which is comparable to EDP (28). When underetching convex corners, (211) was identified as an etch bordering plane with a very high etch rate (3, 27). Among the inorganic solutions, the one most frequently used is based on KOH. The first detailed study on a ternary mixture of KOH, water, and isopropyl alcohol was given by Price in 1973 (17). His major observations were the following: the maximal etch rate occurred at a KOH concentration of 10-15 weight percent (w/o) when no alcohol was added, and around 30% KOH with alcohol. In general, the addition of isopropyl alcohol leads to a decrease of the etch rate. On <100> silicon the activation energy was found to be between 0.52 and 0.69 eV. He found no effect of stirring on the etch rate, indicating that the reaction is not diffusion limited. Under opt imum conditions, Price observed an etch rate ratio of 35:1 for the {100}/{111} crystal planes (17). A much higher anisotropy ratio of up to 500:1 for the <110> to <111> etch rates in a highly concentrated 55% KOH solution was reported by Kendall (29, 30). Further data on the etch rates of <110> and <111> silicon as well as SiO2 were given by Clark and Edell (31). For KOH solutions with a concentration between 9 and 54 w/o, they found the following ranges of activation energies: 0.6-0.8 eV for <110>, 0.4-0.9 eV for <111>, and 0.8-1.0 eV for SIO2. In a work done by Palik et al. (32) the etching process of KOH on silicon was monitored in s i tu by Raman spectroscopy. From these experiments the main reaction species was determined to be OH-. They proposed SiO2 (OH)z-to be the primary etching product with subsequent polymerization. The following overall gross reaction was suggested by them (33) Si + 2H~O + 2 OH--> Si(OH)202-+ 2H2 [3] From experiments done with isopropyl alcohol added to the KOH solution, they conclude that the alcohol does not 3614 J. Electrochem. Soc., Vol. 137, No. 11, November 1990 9 The Electrochemical Society, Inc. participate chemically in the reaction. In a later work (34), Palik et al. included energy level considerations, stating that the etching reaction transfers an electron from OHinto the silicon surface bond and then back to the etch products. A more detailed break down of the reaction equations, including the transfer of charge was suggested by Raley et al. (24). They assumed that four electrons are injected into the conduction band by an initial oxidation reaction, which are later consumed in a reduction step Si + 2 OH--> Si(OH)2 ++ + 4e[4] Si(OH)2 ++ + 4e+ 4H20 ~ Si(OH)6-+ 2H2 [5] In the literature published to date several attempts were undertaken to explain the anisotropic behavior of these etchants. Price indicated that a correlation between the available bond density of different crystal planes and the etch rate could exist (17). However, it is difficult to explain etch rate ratios of about 100:1 when the bond density only varies by a factor of two. Another proposal was made by Kendall who argues that {111} planes get oxidized more rapidly than others and therefore could be covered with a thin oxide layer immediately after immersion into the etchant (29). Palik et al. assume that the anisotropy might be attributable to differences in activation energies and in backbond geometries on different surfaces (33). In this paper, experimental results on the orientation dependence of the silicon etch rate for several solutions as a function of composition and temperature will be given. An attempt will be made to give a model valid for all anisotropic silicon etchants, explaining the underlying mechanism. Furthermore, results on the etch behavior of the most widely used passivation layers SiO2 and Si3N4 will be reported. The effects of dopants on the silicon etch rate will be discussed in an accompanying paper (35).