Lattice strain enhanced acidic etching on as cut sawn silicon wafer

Abstract

The mechanical processing of silicon wafers leads to a heterogeneous lateral strain distribution and various modifications of the silicon, both of which influence the resulting topography after acid etching. In this study we investigate the influence of local strains and the initial topography of slurry and diamond wire saw wafers on the acid etching mechanism. The strain distribution is quantified and qualified by Raman microscopy before and after thermal treatment, while the topography is characterized by confocal microscopy before and after etching. The thermal treatment was used to selectively relax strains and investigate the effect of the individual strains on the etching mechanism. We found that amorphous silicon and compressive strained silicon are mainly present on the top most surface of the saw damage and do not influence the chemical reactivity of acid etching. In contrast, highly reactive tensile strained silicon is found up to 2.7 μm deep in the saw damage and acts as an etching mask. Rapid etching of the tensile strained silicon by HF/HNO₃/H₂SiF₆ leads to the formation of cracks with high local concentrations of intermediate species. These strains induced cracks are etched out together with the original saw damage induced cracks and trenches and form the final surface after etching. Furthermore, we can show how the tensile strain strength must have a relative Raman shift of at least −2 cm⁻¹ to have an effect on the local etch rate. Our data demonstrate how mechanical treatment in combination with thermal treatment and acidic etching can be used to optimize the resulting topography for applications like photovoltaics. In addition, it provides a deeper insight into the acid etching mechanism for non-planar silicon wafers.

Introduction

As has been shown in previous studies, acidic etching of multicrystalline silicon wafer can result in different topographic shapes depending on the type of wafer sawing, the composition and concentration of the etchants [[1], [2], [3], [4]]. The topography of silicon directly influences the quality of applications like photovoltaics, biosensors or microelectronics. Photovoltaics requires a topography that absorbs as much light as possible homogeneous over the entire surface, while increased surface roughness in biosensors increase the density of functional groups on the silicon substrate surface [5,6]. Finally, in microelectronics distinct topographic shapes for functional elements e.g. steps, trenches or line structures are required on a planar semiconducting substrate. Such shapes are generally created by photolithographic techniques with masks that are inert to a subsequent wet chemical etching process. The acidic etching solution with nitric acid (HNO₃), hydrofluoric acid (HF) and hexafluorosilicic acid (H₂SiF₆) is used most frequently. The etching mechanism is commonly described as a two-step mechanism comprising (I) the oxidation of silicon to silicon oxide (SiO₂) and (II) the subsequent dissolution of SiO₂ by HF [[7], [8], [9], [10], [11]].

$$3Si+4HN{O}_3\to 3Si{O}_2+4NO+2H_{2}O$$

$$Si{O}_2+6HF\to H_{2}Si{F}_6+2H_{2}O$$

Based on the extensive research of Acker et al., the etching rate of HF/HNO₃ solutions on silicon can be influenced by three parameters: the concentration of the oxidizing acid, the strain of the sample present and the formation of intermediary nitrogen-based species during the silicon oxidation step [2]. The most important factor is the formation of intermediary species, as it enhances the slowest reaction step, the oxidation of silicon. A higher concentration of HNO₃ leads to a stronger formation of the intermediary species as well as to a faster oxidation and thus a higher etching rate. However, recent studies reveal that undissociated HNO3 is the reactive species in the initial step [11]. In contrast, the aspect of strain has hardly been researched. It is proposed that strained silicon has a lower bonding energy between the silicon atoms, reducing the activation energy required for oxidation and the subsequent dissolution with HF [2,12]. In addition, the amount of strain on a sample can vary depending on the manufacturing and processing technique [3,13]. Usually strain is caused by mechanical treatment or lattice defects [[14], [15], [16]]. Here we focus on the strain caused by mechanical treatment e.g. indenting, scratching, polishing, diamond wire sawing or slurry sawing. The slurry processing can be seen as multiple indenting of the silicon surface, which produces small, roundish pits and localized subsurface cracks [1,3]. In such a case, the strain is distributed as typically seen in indents, with compressive and tensile strain mixed in separated areas. In contrast, diamond wire cutting involves the action of indenting and then scratching along the surface, creating long shallow trenches with local breakouts. Due to the constant breaking out of fragments, most of the underlying subsurface cracks have also disappeared. The strain forms large clusters of compressive strained silicon and amorphous silicon on the elongated trenches and localized areas of tensile strained silicon, especially around larger break outs. In general, most measurements of the strain distribution in large areas are based on Raman microscopy, measuring the shift of the signal for silicon towards lower or higher wavenumbers [2,17,18]. Later optimizations by Herold et al. use a peak deconvolution technique to quantify and qualify localized surface strain [19]. Furthermore, he also showed that thermal treatment can relax certain strain states, making it possible to measure the etching rate only on the basis of a specific strained material.

In addition to the three factors mentioned by Langner et al., we propose that the surface topography in particular has a major influence on the local etching rate [2]. In a previous work Acker et al. discussed the change of different surface parameters after etching of slurry processed wafer based on the initial geometry [3]. Here we continue these experiments and extend them with new techniques for the analysis of strain to discuss in detail the effect of the sample geometry and strain on the local etch rate.