CRYSTAL ORIENTATION DEPENDENCE OF
MACROPORE FORMATION OF P- AND N-TYPE SILICON
IN ORGANIC ELECTROLYTES
M. Christophersen, J. Carstensen, and H. Föll
Materials Science, Faculty of Engineering
Kaiserstr.2, 24143 Kiel, Germany
Phone: + 49 431 77572-501
Email: hf@tf.uni-kiel.de

The crystal orientation dependence of pore formation in Si provides important information on pore formation mechanisms. In addition, the anisotropy of macropores is an important parameter for applications, especially the microstructuring of silicon. The dependence of the macropore morphology on the orientation of p- and n-type silicon samples was studied for various and mostly organic electrolytes containing HF. One of the important electrolyte parameter is the ability to support the oxidation reaction needed for "smoothing" the macropores. A comparison between organic and aqueous electrolytes is given and discussed in terms of the "current-burst-model" [1].

Lehmann and Föll [2] first demonstrated that very regular, smooth and deep macropores can be obtained by anodization of n-type-silicon under back-side illumination in aqueous HF containing electrolytes. Their first model of the macropore formation on n-type-silicon was based on the diffusion of holes generated at the back-side to the pore-tip via a focussing effect of the space charge region surrounding each macropore (so-called "space charge region" model) Several technical applications for example micro-structuring [3, 4] were developed based on the n-type-Si macropore etching technique [5]. 1994 Propst and Kohl [6] found macropores on p-type-silicon by anodizing the sample in water-free acetonitrile containing some hydrofluoric acid and thus demonstrated that the simple space charge region model cannot be the only reason for macropore formation. In the following years, many papers were published dealing with macropores on p-type-Si [7 - 10] and their possible formation mechanism, but the dependence of pore morphologies on crystal orientation [11] remained mysterious. In the present paper we extend our previous work regarding the crystal orientation dependence of macropores on n-type-silicon [11], to macropores obtainable with organic electrolytes in p- and n-type Si and new pore morphologies, never seen before, were obtained. In addition to the orientation dependence, a direct comparison of macropore formation in strongly and weakly oxidizing HF-containing solutions was made. The results are interpreted in terms of the "current burst model of Carstensen et al. [1, 10, 12 – 16]

Since p- or n-type-Si specimen with orientations other than {100} and {111} are hard to obtain, samples for experiments were produced by cutting slices from Si crystal end pieces at predetermined orientations. After polishing the resulting slices with diamond paste of 3 µm grain size, specimens with lateral dimensions of (2 x 2)cm² were cut using a conventional saw. {111} oriented p- and n-type samples obtained from standard wafers (supplier: MEMC) were included in the experiments. In the present investigation no pre-structured nuclei were used.

For pore etching, the specimens were subjected to anodic etching under the following conditions; all parameters (current, voltage and electrolyte temperature) were computer controlled:
i) p-type-samples: Constant (galvanostatic) current density between 0.5 - 2 mA/cm², stabilized electrolyte temperature of 20°C, no illumination was used.
ii) n-type-silicon samples: PC-controlled back side illumination at constant external bias (4 V) and constant current density of 4 mA/cm² was used.

After cleaving the samples, the pores produced were observed with a scanning electron microscope (SEM). For some samples the cleavage plane was polished prior to the SEM investigation.

An optimized electrolyte [10] was used for p-type Si samples with orientations of {100}, {511}, {5,5,12} and {111}. The macropores grow preferentially in the <100>-direction, only on {111}-samples <100> and <113> oriented macropores were found – reminiscent of the results in n-type-Si [12]. But in contrast to n-type-silicon in aqueous electrolytes, the switch-over from <100> to <113> is not as well defined. Examples are shown in Fig. 1.

Next, we used organic electrolytes with (back side illuminated) n-type-silicon. From the current burst model [1, 12, 15], the expectation was that the oxidation reaction was weakened and the direct dissolution reaction would prevail, producing strongly anisotropic and smallish pores with possibly dendritic side pores. Whereas this proved to be true in principle, the real finding was that the strongly enhanced direct dissolution component indeed increased the porosity, however, not by forming many smooth and narrow pores but by excessive branching including new types of morphologies. An example for a {100} oriented n-type sample with comparatively smooth pores is shown in Fig. 2. The macropores tend to form branches in <100> directions and develop "rough" pore-walls, an effect only observed in aqueous electrolytes at much higher voltages. Fig 2b shows that the branches have a high degree of symmetry.

Quite remarkable pore morphologies never seen before (cf. Fig. 3) emerged with n-type Si samples of {1,1,10}, {114}, {113}, {211}, {223} and {111}orientations.

In all cases three qualitatively different kinds of pores could be observed in all samples, see e.g. Fig. 4f). The pores generally exhibit extreme branching, and many "break-through-pores" (in [17] dubbed "dendritic pores") are visible between the macropores. The <100> direction, again, is the main growth direction; in some cases a preferential growth in <113> is visible, too (cf. Fig. 3a or 3f). The directions of the branches (and of some main pores), however, are not always easily identified. In Fig. 3f), for example, the pores also seem to grow in the <111> direction perpendicular to the surface which would be difficult to explain in the current burst model (or in any other model). This will be explained below as a salient feature of a new kind of pore structure and for this the traces of {111}-planes were outlined in some pictures. In higher magnifications (Fig. 4) the nature of the disturbing <111> branches can be identified.

Upon close inspection of the micrographs (bearing in mind that the cleavage surface is no longer necessarily a {110} surface), it can be seen that there is still a main pore growing in a <100> direction, or, on occasion, in a <113> direction. In some cases the pores still emit clearly defined side pores (in <110> or <113> directions), but a new feature emerged (dubbed "wings"), visible whenever a pore seems to be aligned in a <111> direction. Wings are two-dimensional cavities branching out from a main pore and bounded by {111} "ceilings" from which more normal pores emerge like stalactites from the ceiling of a cave. Fig. 4 e) gives a schematic comparison between wings and regular side pores and Figs. 4 a) - d) show relevant micrographs.

The fact that three qualitatively different kinds of pores can be obtained simultaneously together with their "fractal" appearance, is a strong indication that the system is close to a critical point in parameter space [12]. How a specific pore evolves is no longer controlled by the prime parameters of the system, but by the influence of secondary parameters (e.g. space charge regions, source of the hole supply, processes in the neighborhood, modifications of the intrinsic H-passivation kinetics by the specific electrolyte) on the statistical processes controlling the nucleation and growth of pores via current bursts.

The back-side illumination, via the hole dependent oxidation part of a current burst, forces a preferential growth towards the source of the holes (the sample back-side), i.e. perpendicular to the surface, which conflicts with the preferred <100> growth direction for samples not oriented in <100>. While it is rather obvious that the macropore growth on n-type silicon in organic electrolytes is not as well defined as in aqueous electrolytes, using organic electrolytes seems to be a promising way to reduce pore diameters to below 200 nm (Fig. 4c); a task not easily achieved in aqueous electrolytes. In [14], it will be shown that it is possible to obtain a smooth morphology by using control techniques based on predictions of the current burst model and thus to utilize the potential of organic electrolytes in connection with n-type Si.

In all experiments, break-through pores (defined in [18, 19]) were produced simultaneously with the heavily branched macropores. In contrast to the macropores, they show an well-defined orientation dependence, see Fig. 5. The main direction is always <100> as in aqueous electrolytes [19]. The detailed morphology of break-through pores in n-type-silicon has been studied with TEM [17, 19]. The pores were always found to consist of connected (111) octahedra.

The high-resolution SEM micrographs in Fig. 5 confirm this finding for the break-through pores in the present experiments. The {111} planes are the most stable planes independently of the electrolyte. In the current-burst model this is attributed to the intrinsically fast hydrogen passivation kinetics of {111} planes which prevents nucleation of current bursts most efficiently if the direct dissolution current part is dominant.

The simultaneous occurrence of (heavily branched) macropores and break-through pores is a direct consequence of the strongly reduced oxidation part of current burst in organic electrolytes that is further aggravated by the limited hole supply in n-type Si. Oxide growth may no longer be sufficient to quench the current flow generated by local avalanche effects. In addition, the suppression of an anodic oxide leads to easier formation of break-through pores because (i) the voltage does not drop in an oxide, and ii) an oxide smoothes sharp edges which reduces the field strength for hole generation.

It is remarkable that only the break-through pores show strictly the same anisotropic growth in organic and aqueous electrolytes. In the current burst model, break-through pores (or dendritic pores or mesopores) are generated if the direct silicon dissolution is dominant, i.e. under condition of restricted hole supply and/or limited oxygen availability. If the oxidation component is sufficiently suppressed, external parameters (including electrolyte composition) thus do not matter any more; break-through pores are an expression of an extreme case and will always appear in the same morphology. New current bursts will be nucleated as long as planes other than {111} are available which leads to a formation of an octahedron. Eventually, at the tip of the octahedron (which points in a <100> direction), a new octahedron is started. Without any secondary parameters inducing a preferential direction, all tips are equivalent and a dendritic pore with branches in all <100> directions results.

The "current burst" model for macropore formation from Carstensen et al. [1, 10, 12 - 16] postulates that pore formation results from an interaction between direct silicon dissolution and a dissolution via an anodic oxide. The rich morphologies of macropores result from the specific interactions of both reactions.

The direct silicon dissolution is heavily anisotropic and promotes branching, whereas oxidation is rather isotropic and promotes smooth structures. Reducing the oxidizing compont, e.g. by using organic electrolytes, emphasizes the reaming time for H-termination. Differences in the passivation time. Differences in the passivation for dangling bonds get pronounced increasing the anisotropy character of the direct dissolution process.

Macropores form if there is sufficient oxidation to balance direct dissolution. The oxidation process smoothes the strongly anisotropic precursor of the direct dissolution, and, since it needs holes, orients pore growth to the source of holes; i.e. to the sample back side in a typical macropore experiments with backside illumination.

This can be seen in a direct comparison of aqueous and organic electrolytes for n- and p-type-Si. Fig. 6a) shows macropores on {5,5,12}-orientated p-type-Si etched at low current densities in an aqueous electrolyte (etch conditions as described in [11]). The pore growth is not well defined, mirroring the isotropic component of oxidation. The use of an organic electrolyte (reduced oxidizing reactants) leads to well-defined macropores on p-type Si (Fig. 6b)) because now oxidation is weakened and thus the anisotropic part strengthened. With the optimized electrolyte chosen, condition close to the optimal balance between direct dissolution and oxidation were obtained.

Pores in n-type Si behave differently. In organic electrolytes badly branched pores are formed, due to "double" suppression of oxidation (lack of holes and lack of oxidizers) and the resulting dominance of direct dissolution, whereas in aqueous electrolytes the oxidation component is strengthened and, in the case of very smooth macropores is in perfect balance with direct dissolution.

From this a prediction can be made: Changing the balance between direct dissolution and oxidation by e.g. adding oxidizers or protonizers to an electrolyte, should change the pore morphology in a predictable manner. This has worked in other cases [10, 13], here it still needs to be tried.

The use of non-oxidizing solutions leads to a reduction of the pore-diameter (Figs. 4c, 6c and d) and this, together with controlling the branching by optimizing the electrolyte and employing the pore growth control technique described in [12, 14] may be of importance for sub-µm structures, attempted, e.g., for photonic crystals [20].

Conclusions: Organic electrolytes were used for macropore formation on p- and n-type-silicon (under backside-illumination) for different substrate orientations. The observed pore morphologies are often quite different from the ones observed in aqueous electrolytes under comparable conditions. All experimental results can be interpreted within the current burst model by discussing the balance between direct dissolution and oxidation in a current burst.

We thank S. Rönnebeck and A. Feuerhake for the help by the preparation of the silicon-samples and the fruitful discussions. We are indebted to Wacker Chemitronic for supplying bulk Si needed to cut samples with unusual orientations and to our colleagues form the Center of Microanalyse of the Christian-Albrechts - University of Kiel. This work was support by the Deutsche Forschungsgemeinschaft (Förderungsnummer: FO 258/1-2).

References:

pdf-version of this file (5.622KB)