GaAs behaves quite differently as compared with InP and GaP. Namely, the nucleation phase of pores is much more capricious and the nucleation of pores starts only at relatively large defects like pits or scratches. A uniform structure could be obtained if the pores would nucleate easily on predefined defects and grow stable along <111>B directions (see Section 3.3.1).
As it was noted in section 3.3.1, pore growth along <111>B directions may lead to a 3D porous structure. Note that <111>B oriented pores intersect each other without changing their direction of growth or shape. This is essential for producing 3D periodic structures by intersection of pores starting from a common nucleus, by analogy to the pores in Si forming the 'Kielovite' structure . However, experimentally it is not simple to fabricate such 3D structures. This is because the <111>B directions have a high tendency to branch and thus the uniformity will be destroyed as more and more branches appear.
The experiments show that branching can be avoided if the density of nucleated pores is very high, thus the branches will simply have no 'space' for growth. For a dense pore nucleation the current density is critical, however, by increasing the current density the branching tendency increases as well. Thus, a compromising value for the current density should be chosen.
The approach used to obtain a uniform nucleation in GaAs has the following two steps:
The goal of the first step is to create a high density of surface defects, which will serve as nucleation points for pores in the second step. The amplitude and duration of pulses, i.e. the amount of charge per pulse per unit area, is critical and should be optimized for an efficient generation of dense nucleation points.
If the first step is not optimized, domains of crystallographically oriented pores will develop at this stage. Consequently, during the second step no new pores will nucleate between the domains, but branches inside the already existing porous domains (created during the first anodization step) will be nucleated, increasing the already existing domains.
Figure 3.44: (100)-oriented n-GaAs, cross section; a) Top view. Random nucleation of pores on the surface of a (100) n- GaAs sample after the first anodization step. b) A quasi-uniform 3D structure obtained after the second anodization step was applied.
For the optimization of the first step, it should be taken into account that two types of pulses with equal amount of charge per pulse per unit area can be distinguished: a) low current and long time; b) high current and short time. These two cases are presented schematically in Figure 3.45.
Figure 3.68: Two types of pulses with the same amount of charge per pulse per unit area.
It was found that for the case a) the domains tend to be big and less dens (number of domains per unit area). On the other hand, for the case b) the domain size decreases, while their density increases significantly. Thus, in order to obtain uniform nucleation, i.e. to avoid domain formation altogether, high-current/short-time pulses should be considered. The range of currents and pulse widths depends mainly on the doping level of the substrate and electrolyte concentration.
The second anodization step should provide the necessary etching conditions for the pores (nucleated in the first step) to grow into the substrate as deeply as required. Therefore, special care should be taken during the the second anodization step as well. Too low current densities could result in fewer pores growing into the substrate than initially nucleated, whereas too high current densities could result again in domain formation.
An example of a 3D structure obtained at optimized conditions is shown in Figure 3.44. In spite of the fact that the structure in Figure 3.44 seems to be quite uniform - it is not, because the initially nucleated pores are randomly distributed on the surface of the sample. In order to nucleate pores in a highly ordered manner, pre-patterning is required.
Figure 3.46: a, b) Top view at two magnifications of the nucleated pores in photolithographycally defined windows; c) high magnification of the nucleated pores. The two <111>B directions can be observed; d) Cross section view of the nucleated pores. Due to branching, the uniformity defined by lithography is lost into the depth of the structure.
Top and cross sectional views of to the anodized pre-patterned samples are presented in Figure 3.46. One can see (Figure 3.46a, b and c) that the pores indeed nucleate in the windows defined by the electron beam lithography. In each window two small black spots can be observed separated by a white band. The black spots are the nucleated <111>B pores. In spite of the fact that nucleation in predefined windows seems to work, some other problems can appear during the etching process. For example, in addition to the pores nucleated in predefined windows some random pores nucleate, too. Such randomly nucleated pores are marked in Figure 3.46a by black circles. It is likely that random pores nucleate at defects with a lower breakdown voltage, leading to faster nucleation, faster growth, high current densities and successive branching, diminishing the necessity to nucleate or sustain pores in the predefined windows in order to carry the external current. Therefore, even the pores which have started to nucleate in predefined windows can stop growing in favor of the randomly nucleated pores.
In Figure 3.46d the cross sectional view taken from an anodized pre-structured sample is illustrated. One can see that the uniformity of the porous layer does not follow the predefined structure on the surface. The main obstacle here is the branching of pores from the already nucleated pores. A further optimization of the mask as well as of the etching conditions is required in order to obtain uniform 3D structures in GaAs by means of anodization.