The structural characterization of samples etched at different etching times provides additional insight to the mechanism of formation of the mesoporous SiNWs on highly boron-doped Si by the single-step MACE process. In principle, MACE involves two successive processes: surface nucleation of metal catalysts (e.g., Ag) and anisotropic Si etching. Si dissolution takes place through oxidation by H2
and oxide dissolution in HF. Metal nucleation occurs preferentially at surface states and sites around the dopants. Since the oxidation donates four electrons, while Ag+
ion reduction consumes only one electron, a space charge is formed by the excess electrons on the surface that electrically drives Ag+
ions to diffuse toward the nuclei for reduction. Alternate oxidation and nucleation cycles induce sinking of the Ag particles into the Si substrate, resulting in Si etching and SiNW formation. These nanowires are vertical to the Si substrate. The morphology and texturing of the SiNWs depend strongly on the original Si wafer resistivity. SiNWs from resistive Si wafers have in general a smooth surface and a crystalline core without pores. On the other hand, Si wafers with a resistivity of less than approximately 5 mΩ·cm produce mesoporous SiNWs. This was demonstrated for both p-type [11
] and n-type Si wafers [12
]. Since dopants are additional preferential sites for the nucleation of Ag particles, their high density induces porosification of the Si substrate and the formation of a mesoporous layer at the interface between the SiNWs and the crystalline Si substrate. Our experiments showed that the thickness of this porous Si layer increases with the increase of the etching time. It was also deduced from our PL experiments (this is discussed below) that the initial porosity of this layer was lower than that of the SiNWs. Furthermore, the porosity of the SiNWs was gradually increasing from their bottom to their top (different pore and nanocrystal sizes). These observations led us to the conclusion that the formation of the porous Si layer underneath the SiNW arrays precedes the SiNW formation. The SiNWs are thus porous from the beginning, while additional porosification of the nanowires takes place during etching. The higher porosity of the tops of the SiNWs is attributed to the longer time into the etching solution and is responsible for the saturation of the process after a certain time. Indeed, we observed that after the 60-min etching time, it was not possible to further increase the SiNW length. This is attributed to the fact that part of their tops is fully dissolved in the solution when the porosity of this part of the nanowires becomes high enough. From that time on, although the etching process continues on the Si surface, the SiNW length does not increase, since the nanowire tops are progressively dissolved in the solution.
The PL spectra from the mesoporous SiNWs can be understood by taking into account their structure and morphology. As discussed above, the nanowires are composed of assemblies of Si nanocrystals and nanowires interconnected in a Si skeleton, the mean size of these nanocrystals being different along their length. The PL spectra from assemblies of Si nanocrystals are in general broad, and peak position depends strongly on their size distribution and the chemical composition of their surface [21
]. Quantum confinement of the generated carriers is at the origin of the long decay times (in the several micrometer range) [25
]. The recombination mechanism depends on the structural and chemical composition of the nanocrystal surface. In hydrogen-terminated nanocrystals without important structural defects at their surface, free exciton recombination is in general observed [28
], while in oxidized nanocrystals, a significant Stokes shift is observed between the absorption and the PL band peak energy [27
], attributed to an important pinning of the nanocrystal energy bandgap due to localized states at the interface of Si NCs with the surrounding SiO2
]. The same effect can be caused by structural defects at the surface of the nanocrystals. Pump and probe measurements confirmed the above behavior [33
The differences observed from the different samples investigated in this work can be explained, based on the above, by considering the size distribution of nanocrystals and the state of their surface. In the as-grown samples, a number of very tiny nanocrystals that are light emitting are found at the surface of larger nanocrystals. On the other hand, a lot of structural defects exist that quench luminescence (spectrum 1 in Figure ). The tiny nanocrystals (slightly oxidized at ambient atmosphere) are removed by the first HF dip. In addition, some of the structural defects that quench PL are also smoothed out. This is why the PL signal from the SiNWs after the first HF dip is red-shifted compared to that obtained from the as-formed nanowires, and its intensity increases (spectrum 2 in Figure ). The different surface chemistry of the as-formed and HF-treated NWs is confirmed by the FTIR results. In the HF-treated samples, the surface is hydrogen-terminated, while the as-grown sample and the sample after piranha cleaning show mainly Si-O and SiO-H bonds at the surface. The slightly oxidized NWs after piranha cleaning show a blueshift in PL due to a slight shift of the mean nanocrystal size by oxidation (spectrum 3, Figure ). The increase in intensity is again attributed to a further smoothing of surface structural defects that quench PL. Furthermore, light emission from additional nanocrystals, which were dark before due to their large size and are now smaller after oxidation, contributes to the increased PL intensity.
The large increase in intensity after the last HF dip is attributed to the total change of the structure and morphology of the SiNWs, as illustrated in Figure c. This additional HF dip resulted in dissolution of the upper part of the SiNWs. The length of the remaining SiNWs was only the one fourth of their original length. However, even if the SiNW length was significantly smaller, the PL intensity was increased by more than one order of magnitude. To our opinion, PL in this case comes mainly from the mesoporous Si layer underneath the SiNWs. The mean size of NCs in this layer was initially large, while it was reduced by HF/piranha/HF treatments. The peak position is mainly determined by the mean size of the NCs of this layer. Consequently, there is no direct comparison of this spectrum with the three previous spectra.