Influence of the doping level on the porosity of silicon nanowires prepared by metal-assisted chemical etching

SiNWs were fabricated by means of the well-established MACE method as described by Huang et al [7]. The approach involves two steps. First, a lithographically structured noble metal film (Ag, Au, Pt, Pd) is deposited on the Si substrate. Subsequently, the etching is performed in HF and an oxidative agent, typically H₂O₂.

Different commercial Si substrates with doping levels ranging from 10¹² to 10¹⁹ cm⁻³ (p- as well as n-doped) have been used as starting material. The Si surface was cleaned by an RCA-I treatment [15] and rinsed with deionized water. For lithographic structuring, nanosphere lithography was used [16]. Here, monodispersed polystyrene (PS) particles (commercially available from the Microparticles GmbH Company as a 10 wt% solution) were deposited on the cleaned surface by means of a Langmuir–Blodgett technique, resulting in a hexagonally closed-packed monolayer of the particles. Afterwards, oxygen plasma at a low process pressure of 0.06 mbar and an oxygen flow of 160 sccm min⁻¹ (Diener Electronic plasma cleaner 'Nano') was applied in order to reduce the diameter of the spheres. The diameter of the spheres could be controlled by varying the plasma etching time. Subsequently, a 60 nm thick silver film was deposited by thermal evaporation (model B 30.2, HVT Dresden) onto the masked substrate. After removing the PS spheres in an ultrasonic bath consisting of CH₂Cl₂, a silver mask with highly ordered openings remains. Finally, the silver masked substrates were etched for 10 min in a solution consisting of 5.65 mol L⁻¹ HF and 0.1 mol L⁻¹ H₂O₂ leading to the formation of SiNWs. The whole procedure was carried out at room temperature in darkness. The silver was removed by nitric acid.

We performed numerous experiments in the whole parameter space spanned by the H₂O₂ concentration, the HF concentration, the process temperature and the doping type, dopant and doping level of the Si wafers. In general, following the Nernst equation, both an elevated temperature and a high H₂O₂ concentration enhance the redox potential of the etching solution and thus, the etching rate. As a rule, pSi is formed for every doping type and level when these two parameters are high enough. However, for the sake of the stability of the process, it is important to keep the parameters in a moderate regime. The following results pertain to fixed values of the temperature (room temperature) and H₂O₂ concentration (0.1 mol L⁻¹). The concentration of HF was fixed to 5.65 mol L⁻¹, but the latter parameter hardly influences the morphology.

It turned out that for the values given above the morphology is determined by the doping concentration alone. In particular, it does not depend on the dopant type. Therefore, in what follows we concentrate more closely on relation between doping level and morphology.

We began with commercially available n- and p-doped Si substrates with doping levels between 10¹² and 10¹⁹ cm⁻³, listed in table 1. We followed the procedure described in section 2.1 with the parameters of the etching solution specified above. The scanning electron microscope (SEM, using a Philips XL30 FEG ESEM) characterization of resulting structures shows clearly that NWs etched from highly doped Si were always completely porous (figures 1(a) and (b)), while NWs from slightly doped material were primarily single-crystalline (figures 1(c) and (d)), surrounded only by a relatively thin pSi layer [17].

Zoom In Zoom Out Reset image size

The commercially available doped substrates do not allow a gradual variation of the doping concentration and type. Thus, it is not possible to establish whether the transition between crystalline and porous NW morphology is abrupt and at which doping level it occurs. Therefore, we devised an experiment in which we gradually vary the doping level of the NWs, keeping the dopant type unchanged. The remaining process parameters are kept constant as well. Multiple boron-doped Si layers of gradually varying doping level in the range of 6 × 10¹⁷–6 × 10¹⁸ cm⁻³ were grown epitaxially by means of molecular beam epitaxy (MBE, using a Riber SIVA 45 instrument) at 500 °C and a deposition rate of 0.5 Å s⁻¹. The layers were grown on an P-doped Si(100) substrate with a doping level of 1 × 10¹⁵–2 × 10¹⁵ cm⁻³. Prior to the MBE growth, organic contaminants on the surface of the substrate were removed by an RCA-I cleaning [15]. The layers had a thickness of 30 nm each. An additional layer of 140 nm thickness and a high doping level (10¹⁸ cm⁻³) was deposited on the top of the layer stack. It covered the stack to provide protection against etching away the top of the NW during MACE. The boron concentration was measured afterwards by secondary ion mass spectroscopy. The Si layers were separated by additional 0.4 nm thick Ge layers deposited by MBE with a growth of 0.2 Å s⁻¹. The additional Ge layers played the role of markers allowing an easier differentiation between the layers of different doping levels in transmission electron microscopy (TEM, using a JEOL JEM 4010) characterization. The samples with the fabricated boron-doped layers and spacers were used as starting material for the fabrication of NWs exactly as outlined in section 2.1.

Figure 2(a) shows the TEM image of an etched NW. Layers of different boron concentrations separated by thinner spacers are discernible; they are labeled in the figure. The slightly doped Si layers exhibit a crystalline structure (figure 2(b)), while the highly doped Si layers are clearly porous (figure 2(c)). We observe that the transition between crystalline and porous Si is rather abrupt and takes place around the critical doping level between 1 × 10¹⁸ and 3 × 10¹⁸ cm⁻³. It should be noted that the value pertains to our process parameters defined in section 2.1.

Zoom In Zoom Out Reset image size

We also observe that the Ge is etched away more intensively than Si. It is a consequence of the fact that GeO dissolves rapidly in H₂O; a detailed study of this phenomenon is presented in [18].

In what follows, we present the more detailed characterization of the pSi NWs obtained by means of high resolution transmission electron microscope (HRTEM, using a JEOL JEM 4010), energy dispersive x-ray spectroscopy (EDX) and electron energy loss spectroscopy (EELS) using a Philips CM 20 FEG transmission electron microscope with an acceleration voltage of 200 kV. We concentrate on the wires with high doping levels (10¹⁹ cm⁻³). We aim to explain why only highly doped Si results in porous NWs. The pSi obtained by MACE described above resembles closely the pSi originating from well-studied anodic etching, and we assume that similar processes take place in both cases.

To examine the effect of porosity on the crystal quality of the NWs, electron diffraction patterns from bulk Si and from pSi NWs were taken, cf figure 3(a). The diffraction patterns show that the crystalline order of the silicon was not perturbed during the formation of the porous structure, cf the insets (1) and (2) in figure 3(a). We note that the same was observed after anodic etching [14].

Zoom In Zoom Out Reset image size

Furthermore, the NWs were analyzed quantitatively using EDX (along the blue line in figure 3(a)) in four characteristic areas: the crystalline Si substrate, the porous Si layer beneath the silver film, the silver layer, and the porous Si NWs. The EDX intensities of the Si-K_α, O-K_α and Ag-K_α lines are shown in figure 3(b). The measurements show that the porous structure consists of both Si and O. The ratio of Si to O in the area of the pSiNWs amounts to about 1:1.4. The type of the Si–O binding cannot be unambiguously determined by means of EDX, and we resort to EELS for a closer study.

The spectra were taken along the red–green line in figure 3((c), panel 1). The red part of the line comprises the measurements in the area of the pSiNW, while the green line is part of the crystalline Si substrate. Next to this, the resulting EELS spectra are shown after subtracting the background, figure 3((c), panel 2). Three typical edges can be identified: the silicon edge, the carbon edge (caused by the embedding glue of the cross-section sample prepared by a focused ion beam) and the oxygen edge.

The fine structure of the energy loss spectra (electron energy loss near fine edge structure, ELNES) yields information about the binding state of Si. The fine structure of the Si-L_2,3 edge allows us to analyze the binding of an Si atom with its nearest atomic neighbors. In particular, we can differentiate between Si–Si bindings and Si–O bindings.

Figure 3((c), panel 3) shows the fine structure of the Si-L_2,3-edge of two spectra near the border between the substrate and the NW. The green spectrum exhibits a clear Si peak at the energy loss of 100.5 eV, typical for Si–Si bonds [19]. In the area of the NW, the appearance of the spectra abruptly changes. Only a weak shoulder can be seen at 100.5 eV, while a double peak emerges at an energy of about 108 and 114.5 eV. This double peak is typical for Si with an oxygen environment [19]. Altogether, the fine structure points at an overlap of spectra corresponding to Si–Si and Si–O bonds.

The pore size of the pSi was further investigated by means of positron annihilation lifetime spectroscopy. Pulsed monoenergetic positrons were generated by pair production using the 40 MeV electron beam at the ELBE radiation source (Helmholtz Zentrum Dresden Rossendorf) [20]. In mesoporous material, ortho-positronium with a characteristic lifetime can be detected [21]. In pSi, the lifetime amounted to τ = (80.8 ± 19.5) ns with an intensity of 1.66%. With the positron lifetimes measured the pore size is calculated to (6.9 ± 2.8) nm. In agreement with the classification of anodically etched porous materials [11], the Si can be regarded as mesoporous.

With the experimental results at hand, we are now able to give a theoretical model.