|Home | About | Journals | Submit | Contact Us | Français|
In this work, a new approach for the one-step integration of interdigitated electrodes on macroporous silicon substrates is presented. Titanium/gold interdigitated electrodes are used to pattern p-type silicon substrates prior the anodization in an organic electrolyte. The electrolyte characteristics, conductivity, and pH have been found to affect the adherence of the metal layer on the silicon surface during the electrochemical etching. The impact of the metal pattern on size distribution and morphology of the resulting macroporous silicon layer is analyzed. A formation mechanism supported by finite element simulation is proposed.
Silicon anodization in hydrofluoric acid (HF)-based organic electrolyte is a simple and effective technique to obtain macroporous silicon (MPS) on p-type substrates. Macroporous silicon has been demonstrated as an interesting material for a number of applications such as silicon micromachining [1,2], photonic , and sensing [4,5]. The integration of metal electrodes on macroporous silicon surface is of paramount importance in case of sensing applications . In silicon microtechnologies, metal electrodes are usually obtained by patterning metal layers evaporated or deposited on the substrate surface. Due to the high macroporous silicon surface roughness, the patterning of electrodes with dimensions below 5 μm is a challenge. An alternative approach is to pattern the metal electrodes on the silicon surface prior the electrochemical etching. In this case, the presence of the metal on the surface affects pore distribution and morphology. In this work, the macroporous silicon characteristics are investigated as a function of the geometry (width and pitch) of metal pattern and modeled using finite element simulations.
Metal interdigitated electrodes were patterned on the surface of p-type silicon substrates (10 to 20 Ω·cm, (100) orientation) using standard lift-off. The electrodes consist of 1400-μm long Au ribbons (100-nm thick). Au was chosen because it is HF resistant. A 5-nm thick Ti layer was used to promote the Au surface adherence. Ti was chosen instead of Cr because the combination of Cr and Au forms an electrochemical local element, giving rise to the metal layer detachment after few minutes of electrochemical etching . The metal was evaporated all over the wafer using a dual e-beam evaporator (e-gun Balzers, Liechtenstein). The nondesired metal layers on the wafer were removed by dissolving the prepatterned underlying resist in acetone. Three different geometries (widths and pitches, WP (μm)) were tested on the same sample: WP1 = 2-2, WP2 = 5-2, and WP3 = 2-5. The prepared samples were then electrochemically etched to form the macroporous silicon layer. All the experiments were carried out at room temperature, and the solution was stirred to avoid inhomogeneities due to the gas accumulation on the surface. A platinum grid was used as the cathode, while the anode was the silicon substrate. A 200-nm thick aluminum layer was evaporated on the wafer back side, and an electrical contact was made by placing an aluminum plate in contact with it. A mixture of HF and N,N-dimethylformamide (DMF) (HF (48%):DMF = 1:4) was used as electrolyte. DMF is not stable in presence of acids like HF and is, thus, partly hydrolyzed back into formic acid and dimethylamine . The reaction is exothermic and gives rise to the increase of the solution temperature. Solution pH and conductivity were measured after the cooling of the solution at room temperature. A pH of 4 and a conductivity of 2 mS/m were measured. The hydrolysis of the DMF is not limited to the preparation moment. As a consequence, the properties of the solution were observed to change to reach a threshold value. For a 20-day-old solution stored in a closed bottle, it was measured with a pH of 6 and a conductivity of 30 mS/m, indicating that a bigger fraction of the DMF was hydrolyzed. Both as-prepared and 20-day-old solutions were used as electrolytes.
The enhanced conductivity of the old solution allows to use higher current densities for lower applied voltage. In both cases, the macroporous silicon layers were formed, keeping constant the etching current density Jetch at a value lower than the critical current density Jps=182 mA/cm2 at which the silicon surface electropolishing is produced .
When the silicon substrate is prepatterned with an HF-resistant material as the Au mask, the pore formation is expected to be limited to the uncovered areas, as sketched in Figure Figure1a.1a. Figure Figure1b,c1b,c shows the SEM cross-sectional tilted view of a macroporous silicon layer resulting from the etching of a Ti/Au prepatterned surface (WP1 geometry). An as-prepared HF/DMF mixture was used as electrolyte (pH = 4, conductivity = 2 mS/m). The etching current density was set at the constant value 91 mA/cm2 for 25 min. An etching voltage (Vetch)≈6V was necessary to sustain such current. A slow increment of Vetchas a function of the etching time was observed in all experiments. The metal mask was detached by the substrate before the end of the etching. As shown in Figure Figure1c,1c, the underetching of the masked area reduced the contact area between the metal layer and the silicon surface up to the detachment of the masking layer. The observed etching profile induces to suppose a catalytic role of the Ti/Au layer in the etching of the silicon in the close proximity of the metal layer, similar to what happened in the electroless metal-assisted chemical etching of silicon .
Figure Figure22 shows the SEM cross-sectional tilted view of Ti/Au prepatterned samples (geometry WP1 (a) and (b) and geometry WP2 (c)) etched used as an electrolyte of a 20-day-old HF/DMF mixture. As in the previous case the etching current density was set at the constant value 91 mA/cm2 for 25 min.
The different characteristics of the electrolyte (pH =6, conductivity = 30 mS/m) affected the etching parameters as well as the morphology of the resulting macroporous layer. A lower etching voltage, Vetch≈1.5 V, was necessary to sustain Jetch = 91 mA/cm2. Even in this case, a slow increase of the etching voltage as a function of the etching time was observed. A different etching profile was observed in close proximity of the masking layer. The metal layer was sufficiently adherent to the silicon substrate to sustain the etching process and the sample cutting for analysis.
Using as-prepared solutions as well as 20-day-old solutions, the pore distribution was affected by the presence of the Ti/Au mask. More specifically, the pores tend to grow in proximity of the metal edges. As a consequence, the pores grow with a random distribution in the direction parallel to the metal strips, and they grow with a regular spacing in the transverse direction. In the case of geometries with 2-μm pitches, a silicon wall in the middle of the unmasked area is formed. The depth and the high roughness of the top surface of this wall are due to the nucleation phase in which the initiation of the pore growth follows an almost-uniform etching of the surface. The dimension of this wall, as well as the pore diameter, depends on the applied current density. Figure Figure33 shows the top and cross-sectional SEM views of a MPS layer with metal electrodes of WP1 geometry. A 20-day-old HF/DMF mixture was used as electrolyte. The etching current density was set at the constant value of 26 mA/cm2 for 25 min. Both walls, the one sustaining the metal layer and the one between the electrodes, are larger with respect to the samples shown in Figures Figures11 and and22 because of the smaller etching current. On the contrary, the thinner perpendicular wall separating the pores does not show a strong dependence on the applied etching current.
Two-dimensional simulations of the interface silicon/mask/electrolyte were performed using COMSOL MultiphysicsⓇ (COMSOL AB, Stockholm, Sweden) to better understand the etching behavior of how the metal pattern geometry affects the macroporous layer morphology. To model the electrolyte/mask/silicon interface behavior, we considered a semiconductor/metal contact at the front side of a p-type wafer (1015 at./cm3) with a bias of -1 V applied on the conductor border at 30 μm from the semiconductor surface. Figure Figure3c3c shows the simulated cross-section of a metal pattern with geometry WP1. Holes current density (shading) and hole flux streamlines are shown. The flux streamlines follow the electric field lines. The simulation results show that, in the initial stage of the etching, the current density is higher at the metal electrode edges and converges with the field lines towards the metal strip corners. A direct consequence of the current density distribution is the experimentally observed growth of the pores in proximity of the mask edges. Even if during the nucleation phase pores start to grow in all the unmasked area, the pores on the corners of the mask are privileged by the higher current density that leads to a bigger growth rate. In case of smaller pitches, this behavior is reflected in the formation of the middle walls because the space between the metal strips is not sufficient to support the formation of another pore. In case of bigger spacing between the metal strips, the formation of the thicker middle wall is avoided. In fact, as clearly visible in Figure Figure44 that shows the SEM cross-sectional tilted view (a and b) and the simulated cross-section for a Ti/Au prepatterned surface with geometry WP3, the pores on the middle go further the nucleation phase. The higher current at the mask edges in the initial stage of the etching is reflected in a bigger nucleation and pore depth.
A new approach for the integration of metal electrodes on macroporous silicon was shown. The impact of etching parameters as electrolyte characteristics and etching current on the formation of silicon macropores on Ti/Au prepatterned substrate was analyzed. It was shown that the electrolyte consisting of an as-prepared HF/DMF mixture induces catalytic phenomenon at the silicon-metal interface and the consequent metal underetching and detachment. On the other hand, electrolytes consisting in 20-day-old HF/DMF mixtures, in which the DMF hydrolization gives rise to increased conductivity and pH, allow the integration of metal interdigitated electrodes on macroporous silicon. In both cases, the pore distribution and morphology are affected by the presence of the metal pattern. Pores grow faster in close proximity of the metal strip edges, leading to the formation of walls in the middle of the uncovered area for small pitches. The experimental results were in agreement with the COMSOL simulation results that showed a current density distribution influenced by the enhanced electric field at the metal edges.
Future work will be devoted to study the influence of the electrolyte composition, surface state, and metallic materials on the catalytic behavior of the metal layer, with the important consequence on the adhesion of the metal layer at the silicon surface.
The authors declare that they have no competing interests.
GS and MB conceived of the experiments and drafted the manuscript. GS conceived of the COMSOL simulations. LAF supervised the research activity. All authors read and approved the final manuscript.
GS is a PhD student of the Université Catholique de Louvain (UCL). MB is post doctorate research assistant at the UCL. LAF is an associate professor at the UCL. All authors are members of the group Sensors, Microsystems and Actuators Laboratory of Louvain (SMALL) chaired by LAF.
This work is supported by the GaSePoC project funded by INNOV’Iris and by the MINATIS project co-funded by the European Regional Development Fund (ERDF) and the Walloon region of Belgium. GS acknowledges the F.R.S.-FNRS (Belgium) for financial support.