DRIE process optimization to fabricate vertical silicon nanowires using gold nanoparticles as masks

Silicon nanowires have applications in various fields, e.g. vertical transistors, chemical or biological sensors, energy conversion, and storage devices. So far they have typically been obtained by bottom-up methods such as vapor–liquid–solid (VLS) synthesis, starting from metal nanoparticles. In this study, silicon nanowires are fabricated by dry reactive-ion etching using gold nanoparticles as a mask. Starting with the Bosch process, the cycle of SF6 plasma etching and C4F8 plasma deposition was optimized to control the quality of the resulting silicon nanowires.


Introduction
Deep reactive-ion etching (DRIE) is a highly anisotropic etching process used to create deep holes and trenches in silicon wafers with a high aspect ratio and close-to-vertical sidewalls. Originally developed for the bulk micromachining of micro-electromechanical systems (MEMS) [1,2], it is currently also used to create deep trenches for high-density capacitors in DRAM processing, and more recently for creating through silicon visa for 3D packaging [2][3][4].
The most common DRIE process is known as the Bosch process shown in [5,6]. It is a form of PECVD process and includes the cycle of SF 6 plasma etching and protecting the plasma of C 4 F 8 , each phase lasting a few seconds. The fluorocarbon passivation layer protects the entire substrate from further chemical attack and prevents uncontrolled etching. During the etching phase, the directional fluorosulfur ions which bombard the substrate etch away the passivation layer at the bottom of the trench faster than along the sidewalls. They subsequently sputter it off, exposing the substrate for the chemical reaction of silicon with the fluorosulfur species [5][6][7][8].
These etch/deposit steps are repeated many times in sequence resulting in a large number of very small isotropic etch steps taking place only at the bottom of the etched patterns. To etch through a 0.5°mm silicon wafer, for example, 100-1000 etch/deposit steps are needed. The twophase process causes the sidewalls to undulate with an amplitude depending on the process parameters. The cycle time can be adjusted: short cycles yield smoother walls, and long cycles yield a higher etch rate [5,6,9,10].

Experimental
A 3 μm diameter Au disk mask was created by standard photolithography, followed by electron-beam evaporation (see figure 1(a)). The mask thickness was about 25-30 nm.
Alternatively, Au nanoparticles (NPs) were created through the deposition of a thin film by electron-beam evaporation followed by rapid thermal annealing at 1100°C for 30 s. The Au NP size depended on the thickness of the deposited Au thin film. We varied the size of the Au NPs from several tens to several hundred nanometers (see figure 1(b)).
The silicon substrate was subsequently DRIE-etched through the Au mask (figure 2). Observation was carried out using a field-emission scanning electron microscope (FE-SEM) and a 3D microscope of the resulting silicon nanowires. Original content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.       The Bosch process, named after the German company Robert Bosch GmbH which patented the process [5], also known as pulsed-or time-multiplexed-etching, alternates repeatedly between two modes to achieve nearly vertical structures. Starting with a nearly isotropic plasma etching, the plasma contains some ions which attack the wafer from a nearly vertical direction. Sulfur hexafluoride (SF 6 ) is often used for the silicon. This is followed by the deposition of a chemically inert passivation layer-C 4 F 8 (Octa fluorocyclobutane) source gas, for example, yields a substance similar to teflon [5][6][7][8]. Figure 3 shows the relationship between the main parameters of the etching process. The arrows help visualize the influence of the process parameters in relation to each other [11,12].

Results and discussion
First, 3 μm diameter gold masks were used to determine the different ranges of RF power as well as the effect of bias voltage. This served as a reference to determine the parameters of the Bosch process for creating Si NWs using Au NPs as a mask.

Influence of the RF power
The RF power injected into the plasma was varied from 500 W to 100 W with the other parameters being kept constant (see table 1). Figure 4(a) shows that etching at 500 W did not allow the creation of Si NWs-only valley-like structures were    (figure 4(b)). Further reducing the RF power to 100 W allows the gold mask to be kept, and consequently, Si NWs were created ( figure 4(c)). The thickness of the hard mask is determinant to achieving the proper formation of Si NWs.

Influence of the number of etching cycles and duration
As found above, the RF power was kept at 100 W. Two sets of parameters were tested: (i) 150 loops with SF 6 etching for 4 s each, (ii) 300 loops with SF 6 etching for 2 s each. Figure 5 shows that (i) results in a more isotropic etching. The gold mask was fully etched away for (i) while it remained relatively intact in (ii). For equal etching duration, the Si NWs present a controlled height in (ii).  Reducing the etching time as well as increasing the number of etching loops improves the anisotropic etch for long SF 6 plasma total duration.

Influence of the bias voltage
Here, the bias voltage was increased from 50 V to 200 V while other parameters were kept constant as in table 2.
Observation under a 3D microscope (Sensofar, Spain) (figure 6) indicates that the bias voltage has an influence on the process leading to high anisotropy with increasing bias as shown in figure 7.
A higher bias voltage in the etching step is very effective in decreasing passivation layer removal from the sidewalls as it directs the etching species more vertically towards the bottom of the etched patterns [13][14][15].

Creating Si NWs with Au NP masks
The above process optimisation using 3 μm Au masks was applied to the fabrication of Si NWs with Au NP masks (a  The Si NWs obtained with sample D1 present an average diameter of 300 nm and an average height of about 3.1 μm ( figure 8). However, a certain degree of isotropic etching is evident with a sidewall slope of about 70°. Reducing this angle necessitates the reduction in RF power as well as in the SF 6 etching duration. Figure 9 presents the Si NWs obtained with sample D2, with the RF power at 10 W and the SF 6 etching duration at 5 s with the number of process loops and other parameters kept identical with D1. The sidewall slope improved to 85°but the average height was now about 2.5 μm, a consequence of the reduction of RF power and SF 6 etching duration. By increasing the total etching duration and increasing the etching loops to 500, the Si NW height is about 3.2 μm (figure 10).
Sample D4 used the same process conditions as D3, except for a doubling of the bias voltage ( figure 11). Although the height of the Si NWs was unchanged, the sidewall slope was significantly reduced. By increasing the number of etching loops to 900 (figure 12), 150-200 nm diameter Si NWs were obtained with a height of about 4 μm and a sidewall slope from 93°to 95°.

Conclusion
In this study, the DRIE process parameters were optimized to obtain silicon nanowires using a gold nanoparticle mask. In particular, the influence of RF power, the number of etching loops, the SF 6 etching duration and bias voltage were determined. We demonstrated the capability of this process for creating 150-200 nm diameter Si NWs with a height from 2.5 to 4 μm. The Si NWs walls are slightly tapered with a verticality of about 85°. The next step is to reduce the Si NW diameter using smaller nanoparticles that are a few tens of nanometer in size.