Several SiNWs electrodes with each different SiNWs length have been made. The SiNWs were grown in a CVD reactor by VLS method via gold catalysis on highly doped n-Si (111) substrate (doping level (i.e., the number of doping atoms per cubic centimeter of materials, N _d = 5.10¹⁸ cm⁻³). Gold colloids with size of 50 nm are used as catalysts, H₂ as carrier gas, silane (SiH₄) as silicon precursor, phosphine (PH₃) as n-doping gas, and HCl as additive gas. As shown in our previous work 19 20 21 , the use of HCl in our process enables us to reduce the gold surface migration. Thus, the nanowires (NWs) morphology is improved and their length is not limited.

Prior to the growth, the substrates surface has been prepared by successive dipping in (a) acetone, isopropanol and caro (H₂SO₄/H₂O₂, 3:1) to remove organic impurities followed by (b) 10% HF and NH₄F solution to remove the native oxide layer. Then, 50-nm gold colloids are deposited on the surface with 10% HF from an aqueous gold colloid solution (British Bio Cell International Ltd., Llanishen, Cardiff, UK).

The growth has been performed at 600°C, under 3 Torr total pressure, with 40 sccm (standard cubic centimeters) of SiH₄, 100 sccm of PH₃ gas (0.2% PH₃ in H₂), 100 sccm of HCl gas and 700 sccm of H₂ as supporting gas 19 . Our VLS-CVD method enables an easier control of SiNWs parameters (length, density, diameter, doping type, and doping level) and growth on low cost substrates. The doping level of the SiNWs is managed by the pressure ratio: dopant gas/SiH₄. In our setup the ratio can vary from 10⁻⁶ to 10⁻² to obtain doping level from Nd ≈10¹⁶ to ≈10²⁰ cm⁻³ 20 . It was checked by resistivity measurements in four probes configuration 21 22 . The SiNWs length is monitored by the gas injection time. The growth rate is about 500 nm/min under these conditions.

SiNWs morphologies are checked by scanning electron microscopy (SEM) before and after electrochemical cycling. SiNWs density is estimated by counting the number of gold colloids per square centimeters on several SEM images.

All experiments were performed in a glove box at room temperature. The electrolyte was 1 M NEt₄BF₄ (Fluka Chemika, Buchs, Switzerland) in propylene carbonate (Sigma Aldrich, St. Louis, MO, USA). Nanostructured silicon (n-SiNWs) and bulk silicon substrates (n-Si) were always directly used as electrodes.

Micro-ultracapacitors with two identical n-SiNWs electrodes were built by clipping the aluminum current collector, silicon electrodes (Si = 1 cm²), and glass fiber paper as separator. The n-SiNWs with several lengths (5, 10, and 20 μm) were used. In the same way, a micro-EC with two bulk n-Si substrate was built. Electrochemical instruments consisted of Potentiostat/galvanostat equipped with low current channels (VMP3 from Biologic with Ec-Lab software, Slough Berkshire, UK). All SiNWs/SiNWs micro-ultracapacitors were first characterized by cyclic voltammetry with a 100 mV s⁻¹ scan rate between 0.01 and 1 V (Figure 1). Then, ten galvanostatic charge/discharge cycles were performed at ±5 and ±10 μA cm⁻² (Figure 2) and finally, all devices were cycled for 250 galvanostatic charge/discharge at 5 μA cm⁻² (Figure 3).

From 50-nm gold colloids, a NWs density of ≈3.10⁸ NWs cm⁻², with diameters of ≈50 ± 5 nm, has been obtained for all electrodes. With growth times of 10, 20, and 40 min, electrodes with SiNWs lengths of 5 μm ± 10 nm (a in Figure 4), 10 μm ± 10 nm (b in Figure 4), and 20 μm ± 20 nm (c in Figure 4), respectively, have been obtained. Gold colloids are kept on top of the SiNWs (inserted in a in Figure 4) without any influence on the electrochemical behavior. SiNWs length, diameter, and density determination from the SEM images provides an estimation of SiNWs volume and by calculation with silicon density, an estimation of SiNWs mass (respectively, ≈12, 24, and 48 μg for 5, 10 and 20 μm NWs). The developed surface cannot be accurately determined from the SEM images. With the dopant/SiH₄ ratio equal to 4.10⁻³ for all samples, we obtain a doping level of 4.10¹⁹ cm⁻³ 21 .

All devices show a quasi-ideal capacitive behavior. Cyclic voltammetry curves are rectangular. Galvanostatic charge/discharge curves are triangular and symmetrical, which indicates that only very few losses occur between the charge and the discharge. An unexpected lower voltage for 1 M NEt₄BF₄ in PC is used to stay in the system electrochemical stability window (ESW) and avoid side reactions. In fact, the system ESW is smaller than the one obtained on platinum for this electrolyte due to silicon oxidation at a potential below the electrolyte one 15 16 .

Device capacitance increase with the SiNWs length can be seen on cyclic voltammetry curves (Figure 1) and on galvanostatic charge/discharge curves (Figure 2). In fact, for the first one, capacitance is proportional to the current density difference inside the two curves (Formula 1) and for the second one it is inversely proportional to the discharge slope (Formula 2). Capacitance values have been calculated from the galvanostatic charge/discharge experiments for both current densities and reported in Table 1.

Formula 1

C = Δ j 2 * v b ,

with Δj as the current density differences inside the cyclic voltammetry curve and v as the scan rate.

Formula 2

C = j Discharge slope of galvanostatic charge-discharge curve ,

with j the current density used for the galvanostatic charge/discharge.

Devices with the same SiNWs length show similar capacitance values for both current densities. As noticed on the curves, capacitance increases with SiNWs length. This increase is proportional to the length increase between 5 (≈3.5 μF cm⁻²) and 10 μm SiNWs (≈7 μF cm⁻²), but not between 5 and 20 μm (≈9.5 μF cm⁻²). This can be explained by accessible surface losses due to SiNWs constriction when substrates are stacked together. New devices avoiding this constriction will be designed and evaluated. Although previous works on the use of silicon-based electrodes for supercapacitor 10 11 12 13 14 15 reported better capacitance values, the SiNWs length influence in two electrodes devices has never been investigated. Moreover, it could be improved up to the capacitance wanted by increasing the SiNWs length and density and by improving the device design. In fact, SiNWs growth by CVD enables to tune the NWs lengths without any limitation.

Choi et al. 10 reported the use of porous SiNWs as electrode for supercapacitor in such devices but with Li⁺ containing electrolyte. Their capacitance is expressed only in force per gram, so no accurate comparison with our results is possible. Desplobain et al. 12 have obtained devices with 320 μF cm⁻² capacitance by using gold-coated porous silicon but in aqueous electrolyte. SiNWs coated with NiO 13 14 or SiC 15 shows promising performances and cycling ability, but silicon is not the active material and their performances have not been evaluated in the two electrodes devices.

After 250 cycles at ±5 μA cm⁻², each device shows less than 2% capacitance loss (1.8% for 20-μm SiNWs, 0.5% for 10-μm SiNWs, 0.7% for 5-μm SiNWs, and 0.5% for bulk silicon) (Figure 3). Whatever the length, SiNWs are stable after these cycling experiments, as observed on post-experimental SEM images (Figure 4). The top bending that can be observed is due to electrostatic forces occurring during the sample washing with organic solvents before the SEM observation.

Due to the moderate surface capacitance, 20-μm SiNWs-based microdevice only stores 5 μJ cm⁻², i.e., few milliwatts per square centimeter. However, the interest of the device is more directed toward the power density which reaches 1.4 mWcm⁻², which is close to the one of the 5-μm thick activated carbon supercapacitor (5 mW cm⁻²) 7 .