Raman Analysis and Electrochemical Reduction of Silicate Ions in Molten NaCl–CaCl₂

Salt preparation was conducted in a dry Ar-filled glove box (less than 1 ppm O₂ and 1 ppm H₂O). NaCl and CaCl₂ powders (FUJIFILM Wako Pure Chemical Corp., reagent grade) were mixed in a eutectic composition (NaCl:CaCl₂ = 47.9:52.1 mol%). Pre-determined amounts of CaO (FUJIFILM Wako Pure Chemical Corp., 98.0%), SiO₂ (Sigma-Aldrich, 10–20 nm, 99.5%), and CaSiO₃ (Sigma-Aldrich, 200 mesh, 99%) powders were added to the eutectic mixture. CaO was dried at 1273 K for 2 h. SiO₂ and CaSiO₃ were added as received. The added amounts were 0, 0.5, and 1.0 mol% for CaO, 0 and 1.0 mol% for SiO₂, and 0 and 1.0 mol% for CaSiO₃. The mixture was loaded into a graphite crucible (Toyo Tanso Co., Ltd, IG-110 grade, o.d. 55 mm × i.d. 49 mm × height 150 mm) and dried under vacuum at 453 K for 2 d and then at 723 K for 1 d. After the temperature was increased to 1023 K and maintained for 1 d to sufficiently dissolve the additives, the salt was quickly sampled using a borosilicate glass tube (Pyrex^®, o.d. 6 mm × i.d. 4 mm).

Figure 1 shows a schematic drawing of the experimental apparatus used for Raman spectroscopy. The sampled salt was then loaded into a Pt pan (Rigaku Corp., o.d. 5 mm × height 2.5 mm) and placed in an airtight high-temperature stage (Japan High Tech Co., Ltd, 10042). After the mixture was heated to 1023 K, the ionic species of silicates was investigated using a micro-Raman spectrometer (Tokyo Instruments, Nanofinder 30) using a laser source with an excitation wavelength of 532 nm. Origin 2020 software was used to deconvolve the spectra using the Viogtian function.

Zoom In Zoom Out Reset image size

NaCl and CaCl₂ powders were mixed in the eutectic composition and loaded into a graphite crucible (Toyo Tanso Co., Ltd, IG-110 grade, o.d. 100 mm × i.d. 95 mm × height 120 mm). The crucible was placed at the bottom of a graphite vessel in an airtight Kanthal container. The eutectic mixture was dried under vacuum at 453 K for 2 d and at 723 K for 1 d. After the temperature was raised to 1023 K, pre-determined amounts of CaO (0, 0.5 or 1.0 mol%) and CaSiO₃ (1.0 mol%) powders were added to the eutectic mixture.

Figure 2 shows a schematic drawing of the experimental apparatus for the electrochemical reduction of silicate ions. All electrochemical measurements were conducted by a three-electrode method using an electrochemical measurement system (Hokuto Denko Corp., HZ-7000) in a glove box. As the working electrodes, a flag-shaped graphite plate (Toyo Tanso Co., Ltd, 3 mm × 3 mm × thickness 0.5 mm) was used for cyclic voltammetry and a graphite plate (Toyo Tanso Co., Ltd, 10 mm × 10 mm × thickness 0.5 mm) was used for electrolysis. The counter electrodes were glass-like carbon (Tokai Carbon Co., Ltd, diameter: 3.0 mm) for cyclic voltammetry and a graphite square bar (Toyo Tanso Co., Ltd, 5 mm × 5 mm) for electrolysis. A Si square bar (Furuuchi Chemical Corp., 5 mm × 5 mm, 10 N purity) was used as the quasi-reference electrode. The potential of the reference electrode was calibrated with respect to a dynamic Na⁺/Na potential, determined by cyclic voltammetry on a Mo wire (Nilaco Corp., diameter 1.0 mm) electrode.

Zoom In Zoom Out Reset image size

The obtained samples were washed in an 1 M HCl solution prepared from FUJIFILM Wako Pure Chemical Corp., reagent grade, 36 wt.% at 333 K for 15 min to remove the salt adhered to the deposits. The surfaces of the samples were observed using an optical microscope (Thanko Inc., DILITE30) and scanning electron microscope (SEM, Thermo Fisher Scientific Inc., Phenom Pro Generation 5). The deposits were also characterized by X-ray diffraction (XRD, Rigaku, Ultima IV, Cu-Kα, λ = 1.5418 Å, 40 kV, 40 mA).

Figure 3 shows original and deconvolved Raman spectra for molten eutectic NaCl–CaCl₂ containing different amounts of CaO and SiO₂ at 1023 K. The original spectra were shown with background subtracted. Within the wavenumber range of 700–1100 cm⁻¹, no bands were observed for the melts without SiO₂. For the melts containing 0.5 mol% CaO and 1.0 mol% SiO₂ (${r}_{{O}^{2-}/{\text{SiO}}_{2}}$ = 0.5), and that with addition of 1.0 mol% CaO and 1.0 mol% SiO₂ (${r}_{{O}^{2-}/{\text{SiO}}_{2}}$ = 1.0), a strong band was detected near 970 cm⁻¹ and several weak bands at about 820, 920, and 1050 cm⁻¹. The spectra were deconvolved using the Viogtian function. According to previous studies, ^39–41 the symmetric stretch vibrations of SiO₄ ⁴⁻, Si₂O₇ ⁶⁻, SiO₃ ²⁻ and Si₂O₅ ²⁻ are reported at 850–880, 900–920, 950–980, and 1050–1100 cm⁻¹, respectively. Therefore, bands at 920, 972 and 1053 cm⁻¹ are assigned to Si₂O₇ ⁶⁻, SiO₃ ²⁻ and Si₂O₅ ²⁻. The band at 825 cm⁻¹ in our data could be attributed to SiO₄ ⁴⁻ vibration. However, the wavenumber is slightly lower than the reported value, 850–880 cm⁻¹, for SiO₄ ⁴⁻. Wang et al. reported that the Raman spectra for CaO–SiO₂–CaCl₂ slags gradually shift to lower wavenumber with increasing chlorine content, and that chlorine atoms can also be substituted with oxygen atoms. ⁴² Thus, oxygen atoms of SiO₄ ⁴⁻ ion might be partially substituted with chlorine atoms according to the following reaction.

$$\begin{eqnarray}&&{{{\rm{SiO}}}_{4}}^{4-}+{{\rm{Cl}}}^{-}\to {{\rm{SiO}}}_{3}{{\rm{Cl}}}^{3-}+{{\rm{O}}}^{2-}\end{eqnarray} \tag{ 3 }$$

Similar fitting results were obtained for the melt with ${r}_{{O}^{2-}/{\text{SiO}}_{2}}$ = 1.0; there is a strong band at 968 cm⁻¹ for SiO₃ ²⁻, and three weak bands at 822, 920, and 1050 cm⁻¹ attributed to SiO₄ ⁴⁻, Si₂O₇ ⁶⁻, and Si₂O₅ ²⁻, respectively. As the dominant silicate ion was SiO₃ ²⁻ in the melts with ${r}_{{O}^{2-}/{\text{SiO}}_{2}}$ ≤ 1.0, the dissolution limit of SiO₂ is likely to be determined by the O²⁻content. In addition, the weak bands of SiO₄ ⁴⁻, Si₂O₇ ⁶⁻, and Si₂O₅ ²⁻ were observed because of the equilibrium reactions of SiO₃ ²⁻ and Si₂O₇ ⁶⁻.

$$\begin{eqnarray}&&4{{\rm{SiO}}}_{3}^{2-}\rightleftarrows {{\rm{Si}}}_{2}{{\rm{O}}}_{7}^{6-}+{{\rm{Si}}}_{2}{{\rm{O}}}_{5}^{2-}\end{eqnarray} \tag{ 4 }$$

$$\begin{eqnarray}&&{{\rm{Si}}}_{2}{{\rm{O}}}_{7}^{6-}\rightleftarrows {{\rm{SiO}}}_{4}^{4-}+{{\rm{SiO}}}_{3}^{2-}\end{eqnarray} \tag{ 5 }$$

As mentioned above, the ionic species of SiO₂ in melts with ${r}_{{O}^{2-}/{\text{SiO}}_{2}}$ ≤ 1.0 was mainly attributed to SiO₃ ²⁻. To further investigate the stable silicate ions in the melts with various ${r}_{{O}^{2-}/{\text{SiO}}_{2}}$ values (≥ 1.0), CaSiO₃ was used as the SiO₂ source. This is because the dissolution rate of CaSiO₃ is significantly higher than that of SiO₂, which enables the quick preparation of the melts with target compositions. By adding CaO as the O²⁻ source, melts with ${r}_{{O}^{2-}/{\text{SiO}}_{2}}$ = 1.0, 1.5, and 2.0 were prepared and analyzed by Raman spectroscopy. Figure 4 shows the original and deconvolved Raman spectra of the melts. Figure 4(a) was deconvolved to 3 bands at 920 cm⁻¹ for Si₂O₇ ⁶⁻, 972 cm⁻¹ for SiO₃ ²⁻, and 1050 cm⁻¹ for Si₂O₅ ²⁻. For the deconvolution of Fig. 4(b), the band near 830 cm⁻¹ was added and the spectra were deconvolved to 4 bands. Although the band for Si₂O₇ ⁶⁻ is weak, the coefficient of determination, R², for 4 bands is 0.9909, which is higher than that for 3 bands without Si₂O₇ ⁶⁻ band (R² = 0.9842). Therefore, Figs. 4b and 4c were deconvolved to 4 bands. The deconvolution results of the Raman spectra are presented in Table I. As the result, for the melt with ${r}_{{O}^{2-}/{\text{SiO}}_{2}}$ = 1.0, the dominant silicate ion is SiO₃ ²⁻, which is consistent with the melt with the same ${r}_{{O}^{2-}/{\text{SiO}}_{2}}$ prepared from CaO and SiO₂. When ${r}_{{O}^{2-}/{\text{SiO}}_{2}}$ increased to 1.5, the dominant silicate species was SiO₄ ⁴⁻, secondary dominant SiO₃ ²⁻, and minor Si₂O₇ ⁶⁻ and Si₂O₅ ²⁻. Although the ratio of O²⁻/SiO₂ = 1.5 corresponds to Si₂O₇ ⁶⁻ ion, the equilibrium reaction of Si₂O₇ ⁶⁻ (reaction 5) would occur because of its poor stability in the melt. With respect to the melt with ${r}_{{O}^{2-}/{\text{SiO}}_{2}}$ = 2.0, SiO₄ ⁴⁻ is the dominant silicate ion which is consistent with the composition calculated from the O²⁻/SiO₂ ratio.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 5 shows the cyclic voltammograms of a graphite electrode in molten NaCl–CaCl₂ containing different amounts of CaO and CaSiO₃. The black curves show the voltammograms before the addition of CaSiO₃ and CaO (blank). The small cathodic current observed around 0.5 V (vs Na⁺/Na) is likely due to the intercalation of Ca²⁺ into graphite. After the addition of 1.0 mol% CaSiO₃ (${r}_{{O}^{2-}/{\text{SiO}}_{2}}$ = 1.0), as shown in Fig. 5a, cathodic currents increased from 1.0 V, suggesting the electrochemical reduction of dominant SiO₃ ²⁻ that observed in the Raman data (Fig. 4a). A larger cathodic current peak at around 0.50 V is considered as the formation of Na–Si and/or Ca–Si alloys. Figure 5b shows the voltammogram measured after the addition of 0.50 mol% CaO and 1.0 mol% CaSiO₃ (${r}_{{O}^{2-}/{\text{SiO}}_{2}}$ = 1.5). A small cathodic peak observed at 0.95 V is likely due to the electrochemical reduction of secondary dominant SiO₃ ²⁻ that observed in the Raman data (Fig. 4b). Cathodic currents increased rapidly from 0.80 V, which seems to be due to the electrochemical reduction of dominant SiO₄ ⁴⁻. The larger currents compared to Fig. 5a suggest that diffusion coefficient of SiO₄ ⁴⁻ is larger than that of SiO₃ ²⁻. Since SiO₄ ⁴⁻ ions are known to have weaker interaction each other compared with SiO₃ ²⁻ ions, ³⁹ it is reasonable that the effective size of SiO₄ ⁴⁻ is smaller than that of SiO₃ ²⁻. A cathodic current peak at around 0.50 V is regarded as the formation of Na–Si and/or Ca–Si alloys, which is similar to that observed in Fig. 5a. A voltammogram of the melt containing 1.0 mol% CaO and 1.0 mol% CaSiO₃ (${r}_{{O}^{2-}/{\text{SiO}}_{2}}$ = 2.0) is shown in Fig. 5c. A rapid increase of cathodic current from 0.75 V is likely attributed to the electrochemical reduction of dominant SiO₄ ⁴⁻. The current peak at around 0.50 V is ascribed to the same reaction of Na–Si and/or Ca–Si alloys formation, similar to the other two melts in Figs. 5a and 5b.

Zoom In Zoom Out Reset image size

Based on the voltammetry results, potentiostatic electrolysis was conducted at 0.50, 0.70, and 0.90 V with a constant charge density of −14 C cm⁻² in molten NaCl–CaCl₂ containing (i) 1.0 mol% CaSiO₃ (${r}_{{O}^{2-}/{\text{SiO}}_{2}}$ = 1.0), (ii) 0.5 mol% CaO and 1.0 mol% CaSiO₃ (${r}_{{O}^{2-}/{\text{SiO}}_{2}}$ = 1.5), and (iii) 1.0 mol% CaO and 1.0 mol% CaSiO₃ (${r}_{{O}^{2-}/{\text{SiO}}_{2}}$ = 2.0). Optical images of the original graphite substrate and electrolyzed samples are shown in Fig. 6. Deposits in brown or dark brown color were observed at all samples except the sample obtained at 0.90 V in molten salt (iii). No significant deposits were observed at that sample.

Zoom In Zoom Out Reset image size

Figure 7 shows XRD patterns of an original graphite substrate and the electrolyzed samples. In Fig. 7a, the existence of Si was confirmed in the samples obtained at 0.70 V and 0.90 V in molten salt (i). Thus, the increase in cathodic current from 1.0 V in Fig. 5a is confirmed to be the electrochemical reduction of SiO₃ ²⁻ to solid Si.

$$\begin{eqnarray}&&{{\rm{SiO}}}_{3}^{2-}+4{{\rm{e}}}^{-}\to {\rm{Si}}\left(s\right)+3{{\rm{O}}}^{2-}\end{eqnarray} \tag{ 6 }$$

The XRD pattern at 0.50 V indicated the existence of CaSi₂, which was identified by the strongest peak at 32.0 degree (relative intensity 100), and the second-strongest one at 46.8 degree (relative intensity 42). Since CaSi₂ was identified at 0.50 V, the cathodic current peak at 0.50 V corresponds to the formation of CaSi₂.

$$\begin{eqnarray}&&2{\rm{Si}}\left(s\right)+{{\rm{Ca}}}^{2+}+2{{\rm{e}}}^{-}\to {{\rm{CaSi}}}_{2}\left(s\right)\end{eqnarray} \tag{ 7 }$$

$$\begin{eqnarray}&&2{{\rm{SiO}}}_{3}^{2-}+{{\rm{Ca}}}^{2+}+10{{\rm{e}}}^{-}\to {{\rm{CaSi}}}_{2}\left(s\right)+6{{\rm{O}}}^{2-}\end{eqnarray} \tag{ 8 }$$

In molten salt (ii), only Si deposits were confirmed for samples obtained at 0.70 and 0.90 V. As approximately 20% of SiO₃ ²⁻ was confirmed by Raman spectroscopy, as shown in Table I, the small cathodic peak at 0.95 V is considered to be the electrochemical reduction of SiO₃ ²⁻ to solid Si (reaction 6). The cathodic current increase from 0.80 V is attributed to the electrochemical reduction of SiO₄ ⁴⁻.

$$\begin{eqnarray}&&{{\rm{SiO}}}_{4}^{4-}+4{{\rm{e}}}^{-}\to {\rm{Si}}\left(s\right)+4{{\rm{O}}}^{2-}\end{eqnarray} \tag{ 9 }$$

In order to confirm the results shown above, thermodynamic data associated with the electrochemical reduction of SiO₃ ²⁻ and SiO₄ ⁴⁻ were calculated. Generally, the difference in reduction potential of SiO₃ ²⁻ and SiO₄ ⁴⁻ can be calculated from the Gibbs energies for the decomposition reactions, ΔG_d(l, in melt).

$$\begin{eqnarray}&&{{\rm{CaSiO}}}_{3}\left(l,{\rm{in}}\,{\rm{melt}}\right)\to {\rm{Si}}\left(s\right)+{\rm{CaO}}\left(l\right)+{{\rm{O}}}_{2}\left(g\right)\end{eqnarray} \tag{ 10 }$$

$$\begin{eqnarray}&&{{\rm{Ca}}}_{{\rm{2}}}{{\rm{SiO}}}_{4}\left(l,{\rm{in}}\,{\rm{melt}}\right)\to {\rm{Si}}\left(s\right)+2{\rm{CaO}}\left(l\right)+{{\rm{O}}}_{2}\left(g\right)\end{eqnarray} \tag{ 11 }$$

Here, CaSiO₃(l, in melt) and Ca₂SiO₄(l, in melt) are CaSiO₃ and Ca2SiO₄ dissolved into molten salt. However, these Gibbs energies can not be calculated from the reported data. On the other hand, the Gibbs energies for the decomposition reactions for solid CaSiO₃ and Ca₂SiO₄, ΔG_d(s), can be calculated from literature. ^43,44

$$\begin{eqnarray}&&{{\rm{CaSiO}}}_{3}\left(s\right)\to {\rm{Si}}\left(s\right)+{\rm{CaO}}\left(l\right)+{O}_{2}\left(g\right)\end{eqnarray} \tag{ 12 }$$

$$\begin{eqnarray}&&{{\rm{Ca}}}_{2}{{\rm{SiO}}}_{4}\left(s\right)\to {\rm{Si}}\left(s\right)+2{\rm{CaO}}\left(l\right)+{{\rm{O}}}_{2}\left(g\right)\end{eqnarray} \tag{ 13 }$$

Here, we assumed that the difference of the Gibbs energy, ΔG_d(l, in melt) − ΔG_d(s), is almost same for CaSiO₃ and Ca₂SiO₄, so that ΔG_d(s) could be used to estimate the difference in reduction potential of SiO₃ ²⁻ and SiO₄ ⁴⁻. The calculated data are shown in Table III. Since the standard Gibbs energy for CaSiO₃(s) is smaller than that for Ca₂SiO₄(s), the electrochemical reduction of SiO₃ ²⁻ would start from a more positive potential than SiO₄ ⁴⁻. From the Nernst equation for 4-electron reaction, the difference in the reduction potential of CaSiO₃(s) and Ca₂SiO₄(s) was calculated to be 0.16 V, which is approximately consistent with the results of the voltammetry. As the mixture of Si and CaSi₂ was detected at 0.50 V, the cathodic current peak at 0.50 V is attributed to the formation of CaSi₂ expressed in reactions 7 and 8.

Zoom In Zoom Out Reset image size

For samples obtained in molten salt (iii), the existence of pure Si was confirmed at only 0.70 V and the mixture of Si and CaSi₂ at 0.50 V. As only graphite was detected at 0.90 V, the increase in cathodic current from 0.75 V would correspond to the electrochemical reduction of dominant SiO₄ ⁴⁻ to solid Si (reaction 9), and the current peak at around 0.5 V is the formation of CaSi₂ (reaction 7, 8). The identified phases are summarized in Table II.

Figure 8 shows the surface SEM images of the deposits. In molten salt (i), wire-like Si, particle Si and dense CaSi₂ were obtained at 0.90, 0.70 V and 0.50 V, respectively. In molten salt (ii), wire-like Si with different shapes were deposited at 0.70 and 0.90 V, and particle Si was formed at 0.50 V. For samples obtained in molten salt (iii), wire-like Si was observed at 0.70 V and particle Si at 0.50 V. The change in morphology of reduced Si is likely due to the change of overpotential and current density. Generally, when the overpotential is low and current density low, electrodeposits become field-oriented isolated crystals type, such as wires. ⁴⁵ On the other hand, when the overpotential is higher and current density higher, the electrodeposits shift to unoriented dispersion type, such as particles. ⁴⁵ In molten salt (i), the potential of 0.90 V would correspond to the low overpotential and current density, which results in the formation of wire-like Si. The potential of 0.70 V is considered to be classified as the high overpotential and current density, resulting in the formation of particle Si. In molten salt (iii), the electrochemical reduction of SiO₄ ⁴⁻ starts from a more negative potential compared with SiO₃ ²⁻. Then, the potential of 0.70 V corresponds to the low overpotential and current density, giving wire-like Si. The potential of 0.50 V is classified as the high overpotential and current density, resulting in the formation of particle Si. In molten salt (ii), SiO₄ ⁴⁻ and SiO₃ ²⁻ are dominant and second dominant species, respectively. Therefore, by the same logic as above, wire-like Si is formed from SiO₃ ²⁻ at 0.90 V and from SiO₄ ⁴⁻ at 0.70 V. In the same manner, particle Si is electrodeposited from SiO₄ ⁴⁻ at 0.50 V.

Zoom In Zoom Out Reset image size

All the results shown above indicate that the electrodeposition of Si starts from approximately 0.2 V more positive for SiO₃ ²⁻ compared with SiO₄ ⁴⁻. Considering that the formation of Ca–Zn, which reduces the current efficiency of Si–Zn alloy production, has been confirmed to proceed from 0.40 V, SiO₃ ²⁻, which has a wider potential range to Ca-Zn alloy formation than SiO₄ ⁴⁻, is more suitable for the Si–Zn alloy production. Therefore, molten salt (i), in which SiO₃ ²⁻ is dominant and molten salt (ii), which contains SiO₃ ²⁻ as a secondary dominant species, are suitable for the electrodeposition of Si at a liquid Zn cathode.