Process Optimization to Avoid Perfluorocarbon Emission During Neodymium Rare Earth Electrolysis in Molten LiF-NdF₃-Nd₂O₃

The solvent was a LiF-NdF₃ mixture at eutectic composition (77–23 mol%), pretreated under vacuum (8.10⁻² mbar) by increasing temperature up to 600 °C for 6 days to dry the salts directly inside the electrolyser. Pure LiF, NdF₃ and Nd₂O₃ were provided by FoxChemical (99.99%).

Oxide content in the melt was controlled by using square wave voltammetry method previously developed in the laboratory and detailed in Ref. 14.

The working electrode was a 3 mm diameter graphite rod (SIGRAPHINE^® R6510 provided by SGL carbon). The immersion depth was determined after electrolysis to determine the electroactive surface area.

A molybdenum and glassy carbon rod (3 mm diameter) with a large surface area were used as auxiliary electrode and quasi-reference electrode respectively. The glassy carbon quasi-reference electrode was found to be stable during the electrolysis: however, all the I–E curves were referred to the solvent reduction (independent of the oxide content).

All electrochemical studies were performed using Autolab PGSTAT 302 N potentiostat controlled with NOVA 2.1 software.

Gas analysis were performed using TENSOR27 (BRUKER) IR-Spectrometer controlled with OPUS 6.5 software. IR-spectrometer was calibrated at 25 °C in order to measure CO₂ content inside the spectrometer chamber. Calibration curve was realized using a gas mixer (SERV'INSTRUMENTATION) with controlled Ar-CO₂ mixtures (0 − 20 000 ppm of CO₂), made by mixing pure Ar and CO₂ provided by Linde (99.995 vol%).

Figure 2 presents a typical IR-spectrum of anodic gases obtained during electrolysis at 93 mA.cm⁻² in LiF-NdF₃-Nd₂O₃ (1.4 wt%) at 900 °C.

Zoom In Zoom Out Reset image size

On this spectrum, only the characteristic CO_2(g) double bond at 2349 cm⁻¹ is observed. This result indicates that neither CO_(g) nor CF_X(g) were produced in this condition.

In order to verify that CO_2(g) is the only gas produced, theoretical CO_2(g) molar flux ($\dot{C{{O}_{2}}^{th}}$) was calculated according to the Faraday's law (see Eq. 7). As obtained in Eq. 8, this flux is proportional to the imposed current (Eq. 8).

$$\begin{eqnarray}&&Q=I* t=\,{n}_{C{O}_{2}}* \,4\,* F\end{eqnarray} \tag{ 7 }$$

$$\begin{eqnarray}&&{\dot{C{O}_{2}}}^{th}\,=\displaystyle \frac{{n}_{C{O}_{2}}}{t}=\,\displaystyle \frac{I}{4\,* \,F}\end{eqnarray} \tag{ 8 }$$

Where Q is the charge in C, I the current in A, t the time in s, n_CO2 quantity of CO_2(g) produced in mol and F the Faraday constant.

A calibration curve was obtained by varying the CO₂ molar fraction in the gas phase (0–3000ppm in Ar) and measuring the absorption peak surface area (S_peak). Then, experimental CO_2(g) flux was obtained by using the molar fraction of CO_2(g) (yCO₂) measurement:

$$\begin{eqnarray}&&yC{O}_{2}=\,\displaystyle \frac{\dot{C{O}_{2}}}{\dot{C{O}_{2}}+\dot{Ar}}=34.5{* S}_{{\rm{peak}}}\end{eqnarray} \tag{ 9 }$$

where $\dot{C{O}_{2}}$ and $\dot{Ar}$ are the molar flux of CO_2(g) and Ar (mol.s^–1) respectively.

The measured flux is then expressed:

$$\begin{eqnarray}&&\dot{{{\rm{CO}}}_{2}}=\,\displaystyle \frac{\dot{{\rm{Ar}}}{\,* \,\mathrm{yCO}}_{2}}{1-{{\rm{yCO}}}_{2}}=\displaystyle \frac{2.56\,{10}^{-2}{* \,S}_{{\rm{peak}}}}{1-34.5{* \,S}_{{\rm{peak}}}}\end{eqnarray} \tag{ 10 }$$

Experimental and theoretical CO_2(g) fluxes obtained for LiF-NdF₃-Nd₂O₃ (1.4 wt%) electrolysis at 900 °C for different current densities are reported in Fig. 3.

Zoom In Zoom Out Reset image size

This figure is divided in two parts:

• •  
  For i < 220 mA.cm⁻², only CO_2(g) is detected by IR-spectroscopy (see. inset a) and experimental CO_2(g) flux is equal to the theoretical flux,
• •  
  At i = 220 mA.cm⁻², two additional peaks are detected on IR-spectrum of inset (b) at 1283 and 1250 cm⁻¹ corresponding to CF_4(g) and C₂F_6(g) absorption bonds respectively. ⁹ After few minutes of electrolysis, current interruption occurred associated to an anode effect.

These results indicate that only CO_2(g) is electrochemically produced on carbon electrode in LiF-NdF₃-Nd₂O₃ until a critical value was reached. Beyond this critical value, named critical current density ${i}_{critical},$ CF_X(g) are also produced.

As no CO_(g) are electrochemically produced by carbon oxidation, to prevent its formation by Boudouard equilibria, cell atmosphere has to be renewed to limit contact between carbon anode and produced CO_2(g). Compared to GC traditional analysis, the analysis time is faster (150 vs 600 s); no Boudouard reaction occurs as inert material was used and CO wasn't detected.

Carbon oxidation into CF_X(g) is usually associated to oxide depletion nearby the electrode, indicating the solute ($NdO{F}_{5}^{4-}$) diffusion limitation.

This hypothesis was verified by using square wave voltammetry (SWV). According to the SWV method, for a diffusion limited reaction, the differential peak current density (δip) is linearly correlated to the square root of the frequency: ^15,16

$$\begin{eqnarray}&&\delta ip\,\left(A.c{m}^{-2}\right)=nFC\,\displaystyle \frac{1-{\rm{\Omega }}}{1+{\rm{\Omega }}}\sqrt{\displaystyle \frac{Df}{\pi }}\,with\,{\rm{\Omega }}=\,{e}^{nF{\rm{\Delta }}E/2RT}\end{eqnarray} \tag{ 11 }$$

Where n is the number of exchanged electrons, C the oxide concentration in mol.cm⁻³, D the oxide diffusion coefficient in cm².s⁻¹, f the frequency of the square wave signal in Hz, δE the square wave signal amplitude in V, R the ideal gas constant in J.K⁻¹.mol⁻¹ and T the absolute temperature in K.

Typical square wave voltammogram plotted in LiF-NdF₃-Nd₂O₃ (2.4 wt%) at 900 °C on graphite electrode at 100 Hz is presented on Fig. 4.

Zoom In Zoom Out Reset image size

A single peak, followed by a characteristic signal disturbance corresponding to gas bubbling, was observed at 0.82 V Ref and was associated to the carbon oxidation into CO_2(g). The inset of Fig. 4 exhibits a linear relationship between δip and $\sqrt{{\rm{f}}}$ in the 64–225 frequency range: the reaction is then diffusion limited. Thus, the critical current density corresponds to the diffusion-limited current density of NdOF₅ ⁴⁻ noted i_lim ($NdO{F}_{5}^{4-}$).

According to the first Fick's law, i_lim ($NdO{F}_{5}^{4-}$) is given by the expression:

$$\begin{eqnarray}&&{i}_{lim}\left(NdO{F}_{5}^{4-}\right)\,\left(A.c{m}^{-2}\right)=n* F* k* C\end{eqnarray} \tag{ 12 }$$

Where n is the number of exchanged electrons, k the mass transfer coefficient in cm.s⁻¹ and C the NdOF₅ ⁴⁻ content in mol.cm⁻³.

This relation indicates that oxide content and the temperature (though the mass transfer coefficient) are two important parameters influencing i_lim ($NdO{F}_{5}^{4-}$).

Measurement methodology

In order to evaluate i_lim ($NdO{F}_{5}^{4-}$), steady-state voltammogram was realized by measuring during imposed current density electrolysis the stable anode potential. If the applied current density is higher than diffusion-limited one, an anode potential gap should be observed, indicating simultaneous production of CF_x(g) leading to anode effect phenomenon. Figure 5 represents the resultant voltammogram of the LiF-NdF₃-Nd₂O₃ (1.4 wt%) system at 900 °C.

Zoom In Zoom Out Reset image size

This voltammogram is composed of two parts:

• •  
  For i < 210 mA.cm⁻², anodic potential increased with the imposed current density. A typical evolution of the applied potential vs time is presented in the inset (a) of Fig. 5 at 70 mA cm⁻² with a stabilized potential of around 1.2 V Ref.
• •  
  At i = 210 mA.cm⁻², a sudden increase of the potential was observed during the electrolysis from 2.0 up to 4.1 V Ref as observed on inset (b) of Fig. 5.

As detailed above, this evolution allows to determine i_lim ($NdO{F}_{5}^{4-}$) value which is around 210 mA cm⁻² in molten LiF-NdF₃-Nd₂O₃ (1.4 wt%) at 900 °C. This value agrees with the previous observation presented in Fig. 3 where CF_X(g) were produced in the same experimental conditions at around 220 mA.cm⁻².

Influence of Nd₂O₃ content and temperature

The established methodology to measure i_lim ($NdO{F}_{5}^{4-}$) was repeated for several Nd₂O₃ additions at the same temperature (900 °C) and results are reported in Fig. 6.

Zoom In Zoom Out Reset image size

The results highlight a linear relationship between i_lim ($NdO{F}_{5}^{4-}$) and the neodymium oxide content confirming that the oxidation of carbon into CO_2(g) is limited by the diffusion of NdOF₅ ⁴⁻. According to Fig. 6, at 900 °C, the influence of Nd₂O₃ content on i_lim ($NdO{F}_{5}^{4-}$) is expressed as follow:

$$\begin{eqnarray}&&\begin{array}{l}{i}_{lim}\left(NdO{F}_{5}^{4-}\right)\left(A.c{m}^{-2}\right)\\ =\,0.13\,\left[N{d}_{2}{O}_{3}\right]\left(wt \% \right)=290\,\left[NdO{F}_{5}^{4-}\right]\left(mol.c{m}^{-3}\right)\end{array}\end{eqnarray} \tag{ 13 }$$

The conversion of oxide content into mol.cm⁻³ was realized using the data provided by Hu et al. on the molten LiF-NdF₃ density ¹⁷ as a function of absolute temperature:

$$\begin{eqnarray}&&{\rho }_{LiF-Nd{F}_{3}}\left(g.c{m}^{-3}\right)=\,-0.0077* \,T\,(K)+13.893\end{eqnarray} \tag{ 14 }$$

According to Eqs. 12 and 13 at 900 °C, the NdOF₅ ⁴⁻mass transfer coefficient was calculated to be 1.5 × 10^–5 m s⁻¹.

This result is in same order magnitude as mass transfer coefficient determined for other compounds in molten fluorides: around 3–5 × 10^–6 m s⁻¹ between 550 and 700 °C for tritium fluoride in LiF-BeF₂, ¹⁸ or around 2–10 × 10^–6 m s⁻¹ between 550 and 700 °C for several fluoride metals (Np, Pa, Ce, Eu, La) in LiF-BeF₂. ¹⁹

The experiments were repeated in the 850 °C–1050 °C temperature range in LiF-NdF₃-Nd₂O₃ and the influence of the temperature on ${i}_{lim}\left(NdO{F}_{5}^{4-}\right)$ per wt% of Nd oxide is reported in Table I. The mass transfer coefficient k was thus calculated at 850, 900, 950, 1000 and 1050 °C and the obtained values follow an Arrhenius like relationship as evidenced in Fig. 7:

$$\begin{eqnarray}&&\mathrm{ln}\,{k}_{\left(m{s}^{-1}\right)}=0.3852-\,\displaystyle \frac{13350}{{T}_{\left(K\right)}}\end{eqnarray} \tag{ 15 }$$

This equation allows the determination of NdOF₅ ⁴⁻ mass transfer coefficient whatever the temperature in the 850 °C–1050 °C temperature range.

Table I. Influence of the temperature on the limiting current density per wt% of neodymium oxide in LiF-NdF₃-Nd₂O₃ melt.

T (°C)		850	900	950	1000	1050
	$\displaystyle \frac{{\boldsymbol{ilim}}\left({\boldsymbol{NdO}}{{\boldsymbol{F}}}_{5}^{4-}\right)}{\left[{\boldsymbol{N}}{{\boldsymbol{d}}}_{2}{{\boldsymbol{O}}}_{3}\right]}\left(\displaystyle \frac{{\boldsymbol{A}}.{\boldsymbol{c}}{{\boldsymbol{m}}}^{-2}}{ \% \,\mathrm{mass}.}\right)$	0.11	0.13	0.17	0.27	0.43

Zoom In Zoom Out Reset image size