Behaviour of iron and titanium species in cryolite-alumina melts
Abstract
The solubility of divalent iron oxide in cryolite-based melts was studied. Both electrochemical and chemical techniques were employed. To ensure that only divalent iron was present in solution, the melt was contained in an iron crucible under an atmosphere of argon. The experimental work included investigation of the solubility as a function of alumina concentration, temperature and cryolite ratio (CR = NaF/AlF3 molar ratio). The solubility at 1020 ºC was found to decrease from 4.17 wt% Fe in cryolite to 0.32 wt% Fe in cryolite saturated with alumina. FeO and FeAl2O4 were found to coexist as solid phases in equilibrium with the melt at 5.03 wt% Al2O3; the former being the stable solid phase below this concentration and the latter at high alumina concentrations. The standard Gibbs energy of formation for FeAl2O4 from its oxide components at 1020 ºC was determined to be -(17.6 ± 0.5) kJ mol-1. The solubility of FeAl2O4 was found to increase from 0.25 wt% Fe at 981 ºC to 0.36 wt% at 1050 ºC in alumina-saturated melts. By assuming Henrian behaviour, the apparent partial molar enthalpy of dissolution of FeAl2O4 was found to be (64.8 2.5) kJ mol-1. Experiments involving varying cryolite ratio in alumina-saturated melts at 1020 ºC showed a maximum solubility of 0.62 wt% Fe at a cryolite ratio of five. Modelling indicated that divalent iron species were present as NaFeF3 in acidic melts (CR < 3), while Na3FeF5 and/or Na4FeF6 dominated in a basic environment (CR > 3).
The solubility of TiO2 in cryolite-alumina melts at 1020 ºC was measured. The analytical data showed that the titanium solubility decreased with increasing total oxide concentration, up to a concentration of ~3.5 wt% O, while it increased at higher concentrations. The solubility was found to be 3.1 wt% Ti and 2.7 wt% Ti, respectively, in cryolite and in alumina-saturated melts. Modelling indicated that the most probable titanium species are TiO2+ and TiO32-, which coexist in the solution; the former dominating at low alumina concentrations and the latter at high alumina concentrations. Unknown amounts of fluoride may also be associated with the titanium atoms. Determination of the solubility of TiO2 in alumina-saturated melts as a function of temperature showed that the solubility increased from 1.9 wt% Ti at 975 ºC to 2.8 wt% Ti at 1035 ºC. The apparent partial molar enthalpy of dissolution of TiO2 was found to be (88.3 ± 4.1) kJ mol-1, provided that Henry’s law holds.
The electrochemistry of divalent iron in cryolite-based melts was investigated by voltammetry, chronopotentiometry and chronoamperometry. A working electrode of copper was found to be best suited for the study of the reduction of Fe(II), while gold and platinum gave the best results under oxidising conditions. The reduction of Fe(II) ions was found to be diffusion controlled. The number of electrons involved was determined to be two. A discrepancy was observed between the diffusion coefficients obtained by the different techniques. The diffusion coefficient of Fe(II) in alumina-saturated melts at 1020 ºC was found to be DFe(II) = 3.0 x10-5 cm2 s-1 by voltammetry. Experiments performed in an electrolyte with industrial composition at ~970 ºC gave a slightly higher value for the diffusion coefficient. The oxidation of Fe(II) on a gold or a platinum wire electrode showed that the process was diffusion controlled, involving one electron. The reversible potential for the redox couple Fe(III)/Fe(II) was found to be more cathodic than the reversible potential for the oxygen evolution by 350 to 400 mV, depending on the solvent composition and on the temperature.
The electrochemistry of TiO2 in cryolite-alumina melts was studied by voltammetry. The deposition of titanium on tungsten was found to be a three-electron diffusion controlled process. The deposition peak increased with increasing titanium concentration. In alumina-saturated melts two waves were observed prior to the titanium deposition. The potential difference between the cathodic wave closest to the deposition peak and its corresponding oxidation peak indicated a diffusion controlled process that involved a one-electron charge transfer. However, in cryolite melts a single wave was observed prior to the titanium deposition. It is suggested that these cathodic waves might have been caused by underpotential deposition of titanium, and subsequent alloying with tungsten. It cannot be ruled out that redox reactions take place between tetravalent titanium and the titanium alloyed with tungsten, thereby forming trivalent titanium prior to the metal deposition.
In order to determine thermodynamic properties of FeAl2O4, a solid electrolyte galvanic cell was used. Cryolite was present in the half-cell containing FeAl2O4 to ensure that alumina of the alpha modification was in equilibrium with FeAl2O4. An oxygen ion conducting yttria-stabilised zirconia tube served as the solid electrolyte. The EMF was measured in the rage 1245 to 1343 K. By using literature data at higher temperatures, thermodynamic properties for the reaction Fe(s) + ½O2(g) + Al2O3(s,α) = FeAl2O4(s) could be calculated, i.e. ΔHº1600K = –(270615 ± 1387) J mol-1 and ΔSº1600K = -(56.759 ± 0.856) J K-1 mol-1. New thermodynamic data for FeAl2O4 were also calculated, and a predominance area diagram for solid iron phases at 1293 K was constructed. The standard potential of the redox couple Fe(III)/Fe(II) as a function of the alumina content was derived from the solubility data of Fe(II) obtained in the present work and literature data for Fe(III). When the standard potentials are put into context of the Hall-Héroult process, the results indicate that neither the CO2/CO anode gas nor the carbon anode itself can oxidise Fe(II) to Fe(III).
The mass transfer of the impurities Fe, Si and Ti between bath and aluminium in industrial Hall-Héroult cells was investigated. The experiments were performed in several types of cells with prebaked anodes. The impurities were added to the bath in the form of Fe2O3, SiO2 and TiO2. Bath and metal samples were collected periodically before and after the addition was made. With the criterion that the mass transport was diffusion controlled, a model involving first order reaction kinetics was used to calculate the mass transfer coefficients for transfer into the metal phase. Large scatter were observed in the obtained mass transfer coefficients, but the general trend seemed to be kFe > kSi > kTi. By averaging the data obtained, it was found: kFe = (10 ± 3) x 10-6 m s-1, kSi = (7 ± 3) x 10-6 m s-1, and kTi = (5 ± 2) x 10-6 m s-1.