The redox chemistry of niobium(V) fluoro and oxofluoro complexes in LiF-NaF-KF melts

The electrochemical behavior of niobium(V) fluoro and oxofluoro complexes in eutectic LiF-NaF-KF (FLINAK) melts at 700°C has been studied by cyclic voltammetry. The fluoro complexes NbF, introduced into the melt by the addition of K2NbF7, can be reduced to niobium metal in two reversible steps involving one and four electrons, respectively. At 700°C the diffusion constants of the fluoro niobate complexes involved in these reduction steps, i.e., NbF and were determined to be 8.3 X 106 and 3.4 X iO cm2/s, respectively. Titration with equivalent amounts of oxide ions, introduced as Na2O, leads to a conversion of NbF to oxofluoro complexes of the type NbOF3 and NbO2F. At 700°C the conversion of NbF to NbOF3 is not complete, and the degree of conversion is shown to depend stron ly on temperature. Thus, at 645°C the conversion is more nearly complete than at 700°C, while the presence of NbOF complexes cannot be identified in cyclic voltammograms obtained at 795°C. It is concluded that the degree of conversion decreases with increasing temperature. At Na20/K2NbF7 molar ratios equal to three, electroactivity is still observed in the melt, indicating the presence of solute secies. The products of reduction of the oxofluoro complexes have not been identified because the reduction of NbOF,° ions cannot be obtained without simultaneous reduction of Nb(IV)F°4 ions, and at Na2O/K2NbF7 molar ratios exceeding two, no deposits are obtained. The reduction of the oxofluoro complex NbO2F, and complexes formed at Na2O/K2NbF7 molar ratios exceeding two always proceed in one step. Introduction It has been known for some time that it is possible to plate with refractory metals, such as niobium and tantalum, by molten salt electrolysis. As solvents, both chloride and fluoride melts can be used. By using fluoride melts, plating with niobium is possible even when a certain amount of oxide is present in the molten salt bath.1 This is important from an industrial point of view, because the need for large amounts of expensive, very pure chemicals for the solvent melt can be avoided. Therefore, the electrochemistry of oxofluoro complexes of niobium is of considerable practical interest. Furthermore, this matter is also of interest in relation to investigations of the steps in the reduction of niobium fluoro complexes, as was discussed in a previous paper.' Many discrepancies concerning the electrochemical reduction mechanism of * Electrochemical Society Active Member. Nb(V) fluoro species appear in the literature.' In our previous publication it was demonstrated' that it is likely that a number of the earlier investigations came to incorrect conclusions, because the melts used for the experiments were contaminated with oxygen from the atmosphere or from the chemicals used. The experimental difficulties in avoiding oxide in the melts are greatest when very low concentrations of Nb(V) are used. In the present work special care has been taken to keep the O:Nb ratio sufficiently low, when there is no oxide added deliberately, and always to maintain a good control of this ratio. We have recently investigated the influence of oxide on the electrochemical processes of tantalum in eutectic LiFNaF-KF (FLINAK) melts.2 In this investigation,2 a simultaneous monitoring of the total oxide content in the melts was carried out by voltammetry using a glassy carbon working electrode. It was shown that the addition of oxide to a melt containing TaFt ions leads to the stepwise forDownloaded 28 Jun 2010 to 192.38.67.112. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp 1794 J. Electrochem. Soc., Vol. 143, No. 6, June 1996 The Electrochemical Society, Inc. mation of TaOFr and Ta02F1 complexes. In addition to TaFt, TaOFr was observed to reduce to tantalum metal in one multielectron reduction step. Concerning niobium, several papers on the reduction mechanisms both in pure fluoride,"3-7 and mixed chloridefluoride'-'2 melts have appeared in the literature. However, only a few of these"9"2'13 deal with the influence of adding oxide to the melts. From Raman spectroscopy, it is known that Nb(V) forms a number of oxofluoro complexes (e.g., NbOFr and NbO2Ffl in FLINAK melts.'4 In NaCl-KC1 melts with K2NbF, and oxide added, the same complexes seem to be formed." In this case, it has been found" that even a quite small oxide content in the melt disturbs the electrodeposition of niobium metal, causing formation of impure coatings containing niobates. In contrast to this, using FLINAK as a solvent melt, it seems to be possible to deposit pure niobium even with 0:Nb ratios up to one to one.' Thus important differences between the two types of melts caused by solvent effects must exist. In our previous paper,' preliminary results have already been given concerning the reduction of the above mentioned complexes in FLINAK melts. The present work comprises a more detailed study of the redox reactions taking place in FLINAK melts with K2NbF7 and with and without added oxide. We have extended the investigated 0/Nb range beyond three and have taken the influence of the temperature into consideration. Experimental Analytical grade alkali fluorides from Merck were purified separately in a platinum crucible by slow recrystallization from the molten state, as previously described.16 K2NbF7 was prepared as follows: hot solutions of Nb201 (from CERAC, "99.95% purity") and KF (from Merck, analytical grade) in hydrofluoric acid were mixed. A white precipitate formed. The precipitate was collected by filtration and was then recrystallized in 40 weight percent (w/o) hydrofluoric acid. As an alternative, after mixing solutions containing the desired stoichiometric amounts of niobium and potassium, the liquid phase was removed by evaporation. Analysis by infrared spectroscopy as well as by potentiometric determination of the fluoride content showed no difference between the products prepared in the two different ways. Na20 was prepared by heating analytical grade Na202 from Merck in an alumina crucible under vacuum at 600°C for 12 h. Analysis by titration with hydrochloric acid gave typically 98 w/o of the theoretical amount expected for Na20. After preparation and purification, the Na20 and the alkali fluorides were stored in glass ampuls sealed under vacuum. All handling and weighing was carried out in a glove box with an atmosphere of dry nitrogen (dew point approximately —45°C). The K2NbF7 salt was stored in closed polyethylene bottles in the same glove box. All voltammetric experiments were carried out in a closed furnace under an argon (99.99%) atmosphere. The furnace was equipped with a nickel tube with water cooled end covers. The upper end cover was a part of the cell, i.e., fixed to this cover were a holder for the crucible as well as three or four electrodes and a stainless steel tube for the addition of chemicals (Na20 or K2NbF7). The cell was constructed of stainless steel and alumina ceramics for electrical insulation. The setup has been described previously in more detail.' Glassy carbon crucibles (V25, Carbone-Lorraine) served as containers for the melts. The typical amount of melt used in each experiment was 65 g. The heating procedure before each experiment was as follows: after introduction of the salts into the furnace, the furnace chamber was evacuated to a pressure of about 0.1 mbar and heated to 400°C before a flow of argon gas was established. When the working temperature of 7 00°C was reached, the flow of argon gas was replaced by a small overpressure of argon gas (20 kPa), which was then maintained during the rest of the experiment by an automatic pressure controlling system also described previously.' In this work, it was observed that the lowest impurity level was reached when K2NbF7 was heated together with the alkali fluoride chemicals. Before additions of Na20 were made, the chemicals were pressed into pellets inside the glove box. These pellets were then broken into smaller pieces of the desired weight. In order to avoid contamination of the furnace atmosphere during the addition, an argon gas flow was established so that argon gas flowed out of the furnace when the lid of the addition tube was opened. A Schlumberger potentiostat, Solartron 1286, was used for the voltammetric measurements. Data were recorded on either a computer or an X-Y-recorder (Graphtec WX 3000). For high scan rates, a digital storage oscilloscope (Gould 1602) was used. Electrode materials.—Glassy carbon working electrodes (for oxide determinations) were obtained from State Research Institute of Graphite, Moscow, (1 mm in diam, Type SU-2 500). A platinum wire (1 mm in diam) was used as a quasi-reference electrode. The counterelectrode was a 4 cm2 platinum foil. Working electrodes were made of platinum (99.99%, 0.5 mm diam) or silver (99.9%, 0.125 mm in diam) and were obtained from Goodfellow Cambridge, Limited. No differences between voltammograms obtained using platinum and silver working electrodes were observed. This was also the case for melts to which no sodium oxide had been added. We mention this because it has been reported recently17 that in chloride melts (LiC1/KC1 eutectic) the chloro complexes of Nb(V) and Ta(V) are reduced to Nb(IV) and Ta(IV), respectively, by silver. We decided to use a simple platinum wire as quasi-reference electrode. In a previous investigation,' a reference electrode based on the Ni/Ni(II) redox couple was used. The reference melt in this electrode consisted of FLINAK saturated with NW2 and was kept inside a porous boron nitride tube, which acted as a diaphragm. However, during the present work it was found that boron nitride reacts with Nb(V) in FLINAK. [The boron nitride was also found to react strongly with Ta(V) in FLINAK.] This was concluded from the fact that the shape of the voltammograms was changed after immersion of this electrode, compared with the voltammograms obtained with the platinum quasi-reference electrode. Thus, even though this electrode might give stable potentials, its presence changes the melt system, and therefore it was not used. Results and Discussion In Fig. 1 cyclic voltammograms of a FLINAK melt containing 1.00 mole percent (m/o) K2NbF7 and different concentrations of oxide are shown. The oxide concentrations given in the figure legend have not been corrected for the residual oxide content of the melt. When prepared as described in the Experimental section, the residual oxide content in a FLINAK melt containing 1.00 m/o K2NbF7 is 0.03 m/o. This value has been determined from the height of the anodic waves on the voltammograms obtained using glassy carbon working electrodes. This method of oxide determination has been described earlier.2"8" At each change of the oxide concentration by the addition of controlled amounts of Na20, the total oxide concentration of the melt is increased slightly more than the amount added because of experimental difficulties. Thus, the oxide/Nb(V) molar ratios corresponding to the voltammograms shown in Fig. 1 actually vary from approximately 0 (0.03) to approximately 1. In Fig. 1, curve A shows the voltammogram of a FLINAK melt with 1.00 mb K2NbF7 added. Two reduction waves, R, and It7, with peak potentials at —250 and —975 mV, are observed. As is demonstrated later, It, is due to a one-electron process. Potentiostatic electrolysis at a cathodic potential corresponding to the reduction wave It2 results in deposition of niobium metal on the cathode. By Raman spectroscopy it has been shown that K2NbF7 dissolves in FLINAK forming NbF ions.'4 The reduction of NbFt ions to niobium metal at 700°C thus proceeds in two steps involving one and four electrons, respectively. Downloaded 28 Jun 2010 to 192.38.67.112. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp J. Electrochem. Soc., Vol. 143, No. 6, June 1996 The Electrochemical Society, Inc. 1795