3004
Chem. Mater. 2005, 17, 3004-3011
Sn2-2xSbxFexO4 Solid Solutions as Possible Inert Anode Materials in Aluminum Electrolysis V. A. Govorov, A. M. Abakumov,* M. G. Rozova, A. G. Borzenko, S. Yu. Vassiliev, V. M. Mazin, M. I. Afanasov, P. B. Fabritchnyi, G. A. Tsirlina, and E. V. Antipov Department of Chemistry, Moscow State UniVersity, Moscow 119992, Russia
E. N. Morozova and A. A. Gippius Department of Physics, Moscow State UniVersity, Moscow 119992, Russia
V. V. Ivanov Engineering and Technological Centre, Ltd., Krasnoyarsk, Russia
G. Van Tendeloo EMAT, UniVersity of Antwerp, Groenenborgerlaan 171, B-2020 Antwerp, Belgium ReceiVed October 22, 2004. ReVised Manuscript ReceiVed April 4, 2005
Single-phase samples of the Sn2-2xSbxFexO4 solid solution were prepared by solid-state reaction in air at 1300 °C for 0.26 e x e 0.66. The crystal structure of the Sn2-2xSbxFexO4 phases was studied by electron diffraction and X-ray powder diffraction. The compounds crystallize with the rutile type structure with a disordered arrangement of cations (P42/mnm space group, a ) 4.7127(6) Å, c ) 3.1595(4) Å, RI ) 0.020, RP ) 0.019 for Sn1.48Sb0.26Fe0.26O4 and a ) 4.6682(8) Å, c ) 3.1147(6) Å, RI ) 0.030, RP ) 0.022 for Sn0.68Sb0.66Fe0.66O4). The valence state of cations in the Sn1.48Sb0.26Fe0.26O4 and Sn0.68Sb0.66Fe0.66O4 samples was determined by means of 119Sn, 121Sb, and 57Fe Mo¨ssbauer spectroscopy. The presence of Sn(II), Sb(III), or Fe(II) low valent state species was not detected in the samples. Resistivity vs temperature measurements revealed a semiconducting behavior from room temperature up to 900 °C with about four orders in magnitude decrease of the resistivity. Solubility tests in the cryolite-alumina melt showed that the steady-state concentration of dissolved tin for the samples with x ) 0.26, 0.36 is significantly lower than that for SnO2. The electrocatalytic activity and solid degradation products of Sn2-2xSbxFexO4 are compared with those of another possible anode material (SnO2 with small amount of the CuO and Sb2O3 additives). These tests allow evaluation of the prospects of the Sn2-2xSbxFexO4 solid solutions as inert anode materials and to formulate an approach to further improvement of their degradation stability.
1. Introduction Further progress in electrolytic aluminum production involves the development of new materials to replace the commonly used carbon-based anodes. Up to now, the electrolytic cells for primary aluminum production are generally equipped with anodes fabricated from ground coke and pitch (Hall-He´roult process). Such anodes are consumed in the presence of oxygen, released upon electrolysis, and react with the components of fluoride-based cryolite-alumina melt, producing large amounts of greenhouse gases, such as CO, CO2, and fluorocarbons. Hence, use of inert anodes instead of carbon materials would solve a number of ecological and economical problems by reducing both the environmental pollution and the cost of raw aluminum. In the course of numerous efforts to replace carbon anodes used in aluminum production, oxide ceramic materials were considered as promising candidates along with metallic alloys * To whom correspondence should be addressed. E-mail: abakumov@ icr.chem.msu.ru. Tel: (095) 939-33-75. Fax: (095) 939-47-88.
and so-called “cermets” (metal particles dispersed in a ceramic matrix). The technological requirements concerning the anode material are rather rigid. It should be resistant to evolving oxygen and cryolite-alumina melt at ∼ 950 °C, i.e., in the melt, which is used as solvent for oxides. The maximum allowable wear rate for a commercial use is evaluated as 1.0 cm/year at a current density of 0.8 A/cm2. The electric conductivity and electrocatalytic activity with respect to oxygen evolution should be high enough to prevent extra losses of electricity. The material should be resistant to mechanical and thermal stress, and its components should not be poisonous for the raw aluminum, to avoid further purification. Tin dioxide SnO2 was considered as a promising material for inert oxygen-evolving anode because of its reasonable stability in a cryolite-alumina melt and high electrical conductivity of SnO2 with up to 1-2% of Sb2O3 and CuO.1,2 (1) Liu, Y. X.; Thonstad, J. Electrochim. Acta 1983, 28, 113.
10.1021/cm048145i CCC: $30.25 © 2005 American Chemical Society Published on Web 05/03/2005
Sn2-2xSbxFexO4 as Anode Materials in Aluminum Electrolysis
Figure 1. Projection of the rutile structure along the c-axis of the tetragonal unit cell. Infinite chains of the metal-oxygen octahedra can be seen. The unit cell is outlined.
Tin solubility in cryolite-alumina melt is among the lowest compared to other elements (150-500 ppm SnO2).3 However, these data concern the tetravalent tin, which, being reduced at the cathode, produces more soluble Sn(II) and metallic tin.4 The accumulation of the latter in aluminum (ca. 0.4 wt. % Sn at an anode wear rate of 1.7 cm/year)5 lowers the quality of produced aluminum even more than a higher content of less poisonous metals, such as Fe. Besides solubility, there are other important aspects which affect the SnO2-based anode prospects in the context of degradation and environmental pollution.6 Therefore, it would be tempting to chemically modify SnO2-based anode material to decrease possible degradation risks while still preserving the advantages provided by tin dioxide. Various metal cations can be hosted by the rutile structure of SnO27 (Figure 1). However, the choice of dopants is significantly restricted by economical considerations, specifications on aluminum contamination, and solubilities of the individual oxides in the cryolite-alumina melt. This work is devoted to the synthesis, structure analysis, and properties of the Sn2-2xSbxFexO4 solid solution as a material for an inert anode for aluminum production. Degradation and electrochemical tests are made to compare the newly fabricated materials with the conventional SnO2based anode. 2. Experimental Section Samples of rutile-based Sn2-2xSbxFexO4 solid solutions were prepared by solid-state reaction using SnO2, Sb2O3, and Fe2O3 (“Reakhim”, “chemically pure” purity grade) as starting materials. (2) Popescu, A.-M.; Mihaiu, S.; Zuca, S. Z. Naturforsch. 2002, 57a, 71. (3) Xiao, H.; Thonstad, J.; Rolseth, S. Acta Chem. Scand. 1995, 40, 96. (4) Issaeva, L.; Yang, J.; Haarberg, G. M.; Thonstad, J.; Aalberg, N. Electrochim. Acta 1997, 42, 1011. (5) Keller, R.; Rolseth, S.; Thonstad, J. Electrochim. Acta 1997, 42, 1809. (6) Thonstad, J. Aluminium Electrolysis. Fundamentals of the Hall-He´ roult Process; Aluminium Ferlag: Du¨sseldorf, 2001. (7) Baur, W. H. Acta Crystallogr. A 1956, 9, 515.
Chem. Mater., Vol. 17, No. 11, 2005 3005 The oxides were taken in required cation ratios, intimately mixed under acetone in an agate mortar, and pressed under 4 ton/cm2 to obtain pellets with a diameter of 8 mm. The pellets were annealed in alumina crucibles at 1100 °C for 50 h and at 1300 °C for 100 h in air and then furnace cooled. Some samples were prepared using FeSbO4. FeSbO4 was synthesized by heating a stoichiometric mixture of Sb2O3 and Fe2O3 at 1100 °C for 50 h in an oxygen flow. The prepared single-phase FeSbO4 was used for further preparation of the Sn2-2xSbxFexO4 samples at 1100 °C for 528 h in oxygen flow. Phase analysis and cell parameter determination were performed using X-ray powder diffraction with a focusing Guinier-camera FR552 (CuKR1 radiation, Ge was used as an internal standard). X-ray powder diffraction data for crystal structure determination were collected on a STADI-P diffractometer (CuKR1 radiation, curved Ge monochromator, transmission mode, linear PSD). CSD (crystal structure determination) program package was used for the Rietveld refinement from powder X-ray diffraction data.8 The profile reliability factors were calculated after background intensity subtraction. Samples for transmission electron microscopy were made by grinding the powder sample in ethanol and depositing it on a holey carbon grid. Electron diffraction patterns were obtained using a Philips CM20 electron microscope. Energy-dispersive X-ray analysis (EDX) was performed with a JEOL JSM 840A and JEOL JSM 5510 scanning microscopes equipped with a PGT IMIX and INCAx-sight attachments on the Sn(LR1), Sb(LR1), and Fe(KR) lines. Resistivity vs temperature dependences were measured in the range from room temperature up to 900 °C using a 4-probe technique. Solubility of the samples in a cryolite-alumina melt was tested in an open system at 950° for 3 h. Typical sample-to-melt mass ratio was ca. 1:50. The initial composition of the melt with a cryolite ratio (m(NaF)/m(AlF3)) of 2.68-2.69 was taken as follows (wt. %): Na3AlF6 17.96%, Na5Al3F14 60.77%, CaF2 5.43%, and NaF 15.84%, with 7 wt. % of alumina additive. The obtained values of steady-state concentrations should not, however, be considered as the equilibrium solubility ones, especially for elements such as Sn and Sb, which could interact with the melt components to produce volatile compounds. The concentrations of the Sn, Sb, and Fe oxides in the samples of the solidified melt were determined by X-ray fluorescence spectroscopy. A SPEKTROSKAN V scanning X-ray spectrometer (Spectron company, Russia), Eagle-III µProbe energy dispersive microfluorescence spectrometer (polycapillary optics; EDAX) and VRA-20 scanning X-ray spectrometer (Karl Zeis, GDR) were used for element analyses. Calibration curves for the constituting elements were obtained by testing standard samples, which were prepared by adding fixed amounts of salt solutions to the mixture of solid melt components, with subsequent heating (950 °C, 1 h). Calibration curves for Sn and Sb were close to linear, whereas for Fe a square polynomial fit was applied. Mo¨ssbauer spectra were recorded using a conventional spectrometer operating in constant acceleration mode. Upon measurements, Ca119mSnO3 or 57Co(Rh) sources were kept at room temperature while Ca121SnO3 source was cooled to 100 K. The hyperfine parameters were refined using least-squares fitting procedure. The isomer shift values refer to CaSnO3 or R-Fe reference absorber, both at 300 K, or Ca121SnO3 at 100 K. (8) Akselrud, L. G., Grin, Yu. N.; Zavalij, P. Yu.; Pecharsky, V. K.; Fundamentsky, V. S. CSD-UniVersal program package for single crystal and/or powder structure data treatment; 12th European Crystallographic Meeting, Abstract of Papers, Moscow, 1989; v. 3, p 155.
3006 Chem. Mater., Vol. 17, No. 11, 2005
GoVoroV et al.
Table 1. Unit Cell Parameters for the Sn2-2xSbxFexO4 Compounds composition
a, Å
c, Å
Sn1.48Sb0.26Fe0.26O4 Sn1.28Sb0.36Fe0.36O4 Sn1.08Sb0.46Fe0.46O4 Sn0.88Sb0.56Fe0.56O4 Sn0.68Sb0.66Fe0.66O4
4.7127(6) 4.7002(4) 4.6903(2) 4.6781(2) 4.6682(8)
3.1595(4) 3.1476(3) 3.1357(2) 3.1232(2) 3.1147(6)
For electrochemical measurements, a laboratory cell with undivided compartments was used, of the same geometry as described by R. Keller et al.5 About 500 g of the melt with the cryolite ratio of ca. 2.7 (alumina-free or containing 7 wt. % of alumina) was placed into graphite serving as the auxiliary electrode. For experiments in an alundum crucible, an auxiliary graphite rod (ca. 5 mm diameter, pretreated by cathodic polarization) was introduced. An aluminum reference electrode with tungsten connector was separated inside a porous alumina tube. All values of potential are referred to the Al2O3/Al redox couple in the same cryolite media. Stainless steel current connectors were mechanically attached to the anode samples and insulated with alumina cement, as described in ref 5. The Sn0.68Sb0.66Fe0.66O4 sample (a rod with a diameter of 8 mm), prepared by two-step pressing (30 and 100 kg/cm2), showed a pronounced open porosity (ca. 20%) prohibiting quantitative study of the oxygen evolution kinetics. The internal area of the solid/ melt interface was estimated from SEM characterization of the meltimpregnated sample with simultaneous local analysis. Only the pores of 1-2 µm and larger appeared to be permeable. High-density SnO2 ceramic samples with an open porosity of 0.1% (Gus Khrustal’nyi glass-works (Russia)) were used as reference in electrochemical tests (samples denoted below as “SnO2”). Their composition was measured using X-ray fluorescence spectroscopy (wt %): 97.5(6) SnO2, 0.81(9) Al2O3, 0.59(5) SiO2, 0.22(1) CuO, 0.22(2) MnO2, 0.19(1) ZnO, 0.14(1) Fe2O3, 0.13(1) CoO, 0.10(1) Cr2O3, 0.08(1) NiO. X-ray powder diffraction revealed only the rutile-type phase in these “SnO2” samples.
3. Results and Discussion 3.1 Synthesis, Crystal Structure, and Resistivity of the Sn2-2xSbxFexO4 Solid Solutions. The samples of the Sn2-2xSbxFexO4 solid solution were prepared for x values ranging from 0.06 to 0.96 with a step of ∆x ) 0.1, with single-phase materials being obtained only for 0.26 e x e 0.66. Samples with x < 0.26 contain an admixture of SnO2, whereas for x > 0.66 a secondary Fe2O3 phase appears. The X-ray powder diffraction patterns of the single phase samples were indexed in a tetragonal unit cell with cell parameters listed in Table 1. The cell parameters decrease with increasing x. The unit cell volume for the Sn2-2xSbxFexO4 solid solution depends linearly on composition inside the homogeneity range (Figure 2). However, above the upper boundary of the homogeneity range (x > 0.66) the unit cell volume still decreases with increasing x, but with a smaller slope than inside the homogeneity range. This indicates that the replacement of Sn by Sb and Fe occurs for x > 0.66, but the cation stoichiometry after synthesis deviates from the initial one. Martinelli et al.9 have demonstrated that the Sn2-2xSbxFexO4 solid solutions decompose at temperature of ∼1300 °C with release of volatile Sb4O6 and formation of an Fe2O3 admixture. With increasing x from 0.5 to 1.0 the decomposition temperature decreases from 1375 to 1318 °C. (9) Martinelli, A.; Ferretty, M. Mater. Res. Bull. 2003, 18, 1629.
Figure 2. Unit cell volume vs composition plot for the Sn2-2xSbxFexO4 samples. The single phase samples are marked as filled squares, the multiphase samples are marked as open squares. The line is a guide for the eye. Table 2. Initial and EDX Determined Cationic Compositions for the Sn2-2xSbxFexO4 Compounds x
initial composition, Sn:Sb:Fe, at. %
determined composition, Sn:Sb:Fe, at. %
0.26 0.36 0.46 0.56 0.66
74.0:13.0:13.0 64.0:18.0:18.0 54.0:23.0:23.0 44.0:28.0:28.0 34.0:33.0:33.0
74.5(4):12.0(5):13.5(3) 64.0(6):17.3(3):18.7(4) 53.8(3):22.6(3):23.6(3) 43.0(5):28(1):28(1) 33.6(3):33.3(3):33.1(2)
Although our samples were prepared at temperatures slightly lower than 1318 °C, depletion in Sb of the samples cannot be excluded. Evaporation of Sb4O6 would lead to the formation of Fe2O3 as an admixture phase in the samples with x > 0.66. The compositions of the single phase samples were tested by EDX analysis. The results, presented in Table 2, show that no significant deviation from the nominal compositions takes place upon thermal treatments. It should be noted, however, that the precision of the EDX measurements is limited by the overlap of the Sn(LR1) and Sb(LR1) lines having close energies (3.444 and 3.605 keV, respectively). The valence state of cations was studied by Mo¨ssbauer spectroscopy for two compositions: Sn1.48Sb0.26Fe0.26O4 and Sn0.68Sb0.66Fe0.66O4. The observed Mo¨ssbauer parameters (Table 3) show that only Sn4+, Sb5+, and Fe3+ ions are present in the samples. Moreover, no hematite phase (exhibiting a six-line room-temperature spectrum) was detected. Hence, the spectra allow the conclusion that no decomposition occurs upon preparation of the samples with 0.26 e x e 0.66, in agreement with the X-ray diffraction data. The homogeneity range of the Sn2-2xSbxFexO4 (0.26 e x e 0.66) solid solution appears to be significantly different from that reported by Martinelli et al.9 (0.6 e x e 1.0). In fact, we have prepared some Sn2-2xSbxFexO4 samples with 0 < x < 0.9 under the same experimental conditions (by annealing of SnO2 and the FeSbO4 precursor in oxygen flow at 1100 °C for 528 h). No single-phase samples were obtained for all x values. The samples consisted of a phase with the rutile structure and lattice parameters varying from a ) 4.7330(4) Å, c ) 3.1854(3) Å to a ) 4.6535(7) Å, c ) 3.0947(6) Å and unmodified SnO2. According to Martinelli
Sn2-2xSbxFexO4 as Anode Materials in Aluminum Electrolysis Table 3. 119Sn,
121Sb,
and
57Fe
Chem. Mater., Vol. 17, No. 11, 2005 3007
Mo1 ssbauer Parameters of the Sn1.48Sb0.26Fe0.26O4 and Sn0.68Sb0.66Fe0.66O4 Samplesa Sn1.48Sb0.26Fe0.26O4
Sn0.68Sb0.66Fe0.66O4
parameter (mm/s)
119Sn
121Sb
57Fe
119Sn
121Sb
57Fe
δ ∆ (eVzzQ5/2 for Sb) Γ
0.00 ( 0.02 0.59 ( 0.02 0.87 ( 0.02
0.04 ( 0.05 5.3 ( 1.1 2.8 ( 0.2
0.38 ( 0.01 0.72 ( 0.01 0.35 ( 0.01
0.00 ( 0.02 0.63 ( 0.02 0.88 ( 0.02
0.02 ( 0.05 4.6 ( 1.0 2.8 ( 0.2
0.38 ( 0.01 0.75 ( 0.01 0.36 ( 0.01
a Note: δ is the isomer shift; ∆ ) 1/2|eV Q | is the quadrupole splitting for the I ) 3/2+ and I ) 3/2- first excited state of 119Sn and 57Fe, respectively; zz 3/2 eVzzQ5/2 is the quadrupole-coupling constant for the I ) 5/2+ 121Sb ground state; and Γ is the full-width at half-maximum of the individual lines of a hyperfine pattern.
Figure 3. Fwhm vs 2θ plot for the Sn2-2xSbxFexO4 samples with x ) 0.26, 0.46, and 0.66. Miller indexes hkl are marked.
et al.,9 the samples with x e 0.5 upon cooling undergo a demixing with formation of two rutile-type phases: Snenriched and Sn-depleted. However, we have observed no signs of demixing in any samples with 0.26 e x e 0.66 prepared at 1300 °C. The demixing into two similar rutiletype phases would result in a broadening of the reflections on the X-ray powder diffraction patterns due to a small difference in cell parameters of the constituting phases. This should result in a different behavior of the full width at halfmaximum (fwhm) of the reflections with increasing 2θ angle for the demixed sample and the single phase one. As shown in Figure 3, the samples with x ) 0.26, 0.46, and 0.66 exhibit essentially similar fwhm(2θ) dependencies without clear difference between fwhms of particular families of diffraction peaks. This allows us to conclude that demixing and lattice distortions do not occur over the entire homogeneity range. One can suggest that the discrepancies between our results and those of Martinelli et al. originate from the different reactivity of the SnO2 used: smaller particle size of SnO2
could allow the solid-state reaction to occur at a temperature lower than 1300 °C, for example at 1100 °C. On the other hand, a higher synthesis temperature (1300 °C) would allow avoidance of demixing and extend the homogeneity range of the solid solution toward Sn-enriched compositions up to x ) 0.26. To detect any short- or long-range ordering of Sn, Sb, and Fe cations over the cationic positions of the rutile structure, the reciprocal lattice of the x ) 0.66 sample was studied by selected area electron diffraction. The electron diffraction patterns are shown in Figure 4. A complete indexing of these patterns was performed on a tetragonal primitive lattice with the cell parameters a ≈ 4.67 Å, c ≈ 3.12 Å. Only reflections with h + l ) even are present on the [010]* diffraction pattern, in agreement with the P42/mnm space group. The presence of the h00, 0k0, h,k * 2n and 00l, l * 2n reflections on the [001]* and [1h10]* patterns is caused by multiple diffraction. This was confirmed by observing intensity changes of these reflections upon tilting the crystal and by the absence of the h00, 0k0, h,k * 2n reflections in the [010]* pattern. No additional reflections or diffuse intensity which could be attributed to long- or short-range ordering were found on the electron diffraction patterns. This conclusion seems to be corroborated by the 57Fe Mo¨ssbauer parameters (see Table 3). The rather large values of quadrupole splitting ∆ relative to the high-spin Fe3+ (as evidenced by the observed isomers shifts) imply the presence of strongly distorted iron sites. The values of Γ, being similar for the Sn1.48Sb0.26Fe0.26O4 and Sn0.68Sb0.66Fe0.66O4 samples, are significantly larger than the value obtained with sodium nitroprusside (SNR) reference absorber (Γ ) 0.21 ( 0.01) mm/s). In fact, the 57Fe doublets could be fitted as a continuous distribution of quadrupole splittings with the same isomer shift. Such a distribution is indicative of a variety of local cationic surroundings of Fe3+ and therefore it is inconsistent with a short-range ordering in the studied oxides. As to 119Sn (121Sb) Mo¨ssbauer parameters, as well similar for both samples, the lack of noticeable difference in the Γ values could be accounted for by lower spectral resolution provided by the concerned isotopes as compared to that of 57Fe. The crystal structures of Sn1.48Sb0.26Fe0.26O4 and Sn0.68Sb0.66Fe0.66O4 were refined from X-ray powder diffraction data. The initial atomic coordinates were taken from the SnO2 structure.10 At the first step of the refinement, attempts were made to refine the occupancy factor for the cationic position. The scattering factor for this position was set to be equal to (10) Bolzan, A. A.; Fong, C.; Kennedy, B. G.; Howard, C. J. Acta Crystallogr. B 1997, 57, 373.
3008 Chem. Mater., Vol. 17, No. 11, 2005
Figure 4. Electron diffraction patterns of Sn0.68Sb0.66Fe0.66O4.
the Sn scattering factor and the occupancy of this position was refined at fixed atomic displacement parameter B ) 0.5 Å2. However, it is found that the occupancy value for the cationic position strongly correlates with the atomic displacement parameter of the oxygen atom. The dependence of the reliability factor RI on B(O) has a very broad minimum that hampers correct determination of both B(O) and the occupancy factor for the cationic position. For the final refinement cycles the occupancy factors for the cationic position were set according to the bulk composition of the samples. The preferred orientation along the [100] axis was taken into account using a March-Dollase function. The refinement resulted in reasonable values of the atomic displacement parameters for both positions and low values of the reliability factors. The crystallographic parameters, reliability factors, atomic coordinates, and the most relevant
GoVoroV et al.
interatomic distances for the Sn1.48Sb0.26Fe0.26O4 and Sn0.68Sb0.66Fe0.66O4 crystal structures are listed in Table 4. The experimental, calculated, and difference XRD patterns for these compounds are shown in Figure 5. The average cation-oxygen interatomic distance in the Sn2-2xSbxFexO4 solid solution decreases with increasing x from 2.065 Å for SnO2 10 to 2.042 Å for Sn1.48Sb0.26Fe0.26O4 and 2.018 Å for Sn0.68Sb0.66Fe0.66O4. This agrees well with the decreasing average ionic radius upon replacement of Sn4+ (r ) 0.83 Å, CN ) 6) by Sb5+ (r ) 0.75 Å, CN ) 6) and Fe3+ (r ) 0.785 Å, CN ) 6, high spin-state).11 Indeed, the charge balance in the investigated Sn2-2xSbxFexO4 solid solutions implies the substitution of two Sn4+ by one Sb5+ and one Fe3+, Sb2O3 being completely oxidized upon annealing in air, as evidenced by 121Sb Mo¨ssbauer spectra. It is worth noting that a replacement of Sn4+ by Sb3+ and Sb5+ up to 12 at. % Sb was observed in SnO2 samples prepared by hydrothermal synthesis at 250 °C.12 It was shown that such a replacement results in an increase of the unit cell volume, in contrast to what was observed for the Sn2-2xSbxFexO4 solid solution. Samples of the Sn2-2xSbxFexO4 (0.26 e x e 0.66) solid solution show a semiconducting behavior of resistivity vs temperature in the range of 20-900 °C (Figure 6). The values of energy gap estimated from the slopes of Arrhenius plots (see Figure 6) are close to ∼0.2 eV. These values are more than 1 order of magnitude lower than typical gap values in dielectric oxides (∼3 eV).13 This result points to the existence of a localized energy level (e.g., d-level of Fe) located in the vicinity of the bottom of the conduction band. Thermal activation of electrons from this level could be the main mechanism of electric conductivity in the Sn2-2xSbxFexO4 compounds. Typical resistivity values at 200 and 900 °C are, respectively, 15 kOhm‚m and 0.3 Ohm‚m for x ) 0.66, 12.8 kOhm‚ m and 0.14 Ohm‚m for x ) 0.46, and 7.5 kOhm‚m and 0.045 Ohm‚m for x ) 0.36. Even the lowest value is 2 orders of magnitude higher than that observed for the high-density “SnO2”, but the resistivity of the Sn2-2xSbxFexO4 solid solutions is not a crucial factor for their prospects as the anode materials. Moreover, this difference could be reduced by densification. For example, a densification of the samples used for electrolysis tests (x ) 0.66) results in a resistivity drop by 2 orders of magnitude, from 0.3 Ohm‚m to 0.004 Ohm‚m. 3.2 Interaction with the Melt under an Open Circuit. Preliminary tests on dissolution in cryolite-alumina melt revealed a satisfactory chemical stability of the solid solutions. After an interaction with the melt at 950 °C for 3h, no visible changes in the tested pellets were observed. Table 5 presents the concentrations of constituent oxides in the melt. In the right column, the unit cell parameters, as determined for the samples taken from the surface areas of the pellets after tests, are listed. Data for the “SnO2” sample and reference Fe2O3 and Sb2O3 oxides, measured under the (11) Shannon, R. D. Acta Crystallogr. A. 1976, 32, 751. (12) Grzˇeta, B.; Tkalcˇec, E.; Goebbert, C.; Takeda, M.; Takahashi, M.; Nomura, K.; Jaksˇic´, M. J. Phys. Chem. Solids 2002, 63, 765. (13) Cox, P. A. Transition Metal Oxides; Clarendon Press: Oxford, 1995.
Sn2-2xSbxFexO4 as Anode Materials in Aluminum Electrolysis
Chem. Mater., Vol. 17, No. 11, 2005 3009
Table 4. Selected Parameters from Rietveld Refinement of X-ray Powder Data for the Sn2-2xSbxFexO4 Solid Solutions formula space group a, Å c, Å V, Å3 Z F, g/cm3 2θ range, step (deg) RI, RP preferred orientation parameter cationic position: occupancy x, y, z, B (Å2) anionic position: occupancy x, y, z, B (Å2) d (cation-O), Å
Sn1.48Sb0.26Fe0.26O4 P42/mnm 4.70557(3) 3.15451(3) 69.848(2) 2 6.746(2) 21-118, 0.01 0.020, 0.019 1.83(2)
Sn0.68Sb0.66Fe0.66O4 P42/mnm 4.66455(4) 3.10913(3) 67.649(2) 2 6.348(2) 21-118, 0.01 0.030, 0.022 1.89(2)
74% Sn, 13% Sb, 13% Fe 0, 0, 0, 0.69(2)
34% Sn, 33% Sb, 33% Fe 0, 0, 0, 0.32(2)
100% O 0.2998(5), )x, 0, 0.72(8) 2 × 1.995(2), 4 × 2.065(2)
100% O 0.3015(6), )x, 0, 0.35(9) 2 × 1.989(3), 4 × 2.033(2)
Figure 6. Resistivity vs. temperature dependences for the Sn2-2xSbxFexO4 samples with x ) 0.36 (squares) and 0.46 (triangles). Insert presents the Arrhenius plot for the same data. Table 5. Concentrations of Dissolution Products in the Cryolite-Alumina Melt and Unit Cell Parameters after Solubility Tests at 950 °C for 3 h for the Sn2-2xSbxFexO4 Solid Solutions concentration, ppm Figure 5. Experimental, calculated, and difference X-ray diffraction patterns for Sn1.48Sb0.26Fe0.26O4 (a) and Sn0.68Sb0.66Fe0.66O4 (b).
same conditions, are given for comparison. It should be noted that no straightforward dependence of the concentration on the initial cationic composition has to be expected since the dissolution equilibrium is hardly established. Nevertheless, a tendency of increasing solubility with increasing x can be noticed for Fe and Sb. The samples with x ) 0.26 and 0.36 demonstrate lower concentration of SnO2 in the melt as compared to the values for the “SnO2” sample. For all samples of the solid solution, the concentrations of the simultaneously present Fe and Sb oxides appear to be lower as compared to the solubilities of the individual oxides. 3.3 Electrochemical Behavior. A pronounced difference of the Sn0.68Sb0.66Fe0.66O4 and “SnO2” samples was revealed by comparing the open circuit potential (OCP) values (Figure 7). For the former sample, OCP was stable and close to 2 V, whereas for the latter the OCP value was unstable and never exceeded 1.40-1.65 V. OCP values of 1.60-1.72 V were reported in ref 1 for less compact SnO2 samples with CuO and Sb2O3 additives. From the anodic polarization curves presented in ref 2, even lower OCP value of ca. 1.2
composition
SnO2
SnO2 Fe2O3 Sb2O3 Sn1.48Sb0.26Fe0.26O4
60 34
134
200 17
Sn1.28Sb0.36Fe0.36O4
22
133
40
Sn1.08Sb0.46Fe0.46O4
75
147
48
Sn0.88Sb0.56Fe0.56O4
44
177
65
Sn0.68Sb0.66Fe0.66O4
78
190
60
Fe2O3
Sb2O3
cell parameters, Å
∼1100a a ) 4.7112(5), c ) 3.1581(4) a ) 4.7015(2), c ) 3.1484(2) a ) 4.6875(3), c ) 3.1336(5) a ) 4.6805(5), c ) 3.1249(5) a ) 4.6782(2), c ) 3.1218(3)
a This steady-state value was achieved after 3 h and remained unchanged for the next 3 h.
V for SnO2 (CuO, Sb2O3) can be deduced. Note that all these data concern the experiments in a graphite crucible. The following remarks can be made as to the nature of electrochemical corrosion processes responsible for the low OCP of “SnO2”. First, the prewave A appears at the steadystate anodic polarization curves when the potential current is shifted to higher values (solid curve 1 in Figure 7). Second, this wave decreases noticeably at the reverse scan (dashed curve 1′ in Figure 7). Finally, it is absent at the curves
3010 Chem. Mater., Vol. 17, No. 11, 2005
Figure 7. Steady-state anodic polarization curves for the high density “SnO2” sample measured in the presence of 7 wt. % of alumina (solid and dashed lines 1 and 1′) and in an alumina-free electrolyte (dotted lines 2 and 2′). Arrows show the direction of the potential shift. Dotted horizontal lines mark the OCP values for the materials under consideration.
measured in alumina crucible, i.e., in the absence of a large free graphite surface, and a significantly higher and rather stable OCP of 1.85 V appears under these conditions. We account for these findings by a fast open circuit reduction of the Sn(IV) species by gaseous CO produced by graphite oxidation, with subsequent formation of Sn(II) in the interfacial region. A similar wave was reported in ref 1 for the SnO2 + 2 wt. % Sb2O3 + 2 wt. % CuO anode material. The potential of ∼1.5 V nominally corresponds to the oxidation of Sn(II).4 A limiting diffusion current (region A in Figure 7), which is proportional to both Sn(II) concentration and geometric surface area, but not sensitive to µmsize porosity, was never observed for Sn0.68Sb0.66Fe0.66O4. Therefore we can conclude that the reductive corrosion occurs at least much slower than in tin dioxide; the quantity of Sn(II) accumulated before measurements (under OCP conditions) being therefore very low. Further studies are needed to clarify the peculiarities of tin oxide reduction by CO. However, our qualitative test shows the Sn0.68Sb0.66Fe0.66O4 solid solution to be more resistant to reduction by CO than “SnO2”. Region B in Figure 7 is related to the oxygen evolution. The ohmic loss values are comparable to the values reported in ref 1 for a similar electrode material. The slopes of the curves at overvoltages above 50-100 mV (without allowing for the ohmic loss) are also very close. The ohmic corrections remain reliable up to a current density of 0.3 A cm-2 (referred to the geometric surface area). For the region of these low current densities, we could trace logarithmic current vs potential plots (so-called Tafel plots). The slope of Tafel plot is an important electrochemical criterion, usually assumed to be determined by the nature of the limiting reaction stage. This slope is found to be of 0.12-0.13 V, as compared to 0.065-0.26 V for various samples studied in ref 1. Taking into account the low accuracy of ohmic correction related to the discussed values, we can consider them as similar. This means that the limiting stage is the same in both cases, which allows comparison of the Tafel plots extrapolated to zero overvoltage.
GoVoroV et al.
We applied this extrapolation procedure to determine the exchange current density, which can be considered as the rate constant expressed in electric units. This value allows estimation of the electrocatalytic activity. Exchange current densities of 30-250 mA cm-2 were found within the extrapolation accuracy (compared to 23-130 mA cm -2 for various samples in ref 1). Again, we can conclude that our “SnO2” sample has nearly the same electrocatalytic properties as the samples of SnO2 with small amounts of the CuO and Sb2O3 additives.1,2 Similar estimates are hardly possible for the porous Sn0.68Sb0.66Fe0.66O4 sample, because of its non-Tafel behavior observed at overvoltage higher than 0.05 V. Such a behavior is assignable to gas accumulation in pores with subsequent drastic increase of the resistance. However, we can provide a rough estimate of the exchange current density from the narrow region in the vicinity of the oxygen potential (note that the same estimate for “SnO2” is impossible because of the prewave discussed above). Thus, estimated values fall into a range of 10-50 mA cm-2, demonstrating the comparable catalytic activity of the Sn0.68Sb0.66Fe0.66O4 and “SnO2” samples. The dotted curves 2 and 2′ in Figure 7 correspond to an experiment in the alumina-free melt carried out in a graphite crucible. This test is important for predicting the behavior of the material under conditions of the anode effect, which frequently appears upon electrolysis because of lowering alumina concentration in the melt upon its consumption.6 It was difficult to avoid alumina cement dissolution, so we cannot exclude that the shift of the reverse curve toward less positive potentials could by due to the appearance of alumina in the melt. However, the overvoltage increase at a given current density, as compared to that of aluminacontaining melt, never exceeded 0.3 V, and no visible changes of the anode appeared after this scan. This result thus provides a qualitative demonstration of the stability of the rutile-type oxides under anodic effect conditions. Electrolysis for 5 h at 1.5 A cm-2 (per visible surface area) was performed to compare the Sn0.68Sb0.66Fe0.66O4 degradation products formed under open circuit and under anodic polarization. No increase of ohmic loss was observed during this short-time test. 3.4 Solid Degradation Products. X-ray diffraction investigation of the degradation products was performed for the samples taken from different zones of the anode. Zone I contacted with the melt, whereas zone II refers to the gasmelt-solid-phase boundary. Figure 8 shows the position of the strongest 110 reflection of the rutile structure for the probes taken (i) from the anode surface of zone I (1); (ii) from a 1-mm deeper, inner part of the anode at zone I (2); (iii) from zone II (3); and (iv) from the anode material before electrolysis (4). The significant shift of the reflection position toward the low angle side due to increasing lattice parameters upon degradation suggests that the depletion of the sample in both Sb and Fe occurs during the electrolysis, but not a replacement of Fe3+ by Al3+ having smaller ionic radius (r(Al3+) ) 0.67 Å, CN ) 6).11 The lattice parameters relative to the different probes are listed in Table 6. Using a linear regression for the x-cell volume dependence based on the
Sn2-2xSbxFexO4 as Anode Materials in Aluminum Electrolysis
Chem. Mater., Vol. 17, No. 11, 2005 3011
pyrochlore Sb6O13 were detected. These phases could appear at a first step of degradation, being then dissolved in the melt that accounts for their absence at the surface of zone I. 4. Conclusions
Figure 8. Positions of the 110 reflection of the rutile structure: (1) the surface of zone I; (2) the inner part of zone I; (3) the inner part of zone II; and (4) the initial Sn0.68Sb0.66Fe0.66O4 sample. The position of the 110 reflection of SnO2 is indicated for comparison by a vertical line. Table 6. Unit Cell Parameters for the Sn0.68Sb0.66Fe0.66O4 Sample after Anodic Polarization and the Estimated x Values in the Sn2-2xSbxFexO4 Formula probe
a, Å
c, Å
estimated x
Zone 1, surface Zone 1, 1 mm inside Zone 2, 1 mm inside
4.712(2) 4.686(2) 4.679(1)
3.171(2) 3.129(5) 3.125(3)
0.22 0.51 0.56
data of Table 1 (x ) 12.4-0.173 V) one can estimate the cationic compositions of the probes (see Table 6). These data reflect the depletion of the anode material in Sb and Fe, which shifts the composition toward the lower boundary of the homogeneity range of the Sn2-2xSbxFexO4 solid solution. In the inner part of zone I admixtures of Fe2O3 and
The example of the Sn2-2xSbxFexO4 solid solution demonstrates that the key properties of the ceramic anode material (stability in the cryolite-alumina melt, resistivity, and capability of catalyzing the oxygen evolution) can be adjusted through chemical substitution in the rutile-type structure. Such a chemical adjustment extends the field of search for new complex oxides meeting severe technological requirements imposed by the existing routine of the aluminum electrolysis. In this respect the Sn2-2xSbxFexO4 solid solution offers interesting prospects as promising anode material. Acknowledgment. This work was performed in frame of the RusAl - MSU project “Search for oxide materials for inert anodes in aluminum production”. A.M.A. and G.A.Ts. are grateful to the Russian Science Support Foundation for financial support. We are grateful to M. I. Borzenko, Z. V. Kuz’minova, and S. N. Putilin for participation in the experiments. Supporting Information Available: Crystallographic information files for subject compounds (cif). This material is available free of charge via the Internet at http://pubs.acs.org. CM048145I