Microstructural Aspects of the Degradation Behavior of SnO2-Based ...

Journal of The Electrochemical Society, 157 共5兲 C178-C186 共2010兲

C178

0013-4651/2010/157共5兲/C178/9/$28.00 © The Electrochemical Society

Microstructural Aspects of the Degradation Behavior of SnO2-Based Anodes for Aluminum Electrolysis Sergey Yu. Vassiliev,a Veronika K. Laurinavichute,a Artem M. Abakumov,a,b,z Vitaliy A. Govorov,a Evgeny B. Bendovskii,a Stuart Turner,b Alexander Yu. Filatov,a Vadim P. Tarasovskii,c Andrey G. Borzenko,a Anastasiya M. Alekseeva,a and Evgeny V. Antipova a

Department of Chemistry, Moscow State University, 119992 Moscow, Russia EMAT, University of Antwerp, B-2020 Antwerp, Belgium ZAO NTTs Bakor, Sherbinka, 142171 Moscow, Russia

b c

The performance of SnO2 ceramic anodes doped with copper and antimony oxides was examined in cryolite alumina melts under anodic polarization at different cryolite ratios, temperatures, times, and current densities. The corroded part consists of a narrow strong corrosion zone at the anode surface with damage of the intergrain contacts and a large increase in porosity, a wider moderate corrosion zone with a smaller porosity increase, and a Cu depletion zone, where the ceramic retains its initial microstructure and a slight porosity increase occurs due to the removal of the Cu-rich inclusions. Mechanical destruction of the anode was never observed in the 10–100 h tests. A microstructural model of the ceramic was suggested, consisting of grains with an Sb-doped SnO2 grain core surrounded by an ⬃200 to 500 nm grain shell where SnO2 was simultaneously doped with Sb and Mn+ 共M = Cu2+,Fe3+,Al3+兲. The grains were separated by a few nanometers thick Cu-enriched grain boundaries. Different secondary charge carrier 共holes兲 concentrations and electric conductivities in the grain core and grain shell result in a higher current density at the intergrain regions that leads to their profound degradation, especially in the low temperature acidic melt. © 2010 The Electrochemical Society. 关DOI: 10.1149/1.3327903兴 All rights reserved. Manuscript submitted December 7, 2009; revised manuscript received January 18, 2010. Published April 1, 2010.

Tin dioxide with different additives is widely studied as an electrode material for various applications. Particular attention was devoted to SnO2-based low consumable 共inert兲 anodes to replace carbon-based anodes in electrolytic raw aluminum production.1-9 Low consumable oxygen-evolving anodes can operate without the production of greenhouse gases and carcinogenic polyaromatic compounds. Three different kinds of anode materials are generally considered: metals and alloys, “cermets” 共metal particles dispersed in an oxide ceramic matrix兲, and oxide ceramics. The latter looks the most promising from a thermodynamic point of view, despite the apparently poor mechanical properties. Among different binary and complex oxides, SnO2 demonstrates the most promising behavior due to the extremely low solubility of Sn共IV兲 species in cryolite alumina melt.10 However, pure SnO2 has a low electric conductivity and sintering ability. Various dopants and additives are used to improve its conductivity and to obtain a dense ceramic. The most conventional additives are Sb2O3 and CuO. The former, being partially oxidized, forms substitutional solid solutions with SnO2. Insertion of Sb5+ cations creates electrons as charge carriers and drastically increases the electric conductivity because of the following equation11,12 • x Sb2O5 → 2SbSn + 4OO + 2e⬘ + 1/2O2

• where SbSn denotes a positively charged Sb atom at the Sn position x and OO is oxygen at its position without charge. The homogeneity range of this solid solution is temperature-dependent 共e.g., up to 14 atom % of Sn can be replaced by Sb by hydrothermal treatment at 250°C,13 whereas at 1200°C the maximal solubility does not exceed 2.3 atom % Sb 14兲. However, conductivity increases with Sb content only up to 0.1–0.2 atom % Sb, with a subsequent plateau and even a slight decrease in conductivity at a high Sb concentration induced by the coexistence of Sb共V兲 and Sb共III兲 in the solid solution.12,15 If CuO is added simultaneously, rutilelike solid solutions Sn1−xCux/3Sb2x/3O2 共x ⱕ 0.25兲 are formed, but their thermal stability is limited until ⬇1200°C.16 The sintering temperature of the SnO2-based ceramic is usually higher 共up to 1400°C兲.17,18 Effective surface diffusion of the Cu atoms at temperatures ⬎400°C results in a Cu distribution on the grain surface of SnO2. When the tempera-

z

E-mail: [email protected]

ture rises up to 800–850°C, the Cu atoms replace the surface Sn positions, thus increasing the concentration of oxygen vacancies

⬙ + VO•• + OOx CuO → CuSn where CuSn ⬙ denotes a Cu atom at the Sn position with a double •• denotes an oxygen vacancy with a double negative charge and VO positive charge. The high concentration of defects near the grain surface causes a viscous flow sintering of the SnO2 particles.19 Above 1092°C in air, a eutectic CuO/Cu2O melt appears in the SnO2–CuO system that makes a liquid phase sintering mechanism possible.17 A rather small quantity of CuO 共0.75 wt %兲 is sufficient to achieve 98% of the theoretical density,18 but usually an excess of CuO up to 1.5–2 wt % is used to ensure a more homogeneous distribution among the SnO2 grains.8 The best combination of density and conductivity is empirically found for a CuO and Sb2O3 content of 1.5–2.0 wt %, and this composition was most actively used for the development of an anode material for aluminum production 共see Ref. 1 for review兲. However, the influence of the ceramic microstructure and electrolysis conditions on the corrosion mechanism and wear rate of the SnO2-based anode was not comprehensively studied yet. The most systematic study reports the data for more acidic melts than the conventional industrial melt: Their acidity 关in terms of cryolite ratio 共CR兲, determined as a molar ratio of NaF and AlF3兴 is below 2, whereas large-scale electrolysis takes place at CR = 2.2–2.4.3 For the melts with such CR, there is a lack of understanding of numerous chemical, physicochemical, and physical processes resulting in a degradation of the SnO2-based anodes. The depletion of the grain surface by copper was assumed to be the principal cause of the anode corrosion, with a subsequent decrease in mechanical strength of the ceramic leading to cracking of the anode.3,6,7 The Cu depletion is a consequence of the 1–2 orders of magnitude higher solubility of Cu共II兲 species in the melt, compared with the solubility of Sn共IV兲 at different CRs.1 The linear wear rate of the SnO2-based anodes estimated by various groups is ⬃20 to 40 mm/year,3,4 which is much lower compared to the wear rate of metallic anodes and looks satisfactory for industrial application. However, this rate is still too high to avoid contamination of raw aluminum by Sn, which is absolutely undesirable because even 200 ppm of Sn dramatically decreases the mechanical properties of aluminum.1 Because the reduction of tin species in the cryolite melt is a slow process,20 the Sn contamination level in alu-

Downloaded 06 Apr 2010 to 195.208.208.146. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp

Journal of The Electrochemical Society, 157 共5兲 C178-C186 共2010兲 minum is controlled by the solubility of tin species in the melt, which decreases sharply when CR and temperature decrease.10 The analysis of the degradation behavior depending on melt temperature and composition can be found in Ref. 3. The lowest degradation rate of the SnO2-based anode occurs for CR ⬇ 1.8,3 whereas a decreasing CR results in a faster degradation. However, in Ref. 3, some foreign cations 共Ca, Li, and Mg兲 were added to sodium cryolite to change the liquidus temperature. As these additives affect the CR value, an unambiguous interpretation of the observed results becomes difficult. In the present contribution, we investigate the effect of various factors on corrosion rate and confirm anomalous degradation enhancement with decreasing temperature. We report a systematic study of the ceramic microstructure evolution 共porosity and Cu depletion兲 and of electric conductivity at a microscopic level for the 97 wt % SnO2 + 1.5 wt % CuO + 1.5 wt % Sb2O3 ceramic anodes tested upon 10–100 h of electrolysis. We demonstrate that nonuniform electric conductivity at a local scale and the semiconducting nature of the ceramic are responsible for the degradation behavior. Experimental Ceramic fabrication and characterization.— The ceramic anodes were prepared in the form of rods with dimensions of 15 ⫻ 15 ⫻ 100 mm. The initial oxides in the 97 wt % SnO2 + 1.5 wt % CuO + 1.5 wt % Sb2O3 proportion were grinded together in a planetary mill using stainless steel bowls and balls. The obtained powder with 4 wt % of 10% water solution of poly共vinyl alcohol兲 was pressed into rods up to 4–4.7 g cm−3 raw density. The rods were fired in air at 1300°C for 10 h on a powdered corundum pad resulting in sintered ceramics with apparent densities of 6.1–6.8 g cm−3. To get more information on the phase composition of the ceramic material, a rod with the 90 wt % SnO2 + 5 wt % CuO + 5 wt % Sb2O3 was also prepared in the same way. The apparent density and the open and closed porosity of the samples were determined by hydrostatic weighing in water. The phase composition of the anodes was tested with X-ray powder diffraction 共XRPD兲. The patterns were collected on a Huber G670 Guinier diffractometer 共Cu K␣1 radiation, curved Ge monochromator, transmission mode, and image plate兲. X-ray fluorescence 共XRF兲 analysis of the ceramic samples was performed using a Thermo ARL Optim’X spectrometer 关50 W X-ray tube with Rh anode, goniometer equipped with three crystals 共AX06, PET, and LiF 200兲, and two detectors 共FPC and SC兲兴. The measurements were done in vacuum 共10−2 mbar兲 according to the requirements of OptiQuant 21 software for the standardless XRF analysis 共124 characteristic and background points in the whole spectrum from F K␣ to Ba K␣兲, and the result treatment was done using the OptiQuant software package. AnySample calibration was used, and the results were normalized to 100% or to the sample area. Microstructural characterization.— The microstructure and local composition of the anodes were investigated with a JEOL JSM6490LV scanning electron microscope 共SEM兲 equipped with an Oxford Instruments attachment for energy-dispersive X-ray 共EDX兲 analysis. The samples were polished before the electron microscopy investigation. The compositionally sensitive imaging in backscattered electrons was used to determine porosity at a microscopic scale. The contrast/brightness settings were adjusted to achieve the maximal contrast and to show pores 共empty or filled with the melt兲 as black areas. The pore area fraction was calculated as a percentage of the black area on the image. The contrast/brightness settings were calibrated on the samples with the porosity calculated from the results of hydrostatic weighing. A series of images with a 1200 ⫻ magnification was taken from the edge of the anode toward its center on a depth of 3–5 mm with a step of 100 ␮m. By averaging the results of three to five series of images, a dependence of the porosity on the distance from the anode edge was obtained. Part of the SEM

C179

investigations was performed with an LEO Supra 60VP instrument with an Oxford Instruments attachment for wave dispersive X-ray 共WDX兲 analysis. The sample for transmission electron microscopy 共TEM兲 was prepared by cutting a disk 0.5 mm thick and 3 mm in diameter from the rod of the 97 wt % SnO2 + 1.5 wt % CuO + 1.5 wt % Sb2O3 ceramic. The sample was mechanically polished and finally thinned to electron transparency by ion beam milling. Bright-field 共BF兲 TEM and annular dark-field scanning transmission electron microscopy 共ADF STEM兲 images were taken with an FEI Tecnai G2 microscope operating at 200 kV. Compositional mapping was performed in scanning transmission electron microscopy mode using EDX with an EDAX attachment. Energy filtered transmission electron microscopy 共EFTEM兲 imaging was performed using a JEOL 3000 microscope equipped with a Gatan image filter 共GIF-2000兲 using the standard three-window method.22 Bulk and local electrical properties.— Resistivity vs temperature dependence was measured ranging from room temperature up to 900°C in air using a four-probe technique utilizing a homemade setup with platinum contacts. Scanning tunneling microscopy 共STM兲 measurements were performed using the commercial ex situ scanning tunneling microscope “UMKA” with extended spectroscopic facilities 共Nanoindustry, Moscow兲 and 5.3 ␮m maximal scanned area. A Pt–Ir tip 共10% Ir兲 0.5 mm in diameter was mechanically sharpened. For topographic and spectroscopic measurements, a ⫺0.5 V tunneling voltage was applied 共negative voltage corresponds to positive tip polarization兲 in combination with a 0.5 nA tunneling current. The measurements were performed on a mechanically polished ceramic surface, washed with ethanol, and dried in air. The standard technique of local tunneling spectroscopy was used to do the measurements at a fixed point 共see Ref. 23兲. To obtain additional information on the local electric properties, an original technique of differential quasi-topographic spectroscopic mapping was applied 共a detailed description is provided in supplementary material兲.24 The results presented in this contribution were obtained at an ac excitation of 50 mV and a frequency of 6 kHz for 128 ⫻ 128 points. Electrochemical life tests.— For life tests, a Kulon-9 switching power supply 共current up to 250 A兲 was used. The tests were performed in a two-electrode configuration in an alumina-saturated melt. Ingots of TiB2-carbon composite material 共15 ⫻ 15 ⫻ 100 mm兲 or carbon samples of similar size covered with TiB2 were used as cathodes. The ceramic anode was fixed at a copper pin 共working as a current collector兲, and the region of contact was isolated with corundum cement. An immersion depth of 60–70 mm provided a working anode area of ca. 40 cm2. For 10, 12, and 64 h tests, a single ceramic anode and a single cathode were polarized at 0.5 A cm−2 current density 共total current ca. 20 A兲 in a graphite crucible with 2.1 kg of melt. For 24 and 100 h tests, electrode assemblies consisted of two anodes and two cathodes or of five anodes and four cathodes, respectively. In these cases, the graphite crucible contained 13.5 kg of melt. The same current density of 0.5 A cm−2 resulted in a total current of 40 and 100 A, respectively. Melts of various compositions were prepared using Na3AlF6, AlF3, and Al2O3 of “pure” grade. Alumina was added at regular intervals 共5–30 min兲 in a certain excess 共the amount was determined under the assumption of 75 or 100% current efficiency, when the real current efficiency never exceeded 68%兲. Voltage–time dependences were recorded during the whole electrolysis period, with periodical short current decreases of up to 25% of the nominal value to estimate the ohmic drop from the corresponding voltage decrease. To systematically check the effects of the current density and ceramic microstructure, a melt of CR = 1.8 was used 共this CR value is discussed below as “standard”兲. To study CR and temperature influence, a current density of 0.5 A cm−2 was fixed. The tin content in the aluminum metal was measured from atomic absorption data 关224.605 nm analytical wavelength,


C180


acetylene–air flame, contrAA 300 共AnalytikJena兲兴. The metal was dissolved in 10% hydrochloric acid. Calibration solutions were prepared by dissolving tin metal 共or tin and aluminum metals simultaneously兲 in the same medium. A linear calibration dependence was found up to 1 wt % of tin in the solution. For both the melt and the metal, at least five probes from different fragments of the obtained aluminum were analyzed, and the results were averaged.

Results and Discussion Microstructure of the initial ceramic.— Typical backscattered SEM micrographs illustrating the microstructure of the 97 wt % SnO2 + 1.5 wt % CuO + 1.5 wt % Sb2O3 ceramic are shown in Fig. 1a and b. The samples with the apparent densities of 6.1–6.5 g cm−3 共Fig. 1a兲 demonstrate the pronounced open porosity. Increasing the apparent density up to 6.8 g cm−3 results in a significant decrease in the amount of pores, most of which become closed 共⬃4% of overall porosity and ⬍1% of open porosity by hydrostatic weighing measurements, Fig. 1b兲. It is difficult to separate the SnO2 grains in the obtained coarse crystalline mass. In both cases, small inclusions can be identified to be positioned mostly at triple points between the grains. In Fig. 1, these inclusions are imaged as darker gray areas; some of them are marked with arrows. Because backscattered electron imaging is sensitive to the chemical composition, one can assume an exsolution of a secondary phase. BF TEM image 共see supplementary material兲24 and ADF STEM image 共Fig. 2兲 reveal long straight grain boundaries. At the triple junction, the grains do not meet each other, forming a pore filled with a Cu-rich phase. The presence of Cu-rich inclusions at the triple points is evident from the elemental distribution in the compositional maps 共Fig. 2兲. The EFTEM image of the grain boundary itself 共Fig. 3兲 clearly demonstrates that Cu is concentrated in a very thin layer 共up to 6 nm兲 between the ceramic grains. ADF STEM image and EDX mapping 共see supplementary material兲24 also support that the thin intergrain layer is Cu-enriched and Sn-depleted. The XRPD patterns of the 97 wt % SnO2 + 1.5 wt % CuO + 1.5 wt % Sb2O3 samples only demonstrate the presence of the SnO2-based phase with the rutile-type structure. The phase composition of the inclusions at the triple points was identified on the 90 wt % SnO2 + 5 wt % CuO + 5 wt % Sb2O3 sample where such inclusions have a larger size and are present in a larger amount 共Fig. 1c兲. The EDX analysis revealed an atomic ratio of Cu:Sb = 83共2兲:17共2兲 for the inclusions, and XRPD identified the phase as Cu4SbO4.5.25 Because the XRPD pattern of the Cu4SbO4.5 phase was reported without indexation and because unit cell parameters of this compound are not known,25 we have performed the synthesis of a single-phase sample and determined the unit cell parameters to ensure an adequate identification of this phase in the multiphase mixtures. Cu4SbO4.5 crystallizes in a monoclinic unit cell with a = 7.174共2兲 Å, b = 11.519共3兲 Å, c = 6.625共1兲 Å, and ␤ = 99.56共2兲°. The inclusions mostly consist of the Cu4SbO4.5 phase with a possible admixture of CuO or Cu2O to account for the Cu:Sb ratio slightly larger than 4:1. The Sb content in the grains of the SnO2-based phase was investigated with WDX analysis. The intensity of the Sb K␣ peak on the spectra obtained from the internal areas and near the grain boundaries of several grains was comparable with the background noise. However, according to the XRF analysis, the Sb content in the samples is very close to the expected values 共ca. 1.2–1.7 wt % of Sb2O3兲. This means that the volatilization of antimony oxides in the course of sintering does not significantly affect the bulk material composition. The total molar Cu:Sb ratio is finally about 1.8:1, and all antimony cannot be situated exclusively in the Cu4SbO4.5 phase. The rest of the Sb is most probably situated in the grains, which agrees with the low electric resistivity of the ceramics 共typically 10−3 ⍀ m at 900°C兲 reflecting sufficient charge carrier concentration. The temperature dependence of ceramic resistivity is typical of SnO2-based materials doped with copper and antimony.8,26,27

Figure 1. Backscattered SEM images of the microstructure of the 97 wt % SnO2 + 1.5 wt % CuO + 1.5 wt % Sb2O3 ceramic with the apparent density: 共a兲 ␳ = 6.4 g cm−3, 共b兲 ␳ = 6.8 g cm−3, and the 共c兲 90 wt % SnO2 + 5 wt % CuO + 5 wt % Sb2O3 ceramic. Arrows point to inclusions of the Cu-rich phase.

Local electric conductivity of ceramic fragments.— Quasi-topographic conductivity mapping in STM configuration performed on the ceramic after the electrochemical life test 共Table I, sample 4兲 revealed the regions of essentially different conductivities on the micrometer scale. Figure 4 presents the simultaneously registered topographic image and two spectroscopic maps: 兩dI/dU兩 and the phase shift between current and voltage. The topographic contrast 共Fig. 4a兲 is rather weak 共the height drop does not exceed 75 nm兲 and only makes a minor impact into the spectroscopic maps. Figure 4b



C181

Figure 2. ADF STEM image of the triple junction point and corresponding compositional maps revealing inclusion of a Cu-rich phase.

demonstrates the existence of inland regions with high conductivity 共high dI/dU signal and bright regions兲. These regions are separated by areas with a much lower conductivity 共dark regions, 200–500 nm

Figure 3. 共Left兲 BF TEM and 共right兲 EFTEM image of the grain boundary demonstrating enrichment of the grain boundary with Cu. The thickness of the grain boundary is approximately 6 nm.

Figure 4. 共a兲 Topographic image, spectroscopic maps for 共b兲兩dI/dU兩 and 共c兲 phase shift between current and voltage for sample 4, and 共d兲 local tunneling I-V characteristics at different points marked in 共c兲.

thick兲. Figure 4c 共a map of the phase shift兲 with a higher contrast looks like the inverted image of the 兩dI/dU兩 map 共an increase in dI/dU decreases the contribution of the reactive impedance component; i.e., it decreases the phase shift兲. Conventional local voltammetry at the grains and intergrain regions 共Fig. 4d兲 confirms the results of the spectroscopic observations. Dark 共more conductive兲 regions in Fig. 4c are associated with asymmetric spectra with high conductivity at a negative tunneling voltage and with a very low conductivity at the positive voltage. This behavior is characteristic of n-type semiconductors. The tunneling gap is a metal–insulator–semiconductor-type diode; i.e., only the tunneling from the sample to the tip is possible. For the intergrain regions, spectra of a different type were observed: practically

Table I. Sample properties, conditions, and results of the electrochemical tests.

Sample 1 2 3 4 5 6b 7 8 9 10 11 12 13 14 15 16 a b

Apparent density, open porosity 共g cm−3, %兲 6.1, 13.0 6.29, 7.0 6.1, 13.0 6.1, 9.0 6.8, 0.2 6.5, 0.5 6.7, 1.0 6.4, 9.0 6.8, 0.2 6.4, 9.0 6.3, 7.3 6.26, 2 6.4, 9.0 6.4, 9.0 6.4, 9.0 6.4, 9.0

CR, temperature 共 °C兲 1.8, 1.8, 1.8, 1.8, 1.8, 1.8, 1.8, 1.3, 1.3, 2.3, 1.8 共10% 1.8, 1.8, 1.8, 1.8, 1.8,

920 920 920 920 920 920 920 750 750 970 KF兲, 860 920 920 920 920 920

Test duration 共h兲

Current density, current efficiency 共A cm−2, %兲

12 24 64 100 10 12 100 12 10 12 12 12 12 12 12 7.5

0.53, 65 0.52, 48 0.42, 45 0.50, 54 0.53, 44 0.53, 49 0.5, 55 0.53, 60 0.43, 51 0.48, 53 0.53, 51 0.67, 68 0.25, 43 0.45, 44 0.97, 40 1.48, 34

Thickness of Cu depletion layer 共␮m兲

Thickness of degradation layer 共␮m兲

550 230 1250 200 1750 450–750 500–2200 500–1000 共60–200兲a 60 60 70 70 500–1000 500–1000 共50–150兲a Complete disintegration of the anode — 600a 50 ⬍20 1500 310 450 70 330 70 550 120 470 420 960 830

Strong corrosion zone. 90 wt % SnO2 + 5 wt % CuO + 5 wt % Sb2O3.


C182


Figure 6. Schematic model of the ceramic based on the results of microstructural investigation and local conductivity measurements: 共1兲 Sb-doped SnO2 grain bulk 共core兲, 共2兲 SnO2 near-surface grain shell doped with Sb-type and acceptor-type dopants 共Cu2+,Fe3+,Al3+兲, 共3兲 intergrain boundary consisting mostly of Cu-rich material, and 共4兲 triple point containing Cu4SbO4.5 and copper oxides.

Figure 5. 共a兲 Topographic image, spectroscopic maps for 共b兲兩dI/dU兩 and 共c兲 phase shift between current and voltage for sample 6, and 共d兲 local tunneling I-V characteristics at different points marked in 共c兲.

symmetric, with a much lower conductivity at a negative voltage 共curves 4 and 5 in Fig. 4d兲. The same situation was found for the initial ceramics before interaction with the melt. Denser ceramics demonstrate a lower conductivity contrast and identical tunneling spectra for different points of the surface 共Fig. 5兲. Conductivity for these materials is intermediate between highly conducting and weakly conducting areas in the more porous ceramic. The scanning area of the equipment used is limited to 5.3 ⫻ 5.3 ␮m, being smaller than the average size of the ceramic grains. This is the reason why we cannot obtain an image of several highly conducting grains. Despite this fact, a location of the areas with different conductivities satisfies the assumption that n-type conductivity is demonstrated by the interior parts of the SnO2 grains. Contrarily, the intergrain regions cannot be assigned exclusively to grain boundaries because the thickness of the latter, according to TEM results, does not exceed several nanometers. Even minor changes in doping level 共changes in donor or acceptor impurity concentration兲 in the semiconductor can induce pronounced changes in the local charge carrier concentration. This could be a possible reason for the changes in local conductivity. Sb共V兲 acts as a donor in the SnO2 rutile-type lattice, increasing the electron concentration and electric conductivity. One can assume that just the interior parts of the ceramic grains 共grain core兲 with high conductivity are Sbdoped 共Fig. 6兲. The lower conductivity of the outer parts 共grain shell兲 of the grains can be attributed to a simultaneous doping with Sb and acceptor-type dopants, which could be the Mn+ cations with n ⬍ 4. Such dopants decrease the electron concentration. Besides Cu2+, Fe3+ and Al3+ can play the role of acceptor dopants. A pollution of the ceramic by iron and aluminum oxides unavoidably occurs due to grinding of the powders in stainless steel bowls and to sintering the ceramic on the corundum pad. Indeed, traces of Fe and Al 共up to 0.2–0.3 wt % M2O3兲 were detected in the ceramic by XRF analysis. Simultaneous doping of tin dioxide with Sb and the acceptor elements can be treated as the formation of the Sn1−xCux/3Sb2x/3O2 or Sn1−xSbx/2Mx/2O2 共M = Fe,Al兲, solid solutions. The amount of Mn+ cations in the grain shell is low; at least it was not observed by the EDX and EFTEM mapping. This reflects that x in the solid solutions is below the detection limit of these local techniques.

Decreasing the electron concentration by acceptor doping is associated with a lower conductivity of the grain shell in comparison with the grain core. For the microstructure imaged in Fig. 6, one can expect the formation of potential Schottky barriers between more and less conductive regions and the appearance of limitations for the charge transport between these regions. Then, the charge transport can occur only along the less conductive continuous fragments 共grain shells, marked as 3 in Fig. 6兲, and the macroscopic resistance of the anode material is determined by the conducting properties of these fragments. One can assume that the interaction between the ceramic and the melt decreases the energy of the barriers, and the current flow through the more conductive Sb-doped SnO2 grain cores becomes possible. Under these circumstances, the total conductivity becomes closer to the conductivity of the grain cores; i.e., it increases. The anodes after the 100 h life test 共sample 4兲 demonstrate a drastic change in electric conductivity vs temperature behavior. The roomtemperature conductivity of the surface layer of the anode 共sample 4兲 increases by 2–3 orders of magnitude from 0.06–0.33 to 14–1000 ⍀−1 m−1. In the core of the ceramic anode bar, the conductivity also rises but less significantly than in the surface layer 共up to 5–50 ⍀−1 m−1兲. At the same time, a decrease in activation energy from 0.53–1.18 to 0.04–0.08 eV was observed. The conductivity increase is accompanied by the appearance of a nonlinearity of the current–voltage 共I-V兲 characteristics, typical of heterogeneous materials with double Schottky barriers. However, a more detailed analysis of the changes in conductivity during the electrolysis is required to reveal the exact nature of the barriers and their evolution with the electrolysis time. Degradation of the anodes.— Table I summarizes the data on the SnO2-based ceramic materials tested at different electrolysis conditions. The majority of the data corresponds to a standard alumina-saturated melt of CR = 1.8 共920–935°C兲. The closest melt composition reported for laboratory tests on similar ceramics3 corresponds to the formally calculated CR 1.78, but 5 wt % of CaF2 added to the melt surely increased this value. Precise electrochemical measurements with potential control in a three-electrode cell confirmed the good agreement of the ceramic electrode behavior with the earlier reported data.2,3,6,27,28 All electrodes demonstrate a rather stable behavior in the life tests: Cell voltage remains practically constant 共a slow voltage increase resulted from the oxidation of the current collectors兲 and does not exceed 5.0–5.5 V for the 100 h tests 共samples 4 and 7兲. Independent of the electrolysis time, for the samples of relatively high porosity 共samples 1–4 with ␳ ⬍ 6.5 g cm−3兲, the exposure of the ceramic to



C183

Figure 7. SEM image of the central part of the 97 wt % SnO2 + 1.5 wt % CuO + 1.5 wt % Sb2O3 anode after 100 h life test 共sample 4兲. Fluorine distribution map is overlaid 共whiter areas, some marked with arrows兲, demonstrating impregnation of pores with the melt.

Figure 9. SEM image of the 97 wt % SnO2 + 1.5 wt % CuO + 1.5 wt % Sb2O3 anode after 100 h life test 共sample 4兲 near the anode surface.

the melt results in a deep penetration of the melt inside the anode body and elimination of the Cu-rich inclusions to a depth up to 2.5 mm. Figure 7 demonstrates the compositional map of the central part of sample 4, where fluorine is seen in the pores demonstrating impregnation of pores with the melt. After 100 h of electrolysis, majority of the pores is completely filled with the melt even in the middle of the anode body. The Cu concentration in the anode demonstrates a steplike dependence on the distance from the anode surface: For sample 4, the Cu content at the first 2 mm from the surface is not detectable by the EDX analysis, whereas at a larger distance, a significant amount of Cu remains in the sample. The Cu concentration map of the area near the Cu elimination front is shown in Fig. 8. Due to the inhomogeneity of the Cu distribution, its amount can only be roughly evaluated as 1.7共7兲 wt % of CuO in the Cucontaining region and 0.15共16兲 wt % of CuO in the Cu-depleted region. On the SEM micrographs of sample 4 共Fig. 7-9兲, an increasing amount and an increasing size of the pores become evident. A narrow surface layer, 50–200 ␮m thick, is subjected to a pronounced degradation 共Fig. 9兲. In contrast to the initial ceramic 共Fig. 1a兲, the degraded surface layer is characterized by a much larger porosity, destruction of the intergrain contacts, and a clearly visible faceting

of the ceramic grains. From a quantitative analysis of the dependence of the pore area on the distance from the surface, three degradation zones can be selected 共Fig. 10兲. The zone of strong corrosion has a pore area fraction up to 40%. In the moderate corrosion zone, the intergrain contacts are not severely damaged. The pore area fraction in this area is about 10%, and the thickness of this zone is about 800 ␮m. This zone is followed by the Cu depletion zone 共up to ⬃2.5 mm from the anode surface兲 where the porosity increases exclusively due to a removal of the Cu-containing inclusions. The pore area fraction in this zone is typically 1–3% larger than the pore area fraction in the initial ceramic. The border between the Cu depletion zone and the unaffected ceramic roughly coincides with the border between the Cu-containing and Cu-depleted regions, as determined by EDX analysis. Similar degradation zones were found in samples with ␳ ⬍ 6.5 g cm−3 tested at 12, 24, and 64 h. The difference in porosity between the degradation zones becomes less pronounced with decreasing test time, so that for sample 1, tested for 12 h, the separation of zones is to some degree subjective. The wear rate of the anode estimated from the amount of tin impurity in the melt 共typical values of 40–60 ppm兲 and aluminum 共typical values 0.15–0.20 wt %兲 was about 4–6 mm/year. However, the shrinkage of the linear dimensions of the anode indicates essentially a faster degradation 共15–17 mm/year兲. These values are close to the

Figure 8. SEM image of the 97 wt % SnO2 + 1.5 wt % CuO + 1.5 wt % Sb2O3 anode after 100 h life test 共sample 4兲. Copper distribution map is overlaid 共whiter areas兲, demonstrating the Cu-depleted region from the anode surface side. The border of this region is marked with white dashed line.

Figure 10. The dependence of pore area on the distance from the anode surface for the 97 wt % SnO2 + 1.5 wt % CuO + 1.5 wt % Sb2O3 anode after 100 h life test 共sample 4兲.


C184


Figure 11. SEM image of the 97 wt % SnO2 + 1.5 wt % CuO + 1.5 wt % Sb2O3 anode after 10 h life test 共sample 5兲 near the anode surface. Copper distribution map is overlaid 共darker areas兲 demonstrating the Cu-depleted region from the anode surface side. The border of this region is marked with white dashed line. The area of EDX analysis is located between two black dotted lines.

wear rate reported.3,4 This difference can be attributed to the high volatility of tin species at the temperature of electrolysis. The corrosion behavior of the ceramics with ␳ ⱖ 6.5 g cm−3 共samples 5 and 6兲 is different from that of the less dense materials. The microstructure of the initial ceramic of sample 5 was unchanged after the 10 h test except for a narrow layer near the surface. The surface layer with a thickness of 60 ␮m is Cu-depleted 共Fig. 11兲 and contains the melt components in the pores. It is also characterized by an increased porosity: The average pore area fraction in this layer is 4.0% in comparison with 1.0% in the initial ceramic. The melt does not penetrate outside this 60 ␮m affected area, which reflects that closed pores prevail in this ceramic. The dense ceramic demonstrates an encouraging behavior in the 10–12 h laboratory test: The tin content in the aluminum metal was only 0.055 wt %, i.e., 3 times lower than that in the ceramic with ␳ ⬍ 6.5 g cm−3 共0.17 wt %兲. Sample 6 with the 90 wt % SnO2 + 5 wt % CuO + 5 wt % Sb2O3 composition demonstrates a similar behavior, but due to a higher Cu content in the initial ceramic, the residual Cu content in the obtained metal is unacceptably high 共0.17 wt %兲. However, the 100 h life test for the 97 wt % SnO2 + 1.5 wt % CuO + 1.5 wt % Sb2O3 dense material 共␳ = 6.7 g cm−3, sample 7兲 does not result in a lowering of the tin content in the aluminum and a decreasing wear rate in comparison with the less dense anode 共0.34 wt %, 17 mm/year兲. The thickness of the degradation layers estimated from the SEM data was rather close to the value observed in the less dense material 共sample 4兲. Ceramic destruction due to a reduction of SnO2 with the metallic aluminum 共in the form of small drops dispersed in the melt or Al dissolved in the melt兲 occurring at the external surface of the anode rod cannot explain the increase in the anode internal porosity. We can assume that the thinner degradation layer found after a short electrolysis of the dense material results from an initially slower melt penetration due to a lower open porosity. However, at a longer electrolysis time, the melt penetration depth becomes comparable for anodes with different densities causing a similar porosity development and Cu depletion. When the electrolysis time is increased 共samples 1–4兲, the depth of degradation increases 共Fig. 12a兲, but the external zone of strong corrosion expands only slightly. However, prolonged electrolysis results in noticeable changes in the anode size 共ca. 200 ␮m loss from all sides兲. One of the reasons can be the mechanical loss of tin dioxide crystals induced by damage of the intergrain contacts. The thickness of the moderate corrosion zone after 100 h polarization attains only 0.5–0.7 mm, but no manifestation of a decrease in wear

Figure 12. 共a兲 Size of degradation zones and 共b兲 relative increase in porosity vs electrolysis time.

rate with electrolysis time is observed. The 10–100 h tests in the CR = 1.8 melt never resulted in the mechanical disintegration of the anode material, but for longer experiments this risk cannot be completely excluded. The dependence of the size of the degradation zones and the relative increase in the porosity ⌬S/S on the time of electrolysis was analyzed 共Fig. 12a and b兲. The relative increase in the porosity was defined as the ratio of a difference in the pore area fraction in the corrosion zone and the pore area fraction in the initial ceramic 共⌬S兲 to the pore area fraction in the initial ceramic 共S兲. One can see that in the moderate corrosion zone and the Cu depletion zone, a relative increase in the porosity practically does not depend on the electrolysis time. Contrarily to that, the strong corrosion zone shows an increase in ⌬S/S from 110% at 12 h to 220% at 100 h 共Fig. 12b兲. The decrease in the linear anode dimensions after 100 h electrolysis is equivalent to a wear rate of 15–17 mm/year, close to the 20 mm/year reported by Vecchio-Sadus et al.3 The Cu depletion zone and the moderate corrosion zone deep inside the material grow noticeably faster 关⬃130 and ⬃76 mm/year, respectively 共Fig. 12a兲兴. The lateral movement of the strong corrosion zone toward the anode center is relatively slow 共⬃26 mm/year兲, and the thickness of this layer increases only slightly with electrolysis time 共⬃10 mm/ year兲. Simultaneously, we noticed no apparent features of the inhibition of internal degradation in the course of a prolonged electrolysis. The integrity of the ceramic and intergrain contacts is not significantly affected in the Cu depletion zone and the zone of moderate corrosion. The degradation degree of the ceramics in these zones is almost time-independent. It is reasonable to assume that the increasing size of these regions does not result in a considerable risk of the mechanical destruction of the anode. Indeed, in our experi-



Figure 13. SEM image of the 97 wt % SnO2 + 1.5 wt % CuO + 1.5 wt % Sb2O3 anode after 10 h life test 共sample 9, CR = 1.3兲 near the anode surface.

ments we did not observe a mechanical deterioration of the anode, in contrast to that reported by Galasiu et al.7 The development of the size of the strong corrosion zone could be an origin of the risk of the mechanical destruction of ceramics. However, the size of this zone does not exceed 200 ␮m and increases only slightly with electrolysis time. The movement of the strong corrosion zone toward the central part of the anode is probably responsible for the decrease in the linear dimensions of the anode 共e.g., by the mechanical loss of separate tin dioxide crystals induced by a damage of the intergrain contacts兲. Thus, the main amount of the Sn impurity in the melt and aluminum occurs due to loss of the ceramic material from the anode surface from the narrow strong corrosion zone that gradually moves toward the anode center, keeping nearly a constant thickness, which results in decreasing linear dimensions. The intergrain corrosion seems to be the most dangerous process leading to damage on the anode surface and finally to aluminum contamination by Sn. To understand which factors can suppress or aggravate degradation, we tested the behavior of ceramics in the melts with acidity higher or lower than in the standard melt discussed above. The CR and temperature effects 共which usually occur simultaneously兲 and the current density effect 共which indirectly reflects the effect of electrode potential兲 were investigated. We also tried to compare ceramic samples of different “qualities,” as indicated by differences in porosity and local electric conductivity. The decrease in CR and temperature 共samples 8 and 9, more acidic melt兲 results in a dramatic degradation of the ceramic. Intergrain regions are most significantly affected 共Fig. 13; additional images can be found in supplementary material24兲. A destruction of the intergrain contacts was observed up to 0.6 mm in the ceramic with ␳ = 6.8 g cm−3 after 10 h of electrolysis 共sample 9兲, whereas for the ceramic with ␳ = 6.4 g cm−3 共sample 8兲, the electrolysis of 12 h results in a complete destruction of the anode, giving no chance to prepare a sample for SEM. The anode surface is covered by dendrites indicating a fast dissolution of the ceramic in the melt and a subsequent redeposition of SnO2 on the anode surface. Redeposition of a substantial amount of SnO2 is only possible if the content of tin species in the melt 共typically more than 100–200 ppm兲 exceeds the low solubility of Sn共IV兲; i.e., if Sn共II兲 species are formed. At the beginning of the electrolysis, the behavior of the ceramic anode in the low temperature melt is similar to that described above for higher CR and temperature. However, after several hours of electrolysis, an increase in cell voltage and a simultaneous increase in tin concentration in the melt are observed, manifesting the beginning of a fast degradation. Thus, some changes in the chemical composition and physical properties of the ceramics, as induced by electrolysis and chemical interaction with the melt, result in an accelerated degradation.

C185

In contrast, the increase in CR and temperature 共sample 10, less acidic melt兲 results in negligible intergrain corrosion. The experiment in a potassium-based low temperature melt 共sample 11兲 demonstrates that intergrain corrosion facilitates as compared to the standard melt. The CR of the potassium-based melt is 1.8 共calculated from the total quantity of potassium and sodium兲, which is the same as for standard melt. However, its working temperature is 850°C, which is lower than that for the standard melt and similar to that for the more acidic potassium-free melt. Thus, the degradation test in the potassium-based melt unambiguously confirms that the specific features of the degradation behavior are determined more by temperature, rather than by the melt ionic composition 共responsible for CR兲. Current density effects were studied in the standard melt. An increase in current density 共samples 13–16兲 increases the wear rate. The most pronounced degradation was observed in the short test at a current density of 1.48 A cm−2 共sample 16兲: A depth of the degradation layer close to 1 mm was found. A comparison of anodes of different open porosities 共samples 1, 5, 12, and 14兲 also demonstrates a clear trend: The lower the open porosity, the smaller the depth of the degradation zone. When the open porosity is below 2% 共10–12 h electrolysis兲, the depth of degradation is limited by the melt penetration 共sample 5兲. With a larger open porosity 共⬎2%兲, the depth of degradation becomes smaller than the melt penetration depth. This means that the depth of the melt penetration is not a key factor in determining the depth of the Su depletion and degradation. The major degradation phenomenon for the investigated ceramic anodes is the nonuniform intergrain corrosion resulting from nonuniform current density distribution. An increase in the intergrain corrosion rate with a decrease in the melt temperature is difficult to explain by decreasing alumina solubility: Al2O3 depletion 共slow diffusion of oxygen-containing species from the melt bulk兲 is expected to induce the uniform dissolution of ceramic material. All degradation tests took place at current densities much lower than the limiting diffusion current. However, for the n-type semiconductors, the maximum anode current density is determined by the concentration of secondary charge carriers 共holes兲, which is usually small. Its deficiency can lead to an increase in the overvoltage for the anodic process and to the degradation of the material. Under thermal equilibrium, the concentration of holes is higher in the low conducting grain shell because of simultaneous acceptor doping. For a fixed current, one can expect a lower overpotential at the grain shells 共if two hypothetic separate materials are considered兲. Hence, a higher local current density is expected at the grain shells under constant potential. In other words, the differences in the carrier 共holes兲 concentration for the grain core and grain shell must unavoidably lead to a nonuniform distribution of the current density at the anode: The electrochemical process predominantly proceeds in the intergrain regions with higher hole concentrations. High local current densities accelerate the degradation of the intergrain regions; as for all electrochemical processes, a higher current density corresponds to a higher potential. Moreover, a certain potential increase can result in the appearance of an additional anodic process: oxygen evolution from tin dioxide 共so-called catastrophic degradation兲. An increase in the electrolysis temperature increases the total amount of charge carriers 共including holes兲 in both more and less conducting regions. The conductivity of the latter regions influences the degradation crucially. This is why the degradation becomes weaker at higher temperatures and increases noticeably at lower temperatures. The aggravation of the degradation with the current density 共varied in relatively narrow interval兲 indicates the important role of a nonuniform current distribution within the ceramic. The difference in the secondary charge carrier 共holes兲 concentration and electric conductivity found in the grain core and the grain shell results in a nonuniform distribution of the anodic current density. The higher current density at the intergrain regions results in their profound degradation. However, the pronounced temperature dependence of


C186


the intergrain degradation allows us to assume that this nonuniform current distribution is induced by an inhomogeneous local electric conductivity and semiconducting properties of the ceramic material, which are temperature-dependent. A less pronounced inhomogeneity of the local conductivity could suppress the intergrain degradation. Thus, the electrochemical tests confirm the key role of local conductivity and semiconducting properties of the ceramic in the degradation behavior. Conclusion Microscopic features of the 97 wt % SnO2 + 1.5 wt % CuO + 1.5 wt % Sb2O3 ceramic anodes after electrolysis were investigated for different polarization conditions in a cryolite alumina melt of various compositions 共CR, temperature, time, and current density were varied兲. The degradation trends can be formulated as follows. 1. After a relatively long time of electrolysis 共⬃100 h兲, the affected area of the anode is clearly separated into zones. A narrow 共50–200 ␮m兲 strong corrosion zone at the anode surface is characterized by a severe damage of the intergrain contacts and by a large relative increase in the porosity 共by 100–200%兲. The moderate corrosion zone is wider 共up to 800 ␮m兲 with a relative porosity increase of 60–80%. In the Cu depletion zone 共up to 2500 ␮m兲, the ceramic retains its initial microstructure, and a slight porosity increase 共⬃20 to 25%兲 occurs solely due to the removal of the Cu-rich inclusions. 2. The thickness of the moderate corrosion zone and the Cu depletion zone increases with increasing electrolysis time, whereas the strong corrosion zone has an almost constant thickness, slowly moving toward the anode center resulting in a decrease in the linear dimensions of the anode, which makes a main impact into the contamination of the produced Al by Sn. 3. The risk of losing the mechanical integrity of the anode is associated only with the development of a strong corrosion zone. Because this zone has a constant narrow width, no catastrophic mechanical destruction of the anode could be expected and, indeed, was never observed. 4. The whole set of experimental observations can be adequately explained, assuming a microstructural model of the ceramic, which consists of grains with an Sb-doped SnO2 grain core surrounded by an ⬃200 to 500 nm grain shell where SnO2 is simultaneously doped with donor-type dopant 共Sb5+兲 and acceptor-type dopants 共Cu2+,Fe3+,Al3+兲. The grains are separated by grain boundaries 共⬃6 nm兲 that are significantly enriched with Cu. The Cu-rich phases 共CuO and Cu4SbO4.5兲 are segregated at the triple points between the grains. 5. The different secondary charge carrier 共holes兲 concentrations and electric conductivities in the grain core and grain shell results in a nonuniform anodic current density with a higher current density at the intergrain regions that results in their profound degradation. Thus, a more detailed investigation of the electric transport inside the grain and between the grains should definitely benefit in understanding the corrosion mechanisms with a further possibility to control it due to an appropriate chemical doping. One can assume

that the electric charge transfer between the ceramic grains is governed by the existence of potential Schottky barriers caused by a chemical inhomogeneity. Electrochemical experiments and measurements of the electric transport properties at bulk and local scales aimed at revealing the exact nature of the potential barriers are currently in progress. Acknowledgment This work was performed in the frame of the RusAl–MSU project “Search for oxide materials for inert anodes in aluminum production.” The authors thank Thermo Techno Ltd. 共Moscow兲 for their assistance in the XRF analysis. The authors are grateful to Galina A. Tsirlina for the brilliant ideas and fruitful discussions. References 1. I. Galasiu, R. Galasiu, and J. Thonstad, Inert Anodes for Aluminium Electrolysis, 1st ed., Aluminium-Verlag, Germany 共2007兲, and references therein. 2. Y. X. Liu and J. Thonstad, Electrochim. Acta, 28, 113 共1983兲. 3. A. M. Vecchio-Sadus, D. C. Constable, R. Dorin, E. J. Frazer, I. Fernandez, G. S. Neal, S. Lathabai, and M. B. Trigg, in Light Metals. TMS Annual Meeting, pp. 259–265 共1996兲. 4. H. Xiao, R. Hovland, S. Rolseth, and J. Thonstad, Metall. Mater. Trans. B, B27, 185 共1996兲. 5. R. Keller, S. Rolseth, and J. Thondstad, Electrochim. Acta, 42, 1809 共1997兲. 6. A.-M. Popescu, S. Mihaiu, and S. Zuca, Z. Naturforsch., A: Phys. Sci., 57, 71 共2002兲. 7. I. Galasiu, D. D. Popescu, R. Galasiu, M. Modan, and P. Stanciu, in Ninth International Symposium on Light Metals Production, J. Thonstad, Editor, NTNU, pp. 273–280 共1997兲. 8. S. Zuca, M. Terzi, M. Zaharescu, and K. Matiasovsky, J. Mater. Sci., 26, 1673 共1991兲. 9. 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, E. V. Antipov, et al., Chem. Mater., 17, 3004 共2005兲. 10. H. Xiao, J. Thonstad, and S. Rolseth, Acta Chem. Scand., 49, 96 共1995兲. 11. K. Uematsu, N. Mizutani, and M. Kato, J. Mater. Sci., 22, 915 共1987兲. 12. S. R. Dhage and V. Ravi, Appl. Phys. Lett., 83, 4539 共2003兲. 13. B. Grzeta, E. Tkalcec, C. Goebbert, M. Takeda, M. Takahashi, K. Nomura, and M. Jaksic, J. Phys. Chem. Solids, 63, 765 共2002兲. 14. T. Kikuchi and M. Umehara, J. Mater. Sci. Lett., 4, 1051 共1985兲. 15. C. Wang, J. Wang, C. Wang, H. Chen, W. Su, G. Zang, and P. Qi, J. Appl. Phys., 97, 126103 共2005兲. 16. O. Scarlat, M. Susana-Mihaiu, and M. Zaharescu, J. Eur. Ceram. Soc., 22, 1839 共2002兲. 17. N. Dolet, J.-M. Heintz, M. Onillon, and J.-P. Bonnet, J. Eur. Ceram. Soc., 9, 19 共1992兲. 18. J. Lalande, R. Ollitrault-Fichet, and P. Boch, J. Eur. Ceram. Soc., 20, 2415 共2000兲. 19. J.-P. Bonnet, N. Dolet, and J.-M. Heintz, J. Eur. Ceram. Soc., 16, 1163 共1996兲. 20. S. Vassiliev, V. Laurinavichute, A. Abakumov, E. Bendovskii, A. Filatov, D. Simakov, A. Gusev, E. Antipov, and G. Tsirlina, in Light Metals. TMS Annual Meeting, pp. 1135–1140 共2009兲. 21. OptiQuant, UniQuant 5 for Thermo ARL Optim’X, Thermo Fisher Scientific Corporation 共2008兲. 22. R. F. Egerton, Electron Energy-Loss Spectroscopy in the Electron Microscope, 2nd ed., Plenum, New York 共1996兲. 23. S. Yu. Vasil’ev and A. V. Denisov, Tech. Phys., 45, 99 共2000兲. 24. See supplementary material at http://dx.doi.org/10.1149/1.3327903 共E-JESOAN157-017005兲 for additional information. 25. S. Simada and K. J. D. Mackenzie, Thermochim. Acta, 56, 73 共1982兲. 26. M. F. Ralea, I. Galasiu, and R. Galasiu, in Proceedings of the International Semiconductor Conference, CAS, pp. 137–140 共1995兲. 27. S. Mihaiu, O. Scarlat, G. Aldrica, and M. Zaharescu, J. Optoelectron. Adv. Mater., 5, 913 共2003兲. 28. M. K. Paria and H. S. Maiti, J. Mater. Sci., 17, 3275 共1982兲.
