melt corrosion of refractories in the nonferrous industry...

$: Melt corrosion of refractories in the nonferrous industry ...iupac.org/publications/pac/pdf/2011/pdf/8305x1093.pdf · in electric arc furnaces approximately 408 kg of refractories$
1093

Pure Appl. Chem., Vol. 83, No. 5, pp. 1093–1104, 2011.doi:10.1351/PAC-CON-10-10-05© 2011 IUPAC, Publication date (Web): 4 April 2011

Melt corrosion of refractories in the nonferrousindustry and the electric arc furnace: Athermochemical approach*

Viktoria Reiter1,‡ and Harald Harmuth2,†

1RHI AG, TCL Leoben, Magnesitstraße 2, 8700 Leoben, Austria; 2University ofLeoben, Peter-Tunner-Straße 5, 8700 Leoben, Austria

Abstract: A thermochemical approach was implemented to study the dissolution mechanismsof a wide range of refractory oxides and silicates in slags of the nonferrous metals industryand electric arc furnaces. First of all, the slags have been characterized regarding their work-ing range, phase assemblage, and melting behavior. Subsequently, the interactions of differ-ent combinations of refractory and slag have been examined, and emphasis has been placedon the determination of possible reaction products formed during dissolution and the solu-bilities of the refractory oxides and silicates in various slags. Direct dissolution, decisive inthe case of high corrosion rates, as well as indirect dissolution have been described. Varyingoperation conditions (e.g., temperature, atmosphere) have been incorporated into the investi-gations. In addition to the thermochemical calculations, the solubility of magnesia in fayaliteslags has been determined experimentally with the quenching method and the calculatedresults have been compared to published corrosion studies of other authors. These studiesrevealed that thermochemical calculations are a suitable tool to examine melt corrosion. Thethermochemical approach provides information that can be incorporated into product devel-opment and in the operation mode, giving a proper choice of process conditions.

Keywords: dissolution; melt corrosion; nonferrous metals; refractories; thermochemical cal-culations.

INTRODUCTION

In steel plants refractory costs account for approximately 2–3 % of total production costs. For example,in electric arc furnaces approximately 408 kg of refractories per 100 tons of steel are consumed. Therefractory wear should be minimized to reduce costs and to increase productivity that is lost by new lin-ings and repairs. To reduce refractory wear, emphasis must be placed not only on product developmentbut also on the steel-making practice with regard to slag chemistry and service conditions. Refractoriesin metallurgical vessels must resist high temperatures, thermo-mechanical stresses (erosion, abrasion),and attack of corrosive atmospheres, hot metals, and slags (chemical wear). For ceramically bondedmaterials melt corrosion proceeds via penetration and dissolution [1]. The penetration depth dependson the pore structure of the refractory material and the viscosity, surface tension, and wetting angle ofthe slag. A thermal gradient prevails from the hot face to the cold end of the refractory lining. Thisresults in fractional crystallization and freezing of the penetrated melt and thus a densifying of the

*Paper based on a presentation made at the 14th International Symposium on Solubility Phenomena and Related EquilibriumProcesses (ISSP-14), Leoben, Austria, 25–30 July 2010. Other presentations are published in this issue, pp. 1015–1128.‡Corresponding author: E-mail: [email protected]†E-mail: [email protected]

microstructure. The dissolution process may be direct or indirect. Indirect dissolution comprises the for-mation of one or more reaction products at the refractory/slag interface that may act as a protectivelayer. However, in the case of high corrosion rates—assuming a high erosion rate due to bath agita-tion—direct dissolution will prevail. First of all, the bonding phases and the matrix will be dissolved,resulting in a loss of bond that in turn favors the erosion of grains. This assumption is based on post-mortem investigations of used refractory samples. In the case of diffusion-controlled dissolution thecorrosion rate essentially depends on the ion flux j which may be expressed according to Fick’s FirstLaw [2] (1).

(1)

The ion flux j is defined by the effective diffusion coefficient (D), the effective thickness of theboundary layer (Nernst) (δ), the saturation concentration (cS), and the concentration of the componentin the slag (c0). The solubility limit is an important parameter in the dissolution process, e.g., slags sat-urated in MgO decrease MgO dissolution. A first approach to rank refractory components with respectto their resistance against slag attack is to compare their solubility limits. Refractory consumption canbe reduced by a proper choice of the refractory depending on slag chemistry in order to minimize theconcentration difference (cS – c0). For many decades phase diagrams have been used successfully toevaluate solubilities and phase equilibria in refractory corrosion [3]. Nowadays, self-consistent thermo -chemical databases are so advanced that computational thermodynamics can be applied successfully tomulticomponent phase equilibria. This study has put emphasis on thermodynamic modeling of disso-lution mechanisms (direct and indirect) of refractory material components in slags of the nonferrousindustry and electric arc furnace slags. Thermochemically self-consistent databases are fundamental foraccurate and reliable predictions. The thermodynamic calculations have been performed with the soft-ware package FactSage as the FactData have been proven to be one of the best optimized commerciallyavailable data sets for oxide systems. The Gibbs energy minimization modules Equilib and PhaseDiagram have been used in combination with the FACT53, FToxid, and FTmisc databases. Proposalsfor possible improvements in the field of process performance, lining design, product selection, andproduct development are provided in order to decrease refractory wear and increase service life.

THERMOCHEMICAL MODELING AND EXPERIMENTAL PROCEDURE

Figure 1 represents schematically the approach for the thermochemical modeling of the corrosionmechanisms.

V. REITER AND H. HARMUTH

© 2011, IUPAC Pure Appl. Chem., Vol. 83, No. 5, pp. 1093–1104, 2011

1094

jD

c cS�= −δ

( )0

Fig. 1 Schematic representation of the applied methodology.

As the slag is the attacking medium and its characteristics have a crucial influence on the corro-sion, first the working range of the slags and their phase constitution have been described in detail.Afterwards, the refractory/slag equilibria have been determined for direct and indirect dissolution. Theinvestigated materials comprise the main phases (oxides and silicates) of basic and nonbasic refractorymaterials (see Table 1). The dissolution behavior of the refractory material has been characterized withthe following parameters: (a) reaction products occurring at the solid/liquid interface including theircompositions, (b) solubility limits, and (c) amount of refractory material dissolved in 100 g slag.

Table 1 Investigated refractoryoxides and silicates.

Oxides Silicates

MgO Mg2SiO4Al2O3 CaMgSiO4Cr2O3 Ca3MgSi2O8ZrO2 Ca2SiO4MgCr2O4MgAl2O4

The calculations have been adapted to several service conditions by varying temperature (gradi-ent from the hot face to the cold end), atmosphere (oxygen partial pressure), refractory/slag ratio, andlocal slag composition. The calculated results have been evaluated by comparison with published resultsof corrosion studies by other authors. These studies comprise microanalytical investigations of refrac-tories after crucible tests. Indirect dissolution was observed as reaction products have been formed atthe refractory/slag interface [4,5]. Additionally, the solubility of MgO has been determined experimen-tally in fayalite slags with the quenching method. Mixtures of MgO and fayalite slag with increasingrefractory/slag ratios have been equilibrated in air at 1550 °C. Subsequently, the samples have beenquenched in water to freeze the phase assemblage at test temperature and if possible to solidify theresidual melt in a glassy state. The phase constitution and the compositions of the phases have beendetermined with microscopy and microanalytical methods and have been compared to calculatedresults.

Slags of the nonferrous metals industry

Copper-converting slags can be classified with respect to the process technology as fayalite slags (dis-continuous process, e.g., Peirce Smith converter) and calcium ferrite slags (continuous converting, e.g.,Mitsubishi process). However, these slags have several drawbacks. Fayalite slags are characterized byhigh viscosity and a high risk of magnetite precipitation, while calcium ferrite slags show a low solu-bility for SiO2 and are very corrosive. The newly proposed ferrous calcium silicate (FCS) slags shouldavoid these drawbacks and combine the advantages of fayalite and calcium ferrite slags [5,6]. In thecopper converter a magnesiachromite lining is common practice as it shows the highest resistanceagainst the mostly used fayalite slags. Nevertheless, there are attempts to replace magnesiachromitebecause of the possible formation of harmful hexavalent chromium. The investigations should point outfrom a thermochemical point of view which refractory materials could be taken into account as replace-ments for magnesiachromite and how the slag chemistry influences dissolution behavior (calcium fer-rite, fayalite, and FCS slags). The refractory components will be ranked with rising corrosion resistance.Furthermore, the advantages of FCS slags with regard to lower dissolution rates are pointed out sincethere is limited data regarding the wear behavior of FCS slags.


Melt corrosion of refractories 1095

Corrosion environmentFour fayalite slags (1–4), one calcium ferrite slag (5), and one FCS slag (6) have been incorporated inthe investigations (see Table 2). The fayalite slags 1–3 and the calcium ferrite slag are converter slagsof different copper mills, slag 4 represents a synthetic fayalite slag. The composition of the FCS slaghas been chosen to assure that the slag is completely liquid at process temperature and lies in the vicin-ity of the primary phase field of dicalcium silicate.

Table 2 Chemical composition of the investigated slags.

[wt %] 1 2 3 4 5 6

Fe2O3 69.5 52.6 48.9 66.4 59.5 40SiO2 24.9 23.2 26.3 30.1 1.3 30CaO 0.3 5.2 0.6 – 19.2 30MgO 0.4 2.1 0.8 2.3 0.5 –Al2O3 2.1 7.4 1.1 0.4 0.5 –Cu2O 2.3 0.7 13.4 – 18.2 –ZnO 0.2 8.0 5.7 – 0.8 –PbO 0.3 0.8 3.2 – – –

The main components of the slags of the nonferrous metals industry are CaO, SiO2, and FeOx.The molar ratio Fe/SiO2 of the fayalite slags ranges from 1.3 to 1.95. The FCS slag has a basicity,expressed by the molar ratio CaO/SiO2, of 1. Processes working with fayalite slags necessitate an oxy-gen partial pressure (pO2) of 10–8–10–10 atm to avoid magnetite precipitation; in the case of calciumferrite slags, the pO2 may rise to 10–5 atm.

Slag characterizationThe working range of the slags of the nonferrous metals industry is mainly defined by the liquidus sur-faces in the three-component system CaO–SiO2–FeOx that are shown in relation to the oxygen partialpressure in Fig. 2. Additionally, the compositions of the investigated slags are inserted. The workingrange of calcium ferrite slags is determined by the liquidus area at the base line of the triangle. The sec-ond liquidus area that is relevant for fayalite and FCS slags decreases in size with increasing oxygenpartial pressure. Besides the main components, other oxides have to be considered as the liquid domainis enlarged by Al2O3, ZnO, PbO, and Cu2O and diminished by MgO.

The valency of iron strongly affects the phase composition and the melting behavior of the slags.Olivine with inclusions of MgO, CaO, and ZnO is the main phase of fayalite slags under reducing con-ditions. With increasing oxygen partial pressure the ratio Fe2+/Fe3+ decreases and the phase composi-tion shifts toward spinel related to magnetite. The copper-containing phase changes with decreasingpO2 from cuprospinel (CuFe2O4) to delafossite (Cu2Fe2O4) and finally to metallic copper. The mainphases of calcium ferrite slags are dicalcium ferrite, calcium ferrite, and a copper-containing phase thatdepends again on the prevailing oxygen partial pressure. The phase composition of the FCS slag is sim-ilar to the fayalite slag, but wollastonite or rankinite are also stable.



1096

Thermochemical modeling of refractory/slag equilibriaIndirect dissolution: The reaction products that may be formed at the refractory/slag interface are sum-marized in Table 3; however, the levels were dependent on the specific slag/refractory component ratio.The investigations yield the following results for the dissolution of MgO: attack of MgO by fayaliteslags results in a complete dissolution of MgO at low MgO/slag ratios, followed by the formation ofolivine, magnesiowustite, and spinel with rising MgO/slag ratios. Zirconia does not form any reactionproducts in combination with the iron silicate and FCS slags, whereas CaZrO3 was formed with the cal-cium ferrite slag.

Table 3 Potential reaction products at the refractory/slag interface. Abbreviations areCaAl4O7 (CA2), CaAl12O19 (CA6), and Ca2SiO4 (C2S).

Fayalite slag Refractory Calcium ferrite slag

Spinel, cristobalite, eskolaite Cr2O3 Spinel, C2S, CaCr2O4, eskolaiteOlivine, spinel, magnesiowustite MgO Spinel, magnesiowustiteSpinel, corundum Al2O3 Spinel, melilite, CA2, CA6, corundumZirconia ZrO2 CaZrO3, zirconiaOlivine, spinel MgCr2O4 Spinel, C2S, CaCr2O4Olivine, spinel MgAl2O4 Melilite, spinel

The amount of dissolved refractory material in 100 g slag is shown for indirect dissolution inFig. 3.



Fig. 2 Liquid domain in the CaO–SiO2–FeOx system at 1250 °C in relation to oxygen partial pressure.

Cr2O3 and MgCr2O4 exhibit the lowest solubility, nearly independent of slag composition due tothe formation of an iron-containing spinel. The amount of dissolved magnesia is 1–3 g. Zirconia showsa lower solubility in the calcium ferrite and FCS slags due to the formation of CaZrO3. In fayalite slagsno reaction products are formed and the solubility of ZrO2 decreases with decreasing Fe/SiO2 ratio. Thesolubility of alumina increases with rising CaO content of the slags and is highest for the calcium fer-rite slag. The solubility limits of the refractory components decrease with decreasing temperature andoxygen partial pressure linearly with the logarithm of oxygen partial pressure. MgO and Cr2O3 solu-bility increases 1–2 wt %/100 °C, whereas the solubility of Al2O3 increases 5 wt %/100 °C. Silicate sol-ubilities in iron silicate slags increase from forsterite (Mg2SiO4) to monticellite (CaMgSiO4) to mer-winite (Ca3MgSi2O8) to dicalcium silicate (Ca2SiO4) (see Fig. 4). In calcium ferrite and FCS slags thevarious silicate solubilities are of the same order of magnitude. Additionally, the silicate solubilities areconsiderably lower in these slags compared to the iron silicate slags due to the high initial CaO contentin the calcium ferrite and FCS slags.

Direct dissolution: The calculated refractory oxide solubilities in all the slags were higher fordirect dissolution compared to indirect dissolution (Fig. 5). Cr2O3 and MgCr2O4 solubilities were thelowest compared to the other oxides in the iron silicate and FCS slags; however, in the calcium ferriteslag they were higher relative to MgO and ZrO2. The solubilities of Cr2O3, MgCr2O4, and MgOdecreased with increasing slag basicity, namely, the solubilities were lower in the FCS slag. Therefore,the use of FCS slags should be favored in combination with refractory products based on these oxides.The solubility of alumina increased in the FCS slag due to its dependence on the CaO content of theslag. Slag 2 shows a high initial Al2O3 content, and thus the solubilities of Al2O3 and MgAl2O4 arelower compared to other fayalite slags.



1098

Fig. 3 Amount of refractory oxides dissolved in 100 g slag at 1250 °C and for indirect dissolution.

Fig. 4 Amount of silicates dissolved in 100 g slag at 1250 °C and for indirect dissolution. Abbreviations areCa2SiO4 (C2S), Ca3MgSi2O8 (C3MS2), CaMgSiO4 (CMS), and Mg2SiO4 (M2S).

Silicate solubilities in the slags are illustrated in Fig. 6. As determined for indirect dissolution, thesolubility of forsterite was significantly lower in the iron silicate slags compared to the CaO-containingsilicates. The silicates showed considerably higher solubilities than the refractory oxides. The dissolu-tion of these silicates results in bonding loss, weakening the refractory microstructure and enhancingerosive wear. Once again, the FCS slag demonstrated lower silicate solubilities compared to the iron sil-icate slags.

In fayalite and FCS slags the solubilities of the oxides can be ranked with increasing solubility:Cr2O3, MgCr2O4, ZrO2, Al2O3, and MgO. The higher basicity of FCS slags leads to a lower solubilitycompared to fayalite slags. FCS may therefore lead to higher corrosion resistance or they may con-tribute to a chrome-free lining. It has been shown that besides proper product selection the process con-ditions, especially the slag composition, have a crucial influence on refractory wear. Refractory wearalso depends considerably on the temperature and oxygen partial pressure. Thus, process control withregard to constant conditions also plays an important role.

Evaluation of corrosion studiesExperimental determination of the solubility of MgO in fayalite slags: The calculated and experimen-tally determined phase composition of the magnesia/slag 4 equilibria is shown in Fig. 7. Here, <A>denotes the mass percentage of magnesia relative to the sum of magnesia plus slag. The reaction prod-ucts formed are spinel, olivine, and magnesiowustite. The homogeneous composition of the phases indi-cates that equilibrium has been reached in the experiments.



Fig. 5 Amount of refractory oxides dissolved in 100 g slag at 1250 °C and for direct dissolution.

Fig. 6 Amount of silicates dissolved in 100 g slag at 1250 °C and for direct dissolution.

The experimentally determined amount of spinel is higher compared to the calculations up to<A> = 25 wt %, resulting in a lower amount of liquid. In the mixtures, magnesiowustite is not stableuntil <A> = 40 wt % and so the amount of olivine is considerably higher. The cooling rate was too lowto freeze the liquid completely glassy, i.e., magnetite, olivine, and a residual glass phase formed fromthe melt during cooling. These phases can be distinguished from the reaction products formed at1550 °C due to their morphology, and the area that represents the former melt at 1550 °C can be definedexplicitly. The experimentally determined and calculated compositions of spinel, olivine, and the resid-ual melt are shown in Fig. 8 in relation to <A>.

The calculated and experimentally determined compositions of olivine agree well, whereas thespinel has a higher magnesia and a lower iron content compared with the calculations. The residual meltshows a higher SiO2 and a lower Fe2O3 and MgO content as calculated. This deviation arises from thehigher amount of spinel, whereby a higher amount of iron from the slag has reacted to spinel leading toan enrichment of SiO2 in the residual melt. Additionally, the magnesia content of the spinel in thequenching tests is higher compared to the calculation leading to a lower amount of dissolved magnesia.The results of the calculations and the quenching tests agree well with respect to the qualitative phaseassemblage and the composition of the reaction products. Tolerable differences exist in the quantitativephase composition and the composition of the residual melt. Therefore, this method can be used to eval-uate the calculated results as well as to determine the solubility of oxides for which no adequate ther-mochemical data are available.



1100

Fig. 7 Experimentally determined and calculated phase composition of the magnesia/slag 4 equilibria in air at1550 °C.

Corrosion studies of other authors: The investigated reaction products formed at therefractory/slag interface as well as the solubilities of magnesia and Cr2O3 agree well with the results ofthe thermochemical calculations (see Table 4). However, the solubility of alumina in calcium ferriteslags is lower than in the calculations due to the different CaO content of the slags used in the corro-sion studies and the calculations.

Table 4 Solubilities [wt%] of MgO, Al2O3, and Cr2O3.

Fayalite Calcium ferrite

Calc. Lit. Calc. Lit.

MgO 2.8–4.1 2–5 [4] 1.8–3.1 1.6–3Al2O3 8–11 9–15 26–30 14Cr2O3 0–0.5 0–0.4 [4] 0.8–3 0.8–2 [6]



Fig. 8 Experimentally determined and calculated compositions of the phases in air at 1550 °C: (a) spinel, (b)olivine, and (c) residual melt. The solid lines and points represent the results of the calculations and the quenchingtests, respectively.

Electric arc furnace slag

The main refractory materials used in the slag line of electric arc furnaces are carbon-containing bricks(MgO–C bricks) and a MgO-based repair mix [7]. The influence of the dissolution rate on the slag com-position (basicity), temperature, and oxygen partial pressure is discussed below.

Corrosion environmentSix slags (medium attacking) from three different electric steel plants were characterized. The threeplants were chosen to cover a wide range of steel-making practices (see Table 5).

Table 5 Chemical composition of electric arc furnace (EAF)slags.

[wt %] A-I A-II B-I B-II C-I C-II

CaO 31.7 34.5 37.6 40.6 39.2 53.0SiO2 22.4 13.5 18.7 17.7 18.6 16.3Fe2O3 25.6 34.1 23.9 23.9 20.6 14.3Al2O3 12.2 9.3 11.5 9.4 7.2 5.1MgO 3.1 3.1 5.2 5.2 12.4 9.2MnO 5.1 5.5 3.1 3.1 2.1 2.1C/S 1.41 2.55 2.01 2.29 2.11 3.25

The chemical compositions of slags A represent average analyses, whereas the slags of client Band C were sampled after tapping. EAF slags typically contain six major oxides (CaO, MgO, SiO2,FeO, MnO, Al2O3). The basicity (C/S ratio) ranges from 1.41 to 3.25. This is an indicator of the dif-ferent methods of EAF refining in steel plants [7].

Thermochemical modeling of refractory/slag equilibriaThermochemical modeling of MgO/slag interactions revealed refractory brick, and mix dissolution iscontrolled by the diffusion of FeO and MnO into the magnesia and the formation of a magnesiowuestiteboundary layer. The formation of a solid reaction product at the refractory/slag interface indicated a dif-fusion-controlled indirect dissolution process. The solubility limit of MgO and the resulting concentra-tion difference as considered in eq. 1 are shown in Fig. 9.

The solubility limit of MgO ranges from 4.4 to 12.6 wt %. These values agree well with publishedsolubility limits [8,9]. The negative concentration difference of MgO for slags C-I and C-II reveals sat-uration in MgO. Plant C uses a slag conditioner providing MgO. No further dissolution of the refrac-



1102

Fig. 9 Solubility limit of MgO at 1650 °C and log pO2 = –9.

tory material will take place except for the silicates. Whilst MgO refractories are severely attacked byunsaturated slags, in the case of saturated slags the oxidation resistance of carbon in MgO–C refracto-ries becomes the dominant factor affecting corrosion resistance. The impact of basicity and temperatureon the solubility limit of magnesia is shown in Figs. 10a,b.

A decrease in slag CaO/SiO2 increased the MgO solubility, and dissolution of MgO would beaccelerated. These results indicated highly basic slags would be less corrosive toward MgO grains. Atemperature rise of 100 °C resulted in a cS increase of 1.2–2 wt %. With one exception the resultsshowed a trend of higher MgO solubility with increasing oxygen partial pressure. Since trivalent iron isless basic than the bivalent form, the basic magnesia is less stable against Fe3+ resulting in higher sol-ubility limits at increased oxygen partial pressures. Additionally, the FeO–MgO system is completelymiscible, whereas the Fe2O3–MgO system shows a miscibility gap. A change of the operation mode,especially the slag basicity, can be recommended to client A to reduce refractory corrosion. A C/S-ratiosimilar to client B and C should be strived for.

CONCLUSION

In the present study, it has been shown that thermochemical calculations are suitable to qualitativelyexamine melt corrosion. The output of the calculations can be incorporated into product developmentand the operation mode. They can be used to make predictions regarding the dissolution levels of refrac-tory materials that may help to reduce costly experimental investigations. The study demonstrates thatthe dissolution rate of refractory components is significantly influenced by slag composition. For exam-ple, the use of FCS slags in the nonferrous metallurgy may reduce the dissolution rate of refractories



Fig. 10 (a) Solubility limit of MgO in relation to basicity at 1650 °C and log pO2 = –9; (b) solubility limit of MgOin relation to temperature at log pO2 = –9.

based on magnesia and magnesiochromite compared to fayalite slags. However, whilst it is importantto consider the impact on refractory dissolution, the primary slag selection criteria are based on metal-lurgical process requirements. Good agreement between the calculated results and the experimentallydetermined phase compositions indicates that thermodynamic modeling is an appropriate method todetermine refractory dissolution mechanisms in metallurgical slags. The calculations supply reliableresults. So far there have been no attempts to incorporate the actual conditions of corrosion occurringin metallurgical vessels, for example, fluid flow, thermo-mechanical stresses, diffusion, and mass trans-port. To get a comprehensive model of refractory wear, thermochemical and thermomechanical consid-erations as well as computational fluid dynamics (CFD) calculations (mass transfer coefficients) mustbe combined. So far the simulation of subprocesses enables a better understanding of erosive and cor-rosive wear. For example, CFD can be used to calculate the mass transfer coefficient and in combina-tion with the concentration difference the dissolution rate can be calculated. The thermomechanicalwear, the crack formation, and joint opening that support the slag penetration and the melt corrosioncan be described with finite element methods. Additional process simulation tools should be applied todetermine the temperature, atmosphere, and local slag compositions in various metallurgical processesas exactly as possible. A challenge for the future is to link different simulation methods for a compre-hensive description of corrosion, qualitatively as well as quantitatively.

REFERENCES

1. W. E. Lee, S. Zhang. Int. Mater. Rev. 44, 77 (1999).2. W. E. Lee, S. Zhang. Int. Mater. Rev. 45, 41 (2000).3. W. E. Lee, B. B. Argent, S. Zhang. J. Am. Ceram. Soc. 85, 2911 (2002).4. J. R. Donald, J. M. Toguri, C. Doyle. Metall. Mater. Trans. B 29, 317 (1998).5. N. P. Fahey, D. R. Swinbourne, S. Yan, J. M. Osborne. Metall. Mater. Trans. B 35, 197 (2001).6. A. Vartiainen, M. Kytö. Scand. J. Metall. 31, 298 (2002).7. Y. Hoshiyama, Y. Ishihara. J. Tech. Ass. Refr. 21, 247 (2001).8. S. Zhang, H. Sarpoolaky, N. J. Marriott, W. E. Lee. Br. Ceram. Trans. 99, 248 (2000).9. A. Sander, F. Verhaeghe, B. Blanpain, P. Wollants, R. Hendricks, G. A. Heylen. Steel Res. Int. 77,

317 (2006).



1104

melt corrosion of refractories in the nonferrous industry...

Documents