8/12/2019 jom-2000_d859_v1.pdf

1/14

Corrosion of Refractory Aggregates by Molten Aluminum

Saied AFSHAR and Claude ALLAIRE

CIREP/CRNF, Dept. of Eng. Physics & Materials Engng.,Ecole Polytechnique of Montreal (CRIQ campus),

8475 Christophe Colomb Street,Montreal, Quebec, H2M 2N9

Abstract:

The use of a non-wetting additive is not always a sufficient solution to improve thecorrosion resistance of refractories against molten aluminum. Once in contact with liquidaluminum, the coarse refractory aggregates, which do not benefit from the non-wettingadditives in the matrix, may be corroded. In some cases, the corrosion of aggregates

promotes the corrosion of the surrounding matrix, even in the presence of a non-wettingagent. The present study evaluates the corrosion resistance of some fifteen different typesof aggregates commonly used in monolithic refractories and analyses their effect within anon-wetting matrix. Based on the present experimental results, the selection of aluminosilicate aggregates that improve the corrosion resistance of castables to aluminumalloys can be made according to their chemical composition.

Introduction:

Because of their low cost and their good thermal and mechanical properties,aluminosilicate refractories have been widely used in the various processes involved inaluminum production. In particular, in the aluminum cast-houses, the aluminosilicatecastables are often employed for the lining of the holding and melting furnaces. Thesematerials can be roughly described as a mixture of aggregates, having differentcomposition and size, and fine reactive powders, composed mainly of alumina, silica andcalcium oxide. Once mixed with the necessary amount of water and dried, the castablesshould be fired at an appropriate temperature to acquire the required properties. Duringfiring, the fine particles of refractory react together to form the new solid phases whichact as a continuous cement between the aggregates and constitute a rigid matrixsurrounding each aggregate. This provides the refractory with the desirable properties to

resist to the thermomechanical abuses during furnace operation.

However, many types of aluminosilicate castables, especially those containing asignificant amount of free silica and/or impurities such as the alkalis, exhibit poor corrosion resistance in contact with molten aluminum. To solve this problem, in the lastdecade, the refractory producers have developed new technologies and products for thefurnaces used in aluminum industry. The most common solution brought to this field isthe use of some additives making the refractory non-wettable by aluminum.

To better identify the factors involved in the chemical reactions of refractory with moltenaluminum, the corrosion resistance of samples, made only from the fine particles of a low

8/12/2019 jom-2000_d859_v1.pdf

2/14

cement castable, has been recently examined via a laboratory corrosion test 1. Theadvantage of such method is to minimize the eventual effect of the coarse aggregates andto better appreciate the chemical stability of the refractory matrix in contact with the

liquid aluminum. Previous works have been investigated on the corrosion of aluminosilicate materials, where effort has been focused on the role of non-wettingagents on the protection mechanism of refractory matrix 1-4 . In fact, due to the usualfiring temperatures of castables, the coarse aggregates of refractory are generally notsubjected to significant phase transformations and, consequently, cannot benefit from theeffect of most non-wetting additives 1. Therefore, through a direct contact with liquidaluminum, the aggregates can be corroded and, under some conditions, this corrosionmay affect the matrix of refractory even in the presence of a non-wetting agent.

The purpose of the present work was to evaluate the corrosion resistance of differenttypes of aggregate usually employed in aluminosilicate low cement castables. The

general aspects of corrosion as well as the effects of different factors such asmicrostructure, size and chemical composition of aggregates on their surroundingrefractory matrix were investigated. Three categories of aluminosilicate aggregates,according to their corrosion behavior are presented here.

Experimental procedures:

Materials:

Three series of aggregates denoted A, B, and C were used to evaluate their corrosionresistance to aluminum attack. Series A consisted of four categories of aggregatesselected from an aluminosilicate low cement castable according to their color. Theseaggregates are designated as A-White, A-Yellow, A-Gray, and A-Black in the text.

Series B corresponded to the samples prepared from the aggregates classified in series A.The sample preparation consisted of grinding the aggregates of each category of series A.In order to obtain an identical shape for all the samples, the resulting powders were thendry-pressed at room temperature and fired at 1200 0C for 5 hours. The series B sampleswere designated, by referring to the original materials, as B-White, B-Yellow, B-Gray,and B-Black. It should be noted that the chemical composition of the starting materialsmust not be affected by the preparation procedure used for series B samples. Using thesame procedure, some other samples were prepared from a commercial high purity silica

powders (designated as B-Silica).

Series C covered 11 types of aggregates, which currently are used in monolithicrefractories. They are designated as C-a to C-k and, similarly to series A, were used asreceived. The chemical compositions of all the aggregates are listed in Table I.

8/12/2019 jom-2000_d859_v1.pdf

3/14

Preparation of samples:

To facilitate the handling of the samples and also to evaluate the potential effect of

corroded aggregates on the corrosion of the refractory matrix, the aggregates weremounted in a relatively inert material. The latter was obtained from the fine particles of alow cement castable containing BaSO 4 as a non-wetting. The previous works 1,4 havedemonstrated that the samples made from these refractory fine components (see Table IIfor chemical composition) present a good corrosion resistance to molten aluminum, whenthe pre-firing temperature is less than 1050 0C.

The preparation details of the tested samples were as follows: The fine components (lessthan 500 m in size) of the above castable, referred here to as refractory matrix, werefirst mixed with about 10 wt.% water and casted on a vibrating table. To ensure a goodadherence, the aggregates were inserted in the refractory matrix immediately after

casting. This was done in order to expose only one face of the aggregates to the moltenalloy during the corrosion test (Figure 1).

All samples were dried for 24 hours in air, then heated for about 10 hours in an oven at110 0C and finally fired at 1000 0C for 5 hours. As mentioned earlier, this firingtemperature preserves the good performance of non-wetting agent to protect therefractory matrix during the corrosion test.

Corrosion test:

The corrosion resistance of the samples to aluminum attack was evaluated using acorrosion immersion test procedure 5 (Figure 2). In this test, an inert crucible was chargedwith 2 identical samples and 2 kg commercial grade aluminum (99.9% purity). Then itwas heated from room temperature to 850 0C, at a rate of 120 0C/h, in a vertical furnace.One hundred grams of magnesium were added to the liquid metal at about 700-800 0C toobtain a nominal Al-5 wt.% Mg alloy. The tests were carried out over a four-day periodwith a daily addition of 40 g magnesium to maintain a constant composition of alloyduring the test.

After carried out the test, samples were taken out of the liquid metal and allowed to coolslowly to room temperature. Visual examination of corrosion was performed on a centralcross section of the tested samples.

Results and Discussion:

Corrosion aspect:

In general, the affected area by molten aluminum in a refractory material is visible bynaked eye due to its black appearance. Sometimes, the affected zone is just limited to adiscolored (black) layer of refractory contacting molten aluminum, without anysignificant change in the microstructure of material. The term corrosion or corroded

8/12/2019 jom-2000_d859_v1.pdf

4/14

area is employed here when a deep modification in the original texture of the sample isdetected, for example, under optical microscope examination. In this case, the corrodedarea appears as a composite material, composed mainly of small grains of corundum

surrounded by an interconnected metallic network. Therefore, a simple electric resistancemeasurement may often permit to determine rapidly whether the affected zone wassubjected to metal penetration (i.e. to corrosion) or not.

Figure 3 shows three different aspects of samples after the corrosion test. In Figure 3-a,virtually no trace of corrosion is observed neither in the aggregate nor in the surroundingrefractory matrix. In Figure 3-b, corrosion is only observed in the aggregate contactingthe molten metal. While, in the last case (Figure 3-c), corrosion is obvious in both theaggregate and the refractory matrix. Considering the fact that the refractory matrixremained intact during the corrosion tests for the two first cases, it seems reasonable tothink that the observed corrosion in the refractory matrix, in the latter case, has been

initiated from the corrosion of the aggregates. This observation confirms the previousresults 1 indicating that, in some cases, the corrosion of aggregates within a refractorycastable may lead to the corrosion of their surrounding matrix.

Microstructure effect:

Figure 4 illustrates the corrosion aspect of the series A samples and their equivalent in theseries B. Among the series A samples, only the A-White resisted perfectly to aluminumattack. For the samples A-Yellow and A-Gray, the aggregates are, respectively, partiallyand deeply corroded but their refractory matrix are only locally discolored. The worstcase is A-Black sample, when both aggregate and matrix are heavily corroded.

The main objective of series B samples preparation was to obtain a regular shape andespecially an identical size for all the aggregate samples. This should make thequantitative comparison of the corrosion results much easier. However, the laboratory

prepared aggregates visibly presented a much better corrosion resistance compared to theseries A samples. Microscopic examination of samples revealed that only a slight layer (about 100 m thick) of corrosion has been produced in B-Black aggregate, whereas the

black area on the top of B-Gray aggregate, as well as in the refractory matrix,corresponds more to a discoloration than a corrosion process.

Although, the chemical compositions of materials are the same for both series A and B(see Table I), the difference in their corrosion behaviors should be related to their difference in microstructure due to the laboratory procedure employed for the series Bsamples preparation. Figure 5 shows the microstructure of A- and B-Yellow samples. Themicrostructure of B-Yellow seems to be more homogeneous and apparently contains lesslarge pores compared to A-Yellow sample. These results imply that, in some cases, thedensification of an aluminosilicate material may significantly decrease the kinetics of corrosion.

8/12/2019 jom-2000_d859_v1.pdf

5/14

Size effect:

As mentioned earlier, the corrosion of aggregates may extend to their surrounding

refractory matrix during the laboratory corrosion test. This behavior can be explained bythe fact that the corrosion products of some aggregates should, at least in the corrodedregion, modify the chemical composition of the alloy, making it more aggressive withregards to the refractory matrix. In such a case, the degree of contamination of the metalshould be proportional to the volume of the corroded aggregate. Moreover, the extent of affected region of the refractory matrix should depend on the size of the aggregatesinvolved during the corrosion.

Figure 6 illustrates the size effect of two poor corrosion resistant aggregates (B-Silica andC-j) on the corrosion aspect of the samples. For both types of aggregates, the surface areaof the affected refractory matrix (dark region) appears larger for the big aggregates than

for the smaller ones.

A similar interpretation can be made on the corrosion aspect of some low cementcastables. Figure 7 shows the cross sections of two aluminosilicate castables, containing anon-wetting agent, after the Alcan immersion test 5. For these experiences, the 5.08 x2.54 x 2.54 cm samples have been cut from the fired cast material, in order to have 3cross sections and 3 original surfaces for the corrosion test. Thus, the specific area of aggregates contacting the metal bath, as well as the volume of aggregates subjected tocorrosion, is quite larger for the cross sections than for the original surfaces. This mayexplain the more significant corrosion took place on the cut surfaces of the refractorysamples.

Chemical composition effect:

The results of the immersion tests for the C series samples are presented in Figure 8. Inthese series, only samples C-a and C-b exhibited great resistance to aluminum alloyattack during the corrosion tests. These samples contain more than 99 wt.% of alumina(see Table I), which may explain the excellent corrosion resistance of such materials. For the C-c samples, despite their high alumina content (more than 90 wt.%), a thin corrodedlayer with a thickness up to a few hundred microns is observed in some aggregates. Thissuggests that the high content of alumina in a refractory material is not always adominant parameter to prevent the corrosion by aluminum. The presence of some oxides,such as alkalis, even at low quantity, may favorably contribute to the metal attack 1,2,4,6 .This is the case for C-d and C-e samples which were totally corroded, most probably dueto their greater alkali content compared to the C-c aggregates. The same argument can bemade to explain the more severe corrosion produced in the C-i in comparison with the C-h (partially corroded), despite the lower alumina content of the latter.

Table III puts together two chemical composition parameters, determined from Table I,and corrosion aspect of aluminosilicate samples of series A and C. To better illustrate thechemical composition effect of the aggregates on the corrosion behavior of the samples,the above results are reported by a graph. On this graph, the ratio values of major

8/12/2019 jom-2000_d859_v1.pdf

6/14

reducible oxides, such as SiO 2, Fe 2O3 and TiO 2 (see Table IV), over the Al 2O3 content of the aggregates are reported on the Y-axis. The X-axis present the alkaline oxides (Na 2O +K 2O) content of the aggregates. According to the above considerations, the corrosion

results of the tested aggregates (series A and C) are shown in Figure 9.

From such representation, three zones, separated by two approximate boundaries, can bedistinguished: For the low values of (SiO 2+Fe 2O3+TiO 2)/Al2O3 ratio and alkali content(zone I: both less than 0.3), aggregates should exhibit a good corrosion resistance. For high values of these parameters (zone III: both more than 0.5), aggregates can becorroded and their corrosion may extend to the refractory matrix. Between these twozones, there is an intermediate zone (zone II), where the aggregates can be subjected to acorrosion process but this corrosion act does not seem to affect the rest of refractory.

It should be noted that the boundaries between these different zones could be shifted

forward the higher values in X and/or Y axis, by improving the microstructure of aggregates, for example through the densification procedures used for series B sample

preparation.

It is clear that, for aluminum furnaces application, the aggregates situated in the zone IIIare not desirable, considering the fact that the extension of their corrosion to therefractory may decrease the service life of the furnace lining.

Acknowledgements:

The authors are very grateful to Narco Canada Inc., Alcan International Ltd. and theCentre Qubcois de Recherche et de Dveloppement de l'Aluminium (CQRDA) for their financial contribution during the realization of this work.

8/12/2019 jom-2000_d859_v1.pdf

7/14

References:

1. S. Afshar, and C. Allaire, The Corrosion of Refractories by Molten Aluminum ,

JOM, 48 [5], pp 23-27 (1996).

2. M. Allahverdi, C. Allaire, and S. Afshar, Effect of BaSO 4 , CaF 2 , and AlF 3 as well as Na 2O on the Aluminosilicates having a mullite like composition , Journal of theCanadian Ceramic Society, 66 [3], pp 223-230 (1997).

3. M. Allahverdi, S. Afshar, and C. Allaire, Corrosion Resistance of AluminosilicateCeramics to Molten Al - 5% Mg Alloy , Advances in Refractories for theMetallurgical Industries II, CIM Proceedings, d. par M. Rigaud et C. Allaire, pp.295-303, aot (1996).

4. M. Allahverdi, S. Afshar, and C. Allaire, Additives and the Corrosion Resistance of Aluminosilicate Refractories in Molten Al-5 Mg, JOM, pp. 30-34, February (1998).

5. C. Allaire, Refractories for the Lining of Holding and Melting Furnaces , Advancesin Production and Fabrication of Light Metals and Metal Matrix Composites, d. par M.M. Avedesian et al., CIM Proceedings, pp. 163-174 (1992).

6. C. Allaire and P. Desclaux, Effect of Alkaline and of a Reducing Atmosphere on theCorrosion of Refractories by Molten Aluminum , Journal of the American CeramicSociety, 74 [11 ], pp 2781-2785 (1991).

8/12/2019 jom-2000_d859_v1.pdf

8/14

Table I: Chemical composition (in wt.%) of the aggregates.

Material SiO 2 Al2O3 Fe2O3 TiO 2 CaO MgO Na 2O K 2OA, B-White 28.9 67.1 0.93 2.56

8/12/2019 jom-2000_d859_v1.pdf

9/14

Table III: Corrosion test results on series A and C aluminosilicate samples.

Sample (SiO 2+Fe 2O3+TiO 2)

Al2O3

Alkalis (wt.%)

Na 2O+K 2O

Aggregate Surrounding

refractory matrixA-White 0.48 0.15 Not corroded Not corrodedA-Yellow 0.61 0.1 Partially corroded corrodedA-Gray o.65 0.11 corroded corrodedA-Black 0.98 0.17 corroded corrodedC-a 0.002 0.21 Not corroded Not corrodedC-b 0.001 0.16 Not corroded Not corrodedC-c 0.08 0.5 Partially corroded Not corrodedC-d 0.12 0.89 corroded Not corrodedC-e 0.12 0.7 corroded corrodedC-f 1.16 0.41 corroded corrodedC-g 0.69 0.1 corroded Not corrodedC-h 0.47 0.1 Partially corroded Not corrodedC-i 0.34 0.22 corroded Not corrodedC-j 0.67 0.36 corroded corroded

Note: Not Corroded = no trace of corrosion is observed. Partially corroded = corrosion zone in the aggregate is less than about 1 mm. Corroded = significant corrosion is observed (more than 1 mm thick).

Table IV: Free enthalpy of some of the oxides in refractory materials.

Oxides 660 oC 850 oC 1000 oCCaO -256 -249 -242MgO -143 -233 -226BaO -225 -215 -208

Al2O3 -220 -210 -203TiO 2 -177 -167 -162SiO 2 -153 -160 -167FeO -98 -92 -87

Note: Oxides listed below aluminum oxide can be easily reduced by molten aluminum unlike the others.

8/12/2019 jom-2000_d859_v1.pdf

10/14

Figure 1: Configuration of the sample.

Figure 2: Corrosion immersion test.

28 mm

20 mm

Refractorymatrix

Aggregate

Al-5%Mg

Samples

8/12/2019 jom-2000_d859_v1.pdf

11/14

Figure 3: Different corrosion aspects of tested samples. a) A-White ; b) C-g ; c) C-j

Figure 4: Comparison of samples in series A (left) and B (right). From top to bottom, White, Yellow, Gray and Black aggregates, respectively.

8/12/2019 jom-2000_d859_v1.pdf

12/14

Magnification: 200Figure 5: Optical micrographs taken from the samples A-Yellow (top) and B-

Yellow (bottom).

Figure 6: Size effect of aggregates on the corrosion of samples. Top: small (left) and large (right) B-Silica aggregates ; Bottom: small (left) and large (rigth) C-j aggregates.

8/12/2019 jom-2000_d859_v1.pdf

13/14

---- Cut surfaces

Figure 7: Cross section of two low cement castables, containing a non-wetting agent,after the corrosion immersion test. The sample on the top shows no corrosionof the refractory matrix enhanced by the corrosion of the exposed aggregates

(on the cut surfaces) unlike the sample at the bottom.

Figure 8: Corrosion aspects of series C samples.

8/12/2019 jom-2000_d859_v1.pdf

14/14

Figure 9: Effect of the chemical composition of the aggregates on the corrosion behavior of the samples.

0

0,2

0,4

0,6

0,8

1

1,2

1,4

0 0,2 0,4 0,6 0,8 1

Na2O + K 2O (% poids)

( S i O

2 + F e 2

O 3

+ T i O

2 ) / A l 2 O

3

None corroded aggregates

Partially corroded aggregates

Totally corroded aggregates

Aggregates and refractory matrix corroded

zone I

zone II

zone III

zone I