Let s Make a Mullite Matrix!

Transcription

1 Let s Make a Mullite Matrix! Published in Refractories Applications and News, in press. The fine art of

2 Introduction: Let s make a mullite matrix! Bjørn Myhre, Elkem Materials P.O. Box 8126 Vaagsbygd, 4675 Kristiansand, Norway In the paper Let s make a castable! 1, we went through proportioning of refractory castables, and the beneficial effect microsilica addition has on casting properties. It is however, not only flow and packing that are influenced by microsilica additions, also high-temperature properties like hot-strength are affected. In the following paper, arguments are put forward why microsilica additions at relatively high levels bring out the best of the alumina-silicate castable. Implicitly this also shows why substitution of microsilica by reactive alumina may not be such a good idea after all. To begin with, let us look at the picture below: These crystals are mullite crystals that are found on the fractured surface of an ultralow cement refractory castable based on white fused alumina and microsilica, with only.5wt% cement after hot-mor testing at 15 C. If the microsilica content is too low or the cement content too high, these mullite crystals will not form. Quite opposite, instead of this strong and robust bond, a liquid forms which causes catastrophic lowering of hot-strength. What happens and why? Alumina-silicate refractories: It is commonly accepted that a mullite bond-phase is beneficial for high temperature properties such as hot-strength and thermal-shock resistance. In bricks, the bond is normally in place when the refractory is installed. In castables on the other hand, the mineralogical make-up of the bond-phase goes through several intermediates during heat-up and hopefully ends up with the desired physical and mineralogical properties at service temperatures. If the composition of the bond phase is incorrect, the result may on the other hand be disappointing, 1 More about the proportioning and making of castables can be found in the paper Lets make a castable, downloadable from also published in Refractories Applications and News vol 13 (3)May/June 28 p Elkem Materials page 1

3 leading to softening and refractory failure at a much lower temperature than anticipated. If one succeeds in creating a bond system with mullite formation though, the softening may be postponed by several hundred degrees, which generally should provide improved service life. Relevant assumptions: In the following paper, it is assumed that the scenario is a refractory castable based on alumina-silicate aggregates with fine alumina and microsilica in the matrix together with cement. The variables are microsilica and cement, and it is assumed that there is always enough alumina available for the mullitization reaction. Typical sources for such alumina are: calcined alumina, reactive alumina, milled fused alumina (e.g.<74micron), milled tabular alumina (e.g.<45micron). Clay and aluminasilicates from the sillimanite group are not that suited since they all contain excess silica and are thus not ideal reaction partners for the microsilica. In the following, microsilica is considered the sole source of silica. If other aluminasilicates are used, parts of their silica content may take part in the reactions. Precise quantities are difficult to estimate however. The binary phase diagram Silica-Alumina: Figure 1: The phase diagram Al 2 O 3 -SiO 2 (Risbud and Pask 1978) 2 In Figure 1, the binary phase diagram after Risbud and Pask is shown with the metastable system superimposed (dotted). The phase diagram shows us that if we heat a mixture of silica and alumina, first of all a liquid appears from approximately C, and from this liquid, mullite is crystallized. At the lower temperatures, this is a slow process due to the high viscosity of the silicate glass and good mullite yield may first be attained after several hours. Figure 2 shows this temperature dependence for a castable with.5wt% cement and 8wt% microsilica 3. The castable was based on white fused alumina and hot-m.o.r. was consecutively measured after different soak times. 2 3 S. H. Risbud and J. A. Pask J. Am. Ceram. Soc., 6(9-1) (1977) B. Myhre and K. Sunde, "Alumina based castables with very low contents of hydraulic compound. Part II." Strength and High-Temperature Reactions of No-cement Castables with Hydraulic Alumina and Microsilica, in Proc. UNITECR 95, Kyoto, Japan, Nov , p. II/ Elkem Materials page 2

4 Figure 2: Hot modulus of rupture as a function of time at 14 and 15 C for a fused alumina based castable with.5wt% hydraulic alumina in combination with.5wt% cement. 8wt% microsilica. At 14 C, the strengthening (which is caused by the mullite formation) continues for perhaps more than a day, while at 15 C the reaction is completed within a few hours at temperature. At 13 C, unpublished data have indicated that the strengthening by the mullitization occurs at a very low speed. Once the mullite bond is established, it is permanent. This means that if e.g. a castable like that in Figure 2 is prefired for 5 hours at 15 C and then tested at 14 C, then the hot- M.O.R. will be higher than for a sample which has been held at 14 C for 5 hours. Around 25-3MPa would probably be obtained in such a test. This example was for a castable with only minute amounts of cement (i.e. CaO) so that we may correlate the strengthening with the predictions from the phase diagram, Figure 1, where a liquid forms and mullite crystallizes from this liquid. Castables normally contain several percent of cement to give green-strength. A certain amount of calcia is thus introduced. Since this lime is localized in the bond phase, which normally makes up around one third of a castable, the lime concentration becomes quite significant and has to be accounted for. The ternary phase diagram Al 2 O 3 SiO 2 CaO incorporates the cement: It is generally accepted that cement in combination with microsilica and alumina gives lower hot-strength with increased cement content. If you ask why this is so, the explanation is normally some vague reference to low melting liquids in the system, which is an explanation that may be good enough for some people, but has never satisfied me. During many years of research in refractory castables, the following conviction has gradually manifested: Basis for all the explanations should be the phase diagram. Particularly for castables based on relatively pure ingredients of silica, alumina and lime, it should be able to explain many observations by interpretation of the phase diagram Al 2 O 3 SiO 2 CaO. One should however always remember that the phase diagram supposes equilibrium, - which is normally not the case in your castable. Elkem Materials page 3

5 peritectic Figure 3: The phase diagram Al 2 O 3 -SiO 2 -CaO after Osborn and Muan. 4 Figure 3 shows the ternary phase diagram Al 2 O 3 -SiO 2 -CaO with additions relevant to the present topic. The gross composition: First of all, when dealing with a mullite bond, the mullite must be stable. This means that the composition of the castable and bond phase must be in one of the two compatibility triangles indicated in Figure 3, either i) corundum, anorthite and mullite; or ii) mullite, anorthite and silica. If the composition is outside, then mullite will not be stable, and if it forms at all, it will dissolve over time. Most alumina castables are fortunately in the corundum corner of the phase diagram. The hatched area in the corner indicates possible compositions (corundum castables) and relevant silica and cement contents are indicated. Cement and microsilica reactions: One may assume that the cement and the microsilica react during heating 6. If the cement composition is indicated on the calcia-corundum join (here a 71%CA cement) and a line is drawn towards the silica corner, one may by the use of the lever rule determine the minimum silica/cement ratio that is needed to enter the stability regions of mullite. Here as indicated, the minimum amount of microsilica as compared to cement is around 35wt% microsilica and 4 A. Muan and E.F. Osborn :"Phase Equilibria Among Oxides in Steelmaking" Addison-Wesley Publ. Comp., Inc., 1965 Elkem Materials page 4

6 65wt% cement. Or said in other words; a castable with less than 3.5wt% microsilica at 6.5wt% cement will start to dissolve mullite when it is heated. One important corollary that may be found in the phase diagram is connected to the peritectic at 1512 C, located close to the composition of Anorthite. Basic knowledge of phase diagrams and crystallization paths tells us that if we have a liquid with a composition within the compatibility triangle Corundum-Anorthite-Mullite (e.g. a molten castable) and cool it, then: 1) Corundum precipitates and the composition of the remaining liquid moves away from corundum, until 2) mullite starts precipitating and, 3) the last liquid disappears at 1512 C at the peritectic composition. For the formation of mullite in a castable, the following reaction pattern has been suggested 6 : Initially all microsilica and cement, possibly with some alumina, create a liquid supersaturated in silica at temperatures from approximately 13 C. The supersaturated liquid crystallizes mullite until a stable composition is attained. This stable liquid has the peritectic composition. One may add to this that upon heating, to 1512 C, the mullite is stable in this environment, after which the mullite starts to dissolve in an opposite pattern as the crystallization described above. Such a peritectic liquid has been detected both directly 5,6 and indirectly 7 in castables with mullite formation. The existence of a liquid from which mullite forms is not only found in cement containing systems, it is also a consequence of the meta-stable binary system shown in Figure 1. The mullite formation is seen in Figure 2 as the strengthening of the castable with time. Practical consequences: Since the phase diagram indicates that we should expect a stable (more or less) liquid that contains calcia, silica and alumina (15wt%CaO, 48wt%SiO 2, 37wt%Al 2 O 3 ), and that the microsilica and cement react to form the origin of this liquid, we cannot expect that mullite will be expelled unless there is an excess of microsilica. I.e. in a LCC with 6wt% CA cement (7wt%A, 3wt%C), more than 5.7wt% microsilica is bound up in the peritectic liquid. If 6wt% microsilica was used in the castable, less than 1wt% mullite should be expected in the final refractory, and total softening should commence below 15 C on heating. More microsilica gives proportionally more mullite and better strength should be expected. It should be emphasized that it is not only the stability of mullite that matters. Since the mullite is expelled from a liquid, the amount of residual liquid is important. To strengthen the castable, each mullite crystal must connect two or more aggregate grains. With more residual liquid more mullite is needed, and the stronger is the influence of increased temperature. A powerful tool to overcome some of these problems is to reduce the level of cement, as will be seen in the following section U. Schuhmacher: Untersuchungen an zementarmen und ultrazementarmen Korundfeuerbetonen. Dr. Ing. thesis, Rheinish-Westfälishen Technishen Hochschule Aachen, Germany B. Myhre, "Hot Strength and Bond-Phase Reactions in Low and Ultralow-cement castables" in Proceedings of UNITECR 93, Oct Nov Sao Paulo, Brazil p B. Myhre, A.M. Hundere, H. Feldborg, C. Ødegård: Correlation between mullite formation and mechanical properties of refractory castables at elevated temperatures Presented at VIII Int. Met. Conf. Ustron, Poland. May 25-28, 1999 Elkem Materials page 5

7 Examples: In this part of the paper, some examples of the mullitization in castables will be shown. One system based on white fused alumina, which serves as a model system, and one based on Chinese bauxite aggregates. Methodology: Hot- Modulus Of Rupture (H-M.O.R.) This picture shows the testing of the hot Modulus of Rupture of a castable sample (25x25x15mm) at a relatively low temperature, around 7-8 C. Hot M.O.R. testing gives relevant data on strength of castable at temperature, but in many cases where standards dictate a tooshort soak, the equilibrium values are never obtained. This applies in particular for some of the mullite forming reactions we deal with here, and we therefore advise a 24 hour soak prior to testing in order to advance some of the slower reactions. Common standards on hot-m.o.r. testing normally open for a wide variation in heating schedules and soak times, and these should always be indicated in reports and figures covering the issue. In Elkem s laboratories, heating rate is normally 3K/h (5K/min) up to temperature with a subsequent soak of some 3 min to equilibrate, or a longer soak depending on scope and thermal history. Refractoriness Under Load (R.U.L): Refractoriness under load is another thermo mechanical technique that is useful in the investigation of refractory castables. Briefly, the test consists of a furnace equipped with a sample holder that allows the measurement of the sample height as a function of temperature at a given load. The temperature is normally increasing by 3K/h, as in the Hot-M.O.R. testing. In the Elkem laboratories a load of.2mpa is normally applied, although some standards advise lower loads for unshaped refractory products. The sample is in the shape of a cylinder with a central bore in which a thermocouple and a measuring rod are placed which transmit the distance between the top and bottom. The readings must be corrected for the thermal expansion of the measuring rods in order to get correct numbers. Elkem Materials page 6

8 between the two lightly yellow alumina discs on top of the set up. Above the sample is the furnace, which is lowered onto the sample and the load is regulated by a set-up with counter weights. It is seen that this sample has been tested to a temperature with significant subsidence, due to the slight drum shape the cylinder attains during testing with deformation. As is the case with hot-m.o.r. measurements, the results can be vastly different depending on thermal history, and unless pre-firing conditions are given, the interpretation of the test may in some cases be very difficult or close to meaningless. In this picture, the experimental set-up is shown. The sample is the cylinder placed Castables based on white fused alumina: In the following section, the effect of microsilica and cement content on mullite formation will be presented for our model system based on white fused alumina. All aggregates are a high grade, white fused alumina. Together with calcined/reactive alumina, cement and microsilica, the castables were composed to have the same particle size distribution and water addition. To keep the particle size distribution as constant as possible, microsilica and reactive alumina with similar PSD were substituted for each other. More results, incl. recipes can be found in earlier papers 6, 8, 9 downloadable through The compositions are given in the Appendix, Tables 1 and 2, and are based on data from Ref. [8]. Some of the results (Figures 4, 7 and 8) are taken from earlier investigations ( ) and some (Figures 5, 6, 9, and 1) are recent (27) remakes of these mixes. As some of the original ingredients are no longer available, these have been replaced by recent replacement materials. As the chemistry and PSD is very similar, it is considered justified to compare the new results (obtained in 27) with the old data (from ). 8 9 B. Myhre and Aase M. Hundere: Substitution of Reactive Alumina with Microsilica in Low Cement and Ultra Low Cement Castables. Part I: Properties Related to Installation and Demoulding in Proc. UNITECR 97, New Orleans, USA, Nov , p Aa. M. Hundere and B. Myhre: Substitution of Reactive Alumina with Microsilica in Low Cement and Ultra Low Cement Castables, Part II: The Effect of Temperature on Hot Properties. Proc. UNITECR 97, New Orleans, USA, Nov. 4-7, 1997 p Elkem Materials page 7

9 Low cement castables 3 Hot M.O.R. (MPa) wt% MS 6wt% MS 4wt% MS 2wt% MS Temperature (C) Figure 4: Hot-M.O.R. of low cement (6wt% cement), fused alumina-based castables as a function of temperature. Castables with different amounts of microsilica. 24 hours at temperature. q=.25, max. particle size 4mm, 13 vol% water for casting ( wt%) From Ref.[7] In Figure 4, results from a low-cement castable with 6wt% cement are shown for several different microsilica contents. We see that at 14 C, strength increases as the microsilica addition increases. Particularly, levels of 6 and 8wt% microsilica seem beneficial to strength. It should be noted that mullite was not detected 7 at 14 C unless 6 or 8% microsilica was used. With 6wt% microsilica only 1wt% mullite was found, which is in good correlation with the reactions and consequences described earlier in this paper. The peritectic liquid, being approximately 11wt% of the castable may, at 14 C, be either highly viscous or even partially crystalline, which explains the strengthening effect seen at 14 C. At 15 C, the peritectic composition melts and starts attacking the mullite bond, with more or less total strength loss as result. In Figure 5, the effect of pre-firing of a low-cement castable on R.U.L results is seen. In this case we see that prefiring at high temperatures lowers the onset of final subsidence. Based on the findings of the sample prefired at 1 C, one could be led to believe that the castable would tolerate a temperature of 16 C. This is however difficult to understand if the hot- M.O.R. results shown in Figure 4 are taken into consideration, and even more if the phase diagram is consulted. The reason for this decline in refractoriness by the pre-firing, can be explained in the following manner: The mullite bond establishes so rapidly that the bond phase does not reach equilibrium. Then as equilibrium slowly commences, the bond phase is attacked and slowly dissolved by the lime-containing liquid; probably only partly, but at least sufficiently to allow a rapid subsidence at 15 C. Elkem Materials page 8

10 3 2 Prefired 1 C Prefired 15 C 1 Expansion [%] Temperature [ C] Figure 5: Refractoriness under load for a white fused alumina-based LCC with 6% cement and 8% microsilica. Samples pre-fired for 24 hours at 1 or 15 C. 2 LCC 4%MS LCC 8%MS 1 Expansion [%] Temperature [ C] Figure 6: Refractoriness under load. LCC based on white fused alumina. Samples with 4 and 8% microsilica pre-fired at 15 C prior to testing. Figure 6 shows that there is practically no difference in the R.U.L behavior for the LLC with 6% cement if 4 or 8% microsilica is used in the mix and if the castable has been pre-fired to 15 C. Even if no additional refractoriness (as per R.U.L.) is gained by use of 8% Elkem Materials page 9

11 microsilica, due to placing properties and hot strength (Figure 4), it may be a good idea, though, to use 8% microsilica and not substitute parts of it with reactive alumina. Reduced cement castables: We did see in Figure 4 that meltdown (at 15 C) of massive amounts (11wt%) of peritectic liquid softened the castable severely. The amount of this liquid may be reduced in two ways, either by lowering the cement content, or by lowering the silica content. If the latter choice is sought, according to the phase diagram, no mullite bond will be established, and also by entering other compatibility triangles (e.g. less than 3.5wt% microsilica for 6.5wt% cement) other eutectic and peritectic phases take over the role of our peritectic. These new phases have even lower melting points, one as low as 138 C. The latter is attained with microsilica contents between.6 and 3.2wt%microsilica at 6wt% cement. One may remove microsilica entirely, but then one excludes the use of alumina-silicate aggregates and fines. To be frank, taking the refractoriness into consideration, it is a better proposal to reduce the cement content. This is of course if the castable is intended to be used at temperatures approaching 15 C or higher, since at lower temperatures, the cement bond has so many positive attributes, including a high green-strength, etc. Figure 7 shows the effect of lowering cement content on hot-m.o.r. These castables all contain white fused alumina and 8wt% microsilica, and the variable is the cement content. All castables exhibit mullite formation, which is seen as an increase in hot-m.o.r. from 13 to 14 C. Typically there is a minimum at 13 C, and at this temperature, some plasticity may also be found. The softening may (at least for the.5wt% cement and no-cement castables) be connected to the metastable liquid formation of the binary system SiO 2 -Al 2 O 3 (Figure 1). At 15 C both the.5wt% cement and the cement-free compositions exhibit superior strength as compared to the low cement (6%) castable. The slightly better performance at 15 C for the cement-free castables as compared to the.5wt% castable, may be attributed to the tiny amounts (1wt%) of peritectic liquid in the latter, which start attacking the mullite bond upon further heating. It should again be stressed that the mullite formation is irreversible at sub-solidus temperatures, (<1512 C) and that the measured strength at 13 C will be higher for a castable that has been pre-fired at higher temperatures. This is because the mullite formation at 13 C is kinetically inhibited, -probably by the high viscosity of the liquid. Elkem Materials page 1

12 3 25 Hot M.O.R. (MPa) wt% cement.5 wt% cement No cement Temperature (C) Figure 7: Hot M.O.R. of fused alumina-based castables with 8 wt% microsilica as a function of temperature. q=.25, max. particle size 4mm, 13 vol% water for casting (4.2wt%). From Ref. [7] Ultralow cement castables: In ultralow cement castables, the amount of peritectic liquid is significantly reduced, and the fluxing effect is consequently much lower. The result is a castable that provided the aggregates are resistant- will be much more refractory than its low-cement counterpart. This difference is clearly shown in Figure 7. In Figure 8, the dependence of hot-m.o.r. on microsilica content is shown as a function of temperature. It is clear that with more microsilica, more mullite precipitates and stronger castables are made. Theoretically the peritectic should only account for.5% microsilica and mullite should be formed from the excess. However, such high amounts are normally not detected 7, which is probably connected to kinetic hindrances. The strong dependence of hot-m.o.r. on microsilica content at 15 C as seen in Figure 8 is again probably an effect of the melting of the peritectic phase. Although the melting point theoretically is 1512 C, early investigations 6 revealed that it contains relatively significant amounts of impurities, notably alkalis, which would lower melting temperature. A melting around 15 C is hence likely. Upon heating, this liquid attacks the mullite and it becomes important to have massive precipitations in order to maintain strength to high temperatures. Elkem Materials page 11

13 Hot M.O.R. (MPa) wt% MS 6wt% MS 4wt% MS 2wt% MS Temperature (C) Figure 8: Hot-M.O.R. of ultralow cement (.5wt% cement), fused alumina-based castables as a function of temperature. Castables with different amounts of microsilica. 24 hours at temperature. q=.25, max. particle size 4mm, 13 vol% water for casting ( wt%). From Ref.[7] The irreversible nature of the mullite formation must be remembered, and the fact that the softening at 13 C is only a metastable phenomenon related to the mullite formation, as is envisaged in Figure 9. Figure 9 shows a comparison of Refractoriness Under Load (5 K/min) for two parallels with different pre-firing conditions. The sample pre-fired at 1 C starts to subside at approximately 13 C, and then from approximately 15 C, mullite strengthens the sample sufficiently to regain strength. Final softening commences around C. With mullitization in place before testing, e.g. with pre-firing at 15 C, the softening does not appear until the castable starts its final subsidence at temperatures around C. All in all, a castable showing remarkably good properties, considering that it contains 8wt% microsilica, and a big improvement as compared to the low-cement castable shown in Figures 5 and 6. Elkem Materials page 12

14 2 ULCC 1 C/24h ULCC 15 C/24h 1.5 Expansion [%] Temperature[ C] Figure 9: R.U.L. (.2MPa) of ultralow-cement castable based on white fused alumina with.5wt% cement and 8wt% microsilica as a function of pre-firing conditions. In Figure 8, hot-m.o.r. of ULCC s with various microsilica contents were shown. Based on these results, a series of R.U.L. measurements were made and Figure 1 shows the results Expansion [%] % 2% 6% -1 4% % Temperature [ C] Figure 1: R.U.L. of ULCC pre-fired 24 hours at 1 C. Castables where parts of the m microsilica has been replaced by reactive alumina. Microsilica content indicated n next to the curves. Elkem Materials page 13

15 The characteristics of these castables are that, except for the microsilica/reactive alumina makeup, everything was sought to be kept constant. Particle size distribution was constant (by replacement of microsilica by reactive alumina), water addition (in vol%) was constant, raw materials were constant, etc. What we can see from Figures 8 and 1 is that a minimum amount of microsilica (4%?) seems to exist, which has to be exceeded in order to get visible effects of mullite formation. This may not only be caused by the reduction of mullite which forms by the lowering of the microsilica content, but may also be connected to kinetic hindrances. Another explanation is that in order to bond adjacent alumina grains a minimum amount of mullite is required. The next thing that Figure 1 tells us is that it may not be a good idea to replace microsilica by reactive alumina, - at least not in this kind of castable based upon white fused alumina. With alumina-silicate raw materials, the picture may not be that clear-cut, but these things have to be further investigated. Castables based on bauxite aggregates: So far, our examples have been based on very pure raw materials like white fused alumina. Since natural raw materials are much more commonly used, it may be interesting to see how use of such non-ideal components influences the mullite formation and strength. Thus, a series of bauxite-based mixes were prepared. The compositions are given in Table 3 in the Appendix. The castables were based on bauxite aggregates, but with milled, fused alumina and calcined alumina together with cement and microsilica in the bond phase. Calgon (SHMP) was used as dispersant. In Figure 11, the development of hot-m.o.r. as a function of time is shown for two of the compositions at 12 C and 13 C. The compositions shown are one with expected mullite formation: (8% microsilica + 2.5% cement) and one with doubtful or no mullite formation: (6% microsilica + 6% cement). The castable with 6% cement shows little development in strength with time, and shows a drop from 12 C to 13 C, while the one with 2.5% cement and 8% microsilica gets stronger both at 13 C and also with time. During the first few hours the 2.5% cement castable is weaker than the 6% cement castable due to establishment of a liquid phase prior to the mullite formation. This occurs at approximately 1 C lower temperatures than in the purer white fused alumina system described above, but in principle the same reactions are occurring. If the castable fulfils the requirements for mullite formation, a liquid establishes, and from this liquid the strengthening mullite precipitates. Typically, many refractory castables are based on a cement/microsilica ratio close to 1 (e.g. 6%cement + 6% microsilica) and as the standards for hot-m.o.r. testing in some countries advise 3 or 5 hours firing prior to testing, one may easily be mislead to choose the 6% cement + 6% microsilica over the 2.5% cement + 8% microsilica castable. Elkem Materials page 14

16 %MS, 6% cem. 12 C 6%MS, 6%cem, 13 C 8%MS,2.5%cem, 12 C 8%MS, 2.5%cem, 13 C 14 hot-m.o.r [MPa] Time [h] Figure 11: Bauxite-based refractory castables. Hot-M.O.R. as a function of time. After 24 hours pre-firing, much of the mullitization is considered completed. Figure 12 shows the hot-m.o.r. of bauxite castables with 3, 6 and 9% microsilica at 6% cement and also one with 8% microsilica and 2.5% cement as a function of temperature (composition is given in Table 3 in the Appendix ). The choice of 6% cement in combination with 6% microsilica is indeed not the best if the hot-strengths shown in Figure 12 are examined. Unless more than 6% microsilica is used, there are no signs of the characteristic strengthening by mullitization. This is in accordance with the mechanism sketched for the white fused alumina-based compositions described earlier in this presentation, but as mentioned earlier, some 1 C lower than in a pure system. The best results are again obtained by a lowering of the cement content, while maintaining a relatively high microsilica level. Here 2.5% cement together with 8% microsilica was used. Elkem Materials page 15

17 %MS, 6%cem 6%MS,6%cem 3%MS,6%cem 8%MS,2.5%cem 25. hot-mor [MPa] Temperature[C] Figure 12: Bauxite-based castables. Hot-M.O.R. as a function of temperature. Castables prefired 24 hours at test temperature. 2 Expansion [%] %MS, 6%cem 6%MS, 6%cem 3%MS, 6%cem 8%MS, 2.5%cem Temperature[ C] Figure 13: R.U.L. (.2MPa) of bauxite-based castables as a function of microsilica and cement content. Samples had been pre-fired at 1 C for 24 hours. Refractoriness Under Load (R.U.L.) testing is often used to assess refractoriness of a material. Together with strength measurements (hot-m.o.r), it is a good tool. Figure 13 shows the Elkem Materials page 16

18 R.U.L. for the castables of Figure 12. All samples had been pre-fired for 24 hours at 1 C prior to R.U.L. testing. We clearly see that all samples soften between 12 and 14 C, but for the castables with more than 6% microsilica, the subsidence comes to a halt around 14 C after an initial drop. This pattern is very typical for castables with mullite strengthening and as the mullitization is irreversible, higher pre-firing changes the appearance significantly, as seen in Figures 5, 9, 14 and % MS, 6% cem 1 C 9% MS, 6% cem 15 C 1 Expansion [%] Temperature[ C] Figure 14: R.U.L. of bauxite-based castable with 9% microsilica and 6% cement as a function of pre-firing. 24 hours at 1 C or 15 C. For the low cement castables based on white fused alumina it was mentioned that pre-firing at a high temperature lowers the onset of the final subsidence (Fig.5), and the explanation that was offered was that the mullitization is too fast for equilibrium to follow, and that the bond is attacked by the cement containing liquid in order to gain equilibrium. This explanation is also in accordance with the results of Figure 14. The reason why castables with lower cement content do not show the same tendency (Figures 9 and 15) is probably best explained by the amount of this attacking liquid. Less liquid gives less attack on larger amounts of mullite bond. Elkem Materials page 17

19 2 8%MS, 2.5%cem 1 C 8%MS, 2.5%cem 15 C 1 Expansion [%] Temperature[ C] Figure 15: R.U.L. of bauxite-based castable with 8% microsilica and 2.5% cement as a function of pre-firing. 24 hours at 1 C or 15 C. Concluding remarks: The intention behind this paper is to show that the reason for the thermal behavior of castables can be found and estimated from careful interpretation of established, fundamental principles like those in phase diagrams. This interpretation is however not always so simple and although it is regarded more like witchcraft by some, correct use is more likened to an art. What we have also shown is that the result of use of microsilica in high-alumina castable systems is very dependent on proper knowledge. Vast differences in behavior are obtained with only small changes in the cement/silica ratio. Unfortunately there has been a trend towards promoting what I would consider sub-optimal solutions, with a minimum of microsilica combined with liberal amounts of cement. Such conditions are promoting rapid melting, as shown above. Understanding mullite formation is critically important, and mullite is one of the significant factors when hot-strength of alumina-silicate systems is considered. It should not be necessary to get suboptimal results because of misunderstandings connected to the stability and formation of mullite. The results presented and the proposed mechanisms are not only valid for fused alumina and bauxite systems; quite a few raw material candidates exist. Members from the sillimanite group are very prominent, but possibly any high alumina system would benefit from this (theoretic) treatise. Elkem Materials page 18

20 APPENDIX Table 1: No/ultralow-cement castables. q-value =.25. Microsilica/reactive alumina ratio (vol%) 1/ 75/25 5/5 25/75 /1 Weight %: Alphabond (hydraulic alumina) Cement CAC 71% Alumina White fused alumina: -74 micron mm mm mm Microsilica Reactive alumina, BET 7.5m 2 /g, D5=.8µm Calcined alumina BET.8m 2 /g, D5= 4.5µm Citric acid (retarder).3 Darvan 811D (deflocculant) Water (13 vol%) Table 2: Low-cement castables with 6% cement. q-value =.25. Microsilica/reactive alumina ratio (vol%) 1/ 75/25 5/5 25/75 /1 Weight %: Cement CAC 71% Alumina White fused alumina: -74 micron mm mm mm Microsilica Reactive alumina, BET 7.5m 2 /g, D5=.8µm Calcined alumina BET.8m 2 /g, D5= 4.5µm Citric acid (retarder) Darvan 811D (deflocculant) Water (13 vol%) Elkem Materials page 19

21 Table 3: Bauxite-based castable compositions 9% MS 6% 6%MS 3% MS 8% MS Sample number cem. 6%cem. 6% cem. 2.5% cem. Chinese Bauxite: 1-4mm % Chinese Bauxite: -1mm % White Fused Alumina: -74micron % Cement :CAC 71% Alumina % Microsilica 971U (Elkem Materials)% Calcined alumina BET.8m 2 /g, D5= 5µm Additive (SHMP) Citric acid.1 water wt% Elkem Materials page 2