If the 180 min entries from Figs. 4, 5 and 6 are compared with equilibrium values calculated by FactSage it can be seen that they are not consistent with the calculated values. Table 2 lists the FactSage and experimental 180 min entries. The Factsage values are calculated using the Fact Oxide- and Fact Light Metals database, at 1600 °C.

From Figs. 4, 5 and 6 it can be seen that while there are some sub-sample pairs which exhibit high degree of variance, the sample setts as a whole there looks to be a high degree of homogeneity between the two sub-samples. This excludes a consistent non-representative sampling of the final product as the culprit for this deviation. The deviation could come from not retrieving enough of the phase in question when separating metal and slag. This might be a reasonable train of thought if the problem lie in the slag, but not for the metal. Another step that might affect the samples could be contamination by the slag when crushing. This does not hold water since the metal and slag were crushed by two different machines.

Operational factors concerning the procedure or the materials used are next in line. The procedure and materials have been tried and tested previously by

Table 2 Al and Ca concentrations, in [ppm], calculated by FactSage, using the Fact Oxide- and Fact Light Metals database, at 1600 °C, in equilibrium with SiO2-CaO-Al2O3 slags, with the experimental values






















Jakobsson [9]. Jakobsson did not look at kinetic aspects however so it could be that inserting a cold sample into the hot furnace, and the extra heating associated with this causes some deviations in the early samples. Since the deviation is also found in the later samples this is not considered likely to be the cause of the drastic effect seen here. If this has a large impact then the initial mass transfer should be lower due to a lower temperature, and not much higher as seen from two of the samples. It should be noted that Jakobsson [9] samples show much better agreement for Al then for Ca when compared with the literature.

Next it would be logical to look at the analysis method. ICP-MS is a common method employed when looking for trace elements in metals and slags. The preparation might cause problems due to the fact that HF reacts with Si and Ca. NIST proposes using property data from Chase [10] for CaF2 and Lyman and Noda [11] for SiF4. From these it can be seen that SiF4 is more volatile than CaF2, so Si should react first. Jakobsson [9] found that the Si loss was not significant, but if the mixture that was used for all of the samples did cause a loss of Si then this could explain the high concentration of Al and Ca. The calibration mixture for the ICP-MS machine could also be off. When looking at the NIST standard samples the analysis has a low variance for Al while the variance of Ca is much higher. From this evaluation it seems that calibration problems, or dissolution of the samples might have caused problems when viewing the exact values for the impurities. If this is the case then while the values might be off they should all deviate by approximately the same amount allowing one to view the relative concentration behavior. Al does not react strongly with HF, compared to Si and Ca, and the analysis showed consistent values comparing for the NIST standard. Due to this the behavior of Al discussed further.

< Prev   CONTENTS   Source   Next >