In terms of analytical accuracy Figs. 4, 5 and 6 show the experimental results with a 95% confidence level. There were no other major contaminants found in any of the slag samples beyond what was already noted. The measured values for each sub-sample by ICP-MS are average values from the detector and are delivered with their own relative standard deviation(RSD). Due to this inherent uncertainty the
Fig. 4 CSA404020 sample with errorbars set at 95% confidence level
Fig. 5 CSA255520 sample with errorbars set at 95% confidence level
Fig. 6 CSA254035 sample with errorbars set at 95% confidence level
STD values are calculated by taking the root of the sum of the variance between sub-samples and the average between the sub-samples own variance. This is expressed in Eq. 6, where the sub-expressions are from the ASTM standard for basic statistics E2586-14.
The experimental method used here has earlier been used by Jakobsson , and the method is simple to execute. This method is beneficial due to the fact that there is no flow in the crucible, and the reaction surface should be close to constant. Since the quench is immediate after leaving the hotzone it can be assumed that very little mass transfer happens during cooling giving higher accuracy. One problem with this method is that there is a high loss of slag when separating the two phases, due to the porosity of the slag when machining away the crucible. It is crucial that there are no small pieces of metal in the slag making the loss unavoidable if they are separated by hand. Another experiment using a crucible with an inner diameter of 70 mm and outer height of 150 mm was also performed, but there were considerable difficulty associated with the sampling. Glass tubes with an inner diameter of 4 mm were used inserted trough a graphite tube inserted into the melt. The problem was that considerable amounts of slag entered the sampling tube blocking the metal from entering. This problem is thought to be linked to the high viscosity of the slag and its surface tension. The amount of metal recovered was too small for any reliable analysis. An induction furnace was used for this experiment which induces flows among other things, making it less accurate for predicting specific properties, but it should be useful as another way of generating data to check and validate findings from other experimental methods. It is recommended that another method for sampling is used however.