The aim of the thesis is to produce high grade silicon from off-grade materials like sculls and process slags from Elkem Thamshavn. The methods investigated are dense medium separation, optical separation, flotation and metallurgical separation by remelting. Dense medium separation trials are conducted using magnetite suspended in water and aim to find the suspension density where only one phase will float. Optical separation experiments are done to identify light intensity thresholds between refractories, silicon and slag. The viability of flotation without surface activators and at neutral pH are investigated by measuring the zeta potential of each phase. Remelting experiments are done, building on previous work by the author, with the aim of investigating whether adding CaO or MgO to the slag will increase the settling efficiency. No results were obtained in the DMS experiments, because the viscosity of the suspension increased to infinity before the density of either slag or silicon was reached. The flotation experiments showed that flotation is not viable at neutral pH without surface activators, as the zeta potential of slag and silicon is nearly identical. The optical separation experiments were a success. The product fraction had an average silicon content of 74 wt%, compared to 52 wt% in the original off-grade material, while the waste fraction contained 7.5 wt% silicon. The separation efficiency was best for coarser grains, which is the expected result based on theory. The remelting experiments concluded that adding CaO or MgO to the slag will increase the settling efficiency. This was established both by chemical analysis and by surface area analysis of slag samples from each experiment. MgO seemed to contribute more to the settling efficiency than CaO, but because of the large variance between samples, this is not conclusive. The quality of the produced silicon was unaffected or improved by adding CaO, but additions of MgO increased the magnesium content slightly.