The aim of this thesis work is to increase the knowledge of inclusion behavior at the steel-slag interface by mathematical modeling and in-situ Confocal Scanning Laser Microscope experiments. Mathematical models based on the equation of motion predicting liquid and solid inclusion behavior was first investigated. Four main forces, the buoyancy force, the added mass force, the rebound force and the drag force, act on the inclusion as it crosses the interface. There are three types of behavior an inclusion at the steel-slag interface can adopt. These are a) pass, which means that the inclusion is separated to the slag, b) remain, where the inclusion stays at the interface without being fully transferred to the slag or c) oscillate, and the inclusion rises and descends at the interface until the motion is dampened out by the interfacial forces. The studies showed the importance of accurate experimental physical property data. Application of the models to industrial conditions illustrated that useful plots could be made showing the industry how to optimize their interfacial properties in the ladle and tundish to obtain maximum inclusion separation. In-situ Confocal Scanning Laser Microscope (CSLM) experiments were carried out in order to study agglomeration of liquid and semi liquid inclusions at the steel-gas and steel-slag interfaces and in the slag. Liquid-liquid inclusion agglomeration at steel-gas and steel-slag interfaces was seen to not occur without using force. However, when already transferred to the slag the inclusions agglomerated freely due to a higher free energy force. Comparison of experimental and theoretical agglomeration force showed good agreement between experiments and theory. The main conclusion of this work is that inclusion separation is a complex field of study and there exist no model that takes everything into account. Here the tendency for inclusion transfer and how to manipulate the physical properties for inclusion separation together with agglomeration experiments have been studied. For the future maybe coupling of models for computational fluid dynamics, agglomeration, inclusion separation, dissolution and slag entrainment in addition with experimental physical property data can provide a better overview and understanding. / QC 20100823