The presence of nonmetallic inclusions in steel, particularly oxides such as alumina and calcia-alumina, contribute to problems during steelmaking and can negatively affect the appearance and mechanical properties of final products. A cleaner steel has better formability and is less prone to fatigue and corrosion. Furthermore, during steel production in the melt state, a cleaner steel is less prone to cause process control problems such as clogging during pouring and teeming.

The removal of inclusions has been studied previously in both experimental dissolution in slags and in steel melt flow models, where the positions of simulated inclusions are tracked computationally. This study is an examination of the steps between these, where inclusions approach the steel/slag interface and separate across it. Because the interface is known to be of high energy, often with a chemistry different from the bulk, the effect on any inclusions close to that interface is of critical importance. The near-interface behavior of inclusions during removal is separated into two parts, (i) the actual separation across the interface, and (ii) the approach to the interface and the interface's resulting deformation.

In this research, previous models for the removal of inclusions, as it relates to the steel/slag interface, are studied and expanded upon to include alternate shapes and reactions between phases. It is observed that aluminum oxide inclusions of 5 to 100 μm are typically not impeded by the interface during the separation step using ladle, tundish, and mold slags of moderate viscosity (0.06 to 0.6 kg/m/s), requiring only microseconds to milliseconds (respectively) to separate. Certain values of the interfacial tension between slag and inclusion, σ_IS, can however cause spherical inclusions to become trapped at the interface due to the change in overall energy (interfacial tensions between a calcia-silica-alumina slag and the liquid metal are on the order of 1 N/m, but are typically 0.2 N/m or less between solid alumina and similar slags). This value is studied experimentally, finding that for typical situations, σ_IS is not large enough to directly cause settling of alumina inclusions at the interfaces of ladle, tundish, or mold slags. The expanded separation model considers the effect of shape on separation for octahedrons and plates, and the effect of dissolution occurring alongside separation. In this manner, the limiting factors and 'lifetime' of an inclusion at the steel/slag interface is described, so that inclusions that do not fully separate (and settle in local energy minimums) are limited by the dissolution due to the inclusion surface exposed to the slag, and does not remain trapped indefinitely. A high slag viscosity can slow an inclusion during this transition to a settled state several times compared with lower viscosities, and the initial speed, which for high values causes a small delay due to the transition to a preferred speed. However, neither of these effects cause separation to reach critically important times. The other shapes examined have similar separation times, though do not have settling points. This would appear to indicate that for most situations, the removal of alumina inclusions are not impeded by their separation at the steel-slag interface, except in unlikely situations.

Because previous models only considered inclusions already in contact with - or very close to - the interface, a description of the initial approach of an inclusion to the interface is described via fluid dynamics, wherein the particle decelerates in the bounded fluid and eventually deforms the interface. Micrometer-sized particles are slowed considerably (1.7 seconds) due to a limited drag field, while larger inclusions cause more significant deformation, which can delay the final point of rupture by almost two seconds, even if their initial slowdown due to a limited drag field is small.

To verify these results, a water-oil tank was constructed and the movement of buoyant particles from a lower phase into an upper phase is studied using a high speed camera. Again, small particles show slowdown but little deformation, while larger particles can cause interface deformation, a related delay before rupture, and even entrapment of lower phase with the upper phase. Recorded movement of the particle is compared with Stokes movement and the bounded drag approach to a static interface. Observed interface deformation is compared with modeled interface deformation, as a function of particle size, and deviations explained.

For a spherical alumina inclusion that is 5 μm in radius, the approach - as modeled in this thesis - has a delay of 1.7 seconds from the time required for terminal velocity, while its separation takes 14 μs, and dissolution (assuming full contact with the slag) requires 18 ms. A similar particle of 100 μm radius is delayed by a tenth of a second during approach, up to two seconds due to interface deformation (assuming small rupture distances and low interfacial tensions), 280 μs during separation, while dissolution requires about 150 seconds.