A comparison of simulated and analysed phase compositions of solidified EAF slag

Transition towards fossil-free iron and steel production increases the role of electric arc furnaces (EAF), whereas the role of blast furnaces (BF) decreases. This has drawn an increasing attention towards the valorisation of EAF slag with the target in the high value applications such as supplementary cementitious materials (SCM). This is supported not only by the declining production of the granulated blast furnace slag (GBFS) currently widely used as a SCM, but also common targets for both cement and steel industry to reduce carbon dioxide emissions. Due to their different chemical and mineralogical composition, structure and properties in comparison to GBFS, the EAF slag cannot be used as SCMs as such. It is widely considered that usage as SCM would require treatments such as modification of the composition towards lower basicity, removal of certain components (e.g. via reduction), controlled and sufficiently fast cooling as well as a feasible way to even out the potential differences due to use of different kind of raw materials in the EAF; that is, steel scrap, DRI reduced with either hydrogen or other reductants, Fe-containing secondary materials such as briquetted dusts and so on. The purpose of this study has been to supplement the experimental studies with thermodynamic simulations in which both equilibrium and Scheil-Gulliver calculations were made for different slag systems with varying compositions. Although not being able to simulate the formation of amorphous glass phase obtained with fast cooling of low basicity slag, computational thermodynamics nevertheless offer a tool to quickly estimate the effect of different slag treatments on the stabilities, amounts, compositions and solidification orders of solid crystalline phases as well as solidus and liquidus temperatures. According to the results, the simulations can predict the amount of monoxide phase relatively well, whereas the amount of spinel is overestimated in the simulations. The most probable reason for this is the lack of time for the nucleation and growth of the spinel phase in the rapid cooling of the experimental samples. Concerning the compositions, the compositions of the spinel phases were predicted more accurately in comparison to the monoxide phase, for which the shares of MgO and FeO varied significantly between the samples.