SOLIDIFICATION OF BOF SLAGS MODIFIED BY PYROMETALURGY AND SUBMITTED AT DIFFERENT COOLING RATES

In a scenario of blast furnace slag shortage, the use of BOF slag (Blast Oxygen Furnace) as a partial substitute in cement production is a low cost alternative. In this paper, BOF (Blast Oxygen Furnace) slag samples were melted in an experimental electric arc furnace (EAF), with different modifiers materials (Al top dross reject, slag from FeSi iron alloy production, et al.) with different rates of additions, subject to differentiated cooling in specifically designed experimental apparatus. The objective was to search the formation of different phases formed during a process of BOF slag modification, for different materials additions containing Al2O3 and SiO2 and subjected to different cooling conditions. Modified slags, with basicity (CaO/SiO2) below 1.1 and Al2O3 contents less than 5%, subjected to cooling rates up to 21°C/s, favor the formation of crystalline phases; no amorphous phases were detected in this trial. Modified slag tests with basicity below 1.2 and Al2O3 content between 10.0% and 12.5% favor both the formation of crystalline (majority) and amorphous phases (between 25% and 30%). The influence of the cooling rate up to 21°C/s on amorphous phase formation in this trial was not conclusive.


Design of Experiments
The strategies used in the design of experiments for slag modification were as follows: avoid CaO-free and MgO-free formation; reduce basicity, aiming to replicate blast furnace slag by adjusting the SiO 2 content; adjust the Al 2 O 3 content of the slag, targeting blast furnace slag level.
The starting point for dimensioning the experiments were the ternary/ quaternary diagrams of the CaO-SiO 2 -MgO system [2], taking into account contents of 0%, 5% and 10% of Al 2 O 3 .With these parameters, the CaO-free and MgO-free fields disappear with the increase of Al 2 O 3 .In ternary diagrams for 15%, 20%, 30% and 35% Al 2 O 3 this trend continues.Another observation is that above 10% of Al 2 O 3 , Monticellite is no longer the most stable phase, turning into Spinel, this phase being the predominant, from Al 2 O 3 contents above 30% in the CaO-SiO 2 -MgO system.
Thus, it was possible to predict trends for mineralogical phases that would be present in the modified slag, as shown in Table 1.In this way, the criteria that met the premises mentioned in this topic were established in the design of experiments: basicity lower than 1.1 and MgO content less than 10% and increase in Al 2 O 3 content.
A total of fifteen modification tests were carried out in BOF slags, and these were divided into two groups: Crystalline trials (12 heats), which aimed to reduce the initial basicity slag to lower values than 1.1.Second group called Amorphous trials (3 heats), which had as objective to increase the content of Al 2 O 3 above 10% maintaining also in this case, the reduction of basicity to values below 1.1.

Slag Modification Pyrometallurgy
A batch of 5t of solid BOF slag and granulometry between 10 and 30mm, was obtained from COSIPA steel mill, located in São Paulo, Brazil.This material was re-melted in an experimental EAF with 300kg nominal capacity, using

INTRODUCTION
Slag from the BOF process usually accounts for between 9 and 15% of liquid steel production, and is an important component in the primary refinery stage of a steelworks.However, in the current state-of-the-art, the vast majority are discarded, although there is a large literature on the subject of their reuse, a robust and consistent solution is not yet agreed [1][2][3][4][5].The use of industrial waste as a raw material for the manufacture of cement has stood out in the cement industry, which research viable alternatives and the slag BOF process is one of the most investigated.Blast furnace slag is the most used in cement production, due to the favorable proximity of its chemical composition to that of Portland cement, besides having satisfactory hydraulic reactivity.In Brazil, steel production is stagnant, but civil construction, a major consumer of cement, is being demanded, so that the substitution of blast furnace slag for BOF slag would be a low cost strategic response.
The BOF slags, although they present chemical components common to Portland cement have serious restrictions regarding their use in the construction industry due to the expansion caused by the oxidation of Iron (Fe 0 and Fe +2 ) and CaO-free and MgO-free expansion effect after hydration [6].In the HBM process described by Yao-Hung Tseng et al., oxygen and SiO 2 are injected into the liquid slag in order to reduce basicity and inhibit the formation of CaO-free and MgO-free, enabling this material for many applications [7].
The BOF slag is of a basic and oxidizing nature, and for the Brazilian current steel mills conditions, it has in its chemical composition: CaO 35-40%, SiO 2 10-15%, MgO 9-15%, P 2 O 5 1-2% and total iron (in the form of Fe +2, Fe +3 and Fe 0 ) of the order of 19 to 25%.Certain crystallographic phases may favor the cementitious properties of BOF slags like Belite (2CaO.SiO 2 ) and Alite (3CaO.SiO 2 ).Another factor is the influence of the different cooling rates of the liquid slag, i.e., BOF slags that are air-cooled, are generally inert due to the crystallization of their oxides.Blast furnace slag has significant cementitious properties, because when submitted to granulation by high cooling rates, besides the high proportion of amorphous phase, they it contains Alite and Belite.
The major pure solid precipitates that can be observed within a BOF slag are: Merwinite (Ca After reaching the temperature of 1680°C and 1700°C, the charge is liquid and the melting stage completed; the melt was then poured into an experimental slag pot and then transferred by a crane to a turntable and tilting platform.Through an additional rail with digital analytical balance weighing control, it was possible to add several types of additives and granular stabilizers based on SiO 2 and Al 2 O 3 to adjust the chemical composition, thus modifying its basicity and alumina content.
After this "slag pot metallurgy" step, the samples were tapped manually through a metallic shell into specific molds/devices, with different cooling rates.Figure 1 shows experimental process flow to obtain molten BOF slag modified, with the respective orders of magnitude of the temperatures involved in each step.

Experimental Devices for Slag Cooling and Evaluated Cooling Rates
Modified slags produced in the metallurgical reactor through the addition of modifying agents had samples cooled in different ways in order to investigate their effect on the crystallization of the modified slag.Three experimental cooling devices were used as follows:

Device 1 -refractory sleeve, thermally insulated at base -slow cooling (1° C/s)
It is a cylindrical ceramic sleeve cast at both ends, supported on a refractory base, as shown in Figure 2, having been prepared on the laboratory bench.This device is characterized by slow cooling, and through measurements by thermocouples, a cooling rate of less than 1 °C/s was evaluated during the experiments.

Device 2 -refractory sleeve, cooled in a copper base -medium cooling (5°C/s)
The slag is poured into a system comprising a ceramic mold, which receives liquid slag disposed on a water-cooled chute.The same cylindrical ceramic sleeve in device 1 is used, but now rests on a copper base cooled internally by water.
Thermocouples were inserted into the mold, in order to record the cooling rates of the modified slag from high cooling rates (near the shell) to lower rates (region away from the shell).Figure 3 shows a schematic in the glove placed on the mold and filled with slag.

Device 3 -circular cooled mold -quick cooling (21 °C/s)
Figure 4 shows the circular cooled mold for quick cooling, where it can basically be described as a copper mold, subjected to intense heat exchange by cooling with water, during the collection and pouring step of the liquid slag sample.For the calculation of the cooling rate of liquid slag cast in the circular shell, a computational simulation was performed with COMSOL Multiphysics™ software.

Chemical analysis
Part of the aliquots obtained from the samples, which, after milling and homogenizing, were prepared for analysis by the X-Ray fluorescence spectrometry method  using PANalytical™ MiniPal equipment.The Fe 3+ content was calculated by the difference between the total iron, Fe 2+ and metallic iron.

X-ray diffraction Rieteveld (XRD-Rieteveld)
The remaining part of the aliquots described in 2.4.1 Chemical Analysis, in the form of powders, were analyzed by X-Ray diffraction (XRD) in Rigaku™ WindMAX 1000 diffractometer, using Cu Kα, 40 kV, 20 mA and scanning of 0.2°/min.

FactSage  Thermodynamic Calculation
Thermodynamic calculations were performed in software FactSage  7.0 to predict the phases that form during the solidification of the slag.In these calculations, the tool used was the Equilib with FToxid and FactPS data bases.FToxid-SlagA database was used for the calculation of liquid solutions.For solid solutions FToxid-MeO_A presented as RO phase in the results; FToxid-bC 2 S, FToxid-aC 2 S and FT-oxid-Mel presented as Melilite; and FToxid-OlivA presented as Olivine.
The calculations were based on the chemical compositions presented in Table 2.The gas phase was not considered in the calculations.The results obtained by the thermodynamic simulation were compared with the results of the XRD analysis.

Chemical Composition
The chemical compositions for the fifteen tests slag modification are presented in Table 2, parameterizing the quantities of CaO, SiO 2 , FeO, MgO, MnO, Fe 2 O 3 and Al 2 O 3 values in 100%, which are used for thermodynamic simulation as previously stated.The values of basicity for the Crystalline group trials are in the range of 0.7 to 1.3 and the Al 2 O 3 content in the range of 2.2 to 4.7%; for the Amorphous group trials, basicity was obtained in the range of 0.9 to 1.2 and Al 2 O 3 levels in the range of 10.0 to 12.5%.Table 3 shows the cooling conditions of the slag for each heat performed.work.It is shown in Figure 6, where the XRD-Rietveld of Heat #1 (Crystalline trials) the phases present, where no amorphous is detected.Same in Figure 7, for Heat #13, (Amorphous trials), where a 25,1% amorphous was detected.
Using this same comparison procedure for the other heats tested, with all data compiled,the results for Merwinite, Melilite, Monticellite and Wüstite are shown in Figures 8, 9, 10, and 11, for comparative visualization.

Thermodynamic Calculation and XRD-Rietveld analysis
In Figure 5, the thermodynamic simulation of Heat #1 (Crystalline trials) is presented, showing the phases present between liquidus and solidus temperature, assuming thermodynamic equilibrium, in a theoretical process of infinite time solidification.This pattern of thermodynamic solidification has been followed in all heats tested in this      The comparison of the phase prediction by the thermodynamic equilibrium model and XRD-Rietveld was shown to be partially consistently coherent, with reproducibility of the thermodynamic theoretical results, reinforced with the physical verification of the XRD-Rietveld.Quantitatively, however, the comparative results between the phases showed erratic dispersions between the percentage obtained by simulation and XRD-Rietveld.This can be explained by the fact that the solidification in the devices is not done under equilibrium conditions (as opposed to the thermodynamic simulation) and also because of the heterogeneities caused by the variations in the cooling velocity that occurs between the (cooler) wall and the nucleus (warmer) forming crystals of different phases and sizes.

Crystalline trials
All heats in this group had only crystalline phase, independent of the cooling rate applied.In this group, the basicity was between 0.7 and 1.3 and the maximum content of Al 2 O 3 was 4.7%, where the XRD-Rietveld do not indicate the presence of amorphous phase.

Amorphous trials
In this case, all the heats of this group presented both crystalline phase and amorphous phase, for contents above 10% Al 2 O 3 .In this group, the basicity was between 0.9 and 1.2, similar to the crystalline trials and the Al 2 O 3 content was between 10 and 12.5%, where the diffractograms indicated the presence of amorphous phase in heats #13, #14 and #15.The influence of the cooling rate of 21 °C/s on amorphous phase formation is not conclusive.

CONCLUSIONS
The main conclusions can be summarized as follows: 1. Solidification of BOF modified slag resulted in several predominant crystalline phases: Olivine solid solution, Melilite and RO phase, and solid precipitate of Merwinite.
2. The phases found in the XRD-Rietveld analyzes confirm the thermodynamic simulation for this series of experiments in a qualitative way; but in a quantitative way, the results showed erratic dispersions, with a lower percentage bias for the XRD analyzes.
3. Heats tested with basicity below 1.1 and Al 2 O 3 content below 5%, favor the formation of crystalline phases and absence of amorphous phase.With cooling rates up to 21 °C/s, for this group (Crystalline trials), there is no evidence of amorphous phase formation.

Theoretical Thermodynamic Slag Solidification
During the solidification process by thermodynamic simulation, represented schematically by Figure 5, Heat #1 (crystalline trials), Olivine solid solution has a predominance below 1300°C; Monticellite and Kirschsteinite are the predominant solution phases in this Olivine solution.Ca 2 SiO 4 and Glaucochroite are also present at lower temperatures, and at 1200°C, these phases are distributed around 40% for Monticellite, 28% for Kirschsteinite, 15% for Larnite and Glaucochroite in Olivine solid solution.
The solid solution of Melilite begins to form below 1250°C, and contains predominantly Akermanite and Gehlenite in solution, and at 1200°C an average proportion of 37% is estimated for Akermanite and 21% for Gehlenite; at this temperature, we reported the possibility of precipitation of Iron-Akermanite (Ca 2 FeSi 2 O 7 ) in the proportion of 22%.At the temperature of 1200°C the solid oxide solution, RO phase becomes stable with predominance of FeO (53%) and Fe 2 O 3 (32%).

XRD-Rietveld in Comparison with FactSage  Thermodynamic Simulation
It was considered as a criterion of comparison between the methods, that the majority phase within a solution by thermodynamic simulation would be the constituent comparable to the phase detected by XRD-Rietveld.
In Figures 8, 9, 10, and 11 the results obtained from the precipitated phases are compared by FactSage  thermodynamic simulation in T solidus and XRD-Rietveld diffractograms.The typical crystalline phases found in the thermodynamic simulation for this series of experiments are confirmed by XRD-Rietveld.These phases are: solid precipitate from Merwinite, Olivine solid solution containing Monticellite, solid solution of Melilite containing Akermanite and Gehlenite, solid RO phase solution containing Wüstite.
Andradite and Anortite solids precipitates were not detected XRD-Rietveld, and a small amount of Akermanite was detected in comparison with the thermodynamic simulation, and as a result, the simulation did not predict Gehlenite, unlike XRD-Rieteveld, which did not detected Wolastonite.Anortite, 5% and 7% (in Heats 2 and 15) and Andradite, 22% (Heat # 2) and Wollastonite solid precipitates, 3% (in Heat # 2), CaFe 2 O 4 Spinel, were predicted by thermodynamic simulation, but these precipitates were not detected in XRD-Rietveld.

Figure 4 .
Figure 4. (a) and (b) Construction details of the circular cooled mold, highlighting the inlet and outlet hoses; (c) Modified and totally solidified slag sample.

Table 3 .
Slag cooling conditions for each test