Research on Magnesium Reduction Slag for Dephosphorization of Low-Silicon Hot Metal in Steelmaking Process

Abstract

The melting temperature and viscosity of magnesium reduction slag were calculated by using Factsage thermodynamic software. The composition range of the magnesium-slag-based dephosphorizing agent was analyzed by drawing a multiphase diagram of the slag system. The Box–Behnken high-temperature dephosphorization experiment was designed to study the effect of different composition of magnesium-slag-based dephosphorizers on the dephosphorization rate of the steelmaking process. The results show that magnesium slag can be used as a slag-forming agent for smelting low-silicon hot metal to promote slagging, and the effect of each factor on the phosphorus removal rate is ranked, and the results are ω(Fe₂O₃) > basicity > ω(Al₂O₃): ω(Al₂O₃) has no significant effect on the rate of phosphorus removal. When the basicity was 2.8, ω(Fe₂O₃) was 25.94%, ω(Al₂O₃) was 6.73%, and ω(MgO) was 6%, the dephosphorization rate reached a maximum of 96.7%, and the error was experimentally verified to be 2.6% from the predicted value, indicating that the model can be optimized to determine the best magnesium-slag-based dephosphorization agent and has a good prediction of dephosphorization effect.

1. Introduction

Magnesium reduction slag (referred to as magnesium slag) is a by-product of the smelting of metallic magnesium by the Pidgeon method, with an annual output of more than 5 million tons [1]. It contains a high content of CaO and SiO₂, and some Fe₂O₃, Al₂O₃ and MgO, which are secondary resources that can be used.

At present, the resource utilization of magnesium slag is mainly concentrated in the preparation of building materials [2,3,4,5,6], industrial desulfurization [7,8,9], agricultural fertilizers [10,11] and mine filling [12,13]. In terms of building materials, magnesium slag is used to prepare concrete: when the content of magnesium slag is 30%, the durability and long-term performance of concrete are improved [6]. Magnesium slag is used as a hot metal desulfurizer, and the desulfurization efficiency can reach more than 80% in industrial desulfurization [8]. In terms of agricultural fertilizers, magnesium slag is used as a raw material to prepare a silicon–potassium fertilizer, which has achieved a good application effect on calcareous cinnamon soil in northern China [11]. The modified magnesium slag is prepared into a cementitious material, which can be used to replace cement in mine filling [12]. However, the above utilization methods all have problems such as low consumption of magnesium slag, high-processing costs and a low comprehensive utilization rate. Therefore, it is very meaningful to study the resource utilization of magnesium slag.

The composition structure of magnesium slag is similar to that of steel-making slag. Steelmaking lime consumption can be reduced by CaO in magnesium slag and furnace-lining erosion can be reduced by MgO. SiO₂, Al₂O₃ and Fe₂O₃ are conducive to the formation of a multi-component composite slag system, and they reduce the melting point of the slag system. The oxidizing property of slag can also be improved by Fe₂O₃, which is beneficial to dephosphorization in the early stage of smelting. Especially for low-silicon molten iron or semi-steel after vanadium extraction and dephosphorization operations, the low acidic fraction of slag during smelting makes it difficult for lime to be melted and the dephosphorization efficiency is deteriorated [14]. At present, by adding slag-forming materials such as river sand, silica, ferrosilicon, glass and bauxite to steelmaking enterprises, the structure of the steel-dephosphorization slag system is improved, lime melting is promoted, and the efficiency of dephosphorization is increased [15,16,17,18].

In this paper, the composition range of the magnesium-slag-based dephosphorization agent was determined by studying the composition and phase structure of magnesium slag, combined with Factsage thermodynamic software to draw a multi-component slag system phase diagram. The Box–Behnken (part of responsive surface design) high-temperature dephosphorization test was designed by Design-Expert software (for design of experiments (DOE)) to analyze the influence of magnesium slag components on the dephosphorization efficiency of low-silicon molten iron, and to provide a theoretical basis and technical support for magnesium slag as a steelmaking slag material.

2. Physical Properties of Magnesium Slag

2.1. Phase Analysis and Composition

The magnesium slag used in this paper were obtained from enterprise A and enterprise B, which are by-products of the Pidgeon process (silicon-thermal reduction process) magnesium metal. The main components are shown in

Table 1. Composition of magnesium slag (mass fraction %).

No.	CaO	SiO2	MgO	Al2O3	Fe2O3
A	58.7	33.5	2.5	1	4.3
B	58.5	31.1	5.7	1	3.7

.

The physical phase analysis of magnesium slag A and magnesium slag B were carried out by XRD (X-ray diffraction), and the experimental results were analyzed by HighScore Plus software. The physical analysis of magnesium slag A and B were obtained as shown in Figure 1, showing the phase analysis of magnesium slag A and B. The main forms of magnesium slag were Ca₂SiO₄, MgO, Al₂O₃ and Fe₂O₃, where the crystalline phases of Ca₂SiO₄ wer β-Ca₂SiO₄ and γ-Ca₂SiO₄, respectively. The phase difference of magnesium slag A and B were mainly reflected in the difference of β-Ca₂SiO₄ content between them. The increase in the phase β-Ca₂SiO₄ can make magnesium slag have a higher hydration activity and porosity [19]. In order to ensure the good physical and chemical properties of the magnesium slag used in the experiment, magnesium slag A was selected as the experimental raw material.

2.2. Melting Temperature and Viscosity of Magnesium Slag

The magnesium-slag-melting temperature was measured by the CQKJ-II-type slag melting temperature characteristic tester. After the equipment reached the upper temperature limit of 1600 °C, the magnesium slag did not show obvious melting characteristics. Factsage 7.1 thermodynamic software was used to conduct the thermodynamic analysis of magnesium slag, and it was calculated that the initial melting temperature of magnesium slag was 1259 °C, while the complete melting temperature was 2008 °C. Figure 2 shows the change of phase of magnesium slag with the temperature, and Figure 3 shows the variation in viscosity and temperature of magnesium slag. When the temperature was raised from 1000 °C, the mass fraction of Ca₃MgSi₂O₈ and melilite in magnesium slag was gradually reduced, and the mass fraction of dicalcium silicate (Ca₂SiO₄) was slowly increased before the liquid phase was generated. When the temperature reached 1259 °C, the liquid phase in the system began to be generated. At this time, as the temperature increased, the mass fraction of dicalcium silicate (Ca₂SiO₄) was gradually reduced, and the slag viscosity also showed a downward trend. When the temperature was raised to 1350~1450 °C, the viscosity of magnesium slag was still at a high level.

The pseudo-pentad phase diagram of CaO–SiO₂–Fe₂O₃–MgO–Al₂O₃ is shown in Figure 4; point A is the composition point of magnesium slag A. When the ω(MgO) and ω(Al₂O₃) in magnesium slag were kept unchanged, the basicity was adjusted by 1.5–3.5, and ω(Fe₂O₃) was adjusted to 10–30%, the smelting conditions were still not satisfied. Therefore, the mass fractions of ω(MgO) and ω(Al₂O₃) needed to be adjusted simultaneously to improve the slag system structure to fulfill the smelting requirements.

3. Effect of Different Components on the Liquid-Phase Region

Figure 5 shows the ternary phase diagram of CaO–SiO₂–Fe₂O₃ drawn by Factsage software. When ω(Fe₂O₃) was between 10% and 30%, the phase changed from CaSiO₃ to 3CaO–2SiO₂ as the basicity was increased from 1.0 to 1.5, and when the basicity was raised to 3.0, the main phase was Ca₂SiO₄. With the increasing basicity, the complete melting temperature of the slag system was gradually raised, and the area of the liquid-phase zone was gradually reduced. The structure of the slag system needed to be improved by Al₂O₃ and MgO, and the melting point was also lowered.

3.1. Effect of MgO on the Liquid Phase Region

Figure 6 shows the effect of MgO on the liquid-phase zone of the CaO–SiO₂–Fe₂O₃ slag system at 1450 °C. When the MgO content increased from 1% to 3%, the area of the liquid phase did not change significantly, and when it continued to increase to 9%, the area of the liquid phase was rapidly reduced. The excessive MgO content resulted in the presence of a large number of unmelted MgO particles in the slag, which affected the melting temperature and viscosity of the slag and was not conducive to the melting of the slag [20]. When the MgO content was 6%, the slag system still had high oxidizing properties, and the liquidus line tended to the high SiO₂ region, and the slag with low basicity in the early stage of smelting was accelerated to melt. Additionally, the MgO content in slag should be kept in a reasonable range to avoid serious erosion of the furnace lining due to the low MgO content, in which the life of the converter will be affected. Therefore, the MgO content in the magnesium-slag-based dephosphorization agent was set to 6% in this study.

3.2. Effect of Al₂O₃ on the Liquid-Phase Region

Figure 7 shows the effect of Al₂O₃ on the liquid-phase zone of the CaO–SiO₂–Fe₂O₃ slag system at 1450 °C. With the increase in Al₂O₃ content, the liquid-phase line gradually tended to the high CaO region, and the dense Ca₂SiO₄ outer layer in the slag was loosened by Al₂O₃, which was conducive to the diffusion of Fe₂O₃ and SiO₂ into the lime [21], and the melting of the slag was accelerated. When the Al₂O₃ content was in the range of 1–12%, the slag system had a wide range of basicity and a high oxidizability, which was consistent with the composition of slag in the pre-phosphate removal stage.

The pseudo-quinary phase diagram of CaO–SiO₂–Fe₂O₃–Al₂O₃–6%MgO is shown in Figure 8. The increase in Al₂O₃ content led to a reduction in the liquid-phase region, but it was conducive to the melting of high-basicity slag, and the minimum content of Fe₂O₃ in the liquid-phase region was still greater than 30%. This met the dephosphorization conditions for steelmaking. At this time, the basicity, melting temperature and oxidizability of the slag system could be satisfied for the dephosphorization conditions of steelmaking.

Based on the above research, the composition and content of the magnesium-slag-based dephosphorization agent studied in this paper were determined as shown in

Table 2. Composition range of magnesium-slag-based dephosphorization agent (mass fraction %).

CaO	SiO2	Fe2O3	Al2O3	MgO
30~64	10~33	10~30	2~12	6

.

4. High-Temperature Experiment of Magnesium Slag for Dephosphorization in Steelmaking Process

4.1. Experimental Scheme

The experimental design was carried out using the Box–Behnken module of the Design-Expert software, as shown in

Table 3. Box–Behnken analysis factors and levels.

Factor	Level
	−1	0	1
A (Basicity)	1.5	2.5	3.5
B ω(Fe2O3)	10	20	30
C ω(Al2O3)	2	7	12

, and three factors (basicity, ω(Fe₂O₃%) and ω(Al₂O₃%)) were selected as independent variables to design a three-factor, three-level experiment with the response value of the dephosphorization rate and response curves used to determine the optimal dephosphorizer composition.

4.2. Experimental Materials and Methods

The magnesium slag used in the experiments was obtained from enterprise A. The low-silicon hot metal used was taken from the semi-steel of a steel company after vanadium extraction, and its composition is shown in

Table 4. Composition of semi-steel (mass fraction%).

C	Si	Mn	P	S
3.966	0.004	0.011	0.123	0.062

. The other oxides used in the preparation of the magnesium slag were all analytical pure reagents.

As is shown in Table 4, the silicon mass fraction of the low-silicon hot metal used was 0.004%, and the SiO₂ required for steelmaking slag coul not be produced by iron oxidation; rather, it came from the magnesium slag.

(1) 

Preparation of dephosphorization agent:

(1) The components were prepared in accordance with their respective ratios of the phosphorus removal agent required for the experiment and were fully mixed, weighing 100g into the corundum crucible;

(2) The dephosphorization agent was pre-melted in a high-temperature box furnace and the temperature was raised to 1400 °C, and then kept for 30 min. After the sample was completely melted, the heating was stopped, and the temperature was naturally cooled;

(3) The cooled dephosphorized pre-melted slag was crushed to 200 mesh by the crusher.

(2) 

Dephosphorization experiment:

(1) A total of 200 g of semi-steel and 20 g of dephosphorized slag were weighed, mixed uniformly and placed in a corundum crucible, and were then put into a high-temperature box furnace to start heating up;

(2) The semi-steel melting point was calculated to be 1140 °C; therefore, the experimental temperature was set to 1450 °C, which ensured the complete melting of semi-steel and the maintenance of a continuous temperature for six hours. It also ensured the complete reaction of the slag-steel interface;

(3) After the sample was cooled to room temperature, the slag and iron were separated, and the phosphorus content in the steel was analyzed by ICP.

4.3. Experimental Results and Analysis

4.3.1. Analysis of Model-Fitting Degree

The experimental data in

Table 5. Response surface experimental design and results.

No.	A. (Basicity)	B. ω(Fe2O3)	C. ω(Al2O3)	D. Dephosphorization Rate (%)
1	1.5	20	12	69.91
2	1.5	20	2	66.73
3	1.5	30	7	83.24
4	1.5	10	7	34.21
5	2.5	30	2	90.13
6	2.5	30	12	89.72
7	2.5	10	2	58.91
8	2.5	10	12	55.54
9	2.5	20	7	91.27
10	2.5	20	7	90.13
11	2.5	20	7	89.73
12	2.5	20	7	91.69
13	2.5	20	7	92.85
14	3.5	20	2	88.34
15	3.5	30	7	89.79
16	3.5	10	7	58.01
17	3.5	20	12	86.04

Note: ω(MgO%) = 6%.

were fitted by Design-Expert with a quadratic multiple regression, which resulted in a quadratic multiple regression model of D (dephosphorization rate) versus A (Basicity), B (ω(Fe₂O₃)) and C (ω(Al₂O₃)) represented by:

$$\mathrm{D}=91.24+9.19\mathrm{A}+18.28\mathrm{B}\hspace{0.17em}-\hspace{0.17em}1.04\mathrm{C}\hspace{0.17em}-\hspace{0.17em}4.31\mathrm{AB}\hspace{0.17em}-\hspace{0.17em}0.7275\mathrm{AC}+0.74\mathrm{BC}\hspace{0.17em}-\hspace{0.17em}10.7{\mathrm{A}}^2\hspace{0.17em}-\hspace{0.17em}14.23{\mathrm{AB}}^2\hspace{0.17em}-\hspace{0.17em}3.43{\mathrm{C}}^2$$

(1)

The model determination coefficient R² was 0.9866, and the fitting degree was relatively high. The experimental variance analysis results are shown in Table 5.

As shown in

Table 6. Box–Behnken test ANOVA results.

Source of Variance	Sum of Squares	Degrees of Freedom	Mean Square	F-Value	p-Value	Significance
Model	4698.37	9	522.04	57.48	<0.0001	**
A (Basicity)	579.53	1	579.53	63.81	<0.0001	**
B ω(Fe2O3)	2672.17	1	2672.17	294.20	<0.0001	**
C ω(Al2O3)	1.05	1	1.05	0.1157	0.7437
AB	74.39	1	74.39	8.19	0.0243	*
AC	7.51	1	7.51	0.8266	0.3935
BC	2.19	1	2.19	0.2412	0.6384
A2	406.67	1	406.67	44.77	0.0003	*
B2	826.18	1	826.18	90.96	<0.0001	**
C2	27.71	1	27.71	3.05	0.1242
Residual	63.58	7	9.08
Lack-of-fit	39.22	3	13.07	2.15	0.2370
Pure error term	24.35	4	6.09
Total dispersion	4761.95	16

Note: *. Significant difference (p < 0.05); **. Very significant difference (p < 0.0001).

, the model was very significant with p < 0.0001, and the lack-of-fit model was greater than 0.05, indicating that the quadratic polynomial regression equation was well fitted.

According to the various p values, the ranking results of the influence of various factors on the dephosphorization rate were ω(Fe₂O₃) > basicity > ω(Al₂O₃), and the effect of ω(Al₂O₃) on the dephosphorization efficiency was not significant. In the quadratic term, A² and B² had significant effects on the dephosphorization rate, while C² had no significant effect on it, indicating that the relationship between the experimental factors and the response values were not a simple linear relationship. The interaction between basicity and ω(Fe₂O₃) had the largest F value compared with the other two, and p was less than 0.05, indicating that its interaction had a more significant effect on the dephosphorization rate. The p values of AC and BC were greater than 0.05, but the F value of AC was larger than that of BC, so the effect of AC interaction on the dephosphorization rate was greater than that of BC.

4.3.2. Interaction Analysis among Various Factors

The normal probability distribution of the residual dephosphorization rate is shown in Figure 9, and the comparison between the measured value and the predicted value is shown in Figure 10. The normal probability-distribution points of the residuals were located on or immediately on both sides of the straight line: the measured and predicted values were in good agreement, and the correlation between them was high. It showed that the selected model could be used to guide the composition optimization of the magnesium slag dephosphorization agent.

Figure 11 shows the contour map and 3D response surface of the interaction between the two factors, and the relationship between the factors and the response value is visualized by the 3D coordinate system. The greater the degree of distortion of the response surface, the more significant the corresponding factors had on the response value, and a darker color in the area of the figure indicates that the result is more significant and the dephosphorization rate is higher.

Al₂O₃ did not directly participate in the dephosphorization reaction, but it reacted with high-melting-point compounds to reduce the melting point of the slag, improve the fluidity of the slag, and promote dephosphorization. However, the high content of Al₂O₃ causes the basicity and ω(Fe₂O₃) to be reduced, and the phosphorus content in Ca₂SiO₄ will also be reduced [22], resulting in a lower rate of phosphorus removal. Therefore, the interaction between basicity and ω(Al₂O₃), ω(Fe₂O₃) and ω(Al₂O₃) had no significant effect on the rate of phosphorus removal, so the distortion of the response surfaces of (c) and (e) were low. It could also be seen from the contour plots of (d) and (f) that the dephosphorization rate was less affected by the change in Al₂O₃ content.

4.3.3. Optimization Results

Based on the influence of the interaction among the factors on the dephosphorization rate, it was observed that the rate of phosphorus removal increased and then decreased under the influence of the interaction among the factors. Therefore, there was an optimal ratio between basicity, ω(Al₂O₃) and ω(Fe₂O₃), which maximizes the dephosphorization rate. The optimal ratios among the three were obtained by simulations conducted by Design-Expert as the basicity was 2.8, ω(Fe₂O₃) was 25.94% and ω(Al₂O₃) was 6.73%, at which the maximum phosphorus removal rate of 96.7% was achieved.

In order to test the accuracy of the response surface model, three groups of experimental verifications were carried out using the optimized magnesium-slag-based dephosphorization agent, and the results are shown in

Table 7. Optimization experimental results.

No.	A. (Basicity)	B. ω(Fe2O3)	C. ω(Al2O3)	D. Dephosphorization Rate (%)
1	2.8	25.94	6.73	94.1
2	2.8	25.94	6.73	93.5
3	2.8	25.94	6.73	94.7

.

As can be seen from Table 7, in the optimized experimental results, the dephosphorization rates were 94.1%, 93.5% and 94.7%. Additionally, the experimental repeatability was good, and the error between the results and the predicted results was 2.6%. This model was shown to effectively optimize and determine the optimal composition of magnesium-slag-based dephosphorization agents.

5. Conclusions

(1) Magnesium slag can be used as a slag-making material for smelting low-silicon hot metal in the steelmaking process, as well as reducing the consumption of steelmaking lime. The slag structure was improved by the SiO₂, Al₂O₃, Fe₂O₃, etc. brought in by it, and the efficiency of phosphorus removal in the pre-smelting stage was improved.

(2) The multi-component slag system phase diagram was drawn by Factsage thermodynamic software, and a reasonable composition range of the magnesium-slag-based dephosphorization agent was designed: CaO is 30~64%; SiO₂ is 10~33%; Fe₂O₃ is 10~30%; Al₂O₃ is 2~12%; and MgO is 6%.

(3) The Box–Behnken response surface analysis showed that the effects of the model factors on the phosphorus removal rate were in the following order: ω(Fe₂O₃) > basicity > ω(Al₂O₃). Additionally, the effect of ω(Al₂O₃) on the rate of phosphorus removal was not significant, while the interaction between basicity and ω(Fe₂O₃) had a significant effect on the rate of phosphorus removal.

(4) When the basicity was 2.8, ω(Fe₂O₃) was 25.94%, ω(Al₂O₃) was 6.73% and ω(MgO) was 6%, the maximum phosphorus removal rate was 96.7%, and the error of the predicted value was experimentally verified to be 2.6%, indicating that the model can be optimized to determine the best magnesium-slag-based dephosphorization agent, and that it has a good prediction of dephosphorization effect.