Crystallization Behavior and Growing Process of Rutile Crystals in Ti-Bearing Blast Furnace Slag

Materials

The slag used in this work was provided by the Panzhihua Iron and Steel Company. Table 1 shows chemical analysis results of the slag.

As shown in Table 1, the slag mainly consists of TiO₂, CaO, SiO₂, Al₂O₃ and some other oxides, such as MgO and Fe₂O₃, which were inevitably included in the slag, accounted for small proportion.

Experimental procedures

Elevated temperature experiments

All elevated temperature experiments were carried out isothermally in a vertical furnace that is controlled by the Shimaden temperature program (FP93). The overall absolute temperature accuracy of the experiment was ±3 K. The working thermocouple was calibrated against a standard thermocouple. The experimental materials were placed in a crucible and taken into the furnace. The furnace was heated from room temperature to the experimental temperatures and then held at the experimental temperatures for about 30 min to facilitate the melting of experimental materials. Oxygen gas was introduced into the melts at a certain rate to make the low chemical valence titanium components oxidized.

To reveal the crystallization and growing process of rutile crystals at elevated temperatures, it is necessary to take a portion of the reacted samples out of the liquid slag. Typically, a small portion of slag was taken out through quartz tube immediately and then water quenched at once. There may be some uncertainties during elevated temperature experiments, each experiment repeats for five times in order to make the experimental results more reliable. Figure 1 is the experimental apparatus of elevated temperature experiments.

Figure 1:

Experimental apparatus of elevated temperature experiments. Note: 1, quartz tube; 2, refractory bricks; 3, MoSi₂ heating elements; 4, crucible; 5, temperature controller; 6, thermocouple.

Analysis on nucleation mode and growth mechanism of rutile crystals

A portion of samples were taken out of the liquid slag and quenched at high temperatures in order to reveal the nucleation modes of rutile crystals. Figure 2 is the optical microscopic of quenched slag at 1,723 K. The white crystals in Figure 2 are rutile crystals.

Figure 2:

Optical microscopic of the quenched slag at 1,723 K.

Figure 2 implies the crystallization temperature of rutile crystals in this work was higher than 1,723 K and the nucleation time was less than 30 s, which demonstrated that the nucleation process of rutile crystals was heterogeneous nucleation process. Compared with heterogeneous nucleation process, homogenous nucleation process needs high degree of supercooling. Some research results indicated that the necessary supercooling of most metals is 20% of their melting points [4–6]; this may be helpful for analyzing the crystallization temperature of rutile crystals. According to the quaternary phase diagram of CaO-TiO₂-SiO₂-Al₂O₃ system [5, 6], the lowest temperature of liquidus line is 1,473 K; thus, the supercooling of rutile crystals during the homogenous process was around 295 K, and the corresponding nucleation temperature was around 1,178 K. The molten system was complex mixture system; thus, the crystallization temperature of rutile crystal was lower than 1,178 K. However, the crystallization temperature of rutile crystals in this work was higher than 1,723 K. Thus, we can confirm that the crystallization process was a heterogeneous nucleation process. According to the previous research results [3–6], there was a small part of high melting point material in the slag, such as Ti (N, C), served as the cores of rutile crystals, which made the rutile crystals could crystallize rapidly under appropriate degree of supercooling. This was proved in this work.

Figure 3 is the backscattered electron image of quenched slag at 1,708 and 1,658 K, and Table 2 is the corresponding EDS analysis results in Figure 3.

Figure 3:

Backscattered electron image of the quenched sample at (a) 1,708 K and (b) 1,658 K.

As shown in Figure 3 and Table 2, the titanium content is 10.086% in matrix phase (gray area in Figure 3); as a matter of fact, according to our experiment, there was about 3 wt% titanium components in the matrix phase after reaction process. Thus, the titanium component has not precipitated completely at 1,708 K; it would be continued to precipitate with the decrease in temperature. To reveal the lowest precipitate temperature in the system, we take samples at lower temperatures. The titanium content in the matrix phase in Figure 3(b) was 3.539%, which demonstrated that the crystallization process of titanium component was nearly complete.

The rutile crystals in Figure 3 were strip, which indicates that the coarsening process of rutile crystals was not completely at 1,658 K. At present, the theory about the crystal growing process is almost limited in equilibrium state; it is reported that the growing process of rutile was controlled by the diffusion of titanium components [4–6]. This process can be expressed by Lifshitz, Slyozov, Wagner (LSW) equation [4–6]; the relationship between average size of particles and growing time t is

where rˉ0 is the incipient average radius of rutile crystals; t is the growing time; rˉ is the average radius of rutile crystals at a certain moment; k is activation energy for diffusion process. Samples were taken out from the liquid slag and quenched at once at 1,683 K. The experimental results were subsequently fitted via linear method according to eq. (1). Figure 4 illustrates the relationship between r³ and t of rutile crystals.

Figure 4:

The relationship between r³ and t of rutile crystals at 1,683 K.

As shown in Figure 4, r³ and t have approximately linear relationship, which implies the LSW equation was suitable to describe the growing process of rutile crystals; thus, the precipitation and growing process of rutile crystals were corroborated as controlled by diffusion of TiO₂.

Crystallization and growing kinetics of rutile crystals under high temperature conditions

To reveal the growing process and morphology changing of rutile crystals during the experimental process, we take samples at elevated temperatures and observe the microstructure of the quenched slag. The optical microscopic of the quenched slag at 1,638 K with different temperature holding time is shown in Figure 5.

Figure 5:

Optical microscopic of the quenched slag at 1,638 K with different times: (a) 0 min, (b) 20 min, (c) 60 min and (d) 120 min.

As shown in Figure 5, when the temperature holding time was 0 min, there was no crystal in the molten which demonstrated that the crystals in raw slag was completely melt and the new crystals were not formed. When the time reached 20 min, there was already small feathery rutile crystals appeared, when the time reached 60 min, rutile crystals coarsened clearly. According to the theory of crystal growth [6, 7, 12], the crystal growth mode can be divided into three different types: continuous growth, lateral growth and growth from defects. When the temperature holding time was less than 20 min, the rutile crystals grew mainly via continuous growth, after 20 min, the main growth way was lateral growth and eventually forms a quadrilateral.

The liquid slag can be considered as ideal solution, when the rutile crystals and the liquid slag reached the equilibrium state under certain conditions (P, T, C₀), set C₀ as the saturated concentration of rutile under this condition. The chemical potential of rutile in liquid slag was the same as it in rutile crystals. It can be expressed by the following equation [5, 8]:

When the temperature quickly dropped from 1,738 K (experimental temperature) to 1,723 K, the concentration of rutile can be considered constant as C, the chemical potential of rutile in solution is

Due to C > C₀, the liquid slag was under the hyper-saturated state, and the different value can be represented as Δμ=RTlnCC0. ∆μ would be close to zero when it was in the insulation process. During this time, the crystallization quantity of rutile crystals increased gradually.

Assuming X can be expressed by the following equation [5, 8–11]:

where f(T, t) is the volume fraction of rutile at a certain moment at temperature T and f (T, equ) is the volume fraction of rutile at the equilibrium state. Figure 6 shows the curves of the relationship between titanium transform fraction (X) and heat temperature holding time (t).

Figure 6:

Relation between titanium transformed fraction (X) and temperature holding time (t).

As it can be seen from Figure 6, at the beginning of isothermal process, the transformed fraction was unequal to zero at different temperatures, which indicates that rutile crystals begin to crystallization during the rapid cooling process. When the holding time was 60 min, the transformed fraction of titanium components was almost invariant. In another word, the system was close to equilibrium state at the moment. The crystallization process of rutile under isothermal conditions can be expressed by Avrami equation, and the relationship between X and t is [12–16]

where k is the kinetic constant, n is Avrami factor and the fitting results are shown in Table 3.

Avrami factor reflects the nucleation rate, when the nucleation process was controlled by diffusion, the value of n should be 1.5; in this study, n = 1.16, that is because the fine particles combined with each other during the growth process.

Figure 7 shows the relation curves of r3‾ and t; as can be seen in it, the crystal grew rapidly in initial stage of isothermal and 60 min later, the volume growth rate was gradually slowed down which implied the rutile crystal was of self-growth in the initial stage and then coarsened in the anaphase.

Figure 7:

The variation of r³/X with time.

According to references [12–16], the growth of rutile crystal under nonequilibrium state can be expressed by the following equation:

where u is the linear growth rate, which was attributed to super saturation; X is the transform fraction of rutile; k(X)=3DδVBC(∞,X)2kBT; D is the diffusion coefficient of rutile molecular in liquid slag; δ is effective interfacial energy; V_B is the volume of rutile molecular; k_B is the Boltzmann constant; T is the system temperature; C(∞,X) is the equilibrium concentration of rutile molecular at quasiequilibrium. According to the previous analysis, rutile crystals begin to precipitate during the rapid cooling process. The nucleation rate could be approximately regarded to be zero, and then eq. (6) can be written as

The volume fraction was very small at experimental temperatures; thus, the concentration corresponding to X could be

where ϕ is the volume fraction of rutile under equilibrium state; C(∞,1) is the equilibrium concentration of rutile molecular. Substituting eq. (8) into eq. (7):

where k1=k01+ϕC(∞,1); k2=k0ϕC(∞,1); k0=3DσVBC(∞,1)2kBT. The relationships between ddtr3‾X and 1X were linear ships that were shown in Figure 8. The results were consistent with the rule of eq. (9), as long as the relation between X and t; the relation between average particle size and t can be given by eq. (9).

Figure 8:

The relationship between r³/χ and 1/χ.

Investigation on rutile crystals growing process

Effect of melting temperature on crystallization behavior of rutile

Figure 9 shows the backscattered electron images of slag under different melting temperatures.

Figure 9:

SEM images of the slowly cooled slag at different melting temperatures: (a) 1,703 K; (b) 1,713 K; (c) 1,723 K and (d) 1,733 K.

As shown in Figure 9, the obtained rutile crystals at 1,703 K were relatively small, most of the crystals connected with one another, which implied the slag was under high viscosity condition at 1,703 K. The average grain size of rutile crystals increased to round 80 μm as temperature rose.

Figure 10 summarizes the variation of volume fraction and equivalent diameter of rutile versus melting temperatures.

Figure 10:

Precipitation amount and average grain size of rutile versus different melting temperatures.

As shown in Figure 10, the precipitation amount of titanium components significantly increased as melting temperatures rose. As elaborated in this work, the precipitation and growing process of rutile crystals were controlled by diffusion of TiO₂, and the viscosity of liquid slag had great influence on the diffusion of TiO₂; for most of the melts, the viscosity is in direct proportion to temperature. As such, the viscosity of liquid slag would rapidly fallen down with temperature rose, which was quite benefit to the diffusion of TiO₂.

Effect of SiO₂ on the crystallization behavior of rutile

The silicon dioxide serves as an agent in this work; it can combine with CaO, Al₂O₃ and MgO in order to make the titanium components transform into rutile phase. This can be represented as follows [6, 9]:

(10)CaO+2SiO2+MgO=CaMgSi2O6

(11)CaO+2SiO2+Al2O3=CaMgAl2Si2O8

If SiO₂ was not enough, the titanium components in the molten slag would react with aluminum and magnesium oxides in the system which can be described by eqs (3) and (4):

(12)(Al2O3)+(TiO2)=(Al2TiO5)

(13)12(MgO)+(TiO2)=12(MgTi2O5)

Figure 11 shows the precipitate quantity and average grain size of rutile crystals with different addition amounts of SiO₂.

Figure 11:

Precipitate quantity and average grain size of rutile with different addition amounts of SiO₂.

Some research results on CaO-SiO₂-TiO₂-Al₂O₃ system indicated that titanium dioxide is an amphoteric oxide [6, 9]; it presented reticular structure in the liquid slag and formed a complex of the ion cluster, which made the viscosity of the slag to increase. There were two forms of structure for titanium components in the melt: one was four coordinated intermediate tetrahedron (Ti(4)), which was acidity; the other was six coordinated intermediate octahedral (Ti(6)), which was alkaline [6, 9]. When SiO₂ increased, Ti(6) would increase, owing to low basicity of the slag (0.4–0.5). The complicated structure of Ti(6) was the main form of TiO₂ in the melting slag. It was large in size and difficult to move, which was not beneficial to the coarsening of rutile. However, SiO₂ is a high melting point compound; if extensive SiO₂ was added into the slag, the viscosity of the system would increase, which was not conducive to the growth of rutile. The optimum dosage was 40%.

Effect of cooling rate on crystallization behavior of rutile and the matrix phase

The experimental results implied that cooling rate had great influence on the growing process of crystals; Figure 12 illustrates the precipitate amount and grain size of rutile with different cooling rates.

Figure 12:

Precipitate amount and average grain size of rutile with different cooling rates.

When the cooling rate was fast, the molten viscosity of the molten were also quickly increased which was not beneficial to mass transfer of titanium. That was the reason why the rutile crystals were smaller under the fast cooling rate than that was under the slow cooling rate. Figure 13 shows the XRD pattern of the cooled slag with the cooling rate of 0.5 and 4 K/min.

Figure 13:

XRD pattern of the modified slag: (a) 0.5 K/min and (b) 4 K/min.

As shown in Figure 13, when the cooling rate was 0.5 K/min, there were two phases in cooled slag (rutile and Ca(MgAl)(SiAl)₂O₆), while the cooling rate was 4 K/min, there was only rutile phase in the slag, and the matrix phase was amorphous. That was because the viscosity of the liquid slag would be changed rapidly under fast cooling rate conditions, which made the matrix phase do not have enough time to crystallize, it solidified in the form of glass in the slag. The matrix phase of the slag was silicate; the existing form of Si–O anion group was tetrahedron. When the matrix phase was amorphous, the Si and O component presented as unstable net structure, while the matrix phase was pyroxene, the Si and O component presented as chain structure, which was more stable than the net structure. According to references [6–8], the Ti–O component would combine with the Si–O tetrahedron which can strengthen the net structure of the amorphous phase. When it was under the fast cooling rate, the crystallization amount of rutile was little; thus, there was more Ti component in the matrix phase, and the Ti–O combined with Si–O tetrahedron and formed the amorphous phase. On the contrary, when the cooling rate was slow, the rutile crystallization amount was large; the Ti–O cannot combine with Si–O tetrahedron, thus pyroxene was formed.

Effect of crystal seed addition on the crystallization behavior of rutile

As proved in this work, the nucleation mode was heterogeneous. To enhance the precipitation process of rutile crystals, nanometer TiO₂ powder was added into the liquid slag as crystal seed. Figure 14 shows the precipitate amount and the average grain size of rutile versus different addition amount of crystal seed.

Figure 14:

Precipitate amount and average grain size of rutile versus addition amount of crystal seed.

As shown in Figure 14, the crystal seed addition had significant effect on the grain size of rutile crystals. The nucleation energy can be expressed by the following equation [17]:

(14)ΔGheterogeneous=ΔGhomogeneous×2−3cosθ+cos3θ4

where θ changes between 0 and π, thus 0≤2−3cosθ+cos3θ4≤1; according to eq. (14), the nucleation energy of heterogeneous nucleation was less than homogeneous nucleation, thus heterogeneous nucleation of rutile was much easier than homogeneous nucleation. The relationship between nucleation rate and nucleation energy can be expressed by the following equation [17]:

(15)N=KV×eΔGKRT×eΔGART

where K_V is proportionality constant, ΔG_K is nucleation energy, ΔG_A is the activation energy, R is the Clapeyron constant and T is the absolute temperature. The addition of crystal seed made ΔG_K decreased, thus the N value of rutile in the formula increased significantly. That was the reason for the addition of crystal seed was benefit to the nucleation of rutile in the same supercooling degree. However, if excessive crystal seeds were used during the experimental process, the nucleation density in the slag would too large, which was not conducive to the growth of rutile crystals.

Effect of oxygen flow on the crystallization behavior of rutile

The oxygen gas served as the oxidizing agent in the slag. Figure 15 shows the precipitate amount and average grain size of rutile crystals versus oxygen flow. The experimental results in Figure 15 indicated that the oxygen flow had little influence on the precipitate quantity of rutile crystals; however, the grain size of the rutile crystals increased with oxygen flow. This is because more oxygen flow can promote the oxidation process of Ti³⁺ and Ti²⁺, which made the temperature of the slag raised and the viscosity decreased, and this is conductive to the growth of rutile crystals.

Figure 15:

Precipitate amount and average grain size of rutile versus oxygen flow.

Distribution of rutile crystals in elevated temperature molten slag

The experimental results show that the rutile crystals distribute unevenly in the melts, the crystals in the bottom of the molten were much more than that in the top position. This is because rutile crystals in the melt were in motion state due to the action of gravity. The distribute situation of rutile in the cooled slag is shown in Figure 16.

Figure 16:

Diagrammatic drawing of the distributed situation of rutile in the slag.

The movement of rutile in liquid slag can be considered as the movement of solid particles in liquid system; according to reference [18], it can be described by the following equation:

where u_t is the settlement speed, m/s; C_d is damping parameter; γ_s is heaviness per unit volume of particles, N/m³; γ is heaviness per unit volume of melt, N/m³; d is particle diameter.

The damping parameter of settlement process of rutile in the melt can be calculated by the following equation [18]:

While according to the Stokes formula, the u_t in formula (9) can be expressed by the following equation:

where μ is the melt dynamic viscosity, MPa · s; ν is the melt kinematic viscosity, m²/s; ρ_s and ρ are the density of rutile crystal and melt, kg/m³. The relation between μ and ν is [18]

Equation (18) is appropriate for round particles, the rutile crystal was virgulate and the equivalent diameter can be calculated as

where D_s is the equivalent diameter; D_b and D_c are diameters of principal axes which are perpendicular to D_s, m.

According to the experimental results of this work, D_a is about 60 μm, D_b and D_c were about 4 μm; thus, the value of D_s is

When the temperature is 1,733 K, the values of ν and u_t are

The settlement speed of the rutile crystal was about 1.536 m/s, if the depth of liquid slag was 0.457 m, the corresponding settlement time is

This indicated that the settlement time of rutile from the top of the liquid level to the bottom of the crucible was 0.30 s.