Effects of MoO₃ + C on Crystallization and Radiative Heat Transfer of CaO–SiO₂–B₂O₃-Based Glassy Fluoride-Free Mold Fluxes

Abstract

The fluorine in traditional mold fluxes could be harmful in steel-plant environments. Accordingly, fluoride-free mold fluxes have received great attention in recent years. In this work, a method to adjust the crystallization and radiative heat transfer of fluoride-free mold fluxes is proposed. MoO₃ and C mixtures (mass ratio: 4:1) were added into CaO–SiO₂–B₂O₃-based mold fluxes and produced MoB, Mo₂CB and Mo₂C foreign particles. The influences of foreign particles on the radiation of glassy CaO–SiO₂–B₂O₃-based mold fluxes were investigated by measuring the transmissivity of a glassy disk for light in the wavenumber range of 300 to 2500 nm. It was found that transmissivity in all wavenumber ranges were reduced and extinction coefficients were enhanced by the scattering of foreign particles. The effect of foreign particles on crystallization (devitrification) of bulk glassy CaO–SiO₂–B₂O₃-based mold fluxes was also investigated. The crystallization mechanism of glassy mold fluxes disks is mainly surface crystallization. The introduction of foreign particles induced heterogeneous nucleation and the crystallization mechanism of the bulk sample with MoO₃% = 2% changed into bulk crystallization.

1. Introduction

Mold fluxes are fed from the top of molten steel and then infiltrate into the gap between mold and steel strands during continuous casting [1]. Multiple layers, including glassy, crystalline and liquid layers, could be found between the mold and steel shell. One of the most important functions for mold fluxes is to provide optimal horizontal heat transfer. Multiple flux layers could be employed to tailor the horizontal heat transfer. For some steel grades, e.g., medium carbon steel, mild cooling should be maintained to avoid longitudinal surface cracks on the slabs [2]. It is well accepted that increasing crystallinity in the solid layer is the most effective way to increase the effective heat resistance between the mold and the steel shell [3,4]. There are two mechanisms by which the crystallization increases the heat resistance across the mold fluxes’ layers [5]: (1) the crystallization of mold fluxes would increase the surface roughness of solid layers [6] and thereby promote the formation of an air gap between the mold and solid layer, which increase the interfacial thermal resistance [7]; (2) crystallizations would lower the transmissivity of flux layer and reduce the radiative heat transfer across the layer [8,9,10]. Therefore, the overall heat resistances are enhanced. It was reported that high-basicity mold fluxes could have stronger crystallization ability and be suitable for the casting of medium carbon steel [11].

The fluorine in traditional mold fluxes could be harmful to steel-plant environments and could even be leached into groundwater [12]. Accordingly, fluoride-free mold fluxes have received great attention in recent years. CaO–SiO₂–TiO₂-based mold fluxes [13,14,15] are considered to be potential fluoride mold fluxes to replace traditional mold fluxes, with CaTiO₃ and CaTiSiO₅ being substitutes for cuspidine in mold fluxes. However, the lubrication functions of CaO–SiO₂–TiO₂-based mold fluxes still need to be improved [16]. Recently, CaO–SiO₂–B₂O₃-based mold fluxes were proposed by many researchers to be the most promising fluoride-free mold fluxes [17,18,19,20]. Industrial trials have been successfully made on continuous castings of steel billets and slabs [18,19]. Boron oxide could improve the lubrication function of mold fluxes. However, it is reported that boron oxide in fluoride-free mold fluxes would impair the crystallization ability of mold fluxes [21], which makes this kind of fluoride-free mold fluxes unsuitable for casting some crack-sensitive steel grades.

The introductions of foreign particles into glass to modify the crystallization and optical properties of glass were investigated by many researchers [22,23,24]. It is a well-accepted practice in glass ceramics technology to introduce some nucleating agents to provoke internal crystallization [22]. These nucleating agents could be metallic particles or some oxides, e.g., TiO₂, ZrO₂, etc. A new technology was proposed by researchers at Pohang University of Science and Technology to reduce the horizontal heat transfer across the mold fluxes by introducing metallic particles less than a few micrometers in glassy mold fluxes [23,24]. They explained the scattering of metallic particles in glass using Mie theory.

Inspired by these studies, foreign particles were introduced into fluoride-free mold fluxes by in situ reduction of MoO₃ with carbon in the present study. Several kinds of particles, e.g., MoB, formed during the melting of mold fluxes. These particles could scatter light and reduce radiative heat transfer. On the other hand, these particles could also provide potential sites for heterogeneous nucleation and promote the crystallization of fluoride-free mold fluxes. The promoted crystallization and suppressed radiative heat transfer could be beneficial to the application of fluoride-free mold fluxes in the casting some crack-sensitive steel grades.

To validate this idea, the transmissivity of mold fluxes disks with different MoO₃ + C contents were measured and extinction coefficients were calculated in the present work. In addition, the crystallization of mold fluxes with different MoO₃ + C contents were investigated by differential scanning calorimetry (DSC), scanning electron microscopy equipped with energy dispersive spectral (SEM-EDS) and X-ray diffraction (XRD).

2. Experimental Procedure

2.1. Sample Preparation

Reagent powders of CaCO₃, Al₂O₃, H₃BO₃, MoO₃, Na₂CO₃, SiO₂ and C were employed to prepare mold fluxes samples. The mass ratio of MoO₃ to C was set to be 4, which is the theoretical MoO₃/C ratio for following reaction: MoO₃ + 3C = Mo + 3CO(g). CaCO₃ powders were heated at 1100 °C for 8 h in a muffle furnace to decompose into CaO. Al₂O₃, Na2CO₃ and SiO₂ powders were heated at 300 °C for 4 h to remove moisture. The reagent powders were mixed well in an agate mortar and put in a graphite crucible. The melting of mold fluxes was performed in a vertical tube furnace with MoSi₂ heating elements. To protect the oxidation of crucibles and samples, the furnace tube was purged by argon (purity > 99.999%). After melting, fluxes in crucibles were poured and quenched in an iron plate. The glassy samples were machined into a thin disk with a thickness of 3.5 mm. Chemical compositions (mass%) of the investigated mold fluxes are shown in

Table 1. Chemical composition of the studied slag system.

Sample No.	Chemical Composition (wt. %)
	CaO	SiO2	B2O3	Na2O	Al2O3	MoO3	C
1	39.000	39.000	10.00	8.0	4.0	0	0.0
2	38.025	38.025	9.75	7.8	3.9	2	0.5
3	37.050	37.050	9.50	7.6	3.8	4	1.0
4	36.075	36.075	9.25	7.4	3.7	6	1.5

. A part of the quenched samples was ground to fine powders and subjected to XRD. Another part of the quenched samples was coated with gold, and then subjected to SEM-EDS examination. SEM-EDS examinations were carried out using an FEI MLA 250 (Thermo Fisher, Waltham, MA, United States) unit equipped with Bruker SDD-detector (Bruker, Billerica, MA, USA) for EDS analyses. The working voltage was 20 kV. XRD analyses were performed on an 18 kW X-ray diffractometer (model: RIGAKU TTRIII, Rigaku, Tokyo, Japan) with Cu-Kα radiation.

2.2. Radiation Experiment

The glassy disks were polished and then subjected to the radiation experiment. Transmittances of samples were measured by using an Ultraviolet-Visible-Near Infrared spectrophotometer (UV-Vis-NIR, Lambda750, Perkin-Elmer, Waltham, MA, United States) with an integrating sphere. The wavelength range of measurements was from 300 to 2500 nm. All measurements were performed at room temperature in air. Cho et al. [4] measured transmittances of disk mold fluxes samples at 300~773 K in a high-temperature cell using a Flourier transformation infrared (FT-IR) spectrometer. They found little change of shape and height of absorption, indicating that there is no large difference between high-temperature and room-temperature spectra.

2.3. Crystallization Experiments

Differential scanning calorimetry was employed to determine the crystallization temperature of samples during heating. The samples were pulverized and subjected to the DSC using argon as a purge gas with a thermal analyzer (STA 409C, Netzsch-Gerätebau GmbH, Selb, Germany). The samples were heated to 1350 °C with a heating rate of 20 K/min in a corundum crucible, using alfa-Al₂O₃ as a reference material. Approximately 25 mg of the powder sample was employed for each DSC run. Before measurements, temperature calibration, sensitivity calibration and balance calibration were performed.

To investigate the mechanism of crystallization of mold fluxes, quenched disk samples No. 1 and No. 2 were heated to crystallization temperature (peak temperatures for crystallization peaks) and soaked for 2 h. Then the samples were quenched in air and cut in perpendicular direction. Then these heat-treated samples were embedded with resin. The sections were ground and polished. After sections were coated with gold, the samples were subject to SEM-EDS examination. Parts of the samples were crushed and ground into fine powder and subjected to XRD measurement.

3. Results and Discussion

3.1. Microstructure of Samples

The micrographs of prepared disk samples are shown in Figure 1. The XRD patterns of powdered samples are shown in Figure 2. As can be seen in Figure 1, the Mo-free sample is homogeneous. No crystals but a glassy matrix was found in micrographs of the Mo-free sample. With the addition of MoO₃ and C, there were many particles found in micrographs of samples 2, 3 and 4. The particles agglomerated and formed large clusters. In the sample with MoO₃ = 2%, there were still many isolated spherical particles apart from clusters. The agglomeration of particles was promoted by increasing MoO₃ content and some coarser particles formed due to the agglomeration and growth of particles. There were no crystals found in all disk samples with MoO₃ + C addition, indicating that no crystallization took place during quenching.

Combining EDS results with the XRD pattern in Figure 2, the particle phases in micrographs could be determined. For the sample with MoO₃ = 2%, EDS showed a chemical composition of particles contains B and Mo. XRD confirmed that there was only one crystalline phase, MoB, in the sample with MoO₃ = 2%. MoB has a very high melting point of 2873 K [25], and this persisted in a solid state during the melting of mold fluxes. The possible reaction for formation of MoB could be as follows:

9C + 2MoO₃ + B₂O₃ = 2MoB + CO(g).

(1)

With increasing MoO₃ addition, there was C found in the EDS results of samples. XRD showed that the particle phases were shifted from MoB to Mo₂BC and Mo₂C. The melting point of Mo₂C is 2793 K [26] and the melting point of Mo₂BC is 3073 K [27]. These compounds were also solid during the melting of the studied slag system. The formation of Mo₂BC could be due to the following reaction:

17C + 4MoO₃ + B₂O₃ = 2Mo₂BC + 15CO(g).

(2)

The possible reaction for Mo2C formation could be as follows:

14C + 4MoO₃ = 2Mo₂C + 12CO(g).

(3)

3.2. Radiation Properties of Samples

Results from the transmittance measurements are shown in Figure 3. It can be clearly observed that transmissivities in all wavenumber ranges were heavily reduced by the introduction of Mo-containing particles. The reduction of transmittance by Mo addition became mild with the further addition of 6% MoO₃. The extinction coefficients of various samples could be calculated from transmittance values by employing the Lambert–Beer law:

$$\tau =\exp \left(-Ed\right)$$

(4)

where E is the extinction coefficient of the sample; d is the thickness of the sample; $\tau$ is the transmittance of the sample.

The calculated extinction coefficients of various samples are shown in Figure 4. Strong absorptions below 400 nm were found in all samples, which could be associated with a charge transfer band [10]. There are some new peaks around 400–600, 800–900 and 1600–2000 nm found in samples with MoO₃ addition, which could be due to the introduction of MoO₃ in the structure of glass [28]. The Mo-free sample had extinction coefficient values between 150 and 300 m⁻¹. The extinction coefficients from the Mo-free sample correspond to the lowest values in the references [23]. This is due to the fact that there were minor transition metal oxides, e.g., Fe₂O₃, in the composition of glasses reported in the references [23]. The transition metal oxides could increase the apparent absorptivity and decrease the apparent reflectivity, thereby increasing the absorption and extinction coefficients [29]. With the introduction of foreign particles, the extinction coefficients could be enhanced to near 1000 m⁻¹. The extinction coefficients increased with increasing MoO₃ content in the samples, indicating that the scattering due to particles was enhanced.

Yoon et al. [23,24] measured the extinction coefficients of mold fluxes containing iron metallic particles and found that the extinction coefficients were enhanced by the scattering of iron particles. They employed Mie’s theory [30] to calculate the scattering coefficients and the calculated values were found to be in good agreement with the measured values. In the present work, the extinction coefficients were also enhanced by the scattering of particles. However, there are some differences between the present work and Yoon’s work. The sizes of particles in the present work are larger than those in Yoon’s work. In particular, large particle clusters were found in the samples with the addition of MoO₃ + C. Mie’s theory could only be applied to light scattering by particles whose size is comparable with the wavelength of the incident light. Due to the larger size and agglomeration of crystals, the quantification of scattering due to particles by using Mie’s theory is difficult. However, the qualitative explanation of the extinction coefficient change due to the scattering of particles could be given by employing Mie’s theory. The scattering coefficient (${\mu }_{\mathrm{scat}}$) could be calculated from the total scattering cross-section (${\sigma }_{\mathrm{scat}}$) [23]:

$${\mu }_{\mathrm{scat}}={\rho }_s{\sigma }_{\mathrm{scat}}$$

(5)

where ${\rho }_s$ is the density of spherical particles in the glassy matrix.

The total scattering cross-section (${\sigma }_{\mathrm{scat}}$) could be calculated in terms of Mie’s theory:

$${\sigma }_{\mathrm{scat}}=\frac{{\lambda }^2}{2\mathsf{\pi }}{\displaystyle \sum }_{n=0}^{\infty }\left(2n+1\right)\left({\left|a_n\right|}^2+{\left|b_n\right|}^2\right)$$

(6)

where a_n and b_n are parameters defined as:

$$a_n=\frac{{\psi }_n\left(x\right){\psi }_n^{\prime }\left(mx\right)-m{\psi }_n\left(mx\right){\psi }_n^{\prime }\left(x\right)}{{\xi }_n\left(x\right){\psi }_n^{\prime }\left(mx\right)-m{\psi }_n\left(mx\right){\xi }_n^{\prime }\left(x\right)}$$

(7)

$$b_n=\frac{m{\psi }_n\left(x\right){\psi }_n^{\prime }\left(mx\right)-{\psi }_n\left(mx\right){\psi }_n^{\prime }\left(x\right)}{m{\xi }_n\left(x\right){\psi }_n^{\prime }\left(mx\right)-{\psi }_n\left(mx\right){\xi }_n^{\prime }\left(x\right)}$$

(8)

where ${\psi }_n$ and ${\xi }_n$ are Ricatti–Bessel functions. x is the particle size parameter which is the ratio of the subsidence perimeter of the spherical particle to the frequency of the incident beam in the medium:

$$x=\frac{2\mathsf{\pi }r{n}_{\mathrm{med}}}{\lambda }$$

(9)

where r is the particle radium; n_med is the refractive index in the medium; λ is the wavelength of the incident rays.

It could be seen from Equation (5) that the scattering coefficient is proportionate to the density of spherical particles. In the present samples, with increasing MoO₃ content, as can be seen in Figure 1, the number of particles increased. Accordingly, the scattering coefficient increased with increasing MoO₃ content. The extinction coefficient includes contributions from scattering and absorption. Since the composition of the base glass is the same, absorption coefficients of different samples are fixed. Therefore, with increasing MoO₃ content, the extinction coefficient of samples increases.

3.3. Crystallization of Samples

The DSC curves for the powder Mo-free sample and the powder sample with MoO₃ = 2% are shown in Figure 5. After the step for glass transition, an exothermic thermal peak could be found in each curve, which was due to the crystallization of the glassy sample. The glass transition temperatures T_g for the Mo-free sample and sample with MoO₃ = 2% were obtained as the temperature at half the step height in the DSC curves. The crystallization temperatures of samples T_c were determined as the peak position of the crystallization event. Thermal stability, T_c−T_g, could be employed to measure the crystallization ability of the sample. The larger the T_c−T_g value, the weaker the crystallization ability. The T_c−T_g values for the Mo-free sample and the sample with MoO₃ = 2% were 77.9 and 73.5 °C, respectively, which are close to each other. This indicates that the two powder samples have near crystallization ability.

The bulk Mo-free sample and the sample with MoO₃ = 2% were heated at the crystallization temperature determined by DSC for 2 h and quenched. The XRD patterns for samples are shown in Figure 6. It could be seen that only one kind of crystal, CaSiO₃, precipitated from all glassy samples. The ratios of CaO to SiO₂ for all samples were equal to 1. According to the phase diagram of CaO–SiO₂–B₂O₃ [31], the compositions of the present fluxes were within the primary crystal field of CaSiO₃. Therefore, the firstly precipitated crystals should be CaSiO₃. In our previous work on the crystallization of CaO–SiO₂–B₂O₃ [15], the crystallization product was also CaSiO₃, which is in line with the present work. It could be also seen that the crystalline peaks in the XRD pattern of the sample with MoO₃ = 2% were much sharper than those of the Mo-free sample. All these indicate that the crystallization of mold fluxes was promoted by the addition of 2% MoO₃.

The micrographs of the Mo-free sample and the sample with MoO₃ = 2% are shown in Figure 7 and Figure 8. In the micrograph of the Mo-free sample, it can clearly be seen that crystal nucleation and growth can be started from the edge of the sample, which corresponds to the surface of the sample. In enlarged micrographs, columnar crystals could be easily found. The growth direction for all crystals is from the edge to the inner part of the sample. The compositions of crystals were determined by EDS to be CaSiO₃, which is consistent with the XRD results. In most of the inner part of sample, there were no crystals found. All these indicate that the main crystallization mechanism for Mo-free samples could be surface crystallization. Surface crystallization was frequently found to be the crystallization mechanism for glass [32] and glassy mold fluxes [15,33]. Wang et al. [33] investigated the crystallization kinetic of glassy mold fluxes with different TiO₂ contents and suggested that the crystallization mechanism for all samples should be surface crystallization.

The micrographs of the sample with MoO₃ = 2% showed MoB clusters and many strip-like crystals. The EDS analysis showed that the composition of the strip-like crystals was CaSiO₃. The crystallization of CaO–SiO₂–B₂O₃ with various TiO₂ contents was investigated by DTA and SEM-EDS [15]. Stripe-like CaSiO₃ crystals were also found in all samples. The strip-like morphology of CaSiO₃ could be found in natural minerals. It is well known that CaSiO₃ has a linear chain-like structure. The strip-like crystals of CaSiO₃ may stem from its chain-like structural characteristic.

The sample with MoO₃ = 2% was well crystallized and very few glass matrices could be found. This indicates that the crystallinity of the sample with MoO₃ = 2% was much higher than that of the Mo-free sample. In addition, the crystal number and size in the sample with MoO₃ = 2% were much higher than those in the Mo-free sample, indicating that the nucleation and growth of crystals were promoted by the addition of Mo-containing particles. It is possible that there were some bubbles on the surface of the samples, which could also be beneficial to the nucleation of crystals. Crystals could be found not only in the edge (surface) but everywhere in the inner part of the sample with MoO₃ = 2%. This means that the crystallization mechanism of the sample with MoO₃ = 2% was critically altered. The crystallization mechanism was changed from surface nucleation and growth to internal crystallization.

Both surface crystallization and bulk crystallization could be found during the devitrification of glass. Zanotto et al. [32] pointed out that only glass systems that have T_g/T_L < 0.58 (T_g: glass transition; T_L: liquidus temperature) display measurable internal nucleation rates on a laboratory time-scale. It is a common practice in glass science to add some foreign particles or other nucleating agents to induce internal crystallization for the fabrication of glass ceramics. Heterogeneous nucleation fabricated by nucleating agents is usually responsible for the required internal crystallization [22].

When the melt contains solid particles or is in contact with a crystalline crucible or oxide layers, nucleation may be facilitated, which is known as heterogeneous nucleation [34]. Nucleation could be greatly facilitated by an area of low-energy solid/solid interface. The magnitude of the effect f(θ) could be calculated as a function of the contact angle, θ:

$$f\left(\mathsf{\theta }\right)=\frac{\left(2+\cos \mathsf{\theta }\right){\left(1-\cos \mathsf{\theta }\right)}^2}{4}.$$

(10)

The contact angle θ in Equation (10) is the contact angle between solid nuclei and solid particles with the presence of melts. However, it is difficult to obtain the contact angle due to significant data limitations. According to the theory advanced by Turnbull and Vonnegut [35,36], a nucleating agent will be effective in promoting nucleation when the lattice parameters in the low-index crystallographic planes of both the substrate and the nucleated solid are similar. The precipitated CaSiO₃ crystal is mainly pseudowollastonite and has a triclinic symmetry (JCPDS-PDF, card No. 74-874). However, the degrees of the three angles are approaching 90°. The lattice parameters are a = 6.85300, b = 11.89500 and c = 19.67400. The MoB particles have a tetragonal symmetry and lattice parameters of a = b = 3.11245, c = 16.9560 (JCPDS-PDF, card No. 51-940). It was found that the c parameter of MoB is close to that of CaSiO₃, indicating that MoB could be effective in provoking heterogeneous nucleation.

It seems that the promotion of crystallization by MoO₃ + C in the bulk sample is inconsistent with the observation in DSC. This could be due to the different nature of samples. In the powder sample investigated by DSC, due to fine size and defects caused by crushing, the crystallization mechanism could be totally different from the bulk sample. It was reported that an even particle size of the powdered samples would influence the crystallization mechanism of glass determined by DSC [37].

It was observed that the introduction of MoO₃ + C would decrease the extinction coefficients of fluoride-free mold fluxes. The radiative heat transfer of the glassy layer would be weakened by the scattering of Mo-containing particles. On the other hand, the introduction of MoO₃ + C would also promote the crystallization of mold fluxes. The enhanced crystallization of mold fluxes would also weaken the radiative heat transfer of mold fluxes by the scattering of precipitated crystals. Moreover, the well-developed crystalline layer would increase the interfacial thermal resistances. In conclusion, the introduction of MoO₃ + C would weaken the heat transfer between the mold and the steel shell, and therefore lead to mild cooling, which is beneficial for the continuous casting of some crack-sensitive steel grades.

4. Conclusions

MoO₃ and C mixtures (mass ratio: 4:1) were introduced into CaO–SiO₂–B₂O₃-based mold fluxes to adjust the radiative and crystallization of fluoride-free mold fluxes. The influences of particles on the radiative and crystallization of glassy CaO–SiO₂–B₂O₃-based mold fluxes were investigated by UV-Vis-NIR, DSC, XRD and SEM-EDS. The following conclusions could be drawn:

(1)

Particles of MoB, Mo₂BC and Mo₂C with very high melting temperatures could be generated by carbon reduction in mold fluxes. The particles collided and agglomerated into clusters.

(2)

Transmittances in all wavenumber ranges were heavily reduced by the introduction of Mo-containing particles. With the addition of Mo-containing particles, the calculated extinction coefficients were enhanced.

(3)

Both the Mo-free sample and the sample with MoO₃ = 2% precipitated CaSiO₃ crystals. The crystallization of fluxes was promoted by the introduction of Mo-containing particles.

(4)

The crystallization mechanism for the Mo-free sample is mainly surface nucleation and growth. The introduction of Mo-containing particles facilitates heterogeneous nucleation, and the crystallization mechanism changes from surface nucleation and growth to internal nucleation and growth.