New approach of recycling vanadium‐bearing slags as a binder in high‐alumina refractory castables application

Abstract

Handling the massive quantities of by‐products from metallurgical processes has become a major concern in recent decades. Efforts to develop sustainable alternatives for these secondary resources are ongoing to achieve the transition to climate neutrality. This study has investigated the potential of employing vanadium‐bearing slag as a new value‐added binder in refractories, aiming to replace virgin raw materials. Two types of vanadium‐bearing slags from BOF, each containing <2 wt.% vanadium were studied. Low‐cement vanadium slag‐based castables were prepared by gradually substituting 0, 2.5, and 5 wt.% of the commercial calcium aluminate cements (Secar71 and CMA72) by the slags. The flow values of the mixes containing 5 wt.% of slag decrease significantly from about 90% to 30% after 30 min, showing poor ability to flow and thus are not considered as self‐flow castables. Castables containing 2.5 wt.% of slag present a cold crushing strength value, in the range of 71–116 MPa while values for castables containing 5 wt.% of slag fall into the range of 53–68 MPa due to the lower packing properties leading to higher porosity and reduce in strength. Similar observation was concluded for cold modulus of rupture. The micrographs of the samples containing both slag and cement show promising compatibility between the binder and aggregates after sintering at 1500°C. Overall, characteristics obtained with the samples containing slag show promising alternatives as a refractory lining.

INTRODUCTION

The steel-making industry has the highest demand for refractory materials, accounting for over 70% of the global supply.¹ These refractories are used as lining systems throughout the entire process from the blast furnace to the submerged entry nozzle for continuous casting. They are determinant for steel cleanliness by controlling the solute elements (carbon, hydrogen, oxygen, and nitrogen) and preventing inclusions. Understanding the interactions between metal, slag, and refractories is essential for maintaining control over the quality of steel. Refractory castables have evolved and are increasingly employed in the steelmaking industry for lining vessels in high-temperature applications over the past six decades.² As compared with conventional bricks, these refractories present faster, easier, and cheaper installation (self-flow, casting, and gunning by shotcreting). As they are essential for high-temperature processes, they necessitate high performing top-quality raw materials, aggregates, and binders, to guarantee their properties.

The large-sized (up to centimeters) aggregates represent 60–80% of the volume of the ceramic body. In order to ensure bonding between the aggregates, they are held with a finer binder phase, usually a calcium aluminate cement (CAC) in the castables. This hydraulic binder is capable of forming in situ bonds in castable refractories by reacting with water to form hydrated calcium aluminate phases.³ The refractory castables bonded with CAC have a rich history of nearly 90 years.⁴ Firstly, conventional castables were considered with 15–30% of CAC, rapidly overtaken by low cement castables with 1.0% < CaO < 2.5% in the 1970s to decrease the water demand and attaining lower porosity with improved mechanical strength.⁴ In the 1980s, ultra-low cement castables were developed with 0.2% < CaO < 1.0%. Cement-free castables with CaO < 0.20% have been developed in the early 2000s and are nowadays the subject of ongoing research because of their rapid dryout and favorable heating properties. Reducing the calcium oxide content in the matrix enables superior corrosion resistance from metal and slags but reduces mechanical properties. Generally, the production of 1 tonne of cement generates approximately 1.2 tonnes of CO₂ equivalent emissions, which represents 1% of the global CO₂ emissions.⁵ Moreover, in terms of energy consumption, the lime production route is an energy-intensive process. As a result, and to comply with the European Green Deal ensuring climate neutrality by 2050, the usage of CAC in the refractory castables industry is projected to decrease substantially over the next decades. Nowadays, it has become widely recognized that the refractory industry significantly relies on imported natural raw materials.⁶ The increase of industrial production and the reduction in available natural sources both contributed to increase ecological problems and costs of mineral raw materials.⁷ Thus, a shift to a more sustainable management of natural resources and energy represents a key component of protecting the environment and reducing global CO₂ emissions.⁸

The steelmaking sector is one of these heavily impacted industries, producing a variety of wastes and by-products during the many stages of steel manufacturing. Within the steelmaking process, one tonne of steel in a steel plant generates half a tonne of by-products such as slags, dusts, and sludges.⁹ Annually, the global steel production totals approximately 2 billion tonnes, implying a consequent production of at least 1 billion tonnes of waste.¹⁰ Dealing with such massive quantities of waste is certainly a major concern. Landfill is the easiest disposal waste management. In order to overcome this issue, the scientific community has focused on finding alternatives for these valuable materials.

The highest residual component is steelmaking slag from electric arc furnaces (EAF) and basic oxygen furnaces (BOF). Depending on the steel production route, different types of slags are produced with different compositions.¹⁰ BOF's composition presents a high amount of Al₂O₃ content and a significant amount of lime that is suitable for binding application.⁹ Slag composition (mainly lime, and magnesia) and their hydration capacity to form hydrates (C₃AH₆, C₂AH₈, CAH₁₀, and AH₃) are the main parameters that affect the feasibility of slag as cement.^10,11 Commonly, iron and steelmaking slag has been widely used as the bonding system in the production of refractories.

For specific applications requiring strengthening the steel, vanadium metal at very low level (<1 wt.%) is added to hot metal in the converter. Consequently, vanadium is found in BOF slags. As a by-product, vanadium-bearing slag presents a significant amount of lime (CaO), alumina (Al₂O₃), and magnesia (MgO) making it suitable to employ as a secondary raw material in refractories formulations for linings applications. When combined with a deflocculant, it has qualities that are comparable to those of calcium aluminate cement in terms of hardening and setting.¹² The vanadium content in the slag depends on the molten steel, and the technologies used to obtain it. Currently, only a few studies exist with vanadium-bearing slag feasibility in high-temperature applications. Therefore, the present study addresses the use of vanadium slag in high-alumina refractory castable formulations to obtain high-performance refractory castables. Several vanadium slags with variations in composition are employed to enlarge the panel of results and propose a relevant formulation of castables, containing vanadium slag, while respecting the required properties for industrial applications.

This work aims to determine the effects of the vanadium-bearing slags in castables applications and their feasibility as a binder system to substitute calcium aluminate cement. This paper focuses on the resistance and the properties of high alumina castables with both vanadium slag and commercial calcium aluminate cement. The impact of the vanadium slag is compared to that of conventional reference material for comparable applications. The present materials were characterized by X-ray diffraction (XRD) and scanning electron microscopy (SEM). Furthermore, the mechanical properties of the sintered samples at 1500°C were analyzed. Remodeling the processing route of refractory materials used for steelmaking could support the decarbonization of the steel industry. Using secondary resources solves the issue of natural raw materials, reduces costs related to their extraction and processing, and is friendly to the environment.

EXPERIMENTAL PROCEDURE

Starting materials

This study used commercially available calcium aluminate cement (Secar 71, CMA72) as a hydraulic binder reference in refractory castables formulation. Two vanadium slags were given by SPAETER raw materials and Calderys and originate from high-vanadium steel production with a given particle size ≤63 µm. Both slags originated from BOF steel process. Within the present research work, vanadium slags are consequently named slag $$a$$ and slag $$b$$. Both slags were ground to a median grain size d50 of 11–20 µm. High alumina refractory castables were designed. For this purpose, aggregates and fines used were tabular alumina. As ultrafine alumina, a reactive alumina was chosen with a grain size below 1 µm. As a quality standard, the chemical composition of the starting materials was measured by X-ray fluorescence (XRF) (Table 1). The particle size distribution is a key factor controlling the properties of refractory castables. The better the packing, the less binder is required in the formulation. To achieve optimized packing for this study, a theoretical self-flow continuous curve based on the Andreasen model was executed as follows: 1 $$$\begin{equation}\text{CPFT} = 100{{\left( {\frac{d}{D}} \right)}^q},\end{equation}$$$where, CPFT represents the cumulative percent finer than, d the particle size, D the maximum particle size (4000 µm), and q the distribution coefficient.¹³ The q value is determined experimentally and must be lower than 0.37 for optimum packing.¹⁴ The different trials underlined that the factor should be lower than 0.25 for the self-flow formulation of castables. This q-value, being lower than the q-value for optimum packing, indicates the necessity for particle mobility in the castable, rather than solely conforming to the voids.

TABLE 1 Chemical composition of the raw materials.

	Composition (wt.%)
Type	Al2O3	CaO	MgO	V2O5	Fe2O3	SiO2
Tabular alumina	99.9	–	–	–	–	–
Reactive alumina	99.8	0.04	–	–	0.03	0.07
Secar 71	68.6	30.1	0.31	–	0.13	0.32
CMA72	68.3	9.40	21.4	–	0.21	0.31
Slag $$a$$	74.0	6.13	15.7	2.05	0.38	0.85
Slag $$b$$	68.5	25.4	2.86	0.73	0.23	1.34

For the study, castables were formulated using a q equal to 0.25. The particle size distribution of the mixtures was set up to be the closest to the optimum curve with 100% packing density. The particle size distribution of the castables with the slags and with the cement is represented in Figure 1. This study focuses on the resistance and the properties of high alumina castables at high temperatures. Following the Andreasen model, the composition of the refractory samples is shown in Table 2. S71_5 and CMA_5 stand as reference castables made with 5 wt.% of Secar 71 and CMA 72, respectively. Castables with a full substitution of the cement by the vanadium slags are Sa_5 and Sb_5 standing for slag $$a$$ and slag $$b,$$ respectively. Four formulations containing both vanadium slag and cement in half proportions were made as Sa2.5_S712.5; Sa2.5_CMA2.5; Sb2.5_S712.5 and Sb2.5_CMA2.5. A commercial dispersant (provided by BASF) was added to efficiently deffloculate the castables with a low addition of mixing water.

[IMAGE OMITTED. SEE PDF]

TABLE 2 Composition of ultra-low cement castables with substitution of cement by slags in wt.%.

		Sample code
Raw material	Grain size	S71_5	CMA_5	Sa_5	Sb_5	Sa2.5_S712.5	Sa2.5_CMA2.5	Sb2.5_S712.5	Sb2.5_CMA2.5
Tabular alumina	1–3 mm	35	35	35	35	35	35	35	35
0.5–1 mm	10	10	10	10	10	10	10	10
	0.2–0.6 mm	10	10	10	10	10	10	10	10
	0–0.3 mm	17.5	17.5	17.5	17.5	17.5	17.5	17.5	17.5
	0–0.045 mm	10	10	10	10	10	10	10	10
Reactive Alumina	<0.001 mm	12.5	12.5	12.5	12.5	12.5	12.5	12.5	12.5
Secar 71		5	–	–	–	2.5	–	2.5	–
CMA 72		–	5	–	–	–	2.5	–	2.5
Slag a		–	–	5	–	2.5	2.5	–	–
Slag b		–	–	–	5	–	–	2.5	2.5
Water		5	5	5	5	5	5	5	5
Deflocculant		0.1	0.1	0.1	0.1	0.1	0.1	0.1	0.1

Castables preparation

Ultra-low cement refractory castables were prepared using the composition provided in Table 2. The dry raw materials were mixed (except for the dispersant) in a conventional mortar mixer for 2 min at a constant rotation speed of 140 r.p.m. Meanwhile, the 5 wt.% of water was mixed with the dispersant and added to the dry mix after 2 min of mixing followed by mixing for 5 min. Once both mixing times concluded, the bottom of the pot was scraped with a spoon and mixed again for 2 min to obtain high homogeneity and desirable flowability. The castables were poured into steel molds without a vibrating process. After casting, the samples were cured in a wet chamber (relative humidity >90%) at ambient temperature for 24 hours. After removing the samples from the molds, they were cured again in the wet chamber for 24 hours, followed by curing in a dryer oven at 100°C for another 24 hours. Finally, the green samples were fired in an electric furnace following a thermal treatment with a first heat-up to 500°C and a plateau of 6 hours, followed by another heat-up to 1500°C and a plateau of 6 hours. After sintering, cooling was performed to standard temperature. The rate was applied to 2°C/min for heating and cooling.

Characterization methods

The crystalline phase identifications of the slags and cements as well as the different castables samples were determined using an X-ray diffractometer (D8 Advance, Bruker) with Cu-Kα radiation source (λ = 1.54059 Å) in the 2θ range of 5°–90°. The diffraction data were analyzed with Diffrac.eva software. The particle size distributions of the slags and cements were obtained using a laser diffraction analysis under dry dispersion.

Differential scanning calorimetry (DSC) technique was used to undertake thermal analysis (Netzsch DSC 404). The experiments were conducted in a platinum sample holder at a heating rate of 5°C/min in an air and argon environment. The DSC technique provides information about exothermic or endothermic reactions related to crystallization or melts.

Scanning electron microscope (SEM) using Gemini SEM 500 (Zeiss Instruments) equipped with an energy dispersive spectrometer (EDS, XMax 80) was used to perform microstructural examinations of the vanadium slags and the heat-cured castables. The samples were then cast in epoxy (Cloeren) cold resin under low pressure. Polishing was conducted, following the protocol for refractory materials. A carbon-coated layer was applied to the samples for electrical conductivity.

The bulk density and apparent porosity of the castables were measured according to Archimedes' principle and DIN EN 993-1. The true density, defined as the ratio of mass over the volume of the sample, was obtained with the DIN EN 993-2 standards. A helium pycnometer was used for true density measurement.

Heating microscopy was used to determine the dimensional changes as a function of temperature during heat treatment of the slags. Information on sintering, swelling, or melting as well as contact angle measurement were derived from the dilatometric measurements. These tests were performed using a Carl Zeiss microscope and the EMI III software according to DIN 51730. The slag was analyzed under a heating rate of 80°C/min up to 1600°C.

Self-flow measurements were considered to evaluate the performance of self-flowing castables. The flowability was determined with the same recipes as for castable preparation with a fixed water content of 5 wt.% based on dried mixes. The mixtures were tested at 10, 20 , and 30 min after water addition. Then, the mix was poured into a standard cone with a 100 mm base diameter and a 70 mm upper diameter. The cone was removed, and the wet mix was able to spread out on an oily surface. Diameters were measured after 1 day of flowing. The flow factor was determined for each test following the Equation (2): 2 $$$\begin{equation}\text{Flow\ factor}\;\left( \% \right) = \frac{{\bar{d} - {{d}_0}}}{{{{d}_0}}} \cdot 100,\end{equation}$$$where, $$\bar{d}$$ is the average diameter after spreading and $${{d}_0}$$ is the initial base diameter (100 mm). When the flow factor value is within the range of 80–110% of the base diameter, the castable is considered self-flowable.¹⁵

Testing procedure and methods

Cold crushing strength (CCS) represents the ability to resist failure under compressive load. For high alumina castables, typical CCS values are in the range of 50–100 MPa.^16,17 CCS was determined using a uniaxial compression testing machine following EN 993–5. The tests were carried out at a constant load rate of 1.0 MPa on 40 × 40  mm cylindrical samples. The presented results are the average of 3 values for each composition. The CCS (σ) is given by the following expression in MPa: 3 $$$\begin{equation}\sigma = \frac{{{{F}_{{\mathrm{max}}}}}}{{{{A}_0}}}\ ,\end{equation}$$$with $${{F}_{\mathrm {max}}}$$, the maximum load recorded (N) when the failure occurs, $${{A}_0}$$ is the initial cross-section area of the sample over which the load is applied.

Refractory materials are defined by mechanical resonant frequencies related to the mass, dimensions, and elastic properties. The measurement of the resonant frequencies can give the elastic properties, E, G, and ν. Poisson's ratio is a dimensionless factor connecting E and G. To measure E and G, the resonant frequency damping analysis (RFDA) is used (ASTM E1876-22) with an IMCE testing device. This analysis has the advantage of being non-destructive. To do so, the sample is excited by a mechanical impulse, which results in a vibration of the sample at its resonance frequency. The acoustic signal is recorded with a microphone and the frequencies are analyzed to determine E and G modulus. The results depend on the geometric dimensions of the sample and its elastic behavior. The tests have been performed with 150 × 25 × 25  mm samples. To measure the modulus of elasticity E, the sample must be in a flexural mode situation, which means that the sample is held on two parallel wires and the excitation occurs in the middle of the lower face. The vibration is recorded in the middle of the upper face. Using a given Poisson's ratio equal to 0.275 in this study, E can be determined according to the following equations¹⁸: 4 $$$\begin{equation}E = 0.9465\left( {\frac{{m{{{\left( {{{f}_f}} \right)}}^2}}}{b}} \right)\left( {\frac{{{{L}^3}}}{{{{t}^3}}}} \right){{T}_1},\end{equation}$$$where, $$m$$ is the sample weight, $${{f}_f}$$ the flexural resonant frequency, $$t$$ the thickness, $$b$$ the width, $$L$$ the length of the sample, and $${{T}_1}$$ a dimensionless correction factor. $${{T}_1}$$ depends on the ratio $$\frac{L}{t}$$ and the Poisson's ratio was chosen to be 0.275.

The microstructure, pores, and cracks influence the vibration of the sample. To determine the shear modulus G, the sample must be in a torsion mode situation, meaning the excitation point and the measurement point are opposite each other across the spatial diagonal of the sample. Thus, the torsional vibration frequency is measured, and the shear modulus is calculated with the Equation (5). 5 $$$\begin{equation}G = \frac{{4Lmf_t^2}}{{bt}}\ R,\end{equation}$$$where, $$m$$ is the sample weight, $${{f}_t}$$ the torsional resonant frequency, $$L$$ the length of the sample, $$b$$ the width, $$t$$ the thickness, and $$R$$ the correction factor.

Using the standard DIN EN 993–6, the cold modulus of rupture (MOR) of the castables fired at 1500°C was determined. The MOR uses a three-point bending method applied on a rectangular piece (150 × 25 × 25  mm) to determine the maximum stress that the sample can withstand. Each composition was tested on 3 samples, giving the following average values. The modulus of rupture $$\sigma ( F )$$ depends on the ratio of the sample's failure load $$( {{{F}_{{\mathrm{max}}}}} )$$ and it is calculated as follows: 6 $$$\begin{equation}\sigma \left( F \right) = \frac{3}{2} \cdot \frac{{{{F}_{\max }} \cdot {{L}_s}}}{{b \cdot {{h}^2}}},\end{equation}$$$where $${{F}_{{\mathrm{max}}}}$$ is the maximum force exerted on the sample (N); $${{L}_s}$$ is the distance between the points of support (mm); $$b$$ is the breadth of the sample (mm) and $$h$$ is the height of the sample (mm).

RESULTS AND DISCUSSION

Material characterization

The crystalline phases for the two calcium aluminate cements and the vanadium-bearing slags are shown in Figure 2. The amount of each phase increases with the number of “+” signs, which also increases the diffraction peak's strength. Spinel phase MgAl₂O₄ is the main phase in slag $$a$$ while in slag $$b$$, krotite (CaO.Al₂O₃) is the main crystalline peak. This is explained by the small amount of MgO, <3 wt.% for the slag $$b$$. Hibonite (CaO(Al₂O₃)₆) is only encountered in the slag $$a$$. Grossite (CaO(Al₂O₃)₂) is present in all the binders in different proportions. Cement Secar 71 doesn't contain spinel phase as confirmed by its chemical composition. In contrast, spinel appears as the main phase for the CMA72 cement. As shown by the chemical composition of the slags, the very low content of vanadium oxide has not led to the detection of a related crystalline phase by XRD. Further microstructural examination is required to detect and quantify the vanadium.

[IMAGE OMITTED. SEE PDF]

Backscattered electron SEM micrographs of the Vanadium-bearing slags are displayed in Figure 3. The grain shapes of the slag $$a$$ revealed that it contains different crystalline phases with a wide particle size distribution. The particle sizes start from fine particles of 1 µm and extend to particles of several dozen micrometers. The grain shapes of both slags are identical to the grain shape of the cements overall, displaying calcium aluminate phases (CA). Vanadium was detected in brighter spots using EDX for the quantification and identification, included in spinel phases (Mg(Al,V)₂O₄) as detected by XRD analysis. However, the vanadium was not detected in slag $$b$$. Phase precipitation during the cooling process of the slag resulted in the formation of a stable phase consisting of spinel, as indicated by the observed microstructure. Typical thermal expansion coefficient for spinel is in the range 8-9 ppm/K. The presence of vanadium in the spinel structure has probably minor effect on the expansion behavior of the grain, especially as it is found as an impurity in low amount. However, free vanadium oxide has not been found in the slags which suggests a potential entrapment of the vanadium within other oxides such as alumina, calcium, and magnesia. This reduces considerably the hazardous risks related to vanadium slag management and disposal.

[IMAGE OMITTED. SEE PDF]

The DSC measurement curves of the slags and the commercial cements measured in air atmosphere are presented in Figure 4. The DSC measurement was conducted in a range of 30°C to 1450°C. The small peak indicated in Figure 4 indicates endothermic activity that was observed at 258.6°C for slag $$b$$. This is associated with the removal of impurities such as organic components and carbon through dehydration. Thus, there is less stability for the slag $$b$$ compared to the slag $$a$$. Above 1250°C, two distinct behaviors are observed. Slag $$b$$ and CMA72 exhibit an endothermic behaviour as indicated by a rise in their curves, which is associated to the presence of spinel in the composition, while slag $$a$$ and cement Secar71 show an exothermic behaviour as indicated by a fall in their curves. The behaviors of the slags were found to be similar to the commercial cements Secar 71 and CMA72. This observation leads to the conclusion that the presence of vanadium in both slags has no impact on the DSC curves and therefore on the behavior of the slags at high temperatures.

[IMAGE OMITTED. SEE PDF]

The height variation of the samples as a function of the temperature of the slags $$a$$ and $$b$$ and the cements are presented in Figure 5. This method indicates the sintering temperature. The beginning of the sintering process is indicated by the temperature Ts, which corresponds to the first point where the area decreases by 2% from its initial value of 100%.¹⁹ This temperature is determined as the onset of the sintering process. The sintering temperature for the slag $$b$$ was found to be 1470°C. When heated up, slag $$a$$ and $$b$$ behaved contrary. While $$a$$ was found to expand, $$b$$ did not show any significant expansion. This observation can be explained by the expansion behavior of the spinel phase. The temperature of sintering for the slag $$a$$ was not precisely determined but the curve's trend suggests that the value is at least 1550°C or higher.

[IMAGE OMITTED. SEE PDF]

Castable characterizations and properties

As indicated in Figure 6 the variations of self-flow values of the castables over time following water addition were shown. It can be seen that the flow values are similar for the mix incorporating both vanadium-bearing slags and cements. The same conclusion can be drawn for the reference mixtures containing commercial cement. However, a slight increase in flow is reported for the mixes using slag $$b$$ in the bonding system from 10 to 20 min before decreasing again at 30 min. Both mixes with 5 wt.% of slag $$a$$ and slag $$b$$ demonstrate lower flow values, likely due to the lower specific surface area (SSA) of the slags compared to the cements. Reduced SSA, leads to decreased dispersion of the castables. Indeed, castables with a higher SSA require more water to achieve proper dispersion, as discussed elsewhere.²⁰ Self-flow values must be in the range 80–110% of the base diameter, to meet the criteria for a self-flow castable.¹⁵ Based on this experiment, the mixes containing 5 wt.% of slag are unable to flow under their weight and thus do not qualify as self-flowable. By strategically adjusting the recipe using the Andreasen model, it may be possible to enhance the rheological performance of the castables. A new type of refractory castables may be defined to obtain such results.

[IMAGE OMITTED. SEE PDF]

The chemical compositions and the XRD patterns obtained for regular castables and vanadium slag-containing castables up to 5 wt.% are represented in Table 3 and Figure 7, respectively. The corundum (α-Al₂O₃) phase remains the major crystalline peak in all the castables consequently. Spinel (MgAl₂O₄) and hibonite (CaAl₁₂O₁₉) were detected in only small amounts. The greater the amount of slag in the castables, the more the intensities of corundum peaks increase slightly compared to the reference samples. The sample CMA_5 presents the highest amount of MgO, 1.02 wt.%, which results in a spinel phase in the diffraction graph. This is linked to the presence of spinel in the binder CMA72. The existence of spinel in the cement CMA72 results in spinel formation in castables CMA_5 with 1.02 wt.% of MgO; Sa2.5_CMA2.5; Sb2.5_CMA2.5 exhibiting an intense peak for castables with slag $$a$$ attributed to the high presence of spinel in the slag $$a$$, also observed in the castable Sa_5. All the castables share a chemical composition with at least 97 wt.% of Al₂O₃. The binary phase diagram of the CaO–Al₂O₃ system at this composition and a firing temperature of 1500°C, underlined the presence of hibonite (CA₆). During the firing process, the grossite and krotite disappear in favor of hibonite as indicated in Figure 6. Indeed, the CAC reacts in the presence of water to form calcium aluminate hydrates. Upon heating to 1500°C, all the cement hydrates convert into CA₆ in the presence of a suitable source of Al₂O₃ as investigated by many researchers.^21–23 The samples Sa_5; Sa2.5_CMA2.5 and CMA_5 contain CaO bellow 0.50 wt% which explains the absence of hibonite peaks in their XRD data. The V₂O₅ content is insignificant for castables containing slag $$b$$ compared to the castables containing the slag $$a$$. This can be explained by the higher presence of V₂O₅ in slag $$a$$ composition than in slag $$b$$. Also, it is impossible to detect such minor phases in the present powder XRD analysis with the device available.

TABLE 3 Chemical composition of the castables.

	Composition (wt.%)
Sample code	Al2O3	CaO	MgO	V2O5	Fe2O3	SiO2
S71_5	97.9	1.48	–	–	0.04	0.1
CMA_5	97.8	0.46	1.02	–	0.05	–
Sa2.5_S712.5	98.1	0.92	0.3	0.06	0.05	0.19
Sa2.5_CMA2.5	97.8	0.39	0.96	0.05	0.06	0.13
Sb2.5_S712.5	97.8	1.37	–	0.02	0.05	0.14
Sb2.5_CMA2.5	97.8	0.85	0.54	0.02	0.05	0.11
Sa_5	98.3	0.32	0.67	0.1	0.06	0.14
Sb_5	97.8	1.25	0.08	0.05	0.05	0.15

[IMAGE OMITTED. SEE PDF]

The apparent porosity, the bulk density, and the true density of the slag-based castables and the reference are shown in Table 4. For each composition, three samples were used to determine both the apparent porosity and the bulk density. The true density has been measured on two samples. The results display the averages of the values. The values of the true density are of the same order for the castables containing slag and the references, approximately 3.95 g/cm³. Table 4 features that the addition of slags increases the porosity and so decreases the bulk density. However, the changes in porosity were found to be less significant thanks to the ceramic bonds being formed at 1500°C by densification. As seen in the Figure 6, the samples Sa_5 and Sb_5 display lower flow values compared to the other castables. This lower rheological property is indicative of poorer packing of the materials, explaining the higher apparent porosity and the smaller bulk density of the hardened samples. An abrupt increase of porosity is observed for castable containing 5 wt.% of slag $$b$$ which can be correlated to the lack of hydraulic phases formed during the hydration of slag resulting in lower strength development. Moreover, it can also be explained by the presence of hibonite in the XRD patterns which are known for their volumetric expansion behaviour.²¹ For the vanadium-slag-containing castables, fewer hydrates are formed due to the combination of CAC and slag, resulting in a decrease in hydraulic properties. As a consequence, the bonding between the grains is arduous which results in a porous and less dense ceramic body providing less mechanical resistance. The higher porosity observed in the sample CMA_compared to sample S71_5 was attributed to the volume changes associated with the spinel phase present in the CMA72 cement, absent in the Secar71 cement.

TABLE 4 Apparent porosity, bulk density, and true density of the samples.

Sample code	Apparent porosity (%)	True density (g/cm3)	Bulk density (g/cm3)
S71_5	17.5	4.02	2.93
CMA_5	18.1	3.94	3.10
Sa2.5_S712.5	18.8	4.00	3.07
Sa2.5_CMA2.5	19.5	3.95	3.10
Sb2.5_S712.5	22.3	3.90	2.92
Sb2.5_CMA2.5	18.9	3.93	3.07
Sa_5	20.9	3.96	2.93
Sb_5	25.7	3.93	2.80

Figures 8–10 show the SEM micrographs of the sintered castables containing slags $$a$$ and $$b$$ as well as the commercial cements. The typical microstructure of a refractory material is well observed in all the samples. This includes coarse sharpy grains of alumina (≥1 mm) surrounded by the matrix, which was composed of fine grains (≤50 µm), including the bonding phase and pores. The dense microstructure displays different types of pores. Large spherical pores (10 µm) observed in Figure 8A or in Figure 10B represent the intergranular porosity. The small interconnected and closed pores located scattered within the aggregates are also noticed in Figures 8A,9A and 11B.

[IMAGE OMITTED. SEE PDF]

[IMAGE OMITTED. SEE PDF]

[IMAGE OMITTED. SEE PDF]

[IMAGE OMITTED. SEE PDF]

The presence of microcracks that separate the aggregates from the matrix is well observed in Figure 8B; Figure 10A and Figure 11B. This observation was explained as follows factors: the high presence of spinel phases which during the sintering process leads to expansion and additionally, the anisotropic grain growth of CA₆ during dehydration gives rise to hexagonal plate crystals.^7,21,22 The porosity in samples Sb_5 and Sa_5 is notably elevated, as observed in their microstructure. This increase in porosity underlined the poorer flowability of these samples, as illustrated in Figure 6. Consequently, the packing properties of the materials are decreased, as reflected in Table 4. The sintering process has induced needle and platy shapes for the fine particles in the matrix, particularly well observed for samples containing the slag $$b$$ in Figures 10A,B and 11B, increasing the mechanical strength of the samples.²⁰ The EDS analysis supported the previously determined mineral phases validated by XRD in Figure 7.

The micrographs in Figure 8A,B of the reference samples containing commercial cements demonstrate a dense, homogeneous distribution of the aggregates embedded within the matrix. This indicates a high level of compatibility between the aggregates and the binder.

The sample Sa_5 in Figure 11A displays a grain within the matrix, characterized by layers measuring 40 µm in size. This grain contains a vanadium oxide phase at a concentration of about 0.7 wt.%. A similar observation is made for sample Sa2.5_S712.5 in Figure 9A, which also exhibits a vanadium oxide phase concentration of 0.6 wt.%. Vanadium phase is observed within magnesium, aluminum, oxygen and calcium suggesting the formation of a spinel phase incorporating vanadium represented as Mg(Al,V)₂O₄. This phase remained undetected by XRD measurements due to its lower concentration compared to the higher concentration of the alumina spinel phase.

The substitution of the cement by the slag seems to have a detrimental impact on the microstructure after the firing process. In contrast to the reference samples, the aggregates exhibited less embedding within the matrix when containing slag, as evidenced in Figures 9A,10A and 11B. This is mainly attributed to the poorer bonding properties of the slags, characterized by lower CaO content and consequently reduced capacity to react and form hydrates with alumina and water. Nevertheless, a homogeneous dispersion of the slag is observed in the matrix. The matrix significantly influences the mechanical properties of the materials, as discussed in subsequent sections. Indeed, the formation of CA₆ during sintering, establishes bond linkages between the particles in the matrix and the aggregates.²⁰

As indicated in Figure 9A, sample Sa2.5_S712.5, the alumina grain is clearly identified. During sintering, this grain undergoes attack by the calcium aluminate binder (Secar 71) leading to the formation of calcium aluminate phases such as CA₆. The vanadium-containing particle next to the alumina grain does not provoke any reaction with the alumina grain; instead, it remains within the matrix. Throughout the entire microstructure, vanadium particles are predominantly confined to the matrix and do not interact with the aggregates as the calcium aluminate cement. Consequently, if the castable would be recycled by extracting the alumina grains, they would likely be free from vanadium contamination.

Figure 8A presents a microstructure featuring a region containing a high concentration of 9.8 wt.% and 12.5 wt.% of vanadium oxide phase. In contrast to Figures 9A and 11A, the microstructure does not present a distinct grain with a layered structure. Instead, it shows alumina grains being attacked by vanadium-containing particles. The brightness caused by mass contrast of the image indicates the presence of vanadium in a metallic state. This results in white dispersed spots near metallic droplets. Neither XRF measurement nor XRD analysis detected the high presence of Na₂O and SO₃. This observation was limited to this specific area of the sample. The slag is not present as a spinel phase in the castable when mixed with CMA72, and therefore might lead to inferior mechanical properties.

Mechanical response

The ability to resist failure under compressive load decreases significantly when adding slag to the formulation as observed in Figure 12. It can be observed that there is a big strength difference between the reference sample containing 5 wt.% of Secar71, 259 MPa, and the one containing 5 wt.% of CMA72, 133 MPa. According to CCS values, the use of slag (both $$a$$ and $$b$$) decreases significantly the average compressive strength. The values fail by up to three times the reference value when 5 wt.% of slags are added. When slag is present in the castable, the hydrate bonding effect of the calcium aluminate cement is attenuated. Indeed, a smaller quantity of hydrates can be formed which results in lower mechanical resistance of the refractory castable at room temperature. Therefore, the dehydration mechanism affects the pre-firing strength of the castable by volumetric change. After the firing steps, larger pores are observed due to evaporation of gaseous water, leading to higher porosity as seen in Table 4. A correlation can be drawn between the crushing strength of the castables containing slag and the apparent porosity values which slightly increase with increasing content of slag. However, the reference with Secar 71 represents high alumina castables which are rarely employed in industries. In operation, 260 MPa exceeds the practical limits. Castables with 2.5 wt.% of slag present an acceptable level of cold crushing strength. The micrographs of the samples containing a combination of slags and cements reveal excellent compatibility between the binder and aggregates after firing at 1500°C. With additional improvements in the compatibility between binder and aggregates, along with the utilization of appropriate additives, improved crushing strength could be achieved.

[IMAGE OMITTED. SEE PDF]

RFDA measurements were performed on three samples to estimate the dynamic modulus of each castable. The results are presented in Figure 13. The reference samples S71_5 and CMA_5 present similar E modulus, close to 135 GPa, and similar shear modulus, close to 60 GPa. As soon as slag $$a$$ is incorporated to the composition, the samples follow a similar evolution for the elastic and shear modulus, consisting on a decrease to 100 GPa and 50 GPa, respectively. The RFDA was shown to be microstructure-dependent and is thus influenced by the effect at the grain boundaries, the pores, and their interaction. The microstructure of the sample Sa2.5_S712.5 exhibits favorable interaction between alumina grains and calcium aluminate cement. Vanadium-bearing phases are observed around the alumina grains and therefore are not well embedded with the coarse grains. These phases influence the interactions between the binder and the aggregates, which explains the drop in the values for the moduli. However, there is no significant deviation between 2.5 and 5 wt.% of slag, indicating that adding more slag does not affect the microstructure in terms of elastic and shear modulus. The dispersion remains consistent regardless of the amount of slag incorporated into the composition. The samples containing 5 wt.% of slag $$a$$ show significantly higher elastic modulus and are thus stiffer than the samples with 5 wt.% of slag $$b$$. However, it is interesting to notice that sample Sb2.5_CMA2.5 is the only one presenting similar elastic and shear moduli as the reference samples while containing slag.

[IMAGE OMITTED. SEE PDF]

Figure 14 shows the cold modulus of rupture (CMOR) of the castables fired at 1500°C for 6 h. The highest flexural strength values are observed for the reference samples S71_5 and CMA_5 (42 MPa and 38 MPa respectively). This observation is in agreement with the results obtained for CCS and RFDA experiments and agrees with the properties of the refractory materials. The variation of CMOR is similar to the variation of CCS. This consistency in the trend aligns with previous experiments, when addition of slag to 5 wt.%, the modulus of rupture value fails from 40 MPa to around 15 MPa for both slags. Samples Sa_5 and Sb_5 exhibit only minimal differences. However, the sample Sb2.5_CMA2.5 presents a similar flexural strength value, 40 MPa. The measurement for this material indicated the highest CMOR value when compared with the other five castables. This translates into the highest sintering efficiency process. It is perceptible that the combination of slag $$b$$ and CMA72 works quite well in terms of strength (MOR; RFDA) and in terms of rheological properties thanks to the presence of spinel in the composition. When the slag $$a$$ is in combination with the Secar71, a drop is observed and even bigger with CMA. When slag replacement occurs, the flexural strength progressively decreases except for Sb2.5_CMA2.5. The material failure is attributed to the growth of defects followed by the addition of slag. On the other hand, the addition of slag $$a$$ leads to less bonding between the grains in the castable. Furthermore, it validates the observation made in Figure 9B when slag is mixed with CMA72, the presence of metallic droplets decreases the quality of the castables.

[IMAGE OMITTED. SEE PDF]

The microstructure observed for sample Sb2.5_CMA2.5 with needle shapes and a well-distributed binder, along with grains effectively embedded in the matrix, could account for the decline in the results.

CONCLUSIONS

High-alumina refractory castables were elaborated with the substitution of the calcium aluminate cement binders by vanadium-bearing slag. The current study investigated the effect of vanadium-bearing slag on the properties of the refractory samples and compared them with conventional alumina refractory castables. It can be concluded that slag $$a$$ and slag $$b$$ present different chemical composition, mineral phases, and microstructure. However, no major deviations were observed in the DSC curves and shrinkage curves when comparing reference cements with vanadium slags. Flowability measurements gave a first prediction for the mechanical behavior of the castables. The castables containing 5 wt.% of slag depict poor flowability and thus do not satisfy the criteria for self-flowability. However, the microstructure and the mechanical properties obtained with the samples containing 2.5 wt.% of slag show the most promising sustainable alternative as a refractory lining. Additional adjustments of the recipe using the Andreasen model and the mechanical results could be used to enhance and optimize the castables' performance and installation success. The hazardous risk related to the presence of vanadium in the slags must be considered but does not represent an issue as the vanadium is not present in free vanadium oxide phase, but is entrapped with other oxides to form a spinel phase. It reduces the risk associated with the disposal and management of vanadium-bearing slag. The substitution of traditional raw materials with industrial by-product offers environmental benefits that must be intensified for a green transition to climate neutrality in the framework of the European Green Deal.

ACKNOWLEDGMENTS

This project has received funding from the European Union's Horizon Europe Research and Innovation Program under grant agreement no. 101072625. CESAREF (Concerted European action on Sustainable Applications of REFractories).

References

Horckmans L, Nielsen P, Dierckx P, Ducastel A. Recycling of refractory bricks used in basic steelmaking: a review. Resour Conserv Recycl. 2019;140:297–304. [DOI: https://dx.doi.org/10.1016/j.resconrec.2018.09.025] ISBN: 978‐0‐08‐100773‐0