Understanding the Structure and Structural Effects on the Properties of Blast Furnace Slag (BFS)

Abstract

Evolving knowledge of the structure and physical properties of metallurgical slags is summarized in current review. Slag structure, compositional effects, role of cations in structural modifications, parameters used to represent the structure, structural analysis techniques and effects of structure on properties of blast furnace slag (BFS) studied in details. The basicity, polymerization (Q) or depolymerization (NBO/T), optical basicity, Qⁿ values, concentrations of bridging O’s (O^o), non-bridging O’s (O⁻) and free O’s (O²⁻) in slag are useful to represent the structure of slag. Methods and techniques utilized to study the slags are also discussed. The BFS is characterized by using X-ray Diffraction and Spectroscopy, Raman Spectroscopy, Fourier Transform Infrared Spectroscopy (FTIR), X-Ray Photoelectron Spectroscopy (XPS) and Nuclear Magnetic Resonance (NMR) Spectroscopy. The physical properties such as surface tension, viscosity, density, thermal expansion and diffusion, electrical conductivity and resistivity of slags are reviewed thoroughly which are heavily dependent on structure of slag. Viscosity is affected by polymerization or depolymerization of slag structure and cation size; electrical resistivity depends on Q, size of cations and number of available cations; thermal expansion depends on Q and cation field strength (i.e. z/r²); thermal conductivity is linked with rigidity of slag network which is also dependent on Q and metal-oxygen (M–O) bond strength. Degree of polymerization or depolymerization of slag structure also effect the surface and interfacial tension, it decreases as metal-oxygen (M–O) bond strengths (i.e. z/r², cation field strength) decrease.

1. Introduction

An important by product of iron and steel casting industry is slag. It is an important component of steel making process. Properties of slag determine the quality of end product. Its importance can be understood by an old adage “Take care of the slag and the metal will take care of itself”.¹⁾ Most abundant and studied slag in the iron and steel making is blast furnace slag (BFS). It contains useful raw materials and also a potential source of waste heat energy. The primary BFS constituents are oxides i.e. SiO₂, CaO, Al₂O₃, MgO, TiO₂, K₂O, Na₂O etc., along these components it also contains some impurities like FeO, MnO, P₂O₅, and S, which generally come from mine limp, sinter, flux material and coke. In BFS, the most abundant components are SiO₂, CaO, Al₂O₃ and MgO. The approximate percentage of slag constituents is shown in Table 1. Now these days, ironmaking has become a costly process due to high global consumption of iron in different industries and increasing cast of electricity. In order to overcome the cast issue, iron producers utilized cheap, low grade raw materials related with ironmaking. Low grade raw materials feed into blast furnace change the composition and properties of slag due to increased amount of aluminum oxides, titanium dioxides, alkali oxides, alkaline earth oxides. Sulphur and Phosphorus are also among the important constituents of slag. The changes in slag composition causes changes in slag properties which as a result have effects on the productivity and the quality of end product i.e. iron.²⁾ A large amount of carbon dioxide is produced during iron production process from blast furnace. CO₂ is a greenhouse gas causing serious environmental issues such as ozone layer depletion and global warming for many years. A significant amount of sensible heat contained in slag goes as a waste heat into environment adding to severity of global warming phenomenon.³⁾ In order to resolve environmental issues connected with blast furnace operation, researchers are recently trying to utilize sensible heat from slag and reduced the carbon dioxide production by utilization of slag into Portland cement material, roadbed material, and concrete aggregate. Although BFS has environmental issues but it preforms some important functions, such as it seals the metal from oxygen and prevents the oxidation during the manufacturing of iron. It also removes the undesirable elements (i.e. Phosphorus and Sulphur), nonmetallic inclusion and works as insulation against heat losses from metal surface.⁴⁾ It plays significant role in the production and refining of iron, so through better understandings of properties of slag at high temperatures not only tends to lower the cost issues, the higher efficiency of ironmaking and provide insight into the kinetics and thermodynamics of slags.⁵⁾ On contrary, the properties are dependent on the structure of slag and structure affect the slag properties, such as viscosity, conductivity, density, etc. Under these circumstances, slag structure has great importance in metallurgical processes.⁶⁾ The silica tetrahedral (SiO₄)⁴⁻ is assumed to be the basic structural unit in silicate and aluminosilicate based slag.⁷⁾ The main factor which affect the structure of slag is the degree of polymerization and depolymerization of the silicate network in slag structure. Also the addition of certain cations (e.g. Ti⁴⁺, Al³⁺) into the silicate slag, nature of network modifying cations (e.g. Mg²⁺, Na⁺)⁸⁾ and physical nature of slag (i.e. crystalline, glassy, or liquid phase),⁹⁾ all these factors have great influence on slag structure. The utilization of diffraction and spectroscopic structural analyses techniques have deepened the knowledge about structure of slag. X-ray, electron, neutron diffraction, FTIR, NMR and Raman spectroscopy are advanced scientific techniques for characterization of slag structure. The recent advances in computer simulation techniques, together with diffraction and spectroscopic analyses, enhanced the understanding on the role of different cations, aluminosilicate and titanosilicates slag structure.¹⁰⁾ Mysen et al.^11,12,13,14) and Seifert et al.¹⁵⁾ studied the structure of silicates in magmas. Their work is helpful to understand the structure of slag. Mills et al.¹⁶⁾ carried out an investigation to find any relationship between slag structure and its physical properties. Waseda et al.^17,18) has reviewed the structural properties of metallurgical slags.

Table 1. Composition range of BFS constituents.^2,95,96)

Constituent	Range %	Constituent	Range %
SiO2	30–40	CaO	34–40
Al2O3	10–25	MgO	05–10
TiO2*	0–2.0	K2O	0–0.7
Na2O	0–0.3	FeO	0.1–1.9
MnO	0.01–1.2	S	1–1.9
P2O5	< 0.1

*  TiO₂ content up to 22–25% in Ti-rich ores.³⁹⁾

The composition of slag formed in ironmaking is not homogeneous and it changes with the composition and quality of raw materials fed into blast furnace. So the nonhomogeneous behavior of slag composition affect the yield and efficiency of the ironmaking processes. These facts make it compulsion to predict the properties of slag because they control the blast furnace operation and the quality of product. So in order to make reliable predictions for properties of slags, more rigorous studies of slag structure, factors affecting the structure, and effects of structural modifications on its properties, are required.⁸⁾ The structure and properties of slag which contains silicates, aluminosilicates, titanosilicates, alkali, and alkaline oxides are difficult to understand and measure at high temperature. Keep focused on silicates, aluminosilicates, titanosilicates, alkali, and alkaline oxides, this review gives the general introduction to the slag structure, compositional effects, role of cations in structural modifications, parameters used to represent the structure, structural analysis techniques and effects of structure on properties of blast furnace slag (BFS).

2. Structural Modification

2.1. Slag Structure

According to structural research on magmas, glasses and slags it is imagine that the structure of BFS is a three dimensional interconnected network in which, silica (Si–O) tetrahedral is a basic building unit where each Si⁴⁺ is bonded with four oxygen (O⁻) ions. The addition of IA or IIA cations (e.g. Na⁺, Ca²⁺) into slag causes Si–O–Si bond cleavage and create ionic (O⁻–Ca²⁺) bonds. The oxygen ion connected ionically with cation is referred to as non-bridging oxygens (NBO or O⁻). The further addition of metals cations will increase the degree of depolymerization of slag structure. Depolymerization of slag network, generates free oxygen ions (FO) represented by O²⁻ which have no bonding with any Si⁴⁺ ions. In Si–O tetrahedral, each oxygen ion is bonded covalently to another Si- tetrahedral (Si–O–Si) known as bridging oxygen (BO or O^o). So the blast furnace slag contains both ionic and covalent bonds. Bridging of oxygen between Si-atoms (Si–O–Si) forms a three dimensional polymerized structure. In short SiO₂ is considered as network former and shows fourfold coordination. Figure 1 represent the bridging, non-bridging and different type of structural units present in BFS. The polymerization of silica network forms different type of structural units, such as (Si₂O₇)⁶⁻, (Si₂O₆)⁴⁻, (Si₂O₅)²⁻, and SiO₂. These anionic units with different numbers of atoms may exist as sheets, rings and chains. Slag properties are dependent on the level of polymerization (or depolymerization) in the slag structure.^1,8,9)

Fig. 1.

Schematic representation of silicate network in blast furnace slag (a) possible anionic structural units, (b) structure with different concentration of cations and alumina. (Online version in color.)

The structure of crystalline and glassy phase is different as compared to slag phase. In crystalline and glassy phase the structure is represented in terms of short, intermediate, and long range order. The crystalline phases show periodic and symmetric arrangement of atoms and within the unit cell all the atoms are equivalent. The slag phase is complicated as compared to crystalline phase. In case of molten slag, atoms are not equivalent and no long range order exist. The molten slag does not exhibit periodicity and disordered slag network affect its physicochemical properties.¹⁹⁾ The BFS contains a variety of constituents, cations and structural units with different degree of polymerization. Factors affecting slag structure and its properties are explained in proceeding paragraphs.

2.2. Effects of Al₂O₃ on Slag Structure

Currently the amount of Al₂O₃ in BF slags is more than 25% by mass, which have effects on geometry and physicochemical properties of some types of slag. The increasing content of Al₂O₃ in slag would change the slag composition to alkali or alkaline aluminosilicate based slag, where the Al³⁺ ions would be in a tetrahedral bonding with oxygen. As the ionic radius of Al³⁺ is 0.47Å and for Si⁴⁺ is 0.34Å, so comparison between ionic radii suggest that Al³⁺ will be suitable for the substitution of Si⁴⁺.²⁾ Addition of Al₂O₃ into the silicate based slag cause changes in silicate network due to absorption of Al³⁺ ions into silicate network, because Al³⁺ ions exhibits fourfold coordination like Si⁴⁺ ions. The charge difference between Al³⁺ and Si⁴⁺ ions, is balanced by adding extra amount of IA or IIA cations in slag, these ions provide electrical charge balance. Presence of Na⁺ ion near the Al³⁺ ions and Ca²⁺ or Mg²⁺ ion sited between two Al³⁺ ions are important to achieve the balance of neutrality within the tetrahedron.^20,21) The lowest valence cations are used for charge balancing. Those cations which are involved in charge balancing cannot act as network modifier.²²⁾ The Al₂O₃ triggers some changes in slag structure by softens/weakens the BO bonds and it also changes the role of cations in slag network. The cations in slag behave as network breaker but the addition of aluminum needs one IA or half IIA cation for Al³⁺ charge compensation. Provision of an electron to one of Al based tetrahedron of oxygen atoms serves as charge compensator. As a result a covalent bond is formed between the central Al and each oxygen atom.¹⁰⁾ Principally additions of Al₂O₃ work as network formers (NWF), which makes four covalent Al–O bonds and show fourfold coordination. In case of presences of large amount of Al₂O₃ in the slag, the Al³⁺ ions can also act as network modifiers (NWM) and exhibit five or six fold coordination with ionic bonds. This is the reason due to which Al₂O₃ is often referred to as an “amphoteric”.¹⁹⁾ The amphoteric behavior of Al₂O₃ is the stoichiometric ratio between Al ion and charge compensating cations.¹⁰⁾ Addition of Al₂O₃ in molten silicate results into formation of Al–O–Al and Si–O–Si bonds or alternatively, Al–O–Si bonds in order to absorb Al³⁺ ion into the silicate network.¹⁹⁾ According to Lowenstein²³⁾ in two tetrahedral bridged by one oxygen bridge, central atom of only one tetrahedral can be Al; another center must be having Si atom, or any small tetravalent ion i.e. P. Similarly, whenever an oxygen atom is in vicinity of two Al ions, one of them must have a coordination number five or six, towards oxygen. According to Al-avoidance (Q) rule, feasibility of Al–O–Si bond formation depends on energy levels. The Al–O–Si bond formation is highly dependent on nature of cations available for charge balance of Al units. According to Lee et al.²⁴⁾ Q is defined as the relative lattice energy differences among the linkages Si–O–Si, Si–O–Al, Al–O–Al, and for perfect Al avoidance Q = 1. Reported best fit for calcium aluminosilicates, is 0.8 ≤ Q ≤ 0.875, and for the sodium aluminosilicates, is 0.93 ≤ Q ≤ 0.99. Low field strength cations (low z/r²) show better agreement with Al-avoidance rule than cations with greater field strength (i.e. high z/r²). The structural investigation of Al₂O₃–SiO₂ and MO–Al₂O₃–SiO₂ systems reveals that Al³⁺ ions has four-fold coordination with tetrahedral structure. In less polymerized slag systems Al₂O₃ addition give rise to less polymerized structural units by depolymerizing the slag network. But for highly polymerized slag, alumina gives preference to Al–O–Si bonds and forms more polymerized structural units. Also in high polymerized slag some fivefold coordination of Al³⁺ and triclusters are exists.²⁵⁾ Tri-clusters consist of three BOs in tetrahedral configuration, which are rearranged around an Al³⁺ ion so as to produce 3-coordinated oxygen. The slag systems having alumina content is greater than IA or IIA metal oxides, the Al³⁺ ions show five or six fold coordination or formed tri-clusters to charge balance the Al ion,²⁶⁾ and alumina in higher coordination depolymerize the slag network. If excessive amount of Al₂O₃, is added into molten state of slag, Al³⁺ does not exists in tetrahedral coordination. It is substituted with Si⁴⁺ and acts as network modifier with octahedral coordination. An important configuration for BFS is SiO₂–Al₂O₃–CaO–MgO system. In this system deficiency of Ca and Mg cations causes the Al³⁺ ions to work as network modifier and depolymerized the slag structure. The Ca²⁺ or Mg²⁺ ion with low concentration preforms the charge balancing of (AlO₄)⁵⁻ tetrahedral instead of Si–O covalent bonds breaker.²⁾ Acidic or basic nature of Al₂O₃ depends upon the amount of alkali oxides. Al₂O₃ behaves as an acidic oxide if enough number of Ca²⁺ and Mg²⁺ cations are available to balance the (AlO₄)⁵⁻. Acidic Al₂O₃ shows tetrahedron coordination when incorporated into the silicate network. In the case of insufficient basic oxides, the behavior of Al³⁺ is same as that of Ca²⁺ or Mg²⁺ in breaking (SiO₄)⁴⁻ network.²⁷⁾ Neuville et al.^28,29,30,31) studied CaO–Al₂O₃–SiO₂ slag system and summarized important findings about composition effects on Qⁿ species, degree of polymerization or depolymerization and Al coordination at high temperature. Kelsey et al.³²⁾ study the Ca, Mg cations mixing in aluminosilicates, also discussed the effects of composition on Al coordination and cation distribution around NBO. Allworth et al.³³⁾ used NMR spectroscopy to evaluate the role of cations in charge compensation and network modification.

The role of Al₂O₃ is as network former in Al₂O₃–SiO₂–CaO slag system although it is an amphoteric oxide. Alumina is a network former in [AlO₄]-tetrahedral state and behave as network modifier oxide in its [AlO₆]-octahedral state. Figure 2(a), shows that variation of Al–O intensity has least effect on the absorption of [SiO₄]⁴⁻ tetrahedral fluctuations. The peak of [AlO₆] is not affected much with increasing absorption of [AlO₄], increases due to increased amount of Al₂O₃ content. The Si–O–Al bending has positive effect on relative absorption intensity of IR band. For these findings it can be concluded that viscosity of slag increases with increasing amount Al₂O₃ content.³⁴⁾

Fig. 2.

FTIR analysis of CaO–SiO₂–Al₂O₃ with different Al₂O₃ percentages at 1773 K³⁴⁾ (a) 5 mass percentage of MgO, 5 mass percentage of FeO slags, with C/S = 1.45 (b) mass% CaO/SiO₂ ratio 1.3.³⁵⁾

The [SiO₄]⁴⁻ tetrahedral has transmittance band between 1170 and 760 cm⁻¹, where NBO/Si have strong influence of ratios 1 and 2, Fig. 2(b). It indicates that increased amount of Al₂O₃ cause the polymerization of the silicate network structures. If Al₂O₃ is increased beyond 10 mass%, FTIR spectra shows a shift in its center of gravity toward wavenumbers from 960 to 800 cm⁻¹, which is an indication of simpler silicate structures with 3 and 4 NBO/Si ratios. The Depolymerization phenomenon of slag network is associated with NBO/Si ratio. Higher NBO/Si ratio (higher Al₂O₃ content) is a sign of depolymerization of slag network where Al₂O₃ act as a network modifying oxide.³⁵⁾

2.3. Effects of TiO₂ on Slag Structure

In Panxi region, China there is a lot of vanadium-titanium-magnetite (VTM), Ti-rich ore forms high titanium slag with varying TiO₂ content from 22 to 25%.³⁹⁾ As the global iron and steel consumption continuously increased, the high quality ores required for the BF operation has been replaced with low quality ores, therefore the slag not only contains high Al₂O₃ but also large amount of TiO₂. Being a basic transition metal oxide, TiO₂ also exists in other ionic states i.e. Ti₂O₃ and TiO.²⁾ The Ti⁴⁺ ion has tendency to substitute the Si⁴⁺ in silicate network. It acts similar to SiO₂ to work as network former by polymerizing the slag structure. But the addition of TiO₂ effect the slag properties e.g. it decreases the slag viscosity,³⁶⁾ which suggests that Ti⁴⁺ ion behaves as network modifier. FTIR and Raman spectroscopic analysis revealed that depolymerization of slag structure depends on TiO₂ where it breaks the silicate network.³⁷⁾ It acts as network modifying oxide and depolymerized the slag structure by breaking the silicate network without affecting aluminate network in slag. According to structural studies of higher titanium slag, the Ti⁴⁺ may have four, five and six fold coordination. In slag system TiO₂–SiO₂ titanium show fourfold coordination but fivefold coordination of Ti occur for concentration greater than 3.6% of TiO₂. In M₂O–TiO₂–SiO₂ system most of the investigations show the both four and fivefold coordination exist but fivefold coordination dominated for higher titanium content. Similarly for MO–TiO₂–SiO₂ system both the four and fivefold coordination have been observed with fourfold exist in both low and high concentration of TiO₂. These studies indicate that larger cations promote the fourfold coordination of Ti⁴⁺ in slag systems.²⁵⁾ Henderson et al.³⁸⁾ suggested that titanium coexist in four and five fold coordination states and it might be the fivefold Ti coordination state which is responsible for the depolymerization of slag network and decrease in slag viscosity with addition of TiO₂. It has been reported that like Al₂O₃, the TiO₂ may be consider as amphoteric oxide.³⁹⁾ The viscosity of the CaO–SiO₂–Al₂O₃–TiO₂ slag system shows an increasing trend due to polymerization of slag network caused by TiO₂.⁴⁰⁾ It is believed that TiO₂ is a weak acidic oxide and it behaves as a network former. Contrary to about discussion, Sohn et al.⁴¹⁾ and Liao et al.⁴²⁾ researches show that viscosity of slag system decreases due to presences of TiO₂. As a result it acts as basic oxide causing depolymerization under neutral conditions in CaO–SiO₂–MgO–TiO₂–Al₂O₃ slag systems with unsteady basicity and varying amount of TiO₂ although the amount of MgO and Al₂O₃ are not identical in their systems. The Ti⁴⁺ ions work as network modifier in silicate slags to break the silicate network as addition of TiO₂ decreases the viscosity and the structure of slag become weak due to Ti–O–Ti bonding with monomers of TiO₄⁴⁻.⁴³⁾ The Ti⁴⁺ cation has large size but small electronegativity (weak field strength of cation, z/r²) as compared with Si⁴⁺ cation. Due to weak field strength, Ti–O bonds are weaker than Si–O bonds. As a result of presence of Ti₄⁺ cations structural network strength of silicate will decline with decreasing viscosity.⁴⁴⁾ On the basis of FTIR and Raman spectroscopy results, Zheng et al.⁴⁵⁾ suggested that, the function of Ti⁴⁺ in silicate slag is as network former and presence of Ti₂O₆⁴⁻ chain units boosts degree of polymerization (DOP) in silicate network. Addition of TiO₂ has negative effect on viscosity due to formation of TiO₄⁴⁻ monomers and weak silicate network, even though DOP of slag structure has increased. Figure 3(a) shows Raman spectra of CaO–SiO₂–Al₂O₃–MgO slags with TiO₂ ranging from 0–10 mass%. The detailed qualitative analysis of the Raman spectra reveled several [SiO₄] - tetrahedral peaks. These tetrahedral peaks showed symmetric stretching vibrations related to NBO/Si of 1, 2, 3, 4 and asymmetric stretching vibrations of NBO/Si 0 near 1040 and 1170 cm⁻¹. Stretching vibrations of Ti–O–Ti are present near 830 cm⁻¹. A semi quantitative value of different silicate species as a function of TiO₂ content can be obtained by combining different segments of Raman spectra. An addition of TiO₂ decreases the sum of the NBO/Si of 1 and 3 and increases the sum of the NBO/Si of 2 and 4. The depolymerization of silicate structure is associated with decrease in the sum of NBO/Si of 1 and 3.³⁷⁾ The [SiO₄] tetrahedra symmetric stretching vibrations, [AlO₄] tetrahedra asymmetric stretching vibrations and Si–O symmetric bending vibrations bands etc. are represented in Fig. 3(b). Addition TiO₂ makes the [SiO₄] stretching vibration less significant which is observed from decrease in depth of the transmittance. We cannot observe any significant difference associated with asymmetric stretching bands of [AlO₄] tetrahedra. Bending vibration band of symmetric Si–O, moves from 600 to 520 cm⁻¹ due to addition of TiO₂. The function of increasing TiO₂ is to depolymerize the slags with is inferred from decreasing depth of the FTIR transmittance with increasing TiO₂. Depolymerization characteristic of TiO₂ suggests that it works as a basic oxide and a network modifier.⁴⁶⁾

Fig. 3.

(a) Analysis of Raman spectra at 1773 K for different TiO₂ contents in CaO–SiO₂–Al₂O₃–MgO–TiO₂ slags at CaO/SiO₂ = 0.8³⁷⁾ (b) FTIR curves of the cooled samples at 1793 K with different amount of TiO₂.⁴⁶⁾

2.4. Role of Cations

As the utilization of low grade raw materials increased into blast furnaces, the sufficient amount of alkali and alkaline earth oxide has occurred in slag. Addition of network modifier oxide (e.g. MgO) into silicates melts, causes metal oxide to form cations (Mⁿ⁺) and anion (O²⁻), where reaction of cations with bridged oxygen of the silicate network, creates non-bridged oxygen. The depolymerization of slag structure is coupled with nature of basic oxides present in silicate melts. Free oxygen O²⁻ reacts with the bridged oxygen O⁰ causing cleavage of already present Si–O–Si bonds. Charge balancing within silicate network structures is achieved by proper distribution of cations. These cations also connect non-bridged oxygen of the various structural units.²⁾

However, considering the field strength, size and bridging with silicate network, cations have a significant role in the structural modification of slag structure. The z/r² (z = charge and r = cation radius) ratio represents strength of bond between a non-bridging oxygen and a cation. Table 2 gives a summary of field strength (z/r²) of different cations. Non-bridging oxygen (NBO) and a cation have an ionic bond with little covalent character.¹⁾ Slag structure is affected by size of cation and amount of electrical charge. High electric charges of the metallic ion make stronger bonds, which results in high melting temperatures of oxides. Furthermore, the coordination number affects the ionic radius of the metal oxide, which depends on slag composition.²⁾ The cations with larger z/r² (i.e. Mg²⁺) have tendency to generate more depolymerized (SiO₄⁴⁻) and polymerized (SiO₂) anionic units. The order of formation of more extreme anionic units can written in terms of z/r² as: K⁺< Na⁺< Li⁺< Ca²⁺<Mg²⁺.⁸⁾ The increasing cation field strength (z/r²) preform some important structural changes in slag, due to increase in cation field, the distribution of polymeric species (Qⁿ) become wider and attraction for NBOs increased. The configurational entropy is also dependent on field strength of cation, higher the field strength, more compact inter-tetrahedral bond angles will result (i.e. Si–O–Si or Si–O–Al), affecting magnitude of disorder which exist in slag network.^25,47) The cation field strength is illustrated in Fig. 4(a). Stebbins⁴⁸⁾ has been reported that the coordination number and the distribution of T–O–T inter-tetrahedral angles are both affected by the field strength (z/r²) of the cations. The cation with higher field strength, promote the AlO₅ and AlO₆ groups as compared to AlO₄. According to Zhang et al.⁴⁹⁾ viscosity of slag is affected by cation size. It has absurd effect on viscosity due to competing influences of interaction and hindrance during viscous flow. His study suggests that percentage of ionic bond of metal oxide may affect the viscous flow. The MgO function as a network-modifying oxide and it has negative effect on melt viscosity of slag up to 20 mass%.⁵⁰⁾ In the presence of sufficient amount of strong basic oxides (i.e. CaO), effect of addition MgO is insignificant if slag network structured has already be depolymerized. In Ca²⁺ ↔ Mg²⁺ substitution, Mg²⁺ with smaller radii tend to play a role in network modification while Ca²⁺ with larger radii perform charge compensation.^32,33) Kim et al.⁵¹⁾ carried out a study to quantify effects of presence of sufficient amount of Na₂O and K₂O on the BFS. The FTIR analysis, shows depolymerization of structural units with Na₂O addition. A blue shift in Si–O–Al bending and AlO₄⁴⁻ tetrahedral bands was observed to presence of Na₂O (i.e. depolymerization of slag network). Addition of K₂O deepened the AlO₄⁴⁻ tetrahedral bands, which means number of AlO₄⁴⁻ tetrahedral increases. The viscosity of the binary Na₂O–SiO₂ melts decreases due to Na₂O additions which results in depolymerization of the silicate network structure.⁵²⁾ Role of different types of oxygen bonding with the silicon cation is also discussed in ref. [52]. These bonding are now known as bridged and non-bridged oxygen. In brief, Na⁺ cation with smaller ionic radius shows network modifying behavior whereas K⁺ with larger ionic radius behaves as charge compensator. Smaller cations with higher field strength (z/r²) tend to have low coordination number and hence result in a higher proportion of NBOs and less, BOs.^25,48) The size of the cation is an important factor and depends on the radius of cation. The viscosity and electrical conductivity of slag are affected by the cation size and magnitude of the ionic bond formed with the NBO, coordination number (N_coord) and movement of cations.⁹⁾ A measure of internal resistance to the motion of silicate structural unit over another is known as viscosity, thus large cations could possibly hinder this movement (larger viscosity). Electrical conductivity involves the transport of cations through the silicate network, so the smaller cations will be more successful than bigger cations in passing through the network and results higher electrical conductivity. The bridging of silicate chain by cations and mixed alkali effect has significant effects on viscosity, electrical conductivity and thermal diffusion.¹⁾ The bridging and effects of cation size on viscosity and electrical conductivity are illustrated Figs. 4(b) 4(c) 4(d). According to Maekawa et al.⁵³⁾ addition of alkali oxides give rise to the Q² and Q³ depolymerized structural unit. Depolymerization of silicate structure is best achieved by sufficient amount of Li₂O comparable to K₂O and Na₂O. It is because, Li⁺ has higher electronegativity compared with Na⁺ and K⁺ and it makes bond with the lowest polymerized Qⁿ species of non-bridged oxygen.

Table 2. Effective ionic radii and corresponding cation field strength.²⁾

Ion	Ionic Radius (nm)	Z/r2 (nm−2)	Ion	Ionic Radius (nm)	Z/r2 (nm−2)
O2−	0.140	102.0	Mg2+	0.072	385.8
K+	0.138	52.5	Al3+	0.053	1068.0
Li+	0.076	173.1	Si4+	0.040	2500.0
Na+	0.102	96.1	Ti4+	0.061	1075.0
Ca2+	0.100	200.0

Fig. 4.

Schematic representation of cations effects on structure of slag, (a) M–O bond strength, z/r² (b) bridging of cations (c) and (d) hindrance caused by cations size in viscosity and electrical conductivity. (Online version in color.)

3. Structural Representation

The structure of blast furnace slag has great importance because it can affect the properties of slag as well as the performance and productivity of furnace. So a variety of different parameters have been used to understand the structure of slag.

3.1. Basicity

It is the ratio between network modifier component and network former component of the slag composition. The SiO₂ in slag behaves as acidic by accepting electrons and alkali or alkaline metal oxides concentrations are denoted as basic because they donate electrons. Thus, network formers are acidic and network breakers are basic in nature. The basicity values will become problematic when dealing with amphoteric oxides like Al₂O₃.¹⁹⁾ A variety of basicity relations have been used to represent slag structure,⁵⁴⁾ some of these are shown below, where X is the mole fraction.   

$$Basicity=X_{CaO}/X_{Si{O}_2}$$(1)

  

$$Basicity=\left(X_{CaO}+X_{MgO}\right)/\left(X_{Si{O}_2}+X_{A{l}_2{O}_3}\right)$$(2)

3.2. Q and NBO/T

A parameter Q is used for slag network to represent its degree of polymerization. The visualization of polymerization is easier than depolymerization, thus preference is given to parameter Q and calculated by:¹⁹⁾   

$$Q=4-\left(NBO/T\right)$$(3)

The ratio of nonbridging oxygen to tetragonally connected ions, is the measure of depolymerization of slag network, which is denoted by NBO/T and where T denotes the ions Si⁴⁺, Ti⁴⁺, or Al³⁺ in tetragonal configurations.^20,21) The NBO/T is like basicity because it is the ratio of available network breaking cations divided by the network forming cations, here available cation means that the total number of cations excluding the charge balancing cations.¹⁹⁾   

$$NBO/T=2\left(\sum X_{MO}+\sum X_{M_{2}O}-X_{A{l}_2{O}_3}\right)/\left(X_{Si{O}_2}+X_{A{l}_2{O}_3}\right)$$(4)

Where X represents mole fraction, MO and M₂O donates the alkali and alkaline metal oxides. For example, MO = CaO and M₂O = Na₂O in CaO–Na₂O–Al₂O₃–SiO₂ slag system. But the parameters NBO/T and Q have been facing a problem that they do not differentiate between different cations e.g. Na⁺ or Ca²⁺ and their effects on slag structure. This problem can be resolved by introducing the concept of optical basicity.⁴⁾

3.3. Optical Basicity (Λ_th)

A degree of measure of electron donor power of different ions relative to that of CaO is known as the optical basicity (Λ_th), i.e. (Λ_th) is the ratio of electron donor power of slag constituents divided by electron donor power of CaO. It was shown that the optical basicity can be calculated from Pauling electronegativity. The link between optical basicity and electronegativity makes the numerical values more understandable.^55,56,57) The optical basicity was introduced to partially resolve the problem of differentiating between different cations and their effects on the structure and then on property estimations.⁵⁴⁾ It is a measure of the depolymerization of the melt and can be express as in Eq. (5);   

$${\mathrm{\Lambda }}_{\text{th}}=\sum X_i{m}_i{\mathrm{\Lambda }}_i/\sum X_i{m}_i$$(5)

Where i = 1,2,3, … represent the oxides, m represents the number of O atoms in oxide, e.g., 1 for CaO and 2 for SiO₂. The symbol Λ_i is correspond to optical basicity value for ith oxide.

The optical basicity parameter has some disadvantages, e.g. it facing uncertainties in the (Λ_th) values for some transition metal oxides and optical basicity does not differentiate between the sizes of cations. The trend in predicted viscosities for different cations is the reverse of that found experimentally.¹⁹⁾

3.4. Concentrations of O^o, O⁻ and O²⁻

The concentrations of bridging O’s (O^o), non-bridging O’s (O⁻) and free O’s (O²⁻) in slag are useful to represent structure of slag. These oxygen ions could be affected by the nature of cations present in slag and temperature can also cause changes in concentrations of O^o, O⁻, and O²⁻.¹⁾ Zhang et al.⁵⁸⁾ suggested a viscosity model based on structure of slag, which represents formation of slag structure through the different types of oxygen ions. In order to describe the structure of slag and effects of oxygen on specific properties, i.e. the increase in non-bridging oxygen (O⁻) content of slag, decreases its viscosity, calculation of three types of oxygen ions are used.⁵⁹⁾ Toop et al.⁶⁰⁾ calculated the equilibrium constant (k) and described the degree of polymerization in terms of bridging, non-bridging and free oxygens.

3.5. Qⁿ Values

The spectroscopic study of slag structure shows the abundance of different structural units present in melt. Usually, Qⁿ represents structural unit of slag, where n denotes number of bridging oxygen (BOs) and degree of polymerization of these units significantly vary with n. A structural unit fully linked with four bridging oxygens is expressed as Q⁴, while a single tetrahedron unit having no BO is represented as Q^o and so on.¹⁹⁾ Masson⁶¹⁾ calculated the ionic distribution as function of silica content for several binary systems and concluded that cation-oxygen interactions can influence the ionic bonding in silicates. Schramm et al.⁶²⁾ studied the Qⁿ species in Li₂O–SiO₂ system by NMR and concluded that Q², Q³, and Q⁴ units exist in binary system but Q⁴ unit contribute to the thermal conductivity and viscosity.

4. Structural Determination

New techniques such as X-ray and neutron diffraction, nuclear magnetic resonance, Raman and XPS spectroscopy played pivotal role to develop knowledge bank about structure of slag in last few decades. Our knowledge about structure of silicates developed rapidly due to precise data from these state of the art technologies which is integrated with structural thermodynamic models and physical property data.⁸⁾ Theoretical attempts such as numerical simulations, (ab initio, molecular dynamics and Reverse Monte Carlo simulations) have also been made to get insights of the possible atomic level structure of slags and their structural analysis. The information from numerical simulation on glasses are useful to study the structure of metallurgical slags. The advantage of these approaches is that they provide access to the melt structure and atomic mobility at temperatures and pressures not reachable with experimental methods.³⁸⁾ The summary of different methods used for structural analysis for slags is shown in Table 3.

Table 3. Various methods provide structural information of slags.¹⁹⁾

Method	Structural information
Neutron, Electron, X-ray Diffraction	Inter-tetrahedral angles T–O–T, Bond length, Coordination number,
Spectroscopy Raman, Infra-red, Ultra-violet	Bond length and angle, Identification and concentration of different anionic species
Nuclear Magnetic Resonance NMR	Bond length and angle, Identification and concentration of different anionic species
X- ray Absorption Spectroscopy XANES, EXAFS	Coordination number of specific atoms or ions, bond length and angle
X-ray Emission Spectroscopy XPS, ESCA	Coordination number of specific atoms or ions, Change in valence shell
Molecular Dynamics MD	Bond strength, T–O–T angles, Coordination (Al is 4-fold in Na2O–Al2O3–SiO2 slags)

4.1. X-ray Diffraction and Spectroscopy

In order to obtain information about the short range and intermediate range order in materials diffraction and spectroscopic methods are used. Most of the information of slag structure such as bond lengths, bond angles, and the coordination orientation and behavior of specific atoms is obtained from X-ray near edge absorption spectroscopy (XANES) and extended X-ray absorption fine structure (EXAFS) using derived values of the radial distribution function (RDF) subjected to Fourier transformation. When X-rays strike a sample, electrons are emitted from it. The energy of emitted electron is determined and the energy levels are utilized to determine both the coordination number and the valence state of atoms.¹⁹⁾ The early use of XRD to get essential structural knowledge of rapidly quenched melts⁶³⁾ of slag was not successful. Development of high temperature XRD, X-ray absorption fine structure (EXAFS) and X-ray scattering (AXS) found to be very useful in structure analysis around the Si element in multi constituent slag systems.⁶⁴⁾ Detailed XRD investigations of binary silicate slags are useful to understand the nature of melt structures.^65,66) Warren et al.⁶⁷⁾ used X-ray radial distribution functions on silicate glasses to determine inter atomic distance and the coordination number of the silicon atoms. Mozzi et al.⁶⁸⁾ used XRD to confirm the presence of SiO₄⁴⁻ unit in silicate glass melts. According to this study, four oxygen atoms were present around each Si atom in tetrahedral shape, with bond length of 1.62Å for Si–O and bond angle of 120° to 180° between Si–O–Si. Li et al.^69,70) used the XANES spectroscopy to investigate the structure of local silicate bonding. The densities of 3s, 3p and 3d unoccupied Si states, length of Si–O, Si–Si bond, coordination number of Si and Si–O–Si angles are analyzed with the help of XANES spectra.

4.2. Raman Spectroscopy

The vibrational, rotational, and other low frequency modes in a slag system are investigated using Raman spectroscopy. The analysis of Raman spectra identified the different structural components of system. Excitation of the sample produces vibrations as well as rotations, with Raman spectroscopy these excitations can be detected in the visible region. The vibrations produce intense and partly polarized Raman lines, or produce weak polarized lines. In crystalline phases the lattice vibrations are dominated by the silicate species and the vibrations produce well defined Raman lines. However, in both quenched glass and liquids, the vibrations produce broad bands which must be deconvoluted into the various polymeric species. Individual bands correspond to individual Q species and the concentrations of these species can be determined by calculating the areas under the individual peaks.¹⁹⁾ In Raman spectroscopy when incoming radiations strike with the molecules, they polarized the target molecules. These molecules re-emit the absorbed energy in the form of various vibrational modes. Raman analysis is useful for quantitative analysis of the various structural units (Q⁰, Q¹, Q², etc.) involved in silicate network structure.²⁾ Typically deconvoluted Raman spectra of BF slag is represented in Figs. 5(a) and 3(a). Raman spectroscopy was used to analyze the effects of TiO₂ on the structure and viscosity of BF slag.³⁷⁾ It was concluded that TiO₂ addition decreases the viscosity by depolymerizing the silicate network of slag. Mysen et al.⁴⁷⁾ used the Raman spectroscopic technique for quantitative analyses and distinction among Q⁰, Q¹, Q², and Q³ units to measure the degree of polymerization of the slag.

Fig. 5.

Deconvoluted (a) Raman⁴³⁾ (b) XPS²⁾ spectra of blast furnace type slag.

4.3. Fourier Transform Infrared Spectroscopy (FTIR)

Fourier transform infrared spectroscopy utilizes infrared spectrum absorption or emission technique to obtained raw data which is further converted into actual spectrum by Fourier Transformation. FTIR technique is normally use for qualitative analysis instead of other analytical techniques because the deconvolution of the spectra is relatively easy. It remains a fundamental tool to identify structural units present in melts. Raman spectroscopy works on the principle of silicates’ polarizability. On the other hand, FTIR utilizes the dipole moment change as working principle.²⁾ The FTIR analysis of different slag systems with varying concentration of composition are shown in Figs. 2 and 3(b). Lee et al.⁷¹⁾ and Kim et al.³⁴⁾ used FTIR to determine the influence of presence of FeO, MgO and Al₂O₃ and basicity on the viscosity of BFS. In addition, together with Raman and FTIR analysis Park et al.,³⁷⁾ concluded that the addition of titanium dioxide depolymerized the blast furnace slag structure.

4.4. X-ray Photoelectron Spectroscopy (XPS)

XPS is a technique used to determine the chemical and physical composition, and electronic state of a constituent element in sample to a depth of 10 nm. In order to obtain XPS spectra, the target materials are exposed to X-rays. The number of electrons coming out of the material surface and their respective energies are recorded. The XPS as compared to Raman spectroscopy or FTIR is a surface material analysis instead of bulk material analysis.

However, it is difficult to determine various NBO/Si ratios using the XPS technique, as a result it become hard to identify type of silicate polymerization because this analysis focuses on the amount of O⁰, O⁻, and O²⁻ in the slag system.²⁾ Figure 5(b) represents deconvoluted XPS spectra of BFS. The bridged and non-bridged oxygen present in BFS can be quantified through detailed analytical analysis of XPS spectra.⁷²⁾ Kim et al.⁷³⁾ studied the effect of CaF₂ and Li₂O on complex silicate structures and concluded that they help to convert these silicate structures into simpler one, by correlating the viscosity data with XPS analysis. Interfacial tension decreases with the increasing number of free oxygen ions in molten slag.⁷⁴⁾ In depth XPS analysis revealed that addition of MgO has no effect on the fraction of free oxygen while it decreases with higher concentration of Al₂O₃. Sohn et al.⁴¹⁾ studied the effect of TiO₂ on the viscosity of calcium silicate melts by utilizing XPS technique and concluded that concentration of bridged oxygen (O^o) decreased while free oxygen (O²⁻) increased on addition of TiO₂.

4.5. Nuclear Magnetic Resonance (NMR) Spectroscopy

Nuclear magnetic resonance is a phenomenon which is observed in nuclei of certain atoms which are placed in a magnetic field to absorb electromagnetic radiations and then exposed to an alternating magnetic field. NMR spectroscopy has been used extensively, for the identification of silicate melt structures.⁷⁵⁾ When a nuclei transition from lower energy state to higher energy state, it absorbs energy by the spin associated with it and vice versa. So absorption or emission of energy from nuclei is a source of NMR signal. The resonance, or the exchange of energy at a specific frequency between the spins and the spectrometer, provides NMR with its sensitivity. On the application of external magnetic field, nearby electrons produce a shielding effect on the nucleus, the magnitude of chemical shift is a measure of shielding effect.²⁾ When a sample is given angular velocity at magic angle relative to applied field, a spectra shown in Figs. 6(a) & 6(b) is obtained showing change in the energy state of the atom or molecule. This energy change is dependent upon magnetic field, internal field of the sample, chemical bond, and molecular motion. The outer electrons shield the nucleus and thus there is a minute change in the resonant energy which results in a chemical shift in NMR absorption peak. The structural information is derived from these chemical shifts. In liquids, sharp peaks are obtained but in solids, interactions cause broadening of the peaks.⁷⁶⁾ Sukenaga et al.⁷⁷⁾ investigated the effects of AI and AII oxides on the viscosity of CaO–SiO₂–Al₂O₃-(R₂O/RO) melts by using the ²⁹Si MAS-NMR and ²⁷Al MAS-NMR spectroscopy.

Fig. 6.

Chemical shift in NMR spectra obtained from (a) 29Si MAS-NMR (b) 27Al MAS-NMR.⁷⁵⁾

4.6. Numerical Simulations

Numerical simulations are also an important tool to study the atomic level structure of slags. The information obtained from simulation is utilized to study the structure of slags. The advantage of these approaches, ab initio and molecular dynamics is that they provide access to the melt structure which is not reachable with experimental methods. Ab initio simulation is an important tool to study the electronic structure and related properties of materials. Ab initio calculations uses the density functional theory (DFT), which is based on Hohenberg-Kohn theorems and Kohn-Sham equation.⁷⁸⁾ Classical Molecular Dynamics (MD) simulation is also a good method which is used to analysis the structure of metallurgical slag at high temperature. The successful MD calculation depends on selection of a suitable potential function. Generally the BMH (Born-Mayer-Huggins) interatomic potentials, which consists on coulombic interaction, repulsion interaction and Vander Waals force, has been used to study the structure of glasses or slags.³⁹⁾

5. Structural Influence on Properties

The metal-slag separation, reaction kinetics, permeability of the intermediate gases used for reduction and heat transfer efficiency are affected by properties of BFS.²⁾ As discussed in previous paragraphs, structure of silicate network strongly affects the properties of slag. The concentration of various oxides, degree of polymerization, nature of the cations presents in slag and temperature dictate the nature of silicate network. The slag properties i.e. viscosity, electrical conductivity, diffusion coefficient, and thermal conductivity etc. are strongly dependent on the degree of polymerization.¹⁹⁾ The way that these various factors affect the properties is discussed in detail below.

5.1. Viscosity

The viscosity of silicate slags is depending upon degree of polymerization of slag structure, cation effects and temperature. Arrhenius or Weymann equations are used to express the relationship between viscosity and temperature, given as follow:   

$$\eta =A_{\eta }exp\left(E_{\eta }/RT\right)$$(6)

  

$$ln\eta =ln{A}_{\eta }+E_{\eta }/RT$$(7)

Where A_η is a pre-exponential term (constant) and E_η is the activation energy.¹⁾

The viscosity of BF slag is probably the most valuable property in industrial high temperature processes. The degree of polymerization has positive effect on viscosity of slag which means that viscosity increases with increasing degree of polymerization of slag structure and vice versa. The viscosity also increases with increasing concentrations of network formers oxides, decreasing concentrations of both network breakers and fluxes. Na⁺ or Ca²⁺ ion is utilized for charge balancing of Al³⁺ in aluminosilicates, which has significant effects on viscosity because the cations which are used for charge balancing cannot work as a network modifier.¹⁹⁾ The viscosity exhibits a small increase with increasing cation size (K⁺ > Na⁺ > Li⁺) for both liquid and super cooled liquid phases but the hierarchical trend is less obvious for MO silicates (where M = Mg, Ca,).¹⁾ In order to understand the phenomenon of viscous flow, it is necessary to explore the relationship between viscosity and structure of silicates. Reformation of bridging oxygen bonds is essential for viscous flow. The energy of Si–O bonds and activation energy of silica rich melts having viscous flow are comparable which is consistent with reformation of bridging oxygen bonds phenomenon.^79,80) As a result, the viscosity might be dependent only on the excess of fully polymerized structural units i.e. Q⁴ at a given temperature. So in binary alkali silicate melts viscosity might positively correlated with the amount of Q⁴ units.⁸¹⁾ As shown in Fig. 7(c), viscosity (ln η) of the silicate melts increases linearly as ln (Q³/Q²) increases at 1873 K and 1773 K. The value of ln (Q³/Q²), can be measured and it can work as a vital polymerization index to determine the relationship between silicate structure and viscosity of the melts.⁶⁾ Figure 7(b) represented the composition dependences of viscosity for SiO₂–Li₂O slag system at 1473 K and 1573 K, the viscosity increases with increasing amount of SiO₂. Due to increasing structural depolymerization NBO/T ration also increases and corresponding decrease in viscosity is observed in Fig. 7(a). The decrease in viscosity of slag at NBO/T = 3.5–4.0 was observed to be independent of NBO/T ration which suggest that structural change in fully depolymerized melts is zero.¹⁰⁾ Addition of Na₂O into BFS having Al₂O₃ in it causes the viscosity to decline, which obey the rule followed by the basic oxide. However, behavior of K₂O is contrary its basic oxide nature.²⁾ Kim et al.⁵⁰⁾ found that a slag system having 20 mass% Al₂O₃ has a decreasing trend in viscosity with decreasing amount of Na₂O. This specific behavior of slag is due to supply of network breaking O²⁻ and its reaction with O⁻. In addition, as Al₂O₃ present in a fourfold coordinated tetrahedral configuration, the Na⁺ cations provide ionic charge balance for Al³⁺ and its surrounding O²⁻ to maintain charge neutrality of slag units.^82,83)

Fig. 7.

Effects of (a) depolymerization (NBO/T)¹⁰⁾ (b) composition (X_SiO2)⁸⁸⁾ (c) polymerization (Q³/Q²)⁶⁾ on viscosity of slag.

5.2. Density

Density of silicate slag is an important parameter to quantify the amount inclusions removed from the metal by flotation. It is also helpful in conversion of thermal diffusivity values to thermal conductivities. The molar volume (V_m) is related to the density via equation ρ = M/V_m and is only slightly affected by the polymerization of the slag. The oxides with higher molecular weight (M) tend to have lower molar volumes (V_K2O < V_Na2O < V_Li2O) and hence higher densities.¹⁹⁾ The molar volume of a binary slag system is affected by cations at 1400°C and relationship of density with polymerization of CaO–MgO–SiO₂ melt at 1500°C are shown in Fig. 8(a) & 8(b). Density is an important property to explain structure of melts. The density of melts has a linear decrease with increasing ln (Q³/Q²) at a given temperature, i.e. increasing polymerization of structure tends to decrease the density of melt.⁶⁾ According to Bottinga et al.⁸⁴⁾ polymerization has positive effect on partial molar volume of silicate melts or the partial molar volume of the bridging oxygen is larger than non-bridging oxygens. As a result, density decreases with structural polymerization. It has been observed that in silicate systems that the formation of a network structure causes an apparent increase in density (i.e. lower molar volume) with higher amount of SiO₂ (or Q). So the molar volume of SiO₂ is a variable function of mass fraction of silicates (X_SiO2).⁸⁵⁾ The density of alumino-silicates decreases with the addition of Al₂O₃ (i.e. higher molar volume) which can be represented by a polynomial equation. These deviations in molar volume reflect the changes in bonding resulting from the polymerization of the silicate units and in an alternative approach the molar volume changes were correlated with the enthalpy of mixing for the silicate system.⁸⁶⁾

Fig. 8.

Effects of (a) cations mole%⁸⁴⁾ (b) polymerization (Q³/Q²)⁶⁾ on molar volume and density of slag, (c) temperature dependences of cations diffusion into silicate network.⁸⁷⁾

5.3. Thermal Expansion and Diffusion

Thermal vibration causes a thermal expansion in slags when temperature of a slag is increased. These asymmetric vibrations are affected by the chemical bonding in the slag and the thermal history (cooling rate) of the slag. The thermal expansion of pure SiO₂ is very low because of the three dimensional silicate network and resultant symmetry of thermal vibrations but thermal expansion in liquid slags increases as cations are added. The degree of polymerization (Q) and cation field strength (z/r²) are used to express the thermal expansion of liquid slags.¹⁾ Diffusion is the process by which one of the ions in the slag move from point to point within the slag network. The diffusing ions are thought to move by a series of jumps from one vacant position to another. Diffusion process is always from directed from high concentration to lower concentration. Increasing temperature causes the “higher amplitude excitation” of the ions in the melt which, in turn, will encourage the movement of the ions into other vacant positions. The silicate network can be regarded as a hindrance to the diffusion process. This hindrance to the movement of ions in the case of the diffusion coefficient is equivalent to those for fluidity (i.e., reciprocal viscosity) and electrical conductivity. There are several factors that affect the diffusion in slags and glasses, such as degree of polymerization of silicate, size of the diffusing species, number of cations and temperature. Thus, the diffusion coefficient can be expressed in the form of an Arrhenius relation.¹⁹⁾   

$$D=A_{D}exp\left(B_D/T\right)$$(8)

Zhang et al.⁸⁷⁾ suggested that there were two factors affecting the diffusion process, first is the detachment (breaking of M–O bond) of the diffusing ion and second is the subsequent movement of ion through the melt. In polymerized melts there is more free space available for cations to move and thus detachment (i.e., bond strength, z/r²) is the more important process and the diffusion coefficient D values are in order Ca > Mg. In contrast, for depolymerized melts there is less free space and thus D is largely determined by cation size. Figure 8(c) illustrated the diffusion of alkali earth cations into silicate melt.

5.4. Electrical Conductivity and Resistivity

The electrical conductivity (κ) is the ability to transport electrons, provided by the cations under the influence of an applied electrical field. The electrical resistivity (R) is the reciprocal of the conductivity (R = 1/κ). The specific electrical conductivity (κ) is the conductivity of a 1 meter cube of the sample and is given by the relation:¹⁾   

$$\kappa =F{\mathrm{\Sigma }}_i{c}_i{z}_i{u}_i$$(9)

Where F = Faraday constant, c_i, z_i and u_i are the concentration, charge and mobility of the ionic species. It is obvious from Eq. (9) that the conductivity of liquid slags has increasing behavior with increasing cation concentration by adding alkali or alkaline oxides. The polymerized silicate network and the size of the cations hinder the mobility of cations. Arrhenius equations are utilized to show the temperature dependence of the electrical conductivity (κ) and resistivity (R) for a liquid slag:¹⁾   

$$\kappa =A_{\kappa }\mathrm{\hspace{0.17em} }exp\left(-B_{\kappa }/T\right),\mathrm{\hspace{0.17em} } ln\kappa =ln{A}_{\kappa }-\left(B_{\kappa }/T\right)$$(10)

  

$$R=A_R\mathrm{\hspace{0.17em} }exp\left(B_R/T\right),\mathrm{\hspace{0.17em} } lnR=ln{A}_R+\left(B_R/T\right)$$(11)

Where A_κ and A_R = constant, B_κ = E_κ⁄R, B_R = E_R⁄R where E denotes Activation energy and R stands for Gas constant (R = 8.314 JK⁻¹ mol⁻¹).

In slags containing two or more cations, the resistivity will be affected by which cations carry out the charge balancing duties. It is assumed that charge balancing is achieved by the largest cations, which has M–O bond strength (lowest z/r²).^16,88) The concentration and mobility of electrons are affected by the number of cations available and the size of cations, respectively. It was pointed out that when Na₂O replaces CaO in a slag there are 2Na⁺ cations for every Ca²⁺ and hence the number of available cations (N) increases. Thus, the resistivity of MO–SiO₂ slags is higher than those for the equivalent M₂O–SiO₂ slags. It should also be noted that when the Al³⁺ is introduced into the Si⁴⁺ network, a cation is needed to charge balance the Al³⁺ ion, it is suggested that cations on charge balancing duties cannot be used as charge carriers. Thus, addition of Al₂O₃ in slags brings about a decrease in the number of available cations.¹⁹⁾ According to Bockris et al.^89,90) network modifying cations contain an electric current, (Mⁿ⁺) which shows partial ionic nature of silicate melts with anions size larger than cations. As a result, the degree of polymerization strongly affects the electrical conductivity of molten silicates. The electrical conduction mechanism in IA or IIA earth oxides, silicate binaries as the network modifying cations being sole charge carriers while silicate polyanions provide conduction path by depolymerization process. The effects of polymerization and depolymerization of silicate network on the electrical conductivity for different systems are illustrated in Figs. 9(a) & 9(b).

Fig. 9.

Effects of (a) polymerization (Q³/Q²)⁶⁾ (b) depolymerization (NBO/T),¹⁰⁾ on the electrical conductivity of slag (c) dependences of surface tension of slag on cation field strength z/r² at 1773 K.¹⁾ {● for alkali MO and ▲ for alkaline earth M₂O metal oxides}.

5.5. Interfacial and Surface Tension

Interfacial and surface tensions are important properties of BF slags because both properties affect many processes between metal and slag. Surface tension (γ) and interfacial tension (γ_msl) differ from other properties in that sense they are surface properties and not bulk properties. The surface tensions of both metals and slags are dependent upon the concentrations of surfactants present. Surfactants tend to be materials with low surface tension, principal surfactants in slags are B₂O₃, K₂O and Na₂O and CaF₂. The gradients in surface tension give rise to Marangoni flows which affect the kinetics of metal-slag reactions. These surface tension gradients caused by variations in temperature, composition and electrical potential.¹⁹⁾ The interfacial tension is also important which affects the removal of inclusions from the liquid metal and mixing of slag in metal or metal in slag.⁹¹⁾ The interfacial tension has been reported to increase as Al₂O₃ content increased or the contents of Na₂O, CaF₂, FeO, and MnO decreased in the slag.⁹²⁾ The equation used to express interfacial tension is given as follow:   

$${\gamma }_{msl}={\gamma }_m+{\gamma }_{sl}-2\phi {\left({\gamma }_m{\gamma }_{sl}\right)}^{0.5}$$(12)

Where γ_m and γ_sl represent surface tensions of the metal and slag phases, respectively and φ represents interaction coefficient.¹⁹⁾

The interfacial phenomena such as slag-metal, metal-refractory, and slag-refractory interfaces in iron and steel making processes are highly affected by surface tension. Surface tension of network modifying monovalent or divalent cations increases with increasing ionization potential. Higher ionization potential has negative effect on surface tension for trivalent cations network. Surface tension has an increasing trend when the ionic interaction between network modifying cations and oxygen is strong. When strength of covalent bond between oxygen and network forming cations increases, the strength of ionic bond at surface of slag decreases, resulting in polymerization of the bulk liquid.¹⁰⁾ Butler⁹³⁾ used Gibbs’ equilibrium surface model to calculate surface tension by utilizing concept of logarithmic difference and surface activities of a bulk of a component. Higher the basicity (i.e. structural depolymerization), higher will be surface tension. Boni et al.⁹⁴⁾ observed that the surface tension increases with decreasing amount of SiO₂ content due to higher number of bonds at the surface layer, while in case of those systems which contain K₂O have reverse trend because K₂O is listed in surface active components having small absolute surface tension. The surface tension is also affected by depolymerization of slag. When the silicate flow unit changes from chain like units to monomer or dimers at NBO/T=2.0, change in surface tension is considerably decrease. Figure 9(c) shows the effect of M–O bond strength on the surface tension of silicate slag. However, the presence of surfactants tends to cause significant reduction of the overall surface tension. The surface tensions of silicate based BFS which contain alkali metal oxides are significantly higher than those having alkaline earth oxides and surface tensions decrease as M–O bond strengths (z/r², cation field strength) decrease.¹⁾

6. Conclusions

In this review the slag structure, compositional effects, role of cations in structural modifications, parameters used to represent the structure, structural analysis techniques and effects of structure on properties of blast furnace slag (BFS) studied in details. The conclusions drawn from this study are as follow:

(1) The Blast Furnace Slag (BFS) is a three dimensional interconnected network in which, silica (Si–O) tetrahedral is a basic building unit where each Si⁴⁺ is bonded with four oxygen (O⁻) ions. BFS contains both ionic and covalent bonds. SiO₂ is considered as network former and shows four folded coordination.

(2) In crystalline and glassy phase the structure is represented in terms of short, intermediate, and long range order. The crystalline phases show periodic and symmetric arrangement of atoms and within the unit cell all the atoms are equivalent. The slag phase is complicated as compared to crystalline phase. In case of molten slag, atoms are not equivalent and no long range order exists.

(3) Increasing content of Al₂O₃ in slag would change the slag composition to alkali or alkaline aluminosilicate based slag, where the Al³⁺ ions would be in a tetrahedral bonding with oxygen. Additions of Al₂O₃ work as network formers (NWF), which makes four covalent Al–O bonds and show four-fold coordination. Al³⁺ ions can also act as network modifiers (NWM) and exhibit five or six fold coordination with ionic bonds. The Ti⁴⁺ ion has tendency to substitute the Si⁴⁺ in silicate network. It acts similar to SiO₂ to work as network former by polymerizing the slag structure. But the addition of TiO₂ effect the slag properties e.g. it decreases the slag viscosity, which suggests that Ti⁴⁺ ion behaves as network modifier.

(4) Cations have a significant role in the structural modification of slag structure. Slag structure is affected by size of cation and amount of electrical charge. High electrically charge ion make stronger bonds, which results in high melting temperatures of oxides. The depolymerization of silicate structure is best achieved by sufficient amount of Li₂O comparable to K₂O and Na₂O. It is because, Li⁺ has higher electronegativity compared with Na⁺ and K⁺ and it makes bond with the lowest polymerized Qⁿ species of non-bridged oxygen.

(5) Different parameters have strong effect on structure and properties of slag. Basicity is the ratio between network modifier component and network former component of the slag composition. A parameter Q is used for slag network to represent its degree of polymerization. The visualization of polymerization is easier than depolymerization (NBO/T), thus preference is given to parameter Q. A degree of measure of electron donor power of different ions relative to that of CaO is known as the optical basicity (Λ), i.e. Λ is the ratio of electron donor power of slag constituents divided by electron donor power of CaO. The concentrations of bridging O’s (O^o), non-bridging O’s (O⁻) and free O’s (O²⁻) in slag are useful to represent structure of slag. Qⁿ values represents structural unit of slag, where n denotes number of bridging oxygen (BOs) and degree of polymerization of these units significantly vary with n. A structural unit fully linked with four bridging oxygen is expressed as Q⁴, while a single tetrahedron unit having no BO is represented as Q^o.

(6) New techniques such as X-ray and neutron diffraction, nuclear magnetic resonance, Raman and XPS spectroscopy played pivotal role to develop knowledge bank about structure of slag. Most of the information of slag structure such as bond lengths, bond angles, and the coordination orientation and behavior of specific atoms is obtained from X-ray near edge absorption spectroscopy (XANES) and extended X-ray absorption fine structure (EXAFS) using derived values of the radial distribution function (RDF) subjected to Fourier transformation. The vibrational, rotational, and other low frequency modes in a slag system are investigated using Raman spectroscopy. FTIR technique is normally use for qualitative analysis instead of other analytical techniques because the deconvolution of the spectra is relatively easy. It remains a fundamental tool to identify the structural units present in melts. XPS is a technique used to determine the chemical and physical composition, and electronic state of a constituent element in sample to a depth of 10 nm. The XPS as compared to Raman spectroscopy or FTIR is a surface material analysis instead of bulk material analysis.

(7) Viscosity, density, thermal expansion and diffusion, electrical conductivity and resistivity, interfacial and surface tension are heavily dependent on structure of slag. Viscosity is affected by polymerization (Q) or depolymerization of slag structure and cation size; electrical resistivity depends on Q, size of cations and number of available cations; thermal expansion depends on Q and M–O bond strength (i.e. z/r²); thermal conductivity is linked with rigidity of slag network which is also dependent on Q and M–O. The surface tension is also affected by depolymerization of slag. The surface tensions of BFS containing alkali metal oxides are significantly higher than those having alkaline earth oxides and surface tensions decrease as metal-oxygen (M–O) bond strengths (i.e. z/r², cation field strength) decrease.

Acknowledgements

The authors wish to express their thanks to the National Natural Science Foundation of China (Grant No. 51234010) and the Fundamental Research Funds for the Central Universities (Project No. 2018CDXYCL0018) for the financial support of this research.

References

• 1)   K. C.  Mills,  L.  Yuan,  Z.  Li,  G. H.  Zhang and  K. C.  Chou: High Temp. Mater. Process., 31 (2012), 301.
• 2)   I.  Sohn and  D. J.  Min: Steel Res. Int., 83 (2012), 611.
• 3)   Y.  Kang,  K.  Nomura,  K.  Tokumitsu,  H.  Tobo and  K.  Morita: Metall. Mater. Trans. B, 43 (2012), 1420.
• 4)   K.  Mills: KenMills-Content.doc, Southern African Pyrometallurgy, Misty Hills, (2011), https://www.pyro.co.za/KenMills/KenMills.pdf, (accessed 2019-05-22).
• 5)   L.  Gan and  C. B.  Lai: Metall. Mater. Trans. B, 45 (2014), 875.
• 6)   J. H.  Park: Metall. Mater. Trans. B, 44 (2013), 938.
• 7)   C.  Han,  M.  Chen,  W. D.  Zhang,  Z. X.  Zhao,  T.  Evans and  B. J.  Zhao: Metall. Mater. Trans. B, 47 (2016), 2861.
• 8)   K. C.  Mills: ISIJ Int., 33 (1993), 148.
• 9)   K. C.  Mills: ISIJ Int., 56 (2016), 1.
• 10)   D. J.  Min and  F.  Tsukihashi: Met. Mater. Int., 23 (2017), 1.
• 11)   B.  Mysen: Eur. J. Mineral., 15 (2003), 781.
• 12)   B.  Mysen: Eur. J. Mineral., 7 (1995), 745.
• 13)   B. O.  Mysen and  P.  Richet: Silicate Glasses and Melts: Properties and Structure, Elsevier, Amsterdam, (2005), 199.
• 14)   B. O.  Mysen,  D.  Virgo and  C. M.  Scarfe: Am. Mineral., 65 (1980), 690.
• 15)   F. A.  Seifert,  B. O.  Mysen and  D.  Virgo: Am. Mineral., 67 (1982), 696.
• 16)   K. C.  Mills and  B. J.  Keene: Int. Mater. Rev., 32 (1987), 1.
• 17)   Y.  Waseda and  J. M.  Toguri: The Structure and Properties of Oxide Melts; Application of Basic Science to Metallurgical Processing, World Scientific, Singapore, (1998), 67.
• 18)   Y.  Waseda and  J. M.  Toguri: Metall. Trans. B, 8 (1977), 563.
• 19)   K. C.  Mills,  M.  Hayashi,  L.  Wang and  T.  Watanabe: Treatise on Process Metallurgy, Vol. 1, Elsevier, Amsterdam, (2014), 149.
• 20)   B. O.  Mysen: Structure and Properties of Silicate Melts, Elsevier, Amsterdam, (1988), 354.
• 21)   B. O.  Mysen: Earth Sci. Rev., 27 (1990), 281.
• 22)   B. N.  Roy and  A.  Navrotsky: J. Am. Ceram. Soc., 67 (1984), 606.
• 23)   W.  Loewenstein: Am. Mineral., 39 (1954), 92.
• 24)   S. K.  Lee and  J. F.  Stebbins: Am. Mineral., 84 (1999), 937.
• 25)   G. S.  Henderson: Can. Mineral., 43 (2005), 1921.
• 26)   D. M.  Zirl and  S. H.  Garofalini: J. Am. Ceram. Soc., 73 (1990), 2848.
• 27)   J. N.  Tiwary,  S.  Sarkar,  B.  Mishra and  U. K.  Mohanty: Emerg. Mater. Res., 2 (2013), 152.
• 28)   D.  Neuville,  L.  Cormier,  D. d.  Ligny,  J.  Roux,  A.  Flank and  P.  Lagarde: Am. Mineral., 93 (2008), 228.
• 29)   D. R.  Neuville,  L.  Cormier,  A. M.  Flank,  V.  Briois and  D.  Massiot: Chem. Geol., 213 (2004), 153.
• 30)   D. R.  Neuville,  L.  Cormier and  D.  Massiot: Chem. Geol., 229 (2006), 173.
• 31)   D. R.  Neuville,  L.  Cormier and  D.  Massiot: Geochim. Cosmochim. Acta, 68 (2004), 5071.
• 32)   K. E.  Kelsey,  J. R.  Allwardt and  J. F.  Stebbins: J. Non-Cryst. Solids, 354 (2008), 4644.
• 33)   J. R.  Allworth and  J. F.  Stebbins: Am. Mineral., 89 (2004), 777.
• 34)   J. R.  Kim,  Y. S.  Lee,  D. J.  Min,  S. M.  Jung and  S. H.  Yi: ISIJ Int., 44 (2004), 1291.
• 35)   J. H.  Park,  D. J.  Min and  H. S.  Song: Metall. Mater. Trans. B, 35 (2004), 269.
• 36)   G. H.  Wen,  S.  Sridhar,  P.  Tang,  X.  Qi and  Y. Q.  Liu: ISIJ Int., 47 (2007), 1117.
• 37)   H.  Park,  J. Y.  Park,  G. H.  Kim and  I.  Sohn: Steel Res. Int., 83 (2012), 150.
• 38)   G. S.  Henderson,  G.  Calas and  J. F.  Stebbins: Elements, 2 (2006), 269.
• 39)   S.  Zhang,  X.  Zhang,  C.  Bai,  L.  Wen and  X.  Lv: ISIJ Int., 53 (2013), 1131.
• 40)   A.  Ohno and  H.  Ross: Can. Metall. Q., 2 (2013), 259.
• 41)   I.  Sohn,  W.  Wang,  H.  Matsuura,  F.  Tsukihashi and  D. J.  Min: ISIJ Int., 52 (2012), 158.
• 42)   J.  Liao,  J.  Li,  X.  Wang and  Z.  Zhang: Ironmaking Steelmaking, 39 (2012), 133.
• 43)   Z.  Wang,  Q.  Shu and  K.  Chou: Steel Res. Int., 84 (2013), 766.
• 44)   L.  Zhang and  S.  Jahanshahi: 7th Int. Conf. on Molten Slags, Fluxes and Salts, SAIMM, Johannesburg, (2004), 51.
• 45)   K.  Zheng,  Z.  Zhang,  L.  Liu and  X.  Wang: Metall. Mater. Trans. B, 45 (2014), 1389.
• 46)   K. X.  Jiao,  J. L.  Zhang,  Z. Y.  Wang,  C. L.  Chen and  Y. X.  Liu: Steel Res. Int., 88 (2017), 475.
• 47)   B. O.  Mysen and  P.  Richet: Silicate Glasses and Melts: Properties and Structure, Elsevier, Amsterdam, (2005), 85.
• 48)   J. F.  Stebbins: Solid State Ion., 112 (1998), 137.
• 49)   G. H.  Zhang,  K. C.  Chou,  G.  Xue and  K. C.  Mills: Metall. Mater. Trans. B, 43 (2012), 64.
• 50)   H.  Kim,  W.  Kim,  J.  Park and  D.  Min: Steel Res. Int., 81 (2010), 17.
• 51)   W.  Kim,  I.  Sohn and  D.  Min: Steel Res. Int., 81 (2010), 735.
• 52)   B. E.  Warren and  J.  Biscoe: J. Am. Ceram. Soc., 21 (1938), 259.
• 53)   M.  Maekawa,  T.  Maekawa,  K.  Kawamura and  T.  Yokokawa: J. Non-Cryst. Solids, 127 (1991), 53.
• 54)   Slag  Atlas,  Verein Deutscher  Eisenhüttenleute,  Verlag  Stahleisen, Düsseldorf, (1995), 9.
• 55)   J.  Duffy and  M. D.  Ingram: J. Non-Cryst. Solids, 21 (1976), 373.
• 56)   J.  Duffy and  M. D.  Ingram: J. Inorg. Nucl. Chem., 37 (1975), 1203.
• 57)   T.  Nakamura,  Y.  Ueda and  J. M.  Toguri: J. Jpn. Inst. Met., 50 (1986), 456 (in Japanese).
• 58)   G. H.  Zhang,  K. C.  Chou and  K. C.  Mills: ISIJ Int., 52 (2012), 355.
• 59)   G. H.  Zhang and  K. C.  Chou: J. Min. Metall. B, 48 (2012), 1.
• 60)   G. W.  Toop and  C. S.  Samis: Can. Metall. Q., 1 (1962), 129.
• 61)   C. R.  Masson: Proc. R. Soc. Lond. A, 287 (1965), 201.
• 62)   C. M.  Schramm,  B.  De Jong and  V. E.  Parziale: J. Am. Chem. Soc., 106 (1984), 4396.
• 63)   R. W.  Wyckoff and  G. W.  Morey: J. Soc. Glass Technol., 9 (1925), 265.
• 64)   Y.  Waseda and  J. M.  Toguri: The Structure and Properties of Oxide Melts; Application of Basic Science to Metallurgical Processing, World Scientific, Singapore, (1998), 15.
• 65)   Y.  Waseda and  J. M.  Toguri: Mater. Trans. B, 8 (1977), 563.
• 66)   Y.  Waseda and  J. M.  Toguri: Dynamic Processes of Material Transport and Transformation in the Earth’s Interior, ed. by F. Marumo, Terra Science Publishing, Tokyo, (1990), 37.
• 67)   B. E.  Warren,  H.  Krutter and  O.  Morningstar: J. Am. Ceram. Soc., 19 (1936), 202.
• 68)   R.  Mozzi and  B.  Warren: J. Appl. Crystallogr., 2 (1969), 164.
• 69)   D.  Li,  G.  Bancroft,  M.  Fleet,  X.  Feng and  Y.  Pan: Am. Mineral., 80 (1995), 432.
• 70)   D.  Li,  G.  Bancroft,  M.  Kasrai,  M.  Fleet,  X.  Feng,  K.  Tan and  B.  Yang: Solid State Commun., 87 (1993), 613.
• 71)   Y. S.  Lee,  D. J.  Min,  S. M.  Jung and  S. H.  Yi: ISIJ Int., 44 (2004), 1283.
• 72)   J. H.  Park and  P. C. H.  Rhee: J. Non-Cryst. Solids, 282 (2001), 7.
• 73)   H.  Kim and  I.  Sohn: ISIJ Int., 51 (2011), 1.
• 74)   E. J.  Jung and  D. J.  Min: Steel Res. Int., 83 (2012), 705.
• 75)   M. J.  Duer: Introduction to Solid State NMR Spectroscopy, Blackwell Publishing, Oxford, UK, (2004), 72.
• 76)   K. A.  Smith,  R. J.  Kirkpatrick,  E.  Oldfield and  D. M.  Henderson: Am. Mineral., 68 (1983), 1206.
• 77)   S.  Sukenaga,  N.  Saito,  K.  Kawakami and  K.  Nakashima: ISIJ Int., 46 (2006), 352.
• 78)   J. D. C.  McConnell,  A.  De Vita,  S. D.  Kenny and  V.  Heine: Phys. Chem. Miner., 25 (1997), 15.
• 79)  J. O’M. Bockris, J. D. Mackenzie and J. A. Kitchener: Trans. Faraday Soc., 51 (1955), 1734.
• 80)  J. O’M. Bockris and D. C. Lowe: Proc. R. Soc. Lond. A, 226 (1954), 423.
• 81)   B. O.  Mysen: Phys. Earth Planet. Inter., 107 (1998), 23.
• 82)   H.  Maekawa and  T.  Yokokawa: Geochim. Cosmochim. Acta, 61 (1997), 2569.
• 83)   B. O.  Mysen: Contrib. Mineral. Petrol., 127 (1997), 104.
• 84)   Y.  Bottinga and  P.  Richet: Geochim. Cosmochim. Acta, 59 (1995), 2725.
• 85)   V. K.  Gupta,  S. P.  Sinha and  B.  Raj: Steel India, 21 (1998), 22.
• 86)   M.  Persson: Ph.D. thesis, KTH, Stockholm, (2007), http://www.diva-portal.org/smash/get/diva2:12029/FULLTEXT01.pdf, (accessed 2017-12-20).
• 87)   Y.  Zhang,  H.  Ni and  Y.  Chen: Rev. Mineral. Geochem., 72 (2010), 311.
• 88)   G. H.  Zhang and  K. C.  Chou: J. Min. Metall. B, 48 (2012), 433.
• 89)  J. O’M. Bockris, J. A. Kitchener, S. Ignatowicz and J. W. Tomlinson: Trans. Faraday Soc., 48 (1952), 75.
• 90)  J. O’M. Bockris, J. A. Kitchener and A. E. Davies: Trans. Faraday Soc., 48 (1952), 536.
• 91)   S.  Feldbauer,  I.  Jimbo,  A.  Sharan,  J.  Harman and  A. W.  Cramb: Proc. 78th Steelmaking Conf., ISS, Warrendale, PA, (1995), 655.
• 92)   A. W.  Cramb and  I.  Jimbo: Steel Res. Int., 60 (1989), 157.
• 93)   J.  Butler: Proc. R. Soc. Lond. A, 135 (1932), 348.
• 94)   R.  Boni and  G.  Derge: Trans. AIME, 206 (1956), 53.
• 95)  Federal Highway Administration: Report, Blast furnace slag – Material description, https://www.fhwa.dot.gov/publications/research/infrastructure/structures/97148/bfs1.cfm, (accessed 2019-05-22).
• 96)  Nippon Slag Association: About iron and steel slag, Chemical characteristic of iron and steel slag, http://www.slg.jp/e/slag/character.html, (accessed 2019-05-22).