Preparation and Properties of Fayalite (Fe2SiO4) Synthetic Copper Slags with Fe/SiO2 Ratios from 1.5 to 2.3

In this work, three types of fayalite slags were prepared with different Fe/SiO₂ ratios of targeted values between 1 and 2.5 which encompass most of the copper slags generated during the converting process in a typical Pierce-Smiths converter.^[1]

The following lab-grade chemicals were used for the preparation of synthetic slags: SiO₂ (96.5 wt pct, POCH, amorphous according to XRD) and lab-grade Fe₂O₃ hematite (Chempur, 97.2 wt pct Fe₂O₃, 0.51 pct Al₂O₃, 0.40 pct MnO, 0.10 pct SiO₂ and 0.10 pct CaO), with a grain size d₅₀ 7.3 µm, LOI +2.668 pct, containing only α-Fe₂O₃ based on 96-210-1170, COD database). The starting chemical composition of the slag based on the purity of the raw materials is presented in Table I. XRF composition was measured using WDXRF Axios mAX spectrometer (PANalytical), equipped with a 4 kW Rh lamp on powdered (< 63 µm) slag samples.

Based on the pseudo-binary phase diagram “Fe₂O₃”-SiO₂ in air,^[32] liquid phase forms in this system at 1740 °C and it decreases to 1140 °C for the Fe₃O₄-SiO₂ system,^[33] proving the key role of the Fe²⁺/Fe³⁺ ratio in these slags. Additionally, as shown by Hidayat et al.^[32] lowering the oxygen partial pressure from 0.21 (air) to 10⁻⁶ will further reduce the formation of liquid from 1400 °C to 1200 °C, respectively. In this work a simple method was developed to produce slags experimentally, through heating the oxide mixture in a graphite crucible, which reduces the partial oxygen pressure through CO formation and allows the oxide mixture to melt faster. In such conditions, Fe₂O₃ easily reduces to magnetite, (Fe³⁺)(Fe²⁺Fe³⁺)₂O₄, which further reduces to wüstite, FeO^[34] followed by its subsequent reaction with SiO₂ to form fayalite, Fe²⁺₂SiO₄, a main component of copper slags targeted in this work. Thus, conditions to produce copper slag in a relatively simple and resource efficient way developed in this work was heating the oxide mixture (Fe₂O₃+SiO₂) in a graphite (C) crucible at temperature above 1200 °C, as described in the next part.

The raw material starting mixture for slags with three different targeted Fe/SiO₂ ratios of 1.2, 1.6 and 1.9, and their respective chemical composition are presented in Table I. The mixture was first homogenized for 2 hour in a ball mill using 10 mm diameter ZrO₂ balls and ZrO₂ bawl, with a material-to-milling media weight ratio of 1:1. 150 g of homogenized loose powder mixture was placed into a graphite crucible (height 100 mm, external diameter 91 mm, internal diameter 61 mm) covered with a 10 mm thick heat-resistance 301 stainless steel.

The empty furnace was preheated at 3 °C/min to 1280 °C under 1 atm of air, then the crucible filled with the Fe₂O₃+SiO₂ mixture (Table I) was quickly inserted into a furnace chamber and held for 30 minutes to completely melt the powder mixture. Then, the crucible was discharged from the furnace, lid was removed, and graphite crucible was carried using a steel tong and the liquid slag was poured onto a stainless-steel plate to air-quench, simulating the solidification process of air-cooled copper slags dumped into cooling pits before grinding.^[9,35] The slag cooling rate during pouring was about 200 °C/min.

Figures 1(a) and (b) show the graphite crucible before and after slag melting. Figure 2 presents macroimages of air-cooled synthetic copper slags produced in this work. Their optical microscopy images are presented in Figure 3, which show elongated arrow-like crystals of fayalite (Fe₂SiO₄) and cubes of magnetite (Fe₃O₄), which will be further confirmed by SEM-BSE and other experimental methods, and the glassy part can be seen on right image of slag S1.5 [Figure 3(a), right magnified image]. The produced slags were dark grey in color and brittle when crushed by hand. The graphite crucible lost about 2 pct of its initial wall thickness after slag melting which was the effect of graphite oxidation at elevated temperatures that lowered the oxygen activity. Pre-trial melting of slag under the same conditions but using 20 mm and 10 mm height pressed pellets instead of a loose powder mixture did not result in melting.

Full size image

Full size image

Full size image

Table II presents the actual chemical composition of synthetic slags determined by XRF. The actual Fe/SiO₂ ratios of molten fayalite slags were found to be higher (1.5, 2.0, 2.3, Table II) compared to the starting values (1.2, 1.6, 1.9, Table I). Excessive temperature or too intense reducing conditions (presence of CO) during slag melting could result in the formation of metallic Fe, which would be unbeneficial in this work, as metallic Fe is not present in typical copper-converting slags. Because no metallic Fe was detected in this work, we conclude that the designed process to produce synthetic slag was successful. Moreover, the obtained range of Fe/SiO₂ ratios (1.5 ÷ 2.3) is relevant and accurately reflects the real Fe/SiO₂ ratios found in copper slags during the converting process.^[10,36] The actual Fe/SiO₂ ratios were subsequently used to designate the produced fayalite slags as S1.5, S2.0 and S2.3.

In the next part, synthetic slags S1.5, S2.0 and S2.3 were subjected to comprehensive analysis and characterization of their structure and microstructure. Room temperature X-ray diffraction (XRD) was applied to evaluate the crystalline nature and identify the phase components in synthetic slags using an Empyrean diffractometer (PANalytical) with a diameter of goniometer of 240 mm, producing Cu-Kα radiation and operated within the 2θ range of 10–90°. Peak matching and phase quantification were performed using the Rietveld refinement method applying X’Pert High Score Plus software.

The microstructures of slags were analyzed using a Scios 2 DualBeam scanning electron microscope (SEM) (Thermo Fisher Scientific) in backscattered electron (BSE) mode. For each slag, panoramic SEM images were registered, covering the entire cross section of the obtained slag from the bottom to the top part of the slag, corresponding to areas in direct contact with a steel plate onto which it was poured and the top part, respectively. The cross-sections of the refractory samples were prepared by traditional ceramographic technique, followed by 10 nm carbon layer sputtering to ensure electrical conductance. Energy dispersive spectrometry (EDS) analysis was conducted to determine the composition of the different phases present in the synthetic slag samples.

⁵⁷Fe Mössbauer spectroscopy measurements under room conditions were conducted on synthetic slags in transmission geometry using the RENON MsAa-4^[37] spectrometer equipped with the LND Kr-filled proportional detector and the ⁵⁷Co(Rh) source. The absorbers were prepared by mixing about 15 mg/cm² of slag sample with 50 mg/cm² of fine B₄C powder to ensure a random distribution of the investigated materials. The data from measured spectra was processed using the Mosgraf software suite within the transmission integral approximation. The isomer shifts (IS) are reported relative to the isomer shift of α-Fe.

The powder samples of synthetic slags, along with pristine starting oxides (Fe₂O₃, SiO₂), were analyzed by X-ray absorption spectroscopy (XAS) at room conditions to investigate the local cation environment using PIRX beamline^[38] at the SOLARIS National Synchrotron Radiation Centre in Kraków, Poland.^[39] PIRX beamline is a bending magnet source beamline that provides soft X-ray energies in the range of 100–2000 eV, equipped with a plane grating monochromator with a resolving power of ΔE/E better than 2.5 × 10⁻⁴. The samples, in the form of powder, mounted on a carbon tape, were measured at Fe L_2,3 edges and Si K edge. The photon energy range used in the experiment was selective for binding energies of Fe: L₂-719.9 and L₃-706.8, and Si: K-1839 eV.^[40] The measurements were conducted in Total Electron Yield (TEY) detection mode, at room temperature and under high-vacuum conditions (10⁻⁸ mbar). XAS spectra are highly sensitive to interactions with neighbouring atoms and their chemical states, so even small changes in the chemical environment (e.g., coordination, oxidation state) are reflected in energy shifts and relative intensity changes of spectral features. The collected spectra were normalized to a unit step using open-source data analysis software Python MultiChannel Analyzer (PyMCA).^[41]

The viscosity of slags was calculated based on their chemical composition (Table I) in temperature range 800 ÷ 1400 °C in air using FactSage^[42] ver. 7.3. This software contains the Viscosity module which allows for precision modelling of viscosity of silicate melts containing iron oxides under oxidation conditions that very well reproduce experimental viscosity data.^[43‐45] The liquid slag viscosity calculated using FactSage is described by the Modified Quasichemical Model (MQM)^[44,45] considering the network-breaking structure of silicate melt.

FTIR spectra were measured in the mid-infrared range at room conditions using a Bruker Optics-Vertex70V spectrometer. The sample was prepared using the tablet method in KBr. The absorption spectra were recorded over 128 scans with a resolution of 4 cm⁻¹, within the range of 1400–400 cm⁻¹.

Raman spectra of phases were obtained on the polished cross-section using LabRAM HR (HORIBA JobinYvon) spectrometer with an excitation wavelength of 532 nm. The diffraction grating was 1800 lines/mm. Raman spectrometer was equipped with an Olympus light microscope which allowed the selection of microarea for local Raman spectra collection. For further interpretation, both Raman and FTIR spectra were processed only to incorporate baseline using the SpectraGryph software 1.2.16.1 applying the adaptive mode of baseline setting.

Figures 4(a) through (f) present XRD patterns of synthetic slags with different Fe/SiO₂ ratios for characteristic reflexes of phases, while Table III summarizes the results of their Rietveld refinements. As can be seen, slags are predominantly crystalline in nature, which results from a relatively low cooling rate (approximately 200 °C/min) during the free cooling process after the molten slag was poured onto the steel plate. This is consistent with many industrial copper slags that are also crystalline after air cooling.^[9]

Full size image

The XRD patterns of slag S1.5 contained only reflexes characteristic for fayalite (Fe₂SiO₄, ICDD PDF no. 04-007-9022), while slag S2.0 and S2.3 revealed additional reflexes corresponding to magnetite (Fe₃O₄, ICDD PDF no. 01-080-6402) with 13.9 and 17.6 wt pct Fe₃O₄, respectively. Their most intensive reflexes were presented in Figures 4(a) and (b). Fayalite is the targeted phase in produced slags, while magnetite is formed due to Fe²⁺ and Fe³⁺ coexisting in equilibrium during cooling of slag in air. In actual copper fayalite slags magnetite content typically reaches up to 17 pct.^[36] Silica was not present in any of synthesized slag as shown by the absence of 100 pct intensity reflex at 26.651° (quartz ICDD PDF no. 00-033-1161), 20.619° (trydymite ICDD PDF no. 01-071-0261) and 21.987° (cristobalite ICDD PDF no. 01-087-7049) as shown in Figures 4(c) through (e). Similarly, there was no 100 pct intensity reflex of ferrosilite (FeSiO₃, ICSD PDF no. 01-082-1832) as shown by Figure 4(f).

The dominant fayalite phase in slags has olivine (M₂SiO₄, M = Ca,Mg,Fe,Mn) orthorhombic structure (space group Pnma). In this structure, oxygen atoms are arranged in a hexagonal closed-packing, with Si⁴⁺ ions occupying tetrahedral sites (4c) while Fe²⁺ locate in two different octahedral sites: position 4a (M1) with symmetry \(\overline{1 }\) and Fe-O bond length of 2.158 Å, and position 4c (M2) with symmetry m and Fe-O bond length of 2.178 Å.^[46,47]

Fayalite phase was generated through the reduction of ferric oxide (Fe₂O₃) at high temperatures to ferrous oxide (FeO), which subsequently reacted with SiO₂ forming fayalite (Fe₂SiO₄) upon cooling below the crystallization point of 1205 °C, according to reactions Eqs. [1] through [3]. The reduction process was facilitated by the graphite (C) crucible which lowered the oxygen partial pressure during melting of a raw mixture (via CO formation) thereby promoting the dissociation of hematite.^[32,34] Higher Fe₂O₃ content (or Fe/SiO₂ ratio) in slag leads to an exponential increase in p_O2 at a specific temperature.^[32] As reported by Jastrzębska et al.^[48] hematite undergoes reduction through the initial formation of maghemite (γFe₂O₃, cubic, iron at 3+ oxidation state only) which then transforms into inverse spinel magnetite (Fe³⁺)(Fe³⁺,Fe²⁺)₂O₄^[48] and can be further reduced to FeO. From values of Gibbs energies of formation at 25 °C based on,^[49,50] reaction Eqs. [1] and [3] are spontaneous (negative ΔG), and this trend is maintained at least up to 1500 °C as visualized in.^[11] So, based on thermodynamics Fe₃O₄ and Fe₂SiO₄ are most likely to crystallize after pouring the slag onto the steel plate and cooling to room temperature, while FeO formation is excluded due to positive ΔG. As visualized by Zhang et al.^[11] the p_O2 stability window of FeO phase at 1300 °C is very narrow from 10^−8.64 to 10^−10.82 atm (and even narrower < 1300 °C), thus it is beyond the typical range of p_O2 in actual converting process of 10^−7.8 to 10^−6.4 atm. At converting temperature of 1300 °C, magnetite is stable in a very wide p_O2 range of 10⁻⁹ to 10⁻² atm.

Figures 5, 6, 7 present SEM microstructures of synthetic slags with different Fe/SiO₂ ratios (S1.5, S2.0 and S2.3) in both panoramic views and selected magnified zones. The bottom and top indications at the images designate the part of the slag that was in contact with heat-resistant steel (10 mm thick, 309 stainless steel) after pouring the liquid slag on, and the upper part exposed to air during cooling, respectively. The panoramic view shows the cross-section of the entire molten slag sample, ranging from the bottom area to the top area. The EDS chemical analysis in microareas of different slag zones is given in Tables IV, V, VI.

Full size image

Full size image

Full size image

All slags were primarily fayalite-based as expected and were previously verified by XRD [Figure 4]. However, in all slags two distinct types of fayalites (O and O_D) were detected, having different Fe/Si ratios based on EDS measurements, which were not distinguished by XRD.

The first type fayalite O, which was the dominant phase in all slags, exhibited a very elongated morphology with crystal lengths exceeding 1 mm due to the characteristic dendritic growth under rapid cooling conditions. The O fayalite grains were more developed in slags with higher Fe/SiO₂ ratios, which can be explained by the higher iron content, which supports the better development of fayalite grains (Fe/Si = 2.0). O_D appeared as a secondary phase represented by dark grey areas between and adjacent to O grains. O phase in all slags had a higher Fe/Si ratio than O_D phase with its average values of 1.6, 1.7 and 1.9 for S1.5, S2.0 and S2.3, respectively; and 1.2 and 1.4 for S1.5, S2.3 and S2.0, respectively. So, both phases (O, O_D) had lower Fe/Si ratios than in the stoichiometric fayalite of 2. The presence of these two phases results from oxidation of Fe²⁺ in fayalite during slag cooling in air. During cooling (from 1280 °C to 25 °C) oxygen activity in air increases, divalent iron in fayalite oxidizes to trivalent iron (Fe²⁺→Fe³⁺). This destabilizes fayalite phase as evidenced by formation of defected fayalite O_D phase (Fe³⁺ and vacancy containing ferrifayalite or laihunite as shown in^[51‐53]), which could also consist of a mixture of very fine defected fayalite,^[51] magnetite and Si-rich glass, similarly as presented in.^[52] 100 pct intensity peak of ferrifayalite at 25.6° (ICDD PDF 01-078-1435) and laihunite at 25.7° (ICDD PDF 01-085-1418) were however not registered by XRD. Nevertheless, the presence of O_D phase was observed in all tested slags so it cannot be explained simply by changes in Fe/SiO₂ ratio in slags that would result in a release of SiO₂ forming glass. In addition, very fine (< 1 µm) dark nanoareas are seen in fayalite O, indicating initiation of fayalite destabilization. Although Fe/Si in both phases (O and O_D) is lower than in stoichiometric fayalite (2), they were assigned to the Fe₂SiO₄ based on results of further examination, and O_D phase was considered highly disordered due to interactions with oxygen during cooling. The presence of SiO₂ (S) and Fe₃O₄ (F) directly adjacent to O_D phase near the top part of S1.5 slag (increased p_O2 compared to bottom part) shows that they are decomposition products of fayalite O_D (Figure 5, right). Zhang et al.^[11] also suggested this course of gradual decomposition of fayalite due to increased p_O2. They also showed for slightly more complex fayalite slag with Fe/Si of 2.0, that Fe₃O₄ content was even 10 times higher when p_O2 increased from 10⁻⁹ to 10⁻⁶ atm reaching a value close to 40 wt pct.

The interaction of oxygen from air with slag is also verified by different thicknesses of magnetite (F) layer on top and bottom sides of slag, as shown in Table VII. The bottom side was in direct contact with stainless steel, while the top was in contact with air, resulting in thickness differences. For S1.5, the thickness of Fe₃O₄ layer was 0.5 ± 0.1 µm on both top and bottom sides (though the destabilization via presence of O_D phase was more pronounced from the middle to the top side of slag). For the rest of slags, Fe₃O₄ layer was thicker at top of 2.4 ± 1.0 and 11.7 ± 4.9 µm, and thinner at bottom of 0.9 ± 0.3 and 1.4 ± 0.5 µm for slags S2.0 and S2.3, respectively. This evidences the impact of increased p_O2 during cooling of slags in air, similarly, as shown in.^[11]

For the slag with the lowest Fe/SiO₂ ratio (S1.5) flower-like or spherical SiO₂ inclusions were observed (Figure 5), though they were not observed by XRD [Figures 4(c) through 4(e)] likely due to its content below detection limit. This results from the excess (15 wt pct) of SiO₂ in S1.5 slag starting composition. For the highest Fe/SiO₂ ratio slag (S2.3), Fe₃O₄ occurred in slag volume also confirmed by XRD [Figure 4(b)] and Mössbauer spectroscopy [Figures 9(a) through 9(c)] resulting from excess iron in this slag. So, presence of Fe₃O₄ in S1.5 slag may indicate that it is the decomposition product of secondary fayalite O_D.

The Fe/SiO₂ ratio in slag significantly influenced the morphology of slag phases with significantly augmented columnar crystals of fayalite (O) for ratio 2.0. The statistics on fayalite (O) grain sizes are attached in Table VII. The slag S2.0 was made up of fayalite grains (O) with an average size of 366 µm, which was between sizes of 240 µm (S1.5) and 401 µm (S2.3), but had the largest maximum grains of 1800 µm and the highest aspect ratio (length to width) of 7. The best developed grains of fayalite in S2.0 slag result from Fe/Si ratio closest to stoichiometric fayalite, Fe₂SiO₄.

This observation has implications for the design of protective high-temperature resistant freeze linings on the hot face of refractory materials, or directly on externally cooled steel walls of vessels, such as those for Cu metallurgy reported by Fallah et al.^[54] and Chen et al..^[55] Interestingly, the morphological distribution of phases from bottom-to-top in the studied simple composition FeO_x-SiO₂ slags is analogous to solidified slag (freeze lining) based on significantly different system Al₂O₃-CaO-FeO_x-MgO-SiO₂-ZnO^[56] and similar to six different lead slags,^[57] with amorphous matrix containing small precipitates from cooled metal side, via zone with equiaxial crystals, columnar olivine long crystals, and ending at partially glassy layer. This pattern confirms repeatable microstructure controlled by thermal gradient of solidified slags, with faster cooling when amorphous and less developed crystals occur.

Based on the obtained results, it could be expected that slags with higher Fe/SiO₂ ratio will produce thicker magnetite, Fe₃O₄, layer at the hot face of refractory in contact with air, like those in copper metallurgy. This high-temperature magnetite (T_m = 1597 °C^[58]) layer can protect refractory against further penetration of aggressive Si-rich liquid copper slag but, on the other hand, it deteriorates the refining function of slag. Overall, formation of Fe₃O₄ on refractory surface may have few implications to real refractory work: (1) increased slag liquidus temperature that causes a problem with slag melting requiring higher heat energy as is observed for excessive magnetite formation during work,^[36,59] (2) increase of slag ‘apparent’ viscosity generating tapping difficulties and entrapment of metal in slag (viscosity of melt slag is proportional to volume fraction of solid particles^[60])^[11,36]; (3) the probability of forming thicker protective coating at a slag-refractory interface, similarly to^[61]; (4) accretion build up on tuyeres. For the sake of slag fluidity supporting Cu refining, magnetite content shall be less than 17 pct, but in practice it is up to 25 pct.^[36]

Based on the obtained SEM-BSE results it could be summarized that, the Fe/SiO₂ ratio in slag ~2.0 will produce best developed fayalite crystals, which are expected to have good stability without silica and magnetite within its volume. Such a system could serve as stable freeze-lining (solidified slag) directly on the cooled reactor wall, or on the refractory surface, protecting it against penetration of aggressive slag elements toward refractory. Whereas slag with Fe/SiO₂ ~1.5 and above ~2.3 will produce free silica and free magnetite in slag volume, respectively, which would diminish integrity of the heterogenous freeze lining, or cause easier penetration of silica to refractory. Also, Fe/SiO₂ ~2.3 in slag could excessively increase magnetite layer thickness, as 11.7 µm-thick magnetite layer on top of examined slag was almost 5 times thicker than for Fe/SiO₂ of 2.0, though freeze-lining requires a minimum of a few millimeters’ thickness.^[62] On the other hand, Si-rich slag will produce a negligible protective layer. In any case, such lining shall well attach to spinel-containing refractory compositions, like Mg₂TiO₄,^[29] (Mg,Zn)[Al,Fe]₂O₄,^[61] or MgO-(Mg,Zn)[Al,Fe]₂O₄,^[63] which can be further studied by, e.g., air or water cooled finger test.^[64,65] Presence of other oxides typical for converting of copper matte, like PbO, CuO_x, CaO, MgO, Al₂O₃, in fayalite slag analogous to studied in this work, was tested in our previous works^[29,30] and showed tendency to formation of Fe-rich spinel layer at interfaces when subjected to contact with refractory spinels (MgCr₂O₄, MgAl₂O₄, ZnAl₂O₄, Mg₂TiO₄) and oxides (MgO, Al₂O₃). What is clearly seen when comparing the current work (slag freezing on steel) with slag freezing on refractory substrate after corrosion,^[29,30] is that the substrate type (metal or ceramics) completely shifts the phase equilibrium, from fayalite to <magnetite+glass>, respectively.

Figures 8 and 9 present the results of ⁵⁷Fe Mössbauer spectroscopy (MS) for α-Fe₂O₃ raw chemical and molten fayalite slags, respectively. The corresponding hyperfine parameters including isomer shift (IS), quadrupole splitting (QS) and magnetic hyperfine field (B), derived from decomposed spectra are listed in Table VIII. Mössbauer spectroscopy is a useful technique to examine Fe-containing materials, such as copper slags, in which Fe is typically the second most dominant element.^[10] Mössbauer spectra provide information on the oxidation state and coordination environment (octahedral, tetrahedral) of iron indicated by specific values of isomer (or chemical) shift (IS). These shifts follow a characteristic order: IS (Fe³⁺_Td) < IS (Fe³⁺_Oh) < IS (Fe²⁺_Td) < IS (Fe²⁺_Oh).^[66]

Full size image

Full size image

The experimental hyperfine spectra of α-Fe₂O₃ chemical which was used for the preparation of fayalite slags, and the spectra of slags were fitted using a conventional refinement routine with a Lorentzian line shape.

The MS spectrum of α-Fe₂O₃ (Figure 8) was resolved into two sextets: a dominant component (S1) at IS of 0.37 mm/s and a minor component (S2) at 0.38 mm/s. Both subspectra correspond to Fe³⁺ in octahedral sites. The dominant subspectrum (S1) represented 98 pct of the total area and exhibited a strong hyperfine magnetic field (B = 51.09 Tesla) typical of ferric ions (Fe³⁺) in the octahedral sites (O_h) of hematite.^[66] The minor component (S2) accounting for 2 pct of the total area, was a red sextet characteristic of paramagnetic ferric ions in a highly distorted phase, with a high QS of 0.76 mm/s. The latter component likely results from impurities in the raw material of hematite which had a purity of 97.2 pct Fe₂O₃.

As shown in Figure 9, the experimental hyperfine spectra of all the slags were decomposed into two quadrupole doublets and two sextets. The isomer shift (IS) of the doublets, at 1.16 and 1.3 mm/s (for all slags), clearly indicates Fe²⁺ in octahedral positions with two different symmetries. Based on the results of Mössbauer spectroscopy (MS) and the SEM/EDS analysis, two possible scenarios for Fe²⁺ occupation arise. The first scenario assumes that Fe²⁺ occupies two different octahedral sites, M1 and M2, within the same fayalite phase. The M1 site, which is more disordered (D1), is characterized by a quadrupole splitting (QS) of 2.82 mm/s, while M2, with higher symmetry (D2), has a QS of 1.4–1.8 mm/s. The higher intensity of D1 (pink) indicates a larger presence of M1 site component in all slags, compared to D2 (green).

Nikolov et al.^[67] studying copper slags with Fe₂O₃ content of 58.4 wt pct and SiO₂ of 29.3 wt pct observed two quadrupole doublets with IS values of 1.14 and 1.17 mm/s and QS values of 2.68 and 2.88 mm/s, which they ascribed to Fe²⁺ in two different octahedral sites (M1 and M2) in fayalite. They also detected additional doublet (IS 1.24 mm/s) related to Fe²⁺ in glassy phase, which was not observed in any of the slags studied in this work due to a lack of fluxing elements compared to their study.^[67] Further MS studies at 64 K^[68] confirmed the presence of two paramagnetic doublets at IS values of 1.23 and 1.30 mm/s, attributed to Fe²⁺ in two distinct octahedral sites with different electrical field gradients around the Fe nucleus. Variations in IS values may stem from differences in slag cooling rates.

The alternative explanation for the Fe occupation involves the higher spectral line width (Γ) observed for D2, which ranged from 0.9 to 1.0 mm/s, compared to D1 with a significantly narrower line width of 0.28–0.29 mm/s. This suggests that D1 and D2 could originate from Fe²⁺ in octahedral sites of different fayalite phases. SEM/EDS analysis revealed two types of fayalites, one with Fe/Si ratio of 1.6–1.9 and the other with a ratio of 1.2–1.4. The narrower line (Γ) of D1 could therefore be attributed to fayalite with a more stoichiometric composition (O), while D2 with a broader linewidth likely originates from Fe-poor defected fayalite (O_D). Based on the literature and the results obtained in this study, the first scenario—where doublets arise from Fe²⁺ in the M1 and M2 sites of the same fayalite phase—is more plausible. The doublets in fayalite should be assigned to Fe²⁺ in the M1 and M2 sites, with the latter occupying a larger volume of octahedra^[69] due to the significantly higher Γ.

The presence of sextets in all samples indicates that a magnetically ordered phase exists in produced slags. The experimental sextet was resolved into two subspectra with isomer shifts of 0.28 and 0.65 mm/s, attributed to Fe³⁺ and Fe^2.5+ in tetrahedral (T_d) and octahedral (O_h) sites, respectively, of magnetite structure.^[70] Magnetite is inverse spinel^[48] having Fe²⁺ (O_h) and Fe³⁺ (T_d, O_h) ions. Assignment of the S1 to partial oxidation state (Fe^2.5+) ion is due to very fast electron hopping between ferrous (Fe²⁺) and ferric (Fe³⁺) ions in octahedral sites of magnetite, which is faster than the sampling time in Mössbauer effect measurement of 10⁻⁸ s.^[71,72] The S2 subspectrum, with an IS of 0.28 mm/s, corresponds to Fe³⁺ in the tetrahedral site (T_d) of magnetite. The higher hyperfine magnetic field (48.5 Tesla) associated with the Fe³⁺ in T_d sites (S2) supports its identification as Fe³⁺, while the lower magnetic hyperfine field (45 Tesla) of the Fe^2.5+ in O_h sites (S1) indicate a mixed valence state. The increasing contribution of magnetite in the composition of produced slags, S1.5 (3 pct), S2.0 (18 pct) and S2.3 (22 pct), is mainly due to higher Fe/SiO₂ ratios in them (Table II) but it is also impacted by increased p_O2 during cooling the slag in air.

The total contribution of Fe²⁺ to fayalite in this work was 97 pct in slag S1.5, decreasing to 81 pct in S2.0, and 78 pct in S2.3, respectively, with rest to 100 pct corresponding to magnetite. This is accompanied by a decreasing Fe²⁺/Fe³⁺ ratio in slags of higher iron content, being 32.3 (for S1.5), 4.3 (S2.0), 3.5 (S2.3). This results from iron content and creating redox conditions for iron oxides, which were different during melting (reducing) and subsequent cooling (oxidizing).

First, fayalite \(\left( {Fe_{2}^{2 + } SiO_{4} } \right)\) was possible to be formed out of Fe₂O₃+SiO₂ mixture melted in graphite crucible at 1280 °C in air, due to effective Fe³⁺ reduction to Fe²⁺ caused by oxygen activity decrease by C from graphite crucible, as confirmed by about ~50 g crucible mass loss after melting. Considering the mass of oxygen (~4 g) contained in a furnace chamber (volume of 0.015 m³) and the loss of ~50 g of graphite, all available oxygen was consumed. Thus, the resulting high CO/CO₂ ratio during heating reduced the oxygen partial pressure creating conditions for iron redox reaction. This enabled the reduction of ferric (Fe³⁺) to ferrous Fe²⁺) ions mainly by CO, with a likely contribution from direct reduction by solid carbon.

Then, during cooling of slags (from 1280 °C to 25 °C) after pouring them onto the steel plate, oxygen activity increased, which affected the slag by destabilizing the primary fayalite phase \(({Fe}_{2}^{2+}Si{O}_{4})\) as evidenced by the appearance of adjacent O_D phase (Figures 5, 6, 7), finally forming Fe₃O₄ and SiO₂ in slag volume (Figure 5, right). The impact of increased p_O2 is also thicker Fe₃O₄ layer on top surface of slags compared to its bottom side (Table VII), with the former being in a direct contact with air. The phase distribution showing magnetite as the only form of iron oxide in all slags (XRD, SEM-EDS, Mössbauer spectroscopy) indicated that the oxygen partial pressure during slag melting was in approximate range from 10⁻⁸ to 10⁻² atm, according to stability diagram shown by Zhang et al.^[11]

The presence of magnetite in slag S2.3 is mainly due to an excess iron in the starting chemical composition (Table I), next to the primary Fe₂SiO₄ phase. In contrast, the presence of low amounts of magnetite in S1.5 and S2.0 slags, and a corresponding content in slag S2.3, is expected to be the effect of the increased oxygen activity during cooling that destabilized the primary fayalite (O) phase.

In terms of practical aspects, the maximum Fe₃O₄ content in the synthetic fayalite slags produced in this work of 16 pct for Fe-richest slag (Fe/SiO₂ = 2.3) is still lower than the recommended maximum content of magnetite in industrial copper slags of 17 pct. This limit is important for maintaining good slag fluidity and, thus, enhancing Cu refining properties, though in practice Fe₃O₄ contents up to 25 pct are met.^[4]

Fe L and Si K edge X-ray absorption spectra (XAS) for pristine oxide raw materials (α-Fe₂O₃, amorphous SiO₂) and synthetic fayalite slags are shown in Figure 10. The energies of spectral features are attached in Table IX (Fe) and Table X (Si).

Full size image

The L-edge spectra derive from spin-orbit interaction between 2p core holes with 3d final states. This interaction causes the peak to split into two distinct peaks, L₃ and L₂. These peaks correspond to the transition of core electrons from Fe (2p) level to the unoccupied 3d orbital.^[73]

Both the Fe L₃ and L₂ edges were split into two subpeaks (L₃: B, C and L₂: E, F). These subpeaks derive from the crystal field splitting of d bands with oxygen bonding, indicating different positions or oxidation states of Fe within the slag phases.

The XAS spectrum of hematite [Figure 10(a)] clearly shows the doublet for L₃ edge, with subpeaks at B (708.5 eV) and C (710 eV), and for L₂ edge at E (721.6) and F (723.2 eV). The shape of these subpeaks is like previously published XAS spectra for hematite.^[74] According to Shen et al.^[75] the doublets in both L₃ and L₂ regions confirm the presence of Fe³⁺ in hematite. In the hematite structure, Fe³⁺ is in distorted octahedral sites surrounded by oxygen anions.^[76] The presence of only Fe³⁺ in hematite was also evidenced by Mössbauer spectroscopy (Figure 8, Table VIII).

For the fayalite slags [Figure 10(a)] the XAS spectra combine the contributions from all Fe atoms in the structure, originating from both occupational sites (M1 and M2). Interestingly, XAS spectra of fayalite slags are similar to hematite, but with additional pre-peaks (a and d) and different subpeaks intensity ratios. Notably, C/B peak ratio is higher for hematite (1.87) and approaches 1.0 for all slags. Also, E/F peak ratio is below 1 for α-Fe₂O₃ and above 1 for all slags, with lower values for the increased Fe/SiO₂ ratio in slag.

The positions of characteristic features in the XAS spectra (Table IX, indicated as: a, B, C, d, E, F) are very consistent, with only marginal shifts observed between α-Fe₂O₃ and slags for the B (Δ = 0.1 eV) and F peaks (Δ = 0.4 eV). No significant shifts in peak positions were observed among the fayalite slags.

Mössbauer spectroscopy, which is a high-resolution method sensitive to iron oxidation state and coordination, indicated that most Fe in slags (97 pct in S1.5, 81 pct in S2.0, 78 pct in S2.3; Table VIII) is Fe²⁺ occupying M1 (more distorted) and M2 (less distorted) octahedrons, with a minor contribution from octahedral Fe^2.5+ and tetrahedral Fe³⁺ in magnetite. So, it can be deduced that B peak (708.4 eV) and C peak (710 eV) at XAS spectra represent the imposition of Fe²⁺ at equivalent M1 and M2 sites, with B primarily deriving from M2 octahedron. This is supported by the observation of a split in Fe K edge into two subpeaks for synthetic Fe₂SiO₄ (dry ice quenched) as shown by Wu et al.^[77] where the peak intensity ratio changed when XAS spectra were measured at 900 °C.

Si K edge spectra for SiO₂ and fayalite slags [Figure 10(b)] show a dominant A peak and minor b and c features. Position of K edge single peak is at 1850.6 eV for SiO₂ pristine sample, and it leftward shifts slightly to 1850.4 eV (S1.5) and 1850.2 eV (S2.0 and S2.3), as attached in Table X. The analogous single peak for Si K edge of fayalite at 1846.6 eV was observed by Li et al.^[78] The difference in peak positions may derive from differences in cooling rates between slag samples, which significantly affects the structural atomic order. Although the pristine SiO₂ used for slag preparation was amorphous (as confirmed by XRD), the molten slags were crystalline.

The peak shift in energy scale is the indicator of reduced degree of polymerization of SiO₄⁴⁻ clusters in slag.^[78] This was proved by the systematic leftward shift in Si K edge peak energy of minerals with different Qⁿ (Q is tetrahedron, n is number of neighboring tetrahedrons linked by bridging oxygen atoms, n = 0 ÷ 4^[35]), such as shift from Q⁰ (forsterite), Q² (enstatite) to Q³ (talc) at 1845.5, 1846.5 to 1846.9 eV, respectively.^[78] Fe²⁺ ions in copper slag are known to break Si⁴⁺–O_D^– bonds, generating more nonbridging oxygen anions and leading to depolymerization of the slag (reducing Q^x).^[11,35] This results in a drop of slag viscosity, which can even be 5 times lower compared to slags with FeO/SiO₂ ratios of 1.5 and 2.5 at 1000 °C, as shown in Figure 11 (simulated using FactSage, Viscosity module^[42]). Thus, XAS spectra in this work provide evidence of a reduced polymerization rate of copper slag with higher Fe/SiO₂ ratios.

Full size image

Figure 12 presents FTIR spectra of pristine SiO₂ and α-Fe₂O₃ used for the preparation of synthetic slags, while Figure 13 shows spectra of molten fayalite slags.

Full size image

Full size image

SiO₂ (amorphous by XRD) contains characteristic broadened bands corresponding to Si-O-Si vibrations at 1214 cm⁻¹ (symmetric stretching), 1093 cm⁻¹ (asymmetric stretching), 820 cm⁻¹ (symmetric stretching) and 463 cm⁻¹ (bending), which agree well with bands reported by Tran et al.^[79] for amorphous SiO₂. Ocana et al.^[80] observed similar bands in amorphous silica at 1225, 1075, 800, 450 cm⁻¹, and comparable values were obtained in.^[79] It shall be noticed that the size, shape and aggregation of SiO₂ amorphous particles can affect the infrared spectra of silica,^[80] which may be responsible for variation of the results.

α-Fe₂O₃ spectrum reveals intensive absorption band at 542 and 473 cm⁻¹, both containing the shoulder. Although these bands are broader than those observed in natural hematite (517 and 438 cm⁻¹, RUFF no. R050300^[81]) they align with the characteristic absorption peaks of hematite. The shape of hematite particles influences the FTIR spectrum: sharp bands (567 and 483 cm⁻¹) were observed for single-crystalline 1 µm size pseudocubic particles of α-Fe₂O₃, while much broader with additional shoulders were observed for ellipsoidal α-Fe₂O₃ with a length of 2.1 µm and an aspect ratio of 2.5.^[82] Moreover, the bands at FTIR spectra of α-Fe₂O₃ shown to broaden with increasing aspect ratio (length to width) of the hematite particles.^[82] The grain size of the hematite used in this work was 7.3 µm (d₅₀) with a bimodal particle size distribution, which explains the character of the obtained spectrum.

Before FTIR and Raman spectra analysis of fayalite slags, it is important to understand the crystal structure and coordination of atoms in fayalite (Fe₂SiO₄). Olivine-structure fayalite contains isolated [SiO₄] tetrahedra linked by [FeO₆] octahedra. Structure is represented by each oxygen in octahedron bonded to a tetrahedrally coordinated silicon, forming edge-sharing [FeO₆] chains parallel to the c-direction which are cross-linked to similar chains of [SiO₄] tetrahedra. Due to this order, some [FeO₆] octahedra are distorted, with smaller (M1) and larger (M2) volumes. Fe-O bond lengths are 2.158 and 2.178 Å for M1 and M2, respectively.^[69]

FTIR spectra of synthetic fayalite slags (Figure 13) are very similar to each other, with characteristic intense absorption bands in ranges 827–962 and 476–563 cm⁻¹, characteristic for Si-O vibrations in SiO₄ units of fayalite. The peaks at 947–874 and 825–827 cm⁻¹ are due to asymmetric (ν₃) and symmetric (ν₁) stretching of Si-O bonds in SiO₄ of Fe₂SiO₄ (SEM, phase O),^[11,25] respectively.

The well-resolved absorption band at 561–563 cm⁻¹ corresponds to asymmetric bending vibrations (σ) in SiO₄ groups of Fe₂SiO₄ similarly to shown in,^[11,25] and it overlaps with band from stretching vibrations of Fe-O bonds in Fe₃O₄, which was observed in slags S2.0, S2.3 by XRD and Mössbauer spectroscopy. The band at 507–476 cm⁻¹ is due to asymmetric bending (ν₄) of Si-O in fayalite phase.

Broad band at 1113–1120 cm⁻¹ is indicative of Si-O-Si vibrations in the amorphous silicate phase,^[67] which is most likely glassy-like fayalite (O_D) with Fe/Si ratio of 1.2–1.4, as seen in SEM images for all three slags (Figures 5, 6, 7). Alternatively, it can be attributed to SiO₂ remnants, which was especially visible at SEM images of the highest silica content slag S1.5 (Figure 5).

Summarizing, most absorption bands come from internal Si-O vibrations in isolated SiO₄ units of Fe₂SiO₄, and peak at 561–563 cm⁻¹ may indicate the presence of Fe-O vibrations in Fe₃O₄, although it overlaps with bending in SiO₄ groups of fayalite. FTIR spectra obtained in this work are most similar to those obtained for industrial copper slags from flash-smelting and converting processes,^[23] confirming the representativeness of synthetic slags produced in this work.

Raman spectroscopy is a complementary technique to FTIR spectroscopy as it is more sensitive to symmetric molecular vibrations due to polarization changes during vibrations, while FTIR is sensitive to antisymmetric vibrations resulting from changes in the dipole moment.^[83] Raman spectra, measured in microareas observed under an optical microscope (at 100× magnification) of slag S2.3, previously identified by SEM as fayalite O, fayalite O_D and magnetite F, are shown in Figure 14. This figure also compares the spectra with the reference spectrum of fayalite (X050077) and magnetite (R060191) from the RUFF database.^[84]

Full size image

Low-frequency Raman modes at 238–240 cm⁻¹ in fayalite are most likely due to SiO₄ translations.^[69] All bands between 600 and 1000 cm⁻¹ represent internal stretching vibrations in SiO₄ (symmetric ν₁, and asymmetric ν₃) for both IR and FTIR spectra.^[85]

Raman spectra of both crystalline (O) and defected (O_D) fayalite phases show similar features, with two comparable high-energy Raman modes at 816 cm⁻¹ (O, O_D) and 839 cm⁻¹ for O, and 843 cm⁻¹ for O_D. The corresponding bands in reference to the Fe₂SiO₄ spectrum (RUFF, X050077) are broad with distinguishable bands at 820 and 843 cm⁻¹. These bands correspond to symmetric vibrations ν₁ in SiO₄ tetrahedra, which were previously observed in FTIR spectra as a small band at 825–827 cm⁻¹ (Figure 13). Neighboring, low-intensity Raman band at 901–903 cm⁻¹ in fayalite most likely corresponds to asymmetric stretching vibrations (ν₃) in SiO₄ of fayalite, which was observed as the strong band in FTIR spectrum (Figure 13, 914 cm⁻¹).^[69] This confirms the complementarity of both experimental methods, with higher intensity bands from symmetric vibrations in Raman, and asymmetric vibrations in FTIR spectroscopy. Interestingly, this last band was not reported for reference fayalite. A small band at 505-503 cm^-1 is due to asymmetric bending vibrations (ν₄) in SiO₄.^[67]

Overall, the bands for both fayalites’ phases are very similar to each other, although those in fayalite O (Fe/SiO₂ = 1.9) are less sharp compared to defected fayalite O_D (Fe/SiO₂ = 1.2). This can be explained by defected nature of the latter one as seen in SEM images, and its nonstoichiometry (low Fe/SiO₂ ratio).

In slag 2.3, magnetite was confirmed by both XRD, MS as well as SEM/EDS analysis through observation of bright, dendritically grown crystals with an approximate size of 10 µm. Raman spectrum [yellow, Figure 14(a)] measured at the brightest area observed under an optical microscope showed a very strong, broadened band characteristic for Fe-O symmetric vibrations at 662 cm⁻¹ (A₁g, stretching of oxygen atoms along Fe-O) and a weaker band at 533 cm⁻¹ (T₂g, asymmetric bending Fe-O) for magnetite. The analogue bands in the reference magnetite red spectrum occur at 672 and 556 cm⁻¹ (R060191), which also agree well with the values of 662 and 530 cm⁻¹ reported by Qu et al.^[86] A small band at 294 cm⁻¹ represents the symmetric bending (Eg) of Fe-O in magnetite.^[87]

The use of Raman spectroscopy confirmed the presence of magnetite and fayalite phases in molten slags, although negligible differentiation was observed between crystalline and defected fayalites. Despite this, a combination of microscopic and structural methods is valuable and useful for precise slag characterization.