Fundamental experimental and thermodynamic modelling study in support of the lead battery recycling slag

Antimony, tin, and other higher value metals, as well as elements such as arsenic and lead, can be found in the slags produced during the lead battery recycling process. Gopher Resources operates lead battery recycling furnaces that use sodium slag systems for which there has been less thermodynamic research than more common slag systems. Better thermodynamic information could help improve the process efficiency and control of these furnaces, as well as the recovery of higher value metals. The present study focuses on experimental research and thermodynamic modeling of slags belonging to the Na-Si-Fe-O system, with minor elements including S, Sb, Sn, Pb, and As. Examples of important systems studied extensively for the first time include Fe-Sb-Si-O and Na-Sn-Si-O. Phase equilibria methods are used to determine all the interaction parameters between the impurity metals and the main components of the slag. The experimental methodology involves equilibration, quenching, and electron-probe X-ray microanalysis of the samples. The modified Quasichemical model is used to describe the thermodynamics of the slags. The model also takes into account possible formation of matte/metal/speiss liquids, and numerous solid phases, which is important for understanding of fundamentals operation of various process units.


Introduction
Pyrometallurgical recycling of lead acid batteries can be performed using at least three different routes, depending on the sulfur removal technology: 1) hydrometallurgical separation of sulfur in a form of Na2SO4 through precipitation of lead basic carbonate followed by pyrometallurgical processing [1]; 2) pyrometallurgical removal of sulfur in a form of SO2 using oxidative smelting, often conducted in top submerged lance furnace [2]; 3) pyrometallurgical removal of sulfur in a form of Na2S-FeS matte during reductive smelting [3].Gopher Resource operates using the first route.Process steps in terms of main chemical reactions are described below (Figure 1).
Step 1: Hydro de-sulfurization: -3PbSO4(aq) + 3Na2CO3(aq) + H2O = Pb3(CO3)2(OH)2(solid)↓ + 3Na2SO4(aq) + CO2 -Lead basic carbonate is contaminated with Na2SO4 and Na2CO3 Step 2. Oxidizing smelting (Reverberatory furnace) of desulfurized paste and grids: ..) Thus, the thermodynamics of slag, metal and matte under reducing conditions is fundamental for improving the overall efficiency of the process and developing possible options for slag cleaning.The purpose of the present project is to develop a computational tool based on thermodynamics.The tool should be able to predict conditions when the phases are liquid or solid.It can estimate the energy required for the process, and optimal amount of fuel.It uses non-ideal solutions to predict the distribution of every element among the phases in the system.The range of conditions targeted in the present study represent the reduction smelting of materials formed during lead acid battery recycling.

Methods and scope of the project
The methodology is based on the integrated experimental and thermodynamic modelling work.The principles are described in the parallel PbZn'2023 paper [6].In brief, thermodynamic models are developed for the potential phases in the system, which are shown in Table 1.Thermodynamic models contain adjustable parameters, which need to be optimized using experimental information.These parameters represent thermodynamic properties of end-members of solution phases, as well as binary and ternary interaction parameters between solution constituents.Many of model parameters for the selected systems have been determined earlier, through consortia programs conducted in PYROSEARCH since about 2013.The remaining model parameters are determined through systematic assessment of chemical systems, selected for the priority studies.They are shown in Figure 2. When available, experimental literature data are used.When data are not available, experiments are designed to fix the particular model parameters.The model for the slag phase has been developed within the modified Quasichemical framework [7][8][9][10][11].For the geometric interpolation of binary parameters into ternary space [12,13], the acidic component SiO2 was treated as an asymmetric one with respect to basic FeO, PbO, Na2O, SnO, etc. oxides.The decisions on geometric grouping of amphoteric components Fe2O3, Sb2O3, SnO2 were made on case-by-case basis, selecting the one that ensures the best approximation of the experimental data with the least number and values of additional ternary parameters.For slags containing sulfur, quadruplet approximation formalism was used, which considers mixing of pure oxide and sulfide endmembers (e.g.Na2O-FeO-Fe2O3-SiO2-Na2S-FeS-Fe2S3-SiS2-...) with possible additional interaction parameters such as Na-Fe-O-S.While two sublattices (cationic for metals and anionic for O -2 and S -2 ) were used in slag model, the matte/metal/speiss model treats all species equally on one sublattice by assigning binary interaction parameters like Fe-O, Fe-S, Na-S, and ternary Fe-O-S, Na-Fe-O, etc.This provides additional freedom in describing non-stoichiometric mattes (with deficiency and excess of sulfur), extending to sulfur-free metal within one model, but has a limitation on lacking accuracy in ternary oxide or sulfide systems, e.g.FeO-Sb2O3-SiO2, for which therefore the slag model is preferred.Experiments of the present study are conducted using the equilibration, quenching, EPMA analysis methodology.The steps required to produce one experimental point are shown in Figure 3.They were described in earlier publications [14,15].Powder mixtures were prepared with an agate mortar and pestle from synthetic reagents (SiO2, Fe2O3, Fe, Sb2O3, Sb, Na2CO3, SnO2, Sn, As2O3, As, PbO, Pb, FeSall at least 99.8% purity) to obtain the desired bulk composition.To prevent explosion of sealed ampoules by released gases, and minimize loss of volatile components on open substrates, some of these reagents were pre-reacted to form master-slags or master-compounds, such as: -Na2CO3 + SiO2 = Na2SiO3 + CO2(g).Na2SiO3 master-compound can be used in closed ampoule.-PbO + SiO2 = "Pb2Si3O8" master-slag (900°C).This master-slag prevents the loss of volatile PbO in during high-temperature experiments; -Na2CO3 + Sb2O3 + O2 = 2NaSbO3 + CO2(g).This reaction binds volatile Sb2O3 together with Na2CO3 into more stable compound; -Fe + As = FeAs (600-700°C).This compound can be heated up to 1200°C, while pure unreacted As sublimates and causes explosion of sealed ampoules.
The mixtures were pressed into 0.3-0.Equilibration experiments were conducted in a vertical tube furnace.Thermocouple based on Pt-Rh 6%/Pt-Rh 30% alloy protected by the alumina shield allowed the temperature control within the furnace of ±1 K.The location of the thermocouple was 2-3 mm from the sample.The working thermocouple was calibrated against a standard one every 1-2 months.The combined uncertainty in calibration, thermal contact, and other factors provide the overall uncertainty of equilibration temperature of 5 K or less.The equilibration time was pre-determined in trial experiments, specific for each chemical system and the range of compositions.It ranged from 10-15 min for highly volatile materials to several weeks for low-temperature viscous materials.To preserve the composition of phases existing at the temperature of equilibration, samples were quenched into the solution of calcium chloride in water.High concentration of calcium chloride prevents freezing and can provide a quenching medium at 253 K.After quenching, samples were cleaned and placed into epoxy resin.The Epoxy resin provides a stable substrate for the polishing and electron microscopy analysis.Silicon carbide paper and diamond paste used during the polishing, which was rub by the TegraPol-31 machine (Struers, Denmark).To prevent the accumulation of charge on the surface of the sample during the electron microscopy, a layer of conductive carbon film was created used the QT150TES apparatus by Quorum Technologies, UK. Electron probe X-ray microanalysis (EPMA) was run using JEOL JXA 8200L machine by Japan Electron Optics Ltd., Tokyo.Elemental compositions of each phase were measured.For quantitative analysis, a Wavelength Dispersive Detectors (WDD) supplied with the machine was used.During the X-ray count, the acceleration voltage of 15 kV and a probe current of 20 nA were used.Collected characteristic X-ray counts were converted to concentration using the appropriate standards: hematite Fe2O3 for Fe, wollastonite CaSiO3 for Si and Ca, albite NaAlSi3O8 for Na, InAs for As, metallic Sb for Sb, chalcopyrite CuFeS2 or pyrite FeS2 for S, cassiterite SnO2 for Sn.The standards were supplied by Charles M. Taylor Co., Stanford, CA, USA, and 71.4 wt.% PbO-SiO2 K456 glass from NIST for Pb.Electron interactions, X-ray absorption, effects of fluorescence and other factors, that differ between the sample and the standard were taken into account by the Duncumb-Philibert ZAF correction.The correction calculations were performed by the software supplied with the machine.The overall accuracy of measurement of compositions for elements in high concentrations was estimated to be at least 1 wt %.The information on the oxidation states, e.g.Fe 2+ /Fe 3+ , cannot be obtained using the current analytical methodology.The following standards were used for EPMA: For the slag phase, non-zero probe diameters of 20-50 micron may be used to decrease the scatter due to dendrite formation on quenching and limit the loss of volatile species (Na) under electron beam.

Na-Fe-Si-O thermodynamic model
Thermodynamic assessment of the Na-Fe-Si-O and its sub-systems has been conducted using experimental data from literature or experimental results produced earlier.The Fe-Si-O system is largely based on the earlier assessment [16], with modifications.The same modelling framework was used, but parameters have been re-optimized to achieve equal or better agreement with collected experimental data [16], including more recent experimental results [17].Thermodynamic properties of liquid SiO2 in slag have been changed in 2022.The modelling framework for the Na-Si-O system was described by Wu et.[18], but model parameters were modified to describe experimental phase diagram data collected in later work by Nekhoroshev [19].Phase diagram is shown in Figure 4.The modelling framework of Nekhoroshev [19] used more complex slag structure using (Na2) 2+ species to describe the cristobalite liquidus at high temperatures, and it is harder to expand in large thermodynamic database.In the present study it was demonstrated that reasonable description of cristobalite liquidus can be achieved by using very high-power parameters on Si 4+ , such as The model of the present study provides inferior results for metastable low-temperature miscibility gap in high-SiO2 glasses compared to Nekhoroshev [19], and for very low-SiO2 liquids (< 0.3 mole fraction of SiO2), but these conditions are of low priority for reductive smelting.The Na-Fe-O system has been modelled by Moosavi-Khoonsari and Jung [20].In their modelling framework, the NaFe 4+ associate was included in for the slag solution.The formation of NaFe 4+ in slag is expected at higher oxygen partial pressures (higher oxidizing potentials), compared to those observed in Step 3 of Figure 1, but for model consistency and possible other uses of the database, the same modelling framework has been adopted in the present study.Model parameters have been re-optimized to achieve consistency with updated Fe-O systems.Models for the Fe-O system have been re-evaluated by PYROSEARCH in 2015-2016 [21,22] to correct the properties of wüstite, but they have not yet been corrected in FactSage FToxid database, including the 8.2 version.Moosavi-Khoonsari and Jung [20] used the Fe-O system from FactSage FToxid database.Selected results of the modelling of the present study are shown in Figure 4.The Na-Fe-O-Si system has been modelled by Moosavi-Khoonsari and Jung [23], who collected existing experimental data.In our investigation, model parameters were reevaluated to reproduce the collected experimental data and the above mentioned changes in Fe-O and Fe-O-Si.Example is shown in Figure 4. Results at higher P(O2) were also included in the model of the present study, but they are not shown because they are far away from operational conditions.

Introduction of Na in matte-metal thermodynamic model
During the course of the project, it became evident that the formation of liquid matte (sulfide-rich liquid) phase is possible in the blast furnace, or during the pyrometallurgical cleaning of lead blast furnace slags.Liquid matte is expected to be close to Na2S-FeS-FeO stoichiometry.Existing literature data were used to model the Na-S and Na-Fe-S system.Examples of modeling results are shown in Figure 5.The model for the Fe-S is based on the thermodynamic assessment of Waldner [24].The As-S [25] and Pb-S [26] were modelled recently.The Sb-S, Sn-S were modelled as well, based on literature data, but the assessments are not published yet.Preliminary model parameters are introduced for the Na-As-S, Na-Sb-S, Na-Sn-S and Na-Pb-S systems to provide reasonable description of the corresponding ternary phase diagrams.

Fe-Sb-Si-O system in equilibrium with Fe-Sb metal
No literature data were available for the Fe-Sb-Si-O system, which is not only important for lead acid battery recycling, but also for pyrometallurgical production of antimony [29].The summary of produced experimental results and example of the microstructure of are shown in Figure 6.These data were used to fix the Fe 2+ -Sb 3+ interactions in slag, as well as ternary parameters Fe ,Si (Sb )//O ijk g in slag.
A new solid compound FeSb2O4 (schafarzikite) was introduced in the database.

Na-Sn-Si-O system, equilibrium with Sn metal and in air
No literature data were available for the Na-Sn-Si-O system.Two experimental series were designed to fix the

Fe-Sb-Sn-O system in equilibrium with Fe-Sb-Sn metal
First experimental results were obtained for the Fe-Sb-Sn-O system.No new ternary parameters were introduced in the slag, but a Sb 3+ -Sn 4+ interaction parameter in slag was optimized using the data in Figure 8. Also, the predicted distribution of Sb/Sn/Fe between slag and metal, and distribution of Sn between slag and spinel were tested.Significant solubility of Sn 4+ in spinel was observed.

Fe-As-Si-O system in equilibrium with Fe-As metal (speiss)
The solubility of arsenic in slag can be important if speiss-containing secondary materials are introduced in the process.The solubility of arsenic in the slag was experimentally measured and compared with thermodynamic database predictions.The results are shown in Figure 9.

Figure 1 .
Figure1.Schematic of the lead acid battery recycling using the lead carbonate precipitation and pyrometallurgical processing route.Furnace pictures are from[4,5]

Figure 2 .
Figure 2. A summary of chemical systems selected for thermodynamic modelling and experimental study.

Figure 3 .
Figure 3. Steps required to produce an experimental point.

Figure 5 .
Figure 5. Examples of thermodynamic modelling results for the Na-S [27] and Na-Fe-S [28] systems based on data collected in literature.Red lines represent thermodynamic calculation using FactSage using model parameters of present study.

Figure 6 .
Figure 6.Example of experimental results and model predictions in the Fe-Sb-Si-O system.Lines represent thermodynamic calculation using FactSage, symbols on the diagram are experimental data obtained in 2022-2023.

Figure 7 .
Figure 7. Example of experimental results and model predictions in the Na-Sn-Si-O system.Lines represent thermodynamic calculation using FactSage, symbols on diagrams are experimental data obtained in 2022-2023.

Figure 8 .
Figure 8. Example of experimental results and model predictions in the Fe-Sb-Sn-O system (slagmetal-solid equilibria).Lines represent thermodynamic calculation using FactSage, symbols on diagrams are experimental data obtained in 2022-2023.

Figure 9 .
Figure 9. Example of experimental results and model predictions in the Fe-As-Si-O system (slagmetal-tridymite equilibria).Lines represent thermodynamic calculation using FactSage, symbols on diagrams are experimental data obtained in 2022-2023.

3. 7 .
Na-Fe-Si-O-S slag/matte/metal systemTo test the newly developed model for Na-Fe-S matte, an experimental series was designed targeting the distribution of sulfur, oxygen and sodium between slag and matte (coexisting with Fe metal and tridymite).The results are shown in Figure10.The agreement is reasonable and allows expanding this direction to analyze the distribution of impurity metals (Sb, Sn, Pb, and As) between slag and matte.

Figure 10 .
Figure 10.Example of experimental results and model predictions in the Na-Fe-Si-S-O system (slag/matte/Fe metal equilibria).Lines represent thermodynamic calculation using FactSage, symbols on diagrams are experimental data obtained in 2022-2023.

Table 1 .
Main phases and thermodynamic models used in the present study I , Fe II , Fe III , Pb II , Sn II , Sb III , As III , O II , S II ) Modified Quasichemical Model in Pair Approximation (MQMPA) Slag (Na +1 , [NaFe] +4 , Fe +2 , Fe +3 , Si +4 , Pb +2 , Sn +2 , Sn +4 , Sb +3 , As +3 )(O -2 , S -2 ) 5 g pellets and placed into a substrate.Several types of holding materials were used: sealed SiO2 ampoules, open SiO2 crucible, Pt-25%Ir foil, or Ir wire in a shape of spiral.The innovation of the present study was the use of rhenium foil as a new type of holding material.It is less stable than other PGMs at oxidizing conditions, but has a high melting temperature It was even possible to re-use rhenium foil for several experiments within the same chemical system.Fun fact: Re was the last stable element to be discovered in 1925.
(second only to W), less expensive, relatively ductile (compared to Ru, Rh, Os, Ir), and more stable against corrosion by liquid metals (Pb, Sn, Sb) at low oxygen partial pressure.The use of rhenium foil allowed successful experiments in the Fe-Sb-Si-O, Na-Sn-Si-O systems equilibrated with Sb or Sn metals.