Microscopic Thermal Behavior of Iron-Mediated Platinum Group Metal Capture from Spent Automotive Catalysts

Abstract

This research investigates the micro-mechanisms and process control associated with the recovery of platinum group metals (PGMs) from spent automotive catalysts (SACs) through iron capturing. High-temperature smelting experiments, complemented by SEM-EDS and XRD analyses, demonstrate that PGMs spontaneously migrate from the slag phase to the iron phase, driven by interfacial energy, where they are captured to form alloy droplets with a PGM content exceeding 4 wt.%. The composite flux (CaO/H₃BO₃) markedly diminishes slag viscosity and enhances the density differential between slag and metal. This facilitates the aggregation, sedimentation, and separation of alloy droplets in accordance with Stokes’ law, thereby lowering the effective capture temperature from 1700 °C to 1500 °C and reducing energy consumption. Additionally, the flux inhibits the formation of detrimental Fe-Si alloys. PGMs form substitutional solid solutions that are uniformly dispersed within the iron matrix. This study provides both the theoretical and technical foundations necessary for the development of efficient, low-energy processes aimed at capturing and recovering Fe-PGMs alloys.

1. Introduction

Platinum group metals (PGMs), primarily including platinum (Pt), palladium (Pd), and rhodium (Rh), are indispensable in vital industries such as automotive exhaust treatment, petrochemicals, aerospace, electronics, and biomedicine due to their exceptional catalytic activity, notable chemical stability, and high-temperature resistance [1,2]. However, the global distribution of PGM resources is uneven, with over 90% of reserves concentrated in a limited number of countries, including South Africa and Russia [3]. This disparity has led to significant resource shortages and supply chain risks for numerous nations, including China [4]. Concurrently, rapid industrialization worldwide has resulted in substantial quantities of spent PGM-containing catalysts [5]. The PGM content and grade in these spent catalysts substantially exceed those found in primary ores [6], rendering them highly valuable “urban mines” for recovery. Spent automotive catalysts (SACs), owing to their multiple heavy metal constituents, are classified as hazardous solid waste in most countries [7]. If stored in open piles without proper treatment, they not only occupy and waste extensive land resources but also pose severe environmental pollution risks [8,9,10]. Even minimal landfill disposal can lead to various environmental and management challenges, including heavy metal leaching, difficulties in collecting leachate, and clogging of the drainage system [11]. The efficient and environmentally sustainable recovery of PGMs from spent catalysts is of paramount importance for protecting national resource security and reducing external reliance. Furthermore, it supports the objectives of a circular economy and the “dual carbon” goals, delivering significant economic and environmental advantages [12].

Currently, mainstream methods for recovering PGMs from SACs can be categorized into three primary categories: hydrometallurgy, pyrometallurgy, and biometallurgy [13,14,15,16]. Among these, iron capturing stands out as an effective pyrometallurgical technique, garnering significant attention due to its relatively straightforward process, high throughput, and environmental advantages [17,18,19]. This method principally involves adding iron collectors to the molten system. By leveraging the strong affinity between iron and PGMs, it concentrates nanoscale PGM particles dispersed within the slag into denser iron alloy droplets, thereby facilitating PGM recovery through slag–metal separation [20]. However, PGMs in SACs present considerable challenges for efficient separation and recovery due to their low concentration (1000–2000 g/t), small particle size (1–10 nm), and high dispersity within the carrier. The addition of iron collectors alters the composition, physicochemical properties, and slag-to-metal ratio of the molten system, which in turn influences the form of PGMs in the melt and their distribution behavior between the slag and metal phases [21]. Hence, an in-depth investigation into the capture mechanism of iron collectors for PGMs is of great significance for practical applications and guidance. Such research is crucial for optimizing smelting process parameters and collector addition strategies, enhancing capture efficiency, and ultimately advancing the industrial application of iron scavenging for PGM recovery from end-of-life automotive catalysts (SACs). Iron capturing, while effective, also presents certain limitations. Its high energy consumption due to elevated smelting temperatures, potential formation of undesirable Fe-Si alloys, and dependency on flux agents for efficient slag–metal separation are notable challenges. Nevertheless, its high throughput and compatibility with large-scale operations make it a promising route for PGM recovery from SACs.

Scholars worldwide have extensively examined the mechanism of iron capturing. For instance, transmission electron microscopy has been employed to analyze the distribution of platinum group metals (PGMs) on alloy matrices and to determine their crystal growth orientations [22]; research has investigated PGMs in iron-captured spent catalysts from the perspectives of electronegativity and crystal structure between PGMs and base metals [23]; simulations of the Johnson–Matthey process for recovering PGMs from spent automotive catalysts (SACs) have been carried out to elucidate the iron-capturing mechanism [17]; and Kirichenko et al. [24] studied the effects of substituting copper with iron as a capturing agent on PGM recovery rates from SACs. The majority of the studies above have focused on theoretical aspects such as crystal structure and electronegativity, in addition to analyzing the capture mechanism of PGMs in SACs by iron through assessments of alloy phase composition and PGM recovery rates. However, there exists a deficiency in systematic, intuitive experimental observations and detailed mechanistic explanations concerning the micro-dynamic migration of PGMs from slag to iron collectors during smelting, the pivotal interfacial reactions, and the real-time influence of slag physicochemical properties on collection efficiency. This knowledge gap impedes further optimization and the large-scale application of iron-based collection methodologies.

To address these issues, this study systematically investigated the dynamic mechanism of iron-mediated PGM capture through a series of precisely timed SAC high-temperature smelting experiments. Temporal microstructural characterization of smelting products at different capture times using SEM-EDS technology directly revealed the microscopic pathways and enrichment behavior of PGM nanoparticles migrating from the slag phase to the iron-capture agent. Building upon this, thermodynamic calculations of slag viscosity and density combined with Stokes’ law quantitatively explained how slag properties regulate droplet coalescence, growth, and sedimentation separation processes. This aims to establish a comprehensive mechanistic chain from microscopic migration to macroscopic separation, providing a solid theoretical foundation and measurable design criteria for developing new, low-energy, highly efficient industrial iron-capture processes.

2. Principal Analysis

2.1. Thermodynamic Calculations

Based on the composition of SACs, the subsequent reactions are presumed to take place during the smelting and capturing process:

$${\mathrm{P}\mathrm{t}\mathrm{O}}_2{=\mathrm{P}\mathrm{t}+\mathrm{O}}_2$$

(1)

$$\mathrm{P}\mathrm{t}{\mathrm{O}}_2+\mathrm{C}=\mathrm{P}\mathrm{t}+\mathrm{C}\mathrm{O}$$

(2)

$$\mathrm{P}\mathrm{t}{\mathrm{O}}_2+2\mathrm{F}\mathrm{e}=\mathrm{P}\mathrm{t}+2\mathrm{F}\mathrm{e}\mathrm{O}$$

(3)

$$\mathrm{P}\mathrm{t}{\mathrm{O}}_2+4\mathrm{F}\mathrm{e}\mathrm{O}=\mathrm{P}\mathrm{t}+2\mathrm{F}{\mathrm{e}}_2{\mathrm{O}}_3$$

(4)

$$\mathrm{P}\mathrm{t}{\mathrm{O}}_2+4\mathrm{F}{\mathrm{e}}_3{\mathrm{O}}_4=\mathrm{P}\mathrm{t}+6\mathrm{F}{\mathrm{e}}_2{\mathrm{O}}_3$$

(5)

$$\mathrm{P}\mathrm{t}{\mathrm{S}}_2+2{\mathrm{O}}_2=\mathrm{P}\mathrm{t}+2\mathrm{S}{\mathrm{O}}_2$$

(6)

$$\mathrm{P}\mathrm{t}\mathrm{S}+{\mathrm{O}}_2=\mathrm{P}\mathrm{t}+\mathrm{S}{\mathrm{O}}_2$$

(7)

$$\mathrm{P}\mathrm{t}{\mathrm{S}}_2+\mathrm{C}+3{\mathrm{O}}_2=\mathrm{P}\mathrm{t}+2\mathrm{S}{\mathrm{O}}_2+\mathrm{C}{\mathrm{O}}_2$$

(8)

$$\mathrm{S}\mathrm{i}{\mathrm{O}}_2+2\mathrm{C}=\mathrm{S}\mathrm{i}+2\mathrm{C}\mathrm{O}$$

(9)

$$\mathrm{S}\mathrm{i}{\mathrm{O}}_2+2\mathrm{C}\mathrm{O}=\mathrm{S}\mathrm{i}+2\mathrm{C}{\mathrm{O}}_2$$

(10)

$$\mathrm{S}\mathrm{i}{\mathrm{O}}_2+\mathrm{C}\mathrm{a}\mathrm{O}=\mathrm{C}\mathrm{a}\mathrm{S}\mathrm{i}{\mathrm{O}}_3$$

(11)

$$3\mathrm{S}\mathrm{i}+2\mathrm{F}{\mathrm{e}}_2{\mathrm{O}}_3=3\mathrm{S}\mathrm{i}{\mathrm{O}}_2+4\mathrm{F}\mathrm{e}$$

(12)

$$\mathrm{S}\mathrm{i}+\mathrm{F}\mathrm{e}=\mathrm{F}\mathrm{e}\mathrm{S}\mathrm{i}$$

(13)

$$\mathrm{S}\mathrm{i}{\mathrm{O}}_2+2\mathrm{F}\mathrm{e}\mathrm{O}=\mathrm{F}{\mathrm{e}}_2\mathrm{S}\mathrm{i}{\mathrm{O}}_4$$

(14)

$$\mathrm{F}{\mathrm{e}}_2\mathrm{S}\mathrm{i}{\mathrm{O}}_4+2\mathrm{C}=2\mathrm{F}\mathrm{e}+\mathrm{S}\mathrm{i}{\mathrm{O}}_2+2\mathrm{C}\mathrm{O}$$

(15)

$$\mathrm{F}{\mathrm{e}}_2\mathrm{S}\mathrm{i}{\mathrm{O}}_4+2\mathrm{C}+\mathrm{C}\mathrm{a}\mathrm{O}=2\mathrm{F}\mathrm{e}+\mathrm{C}\mathrm{a}\mathrm{O}\cdot \mathrm{S}\mathrm{i}{\mathrm{O}}_2+2\mathrm{C}\mathrm{O}$$

(16)

The Gibbs free energy changes (ΔG) for reactions (1)–(16) were calculated using HSC Chemistry 6.0. The calculations were performed under standard state conditions (1 atm), with all reactants and products considered in their pure forms and unit activity. The built-in HSC 6.0 Thermochemical Database, containing data for oxides, metals, sulfides, and gases, was employed for all species. The Gibbs energy minimization routine was used to compute ΔG across a temperature range of 0–1600 °C, which is the upper limit of the database. For analysis at 1700 °C, the values were estimated by linear extrapolation of the calculated trends. The metal phase was treated as pure Fe, the slag was represented by its major oxide components (SiO₂, Al₂O₃, CaO, and MgO), and the gas phases were assumed to be ideal. In Figure 1, solid lines represent calculated data within the database range, while dashed lines indicate extrapolated values for temperatures exceeding 1600 °C.

Using platinum as an example, although it has excellent oxidation resistance, its strong tendency to adsorb oxygen on its surface means it can still oxidize under certain conditions. As shown in Figure 1, platinum readily oxidizes in air at temperatures below 500 °C. Conversely, above 500 °C, the formed oxides spontaneously reduce back to metallic platinum. This shows platinum’s tendency to oxidize at low temperatures and decompose at high temperatures. The typical operating temperature of automotive exhaust catalysts is around 1150 °C, which is well above the oxidation temperature range of platinum. As a result, in spent automotive catalysts (SACs), some platinum forms inert compounds through oxidation and sulfidation, making it hard to create alloy phases with base metal collectors during later smelting and recovery processes. To effectively recover platinum, it must be converted back into its metallic form. The Gibbs free energy changes ($\mathsf{\Delta }{\mathrm{G}}^{\circ }$) for reactions (2) to (8) at standard conditions are negative at temperatures of 0 °C and higher. This shows that C, Fe, FeO, and Fe₂O₃ can spontaneously reduce platinum oxides and sulfides. Therefore, at the specified smelting temperature, platinum in SACs can remain stable in its metallic elemental form.

Special attention must be paid to the fact that at elevated temperatures of 1700 °C and under reducing conditions, reactions (9) and (10) can lead to the reduction of SiO₂ to elemental silicon, which subsequently reacts with Fe to form ferrosilicon alloy (FeSi). It should be noted that Figure 1 displays ΔG up to 1600 °C, while the experimental temperature of 1700 °C is beyond the plotted range. Extrapolation based on the trend suggests that reactions (9) and (10) remain favorable at higher temperatures. The formation of this alloy effectively transforms the iron encapturing process into ferrosilicon encapturing. Platinum-containing ferrosilicon alloy poses considerable challenges in subsequent hydrometallurgical separation processes. Therefore, it is imperative to meticulously control the smelting temperature within an appropriate range to prevent the reduction of SiO₂ to elemental silicon. Based on Reactions (11) to (16), silicon readily reacts with Fe to form ferrosilicon alloy during the smelting process. Concurrently, FeO reacts with SiO₂ in the slag phase at high temperatures to generate iron olivine. In the presence of carbon reducing agents and CaO, this iron olivine can further convert into metallic Fe and wollastonite. Although FeSi alloy formation is undesirable, it may be reduced by carbon and CaO under prolonged smelting conditions, as indicated by reactions (15) and (16). However, this process is kinetically limited and may not fully prevent Si dissolution into the iron phase.

2.2. Migration Drivers of PGMs

During the migration of PGMs, alterations in the system’s free energy are primarily influenced by surface energy. To gain a comprehensive understanding of the migration behavior of PGMs during capturing, it is crucial to analyze their surface tension properties across various phases. According to prior research on inclusion theory [25,26], PGM particles within the molten system can be regarded as spherical inclusions with radius s. Their spatial distribution at a given time is depicted in Figure 2:

At the melting temperature, PGMs in SACs can be reduced to their metallic state. Thermodynamic calculations show that the Gibbs free energy change for the relevant reduction reactions becomes negative, indicating that PGMs exist in a thermodynamically stable metallic form. When the system is fully melted, and PGM particles migrate from the slag to the metallic iron phase, the change in free energy of the system mainly arises from the change in interfacial energy. This process can be represented by Equation (17):

$$\Delta E=\left({\sigma }_{\mathrm{PGMs}\text{-}\mathrm{metal}}-{\sigma }_{\mathrm{P}\mathrm{G}\mathrm{M}\mathrm{s}\text{-}\mathrm{s}\mathrm{l}\mathrm{a}\mathrm{g}}-{\sigma }_{\mathrm{slag}\text{-}\mathrm{metal}}\mathrm{c}\mathrm{o}\mathrm{s}\alpha \right)\times 2\pi l\mathrm{d}\mathrm{s}$$

(17)

In the equation, σ denotes the interfacial tension. Here, “metal” specifically refers to the iron collector phase, while the slag primarily consists of SiO₂, Al₂O₃, CaO, and MgO. Given that the interfacial tension between the iron phase and the slag (${\mathsf{\sigma }}_{\mathrm{slag}\text{-}\mathrm{metal}}$) is significantly larger than that between PGMs and iron (${\mathsf{\sigma }}_{\mathrm{PGMs}\text{-}\mathrm{metal}}$) or between PGMs and slag (${\mathsf{\sigma }}_{\mathrm{P}\mathrm{G}\mathrm{M}\mathrm{s}\text{-}\mathrm{s}\mathrm{l}\mathrm{a}\mathrm{g}}$), the condition ΔE < 0 always holds. This confirms that the migration of PGMs toward the iron phase is a thermodynamically spontaneous process, driven by a reduction in the system’s overall interfacial energy.

Furthermore, from a microscopic electronic structure perspective, delocalized electron theory strongly supports this phenomenon. Chemical bonds in the slag (e.g., Si-O, O-M, where M represents metal ions) exhibit localized electron characteristics, whereas metallic bonds in the iron phase demonstrate electron delocalization. The valence electrons of platinum group metals (PGMs) exist in a dangling bond state, rendering them unable to form effective bonds with the localized electrons in the slag phase. However, they can form bonds with metal atoms such as iron, facilitating their entry into the iron metallic phase and thereby reducing the system’s overall free energy. Consequently, the spontaneous migration of PGMs from the slag to the metal phase is an inevitable outcome as the system endeavors to attain its lowest energy state.

3. Materials and Methods

3.1. Materials

The primary raw material for this experiment was pretreated spent automotive three-way catalytic converter (SAC) powder supplied by a precious metal recycling company in Jiangsu Province. The sample was ground to a particle size of less than 74 μm (with over 90% of particles below this size) and dried at 105 °C for later use. Its main chemical composition is shown in

Table 1. Chemical compositions of SAC used for the experiments.

SiO2	Al2O3	CaO	MgO	MnO	K2O	TiO2	ZrO2	TFe	Pt/Pd/Rh a	LOI b
34.6	31.0	1.7	8.9	0.6	0.3	0.7	4.1	2.5	97.0/642.0/118.0	12.6

TFE: Total Fe; ^a g/t; ^b Loss of mass in ignition.

, and the X-ray diffraction (XRD, D8 Advance, Bruker AXS Co., Ltd., Karlsruhe, Germany) pattern is shown in Figure 3. It can be seen that the SAC carrier mainly consists of SiO₂ and Al₂O₃, with the dominant phase being spinel (2MgO·2Al₂O₃·5SiO₂). ICP-MS (EXPEEC700, Focused Photonics Inc., Beijing, China) analysis of the sample’s PGM concentrations showed Pt, Pd, and Rh contents of 97, 642, and 118 g/t, respectively. The fluxing agents (CaO and H₃BO₃) and iron collector (iron powder) used were all of analytical grade, which were produced by Sinopharm Chemical Reagent Co., Ltd., Beijing, China. Before use, the samples were thoroughly mixed and dried at 105 °C for 24 h.

3.2. Experimental Procedures

This experiment was conducted in a medium-frequency induction furnace (XZ 25, Shanghai Xiangda Co., Ltd., Shanghai, China) with a rated power of 25 kW under standard atmospheric conditions. To facilitate the observation of PGM distribution within the melt, fluxing agents were generally not added in most experiments. The detailed setup of the experiment is shown in

Table 2. The mixtures of materials used in capture mechanism experiments.

Sam	SAC/g	Fe/g	(CaO/H3BO3, g)	Temperature/°C	Time/Min
2-1	50	0	0/0	1700	40
2-2	50	2.5	0/0	1700	0.5
2-3	50	2.5	0/0	1700	1
2-4	50	2.5	0/0	1700	3
2-5	50	2.5	0/0	1700	10
2-6	50	2.5	0/0	1700	20
2-7	50	2.5	0/0	1700	40
2-8	50	2.5	30/15	1500	40

Note: Carbon (as a graphite crucible) served as the reducing agent. The temperature of 1700 °C was selected based on preliminary trials to ensure complete melting and effective capture, despite Figure 1 being limited to 1600 °C. Extrapolation of ΔG trends supports the feasibility of reduction at 1700 °C.

. All raw materials were dried at 105 °C for 12 h until constant mass was achieved, ensuring complete removal of free moisture.

The experimental procedure is shown in Figure 4. First, accurately weigh 50 g of the SAC sample. Based on the experimental design, decide whether to add an iron collector or a fluxing agent, then place it in a graphite crucible with an inner diameter of 40 mm and a height of 80 mm. Melt the sample quickly in a preheated 1700 °C induction furnace. After maintaining at the set temperature for the required time, promptly transfer the crucible to a 600 °C box-type resistance furnace. It is then cooled at a constant temperature for 2 h to relieve internal stresses, aiding in later sample preparation. After the isothermal process, the graphite crucible is removed and allowed to cool slowly to room temperature in air. The condensed product is poured out to check if a metallic phase has formed, and all output is weighed and sampled. Finally, select a cooled, lumpy slag phase sample. Grind it sequentially using 400-grit, 600-grit, 800-grit, and 1200-grit sandpaper. Then, perform fine polishing with diamond polishing compound. After metal spraying, the sample is ready for microstructural analysis with scanning electron microscopy–energy-dispersive spectroscopy (SEM, Ultra Plus, Zeiss, Heidenheim, Germany, EDS, X-Max 50, Oxford Instruments, Oxford, UK).

3.3. Characterization

This study employed inductively coupled plasma mass spectrometry (ICP-MS, EXPEEC700, Focused Photonics Inc., Beijing, China) to quantify the PGM content in raw materials and reaction products. Additional components were analyzed utilizing X-ray fluorescence spectroscopy (XRF, ZSX PrimusII, Rigaku Corporation, Tokyo, Japan). Phase characterization was conducted using an X-ray diffractometer (XRD, D8 Advance, Bruker AXS Co., Ltd., Karlsruhe, Germany). Microstructure examination was performed with a scanning electron microscope (SEM, Ulcapturelus, Carl Zeiss Microscopy GmbH, Oberkochen, Germany), complemented by elemental analysis of micro-regions via X-ray energy-dispersive spectroscopy (EDS). In toxicity leaching assessments, the concentrations of heavy metal ions in the leachate were determined via an inductively coupled plasma optical emission spectrometer (ICP-OES, ICP-5000, China Polytech Co., Ltd., Beijing, China). Furthermore, the thermodynamic calculation software FactSage 7.2 (version 7.2, Thermfact/CRCT and GTT-Technologies, Montreal, BC, Canada and Ahern, Germany) was employed, with the Viscosity and Equilib modules selected to simulate and theoretically evaluate the viscosity variations in the slag and the associated reaction processes during the smelting-capture procedure.

4. Results and Discussion

4.1. The Dynamic Evolution of Iron-Capturing PGMs

Based on the aforementioned thermodynamic principles, we systematically reconstructed the entire dynamic process of iron capturing platinum group metals (PGMs) through a series of successive smelting experiments. Figure 5 presents scanning electron microscope (SEM) images of SAC melted at 1700 °C, with and without the inclusion of a capturing agent. Specifically, Figure 5a depicts an SEM image of slag obtained after 40 min in the molten state without the addition of any capturing or fluxing agents. The PGMs are observed to remain uniformly dispersed as nanoscale particles within the slag matrix, with no notable aggregation or separation. Additionally, only limited precipitation of refractory minerals such as TiO₂ was detected.

The inclusion of metallic collectors is essential for the efficient recovery of PGMs. This is evidenced by the clearly phased evolution of the capture process and the associated microstructural modifications following the introduction of the iron collector, as illustrated in Figure 5. During the initial melting and contact stage (0.5 min), iron powder melts to form minute microspheres that are uniformly dispersed throughout the slag phase, as shown in Figure 5b. At this stage, PGMs have not yet been captured. As shown in

Table 3. Regional EDS analysis of slags obtained from capture mechanism experiments (mass%).

Rig	O	Si	Al	Ca	Mg	Ti	Ce	Fe	Zr	Pt	Pd	Rh
1	46.60	10.87	11.72	3.88	25.27	-	1.65	-	-	-	-	-
2	7.86	1.88	2.25	-	-	82.44	-	-	-	-	-	-
3	-	6.10	0.17	-	-	-	-	93.73	-	-	-	-
4	-	6.64	-	-	-	-	1.66	85.08	2.39	0.08	4.04	0.11
5	-	22.47	-	-	-	4.68	0.10	66.58	6.16	-	-	-

, Energy-Dispersive Spectroscopy (EDS) analysis indicates that these iron droplets primarily consist of metallic iron with a minor proportion of silicon. The iron droplets exhibit a significant increase in size during the effective capturing and enrichment stage (1 min), as illustrated in Figure 5c, with smaller PGM particles commencing to appear within. The PGM content in the iron droplets surpassed four weight percent, as confirmed by EDS analysis, indicating successful migration of PGMs into the droplets. PGMs within the slag infiltrate the iron microspheres as minute particles and coalesce with iron droplets to form larger alloy phases, as visually evidenced in the EDS area scan data presented in Figure 6. When the smelting capture duration was extended to three minutes, some alloy droplets containing PGMs aggregated and enlarged, subsequently settling and separating under the influence of gravity, as depicted in Figure 5d. Iron microspheres devoid of PGMs remained embedded within the slag matrix. As the smelting duration increased from 10 to 40 min, alloy droplets continued to coalesce and grow, as shown in Figure 5e–g. Nevertheless, some iron microspheres remained encapsulated within the slag phase due to the absence of flux additions, which increased the slag’s viscosity and hindered complete slag–metal separation.

4.2. Multidimensional Regulatory Effects of Flux Agents

The previously referenced experiments demonstrate that systems lacking fluxing agents are characterized by elevated slag viscosity, incomplete separation, and a propensity for forming deleterious phases. To address these challenges, this study has implemented a CaO/H₃BO₃ composite fluxing agent. The selection of this composite flux is based on its synergistic effects: CaO effectively reduces the slag melting point and stabilizes the silicate network. At the same time, H₃BO₃ significantly decreases viscosity and promotes slag fluidity, thereby enhancing droplet coalescence and settling. This facilitates comprehensive optimization of the smelting process.

Quantitative analyses demonstrate that the incorporation of fluxing agents diminishes the slag density from 3.08 g/cm³ to 2.52 g/cm³, thereby elevating the slag–metal density disparity by 19.1% (from 3.82 to 4.55 g/cm³). Concurrently, as illustrated in Table 6, the slag viscosity decreased to 0.381 Pa·s at 1500 °C, markedly alleviating mass transfer resistance. According to Stokes’ law (Equation (18) [27]), the combined augmentation in density differential and reduction in viscosity enhances the ease with which the alloy phases settle out.

$$\mathsf{\nu }={\displaystyle \frac{\mathrm{g}{\mathrm{d}}^2\left({\mathsf{\rho }}_{\mathrm{A}}-{\mathsf{\rho }}_{\mathrm{S}}\right)}{18\mathsf{\eta }}}$$

(18)

where v is the settling velocity of the alloy phase, cm/s; g is the gravitational acceleration, g/cm²; d is the diameter of the alloy droplet, cm; ρ_A is the density of the alloy droplet, g/cm³; ρ_S is the density of the slag, g/cm³; and η is the slag viscosity, Pa·s.

Therefore, the settling velocity of ferroalloy droplets primarily depends on the droplet diameter, slag viscosity, and the density difference between the slag and the metal. During the recovery process, because the platinum group metals (PGM) content in the alloy is low (<1 wt.%), the density of the ferroalloy is typically estimated using the density of pure iron. The relationship between the density of pure iron and temperature can be calculated using Equation (19):

$${\rho }_{\mathrm{F}\mathrm{e}}=8.58-{\displaystyle \frac{0.853\times \left(T+273\right)}{1000}}$$

(19)

where ρ_Fe is the density of pure iron, g/cm³; T is the temperature, °C. Xin Jianjiang et al. [28] proposed a density calculation model for the SiO₂-Al₂O₃-CaO-MgO quaternary basic slag system, as shown in Equation (20):

$${\rho }_{\mathrm{S}}={\displaystyle \frac{M}{V}}$$

(20)

where ρ_S is the density of the slag, g/cm³; V is the total molar volume of oxides in the slag, cm³/mol; M is the total molar mass of oxides in the slag, g/mol. The value of V in Equation (20) can be calculated using Equation (21):

$$V=\sum X_{\mathrm{i}}{V}_{\mathrm{i}}$$

(21)

where X_i is the molar fraction of oxide component i in the slag, %; V_i is the partial molar volume of oxide component i in the slag, cm³/mol. The molar volume (V_i) for the majority of oxides can be determined utilizing Equation (22).

$$V_{\mathrm{i}}=V_{\mathrm{i},{1500}^{\circ }\mathrm{C}}+{\displaystyle \frac{V_{\mathrm{i},{1500}^{\circ }\mathrm{C}}\times \left(T-1500\right)}{10000}}$$

(22)

In the equation, $V_{\mathrm{i},{1500}^{\circ }\mathrm{C}}$ is the partial molar volume of oxide component i in the slag at 1500 °C, cm³/mol; V_i is the partial molar volume of oxide component i in the slag, cm³/mol; T is the absolute temperature, °C.

Table 4. Relationship between partial molar volume of different oxides and temperature.

Oxide	Partial Molar Volume/(cm3/mol)	References
SiO2	$27.516\times (1+{\displaystyle \frac{T-1500}{10000}})$	[28,29]
CaO	$20.7\times (1+{\displaystyle \frac{T-1500}{10000}})$	[28,29]
Al2O3	$28.3\times (1+{\displaystyle \frac{T-1500}{10000}})$	[28,29]
MgO	$16.1\times (1+{\displaystyle \frac{T-1500}{10000}})$	[29]
B2O3	$45.8\times (1+{\displaystyle \frac{T-1500}{10000}})$	[30]
TiO2	$19.65\times (1+{\displaystyle \frac{T-1500}{10000}})$	[31]
FeO	$15.8\times (1+{\displaystyle \frac{T-1500}{10000}})$	[32]
ZrO2	$19.65\times (1+{\displaystyle \frac{T-1500}{10000}})$	[4]

lists the partial molar volumes of the components included in the slag system at temperature T (unit: °C).

The principal components of the slag from Experiment 2-7 and the slag from Experiment 2-8 (following the addition of the appropriate fluxing agent) are presented in

Table 5. XRF results of slag obtained under different conditions (mass%).

Sam	SiO2	Al2O3	CaO	MgO	B2O3	TiO2	ZrO2	FeO
2-7	41.54	34.85	2.32	10.68	0	0.45	1.89	3.13
2-8	28.32	26.27	26.28	6.36	7.64	0.43	0.47	1.63

. The density differences between the two slags were calculated independently. Furthermore, the viscosities of both slags were determined utilizing the Viscosity module in FactSage 7.2, with the corresponding results exhibited in

Table 6. The viscosity of slag obtained from experiments 2-7 and 2-8.

Temperature/°C	Slag Viscosity/(Pa·s)
	2-7	2-8
1300	36.305	1.375
1350	19.156	0.966
1400	10.735	0.694
1450	6.341	0.510
1500	3.923	0.381
1550	2.527	0.290
1600	1.688	0.225
1650	1.163	0.176
1700	0.824	0.140

. The FeO content listed in Table 5 was derived from the TFe content in the slag.

More importantly, the fluxing agent reduced the necessary melting temperature for effective capture from 1700 °C to 1500 °C. This decrease in temperature efficiently prevented the reaction whereby SiO₂ reduces to elemental silicon at elevated temperatures, subsequently forming detrimental FeSi alloys with iron. SEM-EDS results depicted in Figure 7 indicate the presence of trace silicon within iron alloy particles encapsulated in the slag at 1700 °C in the absence of flux (Experiment 2-7). Conversely, following the addition of flux and a reduction in temperature to 1500 °C (Experiment 2-8), the slag matrix predominantly exhibited a glassy phase, devoid of visible metallic particles. XRD phase analysis presented in Figure 8 corroborates that the dominant phase in slag without flux is an iron–silicon alloy. In contrast, the inclusion of flux results in a phase primarily composed of metallic iron. This approach not only mitigates difficulties associated with subsequent hydrometallurgical separation but also promotes slag vitrification, thereby minimizing PGM losses attributable to inclusion.

4.3. State of PGMs in Iron Matrix

Figure 9 presents the phase diagram of Fe-PGM binary alloys [30]. This diagram illustrates that, at the melting temperatures employed in this investigation, higher PGM concentrations facilitate the formation of continuous solid solutions between PGMs and Fe, specifically (γ-Fe)Pt, (γ-Fe)Pd, and (γ-Fe)Rh. These solutions subsequently transform into the corresponding (α-Fe)Pt, (α-Fe)Pd, and (α-Fe)Rh solid solutions during the cooling process.

Zheng et al. [19] calculated the formation energies of platinum group metals (PGMs) present as substitutional or interstitial solid solutions within iron (Fe) matrices employing density functional theory based on first-principles calculations. The findings indicate that the formation energies are negative for both substitutional and interstitial solid solutions, confirming that the alloying process between Fe and PGMs is thermodynamically spontaneous. Notably, the formation energy for substitutional solid solutions is lower than that for interstitial counterparts, suggesting a tendency for PGMs to dissolve into the α-Fe lattice via substitution following cooling. Additionally, Ding Yunji [22] conducted high-resolution observations and analyses of PGM-containing iron alloys utilizing transmission electron microscopy. The results demonstrated a relatively homogeneous distribution of PGMs within the Fe matrix, with no distinct PGM precipitates observed. Only the body-centered cubic α-Fe phase was identified, further corroborating that PGMs do not form independent precipitates but rather dissolve within the Fe matrix.

As depicted in Figure 10a, even after a 40-min duration involving melting and cooling processes conducted without fluxing agents, a lumpy alloy can be obtained. Nonetheless, the elevated concentrations of SiO₂ and Al₂O₃ in the SAC raw materials elevated the melting temperature of the slag system and markedly increased its viscosity. This outcome not only impeded the effective capture of PGMs but also obstructed grain growth within the alloy phase. Consequently, the alloys produced in Experiments 2-7 exhibit numerous distinct grain boundaries. To further enhance the separation between the alloy and slag phases, as well as to promote proper grain growth, it is essential to employ higher melting temperatures and extend holding times.

Based on the results from the elemental surface scanning analysis, as shown in Figure 11, the alloy prepared in Experiment 2-7 exhibits distinct regions of silicon (Si) enrichment, with its distribution range closely overlapping with that of iron (Fe). This phenomenon primarily results from the high melting temperature of 1700 °C, during which SiO₂ in the slag is reduced to elemental silicon under a reducing atmosphere, subsequently dissolving into the iron phase to form a silicon–iron alloy. Simultaneously, the excessively high temperature also induces partial reduction of zirconium dioxide (ZrO₂), which enters the iron alloy phase. In contrast, the surface scanning results of the alloy obtained in Experiment 2-8 reveal no detectable distribution of Si and Zr elements, indicating that the lower melting temperature effectively suppresses the formation of ferrosilicon alloy and inhibits the reduction of other impurity oxides. This elemental distribution pattern aligns with the XRD phase analysis results of Experiments 2-7, as illustrated in Figure 8.

4.4. Iron-Capture Mechanism Model

As shown in Figure 12, the entire capture process can be summarized in two stages: first, the migration of PGM particles within the slag phase and their capture by iron-based precipitants, where captured PGMs dissolve into the iron matrix as substitutional solid solutions; subsequently, the separation and precipitation of the PGM–iron alloy phase from the slag phase. These two stages lack strict temporal boundaries and often occur concurrently in practice. Additionally, both stages involve the aggregation and growth of alloy particles. Due to the non-uniform distribution of PGMs and iron collectors within the slag, the capture process results in the rapid capture and aggregation of specific metal particles. These particles also exhibit accelerated grain growth. Upon reaching a critical sedimentation size, these particles begin settling first.

During descent, they continue capturing PGMs along their path while colliding with and aggregating surrounding alloy droplets, further enlarging until they pay at the crucible bottom to form significant alloy phases.

5. Conclusions

This study systematically examined the micro-mechanisms and process influences involved in the recovery of platinum group metals (PGMs) from spent automotive catalysts (SACs) through iron capturing. The research incorporated a series of smelting experiments, complemented by XRD, SEM-EDS characterization, and thermodynamic analyses. The key findings are as follows:

(1)

Without iron-based collectors, PGMs remain nanoscale-dispersed in the slag phase and cannot spontaneously aggregate and settle, confirming that adding metal collectors is essential for efficient recovery.

(2)

With the addition of collectors, PGMs tend to migrate toward the iron phase driven by interfacial energy. They are subsequently captured, resulting in the formation of alloy droplets with a platinum group metal (PGM) content exceeding 4 wt.%. Over time, these droplets undergo aggregation, growth, and ultimately settle and separate.

(3)

The flux-free system demonstrates notable disadvantages: the excessive reduction of SiO₂ at elevated temperatures (1700 °C) results in the formation of FeSi alloy; the high viscosity of the slag (η = 0.824 Pa·s) impedes complete slag–metal separation; and the precipitation of spinel phases obstructs slag vitrification.

(4)

The composite flux (CaO/H₃BO₃) facilitates multidimensional optimization by markedly decreasing slag viscosity and enhancing the density difference between slag and metal, thereby promoting droplet settling in accordance with Stokes’ law. Additionally, it lowers the effective capture temperature to 1500 °C to inhibit FeSi formation, assists in slag vitrification to minimize PGM inclusions, and results in a pure α-Fe-based solid solution alloy suitable for subsequent processing.

This study elucidates the mechanism of iron capturing, from microscopic migration to macroscopic separation, providing theoretical foundations and process optimization strategies for the development of low-energy, high-efficiency platinum group metal (PGM) recovery technologies. Under optimized conditions (1500 °C with CaO/H₃BO₃ flux), the recovery rate of PGMs exceeded 95%, as confirmed by ICP-MS analysis of the separated alloy phase. This study demonstrates that through the investigation of the microscopic behavior of iron-mediated PGMs, better information can be offered for industrial-scale recovery processes.