Batch Flotation of Lithium-Bearing Slag—A Special Focus on the Phase Properties of Engineered Artificial Minerals for Enhancing the Recycling of End-of-Life Lithium-Ion Batteries

Abstract

The increasing demand for lithium-ion batteries (LIBs) and the critical need for lithium make the efficient recycling of secondary resources essential. Synthetic Li-bearing phases, some with lithium contents greater than natural sources (e.g., spodumene), can occur in slags produced by the pyrometallurgical recycling of end-of-life LIBs. This study investigates both the composition of synthetic model slags reproducing LIB recycling and the recovery potential of Li-bearing phases using SEM-based automated mineralogy and batch flotation tests, respectively. In particular, the efficacy of a novel zwitterionic collector, punicine, in contrast to the conventional collector, oleic acid, was evaluated with a focus on recovering Li-aluminate as a key engineered artificial mineral (EnAM). The flotation tests demonstrated that punicine provided a higher degree of selectivity for Li-aluminate over gehlenite, along with improved recovery of fine and well-liberated particles. The enhanced performance is attributed to punicine’s unique frothing properties and phase-specific interactions. Our findings highlight punicine’s significant potential as a collector for lithium-bearing EnAMs to advance lithium recovery from complex slag materials. The applied unique methodology supports the study of reagent regimes in relation to the flotation behavior of EnAM phases and the sustainable recycling of LIBs.

1. Introduction

The demand for lithium-ion batteries (LIBs) has doubled in the past five years due to their importance as energy storage systems for renewable energy and their utilization in the global shift to electric mobility [1]. As a consequence, lithium was declared to be a critical element by the U.S. Department of Energy and the European Commission in 2020 [2]. Thus, it is highly necessary to secure recycled lithium sources from LIBs and feed them into a more resource-efficient and sustainable circular economy to allow for the necessary shift to renewable energies and lower carbon emissions.

Flowsheets for LIB recycling employ a number of new and dynamic recycling strategies, which include mechanical, hydrometallurgical and pyrometallurgical treatments. Industrial-scale LIB recycling facilities, such as the Umicore battery recycling plant in Hoboken, Belgium, utilizes hydrometallurgical and pyrometallurgical processes to recover metals, such as nickel, copper and cobalt. However, major amounts of rather ignoble elements like lithium and manganese are commonly lost in the slag and are rarely recovered. Slags obtained from this recycling route generally contain higher concentrations of Li₂O (5–20)% (w/w) in comparison to natural phases, such as spodumene (8% (w/w) Li₂O [3]. In addition to being sources of Li [4] and Mn [5], slags originating from other processes are known to contain valuable concentrations of copper [6,7,8,9] and rare earth elements [10,11]. Although slags can contain significant concentrations of valuable metals, the complexity of slag phases and microstructures hinders the beneficiation and extraction processes [12]. In this contribution, a novel approach that involves the beneficiation of slag residues from recycled LIBs is presented. The solution is to engineer specific phases in synthetic slags in order to analyze the concentration of valuable elements in a limited number of well-defined phases within a model system. This approach enables the identification of key parameters for a slagging process that can effectively resolve microstructural challenges [13]. The creation of engineered artificial mineral phases (EnAMs) is accomplished by adjusting the concentrations of certain slag additives and cooling rates during solidification [7,14]. The adjustment of these parameters can allow for the precise manipulation of crystal shapes and sizes, thereby increasing downstream beneficiation and mineral extraction efficiencies. When examining slag from Umicore’s pyrometallurgical recycling process, Elwert et al. [4] found that Li crystallizes primarily within Li-aluminate (LiAlO₂) and that gehlenite (Ca₂Al₂SiO₇) is the dominant silicate gangue phase. Li-aluminate is a particularly important phase, not only because it represents the most abundant Li-bearing EnAM, but it also contains the highest molar ratio of Li and thus represents the primary target phase in Li-bearing slags [15]. Although studies relating to the regulation of the crystal shape and size of Li-aluminate have been carried out [16], the crystallization sequence of different EnAMs is not straightforward since it is dependent on the type and cathode chemistry of the recycled batteries. For example, varying the manganese (Mn) content can negatively influence Li-aluminate formation and instead promote the formation of spinel-type phases [5]. This was observed in second-generation LIB slags, which contained additional manganese ions, and resulted in the formation of diverse EnAMs and corresponding crystallization textures [4].

For slag beneficiation, froth flotation, which is regarded as one of the most important separation techniques in mineral processing, can be applied to Li-bearing EnAMs. Flotation reagents, such as collectors, depressants and activators, are crucial drivers for the separation efficiency in froth flotation. The hydrophobic particles will adhere to air bubbles and are recovered as concentrates, while hydrophilic particles will remain suspended in the aqueous phase and are collected as tailings. To date, there has been limited research on the flotation behavior of Li-aluminate and gehlenite, and information regarding their flotation response to various collector regimes is scarce [17,18,19,20,21]. Although pure model materials have been used to investigate potential collectors [21], the flotation behavior of Li-aluminate in real and synthetic Li-slags is poorly understood. Knowledge gained from the flotation of natural spodumene can serve as a valuable reference for the behavior of lithium-bearing minerals in flotation [22,23,24,25,26,27,28,29]. A prominent benchmark anionic collector, oleic acid, or its deprotonated form, sodium oleate, is widely used for spodumene flotation [22,23,27,29]. Sodium oleate was already successfully applied in the flotation of Li-aluminate in Li-slags [13]. Recently, a novel cyclic betaine collector, punicine, was successfully trialed within microflotation and showed similar recovery rates to sodium oleate [17,19]. Punicine is a collector whose charge varies depending on the surrounding pH conditions: at alkaline pH levels of 9–14, it is negatively charged and can function as an anionic collector. In the pH range of 5–9, the molecule exhibits both positive and negative charges, while at pH levels of 2–5, it predominantly carries a positive charge and acts as a cationic collector [30,31]. Additionally, it is a light-sensitive molecule that becomes activated into a radical state under different wavelengths of light. Both of these properties are currently being studied to understand their interactions with mineral surfaces, particularly Li-aluminate, with the objective of making use of punicine as a pH-responsive and photo-switchable collector [17,19,32]. In this study, we examine for the first time the application of the punicine collector in batch flotation tests for the selective recovery of Li-aluminate from a synthesized and engineered lithium-bearing slag. Additionally, the benchmark collector, oleic acid, is utilized for comparison. The flotation efficiency is discussed in the context of EnAM phase composition, phase association, phase liberation and particle size.

2. Materials and Methods

2.1. Sample Materials

A Li₂O-CaO-SiO₂-Al₂O₃-MgO-MnO_x synthetic slag [16] of composition analogous to a real LIB slag system was provided for this study by the IME Process Metallurgy and Metal Recycling group of the RWTH Aachen, Aachen, Germany. The slag production process is described for a smaller batch in more detail in Rachmawati et al. [16]. The specific feed composition can be seen in

Table 1. The feed composition in % (w/w), batch size and applied cooling rate after slagging of the artificial slag.

Batch Size	Cooling Rate	CaO	SiO2	Al2O3	MnO	Li2O
120 kg	25 K/h	16.9%	18.9%	45.3%	10.4%	8.5%

.

For the sample preparation, 80 kg of slag was crushed in a jaw crusher in 30 cm blocks at UVR-FIA GmbH, Freiberg, Germany. The material passed through the crusher three times until a particle size smaller than 10 mm was reached. Afterwards, the material was sent to a screen discharge ball mill, with a sieve size of 100 µm, and milled to a particle size smaller than 100 µm (mill designed and produced at FIA Research Institute, Freiberg, former GDR, [33]).

For fundamental studies on single phases present in the slag system, the model materials gehlenite and LiAlO₂ (Sigma-Aldrich, St. Louis, MO, USA) were purchased. The composition of the gehlenite, acquired from a Romanian mine, contained 58.2% (w/w) gehlenite, 40.6% (w/w) garnet and 1.2% (w/w) kaolinite (XRD analysis).

2.2. Reagents

Solutions of 0.01 g/mL of each oleic acid (OA, Figure 1b) (purity 90%, Sigma-Aldrich, St. Louis, MO, USA) and 1-(2′,5′-dihydroxyphenyl)-4-decylpyridinium chloride (decylpunicine (Pun); Figure 1a) (provided by A. Schmidt, Clausthal University of Technology, Institute of Organic Chemistry, Clausthal-Zellerfeld, Germany) as collectors and methyl isobutyl carbinol (MIBC) (technical-grade, Sigma-Aldrich, St. Louis, MO, USA) and pine oil (technical-grade, Carl Roth GmbH + Co. KG, Karlsruhe, Germany), as frothers were dissolved in ethanol (absolute-grade, Carl Roth GmbH + Co. KG, Karlsruhe, Germany).

2.3. Flotation

The froth flotation experiments were performed in a Partridge–Smith-type cell using the Dynamic Foam Analyzer 100 from KRÜSS GmbH, Germany, with a volume of 200 mL including a foaming filter FL4551 (KRÜSS GmbH, Hamburg, Germany) with a pore size of (12–15) µm and an inner diameter of 40 mm (Figure 2). Sample aliquots of 25 g were used in each experiment. Sample splitting was performed in rotary splitters.

The collector punicine has not been studied within batch flotation before. Thus, preliminary experiments were executed to screen and identify compatible frother reagents. Common frothers, such as MIBC, which is typically used in combination with oleic acid, or Montanol 800, (Clariant, Muttenz, Switzerland), are not able to build a stable froth in combination with punicine. Finally, the frother pine oil (35 g/t) was identified. The reagent is already known from applications in microflotation with punicine [17], in the flotation of Li-aluminate [20,34] and Li-slag flotation [13]. No pH modifier was used since a lack of froth formation was observed when using acids or NaOH with both punicine and oleic acid. Flotation was carried out with a mass fraction of 11% (w/w) solids to 200 mL tap water (Helmholtz Institute for Resource Technology, Freiberg, Germany). This rather low mass concentration of solids in the pulp was chosen based on experiences with material in the given size range and when using the Partridge–Smith 200 mL set-up (KRÜSS GmbH, Hamburg, Germany). In order to ensure the reproducibility of all experiments, the water was sourced from a 40 L sample of tap water taken from the tap in one day. For conditioning, the slag was suspended in 100 mL tap water and oleic acid or punicine was added in concentrations of 500 g/t, 750 g/t and 1000 g/t with a conditioning time of 5 min. The preparation, conditioning and flotation experiments with punicine were performed with special care to minimize light exposure. Afterwards, 70 g/t of MIBC or 35 g/t pine oil was added. The batch size froth flotation was carried out with an air flow rate of 0.8 L/min. Three concentrates (C) were collected after 30 s (C1), 60 s (C2) and after the froth stopped overflowing (C3). The remaining suspension in the flotation cell was balanced as tailing. After flotation, the concentrates and tailing were filtered and dried at 60 °C for approximately 48 h. Every collector reagent combination was tested with 3–5 repetitions depending on the reproducibility of mass and water pull. All results shown are the average of 3 repetitions excluding outliers. Wherever error bars are depicted, they stand for the 95% confidence interval.

2.4. Bubble Particle Attachment Tests

A set-up to evaluate the degree of hydrophobicity regarding a consequent change in bubble particle adhesion was arranged using a high-resolution camera (OCA 50, DataPhysics Instruments GmbH, Filderstadt, Germany), a cuvette (precision cell 50 mm, Hellma GmbH & Co. KG, Müllheim, Germany) filled with tap water and a gas-tight syringe with air (Hamilton^®, Hamilton Bonaduz AG, Bonaduz, Switzerland). In order to simulate flotation conditions, the powder materials (0.1 g) were suspended in tap water (100 mL), the pH was stabilized at pH 10.8 under the optional addition of 0.1 M NaOH for 5 min at 700 rpm, the collectors were added at a concentration of 2 × 10⁻⁴ M and subsequently the material was further conditioned for 5 min at 700 rpm. Afterwards, a bubble with a volume of 7 µL air was injected into the suspension still attached to the syringe and the suspension was stirred for a further 3 min. An image of the bubble was taken after at least 10 min of settling time. The bubble loading was quantitatively estimated by image analysis using ImageJ 1.8.0 [35] and Equation (1) and is depicted in Figure 3, and the average of three repetitions per conditions and 95% confidence interval was used for discussion.

Bubble loading (in %) = A_p/(A_b − A_c) × 100

(1)

where A_b represents the bubble area, A_c is the bubble area hidden by the cannula and A_p represents the bubble area covered by particles. The A_b and A_p calculation assumed a perfect sphere; A_c was calculated as a perfect spherical cap. Note that the calculation is solely based on the covered area of a bubble assuming a particle monolayer, which does not take into account the density of the particle bulk attached to the bubble, nor the aggregates or multilayers [36]. To avoid an impact of Stokes particle velocity, the particle size of both model materials was chosen to reach similar Stokes particle velocity for Li-aluminate and gehlenite to reduce the particle size and density effect in flotation (Figure A5).

2.5. Analytical Methods

Each one of the following processing products represents combined samples of triplicates within each reagent combination.

2.5.1. Mineral Liberation Analyzer Measurements

A total of 22 processing products, including one feed sample, together with a single drill core sample of the slag (2 cm in diameter), were examined.

The sample preparation of processing products consisted of manually deagglomerating 1–2 g of sample material, mixing it with graphite and embedding the mix with epoxy resin. To avoid the influence of particle density, the first batch of epoxy mounts, containing milled material, were sliced vertically, rotated 90 degrees and remounted in epoxy resin (i.e., B-sections) [37,38]. The epoxy mounts were subsequently carbon-coated using a Leica MED 020 vacuum evaporator (Leica Microsystems GmbH, Wetzlar, Germany) to ensure the conductivity of the sample surface.

Mineral Liberation Analyzer (MLA) measurements were performed in a FEI Quanta 650F scanning electron microscope (SEM) (FEI Company, Hillsboro, Oregon, United States) equipped with two Bruker Quantax X-Flash 5030 EDS detectors (Bruker AXS, Karlsruhe, Germany) and FEI’s MLA Suite 3.1.4 for automated data acquisition. The electron beam accelerating voltage was 25 kV. For both sets of samples, the MLA measurement routine was GXMAP and the X-ray step size spacing was set to 6 µm. For the 22 processing products, a resolution of 700 pixels, horizontal field width of 700 µm and dwell time of 16 µs was used for the analyses. The reliability of results related to MLA experiments was ensured by mapping at least 200,000 particles per processing product achieved with the respective reagent combination. Due to uncertainties relating to the densities of EnAMs, the liberation data are reported in area % (% (A/A)). Nevertheless, estimated densities are utilized in Table 2 to allow for the comparison of XRD and MLA in weight % (%(w/w)) (

Table A1. Estimated densities of EnAM phases based on literature and gas pycnometry.

Phases	Density	Source
	g/cm3
Li-aluminate	2.62	[51]
Gehlenite	2.98	[52]
Eucryptite	2.67	[52]
Spinel	3.61	[53]
Mn-rich spinel	4.23	[54]
Mixed phase	3.1	density of the slag measured by gas pycnometry
Glaucochroite	3.4	[52]
Li2MnSiO4	3.10	[55]

).

In addition, backscattered electron (BSE) imaging and energy-dispersive spectroscopy (EDS) analyses were carried out on the drill core samples to investigate the textural characteristics and elemental compositions of the slag phases, respectively.

2.5.2. Powder X-Ray Diffraction Measurements

For the XRD measurements, a PANalytical Empyrean (Malvern Panalytical GmbH, Malvern, United Kingdom) X-ray diffractometer with a cobalt X-ray source, a voltage of 35 kV, a current of 35 mA and an iron filter on the primary beam for suppressing K-beta radiation were used. Prior to the measurement, the samples were split into 2.5 mL aliquots and ground in a McCrone mill (Retsch GmbH, Haan, Germany) with ethanol and zirconium-oxide grinding media to produce particles with a narrow size distribution below 10 µm and an intact crystallographic structure. Following this, samples were dried overnight and homogenized with small steel balls of 4 mm diameter in a MM 400 Retsch mixer mill (Retsch GmbH, Haan, Germany). Using the back-fill method, the powdered sample material was filled into the sample holders. The irradiated area on the sample surface was kept constant at 12 × 15 mm². The measurement was performed from 5 to 80° 2-theta at a step size of 0.0131° 2-theta. The overall measurement time was about 2 h 30 min per sample. After the phase identification with the ICDD PDF-4+ database, the quantitative phase determination was performed via Rietveld refinement using Profex 4.1.0 software. The results need to be considered with a lower detection limit of 1% (w/w) to 2% (w/w).

2.5.3. Particle Size Measurements

The particle size distribution of samples was determined with dry measurement via a HELOS laser diffractometer at an air pressure of 4 bar with the RODOS dispersion unit, both from Sympatec GmbH, Clausthal-Zellerfeld, Germany.

2.5.4. Zeta Potential Measurements

The measurements were conducted with a Zetasizer Nano (Malvern Panalytical GmbH, Malvern, UK). For the measurement, 0.05 g of material was suspended into 50 mL of 10 mM KCl aqueous solution. All materials were previously ground in a McCrone mill (Retsch GmbH, Haan, Germany) to guarantee a particle size of less than 2 µm. The milling procedure was the same used for preparing XRD samples and is described in more detail in Section 2.5.2. The pH was adjusted with HCl and KOH (0.01–0.1 M) starting from pH 3. After dispersion, samples were left undisturbed for 3 min and the supernatant fraction was used for measurements.

3. Results

3.1. EnAM Phases in the Synthesized Slag

3.1.1. Composition

The present slag is defined within the Li₂O-CaO-SiO₂-Al₂O₃-MnO_x system. The slag powder was analyzed for its phase composition using XRD and MLA. The results from these two independent techniques show good agreement (Table 2). In total, four main phases were found in the sample: Li-aluminate (LiAlO₂), gehlenite (Ca₂Al[AlSiO₇]), eucryptite (LiAl[SiO₄]) and alumina spinel-types. Of these phases, the most abundant consisted of spinel-types (36.1% (w/w)) and gehlenite (31.2% (w/w)). Only a small fraction of the sample (6.0% (w/w)) contained the target EnAM Li-aluminate. Additionally, eucryptite, another lithium-bearing phase, was present, with 11.2% (w/w) in the slag. In addition to the main phases, finely disseminated phases accounted for 14.1% (w/w). These phases, which were minute and finely intergrown (<1 µm), were not resolvable at the X-ray spacing (6 µm) and minimum feature size (~1.5 µm) used for the MLA measurements. Consequently, these phases are hereforth referred to as mixed phases. Additional trace phases, identified by MLA analyses (1.4% (w/w)), included glaucochroite, lithium–manganese silicate phase, a Cr-containing Mn-rich spinel, K-bearing phases and a Mn-rich gangue phase CaMn-aluminosilicate.

Table 2. Composition of slag material determined by MLA and XRD (% by weight). The error shows a 95% confidence interval for MLA and XRD.

Mineral	Li-Aluminate	Eucryptite	Gehlenite	AlMn-Spinel Type	Mixed and Trace Phases
Chemical formula	LiAlO2	LiAl[SiO4]	Ca2Al[AlSiO7]	Al, Mn, (Li)—oxide	varying (Si, Li, Mn, Al, Ca, Ti, Cr)
XRD in % by weight	7.9 ± 0.2	9.1 ± 0.1	32.5 ± 0.8	35.5 ± 0.6	14.1 ± 0.5 1
MLA in % by weight	6.0 ± 0.2	11.2 ± 0.3	31.2 ± 0.3	36.1 ± 1.7	15.5 ± 1.8

¹ main trace phases: Li₂MnSiO₄ (13.5 ± 0.4)% (w/w); glaucochroite (CaMnSiO₄) (0.7 ± 0.1)% (w/w).

Table 2. Composition of slag material determined by MLA and XRD (% by weight). The error shows a 95% confidence interval for MLA and XRD.

Mineral	Li-Aluminate	Eucryptite	Gehlenite	AlMn-Spinel Type	Mixed and Trace Phases
Chemical formula	LiAlO2	LiAl[SiO4]	Ca2Al[AlSiO7]	Al, Mn, (Li)—oxide	varying (Si, Li, Mn, Al, Ca, Ti, Cr)
XRD in % by weight	7.9 ± 0.2	9.1 ± 0.1	32.5 ± 0.8	35.5 ± 0.6	14.1 ± 0.5 1
MLA in % by weight	6.0 ± 0.2	11.2 ± 0.3	31.2 ± 0.3	36.1 ± 1.7	15.5 ± 1.8

¹ main trace phases: Li₂MnSiO₄ (13.5 ± 0.4)% (w/w); glaucochroite (CaMnSiO₄) (0.7 ± 0.1)% (w/w).

Further insights into mineral chemical compositions were obtained via MLA, where it was found that mineral chemical compositions varied from their ideal compositions as indicated by XRD. Energy-dispersive X-ray spectroscopy (EDS) analyses indicated that many of the phases contained low concentrations of impurities. Elemental maps, combined with BSE images of the major phases (Figure 4a), are shown in Figure 4 to highlight the distribution of manganese, aluminum and calcium within the different phases (Figure 4b–d).

For example, gehlenite contains minor amounts of manganese (<2% (w/w)) in its structure (Figure 4b). In contrast, Li-aluminate contains a minor amount of silicon (<5% (w/w)) (Figure A1), which was similarly found by Schirmer et al. with the derived sum formula Li_{1 − x}(Al_{1 − x}Si_x)O₂ [39].

Although the Mn content within eucryptite varies, two compositional variants were defined for the purpose of this study: (i) Mn-poor eucryptite (1.5% (w/w)), containing less than 2% (w/w) Mn, and (ii) the most abundant Mn-rich eucryptite (9.7% (w/w)), containing up to 11 (w/w) Mn, which accounts for the bulk of the total eucryptite concentration (Table 2). Elemental maps clearly illustrate the contrasting distributions of Mn and Al within these two variants. Lower Al concentrations within Mn-rich eucryptite may indicate substitution by Mn as the crystallization sequence progresses. Schirmer et al. [40] described a comparable subphase with Mn, as eucryptite-like lithium aluminosilicate with a formula of Li_{1 − x}(Mn³⁺_y,Al_{1 − (x + y)})_{1 − x}(Si_{1 + x})O₄. As with eucryptite, two spinel-types were observed in the slag: on the one hand, a Mn-poor aluminum spinel, with 34.4% (w/w) modal content in the sample and containing up to 11% (w/w) of Mn in its structure, and on the other hand, a Mn-rich aluminum spinel with only 1.8% (w/w) modal content, but with Mn content of up to 47% (w/w) in its structure. Similar phases have been derived in Wittkowski et al. [5] in form of a spinel solid solution (Li(_2x)Mn²⁺_{(1 − x)})_{1 + x})(Al_{(2 − z)},Mn³⁺_(z))O₄.

Two mixed phases were identified by the MLA. These included a Mn-aluminosilicate and a calcium-bearing CaMn-aluminosilicate. The Mn-aluminosilicate mixed phase is the most abundant mixed phase, comprising 10.9% (w/w), while the CaMn-aluminosilicate mixed phase accounts for 3.2% (w/w). Both mixed phases exhibit a high manganese content, ranging from (15 to 23)% (w/w) Mn. The elemental mapping of the mixed phases helped to derive their constituent EnAM phases (Figure 4). The mixed phases exhibit an even distribution of silicon (Figure A2a), while calcium is concentrated locally (Figure 4d and Figure A2b). Thus, Ca-rich minerals like gehlenite, or the trace phase glaucochroite (CaMn[SiO₄]) or CaMn-aluminosilicate, could represent possible mixing candidates. Mn and Al are also evenly distributed but form a porous network-like structure highlighting the possible occurrence of two phases (Figure 4b,c and Figure A2c). High concentrations of manganese are characteristic of Mn-rich spinel-types, in addition to other major spinel-types, glaucochroite and Mn-rich eucryptite (Figure 4b). Likewise, a high aluminum content is a characteristic trait of Li-aluminate and spinel (Figure 4d), whereas eucryptite contains relatively lower Al concentrations. Consequently, for mixed phases composed of Mn and Si, Li₂MnSiO₄, which was identified by XRD analysis, could constitute a major component, along with eucryptite, Mn-rich eucryptite and spinel. Notably, the total amount of mixed phases determined by MLA (14.1% (w/w)) is comparable to the trace phases of Li₂MnSiO₄ detected by XRD (13.5 ± 0.4% (w/w)).

3.1.2. Phase Liberation

Phase liberation by particle composition (% (A/A)) is frequently used to describe the recovery of one or more target phases in a particle context. The liberation criteria, as stated in

Table 3. Degree of liberation in % by volume of the ground flotation feed material of each phase based on % (A/A).

Liberation Class	Locked	Middlings	Liberated	Fully Liberated
% (A/A) in Particle	0% < x ≤ 30%	30% < x ≤ 80%	80% < x ≤ 100%	100%
Li-aluminate	9	41	24	26
Gehlenite	14	38	18	30
Eucryptite	52	35	4	9
Spinel	10	60	19	11
Mixed phases	39	39	7	15

, are defined as follows:

• Fully liberated: target phase accounts for 100% of particle area;
• Liberated: target phase accounts for 80% < x ≤ 100% of particle area;
• Middlings: target phase accounts for 30% < x ≤ 80% of particle area;
• Locked: target phase accounts for 0% < x ≤ 30% of particle area.

After crushing and milling of the feed material, approximately 50% of the target Li-aluminate is liberated, whereas 41% occurs within the middlings fraction and 9% within the locked fraction (Table 3, Figure A3). In total, 26% of Li-aluminate particles are 100% fully liberated, especially below a diameter of 9.6 µm, where 90 to100% of all Li-aluminate particles are liberated (Figure 5a). Gehlenite, the gangue phase, is well liberated, with 48% reporting to the liberated fraction. Overall, gehlenite shows a similar liberation behavior over particle size towards Li-aluminate, shown in Figure 5, as well as in Figure A4. Approximately 40% of these two phases occur within the middlings fraction. The mass of both phases is distributed similarly across the particle sizes present. As a result, the phases do not exhibit any specific or distinct particle or liberation properties that could facilitate separation between valuable minerals and gangue by means of classification. Li-aluminate is strongly associated with eucryptite (38%) and weakly associated with gehlenite (16%) (

Table A2. Association of slag phases with each other in particles of the feed based on MLA results.

Associated Phases	Li-Aluminate	Gehlenite	Eucryptite	Spinel	Mixed Phases	Trace Phases
Li-aluminate	0.00	2.33	5.19	4.10	2.25	1.47
Eucryptite	37.85	22.70	0.00	30.65	44.63	21.16
Gehlenite	16.43	0.00	21.90	43.97	29.06	21.85
Spinel	29.73	45.26	30.43	0.00	20.02	26.28
Mixed phases	14.66	26.91	39.87	18.01	0.00	29.24
Trace phases	1.32	2.80	2.61	3.27	4.04	0.00

). Although a large percentage of eucryptite occurs within the locked class (52%), it is generally associated with other phases, such as gehlenite (22%), spinel (30%) and mixed phases (40%), relative to Li-aluminate (Table A2). In contrast to gehlenite and Li-aluminate, spinel is characterized by a higher proportion of middlings (60%), which is consistent with its homogeneous distribution in the slag. In addition, spinel-containing particles are coarser (Figure 5b).

3.2. Flotation Attachment Abilities and Surface Characterization

The natural pH of the slag sample in aqueous solution is strongly alkaline, 10.7 ± 0.2. The natural pH of the model materials Li-aluminate and gehlenite are also alkaline, 10.8 and 9.8, respectively. Conversely, these phases differ in surface charge: gehlenite is less negatively charged (−3 mV) than Li-aluminate (−36 mV) at their respective natural pH levels (Figure 6). Furthermore, at the natural pH of the slag, Li-aluminate is more negatively charged (−36 mV) than gehlenite (−12 mV), but both exhibit an overall negative surface potential.

The isoelectric point (IEP) of Li-aluminate occurred at pH 7, while gehlenite’s IEP was found at pH 9.0. It has been observed that Li-aluminate particles strongly coagulate at pH levels near the IEP, specifically between pH 6 and 8.5, which could affect Li-aluminate flotation. This effect was not observed with the slag sample at any pH level. Instead, a gelation effect occurred below pH 5 resulting from the reaction between hydrochloric acid and slag, probably caused by silicic acid formation and consequential gelation of those compounds [13,41].

The floatability of model materials with the presence of collectors was investigated with the bubble particle attachment tests. The resulting bubble loading values listed in

Table 4. Bubble loading values at pH 11 in tap water. The error range indicates a 95% confidence interval.

Conditioning	Bubble Loading Value for Single Mineral		Bubble Loading Ratio
	Li-Aluminate	Gehlenite	Li-Aluminate/Gehlenite
No surfactant	4.0 ± 0.7%	8.5 ± 0.5%	0.47
Oleic acid	23.7 ± 12.7%	25.3 ± 1.8%	0.94
Punicine	45.9 ± 7.8%	51.8 ± 3.7%	0.89

give an indication about the interaction degree of particles with bubbles as a level of hydrophobicity. For these tests, the mentioned model materials were used to estimate the response of EnAM phases. The tests performed without collectors show minor bubble loading (<9%), whereby gehlenite particles adhered more strongly to the bubble (8.5%) than Li-aluminate particles (4.0%) (Figure A6a,d). Thus, Li-aluminate particles exhibit a higher degree of wettability (or lower dewettability) in comparison to gehlenite particles. With the addition of collectors for hydrophobization, the bubble loading value increased (>24%) (Table 4). After the conditioning with the novel collector punicine, the bubble loading results were twice as high in contrast to conditioning with the benchmark collector, oleic acid. In general, gehlenite responds with a collector-independent, minimal higher bubble loading in comparison to Li-aluminate (Figure A6). This confirms the overall higher wettability of Li-aluminate particles observed without collectors. The lowest bubble loading ratio towards Li-aluminate was observed with the collector punicine (0.89). More specifically, the bubble loading of gehlenite resulted in 51.8% in comparison with Li-aluminate, with 45.9%.

3.3. Slag Flotation

To ensure clarity in the following evaluations based on MLA results, the following minerals have been grouped according to their main chemistry after data classification: eucryptite (eucryptite, Mn-rich eucryptite), spinel (spinel, Mn-rich spinel) and mixed phases (CaMn-aluminosilicate mix, Mn-aluminosilicate mix).

3.3.1. Flotation of EnAM Phases

Both tested surfactants were mainly present in their anionic form and acted as anionic collectors (Figure 1) during flotation experiments at natural pH 10.7. It was observed that both collectors showed a characteristic maximum cumulative water and mass pull at a concentration of 750 g/t at comparable flotation times (concentrates were collected at similar timed intervals) (Figure 7a). In more detail, the collectors have similar maximal water pulls (38 ± 12)% (w/w), whereas the maximum mass recovery differed for oleic acid (88 ± 10)% (w/w) to punicine (18 ± 4)% (w/w). A strong froth formation was observed at higher oleic acid concentrations (Figure A7b,c). Oleic acid is known to be a strong foaming reagent, especially at alkaline pH levels [42,43,44]. The natural pH during slag flotation gives the ideal conditions for strong froth formation with oleic acid and leads to higher mass and water pulls—consequently contributing to unselective particle recovery via entrainment. This is confirmed by the large mass pull at 750 g/t and 1000 g/t of oleic acid, which yielded comparable compositions and particle size distribution between the concentrate and the feed material (Figure 7b). At 500 g/t of oleic acid, a larger mass of fine particles was recovered into the concentrate with an x_50,3 of (24 ± 1) µm, caused by a dry froth zone with heavily loaded bubbles (Figure A7a), resulting in a lower mass recovery (10 ± 1)% (w/w) and less water entrainment (2 ± 1)% (w/w) (Figure 7a).

The flotation with the punicine surfactant and pine oil is characterized by less froth formation and a rather shallow froth layer, but with a close-knit froth, showing small bubbles which maintain fine particles [45]. Thus, the recovered mass in the concentrates mainly consists of fine particles with x_90,3 smaller than 38 µm and an x_50,3 value smaller than 13 µm for all concentrations (500 g/t–1000 g/t) (Figure 7c). In general, flotation with punicine leads to the concentration of more finer particles relative to oleic acid (Figure 7b,c). The water–mass pull graph (Figure 7a) shows a high water and a low mass pull. This observation is confirmed by the froth characteristics, since the froth was not visibly loaded with particles and was rather wet (Figure A7d–f). In total, the maximum mass pull at 750 g/t punicine resulted in a coarser concentrate in comparison to other concentrates, which were fine.

The enrichment factor (E,

Table 5. Enrichment factor (E) for EnAM phases (j) in concentrates (C) resulting from batch flotation with collector (k) regarding their feed (F) concentration (c): $E={\displaystyle \frac{c_{j,k,0,C}}{c_{j,k,0,F}}}$.

Collector (k)	c in g/t (c)	E for EnAM Phases (j)
		Li-Aluminate	Eucryptite	Gehlenite	Spinel	Mixed Phase
Pun	500	1.70	1.13	1.11	0.77	0.95
	750	1.54	1.20	1.10	0.81	0.88
	1000	1.54	1.14	1.20	0.71	0.99
OA	500	0.45	0.89	1.26	1.03	0.72
	750	1.00	0.98	1.02	1.00	0.98
	1000	0.99	0.99	1.01	1.00	0.98

) for Li-aluminate observed in the flotation concentrates obtained with oleic acid are low. In fact, the concentrates at 500 g/t are enriched in the gangue phase gehlenite (+7.9% grade, 1.3 enrichment factor) and spinel (+1.1% grade, 1.0 enrichment factor), while the grade of the target phase Li-aluminate decreases (−3.3% grade, 0.5 enrichment factor)—as displayed in Table 5, Figure 8 and

Table A3. Grade in % of EnAM phases in concentrate (C) and tailing (T) of flotation of the collectors punicine and oleic acid with the concentration (c). In the table, the following minerals have been grouped: eucryptite (eucryptite, Mn-rich eucryptite), spinel (spinel, Mn-rich spinel) and mixed phases (CaMn-aluminosilicate mix, Mn-aluminosilicate mix).

Collector	c in g/t	C/T	m in g	Li-Aluminate	Eucryptite	Gehlenite	Spinel	Mixed Phase	Others
Punicine	500	C	2.08	10.2	12.4	35.0	28.9	12.4	1.0
		T	23.21	6.0	11.0	31.5	37.4	13.0	1.1
	750	C	4.57	9.7	13.5	34.2	30.4	11.3	0.8
		T	20.81	6.3	11.2	31.0	37.6	12.9	1.0
	1000	C	2.34	9.1	12.8	38.1	26.6	12.5	1.0
		T	22.70	5.9	11.2	31.6	37.6	12.6	1.0
Oleic acid	500	C	2.72	2.7	10.6	38.6	35.1	11.3	1.8
		T	22.62	5.9	11.8	30.7	34.0	15.7	1.8
	750	C	23.20	6.0	11.2	31.5	34.7	15.0	1.6
		T	1.97	6.0	11.4	30.9	34.7	15.3	1.7
	1000	C	22.56	6.0	11.2	31.7	34.2	15.2	1.7
		T	2.08	6.1	11.3	31.4	34.1	15.5	1.7

.

Punicine flotation resulted in minor selective recovery of the target EnAM Li-aluminate against gehlenite, as shown in the Fuerstenau upgrading diagram (Figure 9a) derived from the individual recoveries (

Table A4. Recovery in % of EnAM phases of batch flotation. In the table, the following minerals have been grouped: eucryptite (eucryptite, Mn-rich eucryptite), spinel (spinel, Mn-rich spinel) and mixed phases (CaMn-aluminosilicate mix, Mn-aluminosilicate mix).

Collector	c in g/t	Li-Aluminate	Eucryptite	Gehlenite	Spinel	Mixed Phase
Punicine	500	14.0	9.3	9.2	6.4	7.9
	750	27.8	21.7	19.9	14.6	15.9
	1000	14.4	10.7	11.2	6.6	9.3
Oleic acid	500	4.8	9.6	13.5	11.0	7.7
	750	92.3	90.5	94.0	92.2	90.0
	1000	90.4	90.6	92.7	91.7	90.1

). However, punicine enriched all considered EnAM phases in minor amounts against gehlenite except spinel and mixed phases (Table 5,

Table A5. Fuerstenau coefficient for EnAM phases of batch flotation regarding the gangue gehlenite. In the table, the following minerals have been grouped: eucryptite (eucryptite, Mn-rich eucryptite), spinel (spinel, Mn-rich spinel) and mixed phases (CaMn-aluminosilicate mix, Mn-aluminosilicate mix).

Collector	c in g/t	Li-Aluminate	Eucryptite	Gehlenite	Spinel	Mixed Phase
Punicine	500	1.53	1.02	1.00	0.69	0.86
	750	1.40	1.09	1.00	0.73	0.80
	1000	1.28	0.95	1.00	0.59	0.82
Oleic Acid	500	0.36	0.71	1.00	0.82	0.57
	750	0.98	0.96	1.00	0.98	0.96
	1000	0.97	0.98	1.00	0.99	0.97

).

These enrichment characteristics resulted in an overall selectivity against spinel phases (Figure 9b), which was not observed with the oleic acid reagent regime (Figure 9c). The maximum Li-aluminate grade of 10.2% (Table A3) was observed at 500 g/t with an enrichment factor of 1.7 (+4.2% grade), while gehlenite showed reduced upgrading of 3.5% (1.1 enrichment factor) (Figure 8, Table 5). The optimum conditions to recover Li-aluminate were 750 g/t of punicine with the highest selectivity for the target phase (Figure 9a,b). The mass yield of Li-aluminate is only slightly influenced by the punicine concentration, whereas the oleic acid dosage has a greater impact on mass pull due to the strong froth formation.

3.3.2. Effect of EnAM Liberation on Flotation

To evaluate the liberation effects of EnAMs on the flotation of the two collector systems, the data for the concentration of 500 g/t were evaluated as they had similar mass pulls and showed characteristic selectivity differences towards Li-aluminate. As already discussed in Section 3.1.2, the spinel phase in particular shows different liberation behavior and different flotation properties as compared to gehlenite and Li-aluminate. Therefore, these three EnAM phases were selected for the following evaluations.

In general, the probability of recovering liberated particles increases as the particle size decreases. Punicine facilitated the recovery of both fine-grained particles (shown in Section 3.3.1) and particularly liberated particles, which is expressed by the enrichment factors (L) in

Table 6. Enrichment factor (L) of concentration (c) of each liberation class (l) for each EnAM phase (j) within the concentrate (C) regarding the feed (F) after flotation with collector (k) with $L={\displaystyle \frac{c_{j,k,l,C}}{c_{j,k,l,F}}}$ over the entire particle size distribution.

Collector	EnAM	L of Liberation Class l
k	j	Liberated	Middlings	Locked
Punicine	LiAlO2	1.61	0.42	0.29
	gehlenite	2.13	0.27	0.26
	spinel	2.50	0.41	0.26
Oleic acid	LiAlO2	0.84	0.97	2.28
	gehlenite	1.66	0.67	0.51
	spinel	1.17	0.97	0.69

, as well as the recovery values in

Table A6. Recovery in % of particles with liberation class (l) containing phase (j) in concentrate of flotation with collector (k) at 500 g/t.

Collector (k)	Liberation Class (l)	Recovery of Particles in % Containing Phase (j)
		Li-Aluminate	Gehlenite	Spinel
punicine	liberated	22	20	15
	middlings	6	3	2
	locked	4	2	2
oleic acid	liberated	4	23	13
	middlings	5	9	11
	locked	11	7	8

for liberation classes.

The recovery values displayed are related to particle size in Figure 10. To this effect, a total of 79% of all liberated Li-aluminate and 64% of all liberated gehlenite particles smaller than 6.8 µm were recovered using punicine (Figure 10a and Figure A8a). In the fine particle size range, punicine is notably more selective for liberated Li-aluminate particles, whereas oleic acid is more selective for liberated gehlenite particles (Figure 10b). Collectively, for all particle sizes, 22% of liberated Li-aluminate particles were transferred into the concentrate using punicine. Overall, oleic acid recovered 23% of liberated gehlenite particles into the concentrate (1.7 enrichment factor), but only 4% of Li-aluminate liberated particles (0.8 enrichment factor) (Table 6). Furthermore, it can be noted that oleic acid recovered coarser liberated particles in comparison to punicine (Figure 10a). The punicine regime preferentially recovers liberated particles independently from the considered EnAM phases (1.6–2.5 enrichment factor), while oleic acid recovers, in addition to liberated particles, middlings (25%) and locked phases (26%) (Table A6). This reflects oleic acid’s different characteristics to punicine concerning the recovery of variably sized and liberated particles.

4. Discussion

In agreement with the predicted EnAM phase formation in engineered Li-slags [13], the main EnAM phases in the studied large scale synthetic slag (120 kg) are Li-aluminate, gehlenite, eucryptite and aluminum spinel (Table 2). The Li-aluminate content characterized in the present sample (7.9% (w/w)) by XRD is relatively low compared to other studies on Li-bearing slags, which reported values from 11% (w/w) to 22% (w/w) [13,16,47]. The Li-aluminate content in the slag was reduced by more than half compared to a slag with the same feed composition in a small batch size of 4.5 kg (13.2% (w/w) Li-aluminate) at the same cooling rate of 25 K/h [16].

The composition of EnAM phases identified by MLA are comparable to the composition calculated by Wittkowski et al. and Schirmer et al. from lab-scale experiments [5,39,40]. Li-aluminate, the target Li-bearing phase, contains minor amounts of silicon according to its published chemical formula Li_{1 − x}(Al_{1 − x}Si_x)O₂ [40] and accounts for 6.0% (w/w) of the slag. Additional Li-host phases include spinel and eucryptite (Mn-poor and Mn-rich). The spinel phases are found as a solid solution according to (Li_(2x)Mn²⁺_{(1 − x)(1 + x)}(Al_{(2 − z),}Mn³⁺_(z))O₄ [5]. These form 36.1% (w/w) of the slag sample. Eucryptite and Mn-rich eucryptite, following the formula Li_{1 − x}(Mn³⁺_y,Al_{1 − (x + y)})_{1 − x}(Si_{1 + x})O₄ [40], contain lithium depending on the incorporated amount of Mn. These minerals together compose 11.2% (w/w) of the slag. Furthermore, the silicate phase Li₂MnSiO₄, which was identified by XRD (13.5% (w/w)), could also represent a lithium-host phase in the slag. Li₂MnSiO₄ was also identified by XRD within synthesized slags of different feed compositions [13,16]. The identification of these Li-containing silicate phases confirms the predicted phase formation considering the composition of the feed used for the studied slag [16].

Regarding the modal composition of the Li-slag sample studied, one of the main differences to other systems described in the literature is the abundance of mixed phases, which were which were observed in the MLA results MLA (14.1% (w/w)). These phases represent a challenge for characterization since the mixes contain finely disseminated phases, smaller than 1 µm, which are difficult to resolve using standard MLA measurement settings. Altogether, it is likely that one or more of the major phases represents a component of the mixed phases, and a proportion of the Li-content of the slag occurs within one or more of these mixed phases.

In context of slag beneficiation, it is noteworthy that the liberation characteristics of Li-aluminate and gehlenite are comparable over particle size, making it impossible to separate the target and gangue phases solely based on particle geometric properties (e.g., size or liberation). In contrast, spinel-type phases exhibit a different breakage behavior resulting in a high amount of middlings (60%, Table 3) and coarse particle sizes. These distinctive characteristics could be explored when designing a comminution circuit for this type of material.

A full degree of liberation (>90% (A/A)) of all phases is achieved for particles finer than 9.6 µm. The target EnAM Li-aluminate is liberated to 50% (Table 3) and represents only a minor association partner to other slag phases (<5%, Table A2). Overall, Li-aluminate is slightly associated with gehlenite (16%, Table A2), which can facilitate their selective separation via flotation under the proper reagent conditions. Eucryptite is the main phase associated with Li-aluminate after spinel (Table A2). Since eucryptite and spinel also carry lithium, it is possible that this association does not drastically influence Li recovery but may limit its physical enrichment ability.

In context of EnAM flotation, the two different reagent regimes, oleic acid and punicine, render characteristic recovery trends to the phases of interest as a function of their particle size and liberation degree (Figure 7b). The punicine regime demonstrates a notable tendency to recover fine particles, with an x_50,3 particle size of (8–13) µm, regardless of collector concentration (Figure 7b). The formation of smaller bubbles and a denser froth (Figure A7d–f) increases the efficiency of fine particle capture [45], distinguishing the punicine regime from oleic acid. This froth characteristic could support the typical phase-independent preferred recovery of liberated particles (with a degree of liberation higher than 80% (A/A)) by punicine. Conversely, the oleic acid regime renders a drier and particle-loaded froth (Figure A7a–c), which results in a coarser concentrate (x_50,3 21 µm; 500 g/t oleic acid) relative to punicine (Figure 7c).

Sodium oleate, a deionized form of oleic acid, has shown a higher affinity towards Li-bearing minerals in Li-slag flotation when using a mechanical small Denver-type flotation cell with collector concentrations between 200 g/t and 900 g/t and higher pulp density [13]. Yet, the use of oleic acid in this study resulted in the selectiverecovery of Ca-bearing gehlenite particles in the lower dosage concentration (500 g/t), leading to the reverse flotation of Li-aluminate. At higher concentrations (750 g/t), the strong frothing characteristics of oleic acid at basic pH (10.5) dominates the process, leading to a high mass pull recovery, a similar concentrate composition to the feed and non-selectivity (Figure 8). This indicates that increases in collector concentration do not yield additional benefits. Further investigations regarding the optimization of oleic acid dosage for this slag system should be limited to 750 g/t as a maximum dosage.

In contrast, the punicine regime enables the direct flotation of Li-aluminate from the slag. The highest upgrading of Li-aluminate (+4.2% (w/w)) was observed at 500 g/t. The optimum Li-aluminate selectivity and recovery obtained among the studied conditions was at 750 g/t of punicine (1.54 enrichment factor, 1.40 Fuerstenau coefficient, 28% (w/w) recovery) (Figure 9a,b). However, higher punicine dosages lead to a higher gehlenite recovery. Overall, the mass recovery of Li-aluminate with the collector punicine is still quite low, but similar to oleic acid at 500 g/t. A change in reagent parameters in the future, e.g., by using depressants or activators, has the potential to increase selectivity or recovery. Additionally, the preceding desliming procedures can also improve flotation performance [21], but may potentially compromise the overall recovery.

In conjunction with the previously discussed properties of punicine regarding the recovery of fine liberated particles (Table 6), the system is found to be particularly selective towards liberated Li-aluminate particles smaller than 10 µm (Figure 5b). Besides the real selectivity of the surfactant to Li-aluminate, i.e., true flotation, entrainment effects of fine, light, particles containing a higher number of liberated phases, such as Li-aluminate, can also be a driver of this behavior. To determine whether this effect has an impact on selectivity, the mass ratio of Li-aluminate to gehlenite was plotted across the particle size classes of both the feed and concentrate within a single liberation class. With punicine, the mass ratio increased across all particle sizes compared to the feed (Figure A8b). Notably, the mass ratio within liberated particles in the size fractions of 10 µm < x < 64 µm rose, indicating that the selectivity towards Li-aluminate is not primarily driven by general entrainment effects in finer particle sizes. Eucryptite, as a Li-host phase, is enriched alongside Li-aluminate when using punicine, but not with oleic acid. In addition to microstructure-related factors, such as the strong association of Li-aluminate with eucryptite (Table A2), these findings may also support punicine exhibiting selectivity towards Li-containing EnAMs, particularly since this trend was not observed with oleic acid.

To determine the potential selectivity of collectors, it is important to consider their adsorption mechanism on mineral surfaces, as described by Schmidt and Zgheib et al. [19] for punicine using the model materials Li-aluminate and gehlenite. For this mechanism, under humid conditions or in aqueous media, Li-aluminate forms hydrophilic species on the surface of the particles. In particular, Li-aluminate reacts to form lithium aluminum hydroxide hydrate (LiAl₂(OH)₇ × H₂O) with the release of lithium cations and hydroxide anions, which can explain the high pH of about 11 of suspensions of Li-aluminate in water according to Equation (2). Subsequent reactions with atmospheric CO₂ can result in the formation of carbonates [48]. In addition, hydroxide leads to further conversions into LiAl₂(OH)₇ × H₂O. During this reaction, lithium cations are released under the formation of aluminum hydroxide anions according to Equation (3) [49]. In the presence and addition of punicine, hydroxide groups of LiAl₂(OH)₇ can be substituted by the nucleophilic 2-olat group of the punicine according to Equation (4), which leads to the targeted surface hydrophobization [19,32]. The type of bond between the punicine and the hydrophilic particle surface can be described as a Coulomb interaction, for example, as a salt-like structure, or as a complex formation.

$$2{\mathrm{L}\mathrm{i}\mathrm{A}\mathrm{l}\mathrm{O}}_2+\left(4+\mathrm{x}\right){\mathrm{H}}_2\mathrm{O}\to \mathrm{L}\mathrm{i}\mathrm{A}{\mathrm{l}}_2{\left(\mathrm{O}\mathrm{H}\right)}_7·\mathrm{x}{\mathrm{H}}_2\mathrm{O}+\mathrm{L}\mathrm{i}\mathrm{O}\mathrm{H}$$

(2)

$$\mathrm{L}\mathrm{i}\mathrm{A}{\mathrm{l}}_2{\left(\mathrm{O}\mathrm{H}\right)}_7·\mathrm{x}{\mathrm{H}}_2\mathrm{O}+\mathrm{O}{\mathrm{H}}^{-}\to {\mathrm{L}\mathrm{i}}^{+}+2\mathrm{A}\mathrm{l}{\left(\mathrm{O}\mathrm{H}\right)}_4^{-}+2{\mathrm{H}}_2\mathrm{O}$$

(3)

(4)

Therefore, the hydration, hydroxylation and accessibility of hydroxy groups on EnAM phases can play a crucial role in the hydrophobization of these phases during flotation with punicine. It is important to highlight that the hydroxylated aluminates required for complexation and hydrophobization can form not only on Li-bearing aluminate but also on aluminosilicates [50], such as the gangue phase gehlenite. In addition to punicine, the adsorption of the benchmark collector oleic acid can also depend on the availability of hydroxyl groups. For instance, in the case of sillimanite, it was discussed that adsorption was facilitated by an ion exchange mechanism between oleate ions and surface hydroxyl sites [50]. The discussion on the comparable adsorption capabilities of collectors on aluminates and aluminosilicates is supported by the results of bubble–particle attachment tests. Experiments have demonstrated that both punicine and oleic acid interacted with the model minerals Li-aluminate and gehlenite, leading to an increase in bubble–particle attachment for both minerals (see Section 3.2). Overall, the loading of bubbles with gehlenite particles was higher under all conditions, thus demonstrating a higher degree of dewettability. When comparing the two collectors, the mass pull obtained by punicine was twice as high as oleic acid. Similarly, a greater bubble loading towards gehlenite was observed with punicine. This affinity for gehlenite, which was not observed with punicine in complex slag systems during flotation, suggests that bubble–particle attachment tests using model materials can only provide an indication of a collector’s ability to effectively dewet target minerals. However, these tests cannot offer selectivity information for subsequent flotation experiments in more complex systems, such as slag. In summary, bubble–particle attachment tests confirm the successful hydrophobization of Li-aluminate with punicine, while the results of the flotation tests with complex synthetic slags show a noticeable selectivity of punicine for the EnAM Li-aluminate. Despite these promising findings, further fundamental studies are necessary and flotation experiments need to be reproduced, especially with Li-slag samples that exhibit contrasting compositions and liberation attributes. In particular, the adsorption behavior of punicine on EnAM phases of different surface characteristics requires further investigation.

5. Conclusions and Outlook

As the target EnAM in slags, Li-aluminate was shown to be well liberated at 50% within a particle size range of (1–117) µm—thus offering a size class suitable for separation via flotation. Additionally, the lack of association between the main gangue phase gehlenite and the target phase Li-aluminate supports their separation (2% association). The maximum obtained upgrade of Li-aluminate in a laboratory-scale flotation column is about 4% (w/w) (up to 1.7 enrichment factor) using the novel collector punicine (500 g/t). Furthermore, punicine leads to the true flotation of Li-aluminate with minor selectivity against gehlenite but higher selectivity against spinel-type phases. In contrast, the benchmark collector oleic acid, known from spodumene flotation, does not upgrade Li-aluminate in slag flotation, while gehlenite and spinel-types become upgraded. Oleic acid facilitates the true flotation of gehlenite, promoting, in general, a higher selectivity than the novel collector punicine when used up to a 500 g/t dosage. The potential of the reverse flotation of Li-aluminate with oleic acid in slag systems should be investigated further with optimization methodologies. Furthermore, these studies reveal that a major difference between the collectors can be recognized by their frothing properties. The punicine regime with pine oil enables the recovery of fine particles with a fine froth but low mass recovery, while the strong frothing properties of oleic acid, in combination with MIBC, under alkaline conditions challenges the selectivity of the system and requires sensitive reagent concentration adjustments. In summary, these results support the suitability of punicine as a Li-aluminate collector, not only in microflotation with model materials, but also for the batch flotation of Li-aluminate as a target EnAM as part of complex synthetic slags. Indeed, further research is needed to investigate different types of punicines as Li-aluminate collectors and the use of depressants, activators or different frothers to improve selectivity and yield. Optimization methodologies should be used in this case given the expected interaction between process parameters. From the fundamental investigations, this study shows that bubble–particle attachment with model minerals are consistent with documented adsorption mechanisms in the literature of punicine and oleic acid with hydroxylated aluminates and aluminosilicates. Further fundamental investigations are required to deepen our understanding of the working mechanisms of punicine and the target phases. The unique methodology employed in this manuscript investigates the influence of chemical and microstructural properties in the flotation behavior of different EnAM phases and is advantageous for studying the potential of novel flotation reagents in complex materials.