Speciation of Manganese in a Synthetic Recycling Slag Relevant for Lithium Recycling from Lithium-Ion Batteries

Abstract

Lithium aluminum oxide has previously been identified to be a suitable compound to recover lithium (Li) from Li-ion battery recycling slags. Its formation is hampered in the presence of high concentrations of manganese (9 wt.% MnO₂). In this study, mock-up slags of the system Li₂O-CaO-SiO₂-Al₂O₃-MgO-MnO_x with up to 17 mol% MnO₂-content were prepared. The manganese (Mn)-bearing phases were characterized with inductively coupled plasma optical emission spectrometry (ICP-OES), X-ray diffraction (XRD), electron probe microanalysis (EPMA), and X-ray absorption near edge structure analysis (XANES). The XRD results confirm the decrease of LiAlO₂ phases from Mn-poor slags (7 mol% MnO₂) to Mn-rich slags (17 mol% MnO₂). The Mn-rich grains are predominantly present as idiomorphic and relatively large (>50 µm) crystals. XRD, EPMA and XANES suggest that manganese is present in the form of a spinel solid solution. The absence of light elements besides Li and O allowed to estimate the Li content in the Mn-rich grain, and to determine a generic stoichiometry of the spinel solid solution, i.e., (Li_(2x)Mn²⁺_(1−x))_1+x(Al_(2−z),Mn³⁺_z)O₄. The coefficients x and z were determined at several locations of the grain. It is shown that the aluminum concentration decreases, while the manganese concentration increases from the start (x: 0.27; z: 0.54) to the end (x: 0.34; z: 1.55) of the crystallization.

1. Introduction

Modern technologies, such as renewable energy and e-mobility, demand a new portfolio of technology-critical elements and materials. Limited resources, national policies or monopolies threaten the supply of some technology-critical elements. Hence, the recovery of these elements from waste is crucial. On the one hand, demand for lithium (Li) has increased rapidly due to the popularity and extraordinary performance of Li-ion batteries. On the other hand, Li is produced by mainly two countries, Australia (ca. 55%) and Chile (ca. 23%) [1]. The recycling of some components from Li-ion batteries is already put into practice, e.g., cobalt, in a pyrometallurgical process [2,3]. Pyrometallurgical recycling has the benefit of being adaptable to many waste streams; additionally, the emission of toxic compounds like HF is prevented. Due to its ignoble character, Li is driven into the slag and is usually not recovered.The recovery of Li from pyrometallurgical recycling slags can be accomplished by the targeted formation of “engineered artificial minerals” (EnAM).

The strategy of EnAM formation is to concentrate the elements of interest in a few phases, with a structure and size that allows an efficient separation. Figure 1 shows a scheme of EnAM formation. The target element (yellow triangle) spreads over several different phases (red and green forms) and the matrix (blue) in the unmodified slag (top picture). The target element will be difficult to isolate due to its occurrence in many different phases. The goal in EnAM is to concentrate the target element in a single phase (red pentagon, bottom picture) that differs physically and chemically from the other phases (green hexagon), and allows for efficient separation and further treatment.

Separation of artificial minerals, enriched in valuable elements from remaining slag components, might be carried out by flotation processes. Here, the composition and structure of phases are crucial. It can be expected that none of the artificially produced minerals will show hydrophobic behavior by nature. Thus, the mineral-collector-interaction has to be studied, and this interaction relies strongly on crystal structure and ions, responsible for the adsorption of the active group of collector molecules.

The recovery of Li from slags has been subject to several studies. Elwert et al. [2] found that Li reacts with aluminum to form predominately large lithium aluminate (LiAlO₂) crystals. The aluminum binds Li from the melt uniformly in one phase at an early state of solidifying phase of the molten slag. The idiomorphic to hypidiomorphic lithium aluminate crystals can easily be separated from the matrix by flotation. Hence, lithium aluminate is a promising EnAM candidate. The recovery of Li from LiAlO₂ was successfully conducted by Haas et al. [4]. Elwert et al. [5] investigated a hydrometallurgical process to recover Li from slags with low aluminum compared to [2] and enriched in silicone. However, it was found that the formation of LiAlO₂ is suppressed in manganese (Mn)-rich slags [2]. In these slags, Li is distributed over several phases of low crystallinity. Elements like Mn seem to have a significant influence on the formation of compounds and grain size during the formation of the slag. Due to the increasing amount of Li-ion batteries with Mn-based cathode materials, i.e., Li-Ni_xMn_yCo_zO₂ (NMC), it must be assumed that the Mn content in battery waste streams will increase soon. Accordingly, it is necessary to understand the role of the redox-active Mn on the genesis of crystalline, and especially amorphous grains from the ionic melts. A careful characterization of the Mn-bearing grains will give insight into these processes. The first survey with mineralogical and thermodynamic methods on the formation of Li-EnAMs was conducted by Schirmer et al. [6] on a similar system, but without Mn. Their study shows that formation of spinel solid solutions is a favorable reaction with a thermodynamically proved potential to scavange Li from the slag in an early stage of the solidification. Adding Mn to this system would result in more complex solidification reactions with Mn-containing spinel solid solutions in particular, as numerous spinel-like oxides of Li and Mn are already described [7].

Therefore, in this study, Li- and Mn-bearing phases from a synthetic slag, in the following termed mock-up slag (MUS), of the system Li₂O-CaO-SiO₂-Al₂O₃-MgO-MnO_x, with up to 17 mol% MnO₂-content, are studied. The crystalline components are determined by powder X-ray diffraction (PXRD). The main Mn-containing phase could be identified as an oxide solid solution of spinel-type using electron probe microanalysis (EPMA). From a combination of the virtual compounds LiMnO₂, Mn_0.5AlO₂ (1/2 galaxite spinel) and Mn_0.5MnO₂ (1/2 hausmannite), the formula (Li_(2x)Mn²⁺_(1−x))_1+x(Al_(2−z),Mn³⁺_z)O₄ was calculated. This matches with the PXRD results, as all main reflections of the oxides are located between those of LiAl₅O₈ and MnAl₂O₄. The EPMA and PXRD analysis showed that a high amount of grains are amorphous or of low crystallinity. The PXRD falls short of giving insight into the Mn species in amorphous phases. EPMA allows recognizing individual crystalline and amorphous phases. For amorphous phases, however, stoichiometric information can usually not be extracted from the data. Hence, laboratory-based X-ray absorption near edge structure (XANES) is applied here to determine the bulk Mn species of the crystalline and the amorphous components of the slag. The findings from this independent method confirmed the Mn oxidation state as being between +2 and +3, as well as the presence of Mn spinel structures.

2. Background

The following section gives a short introduction to the relevant binary subsystems of the Al₂O₃-MnO_x-Li₂O ternary system, which itself is not yet published. This will allow putting the presented results into the mineralogical context. In addition, the methods used to investigate the oxidation state of Mn are briefly described.

2.1. The System Li₂O-MnO_x with Focus on Spinel Structures

The stoichiometric spinel LiMn₂O₄ and several other spinel phases of Li-Mn-O are known in the Li₂O-MnO_x system. Paulsen and Dahn [7] created a binary phase diagram of Li-Mn-O in air. They described a spinel with the formula Li_(1+x)Mn_(2−x)O₄ to be stable between 400 and 880 °C. With the rising temperature, x decreases from 1/3 at 400 °C to 0 at 880 °C. Below 400 °C, the only stable spinel phases are Li₄Mn₅O₁₂ and LiMn_1.75O₄. Raising the temperature above 880 °C leads to the replacement of Li by Mn and tetragonal spinel [Li_(1−x)Mn_x]MnO₄ [7].

2.2. The System Li₂O-Al₂O₃

The binary system Li₂O-Al₂O₃ exhibits several phases. Konar et al. [8] described LiAl₅O₈ spinel, LiAl₁₁O₁₇, LiAlO₂ and Li₅AlO₄.

LiAl₅O₈ is described as an inverse spinel structure with Al³⁺ on tetrahedral sites and Li⁺ and the remaining Al³⁺ on octahedral sites. A polymorphic transition to this spinel structure occurs from the stoichiometric phase. LiAl₁₁O₁₇ is a high-temperature phase, only stable above 1537 °C. For LiAlO₂, a polymorphic transition from α- to γ-phase occurs.

Li₅AlO₄ crystallizes below 50 mol% Al₂O₃, in a mixture either with LiAlO₂ or with LiO₂ [8].

2.3. The System Al₂O₃-MnO_x

Solid solutions of spinel oxide type (XY₂O₄) with a cubic and tetragonal crystal system are described by Chatterjee et al. [9] to form in the Al₂O₃-MnO-Mn₂O₃ system. Besides the cubic MnAl₂O₄ (galaxite), additionally, MnMn₂O₄ (hausmannite) with a tetragonal and cubic crystal system (transformation tetragonal -> cubic: 1172 °C in the air) can occur. While the tetragonal form of hausmannite incorporates only a small amount of galaxite, the cubic variety forms a solid solution.

2.4. Speciation of Oxidation States in Spinel Systems

Recycling slags are usually complex elemental mixtures comprising various components and phases. Due to uncontrolled and–in geological terms–fast cooling conditions, only a few components in these slags are crystalline. Crystalline phases are easily accessible by established methods like PXRD.

The amorphous components, however, elude the identification by PXRD. Wet chemical methods like inductively coupled plasma atomic emission spectroscopy (ICP-OES) yield information on the stoichiometric composition, but fall short for the speciation. EPMA allows studying crystalline and amorphous phases. Stoichiometric information can be obtained for crystalline phases with reasonable certainty. Structure predictions on non-crystalline phases, however, are merely estimations. XANES is a convenient method to analyze the species of 3d-elements in crystalline, as well as amorphous materials. Asaoka et al. determined the oxidation state of Mn in granulated coal ash via XANES [10], while Kim et al. used XANES for Mn speciation in steel-making slags [11]. In general, the oxidation state, the coordination sphere, and in some cases the actual compound can be determined. Usually, this method is exclusively available at synchrotron sources (Sy). Here, a laboratory-based XANES spectrometer was used, which allows everyday access to measure several routine samples or improve sample preparation. The set-up is described by Seidler et al. [12,13]. Laboratory-based XANES has been used successfully in catalysis research, as well as for the determination of vanadium oxidation state in catalysts and vanadium redox flow batteries [14,15,16].

3. Materials and Methods

3.1. Synthesis of the Mock-Up Slag

The chemicals used for producing the mock-up slags are lithium carbonate (Merck KGaA purum, Darmstadt, Germany), calcium carbonate (Sigma-Aldrich, reagent grade, St. Louis, MO, USA), silicon dioxide (Sigma-Aldrich, purum p.a., St. Louis, MO, USA), aluminum oxide (Merck KGaA, Darmstadt, Germany), magnesium oxide (98%, Roth, Karlsruhe, Germany) and manganese dioxide (Merck KGaA, reagent grade, Darmstadt, Germany).

The concentrations of the reactants of the slag synthesis are the following: For V-1 32 mol% Al₂O₃ were mixed with 16 mol% CaO, 21 mol% Li₂O, 3 mol% MgO, 7.4 mol% MnO₂, and 22 mol% SiO₂. For V-2, 29 mol% Al₂O₃ were mixed with 14 mol% CaO, 19 mol% Li₂O, 3 mol% MgO, 13 mol% MnO₂, and 21 mol% SiO₂. For V-3, 28 mol% Al₂O₃ were mixed with 14 mol% CaO, 18 mol% Li₂O, 3 mol% MgO, 17 mol% MnO₂, and 20 mol% SiO₂. The chemicals were manually mixed in a mortar and ground in a disc mill for 5 min. Each sample was placed in a Pt-Rh crucible and heated in a chamber furnace (Nabertherm HT16/17, Nabertherm GmbH, Lilienthal, Germany) in an ambient air atmosphere. The temperature program is shown in Figure 2. A heating rate of 2.89 °C/min was initially employed to reach 720 °C, which is the melting temperature of Li₂CO₃. Subsequently, a heating rate of 1.54 °C/min was used to decompose Li₂CO₃ and reach the target temperature of 1600 °C. Finally, the obtained MUS were kept at 1600 °C for two hours. After that, the samples were cooled to 500 °C at a cooling rate of 0.38 °C/min, and quenched in water. Three MUS V-1-3 were obtained. For PXRD and XANES measurements, parts of the obtained slag were ground for five minutes in a disc mill (Siebtechnik GmbH, Mühlheim an der Ruhr, Germany).

The elemental content was determined by ICP-OES (ICP-OES 5100, Agilent, Agilent Technologies Germany GmbH & Co. KG, Waldbronn, Germany). The samples were fused with sodium tetraborate in a platinum crucible at 1050 °C, and then leached with diluted hydrochloric acid to determine Al, Ca, Li, Mg, Ti and Si. To determine other elements, the samples were suspended in nitric acid and digested at 250 °C and under a pressure of 80 bar in an autoclave (TurboWAVE, MLS, Leutkirch im Allgäu, Germany).

3.2. Synthesis of Galaxite via Solution Combustion Synthesis

Artificial galaxite was prepared via solution combustion synthesis. Aluminum nitrate nonahydrate (VWR chemicals, Darmstadt, Germany, analytical reagent, min. 98%) and manganese (II) nitrate tetrahydrate (Merck, KGaA, Darmstadt, Germany, pro analysi, min 98.5%) were used as oxidizers and mixed in stoichiometric ratio 1:2. Aluminum nitrate nonahydrate (7.5 g) was mixed with manganese (II) nitrate tetrahydrate (2.5 g). As a fuel, urea (Merck KGaA, Darmstadt, Germany, pro analysi, min. 99.5%) was added in excess (5 g). All components were dissolved in water. The solution was heated with a Bunsen burner until near dryness. The mixture was ignited with a second Bunsen burner. Purity was verified by PXRD (STOE STADI P, STOE & Cie GmbH, Hilpertstraße 10, 64295 Darmstadt, Germany).

3.3. Powder X-ray Diffraction

The analysis of the bulk mineralogical composition was provided by PXRD, using a PANalytical X-Pert Pro diffractometer, equipped with a Co-X-ray tube (Malvern Panalytical GmbH, Nürnberger Str. 113, 34123 Kassel, Germany). For identification of the compounds, the PDF-2 ICDD XRD database [17], the American Mineralogist Crystal Structure Database [18] and the RRUFF-Structure database [19] were evaluated.

3.4. Electron Probe Microanalysis

The analysis of single crystals and grains was performed with EPMA. EPMA is a standard method to characterize the chemical composition in terms of single spot analysis or element distribution patterns, accompanied by electron backscattered Z (ordinal number) contrast (BSE(Z)) or secondary electron (SE) micrographs. The EPMA measurements were performed on samples, which were embedded in epoxy resin, polished and coated with carbon. They were characterized using a Cameca SX^FIVE FE (Field Emission) electron probe, equipped with five wavelength dispersive (WDX) spectrometers (CAMECA SAS, 29, quai des Grésillons, 92230 Gennevielliers, Cedex, France). To calibrate the wavelength dispersive X-ray fluorescence spectrometers (WDXRF), an appropriate suite of standards and analyzing crystals was used. The reference materials were provided by P&H Developments (The Shire 85A Simmondley village, Glossop, Derbyshire SK13 9LS, UK and Astimex Standards Ltd., 72 Milicent St, Toronto, ON, M6H 1W4, Canada). The beam size was set to zero, leading to a beam diameter of substantially below 1 µm (100–600 nm with field emitters of Schottky-type, e.g., [20]). To evaluate the measured intensities, the X-PHI-Model was applied [21].

Li, one of the key elements in this study, cannot be directly analyzed, since EPMA uses X-ray emission to detect the elements in the sample, and the extremely low fluorescence yield and long wavelength of the Li Kα render the direct determination of this element merely impossible. With the reasonable assumption that other refractory light elements like Be and B are not present in the investigated material and volatile elements and compounds like F, H₂O, CO₂ or NO₃⁻ are effectively eliminated during the melt experiment, Li can be calculated using virtual compounds. Where necessary, the balanced Li concentration was included in the matrix correction calculation. To access the analytical accuracy with respect to the determination of Li, the international reference material spodumene (Astimex Standards Ltd., Toronto, ON, Canada) and the in-house standard LiAlO₂ were analyzed (

Table 1. Recovery of Li-compounds. Spod: spodumene, % StdDev: Relative standard deviation in percent, Repeats: N = 5, R: Recovery, LiAl: LiAlO₂.

Wt.%	Average Spod.	%StDev., Spod.	Ref. Spod.	R (%)	Average LiAl	%StDev., LiAl	Ref. LiAl	R (%)
Al	15.04	0.35	14.4	104	41.24	0.22	40.9	101
Mg	0.00	n.a.	0.0	n.a.	0.01	n.a.	0.0	n.a.
Ti	0.00	n.a.	0.0	n.a.	0.00	n.a.	0.0	n.a.
Mn	0.05	n.a.	0.0	n.a.	0.00	n.a.	0.0	n.a.
Fe	0.02	n.a.	0.0	n.a.	0.03	n.a.	0.0	n.a.
Ca	0.01	n.a.	0.0	n.a.	0.01	n.a.	0.0	n.a.
K	0.00	n.a.	0.0	n.a.	0.00	n.a.	0.0	n.a.
Si	28.71	0.56	30.0	96	0.01	n.a.	0.0	n.a.
Na	0.10	2.83	0.09	112	0.00	n.a.	0.0	n.a.

).

3.5. X-ray Absorption near Edge Structure

The Mn speciation was achieved with XANES. Other than usual, XANES was not conducted at a synchrotron facility but using a laboratory-based device, the easyXES100-extended (short: easyXES100; easyXAFS LLC, Renton, WA, USA). To enable these measurements with high energy resolution comparable to that obtained at synchrotron facilities, a Rowland circle Johann-type monochromator is used, see Figure 3. The instrument comprises an X-ray tube (100 W, air-cooled tube with W/Pd anode, 35 kV maximum accelerating potential), a spherically bent crystal analyzer (SBCA, Si 110) and an SDD detector (AXAS-M1, KETEK, Munich, Germany). The components are positioned on scissor drives.

This enables energy scanning by synchronously and symmetrically moving the detector and the source. The SBCA sits on a passively driven slider in the Rowland circle, which is coupled to the X-ray source stage. This way, the proper positioning of all three components is elegantly achieved. A helium-filled box is installed in the beam path, to lower background absorption and scattering by air. This set-up allows energy scanning with a resolution of about 1 eV. Further information about the set-up is published by Seidler et al. and Jahrman et al. [12,13,22]. At every energy step, an energy-dispersive X-ray spectrum (EDX) is acquired. The area of interest is automatically integrated by the software based on labVIEW2017 [23]. This way, a file with the relevant information on the energy and counts per lifetime is created and used together with the I₀ measurement to obtain the XANES.

Spectra of manganese dioxide (Merck AG, Darmstadt, Germany, for synthesis) were recorded, validated with spectra from the literature, i.e., spectra from Hokkaido University, Japan [24,25] and subsequently used as reference for energy calibration. The scans in the θ-angle space are converted to the energy space (see Appendix A).

The XANES of references and samples were recorded and processed as follows: The Mn K-edge was scanned from 6482 eV to 6800 eV (SBCA, Si 110, n = 4; see Appendix A), with a step width of 0.25 eV. At each step, an EDX spectrum with a 10 s live time was obtained, resulting in a total measurement time of approx. 3.5 h per spectrum. Three samples of MUS V-3 were prepared in parallel by mixing the finely ground slag with Vaseline (ISANA; Dirk Rossmann GmbH, Burgwedel, Germany) in a weight ratio of 2/3 Vaseline and 1/3 MUS powder. The mixtures were placed in washers with a height of 0.1 mm to adjust the thickness of the samples. The washer was sealed from both sides with adhesive tape (tesapack, tesa SE, Norderstedt, Germany) to hold the sample inside, and protect it from environmental influences. Each replicate was analyzed three times. The following Mn species were used as references: manganese (II) oxide (oxidation state of manganese: +2; alfa aesar, Thermo fisher (Kandel) GmbH, Kandel, Germany, 99%); synthetic galaxite (Mn²⁺Al₂³⁺O₄; oxidation state of manganese: +2), manganese (II, III) oxide (average oxidation state of manganese: +2.67; Sigma-Aldrich Chemie GmbH, Steinheim, Germany, 97%); braunite (Mn²⁺Mn³⁺₆[O₈|SiO₄]; formal oxidation state of manganese: +2.85; Friedrichroda, Thuringia, Germany; obtained from Geo collection, Clausthal University of Technology); bixbyite (Mn₂O₃; oxidation state of manganese: +3; Paling Mine, Republic of South Africa; obtained from Geo collection, Clausthal university of technology); and manganese (IV) oxide (oxidation state of manganese: +4; Merck AG, Darmstadt, Germany, for synthesis). They were prepared in the same way as the actual sample.

An additional I₀ spectrum of an empty washer sealed with adhesive tape was separately acquired for every measurement to calculate the absorption coefficient μ(E) according to the equation: $\mu \left(\mathrm{E}\right)=-\ln I·I_0^{-1}$. The average post-edge absorption μ(E) (post edge line) in each spectrum was normalized to unity using ATHENA software [26]. Spectra from the same Mn species were merged in ATHENA and used for further analysis.

4. Results

The results of the characterization of the MUS V-1-3, which are supposed to match Li-ion battery recycling slags, are presented in this section. The composition was chosen to be similar to recycling slags characterized previously by Elwert et al. [2]. The methodology applied to study the MUS chemistry comprises ICP-OES, PXRD, EPMA and XANES.

Initially, the composition of the reactants and the products was determined by ICP-OES. The suppression of the formation of LiAlO₂ EnAM in MUS with increasing Mn content was studied by PXRD. The microstructure and microscopic elemental composition of the MUS with the highest Mn-content (MUS V-3; MnO₂: 17 mol%) were studied by BSE(Z) micrographs, as well as detailed spatially resolved quantitative point measurements and element distribution profiles, recorded with EPMA. From these results, a hypothesis for the genesis of the Mn-rich grains was established. This hypothesis is discussed in detail in Section 4.3. Finally, it was evaluated by studying the Mn species using XANES.

4.1. Bulk Chemistry

The elemental composition of the reactants and the product slags were determined by digestion, followed by ICP-OES. The Mn-content increases from MUS V-1 to V-3, resulting in a concentration of MnO₂ of 7 mol%, 13 mol%, and 17 mol% in the final products. The results are given in

Table 2. Comparison of the bulk chemical composition of the three melt experiments, given in mole percent.

	Raw Mix (Mole Fraction)			Product (Mole Fraction)			Recovery %
	V-1	V-2	V-3	V-1	V-2	V-3	V-1	V-2	V-3
Al2O3	32	29	28	33	31	29	103	107	104
CaO	16	14	14	15	15	14	94	107	100
Li2O	21	19	18	20	18	17	95	95	94
MgO	3	3	3	3	3	3	100	100	100
MnO2	7.4	13	17	6.1	12	17	82	92	100
SiO2	22	21	20	23	22	21	105	105	105

. A loss of about 5% of Li and 0–18% Mn occurred during the melting and cooling of the material. The MUS are therefore close in composition to the actual recycling slags studied by Elwert et al. [2].

4.2. Powder X-ray Diffraction

The crystalline composition of the MUS V-1-3 was studied by PXRD and compared to an Mn-free material with comparable LiAlO₂-content. The compounds gehlenite, spinel and LiAlO₂ were present in all three slags (Figure 4a). The Li-Al-Oxide reflections are best explained with the diffraction pattern of LiAlO₂ (ICDD PDF2 no. 00-038-1464 [17]). A comparison of the reflection height of three mixtures with increasing Mn-concentration shows a negative correlation with the intensity of the LiAlO₂ main reflection. A comparison with an Mn-free material with comparable Li₂O content (Figure 4c) indicates a strong negative influence of the Mn-concentration on the formation of LiAlO₂. The low intensities of the reflection and the comparable high background imply a high amount of amorphous material being present.

The enlarged section of the two-theta region of the main spinel reflections gives an indication of the changing composition of the spinel with the change of the Mn-content (Figure 4b). The main spinel reflections (311) of all three MUS lie between those of galaxite spinel MnAl₂O₄ and LiAl₅O₈. In this range, two reflections of Li/Mn-spinel are located (Li_1−xMn₂O₄ and Li₂Mn₂O₄) [17]. This indicates the presence of a complex solid solution of a spinel-like oxide with the general formula (Li_(2x)Mn²⁺_(1−x))_1+x(Al_(2−z),Mn³⁺_z)O₄, which was derived by EPMA, as discussed below.

In conclusion, the PXRD results show the decrease of crystalline LiAlO₂ content with an increasing Mn-concentration. They also indicate the presence of a Li/Al/Mn spinel-like solid solution. The findings are consistent with the results of the microscopic EPMA analysis, which are presented in the following section.

4.3. Electron Probe Microanalysis

The MUS with the highest MnO₂-content (MUS V-3, 17 mol%) was subjected to EPMA. From the PXRD results, it was expected that the microstructure of this sample would be most conclusive on the processes, resulting in a LiAlO₂-depleted material. The main compounds of the melt experiment V-3, determined with EPMA, were: a spinel phase (Li_(2x)Mn²⁺_(1−x))_1+x(Al_(2−z),Mn³⁺_z)O₄; lithium aluminate (LiAl) with the stoichiometric formula Li_1−x(Al_(1−x)Si_x)O₂; a gehlenite-like calcium-alumosilicate (GCAS) with the stoichiometric formula Ca₂Al₂SiO₇ and with minute amounts of Mg and Na (max. ~0.2 wt.%); and amorphous phases (APh) with various amounts of Al, Si, Ca, Mn, small amounts of Na (<0.3 wt.%), and sometimes with unusual elements like Ba and K (contaminants enriched in the eutectic residual melt).

The LiAl and the GCAS have already been described by Schirmer et al. [6] in a MUS not containing Mn but Li, Ca, Si, Al and Mg. There is strong evidence that the presence of Mn in the slag has an influence on the formation of Li and Al compounds. In particular, the formation of spinel solid solutions suggested by the PXRD is expected to influence the formation of LiAlO₂. Accordingly, the EPMA focused on elucidating the genesis of the Mn-rich grains. Overall, it was found that besides negligible amounts in amorphous phases, the Mn is almost exclusively incorporated into pure oxide (spinel) structures. For this reason, a detailed examination of a representative grain of Mn-containing oxide is presented in the following. Early crystallites of this type are predominantly found throughout the whole sample.

The BSE(Z) micrograph (Figure 5) shows a large grain of predominantly idiomorphic Mn-enriched oxide (spinel) surrounded by idiomorphic/hypidiomorphic grains of LiAl in a matrix of GCAS and accompanied by a grain of an APh enriched in unusual elements (contaminants), e.g., Ti and K probably representing the last eutectic melt composition.

The gradient of the grey shade of the Mn-oxide grain indicates an increase of the mean atomic number from the center to the rim. At the outer edge, segregation lamellae can be observed. The brighter grey shade of these lamellae indicates a higher mean atomic number than in the surrounding grain. To investigate the changing composition from the inner part of the grain to the rim, several points were analyzed. Due to the relatively small features (<1–2 µm) of the segregation lamellae, the emphasis was on the precise determination of the whiskers. Therefore, the recording of a line scan was omitted.

Table 3. Elemental concentrations (wt.%) at the locations depicted in Figure 5, sorted in ascending order according to the Al concentration. For a close-up of the lamellae region, see Appendix C. The distance from the lamellae region points to the nearest rim is given. Regarding the point scans, a virtual line is drawn through all points to the left rim. The distances are given from each point to the intersection of this line with the rim.

Location	Lamellae Region					Point Scan
No.	379	382	383	380	381	374	378	377	294
Distance from rim in µm	2.6	2.2	6.9	5.6	10.4	10.1	46.5	81.8	101.2
Al	0.7	0.7	1.5	1.5	3.1	6.0	11.9	16.7	22.3
Mg	0.1	0.1	0.1	0.1	0.1	0.1	0.2	0.2	0.5
Ti	0.4	0.4	0.3	0.2	0.1	0.0	0.0	0.0	0.0
Mn	69.7	64.4	62.9	64.0	62.2	60.0	52.3	46.9	39.0
Fe	0.3	0.4	0.5	0.5	0.5	0.4	0.3	0.3	0.3
Ca	0.2	0.1	0.1	0.2	0.1	0.2	0.0	0.0	0.0
Si	0.2	0.2	0.2	0.2	0.1	0.1	0.1	0.0	0.0

contains the original data.

The concentrations found for Al and Mn were used to calculate normalized contents of Al and Mn²⁺, Mn³⁺ as well as the fractions of LiMnO₂, Mn_0.5AlO₂ (1/2 galaxite spinel), and Mn_0.5MnO₂ (1/2 hausmannite) fitting the three spinel components to the elemental amounts (

Table 4. Spinel formulas, calculated with the Al and Mn concentrations of Table 3, sorted in ascending order according to the Al concentration. The Mn²⁺/Mn³⁺-ratio was calculated using the total measured Mn concentration in Table 3. The Li concentration is derived from the calculated fraction of LiMnO₂. In the first section of the table, the concentrations of the elements are given in weight percent. In the second section, the calculated fractions of the different virtual components in percent are presented. In the last section, the stoichiometric ratio of the formula (Li_(2x)Mn²⁺_(1−x))_1+x(Al_(2−z),Mn³⁺_z)O₄ is presented. The distance from the lamellae region points to the nearest rim is given. Regarding the point scans, a virtual line is drawn through all points to the left rim. The distances are given from each point to the intersection of this line with the rim.

Location	Lamellae Region					Point Scan
No.	379	382	383	380	381	374	378	377	294
Distance from rim in µm	2.6	2.2	6.9	5.6	10.4	10.1	46.5	81.8	101.2
Concentrations (wt.%)
Al	0.7	0.7	1.5	1.5	3.1	6.0	11.9	16.7	22.3
Mn2+	21.5	12.2	11.5	13.6	14.5	17.9	19.3	21.8	22.7
Mn3+	48.2	52.2	51.4	50.4	47.7	42.1	33.0	25.1	16.7
Li	0.8	3.7	4.0	3.3	3.2	2.4	2.4	2.0	2.1
Calculated fractions (%)
Mn0.5AlO2	2.1	2.2	4.7	4.9	10.0	19.2	38.2	53.5	71.5
LiMnO2	11.0	49.9	53.6	45.1	42.7	31.8	31.9	26.7	28.5
Mn0.5MnO2	86.9	47.9	41.7	50.0	47.3	48.9	29.9	19.9	0.0
Stoichiometric factors (Li2xMn2+(1−x))(1+x)(Al(2−z),Mn3+z)O4
x	0.13	0.54	0.58	0.49	0.46	0.34	0.33	0.26	0.27
z	1.95	1.95	1.89	1.88	1.76	1.55	1.15	0.85	0.54

). The Li-concentration was subsequently obtained from the calculated amount of LiMnO₂. The general formula obtained from the fittings is (Li_(2x)Mn²⁺_(1−x))_1+x(Al_(2−z),Mn³⁺_z)O₄. The minute concentrations of the other elements are omitted in this calculation. The results show that the Al content in the grain decreases during crystallization. The result of this calculation gives an indication that a solid solution of Li-Al-Mn spinel is formed.

4.4. X-ray Absorption near Edge Structure

To verify the structures suggested from EPMA analysis, XANES was conducted on the sample MUS V-3. For comparison, the spectra of known Mn oxidation states were recorded as well. In Figure 6, Mn K-edge XANES spectra of compounds representing oxidation states from Mn²⁺ to Mn⁴⁺ are displayed for comparison with the spectrum of the MUS.

A shift of the edge to higher energies with increasing oxidation state is observed (Figure 6); this is a well-known effect [27]. From the edge shift and the shape of the curve, it can be concluded that the oxidation state of Mn in the MUS is a mixture of +2 and +3. A mixture of +4 and +2 is unlikely. A combination of both would result in a relatively flat curve, which is not observed in the MUS spectrum. For a better overview of the correlation with oxidation states ranging from +2 to +3, see Appendix B.

According to the results of the Mn K-edge XANES of the reference samples, the average oxidation state of Mn has to be between +2 and +3. Additionally, the analysis via EPMA and PXRD strongly supports the presence of spinel structures involving Mn und Al. The XANES spectra show that Mn is not present in the pure form of any of the analyzed oxides. The same is concluded from the EPMA, which in addition has shown that the Al-percentage in these Mn phases is lower than in pure galaxite.

Accordingly, linear combinations of the spinels galaxite and hausmannite, as well as of galaxite and bixbyite, were calculated. From these linear combinations, the model XANES spectra in Figure 7 and Figure 8 were obtained. These mixtures present a lower Al-content than galaxite, but still have a spinel structure. The results obtained from these linear combinations can be seen in Figure 7 and Figure 8. A linear combination of hausmannite with bixbyite was also calculated. These results are shown in Appendix D.

The linear combinations (Figure 7 and Figure 8) of a 50:50 (mass) mixture of both galaxite and bixbyite, as well as galaxite and hausmannite, are quite similar to the measured spectrum of the MUS V-3. A combination of bixbyite and hausmannite is excluded from further investigation, as the shape is significantly different (see Appendix D). Accordingly, the experimental XANES data was obtained from a 50:50 (mass) mixture of galaxite and bixbyite, as well as galaxite and hausmannite. The obtained spectra are shown in Figure 9 and Figure 10. Figure 10 shows a close-up of the edge-jump.

The spectra displayed in Figure 9 and Figure 10 show that the pre-edge, the edge jump and the region after the edge of the MUS and both references are similar. The results suggest that the slag contains a mixture of Mn²⁺ and Mn³⁺, confirming the EPMA analysis. The combinations of galaxite and bixbyite with an average Mn oxidation state of 2.69—as well as of galaxite and hausmannite, with an average oxidation state of 2.46—mostly match the MUS spectrum. In conclusion, galaxite is present in the MUS.

5. Discussion

The experimental investigation of the influence of Mn on the solidification, and especially on the formation of the EnAM LiAlO₂ in slags of the six-component oxide system (Li, Mg, Al, Si, Ca and Mn) is crucial to understand. This is also indispensable for the phase relations, as well as the reactions in this complex system. It will also help to predict the slag composition and improve thermodynamic modeling. Slags, unlike most geological features, are formed on a short timescale and with high cooling rates. Hence, non-equilibrium thermodynamic modeling will have to be consulted to develop a route to create the desired EnAM.

In contrast to the other elements in this system, Mn is redox-sensitive, occurring in several oxidation states ranging from +2 to +7. Due to the moderate to high oxygen fugacity in the slag, the expected oxidation numbers are +2, +3 and +4, and mixtures thereof. The purpose of this research was to study the suppression of LiAlO₂ formation in Mn-rich Li-ion battery recycling slags. The determination of the Mn-species, including the oxidation state formed in slags, is key to understanding this phenomenon.

Investigations with PXRD and EPMA on Mn-rich MUS reveal that besides LiAl and GCAS, the melt contains large grains of Al/Mn-rich oxides. The PXRD results show that these oxides can be best described as spinel-like compounds. The diffractograms exhibit reflections in the range of the main (311) diffraction line of the spinel-structures MnAl₂O₄ (galaxite), Li_1−xMn₂O₄, Li₂Mn₂O₄ and LiAl₅O₈ (Figure 4). Due to the non-direct matching of these diffraction lines, the best explanation is a spinel solid solution with the elements Li, Al and Mn. With increasing Mn concentration within the melt experiments MUS V-1 to V-3, there is a shift of the diffraction reflection towards galaxite, indicating that the galaxite component is increasing.

The amount of LiAlO₂ seems to be suppressed compared to an Mn-free melt with similar Li-concentration (Figure 4c, black line). Due to the high peak to background ratio, a comparable high amount of amorphous phase can be assumed.

The BSE(Z) micrograph observations show large idiomorphous Mn-rich grains (example see: Figure 5), suggesting an early and complex crystallization scenario. EPMA point scan analyses (Table 3) show a distinct decrease in the aluminum concentration from the center to the rim of the predominant Mn-rich crystals. At the edge, the aluminum concentration drops nearly to zero. Additionally, there is a split into two components, one relatively Mn-enriched and one relatively Mn-depleted. If the composition of all measurements is calculated as fractions of the virtual compounds Mn_0.5AlO₂ (½ galaxite), LiMnO₂ and Mn_0.5MnO₂ (½ hausmannite) a general formula (Li_(2x)Mn²⁺_(1−x))_1+x(Al_(2−z),Mn³⁺_z)O₄ can be calculated. From this calculation, a Li-content is derived and used to assess a gradient of the Li-bearing compounds. In accordance with the elemental gradients, a constant decrease of the galaxite fraction from the center to the rim is observed. In contrast, the hausmannite fraction is increasing. The Li-Mn compound fraction stays more or less constant except for a steep increase at the last ~10 µm from the rim. Directly at the rim, a split into a “normal” and a Mn_0.5MnO₂-dominated region can be observed. The increase and decrease of the individual species over the point scans are shown in Figure 11.

This observation indicates that from the beginning to the end of the crystallization, Li is incorporated into the spinel structure. The spinel composition itself changes from a galaxite-dominated to hausmannite-dominated chemistry. Directly at the rim, the oversaturation of the melt with Mn is such that the spinel solid solution segregates (most probably during cooling down to room temperature) to form two different (most probably spinel-like) oxides.

In this respect, it is interesting that at lower temperatures, the hausmannite converts to the tetragonal crystal system with low solubility of the spinel compound galaxite as reported by Chatterjee et al. [9]. This could indicate an exsolution of the hausmannite component due to crystal lattice incompatibility. The hypothesis is backed by the results from the Mn K-edge analysis, which suggests a mixture of galaxite and Mn²⁺, Mn³⁺ oxide spinels. The virtual Li compound would mix into the cubic galaxite-like spinel phase.

By combining the above results, a scenario of the large crystal genesis is established. The crystallization starts with a high aluminum galaxite-like composition that is subsequently enriched in Mn during the crystal growth. At the end of the crystallization, the solid solution becomes unstable, indicated by exsolution whiskers with a higher mean atomic number, surrounded by the massive crystal.

6. Conclusions

In this study, Mn-rich grains in mock-up slags (system: Li₂O-CaO-SiO₂-Al₂O₃-MgO-MnO_x) were characterized to understand the suppression of the LiAlO₂ formation in Mn-rich Li-ion battery recycling slags. The PXRD, EPMA and XANES data suggest that Mn-rich grains crystallize early on as a spinel solid solution. A generic stoichiometry, i.e., (Li_(2x)Mn²⁺_(1−x))_1+x(Al_(2−z),Mn³⁺_z)O₄ of the solid solution was determined assuming a combination of the virtual components Mn_0.5AlO₂, LiMnO₂ and Mn_0.5MnO₂. From the spatially resolved data, it was concluded that the solid solution is relatively Al-rich at the beginning of the crystallization and becomes depleted during the process. The formation of spinel solid solution with Mn and Al seems to scavenge Li from the melt before the LiAlO₂ crystallization can begin.

In conclusion, the experimental evaluation of mock-up slags has provided valuable insights into the Li, Mg, Al, Si, Ca, Mn and O system, and emphasized the benefit to study model melts and slags. In the future, however, an approach allowing for a faster synthesis of variable composition would be desirable. This approach will help to design suitable EnAM. The extraction of the EnAM from the slag and further processing will be part of subsequent studies.

In addition, it is not clear how these early crystals form on a molecular level. The solid solution could be a product of a solid phase process. It could also be driven by the ion-pair formation in the melt. In this respect, it is crucial to evaluate the primary crystallization fields in the system Li₂O-Al₂O₃-MnO_x in the presence of the other slag compounds Mg, Si and Ca. Despite this, the influence of the viscosity and the oxygen concentration on the early formation of Mn-rich compounds needs to be studied. The impact of viscosity changes, and pair formation in the ionic melt could be accessed by molecular dynamic modeling.