*Article* **Early-Stage Recovery of Lithium from Tailored Thermal Conditioned Black Mass Part I: Mobilizing Lithium via Supercritical CO2-Carbonation**

**Lilian Schwich \*, Tom Schubert and Bernd Friedrich**

IME, Institute for Process Metallurgy and Metal Recycling, RWTH Aachen University, 52056 Aachen, Germany; tom.schubert@rwth-aachen.de (T.S.); bfriedrich@ime-aachen.de (B.F.) **\*** Correspondence: lschwich@ime-aachen.de; Tel.: +49-2418-095-194

**Abstract:** In the frame of global demand for electrical storage based on lithium-ion batteries (LIBs), their recycling with a focus on the circular economy is a critical topic. In terms of political incentives, the European legislative is currently under revision. Most industrial recycling processes target valuable battery components, such as nickel and cobalt, but do not focus on lithium recovery. Especially in the context of reduced cobalt shares in the battery cathodes, it is important to investigate environmentally friendly and economic and robust recycling processes to ensure lithium mobilization. In this study, the method early-stage lithium recovery ("ESLR") is studied in detail. Its concept comprises the shifting of lithium recovery to the beginning of the chemo-metallurgical part of the recycling process chain in comparison to the state-of-the-art. In detail, full NCM (Lithium Nickel Manganese Cobalt Oxide)-based electric vehicle cells are thermally treated to recover heat-treated black mass. Then, the heat-treated black mass is subjected to an H2O-leaching step to examine the share of water-soluble lithium phases. This is compared to a carbonation treatment with supercritical CO2, where a higher extent of lithium from the heat-treated black mass can be transferred to an aqueous solution than just by H2O-leaching. Key influencing factors on the lithium yield are the filter cake purification, the lithium separation method, the solid/liquid ratio, the pyrolysis temperature and atmosphere, and the setup of autoclave carbonation, which can be performed in an H2O-environment or in a dry autoclave environment. The carbonation treatments in this study are reached by an autoclave reactor working with CO2 in a supercritical state. This enables selective leaching of lithium in H2O followed by a subsequent thermally induced precipitation as lithium carbonate. In this approach, treatment with supercritical CO2 in an autoclave reactor leads to lithium yields of up to 79%.

**Keywords:** battery recycling; lithium-ion batteries; metallurgical recycling; metal recovery; recycling efficiency; carbonation; lithium phase transformation; autoclave; supercritical CO2

#### **1. Introduction**

The need for lithium recovery from LIBs is a crucial topic in terms of increased electromobility since lithium is and will remain a relevant element also in next-generation batteries. Lithium is currently industrially, not recycled. Hydrometallurgical research focuses on recovering lithium at the end of the processes; thus, impurities from process additives are possible, and moreover, reagents like Na2CO3 are needed for generating a marketable lithium product. The present study aims to present a method to mobilize lithium without using expensive or environmentally harmful additives: The early-stage lithium recovery ("ESLR") method. This "ESLR" particularly requires a suitable thermal pretreatment, and other elements can then be integrated into existing metal refining processes. The full "ESLR" process investigated here are shown in Figure 1.

For this research's purpose, the publication is structured into three parts: first, state-ofthe-art processes for LIBs recycling are contrasted to innovative research for lithium phase

**Citation:** Schwich, L.; Schubert, T.; Friedrich, B. Early-Stage Recovery of Lithium from Tailored Thermal Conditioned Black Mass Part I: Mobilizing Lithium via Supercritical CO2-Carbonation. *Metals* **2021**, *11*, 177. https://doi.org/10.3390/met 11020177

Received: 19 November 2020 Accepted: 18 January 2021 Published: 20 January 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

transformation. Second, our own research results based on experimental studies are presented and subsequently evaluated in terms of lithium yield and purity. Concludingly, the obtained results are discussed by showing their scientific findings and process technology relevance in comparison to the state-of-the-art.

**Figure 1.** General flowchart of the early stage Li-recovery discussed in this study and the process benefits at a glance.

#### *1.1. State-of-the-Art in Recycling Li-ion Batteries*

LIBs recycling comprises different modules and sequences, leading to alternative process paths. Statements regarding future-dominant process pathways are afflicted with uncertainties due to location and know-how aspects and also because of the heterogeneous and changing scrap stream compositions [1]. However, the available processes, until the point of having generated marketable products, can be divided into preconditioning and metallurgical extraction [2]. The pretreatment steps, in turn, can be asserted to different sectors: deactivation/discharging [3], mechanical processing as dismantling of EV modules and packs to cell level, comminution and sorting by size or physical properties [4] and finally a thermal treatment [5]. Within the metallurgical techniques, there are mainly hydro- and pyrometallurgical processes available [6]. They both comprise benefits and drawbacks; for example, in hydrometallurgy also ignoble elements, like Fe, Al and C, can be recovered, but on the other hand, the processing goes along with comparatively slow kinetics [6]. Depending on individual core objectives, the cells can be charged into a smelter without any pretreatment [3], but regarding a circular economy approach, it is beneficial to consider pretreatment steps [7] in order to maximize resource efficiency. Besides conventional industrial treatments, different studies are in place to give an overview also on innovative emerging recycling paths [8,9] and also approaches to evaluate the environmental impacts of different paths [10–12]. First, the available processes for recycling Li-ion batteries are described, and second, innovative processes for CO2-promoted lithium phase transformations are shown. Therefore, first, indirect carbonation principles and studies are outlined, second, literature on direct carbonation is presented and third, the role of CO2 in a supercritical state is pointed out. Goal of this detailed elaboration is a monitoring of gaps in literature regarding efficient lithium recovery from LIBs.

#### 1.1.1. Thermal Preconditioning

Thermal pretreatments can be carried out, for example, as pyrolysis. Here, the cells are deactivated in the absence of oxygen at temperatures of typically 600 ◦C [13]. Pyrolysis (as well as classical incineration) are thermal pretreatments allowing for a safe cell deactivating and facilitating further downstream recycling without risking a so-called thermal runaway [14,15]. Through chemical cracking and such removal as organic compounds

in gaseous form, which originate primarily from binder, electrolyte, and separator, takes place [13,16]. A major advantage of thermal treatments comprises a safe cell deactivating, thus contributing to risk mitigation in the context of fire incidents. This thermal runaway can occur, for example, during scrap transport, storing, but also by mechanical processing due to this mechanical, electrical or thermal abuse [17]. Several studies report the second advantage of thermal pretreatments, namely an improved detaching of black mass from the cell's current collector foils [7,16,18–21]. Additionally, suitable mechanical preconditioning concepts are required for efficient downstream processing, especially for hydrometallurgical treatments [3]. A mechanical process consists of comminution and sorting for splitting black mass and other cell components, such as casing and current collector foils. Hence, by subsequent mechanical postprocessing, aluminum and copper foils, along with the metallic casing, either aluminum or steel, can be separated as marketable products from the black mass. Black mass then contains all electrochemical active electrode materials [22]. Due to different battery systems on the market, black mass always has different chemical compositions [22].

The separation into individual fractions by means of sieving or physical separation techniques contributes to higher yields of the valuable components and, finally, increases process recycling efficiency [14]. In this way, copper, aluminum and steel can be integrated into their specific recycling processes. Regarding the extracted black mass, two processing alternatives are in place: hydrometallurgical and pyrometallurgical treatments. In the following, these two methods and their challenges for lithium recycling are compared.

1.1.2. Lithium Behavior in Pyro- and Hydrometallurgical Recycling Steps and Need for Early-Stage Li-Separation

Smelting of possibly pelletized black mass with the addition of SiO2 as slag additive in an electric arc furnace has shown that lithium accumulates both in slag and flue dust [23]. Due to its ignoble character, extraction via a metal phase is not possible. As can be seen in Figure 2, a negligible proportion of approximately 0.35% of lithium is accounted for in the alloy produced. Depending on the selected slag system and the amount of slag, increased accumulation in the slag or flying dust can be realized (see Figure 2). Since the slag has a solubility limit for lithium oxide, according to Vest [24], the evaporation of lithium takes place when the corresponding concentration is exceeded. Due to re-oxidation processes, lithium oxide is accumulated in the flue dust (see Figure 2 below, according to Vest [24]). When operating at the lab-scale, smaller quantities of slag are generated, leading to a larger proportion of lithium transferred to the flue dust. The two Sankey diagrams in Figure 2 show a broad distribution of lithium between the three phases slag, flue dust and partly alloy, which is valid for both smelting setups. In order to extract lithium from the produced slag, energy-intensive crushing, classifying and hydrometallurgical purifying are required, but the costs for these treatment steps are currently not covered by lithium's raw material price [3].

**Figure 2.** Lithium distribution after the smelting of black mass in an electric arc furnace. Above: process design aiming for a lithium enrichment in flue dust, based on [23,24]. Below: process design aiming for a lithium enrichment in slag [23]. Reproduced with permission from Stallmeister, C., Schwich, L. and Friedrich, B., Early-Stage Li-Removal—Vermeidung von Lithiumverlusten im Zuge der Thermischen und Chemischen Recyclingrouten von Batterien; published by Thomé-Kozmiensky Verlag GmbH, 2020.

When treating the extracted black mass by hydrometallurgical processing, lithium is not always enriched in one single product fraction, neither, but can be distributed during the multi-step precipitation series in all filter cakes [25]. Firstly, a black mass is typically leached in mineral acid. For this purpose, it is beneficial to conduct the thermal pretreatment as described easing the dissolution process [26]. Here, Shin et al. report the binder's (polyvinylidene fluoride, PVDF) property of not dissolving in acidic solution and disturbing the filtration process after leaching [27]. Yang et al. have shown a strategy to separately incinerate and then hydrometallurgically treat spent anode material in order to purify the C-fraction from lithium impurities, recovering lithium by means of Na2CO3 [28]. When treating the black mass from both cathode and anode, the target products such as copper, iron and aluminum, cobalt, nickel and manganese are cemented or precipitated one after the other. Lithium salt recovery, e.g., as carbonate (Li2CO3), is the last process step in the hydrometallurgical process chain, as suggested by Wang et al. [25]. Wang et al. have proven a lithium leaching efficiency of 98.5% [25]. For obtaining Li2CO3 in the last process step, a carbonation additive like sodium carbonate is used. During the precipitation stages, as can be seen from the data in Figure 3, approx. 27% of lithium remains in other filter cakes. Thus, the purity of the obtained copper, Fe/Al and Ni-Co products is reduced, and the yield of recovered lithium suffers. Moreover, a share of lithium remains in wastewater leading to a complex circuit with continuous neutralization salt removal, again including lithium losses. This complex processing is up to now industrially only viable by recovering the valuable metals nickel, copper and cobalt [3].

**Figure 3.** Lithium distribution after leaching trials of extracted black mass (current trials at IME), based on [23] (data used in [23] was based on experiments by Wang et al. [25]). Reproduced with permission from Stallmeister, C., Schwich, L. and Friedrich, B., Early-Stage Li-Removal—Vermeidung von Lithiumverlusten im Zuge der Thermischen und Chemischen Recyclingrouten von Batterien; published by Thomé-Kozmiensky Verlag GmbH, 2020.

Hence, the process's bottleneck lies in lithium extraction as a solid product instead of lithium leachability (leaching efficiency). A wide range of hydrometallurgical studies shows high leaching efficiencies using different solvents [29–34], for example, by using

H2SO4, a lithium leaching efficiency of 96.7% [35], and by HCl 99.2% [36]. In [37], leaching in citric acid and precipitating lithium by sodium phosphate leads to a leaching efficiency of 99% Li and a recovery as Li3PO4 of 89%. In [38–41] also a lithium precipitation of solid Li2CO3 by Na2CO3 is reported, reaching yields of 80% [38], 90% [39] and 99% [40]. In [38], a corresponding purity of 96.97% is reached; in [39], no purity is given; in [40], the lithium filter cake consists of 10.13 wt.% Li. With a molar ratio of Li/Li2CO3 = 18.79%, this would mean a Li2CO3 content of 53.91%, assuming 10.31 wt.% exists as pure Li2CO3. When using cathode black mass only, a recovery of Li as Li2CO3 of 98.22% with a purity of 99.9% is reported [41]. This means that reaching both a high Li2CO3 purity with a high yield is not straightforward but achievable. Nonetheless, lithium recovery always requires additives, which can be avoided by direct H2O-leaching. Moreover, the volume of required leaching agents can be lowered by H2O-leaching before entering conventional hydrometallurgy due to a mass reduction. In this study, H2O-leaching and using supercritical CO2 are assessed for environmentally friendly and additive-free lithium recovery.

#### 1.1.3. Liquid–Gas Carbonation (Indirect Carbonation)

The hypothesis of this work is a mobilizing of lithium by an "Early-Stage Li-Recovery". Different methods may be applied for this phase transformation to Li-carbonate, which are addressed in the following paragraphs. The use of CO2 for carbonation purposes has been examined for non-lithium materials in numerous studies, e.g., [42–45], and even treating of battery materials with CO2 is possible, as shown above by Hu et al. and Zhang et al. [46,47]. Generally, Kunzler et al. investigated the parameters influencing indirect carbonation, which is understood as a reaction between dissolved elements and CO2. In contrast to that, direct carbonation is defined as a gas–solid reaction for generating carbonates [48]. Kunzler et al. found a correlation between extraction efficiency and grain size of the target metals, solid to liquid ratio, concentration and hence pH value of the leaching agent, and temperature [48].

When aiming for indirect CO2-driven reactions, Hu et al. have conducted a combination of a reductive thermal treatment and H2O-leaching in combination with CO2-gas [46]. Here, only cathode material from NMC (Lithium Nickel Manganese Cobalt Oxide)-cells was mixed with lignite as a reducing agent. According to the maximum reported leaching efficiency of 85% for Li, the optimal thermal-treatment temperature is at 650 ◦C, and the optimal s/l ratio is 10 mL/g (1:10 (g/mL)). Zhang et al. have developed this study further using the same parameters but also investigating optimal CO2 flow rate and leaching time, leading to a Li-recovery of 85% [47]. In summary, when treating the cathode mass only and adding reducing agents like lignite or carbon black to the thermal treatment, lithium yields of 85% [46,47]. However, if not applying CO2 during leaching, the leaching efficiency of lithium is 40% [46,47]. This indicates that the mechanism of indirect carbonation is decisive, but in [46,47], no CO2 atmosphere was used during the thermal treatment; instead, a solid C-carrier was added.

A similar approach is pursued by Jandová et al., where a lithium-containing solution, not stemming from batteries, is treated with CO2 [49]. Then, the solution is heated until the lithium concentration reaches 12–13 g/L, and afterward treated with CO2-gas at a temperature of 40 ◦C for 2 h to generate LiHCO3. Lithium hydrogen carbonate provides a higher solubility in comparison to the first, formed Li2CO3. Finally, the lithium solution is boiled to produce Li2CO3 [46,47,49]. Moreover, an indirect carbonation approach for non-battery materials gives insights into general mechanisms when purging CO2 into aqueous solutions [45]. Within these aqueous treatments, CO2 dissolves as [45,50]:

$$\rm{CO\_2 + H\_2O \leftrightarrow H\_2CO\_3 \leftrightarrow H^+ + HCO\_3^- \leftrightarrow CO\_3^{2-} + 2H^+} \tag{1}$$

The more H-cations are released, the stronger is the resulting acidification [50]. These reactions are to be understood as a function of temperature, pressure and pH [45]. With increasing pH, the dominantly existing phases alternate in the following sequence: H2CO3, HCO− <sup>3</sup> and CO2<sup>−</sup> <sup>3</sup> , hence the higher the pH-value, the more H+-ions are released, contributing to a lowered pH. Especially, CO2<sup>−</sup> <sup>3</sup> is dominant in a pH-area of 10 onwards, whereas HCO− <sup>3</sup> is dominant in an area from 6 to 9, as can be seen in Figure 4:

**Figure 4.** Available CO2-based phases in aqueous solution over the pH [45]. Reproduced with permission from Haug, T. A., Dissolution and carbonation of mechanically activated olivine—Investigating CO2 sequestration possibilities; published by Haug, T.A., 2010.

A reaction between HCO− <sup>3</sup> /CO2<sup>−</sup> <sup>3</sup> and lithium requires the presence of Li+ in the solution. In Table 1, possible lithium phases and their solubility is presented. Connected to that, the chemical reaction formula is presented describing the dissolution of lithium phases in an aqueous solution, without and with CO2-gas purging.

**Table 1.** Solubility of selected lithium phases at 20 and 100 ◦C.


<sup>1</sup> [51], <sup>2</sup> [52] at 18 ◦C.

Yi et al. have also reported the conversion from Li2CO3 in aqueous solution into LiHCO3 by CO2-based carbonation, followed by a chemical purification of the solution and subsequent crystallization of Li2CO3 from a LiHCO3 solution by boiling [53].

If lithium is present as Li2CO3, it decomposes according to Equation (2) [54]:

$$\rm Li\_2CO\_3 \leftrightarrow Li^+ \atop \text{(aq)} \quad + \quad CO\_3^{2-} \text{(aq)} \tag{2}$$

The carbonate ion (CO3 <sup>2</sup><sup>−</sup>(aq)) is the conjugate base of a weak acid (carbonic acid) [55]. Hence, H+-ions are attracted at neutral or acidic areas and consumed from H2O, which generally is present in the ionic form [55–57]. This chemical behavior equals the property of Li2CO3 to be alkaline (see Equation (7)) [58]. In combination with CO2-gas, Yi et al. also report the chemical steps between Li2CO3 dissolution consuming H+-ions from the H2CO3-decomposition (see Equation (3)) [53] and precipitation of solid Li2CO3 according to Equation (4) to Equation (7) [53]:

$$\text{CO}\_2 + \text{H}\_2\text{O} + \text{Li}\_2\text{CO}\_3 \leftrightarrow 2\text{LiHCO}\_3\tag{3}$$

$$\text{LiHCO}\_3^- \leftrightarrow \text{HCO}\_3^- + \text{Li}^+ \tag{4}$$

$$\rm{HCO}\_3^- \leftrightarrow \rm{CO}\_3^{2-} + \rm{H}^+ \tag{5}$$

$$\rm{HCO\_3^{2-}} + \rm{H^+} \leftrightarrow \rm{H\_2O} + \rm{CO\_2} \tag{6}$$

$$\rm Li\_2CO\_3 \leftrightarrow Li^+ + CO\_3^{2-} \tag{7}$$

Hereby, the possible recombinations between aqueous CO2-phases and lithium ions in aqueous phases are shown. These combinations can be transferred to other lithium phases, liberating lithium cations in aqueous solution, too:

If lithium is present as LiF is a black mass, it dissolves in aqueous media according to Equation (8) [59]:

$$\text{LiF} + \text{H}\_2\text{O} \leftrightarrow \text{HF} + \text{LiOH} \tag{8}$$

According to the definition of strong acids and bases [60], HF (pKS = 3.17 [61]) is a strong acid, whereas LiOH (pKb = −0.36 [62]) is very strong base. As a resulting pH-value for dissolving 0.26 g/L at 25 ◦C pH = 7–8.5 is reported [63].

If lithium is formed as LiOH in a black mass, it dissociates in an aqueous solution according to Equation (9) [64–66]:

$$\text{LiOH} + \text{H}\_2\text{O} \leftrightarrow \text{Li}^+ + \text{OH}^- + \text{H}\_2\text{O} \leftrightarrow \text{LiOH} \cdot \text{H}\_2\text{O} \tag{9}$$

the following reaction can take place if CO2 is applied to the system [67]:

$$\text{2 LiOH} \cdot \text{H}\_2\text{O} + \text{CO}\_2 \leftrightarrow \text{Li}\_2\text{CO}\_3 + \text{3 H}\_2\text{O} \tag{10}$$

If lithium is present as Li2O in a black mass, it dissociates to LiOH in aqueous solutions according to Equation (11) [68]. LiOH(aq) is generally stable as lithium hydroxide octahydrate (LiOH · 8H2O) [61].

$$\text{Li}\_2\text{O} + \text{H}\_2\text{O} \leftrightarrow \text{2LiOH} \tag{11}$$

This passage has shown that no study on indirect carbonation by CO2 using whole LIBs black mass, meaning anode and cathode material, is in place. In contrast to that, in the study, real industrial heat-treated black mass from anode and cathode was used without adding a reducing agent. Moreover, this gives the first-time overview of all possible lithium reactions when considering battery materials, which is crucial to extract hypotheses on ongoing mechanisms.

#### 1.1.4. Solid–Gas Carbonation (Direct Thermal Carbonation)

Direct carbonation describes solid–gas reactions for generating a carbonate phase [48]. In this study, it will be investigated by using SCO2. There are different studies in literature optimizing a reductive thermal treatment of black mass for mobilizing lithium via subsequent H2O-leaching [46,47,69–74]. It should be recalled that in [46,47], which were already discussed in chapter 1.1.3, a combination of direct carbonation and indirect carbonation was performed: On one hand, a reductive thermal treatment with adding a carbon-reducing agent like lignite or carbon black contributes to the formation of Li2CO3, hence direct carbonation. On the other hand, CO2 was added during leaching or Na2CO3 was used after a first filtration, both representing indirect carbonation. Therefore, a classification into studies with direct and indirect carbonation is not always straightforward.

However, in all reported studies [46,47,69–74], first, the battery cells are shredded, and, after extracting, a black mass is thermally treated. Battery systems used are LCO-cathode based [69,72,74], LMO-cathode based [71,74], or NMC-cathode based [73,74]. In [73], the only cathode material is used. Most studies focus on a thermal treatment in an inert atmosphere, like a vacuum, Ar or N2 [69,71,72,74,75] or in the air [70] instead of CO2. CO2 was only used in [73]. The performed reductive roasting reportedly also contributes to the carbonation of lithium by precipitating it from the black mass matrix [19,70,72,74,76].

Since the focus of this study is the use of NMC-cathode based black mass, the matching studies are reported in detail: CO2-gas purging at 600, 700 and 800 ◦C (direct carbonation) of NMC-cathode material was performed by Wang et al. for 120 min., but lithium yields are not given. Instead, it is stated that 1.735 g lithium of 1.95 g lithium was transferred to

a solution after water leaching [73]. However, no optimal temperature for carbonation is given; instead, a spectrum of 650–800 ◦C is reported, and no discussion on lithium purity is in place. Xiao et al. treated black mass from NMC-cathode based cells. According to the best-case scenario for pyrolysis and H2O-leaching, 66% of the lithium is recovered. The bestcase scenario here implies vacuum pyrolysis at 700 ◦C for 30 min. in combination with H2Oleaching for 30 min. and an s/l ratio of 1:40 (g/mL) (25 g/L) [74]. Nevertheless, no details on the procedure of NCM-black mass are given, especially regarding lithium recovery.

Since [73] is the only study in place using CO2 for direct carbonation, no detailed yields are quoted from the literature.

In terms of chemical reactions involving lithium phases, only a few data are given, mostly based on thermodynamic simulations [76–78]. Nonetheless, gas–solid reactions are known, e.g., from CO2 absorbing for air purification. Here, a reaction between LiOH and CO2 for generating lithium carbonate is targeted [67].

$$2\text{ LiOH} + \text{CO}\_2 \rightarrow \text{Li}\_2\text{CO}\_3 + \text{H}\_2\text{O} \tag{12}$$

A similar reaction is possible if Li2O is present [79]:

$$\rm Li\_2O + CO\_2 \to Li\_2CO\_3 \tag{13}$$

The reaction is rather a surface reaction, after which CO2 diffuses into the inner part of the Li2O particles. This diffusive process is temperature-supported. For example, at 600 ◦C, the diffusion takes place 10 times faster than at 500 ◦C [79]. In addition, once lithium carbonate is formed, it remains stable in the CO2 atmosphere, even though being in a liquid state, until 1611 ◦C, before decomposing into Li2O [80]. This passage has shown that there is a lack of process details in terms of direct carbonation LIBs black mass by CO2-gas purging. Moreover, the available studies focus on shredding before thermal treatment. In this study, the thermal treatment is performed before a mechanical treatment.

Finally, at this point, no study has reported investigating the influence of solid/gas mechanisms (direct thermal carbonation) in contrast to liquid/gas mechanisms (indirect carbonation) in terms of carbonation either by using supercritical or ambient pressure CO2 and by keeping the process parameters equal. The present research aims to give answers to this question.

#### 1.1.5. The Role of Supercritical CO2 (SCCO2)

When using CO2 for phase transformations, the combination of the liquid and gaseous phase properties is advantageous. For CO2, the supercritical state is reached at a temperature of at least 31 ◦C and a pressure of 73.8 bar [81]. Here, the physical properties of CO2 can be described by a density according to the liquid state and as viscosity equally to the gaseous state, enabling a high efficiency [82]. Chen et al. report on using supercritical CO2 (SCCO2) for carbonating spodumene-based lithium; thus, this study refers to primary lithium production [83]. In their study, also sodium carbonate (Na2CO3) is used as a carbonation agent, and CO2 is added, aiming for a higher carbonate dissolution. It is reported to precipitate as Li2CO3 when reducing the liquid volume, according to Equation (14) [83]:

$$2\,\mathrm{LiHCO\_{3(aq)}} \leftrightarrow \mathrm{Li\_2CO\_{3(s)}} + \,\mathrm{\{CO\_2\}} + \,\mathrm{H\_2O\_{(aq)}}\tag{14}$$

Bertau et al. have dealt with a similar research topic: They suggested a treatment of Zinnwaldite, a lithium ore located in Germany, with SCCO2. It is a promising solution for primary lithium recovery as carbonate [84]. Specific benefits comprise the avoidance of additional chemicals, such as Na2CO3, and significant lithium losses, the economic viability due to the low CO2 price, and the high selectivity by transforming only alkali metal compounds [84]. Moreover, Liu et al. investigated the possibilities of recovering LIB-electrolyte components by means of SCCO2 [85]. Grützke et al. also aimed for LIB-electrolyte recycling using SCCO2, but in contrast to Liu et al., who used synthetic

electrolyte components, whole end-of-life NMC-cells were discharged, deep-frozen and manually opened to extract the electrolyte by supercritical CO2 [86,87]. SCCO2, at, e.g., 40 ◦C and 80 bar in flow-through mode, is used for extracting the electrolyte along with the CO2-stream in a cryogenic trap. A patented technique by Sloop describes a treatment of full batteries with SCCO2, where lithium carbonate from the electrolyte is recovered in the frame of electrolyte removal. The electrolyte removal is reached by dissolving it in the stream of CO2 [81,88].

Rothermel et al. rather focused on graphite recycling options and the achievable graphite purity by making use of supercritical CO2-supported electrolyte recovery (SCCO2) [89]. In 2019, Bertau et al. also reported options to recover lithium from battery black mass [90]. The so-called COOL process, consisting of discharging, mechanical extraction of black mass and a SCCO2 treatment, obtained lithium carbonate with a purity of >99.5% and yield of 60%, but the yield is referring to primary ore treatments, so no information on lithium from black mass is in place [90]. In this context, another benefit is highlighted: consuming CO2 instead of producing CO2 in the context of rising industrial, environmentally harmful CO2-emissions [90]. However, no details about the best-case treatment parameters and the resulting lithium yields in terms of black mass carbonation are given. An earlier patent by Bertau et al. describes lithium recovery from so-called lithium-containing battery residues by SCCO2 for obtaining lithium carbonate [91]. For this material, lithium yields of >90% are reported by using electrodialysis and subsequent addition of carbonates like Na2CO3 or K2CO3, but facts on the corresponding treatment details and parameters are not given. An exemplary process with s/l = 1:40 (g/mL) (50 g/2 L), 4 h at 230 ◦C and 100 bar is described, reporting on a leaching efficiency of 95%. The lithium extraction via electrodialysis takes place in a Li2SO4-solution recovering 98% of the lithium in solution [91]. A lithium yield based on the final, solid product is not given.

This passage has shown that supercritical CO2 plays a role in lithium recovery from different materials, but for LIB-black mass, there is a lack of knowledge regarding decisive process details. This article focuses on lithium carbonation from black mass by means of supercritical CO2.

#### **2. Materials and Methods**

#### *2.1. Recycling Concept with Integrated Early-Stage Li-Recovery*

Under the view of the current process-related drawbacks in conventional recycling processes, and especially in terms of the need for lithium recovery, this study suggests the strategy of an early-stage Li-recovery ("ESLR") process. The method "early-stage" is studied here, describing lithium carbonation before entering acidic leaching or smelting, hence at an earlier position in the recycling chain. This treatment prevents lithium distribution, the further use of additives needed for hydrometallurgical treatments, and costly refining from a slag. The "ESLR" process comprises the following steps, as presented in Figure 5b: After the cells have been deactivated by means of a thermal treatment followed by mechanical processing (shredding and sieving to <1 mm), lithium is enriched in the heat-treated black mass, along with other electrode elements, such as Co, Ni, Mn, and C and partly Al- and Cu foil fragments. This heat-treated black mass is then leached in H2O- to transform and extract the lithium phases and, thus, separated them from the other electrode elements.

**Figure 5.** (**a**) Recycling strategy for lithium-ion batteries (LIBs) at IME with a focus on alternative processing using "ESLR", (**b**) detailed process steps of the "ESLR" process by autoclave/supercritical CO2-carbonation.

As mentioned in Section 1, this phase transformation can be realized by treatment with supercritical CO2 as indirect or direct carbonation. Lithium in the heat-treated black mass is converted into water-soluble lithium hydrogen carbonate (LiHCO3) and lithium carbonate (Li2CO3). Indirect carbonation and H2O-leaching occur simultaneously because the heat-treated black mass is fed in the reactor along with deionized water. In direct carbonation, the heat-treated black mass is subjected to neutral leaching in deionized H2O after the phase transformation of lithium compounds. In both cases, lithium dissolves into an aqueous solution, and the Li-reduced, heat-treated black mass is separated by a first filtration. The first filtration's products comprise:


The Li-containing filtrate must finally be boiled since lithium's solubility decreases from 13.3 g/L at 20 ◦C to 7.2 g/L by heating to 100 ◦C [25]. Moreover, thus, the carbonate precipitation is supported. Either a second filtration step separates the solid carbonate (filter cake) from the residual solution, or the solution is further boiled and is left in the air to dry the carbonate.

#### *2.2. Material Characterization*

The black mass used in the present study to validate the "ESLR" process has been generated by thermal treatment of whole NCM-traction cells. Therefore, a real industrial heat-treated black mass was obtained by thermal treatment at different atmospheres: Ar, 95% Ar + 5% O2 and CO2. For Ar-pyrolysis, the temperatures targeted are 509 ◦C and 603 ◦C. For Ar + O2-pyrolysis, the temperature targeted is 501 ◦C. For CO2-pyrolysis, the targeted temperature is 466 ◦C. These atmospheres have been generated by dynamic pyrolysis in a sealed reactor. It should be noted that due to the experimental setup, not all thermal treatment temperatures have been identical. The thermally treated cells were then subjected to a shredding and sieving process for extracting the heat-treated black mass. The composition of the heat-treated black mass is shown below (see Table 2):

**Table 2.** Chemical composition of the heat-treated black mass used for the autoclave trials pyrolyzed in Ar-atmosphere.


Dynamic particle analysis with QICPIC/L02 (Sympatec GmbH, Clausthal-Zellerfeld, Germany) of the heat-treated black mass reveals that according to the distribution sum (Q3), the d99.3 -value is 101.74 μm. This means that 99.3% of the heat-treated black mass has a grain size smaller than 101.74 μm (see Figure 6). Here, the material's distribution density (q3\*), see Equation (15), can also be extracted, reaching a global maximum at ~95 μm. According to DIN ISO 9276-1, the distribution density is defined as the first derivation of the distribution sum:

$$\mathbf{q}\_3 \mathbf{\*} = \mathbf{q} \mathbf{r}(\mathbf{x}) = \frac{\mathbf{d} \mathbf{Q}\_\mathbf{r}(\mathbf{x})}{\mathbf{d} \mathbf{x}} \tag{15}$$

Furthermore, the distribution sum of the particles represents the number of all particles and not their volume share in the powder. Hence, there are many small particles in the heat-treated black mass, but in contrast to that, the volume of big particles (>101.74 μm) could take up more than 99.3%. Moreover, the largest particle detected in the heat-treated black mass comprises a diameter (EQPC) of 653.74 μm with a FERET\_MAX value of even 748.13 μm. EQPC is defined as

$$\text{xFEQPC} = \sqrt[2]{\frac{\text{A}}{\pi}} \tag{16}$$

Hence, it describes the diameter of a circle whose projection surface (shadow) is identical to the particle. FERET\_MAX, on the other hand, detects the maximum diameter of a particle by analyzing it from 20 different perspectives between 0 and 180◦. This value deviates from the EQPC, especially in terms of irregularly shaped particles; hence, FERET\_MAX takes bigger values, especially when being distinct from a circular form. The maximum detectable particle size with this method comprises 1252 μm.

**Figure 6.** Dynamic particle analysis of CO2-pyrolyzed black mass. A total of 32,650 particles were counted and hence considered for the evaluation. The distribution sum and distribution density in this diagram is based on the heat-treated black mass comprises a diameter (EQPC)-value of the particles, which is indicated as particle size *x.*

More details on the quantitative grain size detected are shown in Figure 7 and can be extracted from the chemical composition of the particles. EDS analyses were carried out for elemental mapping, as well as for point analysis on the black powder, using a ZeissGemini-FE-SEM, equipped with an Oxford UltimMax170 detector (Carl Zeiss AG, Oberkochen, Germany). In Figure 7a,b, the results of the elemental mapping (EDX-layered image) are depicted. Here, the appearance of metallic Al-flakes is shown in Figure 7a; second, a heat-treated black mass particle and a graphite particle are shown in Figure 7b. Moreover, third, Figure 7c is a 1 mm-scale distance shot showing the heterogeneity of the heat-treated black mass in terms of metallic aluminum fragments, heat-treated black mass particles and graphite powder. It should be noted that lithium cannot be displayed by this method. In addition, Figure 7c reveals that the liberation of heat-treated black mass from

the aluminum current collectors could be realized since the particles do not show direct physical contact.

**Figure 7.** EDS analysis of CO2-pyrolyzed EV-battery black mass. (**a**) results of the elemental mapping (EDX-layered image) of an aluminum particle, (**b**) results of the elemental mapping (EDX-layered image) of both an NMC-and-graphite particle, (**c**) Distance shot of the heat-treated black mass from elemental mapping (EDX-layered image).

The following Table 3 shows the chemical composition of the taken spectra. Here, the chemical composition of the Al-flake can be seen by Spectrum 38. However, there are some oxide-graphite-mix particles visible, which results in the chemical composition of just 58.28 wt.%. Moreover, spectrum 16 shows that the dominant particle composition is a Ni-Mn-Co-Oxide, hence, heat-treated black mass particles. Spectrum 33 reflects a graphite particle (C = 46.46 wt.%), which also shows a high amount of oxygen. This may be explained by lithium oxides, hydroxides or carbonates, which cannot be shown here. This theory can be supported by the lithium intercalation in the graphite matrix due to charging.

**Table 3.** Chemical composition of the heat-treated black mass spectra taken by EDS-analysis. Only elements >1 wt.% are shown quantitatively.


These findings are essential to understand the properties of NCM-heat-treated black mass: An important result is the revelation of the material's heterogeneity. Hence, each sample taken shows different chemical compositions. Considering the colorful mixture of relatively big Al-particles, cathode material particles and smaller graphite/soot particles, this statement is supported. Moreover, it can be seen that the Al-foils are liberated from the cathode material. Hence, a thermal pretreatment removes binders and therefore loosens adhesions. This enhances the subsequent leaching efficiency, as indicated before. Moreover, it is shown that fluorine enriches the cathode material particles.

#### *2.3. Neutral Leaching Reference Tests in Deionized H2O*

Neutral leaching comprises reference trials to evaluate the carbonation success. It was performed in a 1 L glassware beaker filled with deionized water. Here, 20 g of the heat-treated black mass is inserted and magnetically stirred at 350 rpm for the defined leaching time. After H2O-leaching, a first filtration obtains a C-filter cake and a Li-bearing solution, then boiled to recover lithium as Li2CO3 (→ Li-filter cake). Neutral leaching is performed at room temperature to increase lithium dissolution. Lithium contents in both filter cakes and solutions are measured by ICP-OES (inductively coupled plasma optical emission spectrometry). Li is the crucial indicator for carbonation success, and hence, lithium yields are calculated as follows:

$$m\_{\rm Li} = \frac{\rm Li\_{Li\_2CO\_3\\_k}(g)}{\rm Li\_{total}\,(g)}\tag{17}$$

where Litotal is the lithium mass in the input, and Li-Li2CO3-fc is the lithium mass in the lithium carbonate filter cake. Since the input material shows deviations in terms of its chemical composition, the Litotal value does not equal the share of the original input analysis. For this reason, the lithium values in the input are calculated as follows:

$$\text{Li}\_{\text{total}}(\text{g}) = \text{Li}\_{\text{C}-\text{fc}}(\text{g}) + \text{Li}\_{\text{solution}}(\text{g}) \tag{18}$$

Here, C-fc represents the carbon filter cake, also indicated as "heat-treated black mass without lithium," and "Lisolution" corresponds to "Li-bearing solution". This calculation is to be contrasted to the leaching efficiency (LE). When calculating *n*Li based on Lisolution [g], the yields in this study were ~10% higher. For example, in one trial with a yield of 79% based on Equation (17), the LE was 88%. The reason for this inaccuracy cannot be determined at this point, but since the deviation was systematic, the authors decided to use the lower values for conservative result interpretation.

#### *2.4. Carbonation by Supercritical CO2*

The unit operation used is a batch 1 l Büchi autoclave reactor operated with deionized water (indirect carbonation) or without any liquid (direct carbonation). The maximum operating temperature is 250 ◦C, and the maximum applicable pressure is 200 bar. After sealing, a stirrer constantly mixes the powder (50 g heat-treated black mass per trial (T0–T9) and 20 g heat-treated black mass per trial (T10–T22) or the suspension in the reactor. The gas flows into the reactor occurs via a valve into H2O if the trial is conducted with an aqueous medium. As soon as the supercritical conditions are reached, a processing time of 120 min is started. The defined holding time of 120 min can be attributed to preliminary studies using autoclave-induced carbonation [92]. The general parameters for the autoclave trials can be seen in Figure 8a. Moreover, Figure 8b shows pictures of the unit operation.


**Figure 8.** (**a**). Fixed process parameters and reached pressures for an autoclave trial (combination of heat-treated black mass with deionized H2O and CO2 gas) in case of trials T1–T9. After the starting pressure of 50 bar, the reactor is heated until reaching 230 ◦C, resulting in different Pmax. (**b**): Used autoclave reactor at IME, RWTH Aachen University.

After leaching, either during autoclave treatment, if H2O is used in the autoclave or after the autoclave, if no H2O is used in the autoclave, a first filtration is performed. The experimental procedure is identical to the procedure described in Section 2.3: Boiling is performed to reduce the volume of liquid and to precipitate Li from the solution. Lithium is obtained in the form of solid Li2CO3 in a subsequent (second) filtration step or by full boiling until obtaining a solid Li2CO3 product within the beaker. The (second) filtration

step is performed as follows: The Li-bearing solution is boiled until reaching 100 mL and then is filtrated, obtaining Li2CO3. The Li2CO3-filter cake is then washed with pure ethanol since lithium carbonate does not show solubility in ethanol. It is dried for at least 24 h and then weighed. The full boiling (indicated as "drying in a beaker") is performed as follows: The Li-bearing solution is boiled until no liquid is left, why the weight of the empty beaker is to be measured before and after performing full boiling. In addition, weighing is done after a drying time of at least 24 h, too. The difference in weight equals the solid carbonate obtained.

An overview of the experimental series described in Sections 2.3 and 2.4 is given in Table 4. Parameter set 1.A, 1.O and 1.C (reference trials) represents H2O-leaching without CO2 addition. Hereby, insights into the mass of water-soluble Li-phases already present in the heat-treated black mass are provided. Enhanced leaching efficiencies are obtained by combining neutral leaching with CO2-carbonation. (see experimental series 2.A, 2.O and 2.C). Moreover, carbonation trials were conducted in the autoclave as well by using argon (Ar) as process gas aiming for the same excess pressures as needed for the supercritical state (73.8 bar) (see Table 4, Parameter set 3.A). Ar as inert gas can deliver knowledge on the main mechanism for ongoing phase transformations: Either the presence of CO2 or the extreme pressure. Parameter set 4.A detects the influence of an autoclave setup without H2O and with CO2 to find out whether gas–solid or gas–liquid reactions dominate in carbonation.

**Table 4.** Parameter matrix for combining pyrolysis conditions with autoclave conditions in this study. Reference trials represent H2O-leaching without autoclave or CO2-incorporation.


The labels of the trials are to be understood as follows: "*number.letter*", where the *number* stands for an experimental series: 1 = neutral leaching in H2O without an autoclave, 2 = autoclave operated with SCCO2 + H2O, 3 = autoclave operated with Ar + H2O, 4 = autoclave operated with SCCO2 and without H2O. The *letter* stands for the pyrolysis atmosphere: A = Ar-pyrolysis, O = 95 vol % Ar + 5 vol % O2-pyrolysis, C = CO2-pyrolysis. In this study, only experimental series 1 and 2 take different pyrolysis atmospheres into account. Hence, the fields of experimental series 3.O/3.C and 4.O/4.C are not experimentally conducted yet (n/a).

#### **3. Results**

In this Section, the results of the trial series 1.A—4.A, 1.O/2.O and 1.C/2.C are discussed by lithium yields. Moreover, Sankey diagrams of the lithium distribution and bar charts on the impurities of the lithium filter cake are shown for trials 1.A.1 and T2 (2.A). The evaluation calculations are performed, as described in Section 2.3.

#### *3.1. Neutral Leaching in Deionized H2O*

For a profound understanding of water-soluble lithium phases already present in the heat-treated black mass, the following results can be obtained. In total, 40 experiments were conducted in experimental series 1.A, 1.O, and 1.C. The amount per parameter set comprises one trial, except for trials 1.A.2-1.A.4, 1.A.5-1.A.7, 1.A.28 and 1.A.40. Since the results are in good accordance, the other trials have not been repeated. All trials have in common that when charging heat-treated black mass in H2O, the pH value of the solution has become alkaline (pH = 11–12).

Figure 9 accordingly reveals how lithium yields from leaching heat-treated black mass in H2O (neutral leaching) depend on six main parameters:


**Figure 9.** Lithium yields when applying no carbonation (neutral leaching) dependent on the parameters used. Trials 1.A.26 and 1.A.34 are left out of this overview since they comprise kinetic trials, for which a yield calculation is not possible due to heat-treated black mass losses during sampling.

Here, a variation of the selected parameters has an impact on the lithium yield. To evaluate the key influencing factors, representative trials are extracted and shown in detail. For efficiency reasons, only the results of experimental series 1.A are depicted for the evaluation according to the parameters selected. In terms of these influencing factors, the following conclusions are possible:

#### 1. **Washing of C-filter cake with deionized water:**

The detailed observation of trials whose parameter combination was equal apart from the washing of the C-filter cake shows that washing is highly beneficial. During the filtration of the C-filter cake, there are physical depositions of the Li-bearing solution left in the C-filter cake. Hence, washing with deionized water liberates the C-filter cake from remaining lithium ions (see Figure 10a).

**Figure 10.** Detailed observation on achievable lithium yields by neutral leaching (H2O) of heat-treated black mass. (**a**) Parameter 1: washing of C-filter cake. (**b**) Parameter 2: filtration of Li-filter cake or full boiling.

#### 2. **Filtration of Li-filter cake or full boiling:**

Filtration is, according to Figure 10b, not an adequate tool when applying neutral leaching in H2O. However, this analysis only focuses on lithium distribution. Hence, no information on filter cake impurities, e.g., F is given here.

#### 3. **Leaching time:**

The optimal leaching time cannot be extracted from the performed trials, as can be seen in the range between 5 and 90 min (see Figure 11a) and between 30 and 120 min (see Figure 11b). Here, no significant improvement of dissolved lithium is achieved when comparing trials with constant parameters except for the leaching time.

**Figure 11.** Detailed observation on achievable lithium yields by neutral leaching (H2O) of black mass. The considered parameter is parameter 3: leaching time. (**a**) Dissolution and lithium recovery by comparing trials with a leaching time between 5 and 90 min. (**b**) Dissolution and lithium recovery by comparing trials with a leaching time between 30 and 120 min.

> Reduced lithium shares over time can be explained by slight deviations in the chemical composition. Hence, deviations in lithium yields are also possible, also due to different lithium phases in the heat-treated black mass. Therefore, a kinetic trial can be found in Figure 12. It can be seen that lithium compounds in the heat-treated black mass of the Ar-pyrolyzed battery cells dissolve as ions instantly. Although the lithium yield can be found below Figure 12, it cannot be directly transferred to the other neutral leaching trials since the amount of heat-treated black mass, and therefore the lithium-bearing input is reduced each time a sample is taken. Samples were taken from the leaching liquor by

using a particle filter, why redirecting the lost particles to the liquid was not possible. In an upscale setup, this mass reduction would show a lower impact. The calculation of lithium yields is based on a reduced leaching liquor volume by sample extraction. The last sample is taken, at 125 min, shows increased lithium mass, which can be explained by analytical deviations.

**Figure 12.** Kinetic trial for lithium dissolution in deionized H2O at an s/l ratio of 1:30.

#### 4. **Particle size of heat-treated black mass:**

As already reported, 99.3% of black mass particles have a grain size below 101.74 μm. In order to reduce the grain size of the few particles above this threshold, the heat-treated black mass was ground in a planetary mill. The aim of this approach was to detect the correlation between smaller grain size and an eased liberation of lithium compounds in neutral leaching. In Figure 13a, no difference with or without grinding is detected. In Figure 13b, this trend shows slightly irregular behavior when comparing trial 1.A.19 to trial 1.A.20. However, in no parameter combination and hence, trial pair compared, grinding to <90 μm has shown an improved lithium yield. This can be explained by the grain size distribution shown before: The majority of the particles shows grain sizes below 100 μm.

**Figure 13.** Detailed observation on achievable lithium yields by neutral leaching (H2O) of heat-treated black mass. The considered parameter is parameter 4: particle size. (**a**) Dissolution and lithium recovery by comparing trials with a solid/liquid ratio of 1:22.5. (**b**) Dissolution and lithium recovery by comparing trials with solid/liquid ratio of 1:15.

Comparing trials 1.A.2-1.A.4 with trials 1.A.5-1.A.7 and trial 1.A.8 with trial 1.A.20 reveals that the grain size of the heat-treated black mass does not influence lithium yields. This is particularly interesting for residual lithiation in the anode. It proves that the degree of liberation of lithium is not enhanced by grain size reduction to <90 μm.

#### 5. **Solid/liquid ratio (g/mL):**

At a constant grain size and leaching time, higher lithium yields are obtained at a solid/liquid ratio of 1:30 in comparison to 1:10 and 1:15 (see Figure 14a). In addition, an improved lithium recovery is possible when comparing a solid/liquid ratio of 1:30 and 1:22.5 (see Figure 14b). Although the leaching time has not shown an impact, the highest yield is reached with a solid/liquid ratio of 1:30 for 120 min. (see trial 1.A.25). When considering the solubility product of lithium carbonate in the water at 20 ◦C (13.3 g/L), and a lithium share of 3.7 wt%, a liquid volume of 294 mL is required for full dissolution, assuming an inserted heat-treated black mass weight of 20 g. Thus, if all lithium is present as lithium carbonate, a solid/liquid ratio of 1:15 g/mL is needed. Since the input material (heat-treated black mass) shows deviations in lithium shares and phases, the findings of a 1:30 solid/liquid ratio are supported. This means that an excess of H2O is needed for high lithium dissolution. In Figure 14, examples of the solid/liquid ratio's impact on lithium yields are given. More trial comparisons would be 1.A.11 with a yield of 38% at a ratio of 1:10, and 1.A.10 with a yield of 44% at a ratio of 1:15. There are also trial combinations where an increase of the solid/liquid ratio leads to equal lithium yields (e.g., 1.A.15 with 1.A.9, leading to 45% lithium yield), but generally, yields of >60% can be reached only when having 20 g/600 mL (1:30).

**Figure 14.** Detailed observation on achievable lithium yields by neutral leaching (H2O) of heat-treated black mass. The considered parameter is parameter 5: particle size. (**a**) Dissolution and lithium recovery by comparing trials with a solid/liquid ratio of 1:22.5. (**b**) Dissolution and lithium recovery by comparing trials with solid/liquid ratio of 1:15.

#### 6. **Pyrolysis temperature:**

The pyrolysis temperature plays an important role in lithium recovery, as can be seen in Figure 15. Here, the difference between a 501 and a 603 ◦C pyrolyzed material is pointed out. Reaching higher temperatures leads to different phase transformations within the battery cells. The impact on lithium leaching efficiency and lithium yield as solid lithium carbonate is proven by different scenarios:

Here, both grain size and leaching time do not show a significant impact on the yield. The solid/liquid ratio, along with the washing of the C-filter cake and the solid–liquidseparation method (filtration vs. full boiling), seems to play an important role in this context. Up to 64% of lithium can be recovered as lithium carbonate. In addition, the parameter *pyrolysis temperature* has an impact on the lithium yield. Lithium yields by leaching heat-treated black mass without preliminary pyrolysis were not satisfying; hence, these first trials are not shown in this manuscript. It must be recalled that the pyrolysis trials at Ar-atmosphere were operated at higher temperatures than the CO2 and Ar + O2 pyrolysis trials.

**Figure 15.** Detailed observation on achievable lithium yields by neutral leaching (H2O) of heat-treated black mass. The considered parameter is parameter 6: particle size. (**a**) Dissolution and lithium recovery by comparing trials with a solid/liquid ratio of 1:22.5. (**b**) Dissolution and lithium recovery by comparing trials with solid/liquid ratio of 1:15.

> For evaluating autoclave trials in terms of lithium mobilization, the lithium yields from neutral leaching are to be contrasted to the lithium yields from autoclave trials using the same parameters (see Figure 16). Since the autoclave trials were operated at a holding time of 120 min, the following diagram points out the achievable maximum lithium yields dependent on the pyrolysis temperature/atmosphere and solid/liquid ratio examined in the autoclave trials. Hereby, a direct comparison between neutral leaching (experimental series 1.A, 1.O and 1.C) and autoclave carbonation (2.A) can be performed.

**Figure 16.** Best of lithium yields dependent on the pyrolysis atmospheres and temperatures. 1.A.36 has not been leached for 120 min, yet.

> Since the focus of this study was the Ar-pyrolyzed material since showing the best neutral leaching results, only for this material the solid/liquid ratios were examined in the autoclave trials (1:10, 1:15, 1:30) (series 2.A, 2.O and 2.C). The CO2- and Ar + O2 -the pyrolyzed black mass was only treated in the autoclave carbonation set up with a solid/liquid ratio of 1:10 (2.O and 2.C). Hence, Figure 17 sums up the maximum yields of neutral leaching dependent on the pyrolysis parameters examined so far:

#### *3.2. Carbonation by Supercritical CO2*

Finally, the obtained lithium yields when using autoclave treatments with an s/l ratio of 1:10 for lithium carbonation can be derived from Figure 17.

Hence, a direct comparison between the atmospheres of the thermal treatments shows the following results: The 509 ◦C Ar-pyrolysis, that the autoclave can make a 12% difference in lithium yield. In comparison to the 603 ◦C Ar-pyrolysis, this difference can reach up to 24% with the correct parameter combination (120 min.). For the Ar + O2-pyrolysis, which is here indicated as thermolysis since comprising O2 in the atmosphere, the increased lithium yield comprises up to 27%. For the CO2-pyrolysis, the obtained difference in lithium yield comprises up to 37%. This indicates higher lithium yields for reductive pyrolysis atmosphere (CO2 vs. Ar-atmosphere at ~500 ◦C) and a stronger impact of autoclave carbonation when dealing with a not fully decomposed heat-treated black mass. This correlation needs further investigations in the future.

The elemental lithium distribution and the lithium carbonate impurities are shown exemplarily for the trial series 1.A with a solid/liquid ratio of 1:10. In Figure 18, the largest part of lithium remains in the heat-treated black mass after leaching. Moreover, the main impurity of the recovered lithium carbonate is fluorine, followed by phosphorous. This can be explained by the presence of LiF in the heat-treated black mass. It should be noted that the value "Li in filtrate" does only occur within neutral leaching and autoclave trials T0–T9 and T21 and T22, which have conducted a filtration.

**Figure 18.** Lithium distribution without autoclave carbonation as exemplary data from the parameter set 1.A.1 by ICP-OES. (**Above**): Ar-pyrolysis in combination with neutral leaching at a solid/liquid ratio of 1:10. (**Below**): Matching impurities within the lithium filter cake by ICP-OES and lithium carbonate impurities.

Figure 19 shows the improvement in Li distribution when applying autoclave carbonation. Trials series 2.A was selected since the neutral leaching trials of Ar-pyrolyzed active mass at 600 ◦C has shown the best yields. Trial series 2.A represents Ar-pyrolysis, with a CO2 + H2O autoclave-reaction, and also a solid/liquid ratio of 1:10 in the autoclave. In this case, the share in the residual heat-treated black mass filter cake is significantly lower, which is a proof-of-concept of the carbonation mechanism within the autoclave.

**Figure 19.** Lithium distribution with autoclave carbonation as exemplary data from the parameter set 2.A (T2 2.A) by ICP-OES. (**Above**): Ar-pyrolysis in combination with neutral leaching with carbonation by supercritical CO2 + aqueous medium at a solid/liquid ratio of 1:10). (**Below**): matching impurities within the lithium filter cake.

Hence, impurities in the range of 2–4 wt.% can be derived. An XRD-evaluation gives more information on the arising phases within the heat-treated black mass, the C-filter cake and the lithium carbonate filter cake (see Figure 20). This is represented here exemplarily by the 603 ◦C-Ar-pyrolyzed samples, thus for trial series 2.A, also with a solid/liquid ratio of 1:10. One main finding is the removal of Li2CO3, present in the heat-treated black mass, from the C-filter cake. This is an indicator for the removal of water-soluble compounds. In contrast to Figure 20, XRD-evaluations of CO2-pyrolyzed black mass at 466 ◦C and Ar + O2-pyrolyzed black mass at 501 ◦C also detect LiNiMnO- and the NiO, which stands for an incomplete decomposition of transition metal oxides.

Small amounts of fluorine can be found in the Li-filter cake in the form of LiF. It can be seen that especially fluorine removal is crucial for reaching high lithium carbonate purities fluorine. Figure 20 shows the diffractogram of the heat-treated black mass and the C-and Li-filter cakes (T3 2.A.). X-ray diffraction was performed at room temperature using a STADI P (STOE Darmstadt) powder diffractometer using an IPPSD detector and monochromatic Cu-Kα1 radiation (*λ* = 1.54059 Å; flat sample; 1.5 ≤ 2*θ* ≤ 116◦ step rate 0.015◦ in 2*θ*) with a measuring time of 2 h.

LiF was still present in the C-filter cake; hence, the solid/liquid ratio was optimized. In order to prove influencing factors on the lithium yield by an adjusted solid/liquid ratio, the parameter 1:15 (solubility of 13.3 g/L lithium carbonate at 20 ◦C) and 1:30 (solubility of 7.2 g/L lithium carbonate at 100 ◦C) were tested. Moreover, to prove the mechanism of autoclave carbonation, two parameters were examined additionally: autoclave carbonation by Ar-excess pressure (3.A) and direct and dry autoclave carbonation by CO2-excess pressure (4.A). Again, since the Ar-pyrolyzed black mass has shown the highest yields in terms of neutral leaching and in terms of autoclave carbonation, only Ar-pyrolyzed black mass is chosen for the parameter improvements.

Table 5 sums up the parameters for the second autoclave carbonation with solid/liquid ratios of 1:15 and 1:30.

**Figure 20.** Graphic pattern of the Ar-pyrolyzed black mass at 603 ◦C (**a**), and the corresponding C-filter cake (**b**) and Li-filter cake (**c**) from trial T3 2.A. The XRD-evaluation was performed using the "match!" Software and the COD Inorganics database.



The following illustration (see Figure 21) shows the results of autoclave carbonation with solid/liquid ratios of 1:15 and 1:30:

**Figure 21.** Lithium yields obtained by autoclave carbonation with a solid/liquid ratio of 1:15 (T10–T16, and T21) or 1:30 (T17–T20, and T22). The C-filter-cakes of T15, T18 and T19 were washed, T17 was leached for 5 min and T20 for 90 min. T21 and T22 were filtrated instead of fully boiled.

Thus, the only process window leading to satisfying yields of 79% is an s/l ratio of 1:30 in combination with CO2 carbonation in an aquatic medium. The underlying mechanism seems to be indirect carbonation.

This can be supported by the detected pH value of all trials. Whereas 1.A.1—1.A.40, T10–T13, and T20 (without CO2 purging in the liquid) showed a pH value of 11–12 after charging heat-treated black mass in H2O, the trials T0–T9. T14–T16, T18/T19 and T21/T22 (with CO2 purging in the liquid) showed a pH value of 7–8 after charging heat-treated black mass in H2O. Hence, CO2 was dissolved in the liquid. The generally higher yields in T10–T20 can also be attributed to the avoidance of a second filtration step for recovering lithium. Instead, the solution was boiled until reaching a slurry-state and then was dried in a beaker. Hereby, lithium losses in the residual filtrate are avoided. In addition, when comparing T17 to T20, the advantage of a longer leaching time and washing of the carbon filter cake with deionized H2O is shown. The washing of the carbon filter cake generally leads to higher yields since dissolved lithium remaining in the filter cakes in the solution can leave the system just by washing. However, comparing T15 to T14 and T16 in terms of C-filter cake washing reveals a rather small impact on the lithium yields (max. 2%). When comparing T21 to T14 and to T16, it can be seen that filtrating of the lithium solution is not expedient. This can be explained by the residual lithium dissolution in the filtrate. The comparison between T22 and T18–T19 confirms this relation. T21 and T22 show very low yields. Although the lithium filter cake was filtrated instead of full boiling and the C-filter cake was not washed, their lithium yield shows disproportionally low yields, which can only be explained by heterogeneity in the heat-treated black mass.

#### **4. Discussion**

This study proves the concept of indirect carbonation for treating lithium-ion battery heat-treated black mass with supercritical CO2. The involvement of supercritical CO2 in terms of lithium carbonate generation is supported by yields comparing Ar-excess pressure and CO2-excess pressure. Moreover, indirect carbonation is shown by comparing a dry autoclave process to a liquid-based autoclave process. The lower pH-value of pH = 7–8 when applying CO2 in comparison to H2O-leaching (pH = 11–12) can lead to the following statements:

1. When leaching heat-treated black mass in H2O, the solution is basic. This can be attributed to the dissolution of basic phases in the liquid. → LiF and Li2CO3 could be detected in the heat-treated black mass by XRD; both phases are slightly soluble and therefore are responsible for the elevated pH-value. Although LiOH and Li2O could not be detected via XRD-analysis in the heat-treated black mass, they may be present in small amounts since the SEI-layers consist of Li2CO3, LiF, LiOH and Li2O [93]. However, it was shown that LiF decomposes to HF and LiOH in aqueous solutions, which indicates Li<sup>+</sup> + OH<sup>−</sup> in the solution.

	- a. The formation of carbonic acid and thus the formation of CO3 <sup>2</sup><sup>−</sup> and HCO3 − as acidic leaching agents. CO2 is added to a basic solution; it reacts acidic by the release of H+ ions. This pH-value decrease can be responsible for a higher leaching efficiency by creating quasi-acidic leaching conditions similar to conventional hydrometallurgy.
	- b. Recombination of Li+, stemming from non-lithium carbonate phases like LiF, with present CO3 <sup>2</sup><sup>−</sup> or HCO3 −. This would entail the following suggested equations (see Equations (19) and (20), schematically shown in Figure 22):

$$\rm Li^{+} + \rm CO\_{3}^{2-} \rightarrow \rm Li\_{2}CO\_{3} \tag{19}$$

$$\rm Li^{+} + \rm HCO\_{3}^{-} \rightarrow \rm LiHCO\_{3} \tag{20}$$

c. A combination of both mentioned mechanisms. In this way, the dissolution of lithium phases in the heat-treated black mass is promoted by CO2, more lithium ions can be formed to Li2CO3, and this effect is also promoted by the increased operating temperatures and arising excess pressure.

**Figure 22.** Schematic process visualization of indirect carbonation promoted by supercritical CO2 based on CO3 <sup>2</sup><sup>−</sup> (**a**) and HCO3 − (**b**) in terms of leaching lithium-ion battery heat-treated black mass in deionized water. When increasing the solution's temperature, lithium carbonate is precipitated as a solid lithium salt.

#### **5. Conclusions**

The presented "ESLR" process, consisting of thermal treatment, mechanical comminution and a sorting step, followed by a subsequent carbonation process, results in the following scientific findings:

Carbonation by supercritical CO2 shows an increased lithium yield of around 15%. This value stems from the difference between a maximum lithium yield in neutral leaching of 64% and a maximum lithium yield in autoclave carbonation of 79%. When expressing the yield as leaching efficiency, 88% were reached. The different pyrolysis atmospheres and temperatures show a direct influence on the lithium yield. Further key influencing factors for both H2O-leaching with and without CO2 are solid/liquid ratio, filter cake washing

and the lithium extraction method (filtration vs. full boiling). It can be concluded that the "ESLR" process shows benefits in comparison to simple H2O-leaching and that the mechanism for indirect carbonation is beneficial. Moreover, the "ESLR" process is a separate step to ease Ni/Co/Mn recovery and to enhance the degree of lithium mobilization. Hence, the resulting lithium-reduced filter cake (C-filter cake) can be integrated into existing hydroor pyrometallurgical steps.

The process of technology relevance is shown by the following specific benefits in contrast to the state-of-the-art:


In contrast to other studies, the sequence of thermal and mechanical treatment is inverted. In this study, battery cells are first thermally treated and then shredded to extract heat-treated black mass. This procedure is safer due to the avoidance of ignition during shredding.

Comparing lithium yields by H2O-leaching in this study to literature, the following statements can be given: In [74], 66% of lithium from NMC black mass are obtained by shredding, then thermal treating and H2O-leaching. Here, the authors rather focus on LMO-cells and report on one trial, only reaching 66% [74]. However, in this paper, 64% could be recovered by thermal treatment with subsequent shredding and H2O-leaching at a thermal treatment by 100 ◦C lower than Xiao et al. and without costly vacuum operations. In comparison to [46,47], where 40% of lithium could be recovered by shredding, thermal treatment and H2O-leaching of cathode black mass, the yield in this study are up to 24% higher.

A comparison of lithium yields by H2O-leaching in combination with CO2 (indirect carbonation) is not straightforward since there is no study in place using the whole black mass from NMC-cells for this process. However, by using cathode black mass with lignite, 85% is reached, whereas, in this study, 79% are reached. This difference might be attributable to the neglection of anode material and/or the use of lignite instead in [46,47]. In comparison to [90], the yields in this study are 19% higher (60% vs. 79%), but yield and matching parameters are given based on a lithium ore treatment. Only the transferability to black mass is mentioned. However, this study also uses heat-treated black mass in contrast to [90]. In comparison to [91], and avoidance of electrodialysis in a Li2SO4-solution and of carbonation reagents could be reached.

A comparison of lithium yields by a thermal CO2-treatment with subsequent H2Oleaching (direct carbonation) is hardly possible since the autoclave process in this study worked at Tmax = 230 ◦C, whereas literature focuses on elevated temperatures (~650–800 ◦C [73]) with CO2 as purging instead of excess pressure; moreover, no yield calculation is given [73].

In this study, a proof-of-concept regarding the indirect carbonation using supercritical CO2 in an autoclave could be shown.

The most important follow-up research comprises a further enhancement of lithium yields to a value of >90%, which is necessary to make the "ESLR" a competitive process option. Then, CO2-driven carbonation without supercritical CO2, but by CO2-gas purging instead. This is crucial because the combination of thermal pretreatment and an autoclave treatment comprise high energy requirements. However, as reported in Section 1.1.1, thermal conditioning is also beneficial for hydrometallurgical treatment. Hence, the connected energy demands cannot particularly and only be counted for the "ESLR" process. First trials with CO2-gas instead of SCCO2 have shown lithium yields around 70%. Hereby, insights into the role of excess pressure (73.8 bar) and high temperatures (150 ◦C) are possible. Moreover, this setup would imply economic benefits due to the avoidance of high-pressure operations. This will be one topic of "Early-Stage Recovery of Lithium from Tailored Thermal Conditioned Black Mass Part II: Mobilizing Lithium via gaseous CO2-Carbonation". Moreover, a refining of the C-filter cake by flotation or acidic leaching should be tested. Upscaling is planned for future research to test possible scale effects due to losses on equipment surfaces, for example, on beakers after boiling the lithium filtrate. In addition, a suitable development for removing fluorine from the heat-treated black mass, filtrates and filter cakes would be an important tool for hazardous-free processing, which would not harm the used equipment by developing HF-gas. Moreover, experimental series 3.O/3.C and 4.O/4.C are to be performed. Moreover, the heat-treated black mass-producing pyrolysis was conducted with a holding time of 60 min. This may be optimized as well to find the perfect match in terms of temperature and holding time.

**Author Contributions:** Conceptualization, L.S. and B.F.; methodology, L.S.; validation, L.S., T.S. and B.F.; formal analysis, L.S. and T.S.; investigation, L.S. and T.S.; resources, L.S. and T.S.; data curation, L.S. and T.S.; writing—original draft preparation, L.S.; writing—review and editing, L.S.; visualization, L.S.; supervision, B.F. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was partly funded by the Research Council of Norway in the frame of the LIBRES project (project number 282328).

**Institutional Review Board Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author. The data are not publicly available since being part of active research.

**Acknowledgments:** In cooperation with Access e.V. EDS-analyses were enabled, and in cooperation with the Institute of Inorganic Chemistry (IAC) at RWTH Aachen, XRD-measurements were possible. We are grateful for the support of both partners.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


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

**Alena Wittkowski 1, Thomas Schirmer 2, Hao Qiu 3, Daniel Goldmann <sup>3</sup> and Ursula E. A. Fittschen 1,\***


**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*<sup>x</sup>* 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(2*x*)Mn2+(1−*<sup>x</sup>*))1+*x*(Al(2−*<sup>z</sup>*),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.

**Keywords:** lithium; engineered artificial minerals (EnAM); X-ray absorption near edge structure (XANES); powder X-ray diffraction (PXRD); electron probe microanalysis (EPMA); melt experiments

#### **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

**Citation:** Wittkowski, A.; Schirmer, T.; Qiu, H.; Goldmann, D.; Fittschen, U.E.A. Speciation of Manganese in a Synthetic Recycling Slag Relevant for Lithium Recycling from Lithium-Ion Batteries. *Metals* **2021**, *11*, 188. https://doi.org/10.3390/met11020188

Academic Editor: Mark E. Schlesinger Received: 30 December 2020 Accepted: 18 January 2021 Published: 21 January 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

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.

**Figure 1.** Sketch of engineered artificial minerals (EnAM) formation. On the left, different elements are shown in the liquid slag-matrix: target element (yellow triangle), phase for EnAM (red) and other phases (green). The upper route refers to an unmodified slag–the target element is spread over several different phases and the matrix (blue). The target element is not recovered. The bottom picture illustrates the formation of EnAM. The target element is concentrated in a single phase (red pentagon) 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*x*Mn*y*Co*z*O2 (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, Mn0.5AlO2 (1/2 galaxite spinel) and Mn0.5MnO2 (1/2 hausmannite), the formula (Li(2*x*)Mn2+(1−*<sup>x</sup>*))1+*x*(Al(2−*<sup>z</sup>*),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, laboratorybased 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-MnOx with Focus on Spinel Structures*

The stoichiometric spinel LiMn2O4 and several other spinel phases of Li-Mn-O are known in the Li2O-MnO*<sup>x</sup>* 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−*<sup>x</sup>*)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 LiMn1.75O4. Raising the temperature above 880 ◦C leads to the replacement of Li by Mn and tetragonal spinel [Li(1−*<sup>x</sup>*)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<sup>+</sup> 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-MnOx*

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 [14–16].

#### **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).

**Figure 2.** Temperature program of the chamber furnace. A heating rate of 2.89 ◦C/min was employed to reach the melting temperature of Li2CO3 at 720 ◦C, followed by a heating rate of 1.54 ◦C to the target temperature of 1600 ◦C. The temperature was held for 120 min, and afterward, a cooling rate of 0.38 ◦C/min was employed to reach a temperature of 500 ◦C.

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 SXFIVE 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 K*α* 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).


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

#### *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.

**Figure 3.** Scheme of the Rowland circle monochromator. The outer circle represents the curvature of the SBCA. The inner circle has a diameter matching the radius of the curvature circle. On this inner circle, the X-ray source and the detector are positioned. The sample is positioned right in front of the detector. The components are set on a scissor drive, allowing fine energy scanning.

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+Al2 3+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: *μ*(E) = − ln *I*·*I* −1 <sup>0</sup> . 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. 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].


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

#### *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.

**Figure 4.** (**a**) Powder X-ray diffraction (PXRD) of the solidified melt. G: Gehlenite, S: Spinel, L: LiAlO2. (**b**): Enlarged section of the main spinel peak. \* 1: the position of the main peak of MnAl2O4 from the ICDD-PDF2 no. 00-029-0880 [17], \* 2: the position of the main peak of the Li(1−*<sup>x</sup>*)Mn2O4 spinel from the ICDD-PDF2 no. 00-038-07891 [17], \* 3: the position of the main peak of the Li2Mn2O4 spinel from the ICDD-PDF2 no. 01-084-1524 [17], \* 4: the position of the main peak of the LiAl5O8 spinel from the ICDD-PDF2 no. 00-038-1425 [17], (**c**)**:** Enlarged sections of the first two main LiAlO2 peaks. In (**c**): for comparison, the LiAlO2 main reflection of an Mn-free solidified melt with comparable Li2O content is presented.

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 (Li1−*<sup>x</sup>*Mn2O4 and Li2Mn2O4) [17]. This indicates the presence of a complex solid solution of a spinel-like oxide with the general formula (Li(2*x*)Mn2+(1−*<sup>x</sup>*))1+*x*(Al(2−*<sup>z</sup>*),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 spinellike 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(2*x*)Mn2+(1−*<sup>x</sup>*))1+*x*(Al(2−*<sup>z</sup>*),Mn3+*z*)O4; lithium aluminate (LiAl) with the stoichiometric formula Li1−*<sup>x</sup>*(Al(1−*<sup>x</sup>*)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 Mnrich 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.

**Figure 5.** Electron micrograph (BSE(Z)) of the solidified melt. Light grey grain: spinel; dark gray sections, and dendrites: LiAl, surrounded by Ca-alumosilicate (GCAS, light grey sections); amorphous phases (APh): amorphous grain with unusual elements (K, Ba, Ti): contamination; black: pores or preparation damage. The chemical analysis of the points marked in red (P294–P383) is presented in Table 3.


**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.

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 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, Mn0.5AlO2 (1/2 galaxite spinel), and Mn0.5MnO2 (1/2 hausmannite) fitting the three spinel components to the elemental amounts (Table 4). The Li-concentration was subsequently obtained from the calculated amount of LiMnO2. The general formula obtained from the fittings is (Li(2*x*)Mn2+(1−*<sup>x</sup>*))1+*x*(Al(2−*<sup>z</sup>*),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.

**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(2*x*)Mn2+(1−*<sup>x</sup>*))1+*x*(Al(2−*<sup>z</sup>*),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.


#### *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.

**Figure 6.** Spectra of Mn samples of different oxidation states. The spectra are compared to the spectrum of the mock-up slag (MUS). A shift of the edge accompanies the increase in the oxidation state which is indicated by the arrow. The mean oxidation states of Mn in the compounds are as follows: MnO: +2; galaxite: +2, hausmannite: +2.67; braunite: +2.85; bixbyite: +3; and MnO2: +4.

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 Figures 7 and 8 were obtained. These mixtures present a lower Alcontent than galaxite, but still have a spinel structure. The results obtained from these linear combinations can be seen in Figures 7 and 8. A linear combination of hausmannite with bixbyite was also calculated. These results are shown in Appendix D.

**Figure 7.** Linear combination of hausmannite (H) and galaxite (G). At the edge jump, the MUS spectrum fits the spectrum of hausmannite, whereas after the jump, the spectrum is more similar to galaxite.

The linear combinations (Figures 7 and 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 Figures 9 and 10. Figure 10 shows a close-up of the edge-jump.

The spectra displayed in Figures 9 and 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.

**Figure 8.** Linear combination of bixbyite (B) and galaxite (G). The spectrum of the MUS is similar to the linear combination. At the edge jump, the MUS spectrum fits the middle of the linear combination, whereas in the region after the edge, the spectrum is more similar to that of galaxite.

**Figure 9.** Experimentally derived X-ray absorption near edge structure analysis (XANES) spectra of a 50 wt.% mixture of galaxite and bixbyite, respective of galaxite and hausmannite compared to the spectrum obtained from the MUS V-3. The pre-edge peak, edge jump and course of the spectrum are very similar for all three spectra.

**Figure 10.** Close-up to the edge jump from Figure 9.

#### **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), Li1−*<sup>x</sup>*Mn2O4, 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 Mn0.5AlO2 ( <sup>1</sup> <sup>2</sup> galaxite), LiMnO2 and Mn0.5MnO2 ( <sup>1</sup> <sup>2</sup> hausmannite) a general formula (Li(2*x*)Mn2+(1−*<sup>x</sup>*))1+*x*(Al(2−*<sup>z</sup>*),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 Mn0.5MnO2-dominated region can be observed. The increase and decrease of the individual species over the point scans are shown in Figure 11.

**Figure 11.** Fractions of the virtual compounds Mn0.5AlO2 ( <sup>1</sup> <sup>2</sup> galaxite, Gal), LiMnO2 (Li-Mn), and Mn0.5MnO2 ( <sup>1</sup> <sup>2</sup> hausmannite, Hsm) in the grain presented in Figure 5.

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 galaxitedominated 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 spinellike) 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(2*x*)Mn2+(1−*<sup>x</sup>*))1+*x*(Al(2−*<sup>z</sup>*),Mn3+*z*)O4 of the solid solution was determined assuming a combination of the virtual components Mn0.5AlO2, LiMnO2 and Mn0.5MnO2. 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*<sup>x</sup>* 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.

**Author Contributions:** A.W. conceived the paper. A.W., U.E.A.F. and T.S. conducted the literature review. All melting experiments were designed and performed by H.Q. and D.G. The chemical bulk analysis was executed by the analysis laboratory of the Institute of Mineral and Waste Processing. The phase analysis (PXRD) and the mineralogical investigation (EPMA) were conducted by T.S. The speciation analysis with XANES and interpretation of the spectra was conducted by A.W. and U.E.A.F. Interpretation, discussion and conceptualization were conducted by all authors. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Clausthal University of Technology in the course of a joint research project, "Engineering and Processing of Artificial Minerals for an Advanced Circular Economy Approach for Finely Dispersed Critical Elements" (EnAM).

**Acknowledgments:** We acknowledge support by Open Access Publishing Fund of Clausthal University of Technology. We thank Jörg Wittrock from the Institute of Inorganic and Analytical Chemistry for the idea and implementation of the galaxite synthesis. We thank Joanna Kolny-Olesiak for discussions and proofreading.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study, in the collection, analyses, or interpretation of data, in the writing of the manuscript, or in the decision to publish the results.

#### **Appendix A**

The conversion from *θ*-angle to energy space is done by following Bragg's law, with the order of diffraction (*n*), Planck constant (*h*), speed of light (*c*), interplanar distance (*d*).

$$E = \frac{n \cdot h \cdot c}{2 \cdot d \cdot \sin(\theta)}\tag{A1}$$

For a cubic system, *<sup>d</sup>* is defined as: *<sup>d</sup>* <sup>=</sup> *<sup>a</sup>*<sup>0</sup> <sup>√</sup>*h*<sup>2</sup> <sup>+</sup> *<sup>k</sup>*<sup>2</sup> <sup>+</sup> *<sup>l</sup>*<sup>2</sup> , with lattice spacing (*a*0) and Miller indices *(h*, *<sup>k</sup>*, *<sup>l</sup>*). Therefore, the term *<sup>n</sup>*·*h*·*<sup>c</sup>* <sup>2</sup>*<sup>d</sup>* is dependent on the chosen crystal and the order of diffraction.

#### **Appendix B**

**Figure A1.** Spectra of Mn samples of different oxidation states. The spectra are compared to the measurement of the MUS. Oxidation states ranging for a better overview in contrast to Figure 6 from +2 to +3.

**Appendix C**

**Figure A2.** Close-up of Figure 4 showing the lamellae region.

**Figure A3.** Linear combination of hausmannite (H) and bixbyite (B).

#### **References**


### *Article* **The COOL-Process—A Selective Approach for Recycling Lithium Batteries**

**Sandra Pavón, Doreen Kaiser, Robert Mende and Martin Bertau \***

Institute of Chemical Technology, TU Bergakademie Freiberg, Leipziger Straße 29, 09599 Freiberg, Germany; sandra.pavon-regana@chemie.tu-freiberg.de (S.P.); doreen.kaiser@chemie.tu-freiberg.de (D.K.); robert.mende@chemie.tu-freiberg.de (R.M.)

**\*** Correspondence: martin.bertau@chemie.tu-freiberg.de; Tel.: +49-3731-392384

**Abstract:** The global market of lithium-ion batteries (LIB) has been growing in recent years, mainly owed to electromobility. The global LIB market is forecasted to amount to \$129.3 billion in 2027. Considering the global reserves needed to produce these batteries and their limited lifetime, efficient recycling processes for secondary sources are mandatory. A selective process for Li recycling from LIB black mass is described. Depending on the process parameters Li was recovered almost quantitatively by the COOL-Process making use of the selective leaching properties of supercritical CO2/water. Optimization of this direct carbonization process was carried out by a design of experiments (DOE) using a 33 Box-Behnken design. Optimal reaction conditions were 230 ◦C, 4 h, and a water:black mass ratio of 90 mL/g, yielding 98.6 ± 0.19 wt.% Li. Almost quantitative yield (99.05 ± 0.64 wt.%), yet at the expense of higher energy consumption, was obtained with 230 ◦C, 4 h, and a water:black mass ratio of 120 mL/g. Mainly Li and Al were mobilized, which allows for selectively precipitating Li2CO3 in battery grade-quality (>99.8 wt.%) without the need for further refining. Valuable metals, such as Co, Cu, Fe, Ni, and Mn, remained in the solid residue (97.7 wt.%), from where they are recovered by established processes. Housing materials were separated mechanically, thus recycling LIB without residues. This holistic zero waste-approach allows for recovering the critical raw material Li from both primary and secondary sources.

**Keywords:** lithium recycling; circular economy; lithium batteries; supercritical CO2; black mass

#### **1. Introduction**

Since the market launch in 1991, the global market of lithium-ion batteries (LIBs) has been growing steadily. The global LIB market was valued at \$36.7 billion in 2019 and is expected to reach \$129.3 billion by 2027 [1]. One reason for this strong growth is the rising market for electric mobility. In 2018, 5.12 million electric passenger cars were registered worldwide, which corresponds to an increase of 63% compared to the previous year [2]. Furthermore, rechargeable LIBs are used extensively in the growing market of cableless electronic devices and applied in electric tools and grid storage applications [3]. Since the global reserves required to produce LIBs, as well as the lifetime of LIBs, are limited, efficient recycling approaches are necessary. The chemistry and technology of LIBs are still in development, resulting in a wide variety of different battery types, which in turn makes recycling more sophisticated. Battery recycling is also supported by the directive 2006/66/EC of the European Union, which requires a recycling rate of spent batteries of at least 50 wt.% of whole spent battery [4].

Despite structural diversity, the basic structure of all LIBs is mostly the same [5]. Usually, the cathode is an aluminum foil with an intercalated Li compound, and the anode a copper foil with a graphite coating. The anode and cathode compartments are separated by a porous polyolefin and the electrolyte is a mixture of an organic solvent and a lithium salt. These cells are enclosed by a sealed container made of aluminum, steel, special plastics,

**Citation:** Pavón, S.; Kaiser, D.; Mende, R.; Bertau, M. The COOL-Process—A Selective Approach for Recycling Lithium Batteries. *Metals* **2021**, *11*, 259. https://doi.org/10.3390/met11020259

Academic Editor: Bernd Friedrich Received: 15 December 2020 Accepted: 30 January 2021 Published: 3 February 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

or highly refined aluminum composite foils [6]. Depending on the used cathode materials current commercial LIBs can be categorized into five types [7]:


To simplify the battery recycling the process should be independent of the type of the spent LIBs and should be applicable for mixtures of different LIBs.

Most of the already developed recycling processes are pyrometallurgical and/or hydrometallurgical approaches. Pyrometallurgical processes are associated with high energy consumption, high capital costs, and potential hazardous gas emission, as well as complex extraction procedures [8,9]. Furthermore, the selective recovery of lithium is very difficult [8]. Moreover, recycling of plastics and electrolyte is not possible. As both components make up 40–50 wt.% of the spent battery, it is difficult to meet the required recycling rate of 50 wt.% [9]. Hydrometallurgical approaches allow for recycling lithium, as well as cobalt and nickel with high purity [7]. Leaching procedures with inorganic or organic acids followed by precipitation and/or solvent extraction obtain the desired products with high recycling efficiencies [10–12]. However, the high recycling rates can only be achieved by using high quantities of acid which in turn not only produce high amounts of wastewater [7]. Already the costs of the chemicals for acidic digestion and subsequent neutralization exceed the intrinsic metal value by far. Furthermore, the low leaching selectivity, especially in the case of inorganic acids, necessitates extensive purification steps, which render the entire process complex and costly. A promising alternative is the COOL-Process (CO2-leching), the core step of which is leaching with supercritical CO2 (*sc* CO2).

Supercritical fluids are interesting alternatives to conventional solvents for metal extraction. There are more than 100 plants worldwide that extract using supercritical solvents, thus creating a broad field of application for these processes [13]. Probably the best-known process is the decaffeination of coffee [14]. A study by Rentsch et al. was able to show that the higher investment costs compared to conventional processes are already compensated by low operating costs after about two years. The low operating costs are due to a low chemical requirement and less complex wastewater treatment [15].

In the field of battery recycling, *sc* CO2 currently plays only a minor role, but this is incomprehensible. Only for the recycling of the electrolyte of the LIBs have several studies [14,16–19] been published. The application of *sc* CO2 for the extraction of metals has only been published in one paper on cobalt extraction [20]. The recovery of the electrolyte is a challenging task, especially regarding the different compositions of the LIBs. Several studies have shown that extraction with *sc* CO2 is an efficient way to recycle the electrolyte [14,16–19], but this requires LIBs with the same composition, which is associated with a high sorting effort and therefore does not appear economical. Supercritical CO2 has also been employed for metal extraction from several materials, like ores, resins, and foils. For instance, Bertuol et al. developed a process that allows the recovery of cobalt from LIBs using *sc* CO2 and H2O2 (4% *v*/*v*) as co-solvent. This process allows the extraction of more than 95 wt.% cobalt in a very short time (5 min) [20]. Other metals, such as nickel, manganese, and lithium, are not considered in this study. Research on the recovery of lithium from LIB by means of *sc* CO2 is not published yet.

Originally, the COOL-Process was developed for the production of Li2CO3 from lithium containing ores, like zinnwaldite and spodumene (Figure 1) [21].

**Figure 1.** Flowsheet for the production of Li2CO3 from Lepidolite, Petalite, Spodumene, and Zinnwaldite minerals by direct carbonation [21].

Considering that this direct-carbonation process promises a selective leaching of Li with subsequent precipitation and separation of Li2CO3 without the addition of further chemicals [21,22], the COOL-Process has been applied to recover Li from black mass in the current work as depicted (Figure 2).

**Figure 2.** Recycling process scheme of lithium-ion batteries (LIB) black mass.

#### **2. Materials and Methods**

#### *2.1. LIB Black Mass Pre-Treatment and Characterization*

LIB black mass sample, type battery Li-NMC, was kindly supplied by the Institute of Mechanical Process Engineering and Mineral Processing from TU Bergakademie Freiberg and pre-treated before carrying out the optimization process by multi-stage crushing using a planetary ball mill PM 100 (Retsch GmbH, Haan, Germany) and subsequent milling by a vibrating cup mill (AS 200, Retsch GmbH, Haan, Germany) for grinding the black mass sample to a particle size of <63 μm (d90: 61.18 μm).

The elemental composition of the LIB black mass was determined by atomic emission spectrometry with inductively coupled plasma (ICP-OES, Optima 4300 DV, Perkin Elmer, Waltham, MA, USA) and atomic absorption spectroscopy (AAS, ContrAA 700, Analytik Jena, Jena, Germany). The LIB black mass sample was treated with aqua regia in a liquid:solid ratio (L:S) of 100 at 180 ± 2 ◦C for 30 min using a Microwave MARS 6 (CEM Corporation, London, UK). The procedure was repeated three times.

Carbon measurement was carried out with the vario EL MICRO cube system made by Elementar Analysensysteme GmbH, Langenselbold, Germany, based on a combustion method.

Thermogravimetric analysis (TGA) and differential thermal analysis (DTA) were conducted by using TGA/DSC (Differential scanning calorimetry) 1 with a DSC sensor and mass-flow controller GC 200 (Mettler Toledo, Gießen, Germany) to examine the thermal behavior of the LIB black mass. TGA/DTA was carried out placing 22.36 mg of the black mass in a 150 μL alumina crucible heated from 25 up to 1000 ◦C with a heating rate of 10 K/min under pure oxygen or nitrogen flow of 40 mL/min.

#### *2.2. Optimization*

The aim was to determine the reaction conditions at which the highest yield of the target value can be obtained. However, processes reported in the literature are usually conducted using the one-factor-at-a-time method. The influence of different factors is evaluated by varying one after another, keeping the other ones constant. This method often fails to determine the global optimum because the correlation between different factors is not in consideration. Hence, the obtained optimum is a local instead of a global one, and the process efficiency from an economic and environmental point of view is not properly evaluated. Furthermore, the optimization cannot be considered accurate, because the influence of some factors (binary correlations) on the target yield is often significant, yet not determined. Therefore, the current work employs a statistical experimental design by considering both binary correlations and squared effect in order to determine the global optimum.

#### 2.2.1. 3<sup>3</sup> Box-Behnken Design

To optimize Li recovery from the LIB black mass sample, a 33 Box-Behnken design was used to determine the global optimum by consideration of all the factor combinations. This design of experiment (DOE) requires tests on every half of the edges and in the center, which was conducted threefold to determine the experimental error. The factors investigated were: temperature *T* [ ◦C], residence time *t* [h], and water:black mass ratio *L:S ratio* [mL/g] in a range comprising three levels (Table 1).

**Table 1.** Factors and levels in the 33 Box-Behnken experimental design.


Statgraphics v.18 (Statpoint Technologies Inc., Warrenton, VA, USA) was used as the evaluation statistical software to determine the global optimum, as well as the model equation which describes how Liyield depends on each nine effects (linear, squared, and binary correlations). The model equation was obtained using Equation (1) via multi-linear regression.

$$\mathbf{y} = \mathbf{b}\mathbf{0} + \sum\_{i=1}^{N} \mathbf{b}\_{i}\mathbf{x}\_{i} + \sum\_{1 \le i \le j}^{N} \mathbf{b}\_{i}\mathbf{x}\_{i}\mathbf{x}\_{j} + \sum\_{i=1}^{N} \mathbf{b}\_{i}\mathbf{x}\_{i}^{\ast}\tag{1}$$

where:

y: Target value: Liyield [wt.%]; xi: Factors: T [◦C], t [h], L:S ratio [mL/g]; N: Number of factors (3); b0: Ordinate section; and bi, bij, bii: Regression parameters of linear, squared and cross effects.

#### 2.2.2. Experimental Procedure

Digestion experiments of the LIB black mass, using the conditions in random order, were performed under elevated pressure (100 bar) using the autoclave Hastelloy C4 (Berghof Products + Instruments GmbH, Eningen unter Achalm, Germany) and are depicted in Figure 3.

**Figure 3.** Hastelloy C4 autoclave used to carry out the COOL-Process.

Digestion experiments by the COOL-process, which uses CO2 as a reagent, were carried out adding the LIB black mass sample (d50 = 9.1 μm) in the BR-300 autoclave to a volume of distilled water of 400 mL in accordance with the operating conditions depicted in Table 2. The suspension was heated to a range of temperature of 150–230 ◦C at a heating rate of 5 K/min and 500 rpm. CO2 was added and a pressure of 100 bar was set after reaching the target temperature. The digestion time varied between 2–4 h. Afterward, the reaction mixture was cooled down to T < 30 ◦C under pressure and subsequently decompressed to normal pressure. The suspension was filled up with distilled water to 1 L and the residue was separated by vacuum filtration using an ash-free paper filter MN 640 dd (Macherey-Nagel, Düren, Germany). The leachates were analyzed by ICP-OES to determine the Al, Cu, Co, Fe, Ni, and Mn content and AAS for the Li content.

**Table 2.** Composition of LIB black mass in wt.% analyzed by ICP-OES, AAS, and combustion.


#### *2.3. Li2CO3 Precipitation*

The leachate (100 mL) obtained at T = 230 ◦C, t = 3 h, and L:S ratio = 30 mL/g was heated T = 100 ◦C. Li2CO3 precipitation was complete at V = 2.5 mL. The solid product was separated by filtration and washed with deionized water (5 mL). The liquid fractions were combined and recirculated for the next run in order to not loose residual Li. Product purity was determined as 99.8% by ICP-OES and AAS after dissolving with HNO2 1 vol.%.

#### **3. Results and Discussion**

#### *3.1. LIB Black Mass Characterization*

After discharging the LIB, a black mass was obtained by mechanical treatment involving crushing and magnetic separation. Housing material, polyethylene, aluminum, and copper were separated as initial products. The resulting black mass is a powder consisting of anode and cathode material, coating material, electrode foils, and small parts of aluminum, copper, and polyethylene from the separator foil. During mechanical comminution, the release of highly volatile compounds of the electrolyte (dimethyl carbonate, diethyl carbonate) also occurs. Table 2 shows the mean value of black mass analysis and the standard deviation of the sample set (Std. Dev.).

The Li content was 3.18 wt.% and the main metals were Co (2.37 wt.%), Cu (2.21 wt.%), Mn (23.89%), and Ni (8.31 wt.%). The high Mn content compared to Co and Ni is noticeable, from which was evident that the LIB was a (LiNixMnyCozO2, x + y + z = 1) type battery (Li-NMC). Iron was introduced into the sample through the mechanical processing of the LIB. However, due to the low concentration (0.29 wt.%), no negative impact on the process was to be expected.

To be able to exclude the possibility of toxic fluorine compounds formed during the digestion, a thermogravimetric analysis (TGA) was carried out. TGA under nitrogen and oxygen showed a mass loss of 20.42 wt.% and 33.34 wt.% (Figure 4). The mean value of both measurements (26.88 wt.%) was in accordance with the carbon content measured by the elemental analysis (26.04 wt.%). Under oxygen atmosphere differential thermal analysis (DTA) showed, as expected, an exothermic peak in the range between 450 and 750 ◦C, which correlates with the combustion of the contained carbon. However, heat formation occurring here is only of minor importance for the process, which works at *T* ≤ 230 ◦C. Only at *T* > 800 ◦C a slight release of fluorine compounds was observed. Therefore, it is not expected that volatile fluorine compounds will form during digestion, which means that no additional safety measures or special materials are required to operate the COOL-Process safely.

**Figure 4.** Thermogravimetric (TG) curve (black) and differential thermal analysis (DTA) curve (blue) of black mass heating under O2 (dashed line) and N2 (continuous line) atmosphere from 25 to 1000 ◦C at 10 K/min.

#### *3.2. Optimization*

#### 3.2.1. Significant Influences on Lithium Yield

The Li leaching efficiency of each experiment obtained experimentally following the DOE was determined and listed in Table 3 as Li yield. To evaluate which of the nine effects: linear (A, B, C), squared (AA, BB, CC), and binary correlation (AB, BC, AC) contribute significantly to Li yield, an analysis of Variance called ANOVA was conducted. The results are depicted in Table 4. Terms that are insignificant for the target value were removed by the stepwise method.

**Table 3.** Li yield in each experiment obtained experimentally following the 33 Box-Behnken experimental design (*A*: *T* [ ◦C], *B*: *t* [h], *C*: *L:S ratio* [mL/g]). The experiments shaded in grey correspond to the replicated central point.


**Table 4.** ANOVA results for Li yield from the 33 Box-Behnken design with three central points.


\* Degrees of freedom.

According to the Pareto diagram depicted in Figure 5, the temperature in terms of a linear effect takes the most pronounced effect on Li yield. The water:black mass ratio (*L:S ratio*) showed the second highest effect, whereas the influence of residence time was rather poor. Consequently, leaching is almost fully completed after 2 h, as Li yield was affected only minorly within the residence time range studied (Table 3). This observation indicates that supercritical CO2 is an efficient leaching agent for Li. The fact that residence time has only a slight impact on leaching black mass for Li mobilization has been observed in other studies, too: For inorganic (e.g., HCl, H2SO4, H2O2), as well as organic (e.g., oxalic acid, tartaric acid, citric acid) leaching agents, residence times between 30 and 240 min were reported [10]. Hence, Li is only weakly bound in the matrix of the black mass and it can therefore be easily extracted. This is also supported by the excellent extraction properties of supercritical (*sc*) CO2. In previous studies, it was shown that the optimal residence time for Li mobilization from zinnwaldite through leaching with *sc*-CO2/H2O is 3 h [21], whereas leaching with HCl takes 7 h [22].

**Figure 5.** Pareto diagram with the significant effects on Li yield.

Furthermore, the AC correlation contributes to increasing the Li yield, too. It represents an interaction between temperature and water:black mass ratio. After Urba ´nska et al. had observed the same trend when leaching black mass with H2SO4 and H2O2 [23], this interaction was not unexpected.

The three squared correlations (AA, BB, and CC), as well as AB and BC as binary interactions were removed from Equation (1) by the stepwise method because of their insignificant effect to the Li yield optimization.

#### 3.2.2. Model Equation and Optimum

In accordance with the experimental obtained results from the DOE, a mathematical model equation was determined considering all significant effects on Li yield. This equation allows for predicting the Li yield at any desired point within the investigated factors levels range. Equation (2) predicts that Li yield reaches its all-time maximum with 98.8 wt.% using the following reaction conditions: *T* = 230 ◦C, *t* = 4 h and *L:S ratio* = 90 mL/g.

$$Li\_{yield}(\text{wt.\%}) = 19.6125 + 0.178 \cdot A + 1.9275 \cdot B - 0.43977 \cdot C + 0.00339 \cdot A \cdot C \tag{2}$$

where:

*A*: Temperature (◦C);

*B*: Residence time (h); and

*C*: L:S ratio (mLwater/gblack mass).

To validate the mathematical model, a twofold experiment involving these optimized parameters was carried out. Both experiments provided a Li yield of 94.5 ± 0.33 wt.%. With a difference of <5 wt.%, one may consider the calculated model by employing a 3<sup>3</sup> Box-Behnken design in accordance with the experimental data. However, bearing in mind that only three of the nine effects studied took effect on Li yield, together with the optimum being in a corner of the DOE, an evaluation of the statistical design will be carried out in follow-up work. For instance, by employing a full factorial design, an improvement of the statistical experimental design can be achieved, thus obtaining a better mathematical model that describes how Li yield depends on the three chosen factors: temperature, residence time, and water:black mass ratio.

The model in terms of surface response is shown in Figure 6, where the bold points correspond to the experimental data and the star to the two replicates which were carried out using the optimal reaction conditions (*T* = 230 ◦C, *t* = 4 h and *L:S ratio* = 90 mL/g).

**Figure 6.** Li yield determined by the mathematical model equation varying the temperature and L:S ratio, maintaining the residence time constant at 4 h. Bold points correspond to the experimental data and the star point to the optimum obtained experimentally.

The Li yield (94.5 ± 0.33 wt.%) obtained under the optimal reaction conditions is roughly in line with several studies using inorganic and organic acids. For instance, Takacova et al. achieved a quantitative Li mobilization from the black mass with 2 M HCl (60–80 ◦C, 90 min, L:S ratio = 50) with simultaneous quantitative cobalt mobilization [24]. Similar results were described by Urba ´nska et al. using 1.5 M H2SO4 and 30% H2O2 (90 ◦C, 120 min, L:S ratio = 10). In their study up to 99.91 wt.% Li was leached and 87.85 wt.% cobalt and 91.46 wt.% nickel were co-extracted [23]. A high Li (<90 wt.%) and <30 wt.% Mn yield were obtained by Li et al. with 1 M oxalic acid (95 ◦C, 12 h, L:S ratio = 10) [25]. In other words, Co, Ni, and Mn were co-mobilized. All published results have in common, though, that their Li selectivity is poor, thus requiring additional purification and consumption of chemicals, not to speak of process costs. In contrast, the current work exhibits not only a high Li selectivity but also a high degree of Li mobilization. Under optimal reaction conditions, only Al was co-extracted (Table 5), but the presence of this metal has no influence on the further process of the Li2CO3 precipitation. The reason is for the insensitivity of the COOL-Process towards Al is the inability of Al to form neither carbonates nor hydrogen carbonates under this condition. Precipitation of Al salts with CO2 only occurs from pH 9 and not in the acetic range [26]. With CO2/water being a weak acid only, any other acidic leachate reagent will consequently co-mobilize Co, Ni, and/or Mn, the interaction of which with Al3+ and Fe3+ inevitably requires tedious and complex separation of these metals prior to Li2CO3 precipitation. The latter step is gaining further complexity through the mutually interacting chemistry of these metal cations with hydroxide, thus severely taking effect on the process economy of these approaches. A further advantage of the COOL-Process is the effect that, in contrast to other carbonates, the solubility of Li2CO3 decreases with increasing temperature. For this reason, the digestion solution is heated to 90 to 95 ◦C and the target product is precipitated in the process. The precipitation behavior of Li2CO3 in such digestion solutions has been extensively investigated in previous studies, so it was omitted here [15,21,22,27].


**Table 5.** Co-mobilization of selected elements at 230 ◦C, 4 h, and L:S ratio = 90 mL/g.

Direct carbonation (COOL-Process) of black mass has the advantage of leaving other valuable metals, such as Co, Mn, and Ni, in the leaching residue from where they can be recycled with ease according to established techniques. There exist pyrometallurgical processes for this purpose, so that the COOL-Process can be understood in terms of an enabling technology, which gives way to isolate lithium prior to known pyrometallurgy in a preliminary stage. The CO2 released during the pyrometallurgical recovery of Co, Ni, and Mn can, in turn, be used for carbonization, this way contributing to both a zero-waste approach and circular economy.

Another advantage of the COOL-Process is its efficiency in terms of Li recovery regardless of the composition of the raw material. Particularly in the field of LIB recycling, a broad and robust feedstock variability is a prerequisite to operate the process economically, which in turn is mandatory in terms of successfully establishing a circular economy. Each battery manufacturer uses different compositions, for which reason there are a plethora of different battery types on the market. Most processes for recycling LIBs are specialized in certain compositions, which entails complex sorting processes. This is usually only possible by hand, which renders these processes highly cost intensive. The flammability of damaged LIB is susceptible to danger, what is an issue when hand-sorting. In the COOL-Process, LIB can be processed regardless of their composition. Moreover, previous studies have already shown that this process is suitable for extracting Li from ores, like zinnwaldite, too. The optimum reaction conditions determined in these studies were 230 ◦C, 3 h, and a L:S ratio of 30 and are thus comparable to the conditions determined in the current work [21]. Therefore, it can be concluded that the COOL-Process probably allows for recovering Li from both primary and secondary sources. It is a textbook example of circular resources chemistry, which comprises origin-independent processes for the production of chemical raw materials that do not differentiate between primary and secondary raw materials [28].

However, the maximum is placed in a corner, which raises the question of whether another factor (e.g., pressure) needs to be explored to ensure that the optimization covers all effective factors. From the viewpoint of process engineering, temperature increase appears as the factor of choice to check for higher Li mobilization. This is not possible, though, which is one of the limitations of the materials in contemporary LIB. Conventional sealing material (polytetrafluoroethylene, PTFE) is only stable up to 230 ◦C, so special materials, such as perfluoro rubber (FFKM), would be necessary. Particularly, on an industrial scale, these special sealing materials, together with the energy input required to reach temperatures >230 ◦C, are associated with considerable additional costs. Since a high Li mobilization was already achieved at 230 ◦C, only a small yield increase can be expected from a further temperature increase, which, however, is not justified in terms of additional energy and raw material demand which is in sharp contrast to the plus of Li to be expected. For these reasons, this factor remained unaltered and the maximum level was set to 230 ◦C. Since the second highest effect on the Li yield was provided by water:black mass ratio, the highest factor level was increased to 120 mL/g to evaluate how much the target value can be increased (Figure 7). 98.6 ± 0.19 wt.% Li was recovered by the COOL-Process at 230 ◦C, 4 h, and 120 mL/g. Considering the increase of 4.6 wt.% on the target value using 120 instead of 90 mL/g, the DOE could be improved by redefining the factor levels considering this enhancement on the Li yield. However, the small increase in yield represents a 22 wt.% reduced Li concentration in the digestion solution, which in

turn will necessitate a higher energy input for its concentration prior to precipitating the target product Li2CO3. Again, this additional energy costs, in combination with resulting CO2 emissions for energy generation, may hardly be compensated for the rather small plus in lithium yield.

**Figure 7.** Effect of L:S ratio on the Li yield carrying out the COOL-Process at 230 ◦C for 2 or 4 h as residence time.

With this in mind, the 4.6 wt.% higher Li yield obtained from varying the water:black mass ratio cannot compensate for the lower energy efficiency. This latter issue can be equalized by reducing residence time *t*. Although the theoretical optimum is placed in a corner of the 33 box, the obtained information is sufficient to recognize the potential that lies in reducing *t* without investing in further optimization work. If *sc*-CO2-leaching is done at 230 ◦C, 120 mL/g, and 2 h instead of 4 h, 99.05 ± 0.64 wt.% Li was recovered. As can be seen in Figure 7, the differences between the Li yields (~1%) when the COOL-Process was carried out at 230 ◦C for 2 or 4 h are not significant when using 120 mL/g. Hence, almost quantitative lithium recovery was reached by simply increasing the L:S ratio from 90 to 120 mL/g and conducting the leaching for 2 h. As pointed out before, the economic impact on the entire Li recycling process is an essential factor to be considered. Process performance depends not only on maximizing target values but also on economic efficiency.

It appears evident from these considerations that in general quantitative metal recovery from whatever feedstock may technologically be feasible, yet the bill is paid in terms of higher energy consumption, higher CO2 emissions, and lacking economy. Applied to the circular economy, where the intrinsic metal value of the secondary raw material should exceed process costs, it is obvious that real-world processes always will constitute a compromise between what is desirable, what is feasible, and what is realizable. A way out of this situation is integrated processes, where the (secondary) raw material is converted into marketable products to the most possible extent. This is given here, since Co, Ni, and Mn, as well as housing material, are products, too, and CO2 is re-circulated. A follow-up economical assessment will be conducted to provide the essential information to which extent additional efforts towards quantitative recovery are justified.

#### *3.3. Li2CO3 as a Final Product*

Black mass leaching under the conditions identified optimal in the optimisation study (Section 3.2) to reach the highest Li concentration (T = 230 ◦C, t = 3 h, L:S ratio = 30 mL/g) yielded an aqueous solution of LiHCO3. The final product, Li2CO3, was obtained making use of its solubility anomaly. The carbonate's solubility in water is 13.3 g/L at T = 20 ◦C, while it is 7.2 g/L at T = 100 ◦C. Heating the solution to T = 100 ◦C not only decomposes LiHCO3 to give Li2CO3, it also serves to reduce the solution volume in order to obtain best possible precipitation results. Filtration of the solid product gave pure X which was washed with deionised water with twice the amount of the volume of the residual precipitation

solution (Figure 8). After drying, Li2CO3 was dissolved in HNO3 1 vol.%. Product purity was 99.8% as determined by ICP-OES and AAS.

**Figure 8.** Li2CO3 as a product after precipitation of LiHCO3 by heating up at 100 ◦C.

The concentration of all other cations, such as Al, Mn, Fe, Co, Ni, and Cu, in sum accounted to <0.17 wt.%. With Li2CO3 purity >99.8 wt.% it was shown that the COOL-Process is capable of producing battery grade Li2CO3 as crude product, which needs no further purification.

The remaining liquid fractions from product precipitation and washing were combined and recirculated. They serve as aqueous phase for the next run. Although this way no lithium is lost, Li2CO3 precipitation remains an issue, since precipitation efficiency in our experiments ranged widely between 43 and 85 wt.%. Optimizing product precipitation is therefore matter of follow-up studies. The same applies for Al, which under the given conditions is not susceptible for precipitation from carbonatic solutions. Upon recirculating the aqueous solutions, Al will accumulate and may interfere with the process. Exploratory experiments showed that Al can be eliminated as oxalate. If, however, aluminium oxalate precipitation interferes with Li leaching beyond what is tolerable, purging the solution is an option.

#### *3.4. Industrial Application Feasibility*

The COOL-Process has been successfully tested on a lab-scale and the high Li yield obtained using LIB as secondary sources demonstrates the efficiency of the direct-carbonation process.

The purpose of realizing holistic research approaches with practical relevance that can be carried out by using different raw materials is challenging. According to the obtained results, it can be affirmed that the goal has been reached because COOL-Process has been used for recovering Li from primary [21], as well as secondary raw material. This success essentially contributes to safeguarding the raw material base of the European industry for LIB production.

Furthermore, considering the lack of Li recycling from secondary sources due to the uneconomic methods, this new approach offers selective leaching where Li can be recovered and subsequently precipitated to obtain Li2CO3. Present pyro- and hydrometallurgical processes developed for LIB recycling are focused on the recovery from other valuable metals, such as Co, Mn, and Ni, among others. Li remains in the solid residue and its recovery is cost intensive (if feasible at all). The current alternative allows for efficiently recovering Li and offers the possibility of recycling other metals, such as Co, Cu, Mn, and Ni, since they are not affected by the COOL-Process. Their selective separation can be carried out using different techniques, such as solvent extraction, membrane technologies, and precipitation [10,11,29,30]. Therefore, the developed process shown in Figure 2 contributes to a zero-waste concept, as well as the development of sustainable recycling processes.

#### **4. Conclusions**

The current work shows a selective process to mobilize Li from LIB black mass by leaching with supercritical CO2. Process parameter optimization was done by using a 3<sup>3</sup> Box-Behnken design as DOE. The maximum Li yield of 94.5 wt.% was reached at 230 ◦C, 4 h, and a water:black mass ratio of 90 mL/g. With a water:black mass ratio of 120 mL/g Li yield was almost quantitative (99.05 ± 0.64 wt.%), yet requiring higher energy input. In contrast to all other studies, only Li and Al were mobilized, which allows for selectively precipitating Li2CO3 in high purity without much effort, yielding battery grade-quality (>99.5 wt.%) as the crude product. There is no further refining required. Other valuable metals, such as Co, Cu, Ni, and Mn, remained in the solid residue, which can be separated selectively and recovered by established processes. The CO2 released in these processes can be fed back to the COOL-Process. Therefore, this holistic approach for LIB recycling comes very close to the goals of zero-waste. Last but not least, this approach allows for simultaneously treating primary and secondary raw materials for Li recycling.

**Author Contributions:** Conceptualization, S.P., D.K., and M.B.; methodology, R.M.; validation, R.M.; formal analysis, R.M.; investigation, R.M., S.P. and D.K.; resources, D.K. and M.B.; writing—original draft preparation, S.P., D.K., and R.M.; writing—review and editing, M.B., D.K., S.P.; supervision, S.P. and D.K.; project administration, M.B. and D.K.; funding acquisition, M.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** Financial support by the German Federal Ministry of Education and Research (Grant nr. 033RC020A) is gratefully acknowledged.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** It was followed according to MDPI Research Data Policies.

**Acknowledgments:** Further thanks are owed to Sebastian Hippmann, Andrea Schneider and Jan Walter for conducting TGA/DTA, ICP-OES and AAS analyses and Resource technology & Metal processing Freiberg GmbH (RMF) for providing LIB black mass sample.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**

