**1. Introduction**

Since 1979, when Goodenough et al. finally tested LiCoO2 (short: LCO) as a cathode material, the development and commercialization of electrochemical energy storage based on the lithium-ion technology has been steadily pushed forward [1,2]. Lithium-ion-batteries (LIBs) basically consist of the same components such as anode, cathode, separator or electrolyte as can be found in other battery technologies. This basic principle has not changed since 1979 and therefore also applies to modern LIBs. However, the initially used LCO cathode material is now just one option on a long list of alternatives like NCA (LiNixCoyAlzO2), NMC (LiNixMnyCozO2) or LFP (LiFePO4) materials [3]. The variety of cathode materials is not only based on the fields of possible applications that reach from mobile electronics to e-mobility or stationary storages and their respective demand for performance (energy and power density) or safety aspects, but also on factors like raw material prices, supply risks or social and ecological sustainability. Concerning the development of the LIB market numerous publications can be found. Especially the electric automotive sector will benefit from decreasing costs made possible by mass production and optimized cell chemistry. Berckmans et al. [4] states that by 2030, the cut of fully electric or hybrid vehicles will rise to 25% of the total vehicles sold. In view of the high amount of valuable metals that are contained in LIBs, especially in their cathode materials, and the predicted market demand [5], an efficient recycling process in order to recover the mentioned valuable metals is absolutely necessary.

In general, the recycling of LIBs can be divided into three processing steps, namely pretreatment, metal extraction and metal refining. The recycling chain of LIBs usually starts

**Citation:** Windisch-Kern, S.; Holzer, A.; Ponak, C.; Raupenstrauch, H. Pyrometallurgical Lithium-Ion-Battery Recycling: Approach to Limiting Lithium Slagging with the InduRed Reactor Concept. *Processes* **2021**, *9*, 84. https://doi.org/10.3390/ pr9010084

Received: 26 November 2020 Accepted: 28 December 2020 Published: 2 January 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional clai-ms in published maps and institutio-nal 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/).

with pre-treatment processing which aims to separate battery components like the battery management system or the housing and the corresponding materials such as plastics or iron from the active materials of the battery electrodes. For pre-treatment, various processes can be found which differ more or less from each other. Basically, steps during pre-treatment are sorting, dismantling deactivation and mechanical processing and separating. Said active materials–after pre-treatment they are also known as black matter–mainly consist of lithium metal oxides or lithium iron phosphate, carbon and residues from the electrode conductor foils. Umicore's Valéas process, for example, is an exception since it does not need a usual pre-treatment but uses the batteries directly in their pyrometallurgical process. [2,6–12].

While the obtained metal and plastic scrap can be recycled directly, the produced black matter that contains the valuable metals needs to be further treated in a metal extraction step to recover Li, Ni, Co and Mn, at best in a quality that is suitable for closed loop recycling. Therefore, pyro-, bio- and hydrometallurgical methods can be used. Biometallurgical processes like bioleaching are considered as environmentally friendly and low cost alternatives to conventional hydrometallurgy, capable of reaching recovery rates of more than 98% for Ni and Co and more than 80% for Li but suffering from low kinetics and resulting poor throughput rates [13–19].

Typical hydrometallurgical procedures, used to recover metals from black matter, are leaching, solvent extraction, chemical precipitation or electrochemical deposition, with which a high selectivity and therefore product purity can be achieved [20]. The possible recovery rates for Ni, Mn, Co and Li, as for example reported by He et al. [21], can be close to or even higher than 99%. The obtained salts or concentrates can usually be directly used for the production of new cathode materials as it is the case for the Duesenfeld process described by Elwert and Frank [22]. An indication of the importance of hydrometallurgical recycling of spent LIBs is, among other things, the high intensity of research activities in this field. According to Huang et al. [23], more than half of the recycling processes that are currently under investigation are related to hydrometallurgical processing.

Pyrometallurgical approaches use high temperatures, usually above 1400 ◦C, and reducing conditions to recover valuable metals as a metal alloy. The advantages lie in the experience with and the properties of conventional pyrometallurgical units which are less complex and less vulnerable, e.g., to organic impurities in the black matter, than their hydrometallurgical counterparts. The decisive factor in this regard is the oxygen potential of the contained metals, which is for example low for Ni and Co, leading to a relatively low-effort recovery. On the other hand, the similarity of the oxygen potential between Ni and Co reduce the selectivity of pyrometallurgical processes since they cannot be recovered separately but only as an alloy. The oxygen potential is also responsible for one of the biggest disadvantages of pyrometallurgy. Lithium, which has a much higher oxygen affinity, cannot be recovered as part of the metal alloy but is bound as an oxide in the slag instead [2,20,24–26].

The refining step is usually based on hydrometallurgical methods and aims for a closed loop recycling. Hence, it mainly applies on the metal alloy and slag from pyrometallurgical processing, which without further treatment, cannot be used for the production of new LIBs. The treatment of the metal alloy aims for a separation of the contained metals, while the slag treatment's goal is to recover Li, which is often technically but not economically feasible due to the low Li content in the slag. [2,20,22,27]

However, it can be summarized that there are still a lot of uncertainties in the LIB recycling chain. Not only the development of the waste stream itself, also the number and diversity of pre-treatment processes lead to varying black matter compositions and qualities. For pyrometallurgy, the lack of Li recovery options is a major problem that is not yet solved, but however, gives the desired novel approach with the InduRed reactor a good opportunity to establish itself as an alternative to conventional processes.

The mentioned InduRed reactor might be a possibility to achieve a simultaneous recovery of Ni, Co, Mn as well as Li with a pyrometallurgical process. The existing pilotscale reactor concept, shown in Figure 1a,b, consists of a packed bed of graphite pieces that

is inductively heated by surrounding copper coils. The input material is fed continuously with up to 10 kg/h from the top onto the hot graphite bed. The uppermost induction coil powers the upper third of the reactor where the input material melts and forms a thin molten layer that moves downwards. A second induction coil, placed half way down the reactor, induces enough power so that reduction reactions can take place. Gaseous reaction products are then removed from the reactor via a flue gas pipe whereas the liquid products move further down. The third induction coil makes sure that the temperature within the reactor can be maintained well above the melting temperature of the mixture and enables a continuous flow out of the bottom of the reactor. The advantages of the reactor are the low oxygen partial pressure, the possibility to control different temperature zones, and the big reaction surface due to the graphite bed. Furthermore, the contact time and intensity between gaseous reaction products and the molten phase can be limited because they only need to pass a thin layer or droplets instead of a molten bath like in conventional pyrometallurgical furnaces. Originally, the reactor concept was developed for the recovery of phosphorus from sewage sludge ashes, which is described by Schönberg [28]. The concept was later also adapted by Ponak et al. [29,30] to treat basic oxygen furnace slag with limited iron phosphide formation.

**Figure 1.** (**a**) Schematic illustration of the so called InduRed reactor and (**b**) said reactor operating at a test series for metal recovery from basic oxygen furnace slag. [29,30].

The aim of this work is to investigate if said reactor concept can potentially also provide a solution for LIB black matter recycling.

For the determination of the basic suitability of black matter as an input material for the InduRed reactor, thus its melting and reaction behaviour, heating microscope experiments, thermogravimetric analysis (TGA) and differential scanning calorimetry (DSC) were carried out. Since black matter can have different properties and contents of impurities depending on the pre-treatment procedure, the influence of which on the properties being investigated is difficult to assess, the investigations are also carried out with pure cathode materials. The ability of the InduRed reactor concept to eliminate one of the biggest disadvantages of pyrometallurgical LIB recycling, namely lithium slagging, is finally evaluated by experiments in a lab-scaled batch reactor, which is based on the InduRed concept. The results, in particular the required reaction temperatures and the Li removal rate via the gas phase from the reactor, form the basis on which a decision is made about the fundamental suitability of the reactor to be part of the LIB recycling chain.

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

The cathode materials (LiNi0.8Co0.15Al0.05O2, sample abbreviation: NCA and LiNi0.33 Mn0.33Co0.33O2, sample abbreviation: NMC) which were used for the experiments have been produced by Gelon Energy Corporation in Linyi, China, while the black matter (sample abbreviation: AM) was provided by a LIB recycling facility operated by Redux GmbH in Bremerhaven, Germany. The chemical composition of said materials is summarized in Table 1 below.


**Table 1.** Chemical composition of used materials. (mass fraction, w/%).

<sup>1</sup> Data from ICP-MS analysis. <sup>2</sup> Calculated from the molar composition of the cathode materials.

In experiments with NCA and NMC, where reducing conditions were desired (sample abbreviation: NCA\_C, NMC\_C), fine powdered coke was used as a reducing agent. Since AM already contains 29.5 w/% carbon there was no need to further add a reducing agent.

In order to investigate the general behavior of the cathode materials at high temperatures and under reducing conditions, the work started with two preliminary experimental series. First, heating microscope experiments were conducted in a Hesse Instruments EM 201 with an HR18-1750/30 furnace (Hesse Instruments, Osterode am Harz, Germany) to investigate at which temperatures reactions or transformations in the sample occur. In the heating microscope experiments, black matter and the cathode materials with and without carbon addition were tested at least twice to check the reproducibility of the results. In the reduction experiments, carbon was added in extents of 10 w/% to the NCA and NMC materials. An argon purge with a flow rate of approximately 2.5 l/min was used to inhibit oxidization reactions of the materials. The settings for the heating rate (80 ◦C/min until 1350 ◦C, 50 ◦C/min until 1550 ◦C and 10 ◦C/min until 1650 ◦C with a holding time of 5 min at 1650 ◦C), the used Al2O3 sample plates and the sample size of approximately 0.1 g of powder, pressed in a cylindrical shape, were the same for all experiments.

The second series were simultaneous thermal analyses (STA), more precisely thermogravimetric analysis (TG) and differential scanning calorimetry (DSC), which have been conducted in a Setaram Setsys Evo 2400 at the Chair of Physical Metallurgy and Metallic Materials at the Montanuniversitaet Leoben. The aim of the STA was to confirm the temperature zones in which changes of the materials could be observed in the heating microscope and to further characterize the underlying reaction mechanisms. An argon purge was used to inhibit oxidization reactions of the materials. For carrying out the analyses, graphite crucibles and a carbon addition of 25 w/% were used in order to prevent damages to the analysis hardware. The need for this is due to reactions between standard Al2O3 crucibles and the produced metal alloy when carbon is added to the mixture. The reactions lead to a destruction of the Al2O3 crucible and the thermocouple underneath gets destroyed. The higher amount of carbon in the STA experiments is needed to prevent reactions between the cathode material and the graphite crucible, which would take part as a reductant.

To simulate the conditions of the InduRed reactor and check its suitability, a third set of experiments has been performed in the so-called InduMelt plant (sample abbreviation starts with: IM\_). The InduMelt plant is a single coil induction furnace that is modeled on the InduRed reactor concept and used to perform preliminary experiments. This is due to the fact that the InduMelt plant is easier to use and requires less effort compared to the continuous InduRed reactor but still provides the same reaction conditions. The crucible concept used for these experiments is therefore based on the InduRed reactor and consists of a bed of packed graphite cubes (25 mm edge length) within an Al2O3 ceramics ring (70 mm radius, 100 mm height) and is shown in Figure 2a. In Figure 2b, the setup of the InduMelt plant is presented.

**Figure 2.** (**a**) Schematic illustration of the crucible concept used in the InduMelt experiments [29] and (**b**) setup of the experimental InduMelt plant.

During the preparation of the experiments, the ceramic ring is fixed on a mortar plate with refractory mortar and alternatingly filled with graphite cubes and input material. In the conducted experiments, the initial sample mass was 552.3 g for IM\_NMC\_C, 520.0 g for IM\_NCA\_C and 561.9 g for IM\_AM. The filled crucible is then insulated, using 20 mm thick Cerachrome fiber wool with a classification temperature of 1500 ◦C and placed within the induction coil. The inductive energy input is controlled in such a way that the temperature increases at a maximum rate of 200 ◦C/h. For the measurement of temperatures of the reactor, two k-type thermocouples are used inside of the reactor to control temperatures up to 1200 ◦C. To keep track of the temperature after the k-type couples fail due to the high temperatures, two separate s-type couples are mounted on the outer wall of the Al2O3 ceramics ring. The temperature distribution in the reactor is known from previous experiments with other waste streams and can show a gradient of several 100 ◦C towards the end of the experiment, with the highest temperatures occurring at the top of the reactor. The s-type thermocouples are therefore placed at the lower third of the reactor in order to reach the necessary temperatures in the area in which the material is supposed to accumulate.

After the experiments, the reactor needs to cool down for at least 24 h before the sampling can start. Hereby, every graphite cube was picked from the reactor one after another and checked for any metal or slag depositions, which, if present, were removed from the cube's surface and collected. The difficulty to collect every little metal deposition and its influence on the overall mass balance of each experiment is discussed in the results section of this work.

However, representative samples were taken from the collected products and the content of species of interest was examined using inductively coupled plasma mass spectrometry (ICP-MS). For all ICP-MS measurements, which were conducted at the Chair of Waste Processing Technology and Waste Management at the Montanuniversitaet Leoben, the sample preparation was done by aqua regia digestion according to the ÖNORM EN 13657 standard. The measurement of the respective species was carried out according to the ÖNORM EN ISO 17294-2 standard.

### **3. Results**

### *3.1. Heating Microscope*

In the heating microscope experiments, the relative cross-sectional area (CSA) of the sample, thus the trend of cross sectional area of the sample cylinder during heating in relation to its initial value, was observed to investigate at which temperatures changes in the material occur. In Figure 3a, where the results of the test series with NMC are shown, one can see a significant difference between the graphs of NMC\_1 and NMC\_2 without carbon addition and, respectively, NMC\_C\_1 and NMC\_C\_2 in which carbon was added. In this case, the first change of the CSA for NMC\_C\_1 and NMC\_C\_2 can be observed at approx. 800 ◦C, which is almost 200 ◦C lower than in the tests without carbon addition.

Moreover, the extent to which the change occurs is significantly higher in experiments with carbon addition. The steep decline of NMC\_C\_1 and NMC\_C\_2 at approx. 1500 ◦C was also observed with other cathode materials and can be explained by the melting point of the contained metals. The difference in the trends of the CSA with and without carbon addition can be explained by the origin of the changes. Mao et al. [31] and Kwon and Sohn [32] investigated and described the reaction behaviour of LCO (LiCoO2) with and without carbon addition. According to their findings and due to the fact, that NCA and NMC are structurally identical to LCO, we assume that the changes in experiments without carbon addition are caused by thermal decomposition of the lithium metal oxides, while in experiments with carbon addition, reduction reactions with Li2O formation led to the observed changes. About the reproducibility it can be said that in the repeated attempts the characteristic changes of the CSA appear at the same temperatures to about the same extent.

**Figure 3.** (**a**) Comparison of the cross sectional area of NMC with and without carbon addition in the heating microscope; (**b**) Comparison of the cross sectional area of different mixtures of NMC and NCA, each with carbon addition.

The results, mainly temperature zones and the extent of the correspondence of changes of the CSA, for NCA and NCA\_C are very similar to those for NMC and NMC\_C. However, since future waste streams are likely to consist of mixtures of different cathode materials, another set of experiments was performed in which NCA and NMC in different compositions and carbon were mixed to investigate if the materials influence each other. In Figure 3b, where the changes of the CSA of NCA\_C, NMC\_C and mixtures with varying composition are shown, no direct influence can be seen. The following Figure 4a,b show the NMC\_C sample before and after the heating microscope experiment. In Figure 4b a perfectly molten metal sphere, indicated by the change of the CSA at approx. 1500 ◦C, and a fine white crystalline structure can be seen. The blue colour of the Al2O3 ceramic is most likely caused by reactions with cobalt and was also observed in all other experiments, especially in those with carbon addition.

**Figure 4.** (**a**) NMC\_C sample before and (**b**) after heating to 1600 ◦C in the heating microscope.

In contrast, the black matter material (AM) showed a completely different behavior, as can be seen in Figure 5a,b in which its CSA does not decrease during heating but increase to almost 120% of its initial value. The lack of the first change of the AMs CSA as well as the absence of any sign of melting at temperatures around 1500 ◦C indicates that pretreatment might have a big influence on basic thermophysical properties of the produced black matter.

**Figure 5.** (**a**) Images of the samples AM (i), NMC\_C (ii) and NCA\_C (iii) at temperatures of 20 ◦C, 1000 ◦C and 1500 ◦C taken during the heating microscope experiments; (**b**) Trend of the cross sectional area of the samples AM, NMC\_C and NCA\_C during heating in the heating microscope.

Reasons for the deviating behavior of AM compared to NMC\_C and NCA\_C could lie in impurities, thus residues from the mechanical processing and separation step during pretreatment, like Cu and Al from conductor foils. A closer look at the chemical composition of AM in Table 1 reveals that the mass content of Cu and Al with almost 6% each is much higher than anticipated. Moreover, the carbon content is much higher than would be stoichiometrically necessary for the reduction reactions. An example of a disruptive reaction could be the formation of aluminum oxide which, in the appropriate amount, could form a supporting structure and thereby reduce the informative value of the CSA. On the other hand, it is also possible that the anode graphite has a lower reactivity than the fine powdered coke which is used in NMC\_C and NCA\_C.

The origin of AM, a pre-treatment process that uses thermal deactivation before mechanical shredding, could also cause the observed differences, since some of the reactions might already have taken place if certain temperatures are overcome during this step. By this, the layered structure of the lithium metal oxides could probably have been changed, e.g., due to thermal decomposition which, as can be seen in Figure 3, occurs at

approx. 1000 ◦C and could change the materials properties permanently. However, reliable information about these thermal processes is difficult to access. In our opinion, however, it is quite possible that at certain points in such a process, temperatures above 1000 ◦C can occur and that therefore the possibility of influencing the material must not be excluded.

### *3.2. Simultaneous Thermal Analysis*

The experiments in the heating microscope gave some first impressions on how NMC, NCA and AM behave at high temperatures and under reducing conditions. For further characterization of the underlaying reactions that cause respective changes in the materials and to create a basis for a kinetics model in the long term, thermogravimetric analysis and differential scanning calorimetry was conducted. The results of the STA are summarized in Figure 6a, showing the trends of the relative mass of the samples, and Figure 6b, which shows the corresponding trends of the heat flow. The evaluation of the measurements, which also includes a correction of the data by reference measurements, was carried out in MATLAB.

**Figure 6.** Results of the simultaneous thermal analyses of NMC\_C and NCA\_C with a heating rate of 40 K/min. (**a**) Trend of the relative mass of NMC\_C and NCA\_C during heating. (**b**) Trend of the heat flow of NMC\_C and NCA\_C during heating.

In Figure 6a the beginning of the mass loss at approximately 800 ◦C matches the observations from the heating microscope experiments. The first mass loss first declines slowly before it becomes steeper around 1000 ◦C and stops at approximately 70 % of the initial mass which was 40.1 mg for NCA\_C and 39.8 mg for NMC\_C. At the end of the thermogravimetric curve, the relative mass is about 55% of the initial mass. This means, that additionally to carbon, which had an initial mass content of 25 w/%, also components of the lithium metal oxide, most likely O2 and Li, had been removed from the sample. Another indication for the presence of reduction reactions between 800 ◦C and 1000 ◦C is the trend of the heat flow, shown in Figure 6b. In both samples, the heat flow between 800 ◦C and approximately 1050 ◦C is endothermic with a negative peak around 1000 ◦C where also the biggest slope of the sample mass occurs. The outstanding exothermic peak in the NCA\_C at 700 ◦C heat flow trend could be the result from Al2O3 formation whereby a significant amount of heat could be released. In order to confirm this, the samples must be heated in a controlled manner to or just above this temperature and analysed using XRD analysis, which is planned within the further scope of the research project.

As in the heating microscope experiments, the behaviour of the sample AM differs greatly from that of NCA\_C and NMC\_C. The overall mass loss only accumulates to around 10% and there are no sharp peaks in the heat flow trend. The lower mass loss is on the one hand due to the comparatively lower lithium metal oxide content (<60 w/%) compared to NCA\_C and NMC\_C (75 w/%) and the resulting decreased ability for CO or CO2 generation. Since the heating rate was the same in all experiments, the less steep mass

loss between 800 ◦C and 1000 ◦C and the absence of significant peaks in the heat flow trend indicate a lower reactivity of AM in general. The suspicion from the heating microscope experiments that certain reactions already took place during the thermal deactivation step has gotten stronger.

Finally, the results from the heating microscope experiments and the STA are summarized in Figure 7.
