**1. Introduction**

Over the course of the last few years, the pyro-metallurgical recycling process of electronic scraps, so-called WEEE (waste of electrical and electronic equipment), developed significantly not only because of governmental restrictions and statutory recycling rates, but also due to valuable and economically interesting amounts of metal contents (e.g., Cu, Au, Ag) [1]. Nevertheless, the industrial processes are tailored to recover only valuable metals from e-waste. In general, when talking about e-waste, valuable metals are concentrated mainly on PCBs (printed circuit boards). The metals found in and on PCBs can be classified into five main groups, namely, base metals (BMs), precious metals (PMs), platinum group metals (PGMs), metals of concern (MCs, hazardous), and rare earth elements (REEs) [2,3].


The general pyro-metallurgical recycling process of PCBs (see Figure 1) is based on the combustion of the organic constituents of the material mix fed into the furnace and using the energy of this exothermic reaction to melt the other constituents. Therefore, depending on the desired process and the PCB content in the feed mix, the organic content is used either to substitute fossil energy (organic content ≤20 wt. %. of the feed mix) or to run the entire process autothermal (organic content >20 wt. %. of the feed mix). Thus, an o ff gas containing combusted organics, flame retardants (bromide and chloride compounds), and parts of the zinc as oxide is produced. Currently, processes are able to recover most of the base metals (Cu, Ni, Sn), as well as the precious metals (Au, Ag, Pd, Pt), contained within the input material in a metal phase at the bottom of the furnace. Above this metal phase, a slag phase will form, containing the ceramic compounds of the feed mix, such as aluminum oxide, silicon dioxide, and magnesium oxide, as well as oxidized metals such as iron and aluminum. Unfortunately, there is always some of the base metal alloy, which is entrapped as droplets inside the slag (~1–2 wt. % of the slag). Within these entrapped droplets, some of the precious metals, as well as parts of lead and tin in their metallic state, are lost to the slag. Depending on the process, atmosphere, and slag chemistry, oxidic forms of lead and tin may also occur. The last group of elements, i.e., the scarce elements, are mainly bound in the slag. However, depending on the chemical reactions inside the furnace, halides may also be formed and lost to the o ff gas [4]. Although the recycling process is quite well developed and plenty of research was done in that field, the distribution of scarce elements is still not fully understood, and an e fficient recovery method is ye<sup>t</sup> to be developed.

**Figure 1.** General schematic process of metal recycling from printed circuit board (PCB) scraps [5].

In order to reduce metal losses to the slag, facilitate their recovery, and possibly recover REEs, further investigation must be conducted. In this context, FACTSage ™ (Thermafact/CRCT, Montreal, Canada and GTT-Technologies, Aachen, Germany) [6] o ffers one suitable opportunity to help understand the influence of slag chemistry on the viscosity of the slag, consequently enabling a detailed

understanding and manipulation of the settling behavior to reduce these losses. A possible solution, which is investigated within this study, is the targeted concentration of metal parts, as well as scarce elements, in certain phases inside the slag, achieved through controlled cooling rates and slag chemistry. A positive concentration would facilitate further separation, collection, and subsequent recycling of these elements, enabling the recycling of scarce elements to become technically and economically feasible [7,8].

With respect to a sustainable recycling strategy, a detailed understanding of the elemental distribution within the process, the reduction of metal losses to the slag, and the material properties of the product is essential. Therefore, the analysis of the present slag material in terms of formed structures depending on the cooling rate and the enrichment of metals in these structures is of grea<sup>t</sup> interest. Two main analysis methods were used in the present study, namely, SEM-based mineralogical analysis (MLA) and X-ray computed tomography (XCT).

SEM-based mineralogy (MLA) is a two-dimensional image analysis method based on scanning electron microscopy coupled with energy-dispersive X-ray spectroscopy [9,10]. It is proven to be a practical method for multidimensional characterization of various phases [11–13]. Using MLA allows for an integrated analysis of shape, size, and distribution of various phases within a specimen which can be a sample block, thin section, or a grain mount. By studying processes such as comminution or physical separation, process parameters can be directly related to product characteristics with respect to particle properties [14–17]. For example, the liberation of valuable phases with respect to grinding fineness, as well as the dominating fracture mechanisms, can be studied [18–20].

XCT enables a non-destructive three-dimensional characterization of specimens. As it is based on the X-ray attenuation of the volume elements (voxel), which is a function of the materials' average atomic number, wavelength, material density, and path length, voxels can have di fferent chemical composition but similar attenuation coe fficients [12]. Studying multiphase materials such as ores and slags can, thus, be challenging. Di fferent approaches exist to face this problem. These are direct three-dimensional analysis of the chemical composition using X-ray fluorescence tomography [21,22] or the combination of two-dimensional analysis of the chemical composition with XCT [19,23–25].

The aim of the investigation is to identify if the slag could possibly represent an artificial resource or collecting agen<sup>t</sup> for various valuable metals and rare earth elements. Based on the findings of this study, it is planned to develop an e ffective and e fficient method for the downstream recycling process of the addressed elements. If it is possible to do so, a huge pyro-metallurgical process optimization potential is expected, as this resource can be influenced in various ways and no detailed knowledge is generated so far.

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

The slag, which was further investigated in this study, resulted from a previously conducted PCB melting experiment in a TBRC (top-blown rotary converter) of industrial scale (0.5 m<sup>3</sup> melt volume) [5]. Within the experiment, 560 kg of finely ground PCBs were injected into a synthetic slag phase at 1300 ◦C starting temperature. The main elemental components of this injected material were Cu, Pb, Zn, Al, Ni, and Sn. For more details on the injected material, the reader is referred to Borowski et al. (2018) [5]. The general schematic experimental set-up is shown in Figure 2.

**Figure 2.** Schematic process of waste of electrical and electronic equipment (WEEE) powder injection into the slag phase.

The experiment was conducted with an initial molten slag bath formed by a composition of SiO2, Al2O3, and CaO, chosen so that the lowest possible liquid temperature was achieved. The powder and fines were injected into the slag phase near the metal slag boundary through the injector (see Figure 3).

**Figure 3.** Top-blown rotary converter (TBRC) with lances for injecting WEEE powders and oxygen [5].

The mass flow of the injected WEEE powder was adjusted according to the calculated energy requirement for an autothermic melting process between 1350 ◦C and 1450 ◦C and was based on the calorific value of the organic content of approximately 30 wt. % in the feed material. After 560 kg of material was continuously injected into the slag phase and 30 min of holding time elapsed, both slag and metal phases were tapped at 1450 ◦C into a cold (20 ◦C) cast-iron ladle that was coated with a thin graphite layer (for more detailed information on the smelting process from which the investigated slag originates, see Borowski et al. (2018) [5]). In the ladle (see schematic in Figure 4), the metal was intended to settle at the bottom and to cool relatively quickly while the slag cooled down with different cooling rates from the outside to the center. At the ladle's cold side walls, in a very small zone 1, very high cooling rates of approximately >100 K/s are realized. Here, the slag is intended to form an amorphous, glass-like structure, whereas, in the center of the ladle, in zone 3, a mineral phase with crystalline structures is expected. Here, due to the solidified surrounding slag, the inner core is

literally thermal insulated, and it features a very low cooling rate of approximately <1 K/s. In between these two zones, the biggest zone 2, with an "intermediate" cooling rate of approximately 5 K/s (mean cooling rate of the entire zone), is formed.

**Figure 4.** Schematic illustration of the slag ladle with expected cooling zones and cooling rates.

The main difference in metal concentration is expected between the inner (zone 3) and outer phase (zone 1) of maximal and minimal cooling rates. The slags of intermediate cooling rates (zone 2) will probably not differ much in terms of results from the inner phase (zone 3), as the cooling rates are too close to each other.

#### *2.1. X-Ray Computed Tomography*

In order to analyze the three-dimensional (3D) structure of the slag, the fragments of samples from the slag ladle (see Figure 4) of zone 1 (H), zone 2 (M), and zone 3 (S) were subsampled by drilling small cores (4 mm in diameter) from them. These cores were then glued on sample holders and scanned using X-ray computed tomography (XCT). The used XCT apparatus was a Zeiss Xradia 510 Versa X-ray microscope (Carl Zeiss AG, Oberkochen, Germany), which combines geometric and optical magnification for increased spatial resolution (Figure 5). The total magnification results from the product of geometric and optical magnification. Detailed information on XCT can be found, e.g., in a review article by Hanna and Ketcham (2017) [26].

**Figure 5.** Working principle of an X-ray microscope combining geometric and optical magnification for increased spatial resolution.

XCT scans were done with two sets of parameters to analyze the whole core in a large field of view (FOV, 7.8 μm voxel size) and to additionally ge<sup>t</sup> a high-resolution virtual subsample via an interior tomography using a small FOV (2.77 μm voxel size). The latter enabled a detailed view of small structures formed by the complex intergrowth of the different slag phases. The scan parameters are listed in Table 1.


**Table 1.** Scan parameters from X-ray computed tomography (XCT) analysis.

The projection images were reconstructed using the scanner's proprietary software Scout&Scan Reconstructor (Version 11.1.8043, Carl Zeiss AG, Oberkochen, Germany). The analysis of the final image stack was done with the software VGStudio MAX 3.3 (Volume Graphics GmbH, Heidelberg, Germany).

#### *2.2. Scanning Electron Microscopy-Based Mineralogical Analysis*

In preparation for the investigation via MLA, the samples were crushed and ground down to a particle size of <500 μm. Subsequently, the material was screened into four di fferent size fractions. Samples of these size fractions were analyzed separately by MLA. The preparation of the samples is shown in Figure 6.

**Figure 6.** Preparation of samples H, M, and S for analyses via SEM-based image analyses; after a sequence of jaw and cone crushing, the material is milled down to an upper particle size of 500 μm. The material is then screened into four size fractions.

The samples were analyzed by means of a mineral liberation analyzer (MLA). Grain mounts of the particle samples were prepared by mixing 3 g of material with the same volume of graphite and epoxy resin. The grain mounts were polished and subsequently carbon-coated with a Leica (Baltec) MED 020 vacuum evaporator (Leica Microsystems, Wetzlar, Germany). The MLA consisted of an FEI Quanta 650F (Thermo Fisher Scientific, Waltham, MA, USA) field-emission SEM (FE-SEM) equipped with two Bruker Quantax X-Flash 5030 (Bruker, Billerica, MA, USA) energy-dispersive X-ray detectors (EDX). Identification of mineral grains by MLA involves backscattered electron (BSE) image segmentation and collection of EDX spectra of the particles and grains distinguished in GXMAP (grain-based X-ray mapping) mode. In this mode, EDX spectra of each particle are collected in a dense grid and further classified, using a list of mineral spectra collected by the user. The GXMAP measurement mode was applied to all samples [10]. Data processing and evaluation was done with the software package MLA Suite 3.1.4.686. The scan parameters for measurement are listed in Table 2.


**Table 2.** Scan parameters from mineral liberation analyzer (MLA) analysis. BSE—backscattered electron.
