Introduction

Due to technological advances and the improvement of economic and social conditions around the globe, the number of electronic devices in total and per capita is increasing. Meanwhile, the lifecycle of devices has become shorter as a result of rapid technology updates and increasing disposable income [1]. The United Nations University has estimated that, for example, in 2016, the global generation of WEEE (waste electrical and electronic equipment) was 44.7 million tons [2] and in 2018, 49.8 million tons [3], and it is expected to reach 52.2 million tons by 2021 [4]. WEEE consists of a wide range of materials, some of which have very high economic value. Due to the huge amounts produced every year, WEEE is considered as a valuable urban resource. Various methods, such as physical separation, electrostatic separation, pyrometallurgical, and hydrometallurgical processes have been investigated with the focus of recycling the valuable metals from these secondary resources [5, 6].

Reprocessing WEEE to recover the valuable metals is not a simple task. WEEE is a heterogeneous mixture of metallic and non-metallic fractions, where the components containing valuable metals are strongly bonded to other components that may have little or no value [7, 8]. Furthermore, common electronic devices may include lead, mercury, and arsenic, which are hazardous elements. Proper management of e-waste is, therefore, a must, as it can pollute ground water, acidify the soil, generate toxic fumes and gases in burning, accumulate in municipal disposal areas, and release carcinogenic substances [9].

Waste-printed circuit boards (WPCBs) are an important part of WEEE: they account for 2–3% of its total mass [4, 10]. WPCBs consist mostly of nonconductive insulator materials and several metals, such as copper and precious metals (PMs). The concentrations of PMs, specifically gold, silver, platinum, and palladium, in WPCBs can be considerably higher than in their primary sources [11]. Moreover, as primary precious metal resources are steadily depleting, the demand for these metals is growing in many sectors. Therefore, recycling of precious metals from WPCBs is of significant interest. The value of precious metals comprises 80% of the value of scrap electronic devices, despite their small quantities [12]. For these reasons, manufacturers, environmental agencies, and governments around the globe are attempting to find systematic and environmentally sustainable ways of recycling precious metals from WPCBs.

Pyrometallurgical processes are capable of dealing with complex e-waste streams [13,14,15,16]. WPCBs, in addition to their high precious metal content, also contain more than 20% copper [17]. Considering the economic and technical aspects, copper smelting is one of the predominant routes for WPCB and e-waste recycling [18]. Copper flash smelting is a mature technology that has been studied and employed by industry for decades [19,20,21], and the combustion reactions of chalcopyrite and other sulfides in it have also been widely investigated [22,23,24,25]. WPCBs can be processed along with the primary feed of copper concentrates. However, the elemental composition of WPCBs and other e-waste is often such that it can affect the processing conditions, for instance by changing the liquidus temperature or viscosity of the slag or by making the slag more corrosive towards refractories. Another issue related to WPCB smelting is that the mix of elements in WPCB does not match that in common primary mineral concentrates, for which the smelting technology has been optimized over many years and for which a large database of information has accumulated [7, 26,27,28].

To obtain the optimal process parameters for WPCB smelting and precious metals recovery, the thermodynamics and kinetic behavior of the valuable elements and their reaction mechanisms must be known. Equilibrium research regarding the precious metal distribution between matte and different slags has been intensively carried out during the past few decades. The investigated slag systems include, for example, FeOx–CaO [29], FeOx–SiO2 [30, 31], FeOx–SiO2–MgO [32], and SiO2–CaO–FeOx–MgO [33]. The distribution characteristics have also been studied as a function of matte grade with various sulfur and oxygen potentials [34,35,36]. However, studies on the kinetic behavior of precious metals in the matte-slag system are lacking [37], and more research is urgently needed to optimize the methods to recover precious metals from WPCBs through the copper smelting process.

The time dependency of precious metal (Au, Ag, Pt, and Pd) distribution was experimentally investigated in laboratory-scale copper matte smelting conditions. The results contribute to deeper understanding of PM distribution behavior during the settling process in the flash smelting furnace (FSF) and provide guidance in finding better strategies and methods for recovering the precious and platinum group metals from e-waste via pyrometallurgical routes.

Experimental

Materials

The raw materials used in this study were industrial chalcopyrite copper concentrate, synthetic iron oxide-silica slag, and PM (Ag, Au, Pt, Pd) powders. The concentrate was provided by Boliden Harjavalta, Finland. Its chemical composition was analyzed by X-ray fluorescence (XRF) spectrometry (Malvern Panalytical B.V., Almelo, The Netherlands) and is shown in Table 1.

Table 1 Chemical composition of the copper concentrate (wt %)

The iron oxide-silica slag was prepared by mixing 65 wt% of hematite powder (Alfa Aesar, 99.99% purity) with 35 wt% of silica powder (Umicore, 99.99% purity). The preparation of the slag has been described in detail in a previous study [38]. Instead of real WPCB, pure Ag, Au, Pt, and Pd powders (Alfa Aesar, ≥ 99.9% purity) were used as sources of precious metals in the experiments.

Apparatus and Procedures

The experimental apparatus used in this study is shown in Fig. 1 and has been described in detail earlier [24]. The silica crucible, filled with sample material, was raised into the hot zone (temperature 1300 °C) with a Pt-hanging wire. After a preset time, the reactions were stopped by dropping the sample into a quenching vessel filled with ice water. For the experiments in an argon atmosphere, the work tube was sealed from the bottom, and a flow of 400–500 mL/min argon (AgA Linde, 99.999% purity) was used.

Fig. 1
figure 1

Schematic of the furnace and geometric details of the hot zone

A slag to concentrate ratio of 1.116 was chosen in accordance with industrial operation practice [24]; this corresponds to a SiO2/Fe flux ratio of 0.53. Usually, the PM concentrations in real WPCBs are relatively low (varying from 80 to 3300 ppm) [39]. In this study, the amount of each metal (gold, silver, palladium, and platinum) was measured to be 2.5 wt% of the amount of concentrate (Table 2) considering the detection limitations of EPMA and LA-ICP-MS analyses [30, 34]. All the powders were ground together in a mortar to obtain a homogenous slag–concentrate-PM mixture. The mixture was then inserted into cone-shaped silica crucibles (Finnish Special Glass, Finland) with a mixture sample weight of 0.5236 g for each experiment, measured by an analytical balance (AB204, Mettler Toledo, USA).

Table 2 Mass contents of the raw materials in each experiment

The contact times were chosen to be 10, 20, 30, 60, 150, 300, and 600 s in air. For experiments in an argon atmosphere, 5, 10, 20, and 40 min were chosen. Each experiment was repeated at least twice, and a sufficient number of phase areas was analyzed in every polished section of the samples in order to ensure the reproducibility and reliability of the results.

Analyses

The samples were prepared with basic metallographic methods and first analyzed with a Mira 3 scanning electron microscope (SEM, Tescan, Brno, Czech Republic) equipped with an UltraDry silicon drift energy dispersive X-ray spectrometer (EDS, Thermo Fisher Scientific, Waltham, MA, USA) coupled to NSS microanalysis software (Thermo Fisher Scientific, Waltham, MA, USA). Considering the insufficient detection limits of EDS [40, 41] and PM concentration results from a previous equilibrium study, the chemical compositions of the matte phase were investigated by electron probe micro-analysis (EPMA) and the slag phase was analyzed by laser ablation-inductively coupled plasma-mass spectrometry (LA-ICP-MS). LA-ICP-MS was essential for the slag phase due to its very low PM concentrations.

The EPMA used in this work for matte phase analysis was a Cameca SX100 equipped with five wavelength dispersive spectrometers (Cameca SAS, Gennevilliers, France). The selected accelerating voltage was 20 kV, beam current 60 nA, and beam diameter 100 µm. The analyzed lines and standards (Astimex) used were as follows: Fe Kα and O Kα (hematite), Mg Kα (diopside), S Kα (pentlandite), Cu Kα (Cu), Zn Kα (sphalerite), Pb Lα (galena), Pd Lα (Pd), Ag Lα (Ag), Pt Lα (Pt), and Au Lα (Au).

The equipment used for slag trace element analysis was a Photon Machines Analyte Excite laser ablation system with a 193 nm wavelength 4 ns ArF excimer laser (Teledyne CETAC Technologies, Omaha, USA) coupled to an Nu AttoM single collector sector field ICP-MS (Nu Instruments Ltd., Wrexham, UK). The laser spot size was selected as 65 µm for the air-atmosphere experiments, and 40 µm for the argon-atmosphere samples, due to the limited size of slag areas available. The obtained detection limits of EPMA and LA-ICP-MS are shown in Table 3, and more detailed information about the technique adopted in our study can be found in the references [34, 40, 42,43,44].

Table 3 EPMA and LA-ICP-MS detection limits in ppmw (parts per million by weight)

Results and Discussion

Smelting Process and Sample Microstructures

In this study, copper concentrate, iron oxide–silica slag, and PMs (Au, Ag, Pt, and Pd) were pre-mixed and reacted at a typical copper smelting temperature of 1300 °C in both air and argon atmospheres. The purpose of the air-atmosphere experiments was to simulate the reaction shaft processes of the flash smelting furnace, while the argon atmosphere was used for simulating the reactions between matte and slag in low oxygen partial pressure (typically around 10−8 to 10−7 atm) on the slag–metal interface in the settler [45].

The SEM-backscattered electron micrographs of samples after different contact times in air are shown in Fig. 2. The reactions proceeded quickly, as the matte (A) and fayalite FeOx–SiO2 slag (B) phases already began to form after 10 s, which is consistent with the rapid reaction phenomena occurring in the flash smelting process [46]. The structure of the sample was first extremely heterogeneous and the interfaces between the matte and slag phases were not clear. Unreacted ferric oxide (D) was also detected. After 20 s, the fayalite slag areas had increased in size and started to segregate more clearly from the sulfide matte. The matte was rather randomly distributed, which showed that coalescence and settling had not taken place at this point. After 30 s of contact time, most of the remaining un-melted silica (C) had reacted and dissolved into the slag, and a clear interface between matte and slag was formed. With increasing contact time, phase coalescence and matte settling progressed further. Some areas were detected in the form of a PMs-Cu-Fe droplet clusters (E), and after 30 s, these droplets were found only in the matte phase. From the micrographs in Fig. 2e–f, platinum metal preferred to form these metallic droplets with copper and iron enclosed within the matte phase, while gold, silver, and palladium dispersed more evenly into the matte phase. However, a few Au-, Ag-, and Pd-containing clusters were also detected in the matte phase. In the sample at 150 s, a larger high-platinum content area was found. At 300 s, a large droplet with a platinum core surrounded by a shell formed of copper, iron, and small amounts of PMs was observed in the sample. However, these larger PM clusters and droplets were avoided in the EPMA analyses, which was used to detect the average concentrations of PMs in the matte.

Fig. 2
figure 2

Micrographs of samples treated in air after different reaction times. a Matte, b fayalite slag, c silica, d ferric oxide, and e PMs-Cu–Fe droplet clusters

Figure 3 shows the SEM micrographs of samples after different contact times in an argon atmosphere. As shown in Fig. 3a, the matte (A) and fayalite slag (B) phases had already formed and separated from each other after 5 min. PM-rich clusters (D) were detected with all contact times, and they were randomly distributed in the matte domains. Figure 3b–d illustrates the small change in the structure of the samples with increasing contact time from 10 to 40 min. At 40 min of contact time (Fig. 3d), the PM-rich clusters had coalesced into larger droplets and were specifically found near the slag–matte interfaces. As is to be expected with longer contact times, the PMs have more time to migrate and coalesce to form bigger droplets within the matte.

Fig. 3
figure 3

Micrographs of samples treated in argon after different reaction times. a Matte, b fayalite slag, c silica, and d PMs-Cu–Fe droplet clusters

Behavior of Major Elements in Matte and Slag Phases

The concentrations of major elements in the matte (Cu–Fe–S) and slag (FeOx–SiO2) are shown in Figs. 4 (air atmosphere) and 5 (argon atmosphere) as a function of contact time. In air, the matte grade (wt% Cu) increased continuously, seen also as a decrease in its iron concentration. The sulfur concentration remained relatively stable with a minor decreasing trend as a function of increasing contact time. After 300 s, the reaction rate decreased, and the matte grade reached a value of approximately 50 wt% Cu at 600 s. In the slag phase shown in Fig. 4b, the silica concentration kept increasing until 60 s, after which the slag composition remained constant. Corresponding to the microstructures shown in Fig. 2, during the first 60 s, the primary silica crystals reacted and dissolved into the slag, and more silica was also dissolved from the crucible.

Fig. 4
figure 4

Chemical assays of matte (a) and slag (b) as a function of time in air-atmosphere experiments

Fig. 5
figure 5

Chemical assays of matte (a) and slag (b) as a function of time in argon-atmosphere experiments

In the argon atmosphere, the concentrations of major elements in matte and slag phases remained relatively constant throughout the time series. The matte grade stabilized at a considerably lower level than in air, due to the absence of free oxygen. The slag composition was almost the same as that in air. These results were similar to the results of previous experiments conducted in the same conditions but with different trace elements [40].

Behavior of Precious Metals in Matte and Slag Phases

The concentrations of the PMs Au, Ag, Pt, and Pd in matte as a function of time are shown in Fig. 6. In air atmosphere (Fig. 6a), the Au, Ag, and Pd concentrations in matte consistently increased with longer contact times. The platinum concentration began to decrease after 150 s, which can be explained by the formation of large platinum-rich droplets in the matte, shown in Fig. 3e–f. In the argon atmosphere (Fig. 6b), the PM concentrations remained relatively constant as a function of time. After the longest contact times, all the PM concentrations in matte were lower in argon than in air. The sequence of concentration levels in both atmospheres was Ag > Pd > Au > Pt.

Fig. 6
figure 6

Concentrations of PMs in matte phase as a function of time: a air atmosphere, b argon atmosphere

The concentrations of the PMs Au, Ag, Pt, and Pd in slag as a function of time are shown in Fig. 7. The slag chemistry plays an important role in the elimination of impurity elements and recovery of precious metals in copper smelting. In the FeOX–SiO2 slag used in this study, all precious metal concentrations initially decreased as a function of time and then stabilized at a constant level. The sequence of PM concentrations in the slag was Ag > Au > Pd > Pt in both air (Fig. 7a) and argon (Fig. 7b) atmospheres. However, the PM concentrations were higher in air than in the argon atmosphere after reaching stable concentration levels.

Fig. 7
figure 7

Concentrations of PMs in FeOx-SiO2 slag as a function of time: a air atmosphere, b argon atmosphere

The distribution reaction for precious metal Me between matte and slag can be described using Eq. (1), when metals are deported into the slag phase due to the oxidation reaction:

$$ {\text{Me}} + x/2\;{\text{O}}_{2} (g) = {\text{MO}}_{x}. $$
(1)

Equation (1) can be presented using an equilibrium constant that consists of the activities of the species and the prevailing oxygen partial pressure, shown in Eq. (2). The activities can be expressed with the concentration of the species and its activity coefficient, as in Eqs. (3) and (4).

$$ K = \frac{{\alpha {\text{MeO}}_{x} }}{{\alpha {\text{Me}} \cdot {\text{PO}}_{2}^{x/2} }}, $$
(2)
$$ a_{\text{Me}} = \gamma_{\text{Me}} {{\text{N}}}_{\text{Me}} = \gamma_{\text{Me}} [{\text{Me wt}}\%]\; / {\text{M}}_{\text{Me}} [n_{\text{T}} ], $$
(3)
$$ a_{{{\text{MeO}}x}} = \gamma_{{{\text{MeO}}x}} {\text{N}}_{{{\text{MeO}}x}} = \gamma_{{{\text{MeO}}x}} ({\text{wt}}\% \;{\text{Me}})\;/{\text{M}}_{\text{Me}} (n_{\text{T}} ), $$
(4)

where K is the equilibrium constant, and αMe and αMeOx represent the activities of metal Me and metal oxide MeOx, respectively. γMe is the activity coefficient of Me and γMeO is the activity coefficient of MeOx. NMe is the molar fraction of metal Me, nT is the total number of moles of monocationic constituents in 100 g of each phase, Mme is the atomic mass of Me, and wt% Me is the weight percentage in matte and slag.

In the equilibrium studies, the distribution coefficient of a PM Me between copper matte and slag is defined by Eq. (5), where [Me wt%] refers to the equilibrium concentration of Me in copper matte and (Me wt %) is the equilibrium concentration in slag in weight percentage [30].

$$ L^{\text{m/s}} ({\text{Me}}) = \left[ {{\text{Me}}\;{\text{wt}}\% } \right]/({\text{Me}}\;{\text{wt}}\% ). $$
(5)

Combining Eqs. (1)–(4), Eq. (5) can be expressed as follows:

$$ L^{\text{m/s}} ({\text{Me}}) = \frac{{\left[ {{\text{Me}}\;{\text{wt}}\% } \right]}}{{({\text{Me}}\;{\text{wt}}\% )}} = \frac{{\left[ {n_{\text{T}} } \right](\gamma {\text{MeO}}_{x} )}}{{K \cdot (n_{\text{T}} )\left[ {\gamma {\text{Me}}} \right]{\text{P}}_{{{\text{O}}_{2} }}^{x/2} }}. $$
(6)

As shown by Eq. (6), the distribution coefficient is a practical and independent thermodynamic parameter of the matte–slag system (only influenced by the temperature, oxygen potential, etc.), so the concentrations of the initial precious metals will not influence the distribution results. Even though these noble metal concentrations in this study are higher than those in real WPCBs, the distribution ratios between the matte and slag should be thermodynamically independent. In that case, the distribution coefficients between matte and slag are essential parameters for evaluating the recycling efficiencies of the PMs during copper smelting, by utilizing copper matte as a medium for their collection. In this kinetic study, the distribution coefficients values of PMs were determined at selected times by Eq. (7), where t refers to the contact time after which the sample was quenched and the reactions stopped [45].

$$ L^{\text{m/s}} ({\text{Me}})_{t} = (\left[ {{\text{Me}}\;{\text{wt}}\% } \right]_{t} /({\text{Me}}\;{\text{wt}}\% )_{t} ). $$
(7)

The logarithmic values of distribution coefficients Lm/s (Me) between matte and slag during the oxidation reactions in air are shown in Fig. 8. All PMs were found to deport strongly into the matte rather than into the slag, and with the increased contact time, their distribution coefficients followed a similar increasing tendency until 300 s, after which the values stabilized in the order of Pd > Pt > Au > Ag. At 20 s contact time, the average distribution coefficient values for Au, Ag, Pt, and Pd were 80, 70, 230, and 200, respectively, and at 300 s, these values surged to 600, 100, 1300, and 1200. During this time period, the matte grade also gradually increased. This also helped improve PM migration into the matte phase [36]. Tags a, b, c on the x-axis of Fig. 8 refer to the PM distribution results from equilibration studies between matte and iron-silicate slag where the temperature and matte grade were closest to the parameters in this work (a—Avarmaa et al. [30]; b—Chen et al. [34]; c—Shishin et al. [36]). The equilibration time in their studies was 3 h, 4 h, and 24 h, respectively. After 300–600 s, the distribution coefficients of the PMs obtained in this work agreed well with these recent equilibrium studies. Therefore, it seems that the equilibrium distribution coefficient values for PMs can be reached even in relatively short contact times.

Fig. 8
figure 8

Logarithmic distribution coefficients of gold, silver, platinum, and palladium between matte and FeOx–SiO2 slag as a function of time in air atmosphere

The distribution coefficients of PMs in argon are shown in Fig. 9. The values are lower than those in air, but they still follow the same order of Pd > Pt > Au > Ag. As shown in Fig. 5a, the matte grade (wt% Cu) was relatively constant in the argon atmosphere; however, the distribution coefficients continued to increase as a function of time. This means that sufficient time is needed for the PMs to migrate to the matte phase in low oxygen partial pressures. In air, the obtained distribution coefficients stabilized after 300 s, while the values in argon continued to increase slightly until the longest investigated contact time of 40 min. The average distribution coefficients values after 40 min for Au, Ag, Pt, and Pd were approximately 200, 90, 710, and 750, respectively, i.e., somewhat lower than the corresponding results in air.

Fig. 9
figure 9

Logarithmic distribution coefficients of gold, silver, platinum, and palladium between matte and FeOx–SiO2 slag as a function of time in argon atmosphere

Conclusions

No previous data exist on the kinetic behavior of precious metals (PMs) in copper flash smelting conditions. In this study, the behavior of Au, Ag, Pt, and Pd was investigated in laboratory-scale experiments at a typical smelting temperature of 1300 °C in both air and argon atmospheres. The samples were analyzed by SEM–EDS for visual and preliminary compositional information, and by EPMA (matte) and LA-ICP-MS (slag) for more accurate phase composition data.

All the PMs studied in this work strongly preferred to deport into the matte rather than the slag phase. The novelty of the present research is the experimental proof that the studied PMs migrate to the matte phase almost instantly when the molten matte and slag begin to form. The PMs approached their reported equilibrium distribution coefficients after only 300 s contact time in the air atmosphere. However, it should be noted that the experimental conditions, mainly the gas atmospheres, in the earlier equilibrium studies did not fully correspond to those utilized in this work, and industrial copper concentrate was used in this study. In the argon atmosphere, the distribution coefficient values slightly increased during the entire contact time of 40 min. The calculated distribution coefficients Lm/s (Me), based on experimentally measured element concentrations, followed the order of Pd > Pt > Au > Ag in both air and argon atmospheres.

The distribution coefficients of PMs in air were higher than the results in argon, and PM concentrations in matte increased with longer contact time and higher matte grade. The oxygen potential had an impact on the PM migration rates and distribution coefficients. Based on the results in argon, considerably longer times are needed for the complete transfer of PMs to the matte at low oxygen potentials.

After reaching stable levels, the PM concentrations in the slag were relatively low, at approximately 408 ppm (Ag), 27 ppm (Au), 17 ppm (Pd), and 9 ppm (Pt) in air, and 346 ppm (Ag), 42 ppm (Au), 24 ppm (Pd), and 3 ppm (Pt) in argon. These low concentrations of chemically dissolved PMs in the slag indicated that copper matte can perform as an excellent collector to recover precious metals. This suggests that the highest PM losses in industrial operations most likely arise from mechanical matte entrainment.

These new experimental results regarding the effect of time on PM distribution behavior can be used for updating databases related to secondary raw materials processing via copper smelting and also for CFD models to simulate the behavior of precious metals in smelting processes more precisely.