Introduction

Waste Electric and Electronic Equipment (WEEE)

The demand for cleaner energy sources to overcome the use of fossil fuels and to slowdown climate change due to human activities creates a favorable scenario for photovoltaic technologies, which is considered a promising technology [1]. Photovoltaic (PV) modules are devices that can convert sunlight into electricity without any other source of energy; they can be made of numerous semiconductors materials [2]. However, PV modules have a technical lifespan of 20–30 years and will become electronic waste (WEEE) in the next few years, since the commencement of broad PV installation occurred in the 1990s [3]. End-of-life modules are expected to reach 5.5–6 million tons by the 2050s [4]. Therefore, it is essential to develop recycling technologies to reduce the amount of this waste, taking into consideration the dimensions that it will acquire in the next years. It is also necessary to evaluate the risks of disposing the WEEE, which can generate environmental impacts [5], given that the waste contains hazardous materials that require special handling and can cause possible detrimental effects on human health. Moreover, recycling can recover reusable components and base materials such as copper [6], precious metals [7, 8], and reserves of carbon [9].

Photovoltaic Modules

The composition of PV modules varies according to the technology used. The modules are basically a layer of a semiconductor material placed between tempered glass and/or glass or a polymer as back sheet. Lead, chromium, cadmium, and nickel are among the hazardous metals usually used [10]. Currently, silicon (Si)-based PV modules, such as single-crystalline Si (sc-Si), multicrystalline Si (mc-Si), and hydrogenated amorphous Si (a-Si) PV modules, are playing a vital role in the PV market [11]. A general quantification of the materials present in silicon-based PV modules is shown in Table 1.

Table 1 Typical composition of materials in a silicon-based photovoltaic module. Source: [5, 8, 12, 13]
Full size table

Modules are encapsulated with various materials to protect the cells and the electrical connectors from the environment [14]—the most common being ethylene–vinyl acetate (EVA). The removal of these encapsulating materials is an important step in the recycling of PV modules [15] (Fig. 1).

Fig. 1
figure 1

Adapted from [13] (Color figure online)

Arrangement of components in a typical silicon-based solar module.

Full size image

Thermal and hydrometallurgical processes are prevalent in most of the PV recycling methods, and the encapsulating material can be removed with the aid of thermal decomposition and nitric acid [16]. Jung et al. [17] used a thermal treatment to decompose the EVA layer and to separate the different layers of solar panels. Doi et al. [18] used various organic solvents aiming to dissolve the EVA layer. Radziemska et al. [19] used a thermal process to decompose the EVA, followed by a chemical treatment for the solar cell, which they refer to as a second primary step in PV modules’ recycling. Kang et al. [20] employed a mixture of acid solvents in the process of etching Si solar panels. Zhang and Xu [21] used pyrolysis in a nitrogen atmosphere to remove the EVA layer, and recycle glass and gallium from thin-film solar modules. However, as shown in earlier studies [5], the use of mechanical processes, such as shredding/milling, and sieving, may assist in the recycling of PV panels and reduce the cost of recycling, given that these processes are able to concentrate metals in different fractions according to particle size. Moreover, mechanical processes may be used prior to the thermal and hydroprocesses as a pretreatment step that aids the following recycling steps and upgrades material content [22, 23].

Electrostatic Separation

Electrostatic separation is frequently used in the separation of equipment containing copper, aluminum, and insulating materials, which is also the case of WEEE. It represents a modern and useful technology for the recycling of industrial waste materials [24, 25]. Electrostatic separation sorts substances with different electrical conductivities, which are typically charged before exposure to electrostatic and gravitational forces. In a roll-type separator (Fig. 2), the materials go through a grounded roller and are subjected to electric charge ionization from an electrode; conductive particles discharge due to physical contact with the roller, while nonconductive particles are attracted to the roller due to Coulomb forces. Thus, the particles are eventually separated due to the differences in conductivity and electrostatic properties [26].

Fig. 2
figure 2

Adapted from [27] (Color figure online)

Diagram illustrating the electrostatic separation principle in a roll-type electrostatic separator.

Full size image

Richard et al. [28] applied electrostatic separation to sort out granular metals and plastics from electric wire waste. Recently, the same authors evaluated the use of three different configuration of roll-type electrostatic separation in order to optimize the segregation of PVC (Polyvinyl chloride), aluminum and copper from electronic waste. The three configurations included the use of an (i) elliptical static electrode with corona effect, an (ii) s-shaped plate with plastic trap and (iii) an s-shaped plate with plastic trap and corona effect. They have found that the corona electrode (or ionizing electrode) is necessary for the separation of PVC and copper [29]. Veit et al. [30] also used this method to recover metals from circuit boards and concluded that the use of electrostatic separation was efficient in obtaining high concentrated fractions of metals, in particular, he was able to concentrate 50% of copper, 25% of tin, and 7% of lead. This method has been further improved by adding a second roll and creating a two-step process [31]. Electrostatic separation is considered an efficient low cost mechanical process that requires little energy in comparison to thermal processes [32] and does not generate byproduct effluents, unlike hydrometallurgical processes [33].

As demonstrated, thermal and hydrometallurgical methods are largely implemented in the recycling of waste PV and the combination of both is currently presenting optimal outputs [15, 19, 20]. The use of electrostatic separation in PV recycling has not yet been studied despite the great potential this method has in assisting the recycling through the segregation of different materials present in the modules. Moreover, the use of electrostatic separation could potentially reduce the cost of PV recycling, which is one of the biggest barriers keeping the recycling of WEEE from expanding [34]. Therefore, in this study, the use of electrostatic separation is assessed in order to segregate the main materials of PV panels.

Materials and Methods

The objective of this study is to evaluate the use of electrostatic separation technique to segregate some of the main materials present in silicon-based photovoltaic modules: silver, copper, silicon, glass, and polymers from the back sheet and encapsulating material. The schematic diagram (Fig. 3) outlines the principle of investigations performed in this study.

Fig. 3
figure 3

Schematic diagram illustrating the procedures used in this study.

Full size image

The experiments were performed with five different crystalline silicon modules (c-Si modules). The aluminum frames were manually removed from all modules. The modules were then milled with a SRB 2305 knife mill (Rone, São Paulo, Brasil) with the experimental parameters based on previous works [8]. The milling was repeated four times, the first two using a screen with 4-mm openings and the second two using a screen with 2-mm openings; the output powder weighed approximately 5 kg.

The output powder was quartered and separated in 300-g samples. These samples were named F1, F2, F3, F4, F5, and F6. In the first part of the experiment, empirical studies were performed by attempting to separate the different materials from the sample by varying several parameters at random starting points. The equipment used for all the electrostatic separations was a MMPM-618C (Eriez, Erie, USA) high-tension roll separator (Fig. 4).

Fig. 4
figure 4

Adapted from [35]

Schematic illustration of the electrostatic separator equipment setup.

Full size image

Visual inspection was adopted to determine the performance of a given parameter. A sample considered great would have the majority of polymers on output C, the majority of silicon on output B, and the majority of glass and metals in output A. The changeable parameters are listed in Table 2, which also presents the parameters that were fixed (pinned down) after repetitive attempts of separating the materials and observing the results.

Table 2 Relationship between parameters and variation for the electrostatic separator used in this study
Full size table

It is important to notice that there are only three collection pans (Fig. 4—item 9): conductor, middling, and nonconductor: A, B, and C, respectively. Samples F1, F2, and F3 were used in the empirical study.

The results obtained from the empirical study led to the restriction of the degrees of freedom related to the variables in the separation process. As a result, it was chosen to vary the electric potential difference and the rotation speed and keep the other parameters fixed. The electric potential difference (voltage hereafter) is given in volts and is the difference in electric potentials between wired electrodes (lifting and ionizing) and the roll (grounded). The selected voltages were 24 and 28 kV; the selected rotation speeds were 50, 65, and 80 RPM.

Each 300-g group (F4, F5, and F6) was quartered and separated in six 50-g samples—one for each combination of parameters. 50-g samples were chosen based on previous WEEE electrostatic separation studies [29]. The end result after electrostatic separation is a total of 18 samples (six combinations of parameters with output fractions A, B and C), which were replicated three times (F4, F5 and F6) for further statistical analyses. To evaluate the efficiency of each combination of voltage and rotation speed, a series of experiments were performed: First, each sample was leached in nitric acid (65% concentration) solution using a 10:1 liquid–solid ratio, under magnetic stirring, at room temperature. The solid part from the leaching was filtered, rinsed, dried, and put aside. The solutions containing the leached metals were analyzed by inductively coupled plasma optical emission spectroscopy (ICP-OES) to determine the amount of silver and copper in each sample. The equipment used was a 5110 ICP-OES (Agilent Technologies, California, USA). The filter residue was then weighed and placed in a crucible, which was placed inside a furnace in order to eliminate the polymers in each sample. The dwell temperature was set to 500 °C based on previous works [15]. The heating ramp was 15 °C/min and the dwell time was 5 h per sample. The samples were weighed again and the mass difference was assumed to be the mass of the polymeric fraction contained in each sample.

Finally, the samples were milled in an alumina mortar and sieved through a mesh 400 sieve (37 µm). Each sample was weighted and analyzed by X-ray diffraction (XRD) using a Siemens (Bruker AXS—Germany) D-5000 diffractometer with Cu Kα1,2 radiation (1.54178 Ǻ) and equipped with curved graphite monochromator in the secondary beam. The data were collected in the Bragg–Brentano (θ/θ) geometry between 10 and 75° (2θ) in 0.05° steps, at 1 s/step using a 1° divergence, and anti-scattering slits, and a detector slit of 0.6 mm.

After initial analysis, an internal standard of hexagonal (P63 mc) ZnO (99.9%) was added to the sample. The internal standard was added so that it would represent 10% in weight of the sample. For Rietveld quantitative phase analysis (RQPA), the same angular interval was analyzed in 0.02° steps, at 15 s/step using a 1° divergence and anti-scattering slits, and a detector slit of 0.2 mm. The X-ray tube was operated at 40 kV and 40 mA [36]. The Si and ZnO contents were determined by RQPA using a free-code software MAUD (Materials Analysis Using Diffraction [37]. After RQPA, the amorphous content was determined using Eq. (1).

$$ A = \frac{{1 - \frac{Ws}{Rs}}}{100 - Ws} \times 10^{4} \%, $$
(1)

where (%) is the weighted concentration of the internal standard and (%) is the Rietveld-analyzed concentration of the internal standard [36, 38]. The amorphous phase is assumed to be glass, while the crystalline phase is assumed to be silicon.

The optimal parameter to separate materials from waste PV modules is given by the analysis of the distribution of material in A, B and C (Fig. 4). In the interest of evaluating the effectiveness of electrostatic separation and the optimal parameter combination, a variance analysis was performed, and p-values were generated to determine significance for a confidence level of 95% (α = 0.05). The variance analysis was performed for the silver, copper, and polymer separation. Because of the cost of XRD and RQPA analysis, it was not possible to perform a variance analysis for glass and silicon as this requires replicated measurements.

Results

In the first part of the experiment, an empirical study was carried out in order to restrain the degrees of freedom related to the electrostatic separation. Initially, all parameters varied, and the output of each combination was analyzed by visual inspection. As a result, the fixed parameters were the two splitter angles, the vibratory feeder speed, and the position of the ionizing electrode, the brush and the lifting electrode. The visual inspection revealed a clear separation among the main materials present in waste PV modules. As can be seen in Fig. 5, the nonconductor fraction (C) contains mostly polymers (white particles), the middling fraction (B) contains mostly silicon (gray and blackish particles) and the conductor fraction (A) mostly contains glass. Although glass is an electrical insulating material [39], its particles fall into the first pan along with the metals (conductive fraction: A). This may be due to particles being too heavy (speed gained during rotation is superior to the electrostatic forces acting) and/or due to the influence of metallic particles, which can attach to the glass particles through the encapsulating material.

Fig. 5
figure 5

Visual analysis comparing the three different outputs from the electrostatic separation during the empirical study. Conductor (a), middling (b) and nonconductor (c) (Color figure online)

Full size image

The following experiments determined which would be the values for voltage and rotation speed used in this study. Rotation speed varied from 15 to 85 RPM and the voltage from 10 to 30 kV (voltages higher than 30 kV were not tested because electrical discharges start taking place at this point). Each output was evaluated by visual inspection and classified as poor, fair, good, and great. The results are displayed in Table 3.

Table 3 Qualitative results from empirical study to determine key values of voltage and rotation speed
Full size table

Table 3 shows two results classified as “great”: 28 kV with 30 RPM and 28 kV with 55 RPM. The table also shows that the best results are obtained with a voltage of 24 kV or higher and a rotation speed of 30 or higher. Thus, two voltage values (24 and 28 kV) and three rotation speed values (50, 65, and 80 RPM) were chosen to quantitatively evaluate the separation of materials from waste PV modules.

The distributions of silver and copper for each parameter combination obtained by ICP-OES are displayed in Figs. 6 and 7, respectively.

Fig. 6
figure 6

Average distribution of silver in each fraction for a given combination of chosen parameters (p-values for A, B, and C distribution are 0.662, 0.789, and 0.856, respectively). There were three replicates per parameter

Full size image
Fig. 7
figure 7

Average distribution of copper in each fraction for a given combination of chosen parameters (p-values for A, B, and C distribution are 0.606, 0.552, and 0.209, respectively). There were three replicates per parameter

Full size image

Both Figs. 6 and 7 indicate that metals tend to concentrate in the first fraction (A), followed by the second and third (B and C). It should be highlighted that the metal concentration in C is around 5% on average. Therefore, the electrostatic separation concentrates about 95% of silver and copper in fractions A and B. The copper did not follow this distribution for the combination 24 kV-65 RPM nor 28 kV-50 RPM. While 100% of separation was not achieved at this point, these results indicate that electrostatic separation is able to separate the metal content from photovoltaic waste, but also show that there is no significant difference between the treatments (combinations of parameters) tested in this study in regard to metal separation. Figure 8 supports the statement that there is a significant difference between the three fraction outputs. The detailed distributions of silver and copper are available in Supplementary Tables 1 and 2, respectively.

Fig. 8
figure 8

Electrostatic separation’s influence on silver and copper content distributions (p value for silver and copper distribution is < 0.001)

Full size image

The polymer distribution was measured by gravimetric analysis. The mass differences before and after combustion are assumed to be the mass of polymers present in a certain sample. The variance analysis for the polymer distribution is shown in Fig. 9, and the average polymer distributions in pans A, B, and C are displayed in Fig. 10. Figure 9 indicates that the combinations of parameters tested for the electrostatic separation do not differ significantly among each other.

Fig. 9
figure 9

Average distribution of polymer in each fraction for a given combination of chosen parameters (p-values for A, B, and C distributions are 0.834, 0.051, and 0.933, respectively). There were three replicates per parameter

Full size image
Fig. 10
figure 10

Electrostatic separation’s influence on polymer, glass, and silicon content distributions (p-values of polymer, glass, and silicon distribution are 0.146, < 0.001, and < 0.001, respectively)

Full size image

Moreover, Fig. 10 shows that it was not possible to segregate the polymers in any of the separation pans. It has been reported that photovoltaic panels may have polyvinyl chloride (PVC) in its substrate [15]. Richard et al. [29] have stated that the use of a reverse s-shaped electrode assists in the separation of PVC by this method. The presence of PVC in waste PV may have influenced the separation process. The detailed distribution of polymers is available in supplementary Table 3.

The silicon and glass distributions were measured by XRD with Rietveld’s refinement. The amorphous phase was assumed to be glass, while the crystalline phase was assumed to be silicon. The distributions of glass and silicon in the three pans for all tested parameters are displayed Fig. 10, while the uncertainties associated with the Rietveld’s refinement for the analysis of this study are presented in Supplementary Table 6.

Figure 10 suggests that glass tends to concentrate in the first fraction (conductor: A) as predicted from the empirical study. Moreover, for a significance of p < 0.001, fraction sA and B concentrate approximately 95% of the glass. A similar behavior is found with the silicon, given it concentrated mainly in A, followed by B and C. The standard deviations suggest that it is possible to concentrate silicon in the second fraction (middling) with the appropriate combination of parameters. The influences of each combination of parameters on the material distribution for glass and silicon are displayed in Figs. 11 and 12.

Fig. 11
figure 11

Distribution of glass in each fraction for a given combination of chosen parameters

Full size image
Fig. 12
figure 12

Distribution of silicon in each fraction for a given combination of chosen parameters

Full size image

The detailed distributions of glass and silicon are available in Supplementary Tables 4 and 5, respectively.

Figure 11 shows that for all the combination of parameters tested, glass mostly stays in the first fraction. This is probably due to its weight, i.e., most of the glass particles have a mass such that the gravitational force has a greater influence than the electrostatic force from the process. The smaller glass particles are the ones found in the last fraction (nonconductor: C). As for the silicon, Fig. 12 shows that for different combination of parameters (particularly for 24.50 and 28.80), it is possible to concentrate this material in the second fraction (middling: B), but for the combinations of this study, it tends to remain on the first fraction (conductor: A). Silicon particles in PV waste are distributed in fine particles that can remain attached to particles of other material. This may affect the distribution of silicon particles during the electrostatic separation. The assessment of the different combination of parameters is limited given that a single analysis for each combination of parameters was made for glass and silicon. Therefore, a statistical analysis is not possible for these given parameters. Moreover, the mass distributions in the three pans as a function of the tested parameters are shown in Supplementary Table 7.

Conclusions

The key conclusions from this study are as follows:

  • Electrostatic separation is able to segregate the metallic fraction of waste photovoltaic panels. Metals tend to concentrate in the first separation fraction (conductor). About 95% of the metals in waste silicon photovoltaic modules concentrate in output pans A and B (conductor and middling, respectively) combined.

  • The studied combination of parameters have no statistical differences among each other for the separation of metals. The influence of the parameters was not significant for either silver or copper.

  • Electrostatic separation was not able to concentrate the polymers present in photovoltaic panels. The presence of PVC as one of the polymers present in photovoltaic panels may have contributed to the failure of the electrostatic separation method [15, 29].

  • The studied combination of parameters have no statistical difference among each other for the separation of polymers. The influence of the parameters was not significant.

  • The glass present in waste PV tends to concentrate in the conductive fraction, followed by the middling and nonconductor (p < 0.001).

  • For the majority of tested combination of parameters, silicon tends to concentrate in the first output fraction (conductor). However, it was shown that it is possible to concentrate silicon in the second output fraction (middling) by varying the parameters.

  • Among the tested combination of parameters (24 kV.50RPM; 24 kV.65RPM; 24 kV.80RPM; 28 kV.50RPM; 28 kV.65RPM; 28 kV.80RPM), it is not possible to determine the optimal combination of parameters for separating metals and polymers from photovoltaic modules at this stage.

  • Electrostatic separation has an influence in most of the materials present in waste silicon photovoltaics. This process may assist in the recycling of waste PV.

This study can be improved by means of samples with higher masses for each parameter combination (e.g., 300 g), by evaluating the separation at a lower rotation speed (e.g., 20, 30, and 40 RPM); by using a reversed s-shaped plate so that materials such as PVC are more influenced by the separation (as stated by Richard et al. [29]); and by heating the samples before the separation to reduce the humidity. The improvement of this study is encouraged by the results observed. Electrostatic separation can assist in the recycling of waste photovoltaics, but the parameters for an optimal separation are still uncertain.