Epoxy-Silicon Composite Materials from End-of-Life Photovoltaic Panels

The prospect of using recovered solar cells from end-of-life (EoL) photovoltaic panels (PVPs) to produce composite materials with dielectric properties was studied. The main goal of this research was to reduce the waste originating from EoL PVPs by reusing the semiconductor, thus rendering solar energy an even greener energy source. Solar cells were recovered from EoL PVPs through thermal treatment to remove polymer sheets and screening to separate the solar cells from glass and electrodes. Composite materials were manufactured by reinforcing two different epoxy resins, Araldite LY556 and Resoltech 1050, with varying concentrations of ground solar cells (0–10% w/w). The mechanical and dielectric properties of the composite materials were examined with bending and shearing tests and Broadband Dielectric Spectroscopy (BDS), respectively. The responses from the two different resin matrices were compared. It was found that the produced composite material resulting from Resoltech resin reinforced with solar cells recovered from EoL PVPs had better mechanical and dielectric properties. BDS characterization of the composite materials indicated that the solar cells can be used to enhance the energy storage capacity of the polymeric matrix and thus may be potentially used in the manufacturing of capacitors.


Statement of Novelty
A significant increase in waste originating from end-of-life photovoltaic panels is expected in the upcoming decades, as the world is turning to renewable energy sources.Therefore, a sustainable management plan for recovering and reusing critical materials in photovoltaic panels becomes imperative.Researchers, so far, have focused mainly on material recovery.This study approached the recyclability issue by focusing on utilization of crystalline silicon contained recovered from first generation solar cells and the possibility of

Introduction
In the recent years, energy production through renewable sources has become increasingly competitive in terms of cost and imperative in terms of carbon footprint.One of the dominant renewable energy sources is solar radiation which may be harvested through solar photovoltaic panels (PVPs).By the end of 2015, installed solar PVPs reached a capacity of 200 gigawatts (GW) and it has been estimated to increase to 4500 GW globally by 2050.Since photovoltaic panels have a life span of about 25-30 years, it is expected that in the next decade thousands of metric tons of installed photovoltaic panels will be withdrawn from existing parks as waste.Their potential disposal in landfills will lead to loss of critical and valuable materials that can potentially be reused [1][2][3].
Characterization of this upcoming type of Waste from Electrical and Electronic Equipment (WEEE) has concerned researchers globally, who investigate possible environmental and economic impacts of solar PVPs after their end of life (EoL) [3][4][5].Researchers have focused on material recovery from 1st generation EoL solar PVPs that use monocrystalline and polycrystalline silicon as semiconductor since the beginning of the century.Physical and/or thermal treatment processes have surpassed chemical methods, as avoiding the use of organic solvents to dissolve the polymer sheets in the PVPs is more sustainable, both economically and environmentally [2,6].In a previous work, an integrated hydrometallurgical process for the recovery of pure crystalline Si and Ag from end of EoL Si PVPs has been proposed [7].
The high temperature required for the manufacturing of crystalline silicon solar cells renders it a valuable material to be recovered and reused, despite its vast availability in nature.Research on material recovery and recycling from EoL PVPs is extensive [4][5][6][7][8][9], whereas reports on alternative utilization of recovered materials are limited [10].In a previous work, a potential reuse pathway of first generation PVP waste as aggregate in Portland cement was studied [11].However gas formation in the cement paste led to decreased performance for the case of silicon PVPs.Researchers have used silicon or silica based materials to enhance the dielectric properties of polymeric matrices such as epoxy resins [12][13][14][15].So, in the present study, an alternative valorization of silicon's semiconductor properties is evaluated, by reusing the recovered silicon from end of life PVPs as an additive in a polymer matrix, aiming to produce a material for energy storage applications while providing a sustainable reuse pathway for EoL crystalline silicon PVPs.
Polymers and polymer matrix composites are electric insulators and have dielectric characteristics.With the application of an external field, their electrical response is primarily related to relaxation phenomena, which describe the delay of a physical system to follow an externally applied excitation.The observed relaxation processes are strongly influenced by the positioning of the polymer chains and the existence of polar groups.External factors such as additives also affect dielectric properties.It has been proven that the concentration of conductive inclusions is a critical parameter governing the electrical behaviour of composite materials [16,17].
Bisphenol resins are commonly used as a matrix for composite materials and have been greatly studied for the relaxation phenomena that appear when exposed to external electrical fields [18][19][20].The separation of the positive from the negative charges throughout the volume of the material creates an overall polarization in it.The latter can be distinguished into deformation polarization from the intramolecular displacements, orientation polarization from the permanent dipoles in the material, and interfacial polarization [16].As the systems in this study are heterogeneous, the investigation of their electrical properties and in particular the conductivity difference between the matrix and the filler particle is a significant parameter.When the filler particle is a semiconductor, the effects of the accumulation of electrical loads in the interfaces can be described by the Maxwell-Wagner-Sillars (MWS) effect [16].In addition, the response of a material to an alternating field depends on both the frequency of the field and the temperature of the environment.At higher frequencies, only some of the polarization phenomena have enough time to adapt to the changes and thus the overall polarization is reduced.Heat can either compensate for this loss by providing energy to the dipoles or it can cause further decreases in polarization, as the thermal vibrations of the atoms neutralize their tendency to align with the field [21].Therefore, it is of interest to investigate the response of a used semiconductor as a filler particle in these resin systems, such as recovered silicon solar cells from EoL PVPs.
Thus, the scope of this work is to evaluate the potential of reusing solar cells recovered from first generation EoL PVPs as reinforcement in epoxy resin systems in order to enhance their dielectric properties, so that they may be used in energy storage applications.Two commercial epoxy resin systems were used and compared in their mechanical and dielectric property responses, in order to determine the more compatible matrix.

Solar Cell Recovery from EoL PVPs
First generation EoL PVPs that use crystalline silicon as semiconductor were provided by POLYECO SA.The panels were delivered to the lab after being uninstalled from the field.Typical first generation PVPs consist of glass, polymeric sheets (adhesive), solar cells, Tedlar back-cover and electrodes.Silicon solar cell recovery process is presented schematically in Fig. 1.In order to recover the crystalline silicon solar cells, a thermal treatment was employed.Panels were manually cut into 40 × 30 mm pieces and subsequently heated to 550 °C for 30 min in a furnace in order to remove the polymer sheets.A mixture of glass, solar cells, electrodes, and ash was recovered and separated in a trommel screen.The recovered solar cells were then ground and their particle size was measured and reported.

Epoxy-Silicon Composite Synthesis
Two thermoset epoxy resin systems were used and compared in this study.The firsts consists of diglycidyl ether of bisphenol A (DGEBA) from Resoltech Advanced Resin Technologies (Resoltech).Components A (Resoltech 1050) and B (Resoltech 1054S) were mixed in a 100: 35 weight ratio, respectively, and the appropriate amount of ground solar cell (particle size < 710 μm) was added to achieve concentrations 0, 2.5, 5, 7.5, 10% w/w.After thorough mixing, air bubbles from the mixture were removed via a DS-26S Vacuum Degassing System 26L (Easy Composites Ltd) and the final product was purred into the matrices to cure in room temperature for a week.Similarly, for the second resin system, Araldite System (Araldite), components A (Araldite LY 556), B (Aradur 917, Huntsman), and C (Accelerator DY 070, Huntsman) were mixed in a 100:90:0.5weight ratio and the corresponding amount of solar cell was added.After completing the same degassing procedure as for the Resoltech samples, the matrices were placed in an oven at 80 °C for 4 h and then at 120 °C for another 4 h in order to complete the curing and post-curing processes.

Composite Material Characterization
The mechanical strength of the composite materials was tested according to ASTM D790 [22]standards for bending and ASTM D2344 [23] standards for shear loads.A motorized test stand: SAUTER TVO 500N500S (Kern Balingen) with load cell of 500N and a displacement speed of 5 mm/ min was used.The composite materials were also dielectrically characterized using Broadband Dielectric Spectroscopy (BDS) in a range of temperatures (30-120 °C) and frequencies of the external field (10 -1 -10 7 Hz).Measurements were conducted according to the ASTM D150 specifications [24] via a Frequency Response Analyzer (Novocontrol Technologies).Temperature was varied by means of Novotherm System (Novocontrol Technologies).Finally, the samples were analyzed using Scanning Electron Microscopy (SEM) images in a Hitachi TM3030 plus microscope in order to Fig. 1 Material recovery process from 1st generation crystalline silicon panels [11] examine the surface morphology at the breaking points and the dispersion of the solar cell in the polymer matrix.

Results and Discussion
Particle size distribution of the ground recovered solar cell was measured and reported in Fig. 2. The results showed that the majority of the sample mass (84%) has a particle size ranging from 63 to 500 μm, as 95.7% passes through a 500 μm sieve while 11.7% passes through a 63 μm.This information serves as an indication, as this distribution in particle size might affect the properties of the composite materials and their dispersion in the matrices.

Mechanical Strength
The mean average of bending and shear loads with the corresponding standard deviations are presented in Table 1.Overall, the presence of the solar cell reinforcement does not effectively change the bending force required for breakage, which varies from 57 up to 95 MPa on Resoltech specimens and from 32.7 to 62.6 MPa for Araldite samples.However, the high standard deviation of the measurements is attributed to the random distribution of silicon granules within the composite materials.It is considered that no shear stresses are applied when performing the bending tests and the position of the neutral axis is a straight line when the beam is uncharged.Since the cracks follow the positioning of the cross-sections, it is most likely that the materials have a curved neutral axis, which follows the longitudinal distribution of the additive.Therefore, the calculation of the bending strength, based on the Euler-Bernoulli [21] trends, can be considered as an approximation.Thus, it can be concluded that the addition of solar cell powder does not bring about any significant changes in the bending strength of the epoxy resin matrices.
Comparatively, the Resoltech resin matrix displayed better mechanical performance than the Araldite resin, as it can be observed in Figs. 3 and 4. The resins displayed optimal bending strength on the average when reinforced with 2.5% and 10% w/w silicon, respectively, and optimal shearing strength for 5% silicon for both resins.However, the reinforcement caused approximately a 30% decrease in   resistance to shear loads in Resoltech and 50% on Araldite samples.This behavior is expected, as the presence of the granules in the polymeric matrix can cause force accumulation points, lowering the breakage force required for them to fail.In the case of shearing strength, the responses for Resoltech samples containing the solar cell reinforcement ranged from 7 to 12.2 MPa whereas for Araldite from 4 to 5.8 MPa, achieving approximately double the force required for breakage on the average.

Scanning Electron Microscopy Characterization
The aim of studying the samples with SEM was to obtain a view of the dispersion of the silicon solar cell powder in the polymer matrix and the surface topography.Homogeneous dispersion of the additive is a requirement for any electrical application.However, it was expected that the bottom sides of the matrices would show a higher concentration of silicon.The latter would be the result of the gravitational downward motion of the dust during curing.As shown in Figs.5a and 6a, the samples contain a small percentage of trapped air indicated by their characteristic spherical shape and color change.The silicon dust particles seem to have a homogeneous dispersion in the bulk of the matrix and an increasing concentration on the bottom surfaces.Moreover, the samples with the highest concentration of silicon (Figs.5b, 6b) indicate that the granules do not form a conductive pathway through the matrix, but are almost exclusively surrounded by the insulating material of the polymer matrix.According to Fig. 5, the matrix of the composite when breaking, creates multiple levels that follow the positioning of the granules.The latter, created load accumulation points where the material failed, and the crack was transmitted to the rest of the sample, validating the hypothesis that was discussed in Sect.3.1 (Fig. 4) regarding the cause of the high percentile decrease of shearing load breaking force in both polymeric matrices tested.

Broadband Dielectric Spectroscopy Characterization
Through BDS analysis, the dependence of dielectric variables, ε' = f(log(f)), ε'' = f(log(f)), Μ' = f(log(f)), Μ'' = f(log(f)), σ ac = f(log(f)) and tan(δ) = f(log(f)) on the frequency of the externally applied field, the temperature, and the concentration of the reinforcing phase is examined.From these measurements, the relaxation processes in a material become evident and the dielectric composite materials can be characterized based on their electrical behavior.All samples underwent isothermal scans on a frequency range from 0.1 to 10 7 Hz for a temperature range of 30-120 °C with a step of 5 °C.First, the alternating current conductivity (σ a.c. ) demonstrated by the composite materials, ranges from 10 -10 to 10 -5 S cm −1 leading to the conclusion that they can be classified as insulators throughout the range of temperatures and frequencies.Comparatively, Araldite is found to be a stronger insulator than Resoltech.[25,26].
Comparative graphs of the real part of the dielectric permittivity (ε ) and the imaginary part of the electric modulus (M") were prepared for the five specimens of each kind of composite material (0, 2.5, 5, 7.5% and 10% w / w in silicon) at a constant frequency value and are presented in Figs. 6  and 7, respectively.The frequencies were selected for each system depending on the polymeric matrix so that the relaxations would appear clearly.The selected frequency for the Araldite composite is 10 -1 Hz, while for Resoltech 10 2 Hz.
The real part of the dielectric permittivity (ε') of all composite materials maximized at low frequencies of the externally imposed alternating field (Fig. 7).The fact that ε' receives high values indicates that the existing and the induced dipoles are capable of reorienting along with the alternating field.Composite specimens exhibit higher values of ε' than the unreinforced matrices, because of the presence of filler, which is characterized by higher values of permittivity.As expected, permittivity increases with filler content systematically.At the low frequency edge, dielectric spectra might be affected by the conductivity of the composite materials and the phenomenon of electrode polarization.However, the recorded values at low frequencies and high temperatures do not imply a significant influence of these terms.The energy given in the form of heat to the composite samples enabled the dipoles to follow the field alternations for a wider frequency range than in ambient conditions.This trend appeared until the highest tested temperature of 120 °C.Interestingly, the values of ε' in the Resoltech resin specimens increased stepwise indicating three relaxation processes.Ascending relaxation time, these processes are: β-relaxation due to the reorientation of polar side groups, α-relaxation assigned to the glass to rubber transition of the polymer matrix, and interfacial polarization (also known as Maxwell-Wagner-Sillars effect-MWS) because of the accumulation of unbounded charges at the systems' interface.In Araldite samples, the values of ε' increase at high temperatures in a rather abrupt manner, masking thus the occurrence of relaxations, Fig. 7a (left).Both resins eventually acquire a constant value of ε' for all temperatures, which is related mainly to the synergetic motion of parts of the macromolecular chains and at a lower level to the rearrangement of the polar side groups [27][28][29].On the comparative M" graph for Araldite, the peak located at − 100 °C represents α-relaxation, which appears when the composite material receives sufficient energy in the form of heat to switch from the vitreous to the elastoplastic state.Although the peaks for the composites are completely formed, the relative peak for the neat resin is recorded only as a tendency and it should lie at a higher temperature.Shifting of the peak of α-mode to lower temperature with the addition of filler, indicates a lowering of glass transition temperature and weak interactions between polymer and inclusions.The slower phenomenon of interfacial polarization is expected at even higher temperatures, out of the experimental window.Impurities in the matrices create interfaces for load accumulation in every sample.Weak β-relaxation cannot be distinguished.It should be located at high frequencies and low temperature regions and it expresses the vibrations of the polar lateral groups of the polymer chain.For the case of Araldite, these lateral groups are the oxygen atoms (= O) that can vibrate.It appears that the addition of silicon in this resin significantly alters its dielectric properties.One possible explanation is the cause of differences in the polymerization mechanism of the resin.The Resoltech sample has more broadened peaks for the relaxation phenomena seen at approximately the same temperatures.The lateral hydroxyl groups found in Resoltech contribute to the appearance of β-relaxation.The distinct difference between the resins is the behavior of the blank Resoltech samples which follow the same pattern as the composites, concluding that silicon does not cause significant changes in the dielectric properties of the Resoltech resin [28,29] (Fig. 8).
The energy storage capacity, expressed via the energy density (< u > J/m 3 ) and relative energy density (< u > / < u > 0% w/w Si ) of all the samples was calculated and cross-compared to evaluate the composite's performance under different circumstances and is presented in Figs. 9 and 10 respectively.Energy density, at low strength fields, is proportional to the real part of dielectric permittivity and thus follows its variations with temperature, frequency, and filler content.Energy density rises fast with the applied field and its upper limit is set by the dielectric breakdown strength of the material.It should be noted that the only material property that affects energy density is ε' [30,31].The first noticeable characteristic of the samples is the dependence of the energy density on the temperature (Fig. 10).At higher temperatures, the samples show an enhanced ability to store energy, displaying an exponential rise after 80 °C especially in the case of Araldite resin matrix.This behavior is in accordance with the theoretical background of dielectric properties which states that ε' can be enhanced if energy is externally offered to the molecular and induced dipoles via a source of heat.However, a striking difference is noticed in the behavior of the samples when the temperature is maintained constant and low, while the frequency changes (Fig. 9).At low temperatures, e.g., 30 °C, the u capacity remains almost unchanged for all frequencies.At high temperatures and low frequencies, interfacial polarization increases significantly, leading to increased values of u, which diminish rapidly with frequency, since the permanent and induced dipoles fail to follow the fast alternation of the field [30].Thus, at low frequencies, e.g., 0.1 Hz, the u capacity is greater than at higher frequencies, since the molecular and induced dipoles have adequate time to follow the variation of the field and to be reoriented in accordance with it.At low frequencies, all samples show increased u capacity in comparison with the resin sample without silicon in it.Recovered silicon solar cell as a filler particle enhanced the dielectric properties of both tested polymeric matrices, achieving a maximum of over 3.5 times increased energy storage capacity for 10% w/w Araldite and almost 3 times increased for 10% w/w Resoltech specimens.In the case of Araldite matrix, however, as displayed in Fig. 10a, the rise in energy storage capacity is not significant for temperatures below 90 °C and the u value becomes maximum at a temperature of 120 °C.In contrast, Resoltech specimens displayed at least 1.5 times increase in all temperatures tested (Fig. 10b).Overall, Resoltech composites containing EoL PVP waste achieved significant energy storage capacities and performed consistently in a wide range of temperatures and frequencies compared to Araldite polymeric matrix, which displays significant increases only under specific conditions.Thus, considering the fact that Resoltech also performed better in mechanical strength tests, it can be considered a more suitable polymeric matrix in prospect of future work on this application.

Conclusions
Silicon solar cells were recovered from EoL PVPs and used as reinforcement in two different epoxy resin systems (Resoltech, Araldite) to produce dielectric composite materials, that were characterized through mechanical strength tests, SEM and BDS.Sample response in bending load tests showed that the presence of reinforcement does not affect significantly the polymeric matrices breaking point.Although a high deviation was observed, it is attributed to dispersion of the reinforcement in the matrix.In case of shearing load tests, however, solar cell granules lower the load required for breakage approximately by 30 and 50% for Resoltech and Araldite, respectively, due to formation of force accumulating points.This phenomenon was validated by SEM imaging of the samples that also displayed a homogeneous dispersion in the bulk of the matrix, albeit with an increasing concentration on the bottom surfaces due to the gravitational downward motion during the curation of the resin systems.Regarding BDS characterization, in all the examined frequencies and temperatures, the Resoltech composite materials with solar cell reinforcement demonstrated an energy storage capacity exceeding 1.5 times compared to the unfilled epoxy sample, indicating that they can be used potentially as dielectric material, as they exhibited a high energy storage capacity in wide condition range.
Summarizing the aforementioned, it can be concluded that (a) recovered silicon solar cells can be used as reinforcement to enhance the dielectric properties of a polymeric matrix, (b) Resoltech was found to be a more suitable polymeric matrix for filler particles of recovered solar cells from EoL PVP waste, as it formed a composite with better and more consistent mechanical and dielectric properties.Finally, in prospect of future work on the subject, implementation of a mechanical size reduction pretreatment (e.g., mill, pulverizer etc.) of the reinforcement is proposed, in order to achieve finer particle size and avoid deviations caused by dispersion in the matrix.

Fig. 2
Fig. 2 Solar cell particle size distribution after grinding

Filler 8 Fig. 3 Fig. 4
Fig. 3 Comparison of breaking force for bending loads for Resoltech (left) and Araldite (right) samples in the examined concentration range of recovered solar cells

Fig. 5 Fig. 6
Fig. 5 Scanning electron microscope (SEM) image magnification for Araldite composite with silicon powder content of a 2.5% at × 1000 (left) and b content of 7.5% at 400x (right)

Fig. 8
Fig. 8 Comparative graphs of the real part of the imaginary electric modulus (M") for a Araldite at 10 −1 Hz (left) & b Resoltech at 10. 2 Hz (right)

Fig. 9
Fig. 9 Energy density < u > (J/m 3 ) in relation to frequency for a steady temperature of 30 °C for a Araldite (left) and b Resoltech (right) samples

Table 1
Mean average of bending and shear loads with the corresponding standard deviation