Alkaline hydrolysis of photovoltaic backsheet containing PET and PVDF for the recycling of PVDF

Multiple end-of-life, crystalline silicon PV panels were provided by a waste management and recycling company in Japan. The surface glass from each panel was removed by shot blasting after physically detaching the aluminum frame. The cross section of each panel was observed by scanning electron microscopy and energy-dispersive X-ray spectroscopy (SEM/EDX) (TM4000, Hitachi High-Tech Corporation, acceleration voltage: 15 kV, backscattered electron image, EDX software: The Oxford Instrument AZtec) to determine the presence of fluorine in the PV backsheet and to calculate the thickness of each plastic layer. Next, the fluorine-containing backsheet was manually removed from the panel and Fourier-transform infrared spectroscopy (FT-IR) (Nicolet Summit, Thermo Scientific, attenuated total reflection method, software: OMNIC Specta) was performed to identify the type of fluoropolymers and to confirm PET in the backsheet. The backsheet containing PVDF was then freeze-shredded to a length of 4 mm or less using a cutting mill (SM300, Retsch) to prepare samples for alkaline hydrolysis. The theoretical weight percentage of PET and PVDF in the backsheet was calculated from the densities of PET (1.38 g/cm³) and PVDF (1.76 g/cm³) and the thickness of the sheet as determined from the cross-sectional SEM/EDX image of the backsheet. To investigate whether the composition ratio of PET and PVDF remained constant before and after shredding, the fluorine content in the samples was analyzed by an automatic quick furnace (AQF-2100H, Nittoseiko Analytech, inlet temperature: 900 °C, outlet temperature: 1,000 °C) and ion chromatography (IC) (Dionex Integrion RFIC, Thermo Scientific, column: IonPac AG19/AS19, suppressor: ADRS-600 4 mm, eluent: 45 mM KOH, eluent flowrate: 1.0 mL/min) for statistical variability measurements.

Shredded backsheet (0.5 g) and 25 mL of NaOH solutions (1, 2.5, 5, and 10 M) were placed in a 300 mL stainless steel reactor. The backsheet was then hydrolyzed by heating the reactor in a silicon oil bath at 100 °C for 2 h under a nitrogen flow of 100 mL/min. To promote the reaction between the shredded backsheet and NaOH solution, baffle plates were installed inside the reactor, which was continuously stirred by rotation at 40 rpm while heating. After the reaction, the reactor was cooled to 25 °C, and the unreacted solids were collected by filtration. The filtrate containing Na₂TP was neutralized with 10 M H₂SO₄ to precipitate terephthalic acid (TPA). The precipitated TPA was filtered and dried overnight at 40 °C in a dryer, while the filtrate was stored as a reaction solution for the analysis of fluorine ion concentration.

The weight of the recovered TPA was measured to calculate the PET decomposition rate because of alkaline hydrolysis, defined as follows: where D_PET is the decomposition rate of PET, W_{TPA, r} is the weight of the recovered TPA, and W_{TPA, t} is the theoretical weight of TPA units in PET layer. Recovered TPA was analyzed by FT-IR for material identification. The reaction kinetics of PET hydrolysis were analyzed using the modified shrinking-core model.

The unreacted solids were dried after washing with ion-exchanged water, and image analysis was performed to observe the PVDF and unreacted PET layers using a digital microscope (VHX-7000, KEYENCE). The structural deterioration of the PVDF layer because of alkaline hydrolysis was analyzed by FT-IR.

Finally, IC was performed on the filtrate obtained after neutralization reaction to investigate the fluorine content, which might have increased due to the dehydrofluorination of the PVDF layer during alkaline hydrolysis. The defluorination rate was defined as follows: where D_f is the defluorination rate, W_{f, filtrate} is the weight of fluorine in the filtrate, and W_{f, backsheet} is the weight of fluorine in the shredded backsheet.

Figure 2 shows the cross-sectional SEM image and fluorine mapping of the PV panel used for the alkaline hydrolysis. The backsheet consisted of three layers of plastics, and the fluoropolymers were on the outer sides of the backsheet. The thickness of both fluoropolymers was 30.3 µm, and that of the inner plastic was 260 µm. FT-IR analysis on the backsheet is shown in Fig. S1 in the Supporting Information and it revealed that both sides of the fluoropolymers were PVDF and the inner plastic was PET. The theoretically calculated weight percentages of the PET and PVDF layers of the backsheet were 77.1 and 22.9 wt%, respectively. The fluorine content of the original backsheet was 9.7 ± 0.4 wt% (n = 5), whereas that of shredded backsheet was 9.3 ± 0.3 wt% (n = 5); therefore, the variation of plastic composition between samples originating from the shredding process was not significant. The shredded backsheet used as the samples for alkaline hydrolysis is shown in Fig. 3.

The analytical results for the recovered TPA, shown in Fig. S2 in the Supporting Information, verify the recovery of TPA from PET layer via alkaline hydrolysis and neutralization. Figure 4 shows the effects of the alkaline concentration and reaction time on PET decomposition rate (given by Eq. 1). Alkaline hydrolysis with 1 M NaOH solution resulted in low TPA recovery even after the reaction time was increased, and the PET decomposition rate was 30% upon heating for 2 h. In contrast, in higher concentrations of NaOH, the PET hydrolysis proceeded effectively. PET decomposition rate after heating for 2 h in 5 and 10 M NaOH solution was 86% and 87%, respectively. Alkaline hydrolysis using 2.5 M NaOH was not completed in the predetermined reaction time. These results indicate that the PET layer of the PV backsheet was hydrolyzed quantitatively according to Eq. 3 [28].

PET decomposition proceeded under milder conditions compared to those of previous studies [28, 30] probably because the thickness of the PET layer of the PV backsheet was only 260 µm (Fig. 2a), which enabled facile alkaline hydrolysis within the predetermined time of the experiment. In addition, the alkaline solution reached PET more efficiently than expected because of the partially peeled-off PVDF layer obtained from the shredding process of the sample preparation. Moreover, stirring the reactor during heating may have promoted alkaline hydrolysis. Although alkaline hydrolysis with 5 M and 10 M NaOH solutions efficiently decomposed the PET layer within 2 h, hydrolysis with 1 M and 2.5 M NaOH solutions required longer duration for the completion of the reaction. As previously reported [28, 30], the reaction temperature is a key parameter for the reaction efficiency of alkaline hydrolysis; thus, PET decomposition rate at lower alkaline concentrations could be improved by increasing the reaction temperature.

In addition, the PET hydrolysis kinetics were analyzed by the modified shrinking-core model, which expresses the effective surface area for PET hydrolysis using the degree of hydrolyzed PET (X) and a proportional constant (c) because of the formation of pore and cracks on PET during hydrolysis [32]. The growth and formation of pore and cracks is estimated to increase the effective surface area in this model. where S is the effective surface area and \({r}_{X}\) is the particle size of PET at any reaction time.

Here, the rate of reaction is expressed as follows: where V is the rate of reaction, k is the rate constant per unit surface area, and C_A is NaOH concentration.

If the PET is regarded as a sphere and its mass balance is considered to be solid, then where \(\rho\) is the density of PET.

The degree of hydrolyzed PET, X, can be represented as follows: where r₀ is the initial size of PET particle. The following equation can be derived from Eqs. 4–7.

Integrating both sides, we obtain: where K is the apparent rate constant equal to \(3k{C}_{A}/{r}_{0}\rho\).

The effect of NaOH concentration and reaction time, as predicted by Eq. 9, is shown in Fig. 5. For 10 M NaOH solution, plots were obtained assuming that almost all the PET was decomposed 60 min after the start of the reaction. Straight lines with a correlation coefficient of 0.93 or higher were obtained for PET hydrolysis at any alkaline concentration, indicating that the PET decomposition in this reaction proceeds according to the modified shrinking-core model.

Microscopic images of the unreacted solids after alkaline hydrolysis for 2 h are shown in Fig. 6. The white PVDF layer covered both sides of the transparent PET layer in the original backsheet; however, a part of the PVDF layer was peeled off from the PET layer after freeze-shredding (Fig. 6a). No significant change was observed in either PET or PVDF of the backsheet hydrolyzed with 1 M NaOH; however, after alkaline hydrolysis in 2.5 M NaOH solution, PET became brittle slightly and physical separation between the PET and PVDF layers was partially observed (Fig. 6b, c). Under higher concentrations of NaOH, while some brittle PET remained as unreacted solids, most of it decomposed, thus enabling the isolation of pure PVDF (Fig. 6d, e). These results demonstrated that more than 80% of PET decomposition would be required to the PV backsheet for the promotion of PVDF separation from PET layer. Although PET decomposition and PVDF separation proceeded efficiently in high alkaline concentrations, surface discoloration of the PVDF layer was observed.

The structural changes in PVDF following alkaline hydrolysis were analyzed using FT-IR spectroscopy (Fig. 7). The characteristic bands of PVDF that were identified in the spectrum of the original PVDF layer, including C–H bending (1400 cm⁻¹), C–F stretching (1180 cm⁻¹), C–H wagging (870 cm⁻¹), and C–F bending (830 cm⁻¹) [38], were also observed in that of the hydrolyzed PVDF layer. Micrographs of the unreacted solids (Fig. 6) showed a slight color change on the surface of PVDF treated with NaOH solutions of higher concentrations; however, no significant difference was observed in the FT-IR spectra of the original and hydrolyzed PVDF. Similar to the deterioration of polyvinyl chloride (PVC) [39‐41], the deterioration of PVDF proceeds via dehydrofluorination and C–C double bond formation (Eq. 10) [34, 37]; therefore, we prepared a deteriorated backsheet by heating shredded backsheet in 10 M NaOH at 130 °C for 6 h to compare the FT-IR spectra. The FT-IR spectrum of the deteriorated PVDF layer exhibited a broad band in the wavenumber range of 1750–1500 cm⁻¹ with its maximum intensity at 1650 cm⁻¹. This band was attributed to the C–C double bonds formed through the dehydrofluorination reaction, but it did not appear in the spectrum of the PVDF layer hydrolyzed using 10 M NaOH at 100 °C for 2 h. This result suggests that the deterioration of PVDF with significant structural changes might not progress under the alkaline conditions of the experiment.

The rate of defluorination of the PV backsheet via alkaline hydrolysis (given by Eq. 2) for 2 h was 0.27% in 10 M NaOH solution and significantly lower at lower concentrations of NaOH. As the limit of detection (LOD) of the IC conducted in this study was 0.4 μg/L (S/N = 3), fluorine might have been eliminated from PVDF through deterioration although its concentrations in the filtrate was extremely low. This experiment demonstrated that the PVDF layer could be separated from the PV backsheet by alkaline hydrolysis of the PET layer; however, the deterioration of PVDF via the dehydrofluorination reaction proceeds simultaneously with an insignificant structural change occurring in PVDF.

Although significant PVDF deterioration was not confirmed, slight color changes were observed on the surface of the PVDF layer. The tolerance for color changes cannot be determined as it will depend on the post-recovery usage of the recovered fluoropolymer. However, if the PVDF layer is not deteriorated, the recovered PVDF would meet the quality requirement for secondary plastics and could be used in the plastic recycling market. As surface color changes were not observed for PVDF treated with solutions containing lower NaOH concentrations, recovered PVDF that is of the same quality as virgin PVDF could be collected from the PV backsheet treated under milder and more optimum conditions that would suppress PVDF deterioration and promote PET decomposition.

Based on the results of this study, we propose a fluoropolymer recycling scheme for end-of-life PV panels (Fig. 8). Firstly, the PV backsheet should be shredded before alkaline hydrolysis. The shredding process is effective for making smaller the backsheets and increasing the surface area of the PET layer to improve its contact with the alkaline solution for hydrolysis. Then, shredded backsheet is placed into a heated reactor to promote PET decomposition. Alkaline hydrolysis enables the decomposition of the PET layer in the PV backsheet and promotes the physical separation of the PVDF layer from the PET layer via filtration. Highly alkaline conditions promote PET decomposition but also deteriorate PVDF; therefore, moderately alkaline conditions are required to collect pure PVDF without defluorination and any structural deterioration. The PET layer in the PV backsheet could be decomposed in alkaline solutions with concentrations under 10 M NaOH; however, several brittle PET layers were formed as reaction intermediates, as evidenced by digital microscopy. To address this problem, another approach to separate brittle PET and PVDF might be required to enhance the purity of the recovered PVDF. Sink-float separation techniques based on the differences in the densities of plastics and triboelectrostatic separation based on the differences in surface potential values might be applicable for the later stages of the recycling scheme. As fluoropolymers generally have high alkaline resistance, the suggested recycling schemes are applicable to all PV backsheets containing PVDF and other types of fluoropolymers.