Introduction

With increasing plastic pollution [1,2,3] and a political push [4,5,6] for increasing plastics recycling, more attention must be given to recycling of polymeric composites. Polymer composites offer high stiffness and outstanding strength-to-weight ratio, which is obtained by reinforcing thermosets or thermoplastics with fibres [7]. Reinforced polymer composites are substituting metals in industries [8] such as wind energy [9, 10], transportation [10], aerospace [10], and sports [9], and consumption is steadily growing [9, 11]. Although carbon fibres are utilised in demanding applications, glass fibres are still the most common reinforcement material [12]. For glass fibre reinforced thermoplastics, polypropylene is among the most commonly used [7, 12], and hence it is of paramount importance to enable composite circularity [11].

With the recycling of composite materials, a strong focus has been targeted toward mechanical, chemical, and thermal recycling. Mechanical recycling of thermoset composites include grounding and subsequently sieving into resin-rich and fibre-rich fractions [13, 14]. The resulting fractions have been demonstrated as reinforcement filler for thermoset polyester resins [15], thermoplastic composites [16], or reduced to conventional filler substitution (e.g. calcium carbonate) [11]. For thermoplastic composites, the effect on mechanical properties by applying varying amount (50–100%) of recycled polypropylene composites during production showed inhomogeneous fibre distribution, fibre shortage and breakage in the obtained products [17]. The inhomogeneity of the recycled material could be mitigated by proper preprocessing (size separation, mixing, etc.) and thereby a close loop recycling of automotive parts was demonstrated [14]. Chemical recycling usually uses solvolysis where the heated (and pressurised) solvent degrades the matrix into small molecules and fibre material may be recovered. This is demonstrated for carbon fibre enforced epoxy composites (CFRE) which was dismantled using supercritical CO2 and the fibres could subsequently be extracted [18]. Additionally, the recovery of carbon fibre mats was demonstrated by degrading and washing out the epoxy matrix using supercritical acetone [19]. Further, acid hydrolysis was applied to recover the carbon fibres from a CFRE from an aviation component [20]. High-pressure hydrolysis of a polyester resin at 350 °C resulted in a pure glass fibre fraction. After rinsing, the glass fibres obtained a clean surface and were applied in making new polyester resins [21]. Finally, during thermal recycling, the matrix is pyrolyzed (300–1000 °C) into oils or gasses [22], and in a few cases into monomer units [23]. In this process, the products require subsequent refining before further use [22, 24]. The extensive heating unfortunately also reduces the fibre strength after thermal recycling [8]. Fibres were demonstrated to maintain their fibre strength when heated to 150 °C [25]. However, 150 °C is below the decomposition temperatures for the most thermoset composites (epoxy, polyester, vinyl ester, phenolic resins, etc.) [25]. This was demonstrated on fibres obtained from a polyester composite that resulted in fibres with reduced flexural and tensile strength [26].

In general, chemical and thermal composite recycling usually has fibre recovery as the main objective. Thus, recycling techniques for recovery of both matrix and fibres “that does not significantly changing the chemical structure of the material” [27] are of relevance and can generally be considered recycling at a higher waste recycling hierarchy level [28]. An emerging alternative for thermoplastic composite waste streams is recycling by dissolution. Initially, the composite is reduced in size (i.e. increasing surface area), which minimises the solvent diffusion length [29, 30], before the granulates are immersed in a solvent at optimal temperatures [31, 32]. Ideally, only the polymer will dissolve while inorganic additives (e.g. fibres or pigments) remain as a solid that can be removed [33, 34]. Finally, an antisolvent is added to precipitate the extracted polymers [32, 35]. The benefit of recycling by dissolution comes from enabling the removal of inorganic additives without altering the chemical structure of the polymer and additives [36, 37]. Hence, removal of unwanted additives will enhance the usage of recycled material in products which previously could not apply recycled plastics.

Previous research has investigated the dissolution of polypropylene at various concentrations (50–200 g L−1) in hot (130 °C) xylene. Here, 97–100 wt% of polypropylene was recovered by precipitation in acetone [34] or n-hexane [38]. After dissolution recycling, the mechanical properties indicate a minor increase in elastic modulus, attributed to fractionation during precipitation (i.e. lower molecular weight polymers remaining soluble upon precipitation) [34, 38]. Solubility is a crucial factor for successful dissolution recycling, and work has been conducted to evaluate the potential of dissolving extruded polystyrene in essential oils [39]. A study demonstrated a 94–100 wt% reduction of blue and orange pigments from a coloured high-density polyethylene matrix by optimising the choice of solvents and anti-solvents [40]. Another pigment study furthermore showed that filtration through a layer of Celite reduced titanium dioxide, iron(III)oxide, and chromium(III)oxide pigment levels by 75–95 wt% from a poly(acrylonitrile butadiene styrene) matrix [37]. Few studies report the removal of fibres using dissolution recycling, where the mechanical properties of recovered glass fibres with various amounts of polymer residues have been investigated [41, 42]. Current research emphasises reclaiming fibres when applying dissolution recycling on thermoplastic composites; hence, more research is needed to evaluate the potential for recovery of both fibres and matrices.

This study attempts to recover glass fibres and polypropylene using dissolution recycling. This is achieved by dissolving polypropylene in hot xylene (130 °C) in a custom-designed heating block, and subsequently filtering the hot suspension through a preheated (130 °C) Büchner funnel. Various composite concentrations and filter papers were examined to find optimal separation conditions, which were defined as the highest possible concentrations of PP and GF, whilst maintaining processability and pure fractions. The quality of separation was evaluated by combustion and thermogravimetric analysis (TGA) to determine the inorganic content. Both fractions were chemically characterised by Fourier transformed infrared spectroscopy (FTIR) and compared to pristine polypropylene and glass fibres. Fibre length was determined using optical microscopy, and fibre thickness and surface morphology were determined using scanning electron microscopy (SEM).

Materials and methods

Material preparation

Polypropylene pellets (PP) (400-GA05, INEOS, UK), chopped glass fibres (GF) (E-glass woven roving, EC13-4.5-T437F), and a glass fibre-loaded PP composite (cPP) (POLYfill PPH GF5030 HC, 30 wt% glass fibres, Polykemi, SE) were all used as received. Images of the pristine materials are shown in Supporting Information Figure S1.1 (SI Fig. S1.1). Xylene (xylenes, reagent grade ≥ 75.0% Sigma-Aldrich, DK) was used as solvent and Whatman Grade 1 (W-filter) (11 μm particle retention, Merck, DE) or laboratory paper (L-filter) (Bordpapir AGF 132, Frisenette, DK) were applied as filter media.

Dissolution

For dissolution, 75 wt% PP and 25 wt% GF (PP/GF) or cPP was added to 15 mL xylene in a snap cap vial (30 mL, VWR, USA) sealed with a rubber septum (Z56,473–7, Sigma-Aldrich, USA) at concentrations listed in Table 1.

Table 1 Concentrations (C) of dissolved PP/GF, cPP, and filters used for filtration, where W and L refers to Whatman Grade 1 filter and laboratory paper, respectively
Full size table

Sample ID is abbreviated as PP/GFXXYY or cPPXXYY, where XX is concentration and YY is filter.

The vial (with content) was heated to 130 °C in a preheated custom-made heating block (see SI S2 for details and setup) and stirred at 600 rpm (RCT basic IKAMAG®, IKA, DE) for 30 min. The septum was pierced with a needle (Sterican® Ø0.8 × 120 mm, B. Braun, DK) to release pressure.

Filtration

The vial was emptied after dissolution into a preheated (130 °C) Büchner funnel connected to a Büchner flask for vacuum-assisted (Vacuum pump N 026.1.2AT.18, KNF, DE) filtration (see SI Fig. 3.1 for setup). Filtration was performed with either a W- or L-filter (see Table 1). All samples were filtered once, except cPP27LD where the filtration process was repeated with the filter incl. the filter cake as filter (with preceding heating of filter, filtrate, and funnel). The vial was rinsed with ≈10 mL hot xylene (130 °C), which was subsequently poured into the funnel. Filtrate and filter cake were placed in separate (pre-weighed) ceramic crucibles and dried overnight at 80 °C to evaporate xylene. cPP27L samples were prepared in triplicates.


The yield contributions from dried filtrate (yF) and filter cake (yC) are given in Eq. (2.1).

$$y_{F} = \frac{{m_{d} - m_{c} }}{{m_{{{\text{tot}}}} }}100\% \; \wedge \;y_{C} = \frac{{m^{\prime}_{d} - m^{\prime}_{c} - m^{\prime}_{c} }}{{m_{{{\text{tot}}}} }}100\%$$
(2.1)

where md and m’d are the determined masses of filtrate and filter cake after drying, respectively. mc and m’c are the masses of the empty crucibles, mf is the mass of the filter paper, and mtot is the initial mass of PP/GF or cPP.


The filtrate (fF) and filter cake (fC) fractions were determined using Eq. (2.2).

$$f_{F} = \frac{{y_{F} }}{{y_\text{{tot}} }} \wedge f_{C} = \frac{{y_{C} }}{{y_\text{{tot}} }}$$
(2.2)

where ytot is the total yield thus the sum of yF and yC.

Inorganic content


To determine the amount of inorganic content, samples were placed in an oven (Miditherm AN 202 262, BEGO, DE) at 600 °C in atmospheric air for 1 h. The inorganic content was determined by Eq. (2.3).

$$IC_{F} = \frac{{m_{a} - m_{c} }}{{m_{b} - m_{c} }} 100 \% \wedge IC_{C} = \frac{{m^{\prime}_{a} - m^{\prime}_{c} }}{{m^{\prime}_{b} - m^{\prime}_{c} - m^{\prime}_{f} }}100 \%$$
(2.3)

where ma and m’a are the masses after 1 h at 600 °C. mb and m’b are the masses before heating.

ICF was used to determine the inorganic contents of PP, GF, cPP, W-filter, L-filter, and all dried filtrates, and ICC was used to determine the inorganic contents of the dried samples including filter paper (i.e. filter cakes). In the following, post-heated GF is referred to as GF-IC.

Thermo-gravimetric analysis

TGA of 5–10 mg samples was performed on a TG 209 F1 Libra (Netzsch, DE) in air. Thermal program: 35 to 600 °C at 10 °C min−1, 15 min isothermal at 600 °C, then heated from 600 °C to 750 °C at 10 °C min−1.

TGA was used to determine the temperature at 5 wt% mass loss (T5), the temperature at maximum decomposition rate (Td), and the inorganic content (res), where res was determined as the residual wt% at 750 °C. Samples included pristine PP, GF, and cPP, as well as dried filtrates and filter cakes of cPP27L and cPP27LD.

Attenuated total reflectance Fourier transform infrared spectroscopy

FTIR spectra were recorded using a Nicolet iS5 spectrometer (Thermo Fisher Scientific, USA) equipped with an attenuated total reflectance unit with a ZnSe crystal. The spectra averaged 16 scans with a 2 cm−1 resolution from 4000 to 600 cm−1. Spectra were baseline corrected and corrected for wavelength-dependent penetration depth by OMNIC built-in functions (OMNIC, v. 8.2, Thermo Fisher Scientific, USA). Samples included pristine PP, GF, GF-IC, and cPP, as well as dried filtrates and filter cakes from cPP filtered through L-filter. For pristine PP and cPP, the pellets were cut in half to record spectra on the core of the pellets. PP, cPP, and filtrates were normalised with the band intensity at 2915 cm−1, whereas GF, GF-IC, and filter cakes were normalised to the band intensity at 920 cm−1. W- and L-filter were normalised by the intensity of the band at 1055 cm−1.

Optical microscopy

Images of the fibres were obtained on a Zeiss Stemi SV6 microscope equipped with a camera (Axiocam 105 color, Zeiss, DE) and a light source (CL 9000, Zeiss, DE). Samples included GF, GF-IC, and dried cPP27LD filter cake. Samples were spread across a microscope slide and images were obtained at 5.0 × magnification. Image analyses were performed in ImageJ (v. 1.53q), where images were calibrated via the scalebar. Fibre length was measured on ≈ 400 fibres and fibre types were categorised based on fibre length.

Scanning electron microscopy

Fibre diameter and surface morphology were examined by SEM (Nova NanoSEM 600, FEI). Acceleration voltage 5 kV, spot size 3 in low vacuum 60 Pa H2O using a helix detector in low vacuum detection mode. GF, GF-IC, as well as dried filter cake from cPP27LD were captured at 700–24,000 × magnification. Image analysis was performed using ImageJ (v. 1.53q), where images were calibrated via the scalebar. Fibre diameter was determined by measuring five times at different positions across 6–10 fibres (i.e. n = 30–50).

Results

Dissolution products

All samples were dissolved after 30 min at 130 °C, yielding transparent suspensions with dispersed fibres. After emptying and rinsing the vials, thin films with minor fibre residues remained inside the vials. The processing of the dissolved samples was required to be < 5 min, as gelation was observed upon cooling. Samples with concentrations of 54 g L−1 and below were able to pass through the W-filter, however, increasing the concentration to 67 g L−1 caused filter clogging. Visual inspection of the PP/GF54W filter cakes showed minor PP residues on the fibres, whereas the concentrations below 54 g L−1 yielded visually clean fibres. cPP-suspensions instantly clogged the W-filter but could be filtered using the L-filter. After drying, all filter cakes had loose fibres lying on the filter paper, except for PP/GF54W where fibres were covered in polypropylene, binding the fibres to the filter paper. The determined ytot, fF, and fC are listed in Table 2.

Table 2 Overall yields (ytot), filtrate (fF) -, and filter cake (fC) fractions as well as inorganic content of filtrates (ICF) and filter cakes (ICC). CL refers to clogged filters
Full size table

For all samples, ytot > 87 wt%, except cPP27LD which yielded 80 wt%. Among the PP/GF samples, PP/GF37W had the highest ytot (94 wt%), and among the cPP samples, cPP27L gave the largest ytot (89 wt%).

Inorganic content

The inorganic content of the pristine materials (see SI Table S5.1) show that neither pristine PP nor W- or L-filters had any measurable residue (i.e., 0 wt% inorganic content) after thermal treatment (1 h at 600 °C), thus showing that both materials completely combusted. GF and cPP resulted in 99 and 30 wt% inorganic content, respectively, attributed to glass fibres. In addition, Table 2 shows that all filtrates fully decomposed (≈ 0 wt% inorganic content), and that all filter cakes contained 98–100 wt% fibres, except PP/GF54W with an inorganic content of 78 wt%.

After thermal treatment of PP, filters, and filtrates, a small amount of ash was visible in the bottom of the crucibles (SI Fig. S5.1). Furthermore, occasionally few (< 20) fibres were observed in the crucibles containing filtrates after burning. However, the masses of fibres and ashes were below the detection limit of the measurement equipment.

Thermo-gravimetric analysis

The degradation temperatures and residual masses obtained by TGA for the PP, GF, cPP, as well as dried filtrates and filter cakes of cPP27L and cPP27LD are listed in Table 3. All obtained thermograms can be found in SI S6 as well as a list of decomposition temperatures for W- and L-filter.

Table 3 Temperature of 5% mass loss (T5), temperature at maximum decomposition rate (Td), and residual mass at 750 °C (res) obtained by TGA. The phases (P) of samples cPP27L and cPP27LD are referred to as filtrate (F) or filter cakes (C). Non applicable refer to thermograms without a characteristic decomposition step
Full size table

The T5 of PP, cPP, and the cPP27LD filtrate were all around 300 °C, whereas the T5 of the cPP27L filtrate was 275 °C. The Td of pristine cPP and cPP filtrates were all around 390 °C, whereas PP was 360 °C.

Filters, PP, and the filtrates from cPP27L and cPP27LD went through full decomposition, leaving residues of 0 wt% (see SI S6). GF and filter cakes resulted in ≈ 99 wt% remaining mass, showing little-to-no signs of organic matter decomposition. cPP maintained 29 wt% of its original mass. In all cases, mass changes were finalised prior to the 15  min isotherm at 600 °C.

Attenuated total reflectance Fourier transform infrared spectroscopy

FTIR spectra of PP, GF, GF-IC, cPP as well as dried filtrate and filter cake of cPP27L are shown in Fig. 1. FTIR spectra and assignments of W- and L-filter as well as dried filtrate and filter cake of cPP27LD are given in SI S7.

Fig. 1
figure 1

Normalised FTIR spectra. The first three spectra are polypropylene containing samples: PP, cPP, and dried cPP27L filtrate, followed by GF, GF after thermal treatment (1 h at 600 °C) (GF-IC), and dried cPP27L filter cake. The spectra are shifted to enhance readability

Full size image

The spectrum of PP (Fig. 2) show strong C–H sp3 hybridised C–H stretches at 2950 and 2870 cm−1, and bands at 2920 and 2840 cm−1 assigned to asymmetrical and symmetrical stretches of CH3 and CH2, respectively. The 1460 and 1380 cm−1 bands are assigned to symmetric and asymmetric C–H deformation vibrations, respectively, and the low-intensity bands at 1165, 1000, 970, and 840 cm−1 are attributed to isotactic PP [43].The spectra of cPP27L show (in Fig. 2) all the same bands as PP, whereas cPP differs only by the broad band between 1250 and 600 cm−1 which is assigned to Si–O–Si from the glass fibres (Fig. 2).

Fig. 2
figure 2

Size distribution of glass fibres (n ≈ 400) in GF, GF after thermal treatment (1 h at 600 °C) (GF-IC), and dried cPP27LD filter cake (left). Enhanced detail of the 0 to 0.5 mm distribution (dashed rectangle) is shown in the insert. Examples on the right show images from optical microscopy of GF (top) and cPP27LD filter cake (bottom)

Full size image

The remaining three spectra of GF, GF-IC, and cPP-27 filter cake all have broad, strong intensity bands around 1000 cm−1 and around 800 cm−1 from Si–O–Si stretching and bending, respectively. The GF further show bands at 2920 and 2850 cm−1 assigned to CH2 and CH3 sp3 hybridised C–H stretches, a weak carbonyl/ester band at 1730 cm−1, and a broad O–H stretching band ≈ 3375 cm−1 from glass fibre sizing. The filter cake shows two low-intensity sp3 hybridised C–H stretches at 2890 and 2975 cm−1 and a broad, low-intensity O–H stretch around 3350 cm−1, these bands could be from sizing or cellulose, as the spectrum of L-filter (see SI Fig. 7.1 and Table 7.1) showed the same characteristic bands of cellulose including two sp3 hybridised C–H stretches (2900 and 2850 cm−1) and a broad O–H stretch ≈ 3340 cm−1.

Microscopy

The fibre length distributions of GF, GF-IC, and dried cPP27LD filter cake are shown in Fig. 2 (additional images are given in SI S8).

Figure 2 shows that all measured fibres were sub 5 mm long. GF had an average fibre length of 1.7 mm ± 1.2 mm, hence short chopped fibres [7]. The cPP27LD fibres had an average fibre length of 0.4 mm ± 0.3 mm and no fibres longer than 2.5 mm, hence milled fibres [7].

The GF-IC fibres had an average fibre length of 0.7 mm ± 0.9 mm, with 58% of fibres shorter than 0.5 mm. In contrast, only 12% of GF fibres were shorter than 0.5 mm, indicating that the thermal treatment (1 h at 600 °C) reduced the glass fibre length.

SEM images of GF, GF-IC, and dried cPP27D filter cake from double filtration through L-filter are shown in Fig. 3.

Fig. 3
figure 3

The 1.4k magnified SEM images of GF (A), GF after thermal treatment (1 h at 600 °C) (GF-IC) (B), and dried cPP27LD filter cake (C)

Full size image

Figure 3A shows how GF (as received) were bundles with a smooth surface and had an average diameter of 12.1 µm ± 1.7 μm. Figure 3B shows similar smooth surfaces on the GF-IC and that the fibre diameter (12.9 µm ± 0.8 μm) was unchanged. Fibres from the cPP27LD filter cake show smooth surfaces with no signs of residual polypropylene (Fig. 3C) with a fibre diameter of 13.2 µm ± 1.1 µm. The SEM image of cPP27LD dried filtrate show a homogeneous polypropylene surface (see SI S9). Additional SEM images of GF, GF-IC, and dried cPP27D filter cake are available in SI S9.

Discussion

Evaluation of pristine materials

The compositions of the pristine PP, GF, cPP, and filters were verified by FTIR (Fig. 2 and SI Fig. S7.1) and TGA (Table 3 and SI S6). The FTIR spectrum of PP showed bands at 1170, 1000, 970, and 840 cm−1 which are characteristic of isotactic polypropylene [43]. The inorganic content, determined by TGA and inorganic content measurements, showed full decomposition of PP. For GF, the characteristic bands for silicate glass and bands from the applied sizing were observed by FTIR (Fig. 2). Thermally removing the sizing by heating (600 °C) and TGA showed a ≈1 wt% loss (Table 3 and SI S6), which is in agreement with typical sizing quantities of 0.2–1 wt% [44,45,46]. Further, the removal of sizing was also confirmed by FTIR from the elimination of the sp3 hybridised bands (Fig. 2). The decrease in fibre length (Fig. 3) could be as a consequence of the introduced thermal relaxation [47,48,49], as previous studies showed up to 70% strength degradation upon thermal treatment (600 °C) [48, 49]. However, the decrease in fibre length (Fig. 3) caused by thermal treatment does not challenge the filtration, as thermal treatment was only used post filtration for product analysis.

The FTIR spectrum of cPP (Fig. 2) showed repetitions of the bands found in PP and GF, except for GF’s sizing band at 1730 cm−1. The glass fibre content was determined to be 29–30 wt% (SI S5 and Table 3), which is in accordance with the glass fibre content declared by the producer (30 wt% ± 2 wt% [50]).

Finally, the FTIR spectra of the filters (SI Fig. S7.1) showed the characteristic bands of cellulose. The inorganic content of W- and L-filter were 0.0 and 0.2 wt% (SI Table S5.1), respectively. Thus, the mass contribution from filters upon measuring inorganic contents of filter cakes was considered negligible.

Filtration processing conditions

A way of increasing the profitability and sustainability of dissolution recycling is to work with as high concentrations as possible whilst maintaining processability. Results from Table 2 show that filtration using W-filter was possible for all PP/GF samples with concentrations below 54 g L−1, whereas increasing the concentration to 67 g L−1 clogged the filter. The ICC showed that PP/GF19W and PP/GF37W had higher inorganic contents in the filter cakes (≥ 98 wt%) than PP/GF54W (78 wt%). Furthermore, PP/GF54W had visible polypropylene residues on the fibres after drying (SI Fig. S4.1). These observations can be explained by the increased viscosity of the suspensions as concentrations increase [36]. This increasing viscosity reduces the filtrate flux and thus prolongs the filtration time. The filtration time is of essence, as the decrease in temperature results in increasing viscosity and eventually sets the suspension as a gel upon cooling. Hence, cooling dictates the processability upon increasing concentrations. A continuously heated filtration rig or heated (130 °C) environment could counteract such complications and potentially enable filtration at higher concentrations. However, for the given experimental setup, the conditions of PP/GF37W excels, as this concentration enabled high yields in addition to pure and well-separated polypropylene and GF fractions.

For the polypropylene composite with milled fibres (cPP), all concentrations clogged the W-filters. Comparing the concentrations shows that PP/GF37W filtered nicely while cPP13W clogged the W-filter. These observations are ascribed to the difference in fibre length. GF had an average length of 1.7 mm, whereas the fibres from cPP were on average 0.4 mm. As the fibre diameters are similar, the shorter fibre length results in a closer packing of the deposited fibres. The denser fibre packing reduces the filter cake porosity and thus increase the filter cake resistance. As with increased viscosity, increased filter cake resistance will prolong the filtration time and result in gelation and clogging. This is underpinned by the fact that reducing the filter paper resistance using the more porous L-filter resulted in that all cPP concentrations could be filtered. Despite coarser filter paper and smaller fibre lengths, the L-filter still captured 99–100 wt% of the fibres (Tables 2, 3, and Fig. 3) leaving a pure filtrate (0 wt% glass fibres in the polypropylene). In addition, it was demonstrated that filtering the suspension through the same filter cake again only resulted in loss of polypropylene due to processing without increased purity. Thus, the obtained purities of cPP13L and cPP27L were similar to PP/GF37W demonstrating removal of the most common fibre types (i.e. chopped and milled glass fibres) in thermoplastic composites.

Quality of recycled plastic and fibre fractions

FTIR was used to chemically characterise the dried filtrates and filter cakes of cPP27L and cPP27LD (Fig. 2 and SI Fig. S7.2). When comparing the filtrates to the spectrum of pristine cPP, the curvature arising from GF in the 1250–600 cm−1 range had disappeared, thus supporting the hypothesis of the removal of glass fibres. Furthermore, the spectra of the filtrates and PP were alike, with the only deviation being a weak band at 1730 cm−1. This band probably originates from sizing from the fibres transferred to the filtrate during the dissolution process, as organic solvents has been applied to extract GF sizing in previous studies [51, 52]. The FTIR spectrum of the filter cakes show the characteristic silica bands, like observed in the spectrum of GF, but with shifted sp3 C–H stretches, which could indicate differences in the chemical composition of the sizing used on the GF and cPP-fibres. When comparing the spectra of the filter cakes to cPP, no bands from polypropylene are observed, indicating that filter cakes are pure glass fibres. Heating (1 h at 600 °C) the dried filtrates (PP/GF37W and cPP27L) caused complete combustion (Table 2), similarly observed with pristine PP (SI Table S5.1). In contrast, the dried filter cakes showed a 1–2 wt% mass loss upon heating, similarly observed with pristine GF. This indicates that sizing remained on the fibres upon dissolution recycling. Likewise, the thermograms (SI S6) of cPP27L and cPP27LD showed similarities to those of GF and PP, validating the inorganic content measurements: the dried filter cakes and GF both remained around 99 wt% of their original masses. The 1 wt% mass loss were in accordance with the one observed from inorganic content measuring assigned to sizing removal (Table 2). In contrast, the dried filtrates and PP both fully decomposed in a single-step mass loss remaining 0 wt% of their original masses. The characteristic temperatures of the filtrates (270–305 °C) were close to that of PP (300 °C). SEM images of the cPP27LD filter cake (Fig. 3C) showed clean, recovered fibres with smooth surfaces similar to the surfaces of pristine GF (Fig. 3A and SI S9). Likewise, the SEM images of cPP27LD filtrate showed a homogenous polypropylene surface.

Application areas besides polypropylene

It is a strength of this study that both inorganic additives and polymeric matrix can be recovered for recycling. In addition, the usage of filtration to recover glass fibres can be applied to other common thermoplastic composites, once useful solvents and temperatures are identified. Although not included in this study, solvent recovery is paramount for this technique to become industrially relevant in terms of operating costs and sustainability. Fortunately, the chemical industry is familiar with handling and recovering solvents, and thus this challenge can be solved at an industrial scale. However, with proper solvents and temperatures, filtration might be useful for separation of fibres or inorganic fillers from other prevalent thermoplastic composites.

Conclusion

Dissolution recycling was demonstrated as a recycling technique for thermoplastic composite recycling and it enables recovery of both fibres and matrix without altering their chemical compositions. The fine Whatman (grade 1) filter enabled the separation of short chopped glass fibres from dissolved polypropylene, whereas a coarser filter was necessary to remove milled glass fibres. The filtrate was identified as pure isotactic PP without glass fibres and the filter cake was identified as glass fibres with 1 wt% organic matter ascribed to sizing. Analysis of the recovered glass fibre surfaces by SEM showed clean, smooth surfaces of the recovered glass fibres. Hence, dissolution recycling enabled separation and recovery of pure polypropylene and clean glass fibre permitting utilisation of both matrix and fibres.