Simultaneous recovery of high-purity Cu and poly(vinyl chloride) from waste wire harness via swelling followed by ball milling

Poly(vinyl chloride) (PVC) swelling coupled with ball milling was employed for the simultaneous recovery of high-purity Cu and PVC from waste wire harness under ambient conditions. The experimentally determined performances of 15 organic solvents for PVC swelling and phthalate plasticiser extraction were compared with those predicted considering Hansen solubility parameters. As a result, n-butyl acetate and acetone were identified as the two best solvents for adequate PVC swelling without PVC dissolution and almost complete plasticiser extraction within 60 min. The swelling was concluded to contribute to the control of phthalate plasticisers, the use of which in wire harness has recently been limited by the Restriction of Hazardous Substances (RoHS) directive. Cables swollen with n-butyl acetate or acetone were subjected to dry ball milling for ~ 60 min to completely separate PVC and Cu and achieve the quantitative recovery of these components from 20-cm-long cables. Thus, this work unveils the high potential of recycling the otherwise non-recyclable long and non-uniform waste wire harness cables and is expected to impact the related (e.g., automotive, electrical, and electronics) industries, contributing to the establishment of a more sustainable society.

Results and discussion characterisation of wire harness samples. Waste wire harness assemblies were collected from ELVs ( Fig. 1a) and contained cables of various diameters/textures, connectors, and banding bands. The connectors and banding bands were manually removed, and the cables were retained (Fig. 1b). Cables with diameters of 1.0, 1.2, and 2.0 mm had shares of 15, 70, and 15 wt%, respectively (Supplementary 1), and were all coated with PVC according to the results of elemental analysis (Supplementary Table S1). The quantitation of Cu, PVC resin, plasticiser, and tetrahydrofuran (THF)-insoluble components and plasticiser identification were conducted as described in Supplementaries 2 and 3, with the obtained results summarised in Table 1. The average contents of Cu and PVC were estimated at 71.1 and 28.9 wt%, respectively, taking into consideration the above cable distribution ( Supplementary Fig. S1). 1 H NMR analysis revealed that samples contained DINP as a plasticiser at a content of 21.4 wt% (Supplementaries 3 and 4), and insoluble compounds accounted for 6.7 wt% of the PVC coating. Thus, the present work quantitatively evaluated the composition of the heterogeneous waste wire harness assembly.
Swelling of pVc coating and Dinp extraction by organic solvents. Swelling allows the PVC coating to be easily separated by crushing in a ball mill reactor. Herein, 15 solvents covering a broad range of solvent-PVC HSP interaction distances (R a [S-PVC], calculated by Eq. (1) and summarised in Fig. 2) were selected, with details of each HSPs summarised in Supplementary Table S5-1. Table 2 summarises the abilities of organic solvents to swell PVC at ambient temperature, showing that according to this ability, the solvents can be divided into inadequate (i.e., benzene, diethyl ether, petroleum ether, ethanol, and water, R swel ≤ 2), moderate (ethyl acetate, isobutyl acetate, and isopropyl acetate, 2 ≤ R swel ≤ 3), adequate (acetone, n-propyl formate, n-butyl acetate, 1,4-dioxane, n-propyl acetate, 3 ≤ R swel ≤ 4 ), partial dissolution (4-methyl-2-pentanone (MiBK), 4 ≤ R swel ), and complete dissolution (THF, dissolution) groups. THF and MiBK, which completely and partially dissolved PVC coatings, respectively, were thus considered unsuitable for swelling. Although the order of R swel did not completely match that of R a [S-PVC], solvents suitable for adequate PVC swelling were concluded to have R a [S-PVC] = 6-9. A detailed discussion regarding the effects of HSP on PVC swelling and the corresponding mechanism are provided in Supplementary 5. Notably, shaking did not substantially impact swelling behaviour, i.e., simple submersion was sufficient for obtaining swollen cables. Figure 2 presents HSP interaction distances between each solvent and DINP (R a [S-DINP]). The best plasticiser extraction yield (Y ext ) of ~ 60% was observed for shaking-free 80-min extraction with acetone and n-butyl acetate (Fig. 3a). Although the order of R swel does not perfectly match that of Y ext for each solvent, these parameters are strongly mutually correlated, as PVC swelling assists solvent penetration into the PVC matrix. Therefore, the behaviour of Y ext cannot be simply explained by that of R a [S-DINP]. DINP extraction was substantially accelerated by shaking, with the effects of time on the Y ext of diethyl ether, ethyl acetate, n-butyl acetate, and acetone shown in Fig. 3b www.nature.com/scientificreports/ the Y ext values of the former group (91%) were comparable to those of the latter (97%) at 80 min. Thus, almost complete DINP extraction could be achieved using supportive shaking motion, whereas this motion did not influence PVC swelling. Thus, solvents adequately swelling PVC without its dissolution were concluded to be appropriate for subsequent dry ball milling, while solvents with high DINP extraction abilities were found to be beneficial for the control of RoHS-regulated phthalate plasticisers. Thus, taking the cost factor into account, acetone, n-butyl acetate, and ethyl acetate were selected as inexpensive and safe (i.e., posing low health and environmental hazards) 24 solvents for further experiments.
insights into the mechanism of pVc solvation and plasticiser extraction. Polymer dissolution is controlled by either the disentanglement of chains or their diffusion at the polymer (solid)-solvent interface 25 . Herein, the dissolution of PVC in THF was controlled by chain diffusion, and no swelling interface was observed before dissolution. At sufficiently high solvent penetration and swelling power, the swollen layer starts to be disentangled, as in the case of MiBK. Furthermore, the PVC coating contained encapsulated DINP, which could be released by swelling or chemical interactions. However, swelling is known to expand the PVC matrix and thus facilitate extraction. Hence, boundary layer resistance appears, particularly when the solvent affinity to PVC is low. Solvent molecules interact with incorporated DINP molecules at the outer boundary layer and initiate extraction to create voids (free surface), which allows for the further diffusion of solvent molecules and their interaction with deeper-lying DINP molecules. This process results in weight loss (shrinkage), as the solvent finally leaves the occupied voids. Figure 4 shows how DINP is extracted from the PVC matrix without swelling,  www.nature.com/scientificreports/ while shrinkage (weight loss) is observed after extraction. Solvents with modest swelling power such as acetone, ethyl acetate (EA), n-butyl acetate, n-propyl acetate, and 1,4-dioxane featured a slow swelling extent that was maximised at the end of the swelling time, leaving adjacent PVC chains unaffected. However, EA achieved a very limited swelling extent. Notably, only shrinkage was observed in diethyl ether and petroleum ether, i.e., these solvents engaged in favourable chemical interactions with DINP but not with PVC. To validate shrinkage behaviour of PVC by weight loss into diethyl ether, petroleum ether, and ethanol, herein R swel tests of fully deplasticized were conducted, as results R swel exhibited unity for all three respective solvents after 60-min submersion, indicate that there was no weight loss occurred during swelling tests, while as R swel was tested < 1 for plasticised cables. Moreover, the R swel values of specimens (see Table 2) fully extracted upon shaking were not different from values of samples extracted by conventional submersion (dipping). Notably, the solvent-PVC HSP interaction distance, R a [S-PVC], was positively correlated with the extent of solvation. Polar forces such as those due to δ d and δ h (as opposed to forces due to δ d ) are highly directional and greatly influence solvation. For example, ethanol exhibited δ d = 15.8, which was close to that of THF (16.8); however, THF dissolved PVC (δ d = 18.8), while ethanol completely incompetent solvent in solvation of PVC, makes sense that δ p and δ h substantially influence PVC solvation. Considering δ p and δ h to describe the order of solvation, we note that solvents (e.g., diethyl ether and petroleum ether) with either low or very high δ p and δ h [(2.9 ≤ δ p ≥ 0) and (5.1 ≤ δ h ≥ 0)] induced no significant chemical changes of PVC, which has δ p = 9.2 and δ h = 6.3. Still, shrinkage (R swel < 1) was observed for diethyl ether/petroleum ether because of plasticiser extraction from the PVC matrix. Moreover, while PVC-diethyl ether/petroleum secondary forces are exceptionally weaker for PVC swelling, shrinkage may impart due to fact that induced secondary forces of PVC-PVC attempt to contract chains by occupied free space (i.e., created by leaving plasticiser molecules from PVC matrix). Unlike the above ethers, benzene achieved a certain degree of PVC swelling, which was ascribed to the interaction of the benzene π-electron cloud with PVC molecules. However, this hypothesis is not backed up by evidence. Another viewpoint on polymer solubility 26 that is difference in free volume between the polymer and solvent liquids, i.e., dissimilarity mainly due to differences of sizes of solvent molecules and polymers molecules (i.e., fashioning of chains), may influence polymer solvation, aside of induced contact energy dissimilarities. Likely, Gilbert 27 also highlighted the contribution of molecular size and chemical structure to polymer solvation. The δ p and δ h deviate in small for THF (δ p = 5.7, δ h = 8), and EA (δ p = 5.3, δ h = 7.2), wherein THF dissolve PVC, while EA R swel exhibited < 3, is very far from PVC dissolution. Although, induced forces are important for discussion but hard to explain individual influence of each on PVC solvation. In the same way, n-butyl acetate (δ p = 3.7) achieved a larger swelling extent than EA (δ p = 5.3), although the δ p of the latter is much closer to that of PVC (9.2). Remarkably, 1,4-dioxane (δ p = 1.8) showed an R swel equivalent to that of n-butyl acetate, which was ascribed to the fact that the δ h of the former solvent (9) significantly exceeds that of the latter (6.3). Thus, one decisively cannot predict solvation by considering only δ p . The worthy to note that high electronegativity of chlorine makes the neighbouring carbon atom of PVC more prone to nucleophilic attack by electron-donating groups 16 . The polarisation due to chlorine may take into consideration but cannot be justified distinct behaviour of PVC solvation into respective solvents. Therefore, we employed the sum of δ p and δ h forces (i.e., 'δt p,h ' = (δ p 2 + δ h 2 ) 0.5 ) as a prediction parameter. For two solvents with similar δt p,h , such as MiBK and n-butyl acetate, one can definitely say that δ p has a larger contribution than δ h . However, abovementioned explanations could not directing such significant order for δ p and δ p to define the degree of PVC solvation. Here, we can make a general statement that the different proportionating value of δ p and δ h may show substantial solvation. The solvents which have δt p,h in the range of 7.3-12.5, is about Table 2. Ability of organic solvents to swell and shrink PVC coatings. a Standard deviation determined from ten results of 10 experiments. www.nature.com/scientificreports/ closer or similar to the δt p,h (11.2) of PVC, showed significant solvation of PVC. Thus, water did not achieve PVC solvation or extraction because of the unfavourable secondary force-induced interactions. At this point, it is worth re-emphasising that polar forces (δ p , δ h ) are highly directional, while those due to δ d are omnidirectional. Conclusively, PVC matrix swelling occur when involved secondary polar forces in PVC-solvent would greatly compatible, whereas chains may stretch (i.e., expansion). This phenomena described as (δt p,h ) PVC/DINP ≈ (δt p,h ) solvent for polar solvents. Likewise, the shrinkage condition stated as (δt p,h ) DINP ≈ (δt p,h ) solvent ≪ (δt p,h ) PVC for nonpolar solvents. Figure 5 shows a general representation of PVC swelling and shrinkage under the influence of organic solvents.  indicating that pure DINP could be easily recovered by extraction followed by distillation. At the same time, the solvent (acetone, n-butyl acetate, and ethyl acetate) regeneration efficiency of distillation was determined as ~ 90% (Fig. 6b). Thus, DINP recycling and solvent regeneration could be achieved by simple distillation.    Y sep values of 63 and 21% were obtained for 10-and 20-cm cables after 80 min, i.e., complete separation could not be realised. As expected, no separation was observed for untreated cables. Thus, we concluded that R swel strongly affects the efficiency of ball milling-induced Cu-PVC separation (Fig. 7d). The successful separation www.nature.com/scientificreports/ achieved for 20-cm cables is a substantial advance compared to our recent works, in which full separation could be achieved for cables with a maximal length of 3 cm 19,21 .

Recovery of cu and pVc by ball milling of swollen cables.
The images of ball milled 20-cm cables before and after manual sorting are summarized in Figs. 8 and 9, respectively. The images and Cu wire lengths of 20-cm cables after complete separation show that cables swollen with both n-butyl acetate and acetone were completely separated into PVC and Cu (Fig. 9a, b). In the case of n-butyl acetate, > 90% of collected Cu wire fragments had a length of ≥ 10 cm, which is a substantial advantage for subsequent Cu refining. For acetone, the recovered Cu fragments were slightly shorter because of the longer milling time required in this case. Figure 9c presents the images and length distribution of Cu fragments recovered from ethyl acetate-treated samples after 80-min ball milling, showing that separation was incomplete and that the obtained Cu wires were much shorter than those obtained from n-butyl acetate-and acetone-treated cables because of the longer milling time.
Thus, we successfully developed a novel technique combining PVC swelling with n-butyl acetate or acetone and dry ball milling to achieve the quantitative separation of 20-cm-long cables into high-purity PVC coating, Cu wire, and DINP, highlighting the great potential of this method for the recycling of non-uniform, long, and thin cables.

Materials and methods
Materials. Waste electric cables were obtained from ELVs (Fig. 1a). Connectors were sorted out, and the cables were cut into 5-, 10-and 20-cm-long pieces. Organic reagents were obtained from Kanto Chemical Co., Inc., (Tokyo, Japan). Coating identity was confirmed by elemental analysis (JM-10/CHN, J-Science Lab Co., Ltd., www.nature.com/scientificreports/ Japan and YHS-11/S-FClBrI, Science Glove Co., Ltd., Japan) for cables with different diameters and textures ( Supplementary Fig. S1). The quantitation of Cu, PVC resin, plasticiser, and THF-insoluble components and plasticiser identification were conducted as described in Supplementaries 2 and 3.
HSp calculation. The HSP concept considers the action of three general secondary forces between molecules (nonelectrolytes), namely dispersion interaction (δ d ), polar interaction (δ p ), and hydrogen bonding (δ h ) forces. The cohesion energy for PVC-PVC or PVC-DINP separation depends on the participating polar/nonpolar groups and contributing forces. The HSP interaction distances between solvents and PVC (R a [S-PVC]) and between solvents and DINP (R a [S-DINP]) were calculated and defined as 28 where subscripts 's' and 't' refer to solvent and target material (PVC or DINP), respectively. All HSPs for selected organic solvents, plasticisers, and PVC were calculated using HSPiP software (5th edition, 5.0.06).
Swelling of pVc coating by submersion into organic solvents. The collected waste cables were cut into 7-cm-long pieces and submerged into organic solvents (100 mL) in sealed 110-mL glass bottles at room temperature for 20, 40, 60, and 80 min. For comparison, submersion tests were also conducted using a water bath shaker (NTS-100V, Tokyo Rikakikai Co, Ltd., Tokyo, Japan) with horizontal shaking motion (150 rpm). For each group of tests, ten cables with different textures (Supplementary Fig. S1) were investigated, and the swelling ratio was reported as the corresponding average. The volume of each cable before and after submersion was measured by an electronic densimeter (MDS-300, Alfa Mirage Co., Ltd., Osaka, Japan) to determine the PVC swelling ratio (R swel ) as where V 0 and V s [g/cm 3 ] are the cable volumes before and after the swelling test, respectively. To investigate DINP extraction behaviour during swelling tests, 12.8-g cable samples were submerged into the target solvents (100 mL) in 110-mL sealed glass bottles for 20 or 80 min at ambient temperature. Moreover, cable samples of identical weight were tested with shaking using a water bath shaker for 20, 40, 60, and 80 min at ambient temperature. After the extraction test, the spent solvent was collected into a 250-mL round-bottom flask and evaporated using a rotary evaporator (RE-301-AW, Yamato Scientific Co., Ltd., Tokyo, Japan) at a flask rotation speed of 30 rpm. The water bath temperature was set to 30-60 °C according to solvent volatility. The solvent condenser was cooled to − 5 to − 10 °C by a chiller (CF-300, Yamato Scientific Co., Ltd., Tokyo, Japan) in all experiments. The distillation pressure was controlled using a pressure-controllable vacuum pump (NVP-1000V, Tokyo Rikakikai Co, Ltd., Tokyo, Japan) and was set to 560, 350, 250, 180, 170, 180, or 175 hPa for acetone, THF, ethyl acetate, n-propyl formate, isopropyl acetate, benzene, or ethanol, respectively. For highboiling-point solvents, namely MiBK, n-butyl acetate, 1,4-dioxane, n-propyl acetate, isobutyl acetate, or water, the vacuum pressure was set to 53, 39, 95, 65, 53, or 73 hPa, respectively. The plasticiser extraction yield (Y ext [%]) was determined as where w f [g] and w f0 [g] are weights of the flask after distillation and the empty flask, respectively, while W DINP in PVC [g] is the amount of DINP in the 12.8-g waste cable sample (see Supplementary 4).
The yields of plasticiser recycled from extraction solvents (Y recy ) were determined by 1 H NMR (Bruker Advance, 400 MHz, Rheinstetten, Germany) using fumaric acid as an internal standard 21 , with details provided in Supplementary 4. Ball milling tests for the recovery of cu and pVc from swollen cables. After swelling and simultaneous plasticiser recovery, the swollen cables were immediately transferred to the ball mill reactor for PVC separation, as schematically illustrated in Fig. 1c,d. Ball milling tests were carried out in a stainless steel reactor with a length of 160 mm and an inner diameter of 160 mm. For all experiments, 12.8-g cable samples and stainless steel balls with diameters of 15 mm (26 g/ball) or 20 mm (62 g/ball) were used. Rotation speed was controlled by a pot mill rotator (PM-001, AS ONE Co., Tokyo, Japan). The optimal rotation speed, ball size, and ball number were determined as 35 rpm, 20 mm, and 20, respectively (see Supplementary 7). After milling tests, PVC containing organic solvents was completely dried at 60 °C in a vacuum oven overnight. The dried samples were weighed, and the separation yields (Y sep ) were determined as where m 0 and m s [g] are the total sample weight after the milling test and the total weight of separated Cu and PVC after the milling test, respectively. (1) R 2 a = 4 δ d,s − δ d,t 2 + 4 δ p,s − δ p,t 2 + 4 δ h,s − δ h,t 2 , (2)