A two-stage strategy for upcycling chlorine-contaminated plastic waste

Abstract

Chemical upcycling of polyolefin plastic waste to lubricant, wax and fuel-range hydrocarbons over metal-based catalysts is a crucial technological solution to the enormous environmental threat posed by plastic waste. However, currently available methods are incompatible with chlorine-contaminated feedstocks. Here we report a two-stage strategy for upcycling chlorine-contaminated polypropylene. First, magnesia–alumina mixed oxide at 30 bar H₂ and 250 °C serves as a chlorine trap by rapidly forming solid chloride, resulting in nearly complete chlorine extraction from the polyolefin melt. This enables the upcycling of plastic waste with up to 10% polyvinyl chloride content to lubricants over ruthenium-based catalysts, in the second stage. The strategy is also applicable to chlorinated aromatics and alkanes. The proposed strategy renders hydrocracking and hydrogenolysis catalysts less sensitive to the chlorine impurities in feedstocks while eliminating HCl emissions and chlorine contamination in products. It could incentivize further progress in plastics upcycling.

Main

Since the 1950s, several billion tons of plastic have been generated, with ~60% discarded into landfills¹. The recycling of plastic waste (PW) is relatively minuscule compared with the waste generation², especially of single-use packaging films^3,4,5. Most PW leaks into the environment, destroying ecosystems^2,6. An important problem in PW management is its mixed nature, consisting of several polymers, including polyolefins (POs) (polypropylene (PP) and polyethylene (PE)), polyvinyl chloride (PVC) and polyethylene terephthalate (PET), blended with various additives such as plasticizers, dyes and antioxidants⁷. This makes mechanical recycling very challenging⁸. Pyrolysis of PO is less sensitive to the nature of the material but requires high reaction temperatures and high energy, producing low-value products (char and light gases)⁹. Hydrogenolysis of pure POs (PP and PE) over metal-based catalysts such as Pt/SrTiO₃ (ref. ¹⁰), Ru/C (refs. ^11,12) and Ru/TiO₂ (ref. ¹³) or hydrocracking over metal–acid bifunctional catalysts, such as Pt/WO₃/ZrO₂ (ref. ¹⁴), makes valuable products, such as lubricant, wax and jet fuel. Unfortunately, contaminants in actual PO waste probably deactivate these catalysts. One of the harshest impurities, PVC, is present in PW at ~4% concentration (if easily recyclable polyethylene terephthalate is excluded from the calculation)¹⁵. The Cl atoms poison the Ru and Pt active sites. Decomposition of PVC during the reaction produces HCl, a highly corrosive and toxic compound. Lubricants and fuels contaminated with even several ppm of HCl or chlorinated organics are unsuitable for use.

Moving chemical recycling and upcycling real-world PW forward requires PVC-tolerant catalysts or an efficient Cl removal step. Such technology will also eliminate the necessity of labour-intensive sorting of small amounts of PVC from PW. Recent reports show several approaches to Cl removal based on solvothermal treatment¹⁶ and pyrolysis over a zeolite catalyst¹⁷.

In this Article, we propose a two-stage absorptive dechlorination/hydrogenolysis or hydrocracking (Fig. 1). Instead of extensive sorting, cleaning or extraction of impurities, the feedstock reacts with a dechlorinating agent, magnesium–aluminium mixed oxide (Mg₃AlO_4.5), to trap Cl as inert MgCl₂. Mg₃AlO_4.5 provides a high density of basic sites^18,19 to bind HCl formed from PVC. Importantly, we show that dechlorination protects hydrogenolysis and hydrocracking catalysts from poisoning, prevents Cl leakage into the liquid reaction products and eliminates HCl emissions. The broader scope of halogen-containing plastic additives, such as fire retardants²⁰ and plasticizers²¹, can be treated similarly. Ultimately, the proposed two-step technology makes hydrogenolysis PVC tolerant, bringing the PO-to-products process closer to practice. Sustainable PO hydroconversion will rely on this pretreatment step to eliminate HCl poisoning of the catalyst and its emission or leakage into the product mixture. This will generally reduce the burden for waste feedstock purification, facilitating widespread adoption of waste catalytic hydroconversion. Recently discussed plastic product design strategies²² could create a stable pipeline of relatively clean PO feedstock, further incentivizing development of catalytic upcycling technologies (such as conversion of PP to lubricants¹³ or PE to aromatics²³).

Fig. 1: Pathways from plastic waste to valuable products.

Current scheme of hydrotreatment of PVC-contaminated plastic waste and proposed absorptive dechlorination. LDPE, low-density PE; PS, polystyrene; HDPE, high-density PE.

Results

PP–PVC mixture hydrogenolysis

The Ru/TiO₂ catalyst is highly active for pure PP hydrogenolysis (Fig. 2a), consistent with previous reports¹³. Liquid products formed over Ru/TiO₂ showed lubricant properties, similar to group I base oil¹³. The liquid yield reaches 66%; the major side product is gas (a yield of 28%), consisting of 76% methane, due to secondary terminal C–C bond breaking (Supplementary Table 1). In this context, liquid in PP hydrogenolysis refers to the fluid fraction of the product that is easily separable from the solid residue. For PP mixed with 10 wt% of PVC, Ru/TiO₂ shows almost no activity, with the residual solid yield reaching 99% (Fig. 2a). During PVC thermal decomposition, HCl and –[CH = CH]– polyene form, potentially poisoning the Ru active sites. HCl was also detected in the gas phase of the reactor headspace using Fourier transform infrared (FTIR) spectroscopy (Fig. 2b).

Fig. 2: Catalyst tolerance towards PVC contamination.

a, Product distribution for pure PP over Ru/TiO₂ (i), PP–PVC mixture over Ru/TiO₂ (ii), PP–PVC over Mg₃AlO_4.5 mixed with Ru/TiO₂ (iii), PP–PVC over Ru/Mg₃AlO_4.5 (iv), PP–PVC via the two-stage strategy (v). b, FTIR spectra of the product gas for each case, with the HCl spectrum as a reference. c, ¹H NMR spectra of liquid products formed with pure PP (i) and PP–PVC via the two-stage strategy (v). d, GPC chromatograms of liquid products. e,f, Two-dimensional ¹H–¹³C heteronuclear single quantum coherence NMR of liquid products formed from pure PP (e) and PVC–PP via the two-step strategy (f). Dechlorination conditions: 250 °C, 30 bar H₂, 6 h, 0.2 g PVC, 0.2 g Mg₃AlO_4.5, 1.8 g PP. Reaction conditions: 250 °C, 30 bar H₂, 16 h, 0.1 g Ru/TiO₂, dechlorinated solid.

Basic oxides trap HCl during high-temperature thermolysis of PVC. Various compounds can be efficient Cl traps for pure PVC pyrolysis and dechlorination, including Fe₃O₄ (ref. ²⁴), ZnO (ref. ²⁵), mixed magnesia–alumina oxide and others²⁶. Following ref. ²⁶, we added Mg₃AlO_4.5 with an Al/Mg ratio of 0.33 and a surface area of 205 m² g⁻¹ to the hydrogenolysis reactor (Supplementary Fig. 1). On the basis of stoichiometry for complete chlorination of Mg₃AlO_4.5 to MgCl₂, it should trap ~2.1 g PVC per gram of adsorbent. We used experimental loadings of 1 g g_ads^–1 of PVC since full chlorination does not occur even at high reaction temperatures due to the formation of a stable chloride shell and the slow Cl diffusion in the bulk²⁷. However, the activity did not increase, while the amount of HCl decreased 5–6 times (still detectable). The same happened when Ru was directly deposited on Mg₃AlO_4.5 (Fig. 2a). Thus, a single-step strategy seems inadequate. In the two-step strategy (Fig. 1), the mixture was first treated with Mg₃AlO_4.5 and then with Ru/TiO₂. For dechlorination, the PVC–PP mixture was combined with Mg₃AlO_4.5 in a batch reactor at 30 bar H₂ pressure, 250 °C for 6 h. The resulting solid was mixed with fresh Ru/TiO₂ and reacted at the same conditions for 16 h. Mg₃AlO_4.5 was not separated from the polymer mixture after dechlorination (see Methods for details). The liquid yield reached ~70%, close to pure PP. Interestingly, the gas yield dropped to 2.2% with a gas consisting mainly of isobutane and isopentane (60% selectivity). The molecular weight distributions of the liquids are similar, with a weight-averaged M_w equal to 900 Da for pure PP and 610 Da for PVC + PP mixture (Fig. 2d). Nuclear magnetic resonance (NMR) of liquid products highlights differences in microstructure (Fig. 2c,e,f). For pure PP, the liquid consists mainly of polypropylene oligomers with distinct signals of CH₃, CH₂ and CH groups in the heteronuclear single quantum coherence spectra¹³. A small fraction of (CH₂)_n fragments and other defects in the PP sequence lead to the 41–19 ppm peaks¹³. For PP + PVC conversion, the signals of various CH₃ groups have higher relative intensity due to shorter chain size, consistent with their smaller ¹³C chemical shift of 17.0–9.5 ppm range. The signals of CH₂ groups in (CH₂)_n domains are much more intense than for pure PP. This finding points to a favourable detachment of methyl groups in PP via breaking the CH₃–CH bond. Standard PP hydrogenolysis preferably breaks CH₂–CH bonds in the first 3–16 h, while CH₃–CH bonds are broken later, in 16–24 h. The stark difference in the relative content of CH₂ groups was also quantitatively assessed by ¹H NMR (Fig. 2c), which shows higher intensity in the 1.6–1.0 ppm region, especially the 1.3 ppm broad peak. Despite differences in the microstructure, both liquids correspond to lubricant-range hydrocarbons.

Direct conversion of a PVC-containing feed in a single step completely inhibits hydrogenolysis, consistent with previous reports showing Ru/C deactivation with only 0.1% PVC (ref. ²⁸). The most stable catalyst tested so far is 0.125% Ru/CeO₂, which raised the solid yield from 5% to 45% with 1% PVC contamination²⁹. Our results show that the two-stage strategy effectively retains the Ru catalytic activity and gives 70% liquid yield with 10% PVC content in the mixture by effectively entrapping HCl in Mg₃AlO_4.5 and physically separating the dechlorination from the hydrogenolysis. Tests with high-density PE– (HDPE–) PVC mixture and amorphous PP show that the two-step procedure is applicable to other polymer compositions (Supplementary Fig. 2 and Table 2). Interestingly, the residual Cl content in the liquid depends on the type of polymer mixed with PVC. The amorphous PP–PVC mixture had Cl contamination, while the HDPE–PVC mixture was Cl free.

Dechlorination mechanism

During dechlorination, PVC is decomposed over Mg₃AlO_4.5 while PP remains intact (Supplementary Fig. 3). In the second stage, PP undergoes hydrogenolysis over Ru/TiO₂, with some exposure to minor products of PVC decomposition from the first stage. This may affect the product distribution over Ru/TiO₂. To assess this hypothesis, the dechlorination mechanism of pure PVC was studied by heating it over Mg₃AlO_4.5 at different temperatures at 30 bar H₂.

Figure 3a shows that the initial Mg₃AlO_4.5 contained mostly a MgO phase with a broad (200) reflection and an impurity of κ-alumina. The initial hydrotalcite decomposition leads to the crystallization of MgO, with Al³⁺ being incorporated into the lattice¹⁹. Upon heating with PVC to 200 °C, new diffraction maxima due to MgCl₂ emerge. As the temperature increases, the reflection becomes more refined owing to the growth of the MgCl₂ crystals. The MgO peak decreases in intensity, reaching a residual value at ~250 °C due to saturation with chlorine. Unreacted MgO is left because we employed nearly double the stoichiometric Mg₃AlO_4.5 amount as complete chlorination may not occur due to forming a stable chloride shell and slow Cl diffusion in the bulk²⁷. In practice, excess Mg₃AlO_4.5 increases the cost and may impede mixing. The ¹³C magic angle spinning (MAS) NMR (Fig. 3b) reveals that as MgCl₂ forms, PVC undergoes decomposition, and the two PVC peaks at 57 and 47 ppm (–CHCl– and –CH₂– groups) decrease. Simultaneously, new peaks at 128 and 32 ppm appear after heating at 250 °C (ref. ³⁰). They correspond to –C = C– polyenic chain segments and aromatic hydrocarbons formed after HCl elimination from PVC. The 32 ppm peak corresponds to alkyl fragments closely attached to these unsaturated units. These species do not originate from PP (experiments with mixed PVC–Mg₃AlO_4.5 were conducted).

Fig. 3: Products of pure PVC degradation over Mg₃AlO_4.5.

a, XRD patterns of PVC mixed with Mg₃AlO_4.5 heated at different temperatures. b, ¹³C MAS NMR of the same samples. c,d, Quantitative XPS results on Cl/Mg (c) and Al/Mg (d) ratios as a function of treatment temperature. e, Reaction scheme. Dechlorination conditions: 30 bar H₂, 6 h, 0.2 g PVC, 0.2 g Mg₃AlO_4.5.

The initial PVC–Mg₃AlO_4.5 mixture does not have Cl atoms on the surface, as shown in X-ray photoemission spectroscopy (XPS) (Fig. 3c). At low treatment temperatures, PVC is covered mainly with Mg₃AlO_4.5 particles, leading to an apparent lack of Cl and C (Supplementary Fig. 4 shows Cl 2p and C 1s data). Cl appears on the surface at 200 °C and reaches a stable Cl/Mg ratio of ~1 at 250 °C. According to XPS, the surface Al content is constant, leading to an Al/Mg ratio of 0.35–0.40 at low treatment temperatures. At 250 and 275 °C, Al is absent from the XPS spectra (Supplementary Fig. 4), and the Al/Mg ratio drops to zero. Simultaneously, scanning electron microscope with energy-dispersive X-ray, more sensitive to the bulk, shows a constant Al content (Supplementary Table 3). Mg₃AlO_4.5, upon interaction with PVC and its decomposition products, will cover the surface with MgCl₂ and carbonaceous deposits, while Al is in bulk, probably as unreacted Mg₃AlO_4.5 phase or a separate oxychloride phase, with particles too small to be detected by X-ray diffraction (XRD) (Fig. 3e). Formation of MgCl₂ compared with AlCl₃ is strongly preferred thermodynamically at these conditions, which drives surface segregation. Results of thermogravimetric analysis coupled with differential scanning calorimetry (TGA-DSC) (Supplementary Fig. 5) also show that Mg₃AlO_4.5 traps ~30% of HCl produced by PVC decomposition at 280–300 °C and atmospheric pressure; the remaining ~70% of HCl desorbs in the gas phase. Operation under 30 bar pressure in batch mode reinforces HCl–Mg₃AlO_4.5 interactions and shifts the HCl absorption temperature from ~300 °C (determined by TGA) to 250 °C (Fig. 3). It also increases the Cl trapping efficiency from 30% to ~100% since no HCl was detected in the reactor headspace using FTIR (Fig. 2b).

The high H₂ pressure of dechlorination enables low-temperature operation and efficient Cl removal, which are hard to achieve under atmospheric pressure attempted before³¹. Reports on PVC decomposition over ZnO (refs. ^31,32) show that more-intimate contact between the oxide and polymer reduces the decomposition temperature from 250 °C to 200 °C, preventing direct HCl evolution in the gas phase.

In addition, we tested a set of Mg-containing oxides for Cl trapping (Supplementary Table 4 and Fig. 6). Characterization shows that MgO is the active component in Mg₃AlO_4.5 adsorbent. At the same time, Al₂O₃ does not participate in Cl trapping. This was further confirmed by experiments with MgAlO_2.5 mixed oxide, which shows HCl leakage in the gas phase after dechlorination. Dispersion of MgO on silica also leads to efficient HCl capture. According to literature reports, other oxides, such as ZnO, should have comparable properties to MgO (ref. ³³).

Impact of dechlorination temperature and pressure

Dechlorination with PP–PVC mixtures followed by hydrogenolysis showcased that the reaction outcome strongly depends on the dechlorination temperature (Supplementary Fig. 7). As the dechlorination temperature increased, the accompanying PP solid yield decreased. It requires >200 °C to effectively extract Cl from the PVC to the MgCl₂ phase, consistent with the pure PVC–Mg₃AlO_4.5 results (Fig. 3). Given the high thermodynamic favourability of Mg₃AlO_4.5 chlorination with HCl, the initial thermal decomposition of PVC to polyenes and HCl is probably rate determining. TGA of pure PVC also shows that the peak decomposition happens at ~280 °C (Supplementary Fig. 5).

To elucidate the role of hydrogen in PVC decomposition and Cl trapping, we replaced H₂ with He. The hydrogenolysis stage over Ru/TiO₂ was unaltered. Figure 4a shows that the liquid yield in He is 1.4 times lower. Gel permeation chromatography (GPC) analysis (Fig. 4b) reveals that the liquid has approximately five times higher M_w when using He than when using H₂. These results clearly indicate that the Ru/TiO₂ catalyst is poisoned and produces less liquid with higher M_w when dechlorination occurs in He. Higher H₂ pressure in dechlorination (Fig. 4c) decreases the M_w of the liquid product down to ~600 Da, typical for PVC-free hydrogenolysis¹³, and maximizes the liquid yield to 69% at 30 bar H₂; beyond that, it slightly decreases the liquid due to excessive formation of gas (Supplementary Table 5).

Fig. 4: Role of hydrogen in dechlorination in the performance of the two-stage process.

a,b, Product distribution and M_w of liquid after the two stages with dechlorination conducted in H₂ or He. c, Liquid yield and M_w at various H₂ pressures. d, Reaction scheme for the formation of chlorinated alkanes. Dechlorination conditions: 250 °C, 30 bar He or H₂, 6 h, 0.2 g PVC, 0.2 g Mg₃AlO_4.5, 1.8 g PP. Reaction conditions: 250 °C, 30 bar H₂, 16 h, 0.1 g Ru/TiO₂, dechlorinated solid.

The distribution of Cl in the solid, liquid and gas products of PP is vital for their use. Supplementary Table 6 shows that most Cl is segregated selectively on the solid residue, leading to a low Cl content in the liquid (<0.1 wt%). Overall, optimization of the dechlorination conditions (Supplementary Fig. 7 and Tables 5 and 6) shows that this step requires 30 bar H₂, 250 °C and ~6 h for completion.

The promoting role of H₂ on dechlorination is rationalized in Fig. 4d. PVC decomposition without hydrogen leads to HCl and unsaturated polyenes. HCl is either trapped in MgCl₂ or undergoes a Markovnikov-type addition to C=C bonds of polyenes to chlorinated alkanes or chloroaromatic compounds, emblematic of PVC pyrolysis²⁴. Chloroalkanes are poisonous for noble metals, including Ru/TiO₂ (ref. ³⁴). Hydrogenation of polyenes and other unsaturated compounds during dechlorination is key. This partial hydrogenation rationalizes why the liquid M_w is lower in H₂ than in He. The Mg₃AlO_4.5 mixed oxides stabilize chemisorbed hydrogen atoms, leading to ethanol condensation to butanol, which involves intermediate hydrogenation of butenal¹⁹. A similar interaction between surface H atoms and strongly bonded unsaturated compounds may explain the minor hydrogenating ability of Mg₃AlO_4.5. This is also supported by the ¹³C MAS NMR spectra (Fig. 3b), where a substantial fraction of carbonaceous residues is saturated alkyls rather than the aromatic or polyenic compounds expected in direct PVC decomposition.

The negative role of chlorinated alkanes was further demonstrated with PP contaminated with 10% 1-chlorooctadecane (a model compound; Supplementary Fig. 8) without dechlorination. This decreased the yield of the liquid from 66% to 40%, the gas from 28% to ~1% and the methane selectivity from 76% in pure PP to just 4%; the main gas product was isobutane (52%). Overall, the gas product distribution (Supplementary Fig. 8) is similar to that in PP–PVC conversion.

The C–Cl bond hydrogenolysis in 1-chlorooctadecane over Ru produces HCl, similar to the chlorinated alkanes produced from PVC (Fig. 4d). Binding of HCl to the Ru surface may lead to Bronsted acid sites, catalysing cracking of hydrogenolysis intermediates, produced from PP. A similar process was detected for the HI-promoted Pd/Al₂O₃ catalyst³⁵. HCl-catalysed isomerization of adsorbed hydrocarbon radicals followed by β-scission leads to isobutane, similar to hydrocracking, where isobutane is a major product³⁶. Thus, these experiments with 1-chlorooctadecane show that chlorinated alkanes, formed from PVC, modify the Ru surface, reducing hydrogenolysis activity and gas yield. Dechlorination in H₂ does not eliminate chlorinated alkanes altogether; the residual HCl binds to Ru/TiO₂ and changes the gas formation mechanism.

Air-free XPS characterization of Ru/TiO₂ treated with 1-chlorooctadecane and PP mixture reveals that Ru remains in the metallic state, but the surface contains chloride anion on the TiO₂ support (Supplementary Fig. 9). FITR of adsorbed CO shows the formation of polycarbonyls from CO dissociation on the Ru particles. Interestingly, the peak position for the spent catalyst changes strongly compared with the fresh Ru/TiO₂, probably due to changes in the Ru electronic state. A minor amount of newly formed Brønsted acid sites was detected with DRIFTS of adsorbed pyridine (Supplementary Fig. 10). These sites forming on the support or during the reaction on Ru could catalyse cracking to C₄–C₅ gas products, instead of methane (Supplementary Table 1).

The difference in product yields and liquid microstructure (Fig. 2) observed for pure PP and PP–PVC mixture may be attributed to the Cl-induced modification of the Ru/TiO₂ catalyst, evident from experiments with 1-chlorooctadecane. Previous work also showed that the Ru catalytic activity in the hydrogenolysis of PP is sensitive to the properties of the support³⁷. Overall, variation in dechlorination conditions changes the outcome of hydrogenolysis by introducing polyene and Cl-containing species. Optimized dechlorination leads to high liquid yields, close to pure PP hydrogenolysis¹³.

Previously, two-step pyrolysis was proposed for a low-density PE–PVC mixture to decompose PVC at 300–400 °C (ref. ³⁸). In our case, dechlorination at 250 °C over Mg₃AlO_4.5 traps HCl. Hence, the exhaust gas is free of Cl (Fig. 2b). A separate filtration to purify the pyrolysis oil from Cl impurities in conventional PVC pre-decomposition is unnecessary. A trace amount of HCl or chlorohydrocarbons in the oil will prevent its use as fuel or lubricant.

To draw comparison with previous reports³⁸, mixed PP–PVC was thermally decomposed in He at atmospheric pressure and then converted over Ru/TiO₂ at 30 bar H₂. The liquid yield decreased to 26% (Supplementary Table 5). The ¹³C MAS NMR analysis (Supplementary Fig. 11) of the solid product showed a substantial contribution of polyenes and aromatics after treatment at ambient pressure. Compared with the data obtained after treatment at high pressure of H₂, the relative content of polyenes increased by an order of magnitude. The liquid Cl content obtained after ambient pressure dechlorination (X-ray fluorescence (XRF) data; Supplementary Table 6) was 2.5 wt%, much higher than observed for high-pressure experiments. Cl content decreased to 0.77 wt% when dechlorination was conducted at 20 bar He with Mg₃AlO_4.5 in batch (Fig. 4), indicating the beneficial trapping effect of Mg₃AlO_4.5 and its inability to capture Cl fully without the high pressure of H₂. In stark contrast, dechlorination at 30–50 bar H₂ led to a Cl fraction in the liquid of 0.05–0.00%. High H₂ pressure facilitates HCl trapping in MgCl₂ and prevents trace amounts of Cl from entering the liquid through C=C bond chlorination and hydrogenation (Fig. 4d).

Hydrogenolysis of PP mixed with different chlorinated compounds

PW contains PVC and other Cl-containing species from fire retardants, antioxidants or other additives³⁹. For example, brominated fire retardants during plastic pyrolysis at 450 °C contaminate the pyrolysis oil with bromophenols⁴⁰. We assessed the two-stage Cl removal for PP mixed with polyvinylidene chloride (PVDC), 1,2,4-trichlorobenzene (TCB) and 1,1,2,2-tetrachloroethane (TCE). For TCB and TCE, the catalyst was active, producing 40–60% liquid with a small amount of gas (Supplementary Table 7), giving a wholly dechlorinated liquid. PVDC contamination, however, led to no catalyst activity. Bulk PVDC decomposition produces HCl slowly due to forming stable aromatic structures with several fused rings containing C–Cl groups⁴¹. Its dechlorination may require longer treatment times or higher temperatures to eliminate HCl, but we did not pursue this optimization.

Thus, Mg₃AlO_4.5 is capable of dechlorinating PP mixtures with small, chlorinated molecules, such as TCE and TCB, while failing with PVDC. This result shows that for sustainable catalytic PP conversion to lubricants, PVDC should first be removed. PVDC–PP mixtures can be separated using solvents⁴², or ideally, PVDC should be eliminated from the plastics products²².

In addition, dechlorination was tested with higher loadings of PVC (20 and 30 wt%). Results show (Supplementary Table 8) that the liquid yield drops to 37% with 20% PVC loading while forming Cl-free liquid. Higher amounts of PVC lead to negligible liquid production.

Hydrocracking of PVC-contaminated feedstock with dechlorination

An alternative method for polyolefin upcycling is hydrocracking over bifunctional metal–acid catalysts such as Pt/WO₃/ZrO₂ (ref. ⁴³). Figure 5 shows that pure PP gives 87% liquid after 2 h reaction at 250 °C, with a product that is centred at C₇–C₈ alkanes. The PP was mixed with PVC and dechlorinated over Mg₃AlO_4.5, similar to the hydrogenolysis procedure. Then, the dechlorinated product was mixed with Pt/WO₃/ZrO₂ catalyst and produced 68% liquid with 24% solid residue after 2 h. The liquid product distribution is slightly shifted to C₆–C₇.

Fig. 5: Hydrocracking of PVC-contaminated plastic waste.

a, Product distribution from PVC–PP mixture in the two-stage and pure PP hydrocracking. b, Liquid and gas product distribution by carbon number. Dechlorination conditions: 250 °C, 30 bar H₂, 6 h, 0.2 g PVC, 0.2 g Mg₃AlO_4.5, 1.8 g PP. Reaction conditions: 250 °C, 30 bar H₂, 16 h, 0.1 g Pt/WO₃/ZrO₂, dechlorinated solid.

The position in the maximum of the product distribution (Fig. 5b) is sensitive to the ratio of metal (Pt) and acid (WO₃/ZrO₂) sites of the catalyst⁴³. The decrease in the C₈–C₁₀ yield corresponds to a reduction in the relative fraction of metal sites. Probably, Pt clusters are more vulnerable than acid sites to poisoning by the PVC decomposition products. Nonetheless, the two-stage procedure retains ~80% of liquid alkanes produced over pure feedstock.

Adsorbent and catalyst regeneration

A sustainable two-stage strategy requires the Ru/TiO₂ and Cl trap to be regenerable. A simple reduction in H₂ flow restores the activity of Ru/TiO₂, similar to standard PP hydrogenolysis (Supplementary Table 9)¹³. Here, Ru/TiO₂ was regenerated in a mixture with MgCl₂ and carbonaceous residues formed from the first cycle. This result demonstrates that MgCl₂ from the first dechlorination does not impact the hydrogenolysis activity.

Chloride can be removed from the spent Mg₃AlO_4.5 to enable the reuse of the Cl trap. Calcination in air removes 90% of Cl, and steaming at 550 °C leads to negligible residual Cl content (Supplementary Table 10). Cl removal corresponds to the high-temperature MgCl₂ hydrolysis reaction:

$${{{\mathrm{MgCl}}}}_{2}+{{\mathrm{H}}}_{2}{\mathrm{O}}\to {{\mathrm{MgO}}}+2{{\mathrm{HCl}}}$$

(1)

The HCl removed with steam can be separated from the water vapour in the downstream scrubber, following standard industrial practices. The Cl can be recuperated from the scrubber and recycled in PVC production or elsewhere. Despite changes in crystal size and morphology after steaming, the regenerated adsorbent showed complete HCl elimination from the gas phase, similar to fresh Mg₃AlO_4.5 (Supplementary Fig. 12).

Discussion

Precious metal-based catalysts in the chemical recycling of polyolefins are severely poisoned by HCl and Cl-containing compounds, preventing treating POs that contain up to 1–4% PVC. A two-stage technology was introduced to avoid catalyst poisoning. Mixtures of PO with PVC are treated at high H₂ pressure with Mg₃AlO_4.5 as a Cl trap. During this dechlorination treatment, PVC decomposes to polyenes and HCl, both adsorbed and absorbed on Mg₃AlO_4.5. HCl is absorbed as MgCl₂. Polyenes undergo partial C=C bond hydrogenation in the first stage and hydrochlorination to chloroalkanes. This mixture then reacts over Ru/TiO₂, leading to ≥65% yields of lubricant-range hydrocarbons, similar to the PVC-free case. A remarkable benefit is that methane is strongly suppressed due to the hydrogenolysis of the C–Cl bond and the release of HCl, which modifies the properties of Ru particles and TiO₂ support. Due to Cl-induced modification of the catalyst, liquid products possess different structure than pure PP hydrogenolysis at the same molecular weight.

The catalysts in this study contain expensive noble metals Ru and Pt, an obvious economic obstacle. Indeed, techno-economic analysis⁴⁴ for PE conversion to lubricants showed that precious metal affects the final product cost. Recent studies using Earth-abundant, cheap catalysts for hydroconversion, including nickel⁴⁵, cobalt⁴⁶ and even zirconia-based catalysts⁴⁷, are very promising. While these materials are much cheaper, catalyst active sites would still be vulnerable to Cl poisoning, necessitating dechlorination. Application of the proposed dechlorination strategy to these catalysts is an important future direction.

The proposed dechlorination operates at 50–150 °C lower temperature than previous analogues, saving energy, producing liquids free of Cl and eliminating downstream energy-intensive purification. Unlike dechlorination traps operating at atmospheric pressure in an inert atmosphere, our Mg₃AlO_4.5 trap operates at high H₂ pressure, facilitating HCl absorption and crystallization of MgCl₂ at milder conditions and hydrogenation of polyenes. Importantly, the approach extends to various Cl-containing contaminants, including chloroalkanes and chloroarenes. Absence of HCl emissions in the gas phase helps reducing equipment corrosion, since HCl emission and absorption are conducted in the same vessel. Energy savings and elimination of downstream separations stemming from the liquid purity are important for sustainable manufacturing. Despite having lower dechlorination temperature, more traditional Cl removal procedures, such as pyrolysis³³ or thermal extrusion⁴⁸, could be preferred in some cases. For example, if dechlorination of PP–PVC mixture is followed by pyrolysis at 400–450 °C, some of the Cl from MgCl₂ would leak at this temperature, leading to product contamination. Thus, it is better to dechlorinate at mild conditions before low-temperature catalytic hydroconversion. Importantly, when PP is contaminated with PVDC, low-temperature dechlorination does not work, and pyrolytic Cl removal should be considered^25,49.

Notably, the catalyst and the trap are regeneratable, despite some morphological changes in the crystal size of Mg₃AlO_4.5 after steaming. The high regeneration temperature to steam crack MgCl₂ imposes limitations that can counterbalance the benefits of the lower dechlorination temperature.

The importance of sorting out halogenated compounds in plastic recycling has been highlighted⁵⁰. The ability to operate on mixed contaminated feedstock will make waste sorting substantially less challenging, improving the logistics of waste collection before catalytic upcycling. This in turn would greatly reduce the cost and enable a broader application of waste hydroconversion. Combined with the availability of green hydrogen, this opens the possibility of reducing CO₂ emissions, currently associated with using hydrogen gas in hydroconversion.

Methods

Catalyst and adsorbents preparation

The Ru/TiO₂ catalyst was prepared according to a previously published procedure¹³. Before the reaction test, the catalyst was pre-reduced in a tubular furnace at 300 °C for 3 h in a 50% H₂/He gas mixture (heating rate 10 °C min⁻¹). The Mg₃AlO_4.5 was prepared as follows. Commercial hydrotalcite Mg₆Al₂(CO₃)(OH)₁₆·4H₂O (Sigma-Aldrich) was calcined in a muffle furnace in stagnant air at 550 °C for 6 h (heating rate 2 °C min^–1).

The MgO/SiO₂ was prepared by incipient wetness impregnation of aqueous Mg(NO₃)₂·9H₂O solution (Sigma-Aldrich) on silica granules (Davisil 646, Brunauer–Emmett–Teller surface area (S_BET 275 m² g⁻¹)). After impregnation, the sample was dried at room temperature for 3 h and 70 °C overnight. Then it was calcined at 550 °C in stagnant air for 6 h. The Mg–Al mixed oxides were prepared by nitrate co-precipitation according to ref. ¹⁹. Samples with Mg/Al ratios of 0.5 and 7.1 were calcined at 550 °C for 6 h and tested in dechlorination.

Adsorbents characterization

After calcination, the surface area and pore volume were estimated through Brunauer–Emmett–Teller and Barrett–Joyner–Halenda analysis of N₂ sorption isotherms at −196 °C using an ASAP 2020 instrument (Micromeritics). Before measurements, samples were degassed at 300 °C for 3 h in vacuum. Elemental composition was measured using X-ray fluorescence on a Rigaku Supermini 200 WD XRF machine with a Pd anode.

Polymer feedstock

Several polymers were mixed with PVC to test process feasibility with different feedstocks: isotactic PP (weight-averaged molecular mass (M_w) 250 kDa, number-averaged molecular mass (M_n) 67 kDa; Sigma-Aldrich 427888), amorphous PP (Sigma-Aldrich 428175) and HDPE (M_w 108 kDa; Sigma-Aldrich 427985).

Dechlorination reaction

Granules of PP (M_w 250 kDa, M_n 67 kDa; Sigma-Aldrich 427888) were mixed with PVC (M_w 43 kDa, M_n 22 kDa; Sigma-Aldrich 389293) and loaded in a 50 ml high-pressure stainless-steel Parr reactor equipped with a glass liner. Then the required amount of Mg₃AlO_4.5 was added with a Teflon-covered stir bar; the reactor was sealed, purged with pure hydrogen (or helium) three times and then charged to attain the required pressure. The reactor was placed on a hot plate for magnetic stirring and heated using a 2 inch band heater for 20–25 min to the required dechlorination temperature. After fixed time intervals, the reactor was quenched in an ice bath, and gas products were removed into a 1 l Tedlar gas bag. The solid products were removed from the liner, separated from the stir bar and used later for hydrogenolysis experiments. Gravimetric estimation showed that solid yield in the dechlorination step was close to 98–100% in all cases.

To expand the scope of dechlorination, PVDC (GoodFellow, LS554616), 1,2,4-trichlorobenzene (99%; Sigma-Aldrich) and 1,1,2,2-tetrachloroethane (Sigma-Aldrich) were mixed with PP and Mg₃AlO_4.5.

Hydrogenolysis reaction

Solid products after dechlorination were placed in a separate 50 ml Parr reactor with a Teflon-covered magnetic stir bar and 100 mg of freshly reduced Ru/TiO₂ catalyst. Then the reactor was sealed, purged with H₂ five times and charged with 30 bar H₂. It was heated with a band heater on the hot plate for 20 min to 250 °C and kept at this temperature for 16 h. After that, the reactor was quickly cooled in an ice bath.

After the reaction and product extraction, the reactor was thoroughly washed with hexane, acetone and deionized water. Then it was filled with 0.05 M HCl water solution and stirred for 3 h to remove contamination. Afterwards, the reactor was washed with water and acetone, dried and scrubbed with an abrasive pad.

Experiments with a model compound, 1-chlorooctadecane (98%; Sigma-Aldrich), were conducted. We mixed 1.8 g of PP with 0.1 g of freshly reduced Ru/TiO₂ catalyst in a Parr reactor. Then 0.3 g of 1-chlorodecane were added, and the reactor was closed, purged with pure H₂ and charged with 30 bar H₂. After 16 h of reaction at 250 °C, product extraction and analysis were conducted as usual.

Dechlorination in a flow reactor

For comparison, dechlorination was conducted in a flow reactor at ambient pressure. We mixed PP with PVC and Mg₃AlO_4.5 in 1.8:0.2:0.2 proportions by weight. The mixture was loaded in a quartz boat in a horizontal furnace. It was then heated in He flow (100 ml min^–1) to 250 °C for 6 h. Then the sample was cooled to room temperature and transferred to the Parr reactor, mixed with 0.1 g of Ru/TiO₂ hydrogenolysis catalyst.

Product analysis

Product analysis was conducted according to a previously published procedure³⁷. Briefly, gas from the reactor headspace was taken to a Tedlar gas bag. Liquid and solid products were separated using filtration with CH₂Cl₂, used as a solvent. The solid residue was then dried at room temperature. The liquid was cleaned from the solvent on a rotary evaporator. Gas products were quantified using a gas chromatography with flame ionization detector (HP-Plot Q 30 m column), calibrated with C₁–C₄ gas standards (Supelco 303100-U, 307300-U). The solid and liquid fraction yields were quantified gravimetrically. The yield of the i^th group of products was calculated according to equation (2):

$${Y}_{i}=\frac{{m}_{i}}{{m}_{{{\mathrm{initial}}}}}\times 100 \%,$$

(2)

where m_i is the mass of the gas, liquid or solid, m_initial corresponds to the initial mass of PP and PVC, and the mass of solid was corrected for the mass of catalyst and adsorbent. Repetitive runs show yield deviation within ~10%.

The weight of the solid residue was calculated gravimetrically by subtracting the catalyst and Mg₃AlO_4.5 masses. The total weight of all products was used to calculate the mass balance:

$${{\mathrm{MB}}}=\frac{{\sum }_{{{\mathrm{product}}}}{m}_{i}}{{m}_{{{\mathrm{iniital}}}}},$$

(3)

where MB is the mass balance, m_i is the mass of the i^th product group (gas, liquid or solid residue) and m_initial is the mass of PP and PVC. In all two-step dechlorination experiments, MB was above 95%.

The total Cl content in the product was calculated as the sum of Cl in the solid residue and in the liquid. Tests with Mg₃AlO_4.5 did not produce any detectable amounts of HCl or chlorohydrocarbons in the gas phase. The Cl balance (ClB) was calculated as follows:

$${{\mathrm{ClB}}}=\frac{{m}_{{{\mathrm{liq}}}}^{{{\mathrm{Cl}}}}{\omega }_{{{\mathrm{liq}}}}^{{{\mathrm{Cl}}}}+\left({m}_{{{\mathrm{solid}}}}^{{{\mathrm{Cl}}}}+{m}_{{{\mathrm{RuTi}}}{{\mathrm{O}}}_{2}}+{m}_{{{{\mathrm{Mg}}}}_{3}{{\mathrm{Al}}}{{\mathrm{O}}}_{4.5}}\right)\omega _{{{\mathrm{solid}}}}^{{{\mathrm{Cl}}}}}{{m}_{{{\mathrm{Cl}}}}^{{{\mathrm{initial}}}}},$$

(4)

where ${m}_{{{\mathrm{Cl}}}}^{{{\mathrm{initial}}}}$ is the initial mass of Cl in PVC (equal to 56.8 wt% of PVC loading); ${m}_{{{\mathrm{liq}}}}^{{{\mathrm{Cl}}}}$ is the mass of liquid products, ${\omega }_{{{\mathrm{liq}}}}^{{{\mathrm{Cl}}}}$ is the Cl weight fraction in liquid products estimated by XRF, ${m}_{{{\mathrm{solid}}}}^{{{\mathrm{Cl}}}}$ is the mass of solid residue,$\,{m}_{{{{\mathrm{Mg}}}}_{3}{{\mathrm{Al}}}{{\mathrm{O}}}_{4.5}}$ is the initial mass of Mg₃AlO_4.5 and ${\omega }_{{{\mathrm{solid}}}}^{{{\mathrm{Cl}}}}$ is the Cl weight fraction in solid residue estimated by XRF. In all experiments with Mg₃AlO_4.5 as Cl trap, ClB reached ~90%.

Liquid products were analysed using GPC with high-resolution (HR) 4, HR 3 and HR 0.5 columns, tetrahydrofuran as a mobile phase and a refractive index detector. The retention time was calibrated using a polystyrene standards kit (Waters, WAT058931).

Adsorbent–PVC interactions

The PVC was mixed and dechlorinated directly with Mg₃AlO_4.5 without adding PP. The powder formed after dechlorination was dried overnight at room temperature. The XRD patterns were collected on Bruker D2 Phaser instrument with Cu tube, 0.05° 2θ step size and 3 s dwell. The ¹³C MAS NMR spectra were acquired on an 11.7 T Bruker Avance III NMR spectrometer with a 4 mm HX probe with a magic angle spinning at 14 kHz. Direct ¹³C excitation was achieved with a 3.5 μs 90° pulse. The XPS spectra were measured on a Thermo K-alpha instrument with Al Kα radiation. Due to charging, the peak position was not corrected, and only the intensity was used to monitor the change in elemental composition on the surface. Scanning electron microscope images and energy-dispersive X-ray analysis were acquired on an Auriga 60 CrossBeam high-resolution microscope with accelerating voltage at 10 kV with the sample deposited on carbon tape.

Characterization of spent Ru/TiO₂ catalysts

Changes in Ru/TiO₂ catalyst induced by chlorine were analysed after the reaction. Initially, 1.8 g of PP mixed with 0.3 g of 1-chlorooctadecane reacted over Ru/TiO₂ at 250 °C for 16 h at 30 bar H₂. After the reaction, the solid was separated with filtration and dried. Then the solid was transferred to a glass vial, mixed with 20 ml toluene and heated to 90 °C for 30 min, until all PP was dissolved. The liquid toluene was decanted with a glass pipette. The remaining powder was primarily spent Ru/TiO₂, contaminated with organic residue and Cl. This powder was dried in a vacuum oven overnight at 90 °C to remove residual toluene.

The sample was transferred to a tubular quartz reactor and heated in 10% H₂/He flow at 300 °C for 2 h (ramp rate 10 °C min^–1). Then the powder was moved to a glovebox (<100 ppm oxygen) and loaded on copper tape in a vacuum transfer vessel. Then XPS spectra were measured using Thermo Fisher K-Alpha Instrument. The binding energy standard was O 1s at 529.9 eV, typical for TiO₂. Transmission FTIR spectra of adsorbed CO were recorded on Nicolet 8700 spectrometer with a mercury cadmium telluride detector and a homemade tubular flow Pyrex cell with KBr windows. Samples were pressed in self-supported wafers (1.26 cm², 80 bar cm⁻² pressure). The sample was pre-reduced in 10% H₂/He flow at 300 °C for 1 h, purged with pure He for 1 h and then cooled to 35 °C. Then the sample was contacted with 1% CO/He mixture for 10 min and flushed with pure He for 5 min followed by spectra recording. DRIFT spectra were recorded using a Harrick Praying Mantis cell, following a similar sample pre-reduction. After purging with He for 1 h at 300 °C, the sample was cooled to 150 C, and 2 μl liquid pyridine was injected in a He stream. Spectra are reported as differences between the sample before and after pyridine chemisorption. Sequential injections were used to ensure complete saturation of the surface with pyridine.

Cl distribution analysis

Chlorine has been measured using XRF for liquid and solid samples. For liquids, decalin and 1,2,4-trichlorobenzene (Sigma-Aldrich) were used to calibrate Cl Kα signal intensity in a broad concentration range. For solids, PVC mixed with parent Mg₃AlO_4.5 was used to calibrate Cl, Al and Mg content in the solid residue after the reaction. In the gas phase, the HCl content was estimated using FTIR spectroscopy on a Nicolet iS50 spectrometer with a 100 mm gas cell equipped with 38 mm CaF₂ windows (Pike Technologies). The cell was evacuated at ~1 Torr and connected to a 1 l Tedlar gas sampling bag. Then the cell was filled with a sample gas and closed. Spectra were acquired at 4 cm⁻¹ resolution, 32 scans with a deuterated triglycine sulfate detector.

Catalyst and adsorbent regeneration

To simulate severe Cl contamination, Mg₃AlO_4.5 was treated with pure PVC, and 0.4 g Mg₃AlO_4.5 and 0.4 g PVC were loaded in a Parr reactor, charged with 30 bar H₂, heated at 250 °C for 6 h. Then the reactor was cooled in an ice bath, the gas was vented and the powder was removed and later calcined in a muffle furnace at 550 °C for 4 h in stagnant air. To remove traces of Cl, the sample was steamed after calcination. The sample was packed in a 12.7 mm quartz tubular reactor with quartz wool and purged with pure He (5 ml min^–1) for 30 min. Then it was heated to 550 °C in He with a ramp rate of 5 °C min⁻¹. Liquid water was administered into the reactor using a syringe pump operating at 0.5 ml min⁻¹ at 550 °C for 2 h. After that, the sample was purged with pure He for 1 h and cooled to room temperature.

After the standard hydrogenolysis reaction, spent Ru/TiO2 was mixed with spent Mg3AlO4.5, and the solid residue was washed with 50 ml CH₂Cl₂ and then dried overnight at 100 °C. This solid mixture was transferred to a tubular furnace, heated to 300 °C with a ramp rate of 10 °C min⁻¹ and dwelled for 4 h in a 50% H₂/He mixture. Then the mixture was mixed with fresh 2 g PP and loaded in a Parr reactor. The reactor was charged with 30 bar H₂ and then heated to 250 °C for 16 h. Products were analysed according to the standard procedure.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.