Mechanistic classification and benchmarking of polyolefin depolymerization over silica-alumina-based catalysts

Carbon-carbon bond cleavage mechanisms play a key role in the selective deconstruction of alkanes and polyolefins. Here, we show that the product distribution, which encompasses carbon range and formation of unsaturated and isomerization products, serves as a distinctive feature that allows the reaction pathways of different catalysts to be classified. Co, Ni, or Ru nanoparticles immobilized on amorphous silica-alumina, Zeo-Y and ZSM-5, were evaluated as catalysts in the deconstruction of n-hexadecane model substrate with hydrogen to delineate between different mechanisms, i.e., monofunctional- (acid site dominated) or bifunctional-hydrocracking (acid site & metal site) versus hydrogenolysis (metal site dominated), established from the product distributions. The ZSM-5-based catalysts were further studied in the depolymerization of polyethylene. Based on these studies, the catalysts are plotted on an activity-mechanism map that functions as an expandable basis to benchmark catalytic activity and to identify optimal catalysts that afford specific product distributions. The systematic approach reported here should facilitate the acceleration of catalyst discovery for polyolefin depolymerization.

. In addition to hydrocracking via silica-alumina materials, tandem catalysis approaches combining precious metal clusters immobilized on other solid acid supports (i.e., not silica-alumina structure) have also been reported. For example, Liu et al. were able to convert LDPE to gaseous and liquid hydrocarbons in 9 and 83% yield, respectively, by applying 1 wt% Pt loaded on WO 3 21 . However, the tandem catalysis has not yet been resolved in detail or struck by a consensus as to the silica-alumina-based catalysts and, thus, will not be intensively studied and incorporated in our following studies.
Hydrogenolysis of C-C bonds using Ru-based catalysts has also been used for the selective deconstruction of plastics, limiting isomerization and unsaturated products. The formation of methane is considered an important indicator of a hydrogenolysis mechanism with alkylidyne intermediates, as C 1 intermediates such as the methenium cation are disfavored in (hydro)cracking mechanisms 22 . Ru NPs on activated charcoal (Ru/C) has been shown to cleave PE into shortened hydrocarbons under H 2 (20-60 bar) at 200-225°C 23 24 . Ru NPs immobilized on a reducible solid support such as TiO 2 have also been reported to catalyze PP hydrogenolysis (Table 1, Entry 7) 25 . Pt NPs immobilized on metal oxide (SrTiO 3 ) or fabricated into a mesoporous shell/active site/core structure (mSiO 2 /Pt/SiO 2 ) have also been used in the hydrogenolysis of PE to afford wax-lubricant range products (C 18-40+ ) under 9-14 bar H 2 at temperatures ranging from 250 to 300°C reacting for 24-96 h [26][27][28] .
Despite the abundant literature describing catalysts that deconstruct alkanes and polyolefins, systematic comparisons among the catalysts are largely absent and their mechanistic pathways are often overlooked, which hinders further rational catalyst design. To overcome these limitations, we studied the catalytic deconstructions of nhexadecane (nC 16 -used as a model substrate and functioning as a benchmark reaction) and polyethylene (PE, Mw~4000 g/mol) in the presence of hydrogen using various heterogeneous catalysts prepared using an identical synthetic procedure. We use the product distribution, including the carbon range and distinctive features such as degrees of saturation and the formation of isomerization products, were evaluated 22 , to classify the C-C bond cleavage mechanisms as: 1. Monofunctional hydrocracking-acid site dominated cleavage with a unimodal distribution comprising C 3-4 as major products including significant isomerization and unsaturated products. 2. Bifunctional hydrocracking-combined acid/metal site cleavage with a unimodal distribution with >C 3-4 as major products including isomerization and unsaturated or a bimodal distribution product range in certain cases. 3. Hydrogenolysis-metal site dominated cleavage with a unimodal distribution with C 1 as the major product.
A schematic summarizing the product distributions for each type of mechanism is given in Fig. 1. Furthermore, these studies allowed an activity-mechanism map to be constructed for catalyst benchmarking, which should enable new, superior polyolefin depolymerization catalysts to be developed that provide specific products or distributions.
All SiO 2 -Al 2 O 3 -, Zeo-Y_H-, and ZSM-5_H-based catalysts were analyzed by powder-XRD, confirming the amorphous nature of SiO 2 -Al 2 O 3 -based catalysts ( Supplementary Fig. 1) and the crystalline nature Scanning electron microscopy (SEM) images show that the particle size of all the ZSM-5_H-based catalysts ranges from 6 to 10 μm (Fig. 2a-d) and no significant changes compared to the support material are observed following the immobilization of Co, Ni, or Ru NPs, and the surface morphology is also maintained (cf. Fig. 2a-d with Supplementary Fig. 7). Transmission electron microscopy (TEM) images of ZSM-5_H show a smooth surface with no apparent surface structure ( Supplementary Fig. 8). Well-dispersed NPs were observed on the surface of Co/ZSM-5_H and Ni/ZSM-5_H (Fig. 2e, f), and more aggregated NPs were formed on the surface of Ru/ZSM-5_H (Supplementary Fig. 9). High-angle annular dark-field (HAADF) images also reveal the surface fine-particle structures of Co/ZSM-5_H, Ni/ZSM-5_H, and Ru/ZSM-5_H catalysts. The average particle size of the Co, Ni and Ru NPs was estimated as 3.6 ± 0.5, 3.1 ± 0.4, and 5.4 ± 0.8 nm, respectively (see further details in Supplementary Figs. 9-11). Similar particle distributions of Co and Ni NPs ensure that the further reactivity comparisons are mainly based on the types of metal modification instead of acid/metal site intimacy during their working state (hydrogen atmosphere with heat, see below) 41 . The more aggregated particle distribution, on the other hand, rationalizes the relatively lower Ru surface atomic concentration than Co and Ni via XPS analysis (2.6 vs. 9.8 and 8.3%).

Activity studies
In the initial phase of the study n-hexadecane (nC 16 ) was used as a model substrate 23 for PE, a saturated aliphatic polymer. Each catalyst was evaluated under sufficient H 2 (~1.15 eq.) to quantitatively produce methane, to ensure that selectivity would not be restricted by a lack of hydrogen, and at three temperatures typical of catalytic hydrocracking reactions, i.e., 275, 325, and 375°C 16,17,42,43 . The results from the main library catalysts are summarized in Table 2 and Supplementary Figs. 12-15 provide detailed product distributions, and Supplementary Table 3 provides the carbon balance in weight and hydrogen consumption.

Unmodified silica-alumina catalysts
Under the typical conditions used (nC 16 Entries 8,9). Light hydrocarbon (C 1-5 ) products, with C 3 as the major products, were identified in all the reactions ( Supplementary Fig. 12). Significant portions of isomerization products were detected, and limited methane was obtained confirming a monofunctional hydrocracking mechanism. Under the reaction conditions employed, the reactivity of the silica-alumina materials follows the order: ZSM-5_H > Zeo-Y_H > SiO 2 -Al 2 O. The differences in activity among amorphous SiO 2 -Al 2 O 3 , crystalline Zeo-Y_H, and crystalline ZSM-5_H at different temperatures ( Table 2, Entries 1-9) reveal the significance of local confinements and the corresponding topology of a given silicaalumina material towards the activity and selectivity of the catalyst (see below) 44,45 .
Control experiments were conducted in which the reaction of nC 16 was studied under a nitrogen atmosphere in the presence of the unmodified silica-alumina catalysts at 375°C (Fig. 3a and Supplementary Fig. 16). Interestingly, the nC 16 conversion in the presence of ZSM-5_H was reduced under the nitrogen atmosphere compared to the reaction under hydrogen, and a wider unimodal product distribution was obtained (cf.  Fig. 16-right). The major products consist of short carbon range (C 3-4 ) hydrocarbons with a limited amount of methane, confirming a monofunctional cracking mechanism under nitrogen. Despite the absence of an external hydrogen source when the reaction is carried out under a nitrogen atmosphere, the hydrogen atoms in the hydrocarbon substrate are able to be transferred, to afford saturated products together with the corresponding unsaturated products resulting from the hydrogen transfer 46 . The different applied atmospheres do not lead to significantly different C-C bond cleavage pathways. Nevertheless, under a hydrogen atmosphere, the reaction should be classified as hydrocracking due to the influence of the hydrogen on the kinetics and product distribution, i.e. wider vs. narrower unimodal distribution under N 2 or H 2 atmospheres 47 . Despite the unmodified silica-alumina catalysts lacking metal sites for efficient hydrogenation/ dehydrogenation steps, the hydrogen may affect the concentration and diffusion behavior of surface species 48 .

Immobilized Co and Ni NP catalysts
The Co/SiO 2 -Al 2 O 3 and Ni/SiO 2 -Al 2 O 3 catalysts only result in low conversions of nC 16 , i.e., <5%, similar to that observed with the silicaalumina supports alone under the benchmark conditions at 375°C ( A bimodal product distribution including isomerized hydrocarbons was obtained with Co/ZSM-5_H and Ni/ZSM-5_H, with both affording C 3-4 and C 7-8 hydrocarbons as major products (Fig. 3b- 13,14), which contrasts with a unimodal distribution dominated by C 3-4 products when ZSM-5_H was applied under the typical conditions (cf. Fig. 3a-left with 3b-left, 3cleft). The change in the product distributions may be appreciated by considering the proportion of liquid to gas ratio of the products (L/G ratio = C 5-16 yield/C 1-4 yield), which corresponds to 2.9 for Co/ZSM-5_H (conv. = 49.0 ± 3.7%) and 2.0 for Ni/ZSM-5_H (conv. = 85.6 ± 4.1%) compared to 0.3 for ZSM-5_H (conv. = 98.0 ± 2.0%). The bimodal product distribution and higher carbon range C 7-8 products obtained using Co/ZSM-5_H and Ni/ZSM-5_H implies a bifunctional pathway. A set of controls with a prolonged reaction time of 6 h in the presence of Co/ZSM-5_H and Ni/ZSM-5-H was conducted to establish whether the bimodal product distribution is due to an insufficient reaction time. Both Co/ZSM-5_H and Ni/ZSM-5_H result in near-quantitative conversion of the nC 16 substrate and bimodal product distributions, i.e., a L/G ratio of 1.9 for Co/ZSM-5_H (conv. = 93.1 ± 4.3%) and a L/G ratio of 2.0 for Ni/ZSM-5_H (conv. = 99.8 ± 0.2%), confirming a bifunctional hydrocracking mechanism (cf. Fig. 3a-left with 3b-right, 3c-right). The Co and Ni NPs are unable to cleave C-C bonds efficiently, and therefore the catalytic activity is mainly determined by the acid sites when the reaction is conducted above the onset temperature of the silicaalumina support, evidenced by the proportion of C 3 product for the hydrogenolysis of nC 16 using Co/ZSM-5_H or Ni/ZSM-5_H (Fig. 3b, c). Hence, Co/ZSM-5_H and Ni/ZSM-5_H are more active than the Zeo-Y_Hbased counterparts, i.e., Co/Zeo-Y_H and Ni/Zeo-Y_H, due to the higher activity of the ZSM-5_H support, and despite having identical NP coverages. Such interplay between the immobilized NPs and surface is often referred to as the acid-metal balance and can further affect product selectivity, in these cases not specifically favoring C 4 products 49,50 .
Although the catalysts composed of Co or Ni NPs immobilized on Zeo-Y_H or ZSM-5_ H are less active than the zeolite supports alone ( Table 2, cf. Entries 4-9 with 13-18 and 22-27), the extent of unsaturated products (i.e., alkenes, aromatics, etc., determined using 1 H NMR spectroscopy) is lower (see further details in Table 3). This difference is because the Co and Ni NPs efficiently catalyze the hydrogenation of unsaturated bonds formed during the cracking process. For example, the degrees of saturation of the resulting liquids using ZSM-5_H under the typical conditions correspond to 87.4, 7.9, and 4.8% of saturated, unsaturated, and aromatic hydrocarbons, respectively (  Supplementary Fig. 17). With Ni/ZSM-5_H, 98.6% and 99.9% of saturated products are obtained at 30 and 60 bar of H 2 , respectively ( Table 3, Entries 8,9, and Supplementary Fig. 18). The higher the hydrogen pressure shows the higher the percentage of saturated products, though only to a modest extent. In contrast, the metal modification has a greater influence, significantly increasing the amount of saturated products in the final product distribution. The higher content of saturated products is a distinctive feature of the bifunctional pathway, which reduces potential coking through the suppression of coking precursors such as polyaromatic compounds [51][52][53] . Note that as the concentration of unsaturated products increases, the color of the solution also becomes more intensely yellow in color (Fig. 3d).
The proximity of the acid sites and metal sites was investigated by comparing the products obtained using ZSM-5_H + Ni/γAl 2 O 3 (Nimodified gamma-alumina prepared using identical procedure as Ni/ ZSM-5_H) and Ni/ZSM-5_H + γAl 2 O 3 as catalysts 54 . ZSM-5_H + Ni/γAl 2 O 3 may be considered a low intimacy combination and Ni/ZSM-5_H + γ Al 2 O 3 as a high intimacy combination. The low intimacy combination resulted in near-quantitative conversion with a unimodal distribution indicative of a monofunctional pathway typical of ZSM-5_H (cf. Fig. 3aleft with 3e-left), whereas the high intimacy combination led to a conversion of 93.9 ± 0.3% with a bimodal distribution similar to the bifunctional pathway observed for Ni/ZSM-5_H (cf. Fig. 3c-left with 3eright). These results confirm the importance of acid and metal site proximity in influencing the product distribution as reported elsewhere 41,54,55 . Note that γAl 2 O 3 shows some activity (conv. = 6.8% ± 0.9) in the transformation of nC 16 at 375°C, explaining the slightly higher activity of the Ni/ZSM-5_H + γAl 2 O 3 combination compared to Ni/ZSM-5_H alone. Moreover, the Co and Ni metal sites appear to provide an alternative route facilitating the dissociation of intermediates away from the acid sites that would otherwise be cleaved further by them to afford higher degrees of saturated products in higher carbon ranges.

Immobilized Ru NP catalysts
The three Ru-modified catalysts all result in the near-quantitative conversion of nC 16 with high selectivity to methane, with high conversions even obtained at 275°C (  Fig. 15). The high yield of methane and trace amount of isomerized products in the C 5-12 range (Table 2, Entries 28, 31) indicate that a hydrogenolysis mechanism is dominant 23,24,56 . Note that evaluating metal NPs dispersed on an inactive support material may provide insights on the onset temperature for C-C bond hydrogenolysis of a given metal. Ru NPs immobilized on carbon (Ru/C) were shown to depolymerize PE via a hydrogenolysis mechanism with an onset temperature around 220°C, whereas Pt/C, Pd/C, and Rh/C require temperature ≥280°C 23,24 . As Ru NPs efficiently catalyze C-C bond Reaction conditions: n-hexadecane (1.59 g, 7.0 mmol), catalyst (0.1 g, metal loading = 2.5 wt%), 375°C. Note that degrees of saturation are defined by the ratio of proton integrations in 1 H NMR spectra to indicate the adjacent carbon-carbon bonds (saturated: δ = 0.25-2.0, unsaturated: δ = 2.0-6.0, and aromatics: δ = 2.0-6.0) given the C-H and C-C bond exclusivity of hydrocarbons.
hydrogenolysis, the activity and selectivity are dominated by the Ru NPs, especially where the silica-alumina support is less active, e.g., SiO 2 -Al 2 O 3 , and below the onset temperature of the support. Traces of isomers and chain-end initiated linear alkanes were identified at 325°C (Supplementary Fig. 19), indicative of hydrocracking attributed to the silica-alumina supports (ZSM-5_H in this case). Therefore, the kinetics of the catalysts for methane production follows the order: Ru/ZSM-5_H (88.6%) > Ru/Zeo-Y_H (76.4%) > Ru/ SiO 2 -Al 2 O 3 (54.5%) at 325°C. With the exception of Ru, the influence of a metal on the selectivity of the reaction is less predictable, especially when the hydrogenolysis onset temperature is close to the temperature required by the support to initiate hydrocracking. Thus, the overhead of reaction temperature with respect to hydrogenolysis/ hydrocracking onset temperature critically influences which mechanism prevails. For instance, ultra-stable Y zeolite and beta zeolite modified with Pt NPs (i.e., ≥280°C for hydrogenolysis) behave as bifunctional hydrocracking catalysts at around 300°C 20 .

Impact of the Si/Al ratio
Zeo-Y_H-based catalysts with different Si/Al ratios (SARs) were also evaluated in the deconstructions of nC 16 at 375°C, see Table 4 and Supplementary Fig. 20 for the full product distributions, and Supplementary Table 4 for carbon balance in weight and hydrogen consumption. Zeo-Y_H [60] (Si/Al = 60:1) and Zeo-Y_H [80] (Si/Al = 80:1) result in conversions of 24.6 ± 1.2% and 14.9 ± 0.6% compared to the 26.7 ± 1.4% for Zeo-Y_H (Si/Al = 30:1) under the standard reaction conditions, i.e., 45 bar H 2 , 375°C, 2 h (Table 4, Entries 1-3). As the SAR increases the catalytic activity decreases, which may be associated with their surface acidity, and is in accordance with the literature 21 . The presence of isomerized and unsaturated products confirm that a monofunctional pathway is in operation (Supplementary Table 5, Entries 2, 3) as would be expected. The differences in activity between the different types of zeolites with comparable SARs (Si/Al = 20-30) are more considerable, i.e., Zeo-Y_H results in a conversion of 26.7 ± 1.4% whereas ZSM-5_H results in a conversion of 98.0 ± 2.0% under the standard reaction conditions (cf. Table 2, Entries 6, 9 with Table 4, Entries 2, 3). Such difference also reveals that the confinements originating from the zeolite topology play a more dominant role than the SAR 57 . Furthermore, Zeo-Y_H shows selectivity toward higher carbon range products due to the larger pore sizes whereas ZSM-5_H favors the C 1-4 hydrocarbons which may be attributed to smaller pores, typically referred as shape selectivity and correlated to the zeolite topology 30,58,59 .

PE depolymerization
The catalysts employing ZSM-5_H as the support material are the most active and were further used to depolymerize PE (Mw~4000 g/mol, Sigma-Aldrich). ZSM-5_H results in a conversion of 95.8% (PE:~1.6 g, 112 mmol of carbon, catalyst: 0.2 g, H 2 : 45 bar, reaction time: 2 h, solvent-free) at 375°C. The product distribution is similar to that obtained with nC 16 , i.e., with C 3-4 hydrocarbons as the major products (Fig. 4a) and unsaturated liquid products, confirming a monofunctional hydrocracking mechanism. Under identical reaction conditions, quantitative depolymerization of PE to methane was observed in the presence of Ru/ZSM-5_H, clearly demonstrating a hydrogenolysis mechanism (Fig. 4b).
Low conversions of 13.3 and 27.1% were obtained at 375°C in the presence of Co/ZSM-5_H and Ni/ZSM-5_H, respectively (Fig. 4c, d). Prolonging the reaction time to 16 h increases the conversion of PE to 66.7 ± 4.4% for Co/ZSM-5_H and 87.1 ± 7.5% for Ni/ZSM-5_H (Fig. 4e, f). With the prolonged reaction time of 16 h, unimodal product distributions were observed with C 3-4 hydrocarbons being the major products using Co/ZSM-5_H and Ni/ZSM-5_H under solvent-free conditions. Inherent mass-and heat-transfer limitations of the solvent-free reaction involving a polymeric starting material [60][61][62] , could be mitigated by employing a solvent 63,64 . Hence, nC 16 was applied as a reactive solvent (note that the total carbon molar number was maintained, i.e., nC 16 + PE =~1.4 + 0.2 g corresponding to~112 mmol of carbon, catalyst = 0.1 g). Significant coking was not observed presumably due to the short reaction time and the application of hydrogen (Tables 2, 4),   Fig. 21). No apparent differences in TGA curves above 500°C were observed, indicating a well-closed carbon balance by the gaseous and liquid products due to limited solid residual formation during a reaction. The use of nC 16 a solvent is also expected to prevent coke formations and to preserve catalytic performance for the more challenging polymer substrates over longer-term operation 65,66 , particularly in the cases via highly acidic catalysts [67][68][69] . Nonetheless, regeneration processes for coke removal could be applied if required 70,71 .
Co/ZSM-5_H shows the conversion of 25.4 ± 4.3%, whereas Ni/ ZSM-5_H results in a near-quantitative conversion of 99.3 ± 0.6% at 375°C with a reaction time of 6 h (Fig. 4g, h). No significant differences in the product distribution or selectivity were observed with Co/ZSM-5_H under solvent-free or solvent-assisted depolymerization conditions, presumably due to the general low activity of the applied catalyst (cf. Fig. 4e with 4g). However, the selectivity shifts toward longer hydrocarbons with Ni/ZSM-5_H under solvent-assisted conditions, resulting in C 5-6 hydrocarbons as the main products (cf. Fig. 4f with  4h). The intrinsic activity of a given catalyst, therefore, has more significant effects on transfer issues toward selectivity and should be considered a primary factor. Despite the shift toward a higher carbon range when employing the solvent-assisted condition, it is worth noting that the resulting unimodal distribution with C 5-6 as the major products after PE depolymerization in the presence of Ni/ZSM-5_H differs from the bimodal distribution obtained for nC 16 (cf. Fig. 3c-right  with 4h). Co-and Ni-modified catalysts afford predominantly saturated products following the PE depolymerizations, indicating a bifunctional hydrocracking mechanism for PE depolymerization in the presence of Co/ZSM-5_H and Ni/ZSM-5_H under solvent-assisted conditions. These, however, revealed a more sensitive nature of the bifunctional pathway upon transfer issues than the two other pathways. Recent studies also demonstrated that the migration of intermediates between the acid and metal sites affects reactivity as well as selectivity, and rationalizes the sensitive nature of a bifunctional pathway compared to the monofunctional and hydrogenolysis pathways which have more localized active sites 53,54 .

Activity-mechanism map
The assignment of the principal mechanism for each catalyst is based on the product distribution, i.e., carbon range, isomerization, and unsaturated products. The catalysts were plotted on an activitymechanism map in a polar coordinate system (Fig. 5a) for the reactions of nC 16 deconstructions under 45 bar H 2 at 375°C with a reaction time of 2 h (i.e., the benchmark conditions). Differences in activity (r) and mechanistic pathway (θ) are indicated by the map (see further details in Methods), allowing rapid benchmarking of new catalysts and enabling the prediction of specific product distributions for depolymerizations.
The types of metal modifications strongly influence the principal cleavage mechanism (Fig. 5b-e). All unmodified silica-alumina supports function via a similar mechanism as confirmed by the similar product distributions (cf. Fig. 5a with Supplementary  Fig. 12). Immobilization of Co and Ni NPs on the silica-alumina supports switches the mechanism to bifunctional hydrocracking with the Ni-modified catalysts being more active than the Coanalogs (cf. Fig. 5c with 5d, e). The catalysts containing immobilized Ru NPs operate predominantly via hydrogenolysis as methane is the major product (cf. Fig. 5a with Supplementary Fig. 15). The types of silica-alumina support affect the rate of reaction considerably via the acidic and confinement of the support structure but not the principal mechanism (cf. Fig. 5b-e with Supplementary Fig. 22) [72][73][74][75] . The Si/Al ratio influences the catalytic activity but to a lesser degree than the type of metal NPs used and the topology of the silicaalumina support (Supplementary Fig. 23). In contrast, the applied reaction temperature has shown no apparent patterns for predicting activity and prevailing mechanism ( Supplementary  Fig. 24), unless the onset temperatures of the catalytic components are known.

Discussion
The depolymerization of polyolefins is an attractive way to recycle/ reuse the plastic waste in comparison to mechanical recycling to afford high-quality products [76][77][78] and has a greater impact on sustainability than landfilling 79,80 . Despite the current cost of hydrogen, which is predicted to decrease 81,82 , the ability to fine-tune the reaction mechanism and hence the product distribution is highly valuable. Moreover, as a direct consequence of the reductive environment, the process is expected to be cleaner and result in less coke formation than alternative methods such as incineration and pyrolysis [83][84][85] .
In order to transfer the process from the laboratory scale to the industrial arena, new catalysts with superior activity and selectivity to the current range are required. To this end, we classified depolymerization mechanisms according to the product distribution, which includes the carbon range, isomerization, and degrees of saturation. Based on the study of 18 catalysts, an activity-mechanism map was constructed according to the defined benchmark conditions employing nC 16 as a model substrate. The activity-mechanism map serves three major functions. First, it can be used to classify catalysts that afford a specific product distribution. Second, new catalysts can be plotted on the map to allow their activity and mechanism to be gauged relative to the known catalysts. Third, the optimal components, i.e., support, metal type, etc., to provide a specific product distribution can be ascertained from the map. We expect this benchmarking strategy accelerates the development of new catalysts for both plastic-to-fuel and plastic-to-chemical scenarios.

Typical synthesis of metal-modified catalyst
The support material (1.50 g) and a metal salt (CoCl 2 , 82.7 mg, 0.64 mmol; NiCl 2 ·6H 2 O, 151.8 mg, 0.64 mmol; RuCl 3 ·3H 2 O, 97.1 mg, 0.37 mmol) were placed in a round bottle flask (100 mL) with a magnetic stir bar and then dispersed in deionized (DI) water (36 mL). The suspension was stirred at 500 rpm at 35°C for 1 h. NaBH 4 (70.1 mg, 1.85 mmol for RuCl 3 ·3H 2 O; 120.8 mg, 3.19 mmol for CoCl 2 and NiCl 2 ·6H 2 O; 5 eq. to the applied metal) was dissolved in DI water (36 mL) and added to the suspension in one portion. The suspension was stirred at 800 rpm for 1 h. The resulting solid was vacuum filtered and washed with DI water (3 × 25 mL). The solid was then dried at 100°C in an oven for 18 h. Various batches of the catalyst were prepared, characterized, and evaluated, and no significant differences were observed. Typical reaction of n-hexadecane deconstruction n-Hexadecane (1.59 g, 7.0 mmol; 112 mmol based on carbon) and the catalyst (0.10 g) were added to a glass vial (20 mL) with a glass magnetic stir bar, and the vial was then placed into an autoclave (75 mL, Parr Instrument). The autoclave was purged three times with H 2 and then pressurized to 45 bar and sealed. The pressurized autoclave was placed in a heating block at the desired temperature (275-375°C) and stirred (350 rpm) for the given reaction time (2-6 h). After the reaction, the autoclave was cooled to room temperature in a water bath. The gaseous products were transferred into a gas-sampling bag (1 L) and injected into a specialized gas-sampling gas chromatography-flame ionization detector (GC-FID) for analysis. For GC-mass spectrometry (GC-MS) analysis of the liquid products, n-dodecane (0.085 g, 0.5 mmol) was added to the vial as an internal standard and diethyl ether was used as the solvent if the resulting liquid was insufficient for GC sample preparation. In certain cases, dichloromethane was used to wash (8 mL × 3) the catalyst, and the catalyst was then vacuum dried overnight before conducting the thermogravimetric analysis (TGA).
Typical reaction of polyethylene depolymerization PE (1.59 g, Mw~4000, Sigma-Aldrich;~112 mmol of carbon) and the catalyst (0.20 g) were added to a glass vial (20 mL) with a glass-coated magnetic stir bar, and the vial was then placed into an autoclave (75 mL, Parr Instrument). The autoclave was purged three times with H 2 and then pressurized to 45 bar and sealed. The pressurized autoclave was placed in a heating block at the desired temperature (275-375°C) and stirred (500 rpm) for the given reaction time (2-16 h). After the reaction, the autoclave was cooled to room temperature in a water bath. The gaseous products were transferred into a gas-sampling bag (1 L) and injected into a specialized gas-sampling gas chromatography-flame ionization detector (GC-FID) for analysis. For GC-mass spectrometry (GC-MS) analysis of the liquid products, ndodecane (0.085 g, 0.5 mmol) was added to the vial as an internal standard and diethyl ether was used as the solvent if the resulting liquid was insufficient (<0.5 mL) for GC sample preparation.

Instruments
Analysis of the liquid products was performed using an Agilent 7890B Gas Chromatograph coupled with Agilent 7000C MS triple quad detector with He as a carrier gas, Agilent HP-5ms ultra inert capillary column, with a ramping rate of 7.5°C/min from 40 to 197.5°C (for samples of nC 16 conversion) or from 40 to 295°C (for samples of PE conversions). Analysis of gaseous products was performed using an Agilent 7890B Gas Chromatograph with N 2 as a carrier gas, Agilent PoraPLOT Q capillary column, with a ramping rate of 10°C/min from 35 to 175°C. The corresponding retention times of C 1-5 hydrocarbons was identified by a reference gas mixture obtained from Carbagas comprising H 2 , CO, CO 2 , CH 4 , C 2 H 4 , C 2 H 6 , C 3 H 6 , C 3 H 8 , and nC 4 H 10 in N 2 . FID area signals were calibrated using reference methane, ethane, propane, and n-butane gas bottle purchased from Sigma-Aldrich. 1 H NMR spectra were recorded on a Bruker 400 MHz instrument. Powder X-ray powder diffraction (Powder-XRD) patterns were acquired by Bruker D8 Discover. X-ray photoelectron spectroscopy (XPS) measurements were obtained on a Kratos Axis Supra. Scanning electron microscopy (SEM) images were recorded on a Carl Zeiss Gemini 300 microscope. Transmission electron microscopy (TEM) images and scanning transmission electron microscopy (STEM) mappings were obtained on the FEI Talos microscope and analyzed under 200 keV acceleration energy. Thermogravimetric analysis (TGA) was conducted by Mettler Toledo TGA/DSC 3+ with a ramping rate of 5°C/ min from 35 to 900°C and a flow rate of 20 mL/min under air.

Data analysis and quantification
Conversion of nC 16 and quantification of the C 5-16 products was estimated from the GC-FID signal ratio with nC 12 as the internal standard with documented approaches of effective carbon number (ECN) calculations for correlations among the resulted hydrocarbon products 87 . Quantification of the C 1-4 fraction was determined from the GC-FID signal ratio compared to signals with 100% reference gas injection.
Conversion of PE was calculated using the following equation and quantification of the products was determined from the GC-FID signal ratio using nC 12 as an internal standard: Activity-mechanism map The activity-mechanism map is constructed on a two-dimensional polar coordinate system composed of activity as the radius (r) and mechanism as the theta (θ). The value for "activity" is denoted by the nC 16 conversion at the selected benchmark conditions (45 bar H 2 , 375°C, 2 h as the current case). The benchmark conditions would be further adapted (preferably lower applied temperature) in the case of a new catalyst that outperforms all the tested catalysts by relocating the maximum activity (r = 100%). The value for "mechanism" is denoted by the similarity compared to a perfect mechanism with product selectivity as the major feature: 0º (perfect monofunctional hydrocracking): C 3-4 products only. 45º (perfect bifunctional hydrocracking): >C 3-4 products only. 90º (perfect hydrolysis): C 1 product only.

Data availability
The data that support the figures and findings of this work are provided in the Supplementary Information and in the Source Data files. Source data are provided with this paper.