The integrated system allowing PET depolymerization and conversion coupled three reaction steps: PET methanolysis, methanol dehydrogenation, and DMT hydrodeoxygenation. Experiments demonstrated that the alcoholysis of PET could be directly carried out in methanol at 210°C in the absence of the catalyst to obtain a 100% yield of DMT monomers within 30 min (Figure S1). Therefore, we mainly focused on catalysts for methanol dehydrogenation and DMT hydrodeoxygenation. Table 1 lists the performances of different catalysts in these processes. Almost all the catalysts tested were inactive, except Cu/SiO2 (HT) with a 73% yield of PX, while the yield of the by-product methyl 4-methylbenzoate and 4-methylbenzyl alcohol were 23% and 4%, respectively. Interestingly, Co/SiO2, Ni/SiO2, and Fe/SiO2 showed no activity for the whole reaction. Once PET was depolymerized into DMT monomers, the hydrodeoxygenation reaction stopped since Co, Ni, and Fe active centers hardly catalyzed methanol dehydrogenation, resulting in a lack of hydrogen for DMT hydrodeoxygenation.
Table 1
Catalytic performance of different catalysts used in the PET conversion process.
Catalyst | PET Conv. (%) | EG/DMT | DMT | PX yield (%) | By-product yield (%) | Incremental pressure at RT (MPa) | Gas composition (MPa) |
yield (%) | Conv. (%) | Methyl 4-methylbenzoate | 4-methylbenzyl alcohol | H2 | CO | CO2 | CH4 |
Cu/SiO2(HT) | 100 | 100 | 100 | 73 | 23 | 4.0 | 2.9 | 73 | 24 | 3 | |
Co/SiO2 | 100 | 100 | - | - | - | - | - | - | - | - | - |
Ni/SiO2 | 100 | 100 | - | - | - | - | - | - | - | - | - |
Fe/SiO2 | 100 | 100 | - | - | - | - | - | - | - | - | - |
Cu/TiO2 | 100 | 100 | 100 | 17 | 57 | 27 | 1.2 | 69 | 28 | 2 | 1 |
Cu/ZrO2 | 100 | 100 | - | - | - | - | - | - | - | - | - |
Cu/CeO2 | 100 | 100 | - | - | - | - | - | - | - | - | - |
Cu/SiO2(IM) | 100 | 100 | - | - | - | - | - | - | - | - | - |
Cu/SiO2(DPU) | 100 | 100 | - | - | - | - | - | - | - | - | - |
Cu/SiO2(DPA) | 100 | 100 | 2.2 | - | 2.2 | - | 0.8 | 53 | 35 | 12 | - |
CuNa/SiO2 | 100 | 100 | 100 | 100 | - | - | 3.4 | 75 | 21 | 4 | - |
CuLi/SiO2 | 100 | 100 | 100 | 89 | 8.4 | 2.2 | 3.0 | 79 | 18 | 3 | - |
CuK/SiO2 | 100 | 100 | 100 | 97 | 1.9 | 0.7 | 3.3 | 73 | 23 | 4 | - |
CuRb/SiO2 | 100 | 100 | 100 | 60 | 18 | 22 | 3.0 | 74 | 23 | 3 | - |
CuCs/SiO2 | 100 | 100 | 100 | 67 | 20 | 13 | 3.0 | 73 | 22 | 5 | - |
Reaction conditions: 0.12 g PET, 0.1 g catalyst, 30 mL methanol, 210°C, 6 h. |
HT: Hydrothermal method, IM: Impregnation method, DPA: Deposition–precipitation with ammonia method, DPU: Deposition–precipitation with urea method, RT: Room temperature, EG: Ethylene glycol, DMT: Dimethyl terephthalate, and PX: p-xylene. |
Table of content |
We subsequently used Cu as the metal active center to investigate the influence of the support on this reaction. Unlike Cu/SiO2, Cu/TiO2 (PX yield 17%), Cu/CeO2 and Cu/ZrO2 (PX yield 0%) did not show good performances. We speculated that SiO2 had a more amorphous structure than TiO2, CeO2, and ZrO2 and was easily etched by ammonia to form a strong copper silicate structure. This copper silicate precursor structure was beneficial to partially reduce Cu+ and Cu0 for methanol dehydrogenation. The synthesis methods of Cu/SiO2 were investigated by comparing the effects of various Cu-based catalysts on the conversion of PET. Several Cu-based catalysts were prepared using various methods such as the hydrothermal method (HT), impregnation method (IM), deposition–precipitation with urea (DPU), and deposition–precipitation with ammonia (DPA). Only the Cu/SiO2 catalysts prepared by HT and DPA showed reactivity towards methanol dehydrogenation, producing hydrogen at 2.9 and 0.8 MPa, respectively. As indicated above, this hydrogen can be used for subsequent DMT hydrodeoxygenation. However, hydrodeoxygenation of PET on Cu/SiO2 (DPA) stopped at intermediate methyl 4-methylbenzoate, most likely because of the lack of H2 released during the methanol dehydrogenation to completely convert PET to PX. Unlike the HT and DPA methods, the Cu species of the catalyst prepared by the IM and DPU methods generated Cu0 species after being completely reduced (Figure S2). In previous work20–22, it was proven that a mixture of Cu2O and Cu was the active site for methanol dehydrogenation. This could explain the low activity of Cu/SiO2 (IM) and Cu/SiO2 (DPU) to provide the required H2 for DMT hydrodeoxygenation reactions. Finally, we attempted to introduce alkali metals (e.g., Na, Li, K, Rb, and Cs) in the form of chlorides via a hydrothermal treatment to investigate the catalyst activity. A 100% PX yield was obtained on CuNa/SiO2 (HT) (Figure S3). Thus, the addition of NaCl significantly promoted the conversion of PET to PX on Cu/SiO2 (HT). We tried to use CuNa/SiO2 for cyclic reaction tests (Table S1). The used catalyst can still maintain a PX yield of 96% in the second bath, but when the third bath is performed, the PX yield is reduced to 52.7%. The main reason is that the Cu species is converted to Cu0 in a hydrogen atmosphere. The active sites of the hydrogenation reaction are reduced as a result.
To study the influence of NaCl on the hydrothermal treatment, a series of characterization methods were used on the catalyst with and without the addition of NaCl. Cu/SiO2 (dried) and CuNa/SiO2 (dried) samples prepared by the HT method exhibited X-ray diffraction (XRD) peaks characteristics of Cu2Si2O5(OH)2 (2θ = 19.9, 21.8, 30.8, 35.0, 57.5, and 62.4° (PDF #27-0188)) (Figure 2a). The addition of NaCl reduced the crystallinity of copper silicate and the size of Cu nanoparticles, resulting in more uniform size distribution. The peaks for copper silicate disappeared after air-calcination and reduction, and they were replaced by Cu characteristics peaks (2θ = 43.3° (PDF #04-0836)) and Cu2O (2θ = 36.4, 42.3, 61.3, and 77.3° (PDF #05-0667)) (Figure S4). The N2 adsorption–desorption showed that the specific surface area and mesoporous volume of the Cu/SiO2 (dried) precursor were 277.9 m2/g and 0.08 cm3/g, respectively (Figure 2b). After the addition of Na+, the specific surface area decreased by 5/6 (46.9 m2/g), and the mesoporous volume also decreased significantly (0.06 cm3/g). Overall, the entire structure became denser.
The hydrogen temperature-programmed reduction (H2-TPR) profile of the dried Cu/SiO2 precursor sample showed reduction peaks at 256 and 280°C (Figure S5a), ascribed to the reduction of copper silicate to Cu+ and Cu0, respectively. After the addition of NaCl, reduction peaks appeared at higher temperatures of 274 and 299°C, revealing a stronger interaction between Cu particles and the support. The CO adsorption FTIR spectra of the reduced Cu/SiO2 and CuNa/SiO2 showed absorption vibration peaks of CO probe molecules at 2117 and 2111 cm−1, respectively. Moreover, Cu/SiO2 contained Cu0 (band at 2133 cm−1) species, while CuNa/SiO2 showed the presence of Cu+ (band at 2124 cm−1) (Figure S5b). We measured the FTIR spectra of different states of Cu/SiO2 and CuNa/SiO2 (Figures S5c–e) and found that both dried samples showed a characteristic O-H stretching vibration peak at 669 cm−1, which was ascribed to the copper silicate species, in line with the XRD results. The peak at 795 cm−1 was attributed to the bending vibrations of the Si-O bond of the amorphous silica support. The relative content of copper silicate was determined by the intensities of two peaks (i.e., I667/I795). After air-calcination, the intensity of the characteristic peak of copper silicate decreased slightly, while the peak of the carrier increased. This may be explained in terms of a lower crystallinity since copper silicate lost part of the crystal water during the calcination process. Moreover, the characteristic peak at 669 cm−1 nearly disappeared in the reduced sample, revealing that copper silicate may have been reduced to other Cu species by hydrogen. The thermogravimetric analysis (TGA) profile of the precursor (Figure S6) showed that physisorbed water from the precursor was removed at a temperature lower than 130°C. As the temperature increased to 600°C, crystal water was gradually removed, and the copper silicate decomposed into CuO and SiO2. After the addition of NaCl, the water content of the precursor decreased (9.16%), and its structure was more compact. The significant weight loss at 1000–1145°C corresponded to the decomposition of CuO into Cu2O.
Cu X-ray photoelectron spectroscopy (XPS) and Auger Cu LMM analysis were performed to elucidate the chemical states of Cu. Cu2+ satellite peaks at 920–950 eV revealed an incomplete reduction of the precursor (Figures 2d and f). The characteristic peaks for Cu2O and Cu (952.2 and 932.1 eV) were too close to be distinguished by XPS. We intuitively determined the Cu+/Cu0 ratio by Cu LMM X-ray induced Auger electron spectroscopy (XAES, Figure 2c and e). The higher Cu+/Cu0 ratio (1.86) of CuNa/SiO2 confirmed that after the addition of Na+, copper silicate with a low crystallinity and a dense texture was less likely to be reduced to Cu0. A higher ratio of Cu+/Cu0 was indicative of a higher tendency to both methanol dehydrogenation and DMT hydrodeoxygenation.
Transmission electron microscopy (TEM) images intuitively showed the different morphologies of the two copper silicates formed with and without NaCl introduction during the hydrothermal process. Thus, while the dried precursor of Cu/SiO2 showed a layered copper silicate structure (Figure 2g), the dried precursor of CuNa/SiO2 showed a special state of granular particle accumulation (Figure 2i). After reduction under H2, high-resolution transmission electron microscopy (HRTEM) revealed a Cu particle size distribution in CuNa/SiO2 centered at 3.9 ± 0.9 nm (Figure 2j), while Cu/SiO2 showed smaller Cu particle sizes (5.1 ± 1.5 nm) (Figure 2h). TEM-mapping confirmed that Cu and Na were uniformly distributed on SiO2 (Figures 2o and l).
Based on the characterization results, we propose a mechanism to explain the effect of NaCl addition on the formation mechanism of Cu/SiO2 during the hydrothermal process. In the traditional hydrothermal synthesis process, layered copper silicate is normally formed, with Cu/Cu2O particles being located on the layered structure after calcination and reduction. With the addition of NaCl, a large amount of Na+ occupied the hydroxyl group sites on the SiO2 surface, inhibiting nucleation of layered copper silicate and normally grown of this phase into a complete crystal shape. Thus, copper silicate finally exhibited a state of granular particle accumulation. Compared with traditional Cu/SiO2, the granular copper silicate with poor crystallinity formed after the addition of Na+ had a very dense structure, with a specific surface area of 46.9 m2/g and a mesopore volume of 0.06 cm3/g. H2-TPR revealed that this dense structure was relatively difficult to reduce (Figure S5a), thus resulting in a higher number of active sites (i.e., high Cu+/Cu0 ratio of 1.86). SEM (Figure S7) showed that Cu+/Cu0 particles of CuNa/SiO2 were smaller and more uniformly distributed after air calcination and hydrogen reduction.
We investigated the influence of the different Na+/Cu2+ molar ratios generated during the hydrothermal treatment on the performance of CuNa/SiO2. Thus, Na+/Cu2+ molar ratios of 2.5:1, 5:1, 10:1, and 15:1 were denoted as 2.5 NaCl, 5 NaCl, 10 NaCl, and 15 NaCl, respectively. Based on the conversion tests of PET at 210°C, the yield of PX exhibited a volcano-type distribution (Figure S8a), with a maximum PX yield of 100% reached for the 5 NaCl sample, while samples 2.5 NaCl, 10 NaCl, and 15 NaCl showed lower yields of 78.3, 92.3, and 60.7%, respectively. To further explore the influence of the addition of NaCl on the formation of the catalyst, we carried out a series of characterization tests. The precursor samples all showed the characteristic diffraction peaks of the Cu2Si2O5(OH)2 crystal phase (Figure S9), indicating that the addition of NaCl did not affect the phase composition of the catalyst. In the case of the 5 NaCl sample, the copper silicate had poor crystallinity compared to other samples. TGA tests of the CuNa/SiO2 precursor showed that physisorbed water (2.41%) and crystal water (6.75%) upon addition of 5 NaCl was the lowest among all the samples tested (Figures S10 and S8b). This also confirmed that the copper silicate structure was denser at this ratio, facilitating the removal of water during the calcination process. N2 adsorption–desorption (Figure S11) revealed CuNa/SiO2 to have the lowest surface area (46.9 m2/g) upon addition of 5 NaCl (Figure S8c), indicating that the formed structure was the most compact among the samples tested herein. In the Cu LMM XAES spectra of the CuNa/SiO2 after reduction, the ratio of Cu+/Cu0 still presented a volcano-type distribution (Figure S8d), and the Cu+/Cu0 ratio of CuNa/SiO2 was the highest when 5 NaCl was introduced (1.86, Figure S12).
Based on the above observations, we speculated that, in the traditional hydrothermal synthesis process, Cu2+ in the solution combined with the silanol groups on the SiO2 surface to form copper silicate, which accelerated the layered copper silicate nucleation and growth significantly. This type of layered copper silicate showed a low interface area with SiO2, leading to a low Cu+/Cu0 ratio on the layered copper silicate surface (Figure 3a). A large amount of Na+ occupied the silanol on the surface of the SiO2 upon addition of 5 NaCl, thereby inhibiting nucleation and growth of layered copper silicate. Cu2+ in the solution could only be combined with the remaining silanol on the SiO2 surface to form scattered and isolated copper silicate particles, and the compact structure had a small surface area and poor crystallinity. The formed granular copper silicate showed a large interface area with SiO2. Granular copper silicate was more difficult to reduce compared with traditional layered copper silicate, resulting in a high ratio of Cu+/Cu0 active sites in the prepared catalyst (Figure 3b). However, when the amount of added NaCl was too high, Na+ occupied all the silanol sites on SiO2, resulting in the precipitation of Cu2+ with SiO32− in solution to form copper silicate, which was then deposited on the SiO2 surface. Compared to the catalyst with 5 NaCl introduced during hydrothermal treatment, this type of copper silicate showed better crystallinity (Figure S7a) and was relatively easier to reduce to Cu/Cu2O·SiO2 with a low ratio of Cu+/Cu0 (Figure 3c). In general, the addition of NaCl in the hydrothermal treatment resulted in the formation of granular copper silicate with a lower crystallinity, smaller specific surface area, and denser texture. When NaCl was introduced, Cu+/Cu0 showed a volcano-type curve distribution. When the molar mass ratio of Na+/Cu2+ reached 5:1, the Cu+/Cu0 ratio reached the maximum (1.86), providing significantly more active sites for methanol dehydrogenation and DMT hydrodeoxygenation.
Finally, to ascertain whether the addition of NaCl only affected the formation process of copper silicate or could promote the reaction itself, we prepared a Cu/SiO2-HT-Na-IM sample. We first synthesized Cu/SiO2 by a hydrothermal method, which was subsequently impregnated with NaCl after the formation of layered copper silicate. The new copper silicate precursor has the same loading of Na+ (2.4%) (Table S2) as CuNa/SiO2. XRD (Figure S13) showed that the impregnated Na+ did not affect the formation of copper silicate, although the yield of PX was moderate (65.8%). Thus, we confirmed that the addition of NaCl in the hydrothermal treatment affected the morphology of the copper silicate and therefore the Cu+/Cu0 ratio after reduction. Na impregnation after the formation of copper silicate not only failed to promote the catalyst activity but also inhibited some of the active sites by covering them, resulting in a reduction in the yield of PX.
We used CuNa/SiO2 to conduct a kinetics study on the reaction of DMT (A) and the intermediates at the optimal reaction temperature (210°C) and monitored the distribution of products over time. As soon as the reaction started, the DMT concentration decreased (Figure 4a) at an initial rate of 0.36 g/g/h, revealing a high efficiency for hydrogen production. Hydrogen was produced during the heating process, and it was sufficient to maintain the amount of hydrogen required for the next reaction. When the reaction started, the intermediate methyl 4-methylbenzoate (C) was produced and a maximum yield of 24.3% at 1.5 h. Within 1–1.5 h, the intermediate 4-methylbenzyl alcohol (D) was produced slowly. At 3 h, almost all the DMT was converted, while the yield of the target product PX reached 100% at 6 h.
In this kinetics study, we only observed two intermediates: C and D. Importantly, we did not observe 1,4-benzenedimethanol, which implied that DMT underwent one-sided adsorption on CuNa/SiO2. Therefore, methyl 4-(methylol)benzoate (B) may appear transitorily as an intermediate product. Based on these results, we speculated that the reaction from DMT to PX involved four steps: (1) one-sided adsorption of the ester of DMT on CuNa/SiO2 and subsequent hydrogenation to alcohol, yielding B; (2) alcohol of B underwent hydrogenolysis to methyl groups and desorbed to form C; (3) ester C adsorbed on CuNa/SiO2 and was hydrogenated to alcohol and obtain D; (4) alcohol D underwent hydrogenolysis to methyl groups and desorbed to obtain the target product PX. The kinetics of the three intermediates were studied under the same conditions (Figures 4b, 4c and 4d). Intermediates A, B, C, and D were completely consumed after approximately 5, 1.5, 5, and 1 h, respectively. The experimental results confirmed our proposed reaction pathway. The simulation results performed using MATLAB yielded the rate constants of each step (k1 = 0.0138 min−1, k2 = 0.0424 min−1, k3 = 0.0103 min−1, k4 = 0.0632 min−1). It is worth noting that C was obtained with the largest concentration after the DMT hydrogenation. This was because steps (2) and (3) required intermediate C desorption and re-adsorption on CuNa/SiO2, making the hydrogenation of C the rate-determining step (k3 = 0.0103 min−1) of the overall process. The rate of alcohol hydrogenolysis was about 4–6 times that of the ester hydrogenation.
In-situ FTIR also demonstrated our reaction process of DMT and its intermediates (B, C, and D) on CuNa/SiO2. The different substrate functional groups appeared in four regions in the FTIR spectra: (1) aryl C=C, 1494, 1523 cm−1, (2) aryl C=O, 1594–1664 cm−1, (3) aryl -CH3 stretching, 2950–2863 cm−1, and (4) O-H, 3305–3290 cm−1. It should be noted that the aryl C=C band of the four substrates at 1494 and 1523 cm−1 did not change during the reaction (Figures 4e, 4f, 4g, and 4h), revealing that the aromatic structure remained intact during the process, and aromatic hydrocarbons tended to be generated. The intensity of the C=O stretching vibration peak at 1594–1664 cm−1 for intermediates A, B, and C decreased continuously with time until complete vanishment (Figures 4e, 4f, and 4g). This phenomenon indicated that the hydrogenation reaction occurred continuously under in-situ conditions. A and C, which lacking hydroxyl groups itself, produced O-H vibration peaks at 3305–3290 cm−1 and then gradually disappeared (Figures 4e and 4g), indicating that C=O hydrogenation to hydroxyl groups occurred, then hydrogenolysis took place, in line with the kinetics results. In addition, the bands corresponding to hydroxyl groups of B and C gradually decreased until they vanished as a result of hydrogenolysis. Finally, under in-situ conditions, the -CH3 vibration peak of PX continuously increased with time. This result provided stronger evidence that PX was generated. The in-situ infrared study once again confirmed the path of DMT conversion on CuNa/SiO2, and the results are highly consistent with the kinetics behavior.
Based on the experimental data described above, we conducted a preliminary on-site test of an island using our method (Figure 5). A recent survey of beach sediment along the coastline of the Phuket Island showed that PET (mainly containing beverage bottles, plastic films, and microwave packaging) accounted for ca. 33.1% of the overall plastic sediment23. Our method achieved conversion of this sediment with 100% PX yield at 210°C on CuNa/SiO2 in 6 h. Every ton of plastic sediment contained 331 kg of PET, and thus, 181 kg of PX and 105 kg of ethylene glycol (EG) could be obtained via this route under optimal conditions. The obtained PX and EG could be used as automobile fuel and antifreeze replenishment. As such, we demonstrated herein that CuNa/SiO2 provides a viable option for processing waste PET accumulated on islands without a need of external hydrogen, transforming it into a high-value-added energy supply. This work could also help to solve the problem of waste resource conversion and reuse in the world today.