LM-PdCu electrocatalyst with stacked morphology and lamellar mesoporous structure was synthesized by a simple solution-phase route with NH4PdCl4 and Cu(NO3)2 as metal precursors, cetrimonium bromide (CTAB) as mesopore-forming template, 1,3,5-trimethylbenzene (TMB) as structure-controlled agent, and NaBH4 as reducing agent. Typically, predominant NH4PdCl4 and Cu(NO3)2 were first mixed in the solution containing pre-dissolved CTAB and TMB. Then, freshly prepared NaBH4 was injected into above solution and further reacted at 90 oC for 1.5 h. Finally, mesopore-forming template and structure-controlled agent were removed by H2O/ethanol for several times to expose clean metal sites in mesopores (see the details in Supporting Information). Similarly, monometallic LM-Pd and other bimetallic LM-PdM alloys were prepared accordingly with the same procedures but using different metal precursors.
LM-PdCu electrocatalyst and corresponding counterparts were thoroughly revealed by various advanced characterization techniques. Low-magnification transmission electron microscopy (TEM) image showed that LM-PdCu was homogeneously dispersed with nearly one-dimensional (1D) stacked morphology (Fig. 1a). The average length and width of LM-PdCu were estimated as 210 nm and 120 nm, respectively. TEM image further disclosed high degree of uniformity of products with 2D lamellar mesoporous structure (Fig. 1b). Different to traditional spherical mesopores, LM-PdCu were constructed by interlayer stackings, forming stacked morphology and lamellar mesoporous structure. Enlarged TEM image of single LM-PdCu further revealed open nanochannels of lamellar mesoporous structure (Fig. 1c), which were also confirmed by corresponding equalized rainbow color mode image (Fig. 1d). Above characterizations clearly demonstrated that LM-PdCu was morphologically stacked and structurally lamellar (Fig. 1e).
Atomic structure of LM-PdCu was further characterized. Spatial distribution of Pd and Cu in the product was first revealed by high-angle annular dark-field scanning TEM (HAADF-STEM) and corresponding energy dispersive X-ray spectroscopy (EDS) mapping images (Fig. 1f). Two metal elements, Pd and Cu, were uniformly distributed in whole sample, having atomic ratio of 76:24 (Figure S1), proving they were compositionally alloyed rather than phase-separated. High-magnification TEM image recorded from side view of LM-PdCu further implied lamellar mesostructures with alternative interlayered metal frameworks and 2D mesopores (Fig. 1g). The average framework thickness and mesopore size were calculated as 3.8 nm and 3.1 nm, respectively. High-resolution TEM image further showed a clear lattice fringe with a d-spacing distance of 0.254 nm, which corresponded to the (111) plane of a face-centered cubic (fcc) PdCu alloys (Fig. 1h). In addition to the result from powder X-ray diffraction (PXRD) pattern, our result definitely confirmed successful synthesis of well-alloyed LM-PdCu. More importantly, this strategy can be generally extended to various lamellar mesoporous structures with different metal compositions. Not only monometallic LM-Pd but also bimetallic LM-PdM (M = Ag, Ni, Co, and Fe) were successfully prepared with same stacked morphology and lamellar mesoporous structure as well as well-alloyed composition (Figs. 1i,j and S2-S4). Considering potential applications induced by different metal compositions, this study would provide an opportunity for materials discovery and application exploration.
On the basis of synthetic parameters presented above, we believed the utilization of CTAB as mesopore-forming template and TMB as structure-controlled agent as well as NaBH4 as reducing agent played concurrent roles in precise synthesis of novel LM-PdCu electrocatalyst with stacked morphology and lamellar mesoporous structure. To explore their important effects, all synthetic parameters were investigated systemically. First, we found, in the absence of CTAB, the product was metal nanoparticles without any mesopores. After the addition of CTAB, the structure first evolved into spherical nanoparticles with ring-like lamellar mesopores. With further increase of CTAB concentrations, perfect LM-PdCu with 2D lamellar mesoporous structure and 1D stacked morphology was formed accordingly (Figure S5). Then, TMB as concurrent structure-controlled agent was investigated (Figure S6). In the absence of TMB, non-porous products were obtained. In comparison, 1D stacked morphology was formed immediately after the addition of 0.36 mmol TMB. However, small amount of TMB disabled the formation of perfect lamellar mesoporous structure. With further addition of TMB, 2D lamellar mesoporous structure was produced accordingly and became wider with a width of > 200 nm in the higher TMB concentrations. The results clearly demonstrated the concurrent roles of CTAB as mesopore-forming surfactant and TMB as structure-controlling agent for the formation of LM-PdCu. Besides, we explored the effect of reducing agents for adjusting its reduction kinetics36. Other reducing agents with weaker reduction ability, for example N2H4·H2O and ascorbic acid (AA), formed a stack of nanoparticles, instead of 2D lamellar mesopores (Figure S7), indicating the importance of NaBH4 as reducing agent that can adjust the reduction kinetics for the formation of LM-PdCu.
To gain more insights on formation mechanism of LM-PdCu, we also performed time-dependent structure evolution characterizations by detailed TEM studies. Upon the addition of NaBH4 to reaction solution (30 s), some crystalline nanoparticles with an average size of 5.2 nm were immediately formed (Fig. 2a). As the nucleation proceeded for 5 min, small nanoparticles evolved into lamellar stacked structures with discrete metal aggregates (Fig. 2b). With a longer reaction time of 20 min, discrete metals further attached into bigger crystals (Fig. 2c). Finally, perfect stacked metal frameworks were obtained after being reacted for 1.5 h (Fig. 2d). On the basis of above results, a nucleation and attachment mechanism along lamellar assembled template was proposed for the formation of LM-PdCu (Fig. 2d). In an aqueous solution, CTAB and TMB co-assembled and further stabilized with NH4PdCl4 and Cu(NO3)2 by electrostatic and coordinated interactions (with quaternary head of CTAB) into ordered lamellar structures, in which CTAC and TMB were used as mesopore-forming surfactant and structure-controlling agent, respectively37. After the injection of strong reducing agent (NaBH4), PdCu alloys were immediately nucleated into small nanoparticles and further attached into discrete aggregates along lamellar assembles. With further attachment of discrete nanoparticles, continuous 2D metal frameworks with stacked morphology and lamellar mesoporous structure were produced accordingly. Surface-clean metal frameworks of LM-PdCu were finally obtained by further removing CTAB and TMB.
Unique physicochemical properties of lamellar mesoporosity and bimetallic alloys rendered LM-PdCu promising electrocatalyst for PET upcycling38, 39. Meanwhile, monometallic LM-Pd and bimetallic nanoparticle PdCu (NP-PdCu) were also investigated as controls for sharp comparisons. Before electrocatalysis, waste PET plastic was hydrolyzed into PET hydrolysate (PETH) in KOH, which consisted of terephthalic acid (TPA) and ethylene glycol (EG)40–42. After that, electrochemical upcycling of PETH was probed in a three-electrode system to explore the EG oxidation reaction (EGOR) at anode. In the process, EG was first electrooxidized to glycolaldehyde (GAD) by losing two electrons (2e−) and then electrooxidized to high value-added glycolic acid (GA) by further losing 2e−. After the separation and purification of GA, isolated KOH can be recycled for further PET hydrolysis. Meanwhile, at cathode, H2O was electroreduced into H2 accordingly (Fig. 3a).
Linear sweep voltammetry (LSV) curves were first collected and further compared in N2-saturated 1.0 M KOH with or without PETH containing 0.10 M EG (Fig. 3b). On the LM-PdCu catalyst, an anodic potential of 1.70 V (vs. reversible hydrogen electrode (RHE) hereafter) was required at a current density of 10 mA cm− 2 in KOH, which drove the oxygen evolution reaction (OER). In comparison, the potential shifted negatively to 0.76 V in PETH under the same conditions, indicating that the EGOR in PETH occurred at a remarkably lower potential. Differently, the higher potentials to drive EGOR electrocatalysis were required for LM-Pd (0.86 V) and NP-PdCu (0.84 V), indicating their lower activities for EGOR electrocatalysis in PETH. FE and yield rates of GA were then collected in the optimized potential window from 0.8 to 1.1 V. As summarized in Fig. 3c, LM-PdCu disclosed the highest GA FE of 96.8% and yield rate of 0.129 mmol cm–2 h–1 at 0.95 V. In comparison, the highest FE and yield rate were 92.5% and 0.113 mmol cm–2 h–1 for LM-Pd at 0.90 V and 91.2% and 0.079 mmol cm–2 h–1 for NP-PdCu at 0.90 V, respectively. Meanwhile, LM-PdCu electrocatalyst hold the highest mass activity and electrochemical active surface area (Figures S8-S10). Moreover, the fastest electrocatalytic kinetics, including the lowest apparent activation energy (Ea) and Tafel slope values, and the best electron transfer resistance, were also achieved by LM-PdCu (Figures S11-S13).
In general, noble metal electrocatalysts can be quickly deactivated by poisoning with C–O/C–H intermediates during EGOR electrocatalysis43. Although LM-PdCu electrocatalyst hold the strongest OH adsorption capacity (Figure S14)44, 45, its exposed active sites were still occupied by C–O/C–H intermediates, as being confirmed by the disappearance of adsorption and desorption peaks of OH– in EG solution (Figure S15). This would result in a huge waste of noble metal sites and cause the deactivation of electrocatalysts. To further prove our hypothesis, in situ Raman spectroscopy of LM-PdCu was further performed during EGOR electrocatalysis in PETH under different potentials. As shown in Fig. 3d, with the increase of potentials, the adsorption signals of EG became weaker for 200 s of electrocatalysis46. Remarkably, more active sites were occupied by oxidized intermediates and/or products at high potentials, definitely indicating LM-PdCu can be deactivated during EGOR electrocatalysis. Similarly, open-circuit potential (OCP) tests with different test times showed that adsorption capacity of EG in PETH decreased rapidly at CE mode (Figure S16)47. We thus concluded, on fresh catalyst surface, EG can be progressively electrooxidized to GA, but the desorption of GA and/or other intermediates (for example GAD) was remarkably slow due to serious poisoning of metal sites48, 49. With continued electrocatalysis, the reaction intermediates and products occupied most active sites of noble metals, making difficult for EG to be further adsorbed, which thus deactivated the electrocatalyst for selective EGOR electrocatalysis in PETH at CE mode (Fig. 3e).
Pulsed electrocatalysis (PE) has recently been recognized as an effective route to redistribute surface microenvironment of noble metal electrocatalysts and accelerate removal of poisoning intermediates, thus enhancing electrocatalytic activity and selectivity in various reactions50–52. Therefore, we performed and compared EGOR electrocatalysis from PETH in the potential square wave outputs at PE and CE modes (Fig. 4a). In CE mode, a high oxidation potential was continuously output. In PE mode, differently, the alternating periods of high oxidation potential and low OCP were output repeatably. Remarkably, when being pulsed from high potential to low OCP, a large reversed charging current was generated in a very short time for charging the surface double–layer of electrocatalyst53, 54. Meanwhile, four output forms of square waves were designed with different high potential output times, all of which exhibited high reversed performance (Figure S17). We further compared chronoamperometric performance at CE and PE modes. As presented in Fig. 4b, current density rapidly decreased as the reaction proceeded at CE mode. In comparison, PE mode retained perfectly current density (Figure S18). Remarkably, at PE mode, oxidated intermediates and products (GA) were promptly desorbed from metal surface of electrocatalyst, resulting in pretty clean sites for further GA electrosynthesis from PETH. We also compared electrocatalytic EGOR performance in PETH under various pulsed pattens with different output times. The best FE and yield rate of GA from PETH were achieved when being pulsed at high potential for 1.0 s and low potential for 0.5 s (Figure S19).
Electrocatalytic performance of LM-PdCu was then explored and further compared at CE and PE modes for GA electrosynthesis from PETH (Fig. 4c). After being performed at CE mode, the stability gradually deteriorated with the increase of input potentials, having only 1.8% of current retention rate at 1.1 V for 3600 s. At PE mode, however, the stability became much better even at high potentials, retaining 42.8% of current retention rate at 1.1 V, after being electrochemically pulsed for 5400 s. Typically, PE mode accelerated the desorption of toxic intermediates and/or products on catalyst surface and thus enhanced catalyst resistance to toxification during EGOR electrocatalysis in PETH. Electrocatalytic EGOR performance in PETH was then evaluated at PE mode. As poisoning species can be removed quickly from catalyst surface at PE mode, GA yield rate did not reach the peak at 0.95 V (at CE mode). As presented in Fig. 4d, GA yield rate gradually increased with the increase of input potentials, reaching 0.475 mmol cm–2 h–1 at 1.1 V, which was 10.4 times higher than that at CE mode. More importantly, FE values of GA were remarkably high as > 92% in all potentials tested (Fig. 4e). It was mostly because electrosynthesized GA was timely removed from catalyst surface and thus prevented GA being over-oxidized to other C2 and/or C1 products. We also collected concentrations of reactant (EG) and product (GA) during the electrocatalysis. With an input charge of 900 C, EG concentrations in the PETH gradually decreased, followed by the increase of GA concentrations (Fig. 4f). EG conversion and GA selectivity finally reached as 97.1% and 95.4%, respectively (Figure S20), indicating EG in PETH was almost electrooxidized into value-added GA at PE mode. Moreover, 1H nuclear magnetic resonance (NMR) spectra showed that, with the increase of input charge, EG was gradually consumed and further electrooxidized into GA, while TPA do not engage in the reaction (Figure S21).
Not only superior activity and selectivity but also high stability was achieved by LM-PdCu for GA electrosynthesis from PETH at PE mode. As shown in Fig. 4g, after being continued PE for even 20 cycles, both activity and selectivity were retained perfectly with high GA FE of 92.7% and yield rate of 0.464 mmol cm–2 h–1. Meanwhile, TEM image and HAADF-STEM EDS mapping images revealed that LM-PdCu maintained well fresh structure and composition after being performed stability tests (Figures S22-S23). Meanwhile, when compared with activity and selectivity in EGOR electrocatalysis, LM-PdCu also represented one of the best electrocatalysts for GA electrosynthesis (Fig. 4h and Table S1)15, 30, 33, 55. More impressively, LM-PdCu was most stable for EGOR electrocatalysis even in remarkably high potential of 1.1 V, when comparing with previously reported electrocatalysts (Fig. 4i and Table S2)23, 28, 32, 56–60. These results definitely highlighted high potential of LM-PdCu as a high-performance electrocatalyst for selective GA electrosynthesis from PET upcycling.
In order to further explore how reacted species were adsorbed and desorbed on catalyst surface during electrocatalysis, we performed in situ Raman spectra at CE and PE modes. As shown in Fig. 5a, there were three main Raman signals at 875 cm–1, 1083 cm–1, and 1464 cm–1, which can be exactly assigned to νs (C-C-O), γ (C-O-H), and δ (CH2) of GA on catalyst33, 61. After being electrocatalyzed at CE mode for 5 min, all Raman signals increased dramatically, indicating that a large amount of GA and counterpart intermediates were adsorbed on catalyst surface. After switching to PE mode directly, by contrast, the Raman signals gradually weakened with the reaction time being prolonged to 3 min. Obviously, adsorbed GA and other intermediates were remarkably desorbed from catalyst surface at PE mode. Similarly, 2D in situ Raman spectra showed that oxidated species were rapidly adsorbed on catalyst surface at the high potential of 1.05 V, resulting in a serious poisoning of metal active sites during electrocatalysis at CE mode (Fig. 5b). However, after being switched to low OCP of 0.50 V, δ (CH2) signals of GA gradually became weaker. The result further confirmed that switching to PE mode remarkably desorbed poisoning GA and/or intermediates and thus realized the site clean of metal catalyst surface for further electrocatalysis.
Dynamic adsorption and desorption of reacted species and products during GA electrosynthesis at PE mode were further required by in situ Raman spectroscopy. As presented in Fig. 5c, adsorption signals of GA and intermediates were immediately generated on catalyst surface at high potentials based on electrocatalytic oxidation of EG, all of which can be then desorbed by cleaning step after continuously switching high potential to OCP. More interestingly, 2D in situ Raman spectra visualized dynamic processes of selective electrooxidation of EG and rapid desorption of GA on catalyst surface at three continuous PE cycles (Fig. 5d). We also carried out OCP tests after 30 min of electrocatalysis at CE and PE modes. In comparison to CE mode, PE mode disclosed much better EG adsorption capacity, further highlighting that removal of poisoning species exposed active metal sites for further electrocatalysis (Fig. 5e). On the basis of above observations, we proposed a reliable electrocatalytic mechanism for GA electrosynthesis from PETH at PE mode (Fig. 5f). There were two dynamic yet continuous steps at PE mode, including selective electrosynthesis of GA at high potentials and oxidation removal of GA at low potentials. Clean metal sites of catalyst surfaces facilitated continuous electrooxidation of EG and fast desorption of GA, which accelerated electrocatalytic kinetics and thus achieved high performance for robust GA electrosynthesis from PET upcycling at PE mode.
Considering high performance in GA electrosynthesis from PET upcycling, we finally demonstrated its practical scaled-up application at PE mode from real bottle waste plastics. As presented in Fig. 6a, bottle waste plastics were first hydrolyzed in a KOH solution to obtain PETH. Then, 5.1 g of TPA was separated from PETH by direct filtration. After being diluted, EG was electrocatalytically oxidized into glycolate at PE mode. After being further separated and acidified, 1.53 g of high-purity yet value-added GA was produced accordingly. Impressively, a high yield of 64% for GA electrosynthesis from upcycling of bottle waste plastics was achieved. LSV curves collected from bottle waste plastics and PET in KOH and PETH were almost overlapped, indicating same electrocatalytic performance, which thus confirmed high performance of LM-PdCu for practical GA electrosynthesis from PET bottle plastics at PE mode (Fig. 6b). Moreover, 1H NMR spectra and PXRD patterns of products (GA) verified high purity of GA produced from waste bottle plastics at PE mode (Fig. 6c,6d). The results definitely highlighted high efficiency of LM-PdCu electrocatalyst for robust GA electrosynthesis from real waste plastics at PE mode.