Introduction

Almost 80% of the ever-produced plastics (8.3 billion metric tons) are in the environment, mostly in water bodies and landfills due to improper disposal and recycling. Microplastics (plastic fragments <5 mm) are formed by weathering of synthetic plastics via physical, chemical, and biological processes. They are ubiquitous (24.4 trillion pieces) in the environment, especially in the aqueous medium, and are deleterious to the environment, human health, and biological ecosystem1,2. Hydrolysis of concentrated polymers followed by separation of monomers (i.e. terephthalic acid; TPA and EG for PET) is the current approach to recycle PET. However, the tedious separation of monomers after neutralization and crystallization makes the process expensive3. Among various approaches, hydrothermal approaches can transform plastics into hydrocarbons (gasoline/diesel), however, they require high temperature (350–500 °C) and pressures (20–30 MPa)4. Oxidative photocatalytic reforming of plastics into hydrogen/hydrocarbons is an intriguing approach to mitigate the environmental challenge while creating considerable energy benefits5,6. Due to the challenging hydrolysis of addition polymers, the photocatalysis process is mainly employed for the oxidation of condensation polymers such as PET and PLA7,8.

Graphitic carbon nitride (g-C3N4; CN), a promising photocatalyst due to its remarkable chemical and photophysical (2.7 eV band gap) properties, still suffers from poor absorbance of higher wavelength photons (λ > 450 nm) and prodigious carrier trapping in H-bonded (NH/NH2) nanosheets9,10. The modification of the chemical structure/coordination pattern of C6N7 units is a very effective tool to tune the optoelectronic properties of CN. Lotsch et al. reported a cyanamide (NCN) functionalized carbon nitride (NCNCNx) with improved visible absorption and capabilities of charge storage for delayed photocatalysis under dark conditions11,12. The NCN moieties embedded in the CN scaffold also provide an active site for substrate adsorption and catalytic reaction. Several variants of NCNCNx including reductive cleavage of triazine (C3N3) units have been developed for various photocatalytic applications such as H2 generation, biomass photoreforming, alcohol oxidation, and sulfinates transformation13,14,15,16. NCN/CNx can also catalyze photocatalytic H2 evolution from PET and PLA derivatives reaching a maximum conversion rate of 1.65 and 3.59 μmol g–1cat h-1 (after 50 h)17. However, the process requires NiP co-catalysts which are prone to deactivation under photocatalytic/alkaline conditions. Recently, Zhang et al. demonstrated excellent stability in defect-rich metal phosphosulfate co-catalysts-coupled CdS (d-NiPS3/CdS) catalysts, reaching an H2 evolution rate of 40 mmol g–1cat h–1 18.

In recent years, single atom catalysts (SACs) have garnered significant interest due to maximum atom utilization requiring very low concentrations of metals, while the unique geometric arrangements and undercoordinated sites at the atomic scale ensure high reactivity. Inorganic crytals decorated SACs such as Cu1/TiO2 and Ru single-atom catalyst loaded on Co/Al oxide (RuSA-CoAlO) etc have been reported for plastics/plastic derivatives oxidation to value added chemicals19,20. For instrance, Liu et al. reported various single atom metal decorated on defect rich TiO2 (M1/TiO2) can cleave PET chain to yield bis(2-hydroxyethyl) terephthalate (BHET)21. Carbon nitrides can accommodate high concentrations of single atom (SA) sites via coordination with inner-facing secondary nitrogen present in C6N7 constituted cavities. Effective orbital overlap between metal and cavity nitrogen enables facile charge transfer between the CN scaffold and metal center which can regulate the catalytic performance. Since charge distribution on the SA site is highly dependent on the metal-support interaction, the chemical nature of CN is critical to improve photoactivity. For instance, Chen et al. showed that the oxidation state and reactivity of the Pd SA site vary significantly as the cavity size changes in polyheptazine imides (PHI), polytriazine imide (PTI), and linear melem oligomers (LMO)22. The photoactivity of the CN-based SACs can be further improved by heteroatom doping and surface modification to enhance charge separation and interaction between substrate and catalysts23. Li et al. reported Cu SACs embellished on crystalline carbon nitride (Cu-CCN) can afford almost 100% selectivity toward the conversion of CO2 to CO due to better charge separation24. The utilization of H-bonded macromolecular conjugates of nitrogen-rich precursors (cyanuric acid, melamine, barbituric acid, etc.) has been widely adopted for the synthesis of oxygen-doped CN with improved crystallinity, visible absorption and surface properties25,26,27. The NH2 functionalities of melamine can form H-bonding with keto tautomeric form of cyanurates or barbiturates resulting in an extended 2D H-bonded structure. The preorganized H-bonded structure can condense efficiently and therefore minimize carrier recombination due to reduced inter-sheet H-bonding28. We hypothesized that like mononuclear melamine and cyanuric acid aggregates, three rings containing melem and cyameluric acid units should form a densely packed H-bonded aggregate. The condensation of a close-packed three-membered ring structure will result in relatively low NH3 evolution which concomitantly reduces defects state generation. Furthermore, extended conjugation can further improve the photophysical properties of melem and cyameluric derived carbon nitride. On the other hand, the decoration of isolated IrO2 species should induce the reaction kinetics due to the efficient capture of generated holes29.

Herein, we designed an NCN-functionalized O-bridged carbon nitride (MC) by thermal annealing of H-bonded melem (M)-cyameluric acid (C) macromolecular aggregates. The MC was used as a scaffold for the fabrication of Ir SACs (MCIrSA) which demonstrate improved visible absorption and oxidation kinetics. Advanced characterization using X-ray photoelectron spectroscopy (XPS), X-ray absorption near edge structure (XANES), extended X-ray absorption fine structure (EXAFS), and solid-state electron paramagnetic resonance (ssEPR) reveals reduced electron density on the Ir site due to charge transfer to C6N7 moieties. Under AM1.5 G irradiation, MCIrSA displays a high rate of hydrogen evolution (147.5 and 29.58 μmol g−1cat h−1) using PLA and PET derivatives in alkaline solution, respectively. Experimental and theoretical investigations using EPR and DFT suggest the facile generation of •OH radicals on undercoordinated IrN2O2 sites via O-Ir-OH intermediate under visible irradiation.

Results

Synthesis of catalysts

The NCN-functionalized O-bridged carbon nitride (MC) was synthesized by thermal annealing of melem(M)-cyameluric(C) macromolecular conjugate (Fig. 1a). Melem (2,5,8-triamino-s-triazine) was synthesized by thermal annealing of melamine at 425 °C30. While cyameluric acid was synthesized by KOH-assisted thermal hydrolysis of melon powder (see Methods section)31. The hydrothermal treatment of an equimolar amount of melem and cyameluric acid in DMSO resulted in an extended hydrogen-bonded macromolecular conjugate. Cyameluric acid remains in dynamic equilibrium between various keto-enol tautomeric forms32. The keto form can form a strong hydrogen bond with melem units to give a long-range order macromolecular conjugate. Such macromolecular assemblies have been previously reported between monomeric melamine and cyanuric acid27,28. The thermal condensation of H-bonded melem-cyameluric conjugate due to the presence of keto form resulted in NCN-functionalized O-bridged carbon nitride (MC). The presence of OH terminal groups on cyameluric acid provides O-bridging while incomplete condensation of NH2 and cleavage of heptazine units in melem give rise to NCN functionalities. The formation of O-bridged carbon nitride has been previously reported using cyanuric acid33 as well as melamine-cyanuric acid macromolecular conjugate derived carbon nitride27,28,34. For comparison, pristine melem and cyameluric were also annealed at 550 °C that form bulk carbon nitride (CNB) and O-linked carbon nitride (CNO) respectively (Fig. 1b, c).

Fig. 1: Synthesis of NCN-functionalized O-bridged carbon nitride (MC).
Fig. 1: Synthesis of NCN-functionalized O-bridged carbon nitride (MC).
Full size image

a NCN-functionalized O-bridged carbon nitride (MC) synthesis using melem and cyameluric acid macromolecular conjugate. The cyameluric acid tautomers makes hydrogen bond with melem molecules to form a macromolecular complex which on thermal annealing produces NCN functionalized O-bridged carbon nitride. b Direct annealing of melem resulted in the formation of bulk carbon nitride (CNB). c Thermal annealing of cyameluric acid led to the formation of O-bridged carbon nitride (CNO).

Catalyst characterization

To probe the chemical structure of CN and MC, 13C solid-state nuclear magnetic resonance (NMR) spectroscopy was performed using the cross-polarization magic-angle spinning (CPMAS) technique. Direct 13C pulse MAS spectra of CN exhibited two strong signals of almost equal intensity at 156 and 164 ppm (Fig. 2a). The NMR signal at 156 ppm originated from three bridging CN3 carbons of heptazine units (C6N7). While the signal at 164 ppm was attributed to three carbons linked to bridging Ns connecting heptazine units. The direct 13C pulse MAS spectra of MC also displayed two main peaks at 156 and 164 ppm for CN3 carbons at bay and corner positions (Fig. 2a). A shoulder NMR peak at 172 ppm originated from the carbon connected to the NH/O bridge while a weak hump around 121 ppm was attributed to NCN carbons matched with reported chemical shifts11,35. It should be noted residual N=C-O groups can also contribute to this peak due to the almost similar chemical shifts.

Fig. 2: Solid state NMR, TEM and FTIR analysis.
Fig. 2: Solid state NMR, TEM and FTIR analysis.
Full size image

a Direct excitation 13C ssNMR spectra of CN and MC. b 13C ssNMR spectra of CN and MC after 30 s contact time. c 1H−13C CPMAS spectra of pristine CN and MC showing the evolution of 13C NMR signals at different contact times. Scale color: yellow to blue: 0.10–10 ms. d HR-TEM image of MC showing a porous structure of MC. e HR-TEM image of MC: inset images are enlarged image and iFFT images demonstrating lattice fringes with 0.36 nm d-spacing. f FTIR spectra of CN (yellow) and MC (blue). MC displayed a sharp band around ~2200 cm−1 due to NCN/NCO groups.

The CP experiment with a contact time of 10.0 ms displayed that the signal intensity of CN was close to the value obtained in the direct 13C pulse spectrum which was in accordance with the previously reported C6N7(NH2)3 (melem) model (Fig. 2b)36. On the other hand, MC displayed a signal at 164 ppm that cross-polarizes faster than the 156 ppm signal due to 1H spin-lattice relaxation in the NH-rich structure. 1H-13C CP MAS spectra of CN and MC with 18 contact times (0.1–10.0 ms) demonstrate that the MC peak at 156 ppm peak is still rising while the CN peak reaches a maximum (Fig. 2c)37. According to these observations, both materials possess melem-type structures with a slightly populated NH concentration in MC.

The scanning electron microscopy (SEM) and transmission electron microscopy (TEM) images of MC displayed porous, entangled graphenic structures compared to the stacked structure of carbon nitride (Fig. 2d and Supplementary Figs. 1–4). Apart from C and N, EDX spectra of MC also displayed signals for potassium and chlorine due to residual KCl form neutralization of potassium cyamelurate precursor (Supplementary Fig. S1). After washing MC, the Cl peak almost disappeared while K intensity was reduced deciphering the partial doping of K in the MC scaffold (Supplementary Fig. S2). The condensation of macromolecular H-bonded melem-cyameluric acid conjugate follows by the breaking of H-bonding and removal of NH3 which brings heptazine units closer to form an aggregated structure that concomitantly condenses to a porous structure27,28. High-resolution TEM, corresponding select area electron diffraction (SAED) pattern and fast Fourier transform (FFT) demonstrate that the overall structure of MC was amorphous (Fig. 2e). However, a few short-range orderings were also observed with a 0.32 nm lattice spacing for the (002) plane of the CN structure (Fig. 2e inset).

The FTIR spectra of CN exhibited signature peaks corresponding to triazine ring bending (807 cm−1), heptazine (C6N7) units stretching (1035–1512 cm−1), and cumulative stretch of residual N-H and intercalated O-H in the 2990–3415 cm−1 region (Fig. 2f). MC also displayed all vibrational peak features specific to the C6N7-based carbon nitride structure along with an additional sharp band at 2174 cm−1 for the C-N stretch of NCN groups38. Close observation of NCN signals displayed splitting (2174 and 2150 cm−1) which was due to the asymmetric stretch of NCN functionalities39. The FTIR spectra of melem monomer and H-bonded conjugates do not reveal any sign of NCN stretching suggesting the NCN functionalities in MC emerged during condensation of H-bonded conjugate (Section 3.1 in SI, Supplementary Fig. 5). The XRD spectra of CN and MC exhibited characteristics (100) and (002) peaks for heptazine constituted framework (Section 3.2 in SI, Supplementary Fig. 6)40. However, a slight decrease in (002) peak intensity reveals decreased crystallinity in MC due to the incorporation of NCN/NCO functionalities41. Distinctly, the CNB and CNO samples do not show any (100) peak and very broad (002) band suggesting a highly amorphous nature (Supplementary Fig. 5)33,42. The synchrotron-based wide-angle X-ray scattering (WAXS) 2D detector images of CN displayed two intense rings with Q values of 0.911 and 1.912 Å−1 which correspond to the (002) and (100) planes (Section 3.3 in SI, Supplementary Fig. 7a, b). The corresponding d-spacing was calculated to be 1.6 and 3.2 Å which was in good accordance with XRD results. Interestingly, the (002) and (100) peaks for MC were shifted slightly to higher Q values (0.933 and 1.954 Å−1) suggesting an increase in interplanar distance in MC due to N-C=N functionalization (Supplementary Fig. 7c, d). The XPS survey scan and high-resolution XPS spectra of CN and MC in the C1s and N1s region displayed very close features suggesting the basic C6N7 scaffold in both CN and MC remain almost similar (Section 3.4 in SI, Supplementary Figs. 8, 9). The O1s spectra of CN and MC displayed two major peak components for residual C=O/N=C-O and intercalated -OH functionalities43. Interestingly, MC also displayed an additional peak component centered at 530.9 eV due to the doping of O in the bridging position (Supplementary Fig. 9f). Due to the condensation of the H-bonded macromolecular structure, MC displayed improved surface properties (SBET-17.46 m2 g−1) compared to CN (SBET-8.84 m2 g−1) (Section 3.5 in SI, Supplementary Fig. 10a). The DR-UV-Vis spectra of MC displayed an extended absorption edge compared to CN and melem due to the increased conjugation degree after NCN functionalization (Section 3.6 in SI, Supplementary Fig. 10b)11. The optical band gap value for CN and MC was calculated to be 2.63 and 2.48 eV (Section 3.7 in SI, Supplementary Fig. 11) The Mott-Schottky plot demonstrates n-type conductivity for both CN and MC with a flat band potential of −0.36 and −0.31 V vs NHE at pH-0 (Section 3.8 in SI, Supplementary Fig. 12a). The increased degree of conjugation due to the introduction of NCN in MC also improves photocurrent density (0.15 mA cm−2) and charge carrier mobility evinced by lower charge transport resistance in Nyquist plots (Supplementary Fig. 12b, c).

To improve the photophysical properties of MC, we have pinned Ir single atom on MC (MCIrSA) using hydrolysis decomposition of IrCl3.xH2O followed by coordination in the C6N7 cavity (see Methods Section). The Ir content of MCIrSA catalysts was determined to be 0.031 wt% (1.624 μmol g−1cat) using ICP-OES analysis (Supplementary Table 6). For comparison, CN with Ir SA sites (CNIrSA) and MC with IrO2 nanoparticles (1 wt%) (MCIrNP) were also prepared. The XRD patterns of CNIrSA and MCIrSA displayed characteristic (100) and (002) peaks and no peak associated with Ir NP or Ir oxides was detected suggesting the ultrafine distribution of Ir in CN and MC scaffold (Section 3.9 in SI, Supplementary Fig. 13)44. The (100) and (002) peak intensities decreased slightly due to the exfoliation of sheets during the sonication step. Further, the WAXS 2D map and azimuthally integrated intensity versus Q plot show only the presence of (100), (002), (006) and (004) peaks of carbon nitride framework without any trace of Ir-related bulk species or nanostructures (Fig. 3a, b).

Fig. 3: WAXS, electron microscope and XAS characterization.
Fig. 3: WAXS, electron microscope and XAS characterization.
Full size image

Synchrotron-based WAXS 2D images and Q values of a CNIrSA and b MCIrSA. No peak for any particulate distribution was observed. c HR-TEM image and respective of MCIrSA showing amorphous structure MCIrSA. d AC-HAADF STEM image of MCIrSA, arrows and circles pointing to bright spots for Ir single atom. e EELS elemental mapping for N (orange), O (blue), and Ir (pink). f Normalized XANES spectra of IrCl3 (black), IrO2 (green), CNIrSA (violet) and MCIrSA (red). g FT-EXAFS spectra of Ir metal, IrO2 NPs, CNIrSA and MCIrSA (solid line: experimental data, dots: fit). WT map of h MCIrSA. i Ir foil. j IrO2 NPs.

The FE-SEM images of MCIrNP displayed the clear presence of bright IrO2 NPs while no bright spot or Ir EDX signal was observed for MCIrSA due to the very small amount of Ir SA in the material (Supplementary Figs. 14, 15). The high-resolution TEM images, corresponding FFT and SAED pattern of MCIrSA show amorphous nanosheets with no trace of nanoparticulate distribution (Fig. 3c and Supplementary Fig. 16). STEM elemental mapping confirmed well-distributed Ir elements overlapped on C, N and O elements corroborating the presence of Ir in highly dispersed form (Supplementary Fig. 16). AC-HAADF-STEM of MCIrSA shows a sharp bright spot of Ir due to Z-contrast validating the presence of Ir in an isolated atomic state (Fig. 3d and Supplementary Fig. 17a–d). The EELS mapping and intense N K-edge and O K-edge signals show well-constituted O bridged heptazine framework decorated with Ir SA (Fig. 3e and Fig. S17e-i). The XPS spectra of CNIrSA and MCIrSA remain almost identical to pristine CN and MC and the signal for Ir was below the detection limit (Section 3.10 in SI, Supplementary Figs. 18, 19). The near edge X-ray absorption fine structure (NEXAFS) spectra of CN, MC, CNIrSA and MCIrSA show characteristics π*C=C, π*N-C=N, σ*N-C=N and σ*C-N resonance peak for tri-s-triazine units of carbon nitride (Section 3.11 in SI, Supplementary Fig. 20)45. The N K-edge NEXAFS also show signature π* (π*C–N=C and π*N–C3) and σ* (σ*C–N and σ*C–N=C) for all samples without any detectable variation suggesting Ir decoration does not have any influence of C-N coordination in CN and MC structures (Supplementary Fig. 20)46. No distinct change in CN and MC O K-edge features demonstrate a low concentration of bridging O in MC which remains identical after the decoration of Ir SA (Supplementary Fig. 20)47. The Ir L3 X-ray absorption near-edge structure (XANES) spectra of CNIrSA and MCIrSA displayed a sharp rising edge at 11215.5 and 1121.6 eV respectively due to 2p → 5d (t2g5→eg0) transition (Fig. 3f, Section 3.12 in SI, Supplementary Fig. 21)48,49. The XANES features of Ir SA containing carbon nitrides (CNIrSA and MCIrSA) were closely matched with iridium oxide (IrO2) with slight edge shifting toward high energy corroborating the presence of Ir in 4+δ+ state (Fig. 3f)50. Furthermore, the absence of any pre-edge feature demonstrates the centrosymmetric coordination environment of Ir species. These observations suggest that Ir SA in both CNIrSA and MCIrSA was present in a 4+δ+ oxidation state and coordinated to two oxygens. Interestingly, non-normalized raw spectra of CNIrSA and MCIrSA exhibited increased white line intensity compared to IrO2 NPs suggesting decreased electronic density and increased oxidative character (Ir(4+δ)+) (Supplementary Fig. 21)51,52. Since Ir SA was connected to N atoms in the heptazine-constituted cavity, partial charge transfer to the CN scaffold led to decreased electron density and more states became available in the 5d orbitals53. MCIrSA demonstrates the highest white line intensity suggesting effective coordination of Ir in NCN functionalized O-bridged CN54. The exact valence state of CNIrSA and MCIrSA determined using Ir valence state and absorption edge energy plots was found to be 4.4 and 4.7, respectively (Supplementary Fig. 22). The decreased charge density on the Ir SA site was also confirmed from the Bader charge analysis in DFT calculation (See DFT discussion section 2.6).

Fourier-transformed extended X-ray absorption fine structure (FT-EXAFS) spectra of Ir-incorporated materials were acquired to discern the atomic coordination environment of Ir centers (Fig. 3g and Supplementary Fig. 23). The FT-EXAFS spectra of MCIrSA displayed a sharp first coordination shell at 1.62 Å originating from Ir-N/O scattering (Fig. 3g)55. The absence of any second shell scattering discerns the presence of isolated Ir-N/O species. CNIrSA demonstrated a sharp scattering shell at 1.54 Å for Ir-N/C species while a small peak at ~2.62 Å might have originated from scattering from neighboring nitrogen (Ir-N-C) and partially agglomerated Ir-Ir species55,56,57. In contrast, the Ir-Cl scattering appeared at a higher R(Å) value (~1.84 Å) which excludes any possibility of unreacted IrCl3 in the materials (Supplementary Fig. 23a). On the other hand, Ir foil displayed a strong Ir-Ir coordination peak at 2.56 Å which does not match with CNIrSA and MCIrSA and therefore excludes the possibility of any Ir-Ir interaction in the materials (Fig. 3g). Additionally, the Ir-O-Ir peak for the IrO2 nanoparticles (IrO2 NPs) was absent in both CNIrSA and MCIrSA substantiating the presence of IrO2 as an isolated species in the catalysts (Supplementary Fig. 23). EXAFS data fitting for CNIrSA and MCIrSA demonstrate an Ir-O bond length of 1.77 and 1.80 Å while Ir-N bond length was found to be 2.11 and 2.02 Å respectively which closely matched with the bond length calculated from DFT models of MCIrSA (Supplementary Fig. 24 and Supplementary Tables 1, 2). The Ir-O and Ir-N coordination number (C.N.) of MCIrSA was calculated to be 0.9 and 2.9 (Supplementary Table 1). The overall coordination of IrO2 SA species (3.8) in MCIrSA was significantly lower than IrO2 NPs (C.N. 6) validating the existence of undercoordinated Ir species. Such undercoordinated Ir site formation has also been reported previously on carbon nitride scaffolds55.

Wavelet transform (WT) EXAFS spectra of CNIrSA, MCIrSA, IrO2 NPs and Ir foil were acquired to further evaluate the coordination structure of Ir species (Fig. 3h–j and Supplementary Fig. 23b). The WT-EXFAS spectra of MCIrSA exhibited a sharp zone in K = 5.38 Å−1 and R = 1.62 Å corresponding to first shell scattering from Ir-O/Ir-N bonds (Fig. 3h)58. Distinctly, Ir foil displayed a sharp zone in K = 12.47 Å−1 and R = 2.59 Å attributed to Ir-Ir scattering (Fig. 3i). The observance of a sharp zone in lower K space and the absence of any Ir-Ir scattering zone at a higher K value for MCIrSA confirms the presence of Ir species as single atoms coordinated to lower atomic number O/N atoms. The WT-EXAFS map of IrO2 NPs showed a sharp zone in K = 8.01 Å−1 and R = 1.62 Å corresponded to Ir-O scattering while a relatively weak zone at K = 13.50 Å−1 and R = 3.45 Å was originated from Ir-O-Ir second shell scattering (Fig. 3j). The R space value of MCIrSA (1.62 Å) closely matched with the first shell scattering of IrO2 NPs while the absence of second shell scattering at a higher R-value corroborates the presence of isolated Ir-O species59. CNIrSA also reveals a sharp zone in K = 5.21 Å−1 and R = 1.62 Å confirming the formation of identical IrO2 SA species present in MCIrSA (Supplementary Fig. 23b).

Charge carrier dynamics

The UV-Vis absorption profile of MCIrSA displayed enhanced light absorption compared to pristine MC due to electronic charge transfer from Ir SA sites to the C6N7 framework populating the electron density of the π-conjugated network (Section 3.13 in SI, Supplementary Fig. 25)60. The XANES results (c.a. Ir4.7+) and Bader charge analysis also demonstrate electronic charge transfer from Ir SA sites to the CN framework (See DFT discussion, Section 3.12 in SI, Supplementary Fig. 23). As the oxidation state of Ir was close to ~5 (4.7), it is most likely that upon light irradiation, Ir site will acquire one hole to achieve a stable 6+ state. To validate charge transfer between Ir sites and MC framework, in situ XPS under dark and light irradiation conditions was performed (Supplementary Figs. 26–30, Section 3.14 in SI). The Ir signals cannot be detected in CNIrSA and MCIrSA due to low concentration, however, BE shift in N1s spectra provides significant information. Bare CN and MC do not show any measurable shift in N1s spectra under dark and light suggesting carrier generation does not change the overall charge density of the heptazine ring. On the other hand, CNIrSA showed a slightly negative BE shift suggesting increased charge density on Ir-linked Ns sites in C6N7 moieties due to holes transfer on Ir sites. MCIrSA demonstrated the highest N1s BE shift (0.2–0.3 eV) corroborating efficient hole transfer on Ir sites. Electron paramagnetic resonance (EPR) spectra of materials were collected to understand the electronic structure and free carrier concentration under dark and irradiated conditions (Fig. 4a, Section 3.15 in SI) An identical amount of CN and MC-based catalysts were charged in the EPR cavity and irradiated during measurement for a constant time during the irradiation to estimate the relative charge generation rate. The room temperature EPR spectra of CN under dark displayed a Lorentzian peak at a g-value of ≈2.003 due to the presence of unpaired electrons in the π-conjugated system or defect states61. Under visible light irradiation, the EPR peak intensity of CN increased due to the populated HOMO-LUMO electronic transition62. The EPR signal intensity of MC was relatively higher than CN which was attributed due to the low band gap and extended conjugation in the NCN-functionalized structure concomitantly more unpaired electrons are available in the structure. When Ir SA sites were introduced in the CN and MC, the EPR signal intensity increased in agreement with the UV result, further demonstrating electronic transfer to the CN framework from single atom sites. MCIrSA showed the highest signal intensity due to a more pronounced electronic transition. Notably, under light irradiation, the signal intensity of CN/MC IrSA increased which suggests electrons were transferred to the heptazine framework and conversely holes were migrated to Ir sites. As expected, these results suggest effective hole capturing by IrN2O2 SA sites in MCIrSA. The photoluminescence (PL) spectra of CNIrSA and MCIrSA was almost completely quenched compared to metal-free CN suggesting reduced radiative recombination due to efficient hole capturing by undercoordinated Ir metal centers and facile migration of photogenerated electrons in conjugated CN scaffold and non-radiative intersheet recombination (Fig. 4b, Section 3.16 in SI)63,64. The PL intensity of CNIrSA and MCIrSA can also be quenched by metals due to the heavy atom effect that (1) can quench the excited state by photo-induced electron transfer, and (2) by the heavy atom effect, which promotes ISC into dark triplet states. The PL intensity of MCIrSA was slightly higher than CNIrSA which resulted from higher numbers of charge carrier generation in the former thus leading to more radiative transitions. Time-resolved PL (TRPL) decay traces of carbon nitride-based materials can be best fitted by three components which are consistent with 1) direct band-to-band recombination in C–N sp3 composed σ* to lone pair and sp2 C-N π* constituted hybrid (LP+π*) orbital (conduction band) 2) recombination of charge carrier in C-N sp2 π* molecular orbital to LP+π* 3) σ* to π* orbital intersystem crossing (ISC) and inter sheets carrier migration (Fig. 4c)65,66. The individual peak components and their lifetime are summarized in Supplementary Table 3. The average lifetime (τavg) which is a coherent measure of the overall recombination process was also calculated. The average lifetime of CN and MC was found to be 8.9 ± 0.5 and 6.8 ± 0.4 ns respectively. The decreased PL lifetime of MC compared to CN was indicative of the fast quenching of charge carriers. The prodigious PL quenching might be a result of non-radiative transitions and rapid charge migration on more conjugated MC sheets compared to CN followed by non-radiative recombination. After the decoration of Ir SA sites, the τavg of CNIrSA and MCIrSA was calculated to be 8.4 ± 0.4 and 3.9 ± 0.4 ns respectively. The significantly reduced lifetime for MCIrSA (6.8 to 3.9 ns) demonstrates better carrier separation in MCIrSA due to efficient quenching of holes by Ir sites and rapid migration of electrons on MC sheets followed by non-radiative recombination67. The obtained results suggest incorporation of metal centers induces charge separation because of excitons dissociation facilitated by the capture of holes at undercoordinated metal sites, and is in good agreement with UV-Vis and EPR results.

Fig. 4: Photophysical characterization and gas products analysis of plastic conjugates oxidation.
Fig. 4: Photophysical characterization and gas products analysis of plastic conjugates oxidation.
Full size image

a Solid-state X-band EPR spectra of CN, MC, CNIrSA and MCIrSA solid line: light; broken line: dark. b Steady-state PL spectra of melem, CN, MC, CNIrSA and MCIrSA. c Time-resolved PL spectra of melem, CN, MC, CNIrSA and MCIrSA. Dots: raw data; solid line: fitted data. d H2 evolution rate from PET (KOH) plastic derivative solution using MC and MCIrSA catalyst after 12 h. e H2 evolution rate from different plastic solutions using MC as a catalyst. (PLA (blue): polylactic acid, PET (red): polyethylene terephthalate, PS (yellow): polystyrene, PP (green): polypropylene, HDPE (violet): high-density polyethylene. Error bars: standard errors. f H2 evolution rate. g CH4 evolution rate using CN, MC, MCIrNP, CNIrSA and MCIrSA in PLA plastic derivatives solution. Error bars: standard errors. h H2 evolution rate in PET derivatives solution using MC, MCIrSA, CN and CNIrSA. i H2 evolution rate within 48 h using MCIrSA catalysts. Color: Melem (black), CN (yellow), MC (blue), MCIrNP (green), CNIrSA (violet), MCIrSA (red).

Photocatalytic H2 evolution from plastic derivatives photoreforming

Because of improved visible light absorption and better charge separation kinetics of MC and MCIrSA samples, the hydrogen evolution in 1 M KOH was tested under dark and solar-simulated AM1.5G irradiation. Under dark conditions, no trace of H2 can be detected while a moderate H2 generation was observed under visible irradiation. It can be seen from Fig. 4d, that MC displayed an H2 evolution of 24 μmol g−1cat while using MCIrSA the H2 evolution rate was 45 μmol g−1cat after 12 h. After that raw polyethylene terephthalate (PET) plastic derivative solution containing a micro fragment of PET hydrolysis products (terephthalate; PTA and ethylene glycol; EG) in 2 M KOH was tested for the photocatalytic H2 evolution (Supplementary Fig. 31). The EG concentration of PET microplastic derivative (MP) solution was calculated to be 4.63 mmol/L using HPLC (see Methods section). Interestingly, by using MC and MCIrSA catalysts in PET MP (2 M KOH) solution, the H2 evolution rate reached 86 μmol g−1cat and 355 μmol g−1cat after 12 h, respectively. The observed H2 generation rate using MC and MCIrSA reached 7.16 μmol g−1cat h−1 and 29.58 μmol g−1cat h−1.

Encouraged by these results, we explore the scope of MCIrSA catalysts for photocatalytic oxidation of various other condensations (polylactic acid, PLA; polyethylene terephthalate, PET) and addition polymers (polystyrene, PS; polypropylene, PP; high-density polyethylene, HDPE) derivatives. As can be seen in Fig. 4e, addition polymers such as PS, PP and HDPE were almost inactive toward the photocatalytic oxidation since they cannot be fragmented into oligomers/monomers by KOH treatment. On the other hand, condensation polymers PLA and PET conjugates show time-dependent H2 evolution since monomers lactic acid (LA) and EG can be easily oxidized under photocatalytic conditions. Using MC as catalysts, PLA demonstrated much higher activity (78.5 μmol g−1cat after 4 h) compared to PET derivative (17.90 μmol g−1cat after 4 h) (Fig. 4e). The high H2 evolution rate of PLA derivatives was attributed to a relatively high concentration of LA monomer (137 mmol/L) compared to PET derivatives (EG-4.63 mmol/L) (see Methods section). Since PLA derivatives displayed the highest H2 evolution rate using MC, the time-dependent photocatalytic activity of other catalysts (CN, CNIrSA, MCIrSA and MCIrNP) was also tested using PLA derivative solution under visible light (Fig. 4f). The kinetics of H2 evolution was very slow for metal-free CN and MC catalysts (29.8 and 45.6 μmol g−1cat after 4 h). MCIrNP (with 1 wt% IrO2) reveals a linear increment of H2 generation reaching the highest value of 156.2 μmol g−1cat after 4 h. Embellishing Ir SA site on CN (CNIrSA) does not have much influence on the photocatalytic activity and only 76.7 μmol g−1cat H2 was produced in 4 h. Enticingly, MCIrSA boosted the H2 generation rate which was much higher than MCIrNP and CNIrSA. The maximum H2 yield was 442.6 μmol g−1cat in 3 h which dropped slightly in the 4th hour (425.3 μmol g−1cat). We also tested other possible gaseous products such as CO, CH4, C2H4, etc., which were found to be negligible (Fig. 4g and Supplementary Fig. 32). After that, time-dependent H2 evolution using CN, CNIrSA, MC and MCIrSA catalysts was tested which again gave the highest yield for MCIrSA catalysts reaching an H2 production of 175.3 μmol g−1cat after 4 h while CNIrSA can only afford 40.1 μmol g−1cat in 4 h (Fig. 4h). To investigate the influence of MP nanoparticles on photocatalytic performance, a reaction using unfiltered/agitated PET plastic derivative solution was performed which displayed a total H2 yield of 164.7 μmol g−1cat after 4 h. The observed H2 evolution rate was almost similar to the filtered MP solution (175.3 μmol g−1cat) suggesting small plastic fragments do not significantly influence the catalytic performance of the catalyst. To evaluate the scope of catalysts for real plastic waste oxidation, a PET bottle (in 2 M KOH) was treated in identical conditions using MCIrSA catalysts. After 4 h of irradiation, the evolved H2 was found to be ~9.8 μmol g-1cat while other gaseous products such as CH4 and C2H4 were present in negligible amounts (Supplementary Fig. 33). The small amount of H2 in PET bottles digestate, compared to PET derivative substrate might be due to high polymerization degree and the presence of other additives in PET bottles. The liquid products analysis revealed acetic acid was the main product and a significantly low amount of EG was observed (Supplementary Table 4). The oxidation of PET and PLA in the absence of KOH and atmospheric air yielded an almost negligible product suggesting KOH digestion was necessary for the hydrolysis of polymers. When the reaction was carried out under pressurized O2 (8 bar) using PET derivatives and MCIrSA as a catalyst, explicitly high concentration of EG (8127.02 μmol g−1cat), methanol (3267.27 μmol g−1cat) and ethanol (5257.6 μmol g-1cat) was observed depicting O2 promote photooxidation process (Supplementary Table 4). MCIrSA-mediated oxidation of other condensation polymers like Nylon-6 and PVP-29000 yielded a high amount of methanol and ethanol products deciphering the general scope of MCIrSA catalysts for oxidation of a wide variety of condensation polymers (Supplementary Table 4). Furthermore, the catalytic activity of MCIrSA catalysts toward PET derivatives for a long reaction time of c.a. 48 h was determined under AM1.5 G solar simulated light (Fig. 4i). The MCIrSA catalysts reached a maximum H2 yield of 460.2 μmol g−1cat after 24 h and thereafter the concentration of H2 remains stagnant. Also, we observed that after 24 h, MCIrSA PET conjugate solution becomes transparent compared to pale yellow for the fresh solution. This could be due to the fact that CN exfoliates in alkaline solution over time. To confirm that the catalyst remained stable after the reaction and that suppressed activity was merely due to reduced visible absorption in exfoliated sheets, a series of H2 reactions were carried and results are presented in Supplementary Fig. 34. Firstly, MCIrSA catalysts were dispersed in a 2 M KOH solution and the H2 evolution rate was measured as a function of time under AM1.5 G irradiation (Supplementary Fig. 34a). After the photoreaction, the catalysts turned white due to exfoliation and were separated by centrifugation at 16,770 × g for 5 min. The obtained catalyst was redispersed in 2 M KOH and the H2 production rate was monitored. Interestingly, fresh and recycled catalysts show almost similar H2 formation rates substantiating catalytic stability Supplementary Fig. 34c. In another reaction, MCIrSA catalysts after 4 h photoreaction were centrifuged and washed with water to remove any adsorbed surface impurity and H2 evolution was measured (Supplementary Fig. 34b). The H2 evolution rate in this case also remains almost similar for fresh and recycled catalysts confirming that catalyst remain stable after photoreaction (Supplementary Fig. 34d).

To know if PET monomers solution was transformed into liquid oxygenates compounds or completely oxidized to CO2, we determined CO2 concentration by acid titrimetric analysis using methyl orange and phenolphthalein indicators (total alkalinity method) and performed a carbon balance (Fig. 5a, Supplementary Fig. 35 and Supplementary Table 5). Since evolved CO2 can be captured by KOH solution, to form carbonates, the acid titration gives the amount of carbonates (CO2) evolved during the photocatalytic oxidation. Fresh PET derivative solution containing 4.63 mmol/L EG was irradiated under simulated sunlight and evolved gaseous products were analyzed by GC, while liquid organics were analyzed using HPLC. After a 4 h reaction, the calculated carbonate concentration was found to be 12 mmol/L using PET derivative (Fig. 5a). To further validate the stoichiometric degradation of EG monomer, we carried out a photooxidation reaction using 0.1 mmol/L EG and MCIrSA catalyst (Fig. 5b). After 4 h irradiation, the carbonate concentration was found to be 0.21 mmol/L suggesting complete stoichiometric degradation of EG. A long reaction time of 12 h does not increase carbonate yield (0.24 mmol/L) confirming all substrate was utilized within 4 h. The H2 concentration was not determined due to too low concentration for accurate quantification. For PLA derivative the total carbonate concentration was calculated to be 20 mmol/L. Formic acid and glycolic acid show overlapping peaks in HPLC therefore, liquid product quantification cannot be done confidently. However, it can be seen from the HPLC chromatogram that the contribution of formic acid remains too small, and therefore we considered that the entire area corresponds to glycolic acid (C2 product) (Fig. 5c). Based on this the glycolic acid concentration was calculated to be 1.53 mmol/L. While H2 production was 175 μmol g−1cat for 10 mL PET derivative solution which can be translated to 17.5 mmol/L H2. Therefore 4.63 mmol EG (9.26 mmol/L carbon) gives a total of 15.06 mmol/L carbon which was slightly higher than the theoretically calculated value. The additional carbon in carbonate form might be contributed by the decarboxylation of PTA molecules. On the other hand, 4.63 mmol EG should provide 13.98 mmol/L H2. The obtained H2 (17.5 mmol/L) was close to the theoretical value (Supplementary Fig. 35).

Fig. 5: Plastic derivatives oxidation products analysis and recyclability.
Fig. 5: Plastic derivatives oxidation products analysis and recyclability.
Full size image

a The total evolved carbonate (dissolved CO2) (pink) and change in KOH concentration (blue) of the MP derivative solution after the photoreaction. The evolved CO2 converted to carbonate and total carbonate concentration was monitored by titrimetric analysis. b The carbonate concentration of 0.1 mmol/L EG after 4 and 12 h reaction. c HPLC chromatogram of lower to upper panel: pure TPA (black), KOH treated PET derivatives without irradiation (blue) and PET derivatives after 4 h irradiation (magenta) using MCIrSA catalysts. d NMR spectra of lower to upper panel: KOH treated PET derivative without irradiation, PET derivative after photoreaction using MC and MCIrSA catalysts in 0.5–4.2 ppm chemical shift region and e 7.0–8.5 ppm chemical shift region. The colored dots on structures and NMR spectra displayed the chemical shift of corresponding protons in the product. XPS spectra of MCIrSA catalyst after 12 h reaction in f C1s and g N1s regions. h photocatalyst activity of MCIrSA catalysts in the absence and presence of hole (Na2SO3) and •OH (IPA) radical scavengers.

The PET derivative photooxidation product identification using HPLC only demonstrates the presence of glycolic acid, formic acid and HPTA after 4 h (Fig. 5c). Other possible reaction products were not detected in HPLC probably due to too low concentration. Our effort to identify reaction products using electrospray ionization-mass spectroscopy (ESI-MS) reveals the increased intensity of peaks at m/z values of 82 and 114 corresponding to potassium salt of formate and glycolate (Supplementary Fig. 36). Except for a few small changes, at a lower molecular weight range, the ESI-MS of raw PET-derivative reaction product was almost identical to raw PET solution suggesting minimum utilization of terephthalic acid during photoreaction. Therefore, 1H NMR using the water peak suppression method was employed for the identification of the product (Fig. 5d, e and Supplementary Figs. 37–41). The 1H NMR spectra of PET derivatives after a 4 h reaction exhibit signals for glycolic acid (GA), acetic acid (AA), glyoxal, glyoxalic acid, and formic acid, agreed well with previously reported PET derivative oxidation products18,68,69. The oxidation of EG proceeds by the formation of glycoaldehyde, glyoxal, glycolic acid, glyoxylic acid, oxalic acid, and formic acid and finally mineralization to CO270. Few oxidation products such as glycoaldehyde and oxalic acid cannot be identified, probably due to overlapping of peak and low concentration. The aromatic region of 1H NMR spectra displayed intense PTA signals along with traces of isophthalic acid impurity in PET derivative ascribing PTA have not been oxidized under photocatalytic conditions (Fig. 5e).

Since cyameluric acid was synthesized by neutralization of potassium cyamelurate using HCl, residual potassium and chlorine as KCl and doped K were also present in catalysts. To check the influence of residual K and Cl on photocatalytic activity, the final MCIrSA catalyst was washed with water under sonication to remove residual salt followed by photocatalytic performance evaluation for PET conjugates oxidation (Supplementary Fig. 42). After washing with water, the H2 evolution rate was slightly reduced which indicates the reinforcing effect of K on catalytic performance. The XPS analysis of the washed catalyst displayed moderately reduced K signals and a trace of Cl suggesting KCl salt was removed during the washing and only doped K remains in the MC scaffold (Supplementary Figs. 43, 44). However, the catalytic activity of washed MCIrSA catalyst remains significantly higher than pristine MC, corroborating IrN2O2 SA sites were prime active sites leading to boosted H2 evolution. The Ir content of recycled MCIrSA catalysts after 12 h PET derivatives photoreaction was determined using ICP-OES. The Ir content of recycled catalysts was found to be 0.028 wt% (1.478 μmol g−1cat) which was close to fresh catalysts and only a 9.7% drop in Ir content was observed (Supplementary Table 6). XPS analysis of recycled catalysts remains almost similar to fresh MCIrSA catalysts deciphering the absence of any structural change and metal leaching (Fig. 5f, g and Supplementary Figs. 45, 46). Furthermore, Cl-doped C3N4 (Cl-C3N4) and K-doped C3N4 (K-C3N4) were also tested for photocatalytic PET derivatives oxidation which displayed slightly improved gaseous and liquid products yield for K-C3N4 compared to Cl-C3N4 and pristine CN (Supplementary Fig. 47 and Supplementary Table 7). The photoreaction using bare IrO2 NPs and 1.0 wt% IrO2 NPs supported on inert Al2O3 (IrO2/Al2O3) yielded a small amount of H2 compared to CN and doped CN materials. Interestingly, IrO2 NPs impregnated MC (IrO2/MC) displayed the highest H2 evolution rate and liquid products suggesting support materials play a crucial role in improving the catalytic performance.

Mechanism of photoreforming

To elucidate the plausible mechanism and active species responsible for the photoreforming of plastic derivatives, we have performed photocatalytic experiments using IPA and Na2SO3 as hole and •OH radical scavengers, respectively (Fig. 5h). It was found that both hole and •OH radical scavengers reduced the photocatalytic H2 evolution rate with slightly higher reduction for IPA suggesting hole generated •OH radicals were primary active species responsible for improving plastic conjugates degradation. Furthermore, we have carried out photocatalytic experiments using pure terephthalic acid (PTA) which is known to react with •OH radicals and form hydroxy terephthalic acid (HPTA). The catalytic reaction using MCIrSA and PTA followed by HPLC analysis (UV detector) demonstrated the presence of HPTA which again validates the facile generation of •OH radicals (Supplementary Fig. 48). Our DFT findings also support this observation where photogenerated holes are transferred to IrN2O2 sites followed by the generation of •OH radicals (see DFT discussion). To further validate the origin of •OH and other radical species, low-temperature liquid state in situ EPR spectra (Fig. 6) were measured using catalysts and spin trap agents under dark and AM1.5 G irradiation conditions (see Experimental Section in SI for details). The EPR spectra of CN, MC, CNIrSA and MCIrSA using 2,2,6,6-tetramethylpiperidine-1-oxyl (TEMPO) spin trap under dark conditions displayed fingerprint triplet with an intensity of 1:1:1 (Supplementary Figs. 49–52). TEMPO is a positively charged spin trap label and neutralization of charge by electron trapping generates EPR silent hydroxyl amine of TEMPOH71. Therefore, the decreased signal intensity of TEMPO signals represents the extent of charge carrier generation. After visible light irradiation (5–20 min), all catalysts (CN, MC, CNIrSA and MCIrSA) showed reduced signal intensity as a function of irradiation time corroborating the generation of charge carriers under light. Compared to metal-free CN/MC counterparts, CNIrSA and MCIrSA with isolated Ir sites displayed greatly reduced signal intensity deciphering IrN2O2 sites facilitates better charge separation (Fig. 6a). The highest reduction in TEMPO signals was observed for the MCIrSA catalysts due to populated charge carrier generation and efficient hole trapping by IrN2O2 sites. The EPR experiment for all catalysts using 5,5-dimethyl-1-pyrroline N-oxide (DMPO) as a hydroxyl (•OH) radical trap under dark did not exhibit any signal confirming irradiation was essential for generating reactive oxygen species (ROS). Under light irradiation, a well-resolved quartet for DMPO-OH adduct with 1:2:2:1 intensity confirms the formation of hydroxyl (•OH) radicals under light (Supplementary Figs. 49–52). Interestingly, the highest EPR signal intensity of MCIrSA suggests populated •OH radical generation by IrN2O2 SA sites (Fig. 6b). Since IrO2 is a well-proven hole trap agent, the high •OH radical generation rate must be associated with efficient hole scavenging by IrN2O2 SA sites72. The superoxide radical (O2•-) trapping using DMPO displayed a similar trend with the highest DMPO-O2•- quartet (1:1:1:1) intensity for MCIrSA catalyst (Fig. 6c). These results suggest, under aerobic conditions, superoxide radical O2•- can be another potential ROS to oxidize condensation polymers. These observations are supported by increased product formation under the O2 atmosphere (Supplementary Table 4). However, we carried out our reaction under anaerobic conditions and therefore excluded the possibility of O2•- species as potential ROS for plastic derivatives oxidation. Quantitative measurement of generated ROS with the help of EPR signal counts showed that MCIrSA produced the highest amount of ROS which concomitantly produced the highest H2 (Fig. 6d–f).

Fig. 6: EPR spectra.
Fig. 6: EPR spectra.
Full size image

EPR spectra of reaction product after 20 min irradiation under AM1.5 G irradiation using a TEMPO as photogenerated holes trapping agent b DMPO as hydroxyl radicals and c superoxide radicals spin labeling agent and CN, MC, CNIr and MCIr catalysts. Quantitative analysis of photogenerated d holes e hydroxyl radicals and f superoxide radicals under dark and after 5, 10, 15, 20 min irradiation using CN, MC, CNIr and MCIr catalysts. Color: CN (yellow), MC (blue), CNIrSA (violet) and MCIrSA (red).

The non-emissive trapped electrons significantly influence the catalytic properties of carbon nitrides. To investigate the decay kinetics of the excited states and trace charge-carrier dynamics on a picosecond timescale, we utilized femtosecond transient absorption spectroscopy (fs-TAS) (Fig. 7)73. The 2D pseudo-color maps and corresponding spectra at 350 nm excitation for CN, MC and MCIrSA at 0.5 ps-3ns delay time are presented in Fig. 7a–f. All samples showed a weak absorption band at <475 nm originating from the ground state bleaching (GSB) which was consistent with radiative PL emission (420–470 nm) (Fig. 7a–c). Additionally, strong negative bleaching signals in the 475-700 nm range were likely due to stimulated emission (SE) of trapped electrons74. A similar bleach pattern has been reported previously for NCN/NH2 functionalized carbon nitrides75. Compared to CN, these signals were stronger in TAS spectra of MC suggesting populated non-emissive trapped electrons in extended conjugated structure with NCN functionalization (Fig. 7d, e). Additionally, longed-lived signals in ns time scale substantiate delayed recombination of bound electron-hole pairs in NCN functionalized MC76. MCIrSA displayed a similar pattern with a slightly strong band deciphering increased electron density (due to MLCT transition) on the MC framework (Fig. 7f). The positive bleaching signals at higher wavelengths (650–750 nm) occur from excited state absorption (ESA). As expected, signal intensity for MC and MCIrSA were stronger, suggesting higher density of both photogenerated electrons/holes. The decay kinetics of the photogenerated charge carriers were determined by triexponential fitting of TA signals (Fig. 7g). The relative contribution and lifetime of each component are presented in Supplementary Table 8. The three lifetimes for CN were τ1 = 1.720 ps ± 1.7 ps, τ2 = 19.43 ps ± 4.9 ps, and τ3 = 1.760 ns ± 1.8 ns. The first short lifetime was assigned to hot electrons cooling down to the conduction band while the second and third lifetime were associated with exciton annihilation and the shallow trapping state77,78. Distinctively, the lifetime components for NCN-modified MC catalysts were relatively shorter, τ1 = 1.574 ps ± 1.6 ps, τ2 = 17.10 ps ± 17 ps, τ3 = 1.602 ns ± 1.6 ns compared to CN suggesting NCN functionalization introduced effective pathways for excitons dissociation. After introducing IrO2 single atom sites the lifetime for each component was further reduced (τ1 = 1.437 ps ± 0.47 ps, τ2 = 16.21 ps ± 4.6 ps and τ3 = 1.487 ns ± 0.15 ns) corroborating efficient hole trapping by IrN2O2 sites resulting into electron-hole pairs dissociation74. fs-TAS results clearly indicate NCN functionalization and IrN2O2 site’s introduction CN scaffold facilitates carrier’s dissociation essential for improved photocatalytic performance.

Fig. 7: fs-TAS and in situ FTIR analysis.
Fig. 7: fs-TAS and in situ FTIR analysis.
Full size image

Femtosecond TAS contour plot for a CN. b MC. c MCIrSA. and corresponding TAS spectra of the difference in optical density (ΔOD) in milli seconds as a function of wavelength for various delay times (0.5ps-3ns, scale: green to red) probed at 350 nm excitation d CN. e MC. f MCIrSA. g Transient carriers lifetime decay kinetics for CN (black), MC (blue) and MCIrSA (red). h in situ FTIR of MCIrSA, bottom to top: under He and dark (black), under He, no moisture and irradiation for 5 min (gray) and 10 min (purple), after spraying with 2 M KOH and keeping for 5 min (green) and 10 min (yellow), after spraying with 2 M KOH and AM1.5 G irradiation for 5 min (violet) and 10 min (red).

To trace the change in the chemical state of MCIrN2O2 sites under reaction conditions and understand the reaction mechanism, in situ FTIR was carried out under dark and light conditions (Fig. 7h). The in situ FTIR spectra of pristine MCIrSA under dark show weak signals corresponding to triazine ring stretch and bending vibration in the range of 1150–1515 cm−1. A fainted peak at ~975 cm−1 was attributed to Ir-O stretch IrN2O2 entities. The observance of Ir-O stretch for MCIrSA at high wavenumber compared to IrO2 nanoparticle was attributed to the high oxidation state of Ir (~5+) in IrN2O2 than Ir (4+) in IrO279. After irradiation of samples, the bands corresponding to CN skeleton vibration become intense suggesting a dynamic change in the CN framework due to electronic transitions. After the introduction of KOH containing moisture and under dark, intense vibration bands centered at 3296 and 1630 cm−1 were observed due to O-H stretch and H2O bending vibration. The bands corresponding to C6N7 were almost suppressed due to the high signal intensity of OH and H2O peaks. Interestingly, after irradiation of moist samples under light, the C3N3 stretch and bending vibration become prominent, suggesting better charge transfer dynamics. Additionally, a small band at a relatively high frequency 1080 cm−1 was attributed to the Ir6+-O stretch. The higher value of Ir-O stretch after KOH moisture and light exposure suggests the formation of O-Ir-OH transient species which capture photogenerated holes followed by elimination of •OH radical and leaving Ir6+-O sites80. Li et al. also employed synchrotron-based in situ FTIR during OER and reported a similar Ir6+=O peak at a higher wavenumber (946 cm−1)81. The Ir6+-O band probe the formation of O-Ir-OH transient species under alkaline condition. Similar species have also been experimentally reported on the IrO2 crystal surface during OER82.

To validate our assumption, DFT studies were employed on MC and MC-IrN2O2 models and free energy diagrams (FED) for hydrogen evolution reaction, and the desorption of *OH. Bader charges were further calculated to unravel the charge transfer. The H* adsorption on MCIrSA sites was explored in three possible positions 1) C@NCN 2) C@Nbridge 3) C@Obridge (Fig. 8a). On C@NCN and C@Nbridge sites strong adsorption is observed, which is demonstrated by low energies (~−2.0 eV) (Supplementary Fig. 53). The reason for the strong adsorption on these sites is that the H* is not stable on carbon sites and migrates to the nearby N atoms forming a strong bond. As a result, these sites are highly unfavorable for H2 evolution. However, on C@Obridge site the H* adsorption is relatively moderate (0.31 eV) and energy to overcome the thermodynamic barrier for H2 generation can be supplied by visible light. Therefore, MC can generate hydrogen under visible light and C@Obridge site is the main contributor. After the decoration of IrN2O2 SA sites, the energy barrier for C@Nbridge becomes positive while the value for C@NCN remains highly negative (Fig. 8b). The moderate energy for the C@Obridge site again thermodynamically favors H2 generation on C@Obridge sites under photon energy. On the IrN2O2 site, proton adsorption displayed moderate negative energy (~−1.0 eV). However, after the addition of a proton on the Ir atom at the IrN2O2 site, it immediately transfers to O sites forming O-Ir-OH sites which are usually not considered as H2 evolution sites due to high dissociation energy for the O-H bond. Therefore, C@Obridge sites were the main active centers for H2 evolution. To confirm the feasibility of O-Ir-OH species, the formation energies for MCIrN2O2, MCIrN2O, MCIrN2O-OH, and MCIrN2OH-OH were determined (Supplementary Table 9). Interestingly, the formation energy of MCIrN2O-OH and MCIrN2OH-OH were significantly higher than MCIrN2O2 and MCIrN2O corroborating the stability of Ir-bonded OH species.

Fig. 8: DFT analysis.
Fig. 8: DFT analysis.
Full size image

a MCIrN2O2 model structure with possible adsorption sites (circles). b Free energy diagram of H2 generation on different examined active sites displayed in a. c The free energy of •OH formation on Ir site. d The differential Bader charge density of MC-IrN2O2 and MCIrN2O-OH. The yellow and cyan surfaces correspond to the charge gain and lost regions, respectively. e A plausible mechanism of hydroxyl radical generation on MCIrSA catalyst.

The free energy diagram for the formation of •OH radical from the O-Ir-OH is shown in Fig. 8c. Thermodynamic free energy of •OH radical formation from water is reported by Armstrong et al. as 2.73 eV83. This value has been employed to evaluate a hydroxyl radical formation in one electron water oxidation reaction84. We use this value to estimate the free energy of OH desorption from O-Ir-OH to form •OH radical. Our analysis suggests that at least 2.33 eV energy is required to desorb the *OH in O-Ir-OH and form •OH radical, which can be surpassed with long wavelength photons. In our experiment, the equivalent energy of AM1.5 G radiation is between 2.47 and 2.94 eV, which exceeds the energy required to form the •OH radical resulting in facile formation recovery of the active site. The calculation of bond length and coordination of O-Ir-OH and +Ir-O sites demonstrate O-Ir-OH sites have almost identical coordination while for +Ir-O a third slightly weak Ir-N bond (2.50 Å) was formed demonstrating partial charge transfer from MC scaffold to +Ir-O species (Supplementary Fig. 24 and Supplementary Table 2). Bader charge analysis was performed to gain insight into the charge states of the system and the level of interactions (Fig. 8d). The amount of charge transfer from Ir to MCIrSA is +1.79e, indicating a strong interaction between the Ir atom and MC. The positive charge on the Ir atom is favorable for adsorbing •OH radical intermediates, which can effectively promote the reaction. This observation was well aligned with XANES results which suggest Ir has a slightly higher oxidation state (4.7+) and unoccupied 5d states became available. After the formation of O-Ir-OH, the Bader partial charge on Ir was +1.62e. The partial charge lost (cyan) and gained (yellow) are displayed in Fig. 8d and Supplementary Fig. 54. The charge transfer mainly occurred between Ir and OH on the O-Ir-OH system. Supplementary Fig. 55 shows the projected density of states (PDOS) for the MC (top panel) and MCIrN2O2 (bottom panel). For MC, the frontier p orbitals of N and C are the main contributors toward bonding with adsorbates. The peak close to the fermi level (0 eV) indicates that MC can strongly bind with adsorbates and this peak mainly corresponds to the p orbitals of N. However, as soon as the IrO2 SA sites are doped to MC, the peaks are shifted downward, indicating a strong hybridization of the p orbitals of N and C with the d orbital of Ir. This downward shift relative to the Fermi level suggested that MCIrN2O2 would weakly bond with adsorbates compared to the MC. The shift in the projected density of states effect is manifested by the weaker free energy adsorption of hydrogen on MCIrN2O2 (Fig. 8b) compared to MC (Supplementary Fig. 55).

Based on these observations, a general mechanism was proposed (Fig. 8e). In the first step, protons are selectively adsorbed on O-Ir-O to form O-Ir-OH. Then another proton adsorbed to C@Obridge sites to form H-C@Obridge intermediates. Due to the low energy barrier, under visible light, the site can accept photogenerated hole-forming +Ir-O sites after releasing a •OH radical. It is worth mentioning that ground state Ir with ~5 (4.7) oxidation state in MCIrN2O2 most probably will acquire one hole to reach another stable Ir6+ state. At the same time, the H-C@Obridge site acquires photogenerated electrons and protons leading to H2 generation. The +Ir-O site captures OH- ions from the solution to regenerate O-Ir-OH species to continue •OH radical generation cycles. The generated •OH radicals attack plastic monomer fragments and oxidize them into CO2.

Discussion

We synthesized a new carbon nitride (MC) using melem-cyameluric acid H-bonded aggregate. The detailed physicochemical analysis, including ssNMR, FTIR, NEXAFS, and XPS revealed the presence of NCN functionalization and O-bridging in heptazine-constituted carbon nitride scaffold. Due to the condensation polymerization of macromolecular structure and evolution of NCN/O-ligation on C6N7 structure, MC catalysts displayed improved surface and photophysical properties. The MC cavity was selectively decorated with iridium single atom sites which enabled improved H2 evolution performance from plastic derivatives solution. Only condensation polymers derivatives (PLA and PET) were able to produce H2 due to their breakdown in sacrificial monomer units. XANES and FT-EXAFS revealed the presence of isolated IrO2 sites in a high oxidation state (4.7+) due to electronic charge transfer from Ir to the heptazine-conjugated network. The under-coordinated Ir sites embellished in MC deliver high charge generation (ssEPR)/separation (PL/TRPL) actuating high H2 generation under visible irradiation. DFT studies coupled with in situ EPR/FTIR results probe the formation of O-Ir-OH sites followed by thermodynamically favored •OH desorption under light energy leading to the high density of •OH for oxidation of plastic monomers. Future research should aim to improve the H2 evolution rate and replacement of expensive Ir sites with earth-abundant transition metals for real-life application of the catalytic process. The findings presented in this study will direct future research to develop SACs and even cooperative multi-site SACs with improved activity and selectivity for various applications.

Methods

Synthesis of 2,5,8-triamino-s-triazine (melem) 30,85

For the synthesis of melem (2,5,8 -triamino-s-heptazine), 10 g melamine was thermally annealed at 425 °C for 12 h in an alumina crucible. The obtained tri-s-triazine units containing melem were purified by refluxing in DI water followed by filtration of the hot solution to remove any unreacted melamine and low molecular weight moieties. The obtained white paste was dried under a vacuum at 90 °C and ground well. The characterization results were well-matched with previously reported literature86.

Synthesis of cyameluric acid31

For the synthesis of cyameluric acid, 5 g of melon powder87 was charged in a round bottom flask and 100 mL 2.5 KOH was added to make a suspension. The obtained mixture was refluxed for 4 h. The yellow suspension becomes transparent during this process due to the formation of soluble potassium cyamelurate. The obtained solution was cool down at room temperature and acidified using concentrated HCl below 5 °C to obtain a white colloidal solution of cyameluric acid. The white suspension was centrifuged (16,770 × g for 10 min) and washed several times with water until the pH of the supernatant reach 7. The obtained solid was dried under a vacuum at 60 °C.

Synthesis of melem(M)-cyameluric(C) macromolecular conjugate derived NCN-functionalized O-bridged carbon nitride (MC)

Initially, an equimolar amount of melem and cyameluric acid were added to 30 mL of dimethyl sulfoxide (DMSO) and sonicated for 15 min in a sonication bath. The obtained suspension was transferred to a Teflon® lined autoclave and heated at 180 °C for 12 h. This step generates hydrogen-bonded melem-cyameluric acid macromolecular conjugate. The autoclave was cool down to room temperature and a viscous white solid was diluted with DI water. The resulting suspension was centrifuged (16,770 × g for 10 min) and washed with water 2 times followed by washing with ethanol. The obtained solid was dried under vacuum at 80 °C. The resulting white powder was transferred to an alumina crucible and heated at 525 °C for 2 h with a ramping rate of 3 °C/min. The dark yellow powder was used directly for further steps.

Synthesis of O-linked heptazine based carbon nitride (CNO)33

The O-linked carbon nitride was prepared by direct thermal annealing of cyameluric acid powder in an alumina crucible at 550 °C. The obtained spectroscopic results matched well with the reported literature.

Synthesis of melem-derived bulk carbon nitride (CNB)

The thermal annealing of melem powder at 550 °C in an alumina crucible afford bulk carbon nitride with pale yellow color. The obtained powder was grounded and used for further applications.

Synthesis of graphitic carbon nitride (g-C3N4; CN)88

The bulk graphitic carbon nitride powder was synthesized by thermal polycondensation of melamine at 550 °C. 10 g melamine in an alumina crucible covered with a lid was heated at 8 °C min−1 up to 300 °C and 2 °C min−1 up to 550 °C and finally holding the temperature 550 °C for 4 h.

Synthesis of Ir single atom decorated NCN-functionalized O-bridged carbon nitride (MCIrSA)

The Ir single atom decorated MCIrSA catalysts were prepared by dilution self-deposition approach. In brief, 250 mg of MC was dispersed in DI water by 15 min ultrasonication. Then a separate solution of IrCl3.xH2O (0.1 g/mL) was prepared by considering the Ir content in salt. After that 30 μL of this solution was transferred to 50 mL of DI. Water. The resulting dilute solution of IrCl3 was added dropwise to MC suspension with strong stirring. The solution was kept stirring for 6 h to ensure uniform deposition. Finally, the resulting suspension was centrifuged (16,770 × g for 10 min) washed with water and ethanol and dried under a vacuum.

Synthesis of IrO2 nanoparticles decorated NCN-functionalized O-bridged carbon nitride (MCIrNP)

For comparison, IrO2 nanoparticles decorated MC (MCIrNP) were also prepared by adding commercial 1 wt% IrO2 (M.W.-224.22 Sigma Aldrich) to MC samples followed by sonication for 5 h and stirring for 6 h. Finally, the obtained solution was centrifuged (16,770 × g for 10 min) and dried under a vacuum.

Synthesis of Ir single atom decorated graphitic carbon nitride (CNIrSA)

The CN embellished Ir single atom catalysts (CNIrSA) were prepared by following the exact recipe used for the synthesis of MCIrSA except CN was used instead of MC.

Synthesis of IrO2/MC and IrO2/Al2O3

For the preparation of IrO2/MC and IrO2/Al2O3, 100 mg MC or Al2O3 was dispersed in 100 mL DI water, and sonicated for 30 min. After that, 1 wt% IrO2 (Sigma Aldrich) was added into the suspension and stirred for 6 h at room temperature. Finally, water was evaporated at 80 °C and solid catalysts was collected.

Synthesis of K-doped C3N4 89 and Cl-doped C3N4 90

K-doped C3N489 and Cl-doped C3N490 were synthesized by following literature methods. In brief, a suspension of dicyandiamide (2 g) in water (10 mL) was obstained and 0.05 M KOH solution was add with strirring. The resultsing suspenstion was heated at 100 °C until water evaporate followed by drying solid at 80 °C. The obtained solid was ground and annealed at 520 °C for 2 h (5 °C min-1) to obtained K-doped C3N4. For the synthesis of Cl-doped C3N4, cyanuric chloride (15 mmol) and melamine (7.5 mmol) were mixed in 60 mL acetonitride and charged in a 100 mL of Teflon-lined autoclave and mixed for 12 h. After that mixture was heated at 180 °C hydrothermally for 96 h. The resulting solid was washed with DI water and assigned as Cl-doped C3N4.

Photocatalytic plastic derivatives oxidation studies

The plastic derivatives (MP) solution from different polymers was prepared by following the literature procedure with slight modifications17,91. For the preparation of the plastic derivatives solution, different polymer granules were soaked in liquid N2 and grounded with a grinder and passed through a 100 μm mesh. The obtained powders were added to 2 M KOH and stirred at 60 °C for 4 days. The solution was allowed to settle, and the supernatant solution was used for the photocatalytic hydrogen evolution reaction. The initial amount of PET and PLA derivatives used for digestion with KOH was 10 g/L. Since PLA was completely soluble in 2 M KOH the calculated concentration of lactic acid (considering one less OH in PLA (Lactic acid MW: 90 - OH MW:13 = 73) was found to be 137 mmol/L. While PET was not completely soluble, so generated EG was measured by HPLC. The concentration of EG in PET derivatives was calculated to be 4.63 mmol/L.

The photocatalytic performance of catalysts was evaluated by taking 10 mL of plastic derivatives solution and 20 mg of CN-based catalysts in a Borosil glass vial (Thermo Scientific, Crimp Seals with clear PTFE rubber). The solution was sonicated for 15 min to disperse the catalysts followed by sealing with a rubber septum. The solution was purged with nitrogen using a long needle to remove the dissolve gaseous from the solution and headspace. The obtained reaction mixture was irradiated under AM1.5 G irradiation using a solar simulator. The power density at the surface of the reaction vessel was maintained at 100 mW cm−2 using an intensity meter. During the photoreaction, the gaseous products are withdrawn from the vial using a gas-tight syringe and analyzed using a GC equipped with a Molsieve column and TCD/FID detector. The product quantification (H2, CO, CH4) was done by injecting a calculated amount of standard gaseous mixture (Praxair Canada) into the GC followed by the formation of a calibration curve. After the reaction, the liquid products are neutralized using HCl and analyzed using HPLC, ESI-MS and 1H NMR.