Hydrogen peroxide (H₂O₂) photoproduction in seawater with metal-free photocatalysts derived from biomass materials is a green, sustainable, and ultra environmentally friendly way. However, most photocatalysts are always corroded or poisoned in seawater, resulting in a significantly reduced catalytic performance. Here, we report the metal-free photocatalysts (RUT-1 to RUT-5) with in-situ generated carbon dots (CDs) from biomass materials (Rutin) by a simple microwave-assisted pyrolysis method. Under visible light (λ ≥ 420 nm, 81.6 mW/cm²), the optimized catalyst of RUT-4 is stable and can achieve a high H₂O₂ yield of 330.36 μmol/L in seawater, 1.78 times higher than that in normal water. New transient potential scanning (TPS) tests are developed and operated to in-situ study the H₂O₂ photoproduction of RUT-4 under operation condition. RUT-4 has strong oxygen (O₂) absorption capacity, and the O₂ reduction rate in seawater is higher than that in water. Metal cations in seawater further promote the photo-charge separation and facilitate the photo-reduction reaction. For RUT-4, the conduction band level under operating conditions only satisfies the requirement of O₂ reduction but not for hydrogen (H₂) evolution. This work provides new insights for the in-situ study of photocatalyst under operation condition, and gives a green and sustainable path for the H₂O₂ photoproduction with metal-free catalysts in seawater.

Carbon dots (CDs) have uniquely structural, physicochemical and photochemical properties, suggesting a promising platform for catalysis applications. The in-depth understanding of the structure-activity relationship in the CDs-based catalyst system needs to know the effect of the crystalline core on their catalytic performance. The efficient catalytic oxidation of cyclohexane is an urgent challenge in current chemical industry, in which, adipic acid (AA) plays an important role in industry for synthesis of nylon-6 and nylon-66. Here, we fabricated the pristine CDs by electrochemical etching graphite rod method and derived CDs with high crystalline core (CD-600, CD-800, and CD-1100) through a thermal treatment method in tube furnace. Furthermore, these CDs performed an outstanding catalytic performance for one-step synthesis of AA from cyclohexane. With the help of machine learning (ML), the deep correlations between features (structures of CDs, catalytic conditions) and catalytic performances were investigated by XGBoost (XGB) model. Then under the optimization and prediction of XGB, it was found that high crystalline core preceded the other features and CD-1100 could get the best conversion of 30.696% and selectivity to AA of 92.52% at reaction conditions of 130 °C, 1.5 MPa, and 10 h. This work pioneered the application of ML in industrial issues and demonstrated a comprehensive understanding on CDs as catalyst to realize one-step synthesis of AA.

Halide perovskite nanocrystals are potential catalysts for CO₂ photoreduction, while, the strong radiative recombination and insufficient stability limit their catalytic performance and application. Herein, we report that layered double hydroxide nanosheets activate CsPbBr₃ nanocrystals (CLDH) for enhanced photocatalytic CO₂ reduction. These CLDH heterojunctions show the remarkably enhanced CO₂ photoreduction performance; without cocatalyst and sacrificial agent, the average electron consumption rate of CLDH (49.16 μmol·g⁻¹·h⁻¹) is approximately 3.7 times higher than that of pristine CsPbBr₃. Also, CLDH catalyst exhibits a robust stability after ten cycles over 30 h.

Great attention has been paid to green procedures and technologies for the design of environmental catalytic systems. Biomass-derived catalysts represent one of the greener alternatives for green catalysis. Photocatalytic production of hydrogen peroxide (H₂O₂) from O₂ and H₂O is an ideal green way and has attracted widespread attention. Here, we show a metal-free photocatalyst from cellulose, which has a high photocatalytic activity for the photoproduction of H₂O₂ with the reaction rate up to 2,093 μmol/(h·g) and the apparent quantum efficiency of 2.33%. Importantly, a machine learning model was constructed to guide the synthesis of this metal-free photocatalyst. With the help of transient photovoltage (TPV) tests, we optimized their fabrication and catalytic activity, and clearly showed that the formation of carbon dots (CDs) facilitates the generation, separation, and transfer of photo-induced charges on the catalyst surface. This work provides a green way for the highly efficient metal-free photocatalyst design and study from biomass materials with the machine learning and TPV technology.