Effects of metal-modified catalysts on the desorption performance of mixed amine solutions and machine learning prediction

• Enhancing catalytic desorption performance through metal-modified catalysts. • 2. CO 2 desorption capacity reached 48.96 mmol,36.80% above the blank control. • 3. The relative regenerative load decreased to 73.12% of the blank system. • 4. DF value reaches 1.33 × 10 -6 mol 3 ·kJ -1 ·min -1 , 259.46% higher than the blank. • 5. Spline-KRR achieves an RMSE of 0.00133, with R 2 0.992 and Pearson correlation 0.998. The high energy penalty associated with solvent regeneration is still a major bottleneck in amine-based CO 2 capture. In this work, the effects of five solid acid catalysts on the desorption performance of a mixed-amine solvent were compared, and the HY catalyst with superior desorption behavior was selected and further modified with four transition metals (Co, Mn, Gr and Ce) to enhance its catalytic activity. The findings indicate that the CO 2 desorption capacity and maximum desorption rate of the Co-modified HY catalyst reach 48.96 mmol and 0.02211 mmol/s, corresponding to increases of 36.80% and 35.39% relative to the blank system, while the relative regenerative load decreases to 73.12% of the blank. In the later stage of desorption, the value of the desorption factor (DF) reaches 1.33 × 10 -6 mol 3 ·kJ -1 ·min -1 , which is 259.46% higher than that of the blank, and the catalyst also exhibits good cyclic stability with negligible impact on the absorption performance. Field Emission Scanning Electron Microscope (FESEM), Brunauer-Emmett-Teller (BET), X-ray Diffraction (XRD) and Fourier Transform infrared spectroscopy (FTIR) characterizations indicate that the catalyst does not alter the chemical composition of the absorbent. Furthermore, six machine-learning models were developed to predict the desorption behavior, among which the Spline-KRR model achieves an RMSE of only 0.00133, with R 2 and Pearson correlation coefficients of 0.992 and 0.998, respectively, demonstrating high fitting accuracy without obvious outliers and confirming the reliability of the model. Future efforts will continue optimizing this catalyst system, providing robust technological support for greenhouse gas reduction and climate change mitigation.