Machine learning-assisted thermomechanical coupling fabrication of hard carbon for sodium-ion batteries

This study conducts an extensive investigation into the application of hard carbon in sodium-ion batteries. More than 100 hard carbon samples are meticulously synthesized by precisely controlling temperature and pressure, and their microstructure and performance are comprehensively evaluated. This research introduced innovative standardized parameters, including degree of graphitization, degree of crystallinity and degree of defect, which can be universally applied to the field of amorphous carbon materials. Furthermore, the utilization of machine learning facilitates rapid prediction and screening of high-performance hard carbon materials. As a result, a fully optimized hard carbon anode is successfully developed. At a current density of 0.03 A g⁻¹, the material demonstrates a reversible capacity of 375 mAh g⁻¹ and an initial coulombic efficiency of 92.3 %. At a high current rate of 2 A g⁻¹, it delivers a capacity of 250 mAh g⁻¹. The investigation further provides robust evidence for the pore-filling mechanism of Na clusters by in situ spectroscopic studies. This study not only introduces novel approaches for controlling and quantifying the microstructure of hard carbon but also demonstrates a case study of efficient material development by machine learning.