Designing Heterodiatomic Carbon Hydrangea Superstructures via Machine Learning-Regulated Solvent-Precursor Interactions for Superior Zinc Storage.

Carbon superstructures with exquisite morphologies and functionalities show appealing prospects in energy realms, but the systematic tailoring of their microstructures remains a perplexing topic. Here, hydrangea-shaped heterodiatomic carbon superstructures (CHS) are designed using a solution phase manufacturing route, wherein machine learning workflow is applied to screen precursor-matched solvent for optimizing solvent-precursor interaction. Based on the established solubility parameter model and molecular growth kinetics simulation, ethanol as the optimal solvent stimulates thermodynamic solubilization and growth of polymeric intermediates to evoke CHS. Featured with surface-active motifs and consecutive charge transfer paths, CHS allows high accessibility of zincophilic sites and fast ion migration with low energy barriers. A anion-cation hybrid charge storage mechanism of CHS cathode is disclosed, which entails physical alternate uptake of Zn/CFSO ions at electroactive sites and chemical bipedal redox of Zn ions with carbonyl/pyridine motifs. Such a beneficial electrochemistry contributes to all-round improvement in Zn-ion storage, involving excellent capacities (231 mAh g at 0.5 A g; 132 mAh g at 50 A g), high energy density (152 Wh kg), and long-lasting cyclability (100 000 cycles). This work expands the design versatilities of superstructure materials and will accelerate experimental procedures during carbon manufacturing through machine learning in the future.