Herein, we report machine learning algorithms by training datasets from a set of both succeeded and failed experiments for studying the crystallization propensity of metal-organic nanocapsules (MONCs). Among a variety of studied machine learning algorithms, XGBoost affords the highest prediction accuracy of > 90%. The derived chemical feature scores that determine importance of reaction parameters from the XGBoost model assist to identify synthesis parameters for successfully synthesizing new hierarchical structures of MONCs, showing superior performance to a well-trained chemist. This work demonstrates that the machine learning algorithms can assist the chemists to faster search for the optimal reaction parameters from many experimental variables, whose features are usually hidden in the high-dimensional space.

Water oxidation can lead to a sustainable source of energy, but for water oxidation catalysts to be economical they must use earth abundant metals. We report here 2:1 6,6′-dihydroxybipyridine (6,6′-dhbp)/copper complexes that are capable of electrocatalytic water oxidation in aqueous base (pH = 10–14). Two crystal structures of the complex that contains 6,6′-dhbp and copper(II) in a ratio of 2:1 (complex 1) are presented at different protonation states. The thermodynamic acid dissociation constants were measured for complex 1, and these show that the complex is fully deprotonated above pH = 8.3 (i.e., under water oxidation conditions). CW-EPR, ENDOR, and HYSCORE spectroscopy confirmed that the 6,6′-dhbp ligand is bound to the copper ion over a wide pH range which shows how pH influences precatalyst structure. Additional copper(II) complexes were synthesized from the ligands 4,4′-dhbp (complex 2) and 6,6′-dimethoxybipyridine (complexes 3 and 4). A zinc complex of 6,6′-dhbp was also synthesized (complex 5). Crystal structures are reported for 1 (in two protonation states), 3, 4, and 5. Water oxidation studies using several of the above compounds (1, 2, 4, and 5) at pH = 12.6 have illustrated that both copper and proximal OH groups are necessary for water oxidation at a low overpotential. Our most active catalyst 1 was found to have an overpotential of 477 mV for water oxidation at a moderate rate of k_cat = 0.356 s^–1 with a competing irreversible oxidation event at a rate of 1.082 s^–1. Furthermore, our combined work supports previous observations in which OH/O^– groups on the bipyridine rings can hydrogen bond with metal bound substrate, support unusual binding modes, and potentially facilitate proton coupled electron transfer.

Chitin fibers, prepared by extracting chitin directly from shrimp shell waste and dry-jet wet spinning from the resulting ionic liquid (1-ethyl-3-methylimidazolium acetate) solution in a one-pot process, were surface modified by taking advantage of the insolubility of chitin in common solvents (e.g., water, organics). In this proof of concept example, the chitin fiber surfaces were first deacetylated using aqueous NaOH to make available the primary amine (the functional group of chitosan) on the surface. Further treatment of the fibers allowed for the task-specific tailoring of the functionality (here we appended amidoxime for the extraction of aqueous uranyl ions from seawater). Compositional analysis and physical property measurements (e.g., tensile strength and thermal decomposition) of the fibers before and after surface modification indicated minimal change to the bulk material; however, spectroscopy and sorption studies of uranyl ions from aqueous solution demonstrated surface modification. The lower cost, one-pot process used in this study resulted in weak and brittle fibers, suggesting that additional purification of the chitin before pulling fibers will greatly improve the strength and utility of the resulting material. Overall, a platform has been developed for the surface modification of chitin fibers that provides both the physical properties of chitin and the functional properties of chitosan, resulting in an advanced material from a biorenewable resource with reduced chemical input.