Hypervalent iodine compounds are widely employed in organic transformations, including selective oxidation, C–X bond formation, oxidative functionalisation, and oxidative dearomatisation reactions, among others. Owing to their high selectivity and operational controllability, these reagents have found broad application in both stoichiometric and catalytic manifolds. Among iodine(V) species, o-iodoxybenzoic acid (IBX) is one of the most extensively studied and commonly used reagents. However, despite its advantages, the intrinsic reactivity of IBX is often limited, resulting in slow reaction rates or the inability to promote a range of demanding oxidative transformations, particularly in comparison with more reactive reagents such as Dess–Martin periodinane (DMP).
To overcome these limitations, significant efforts have been devoted to the development of more reactive IBX derivatives. Previous studies have shown that sterically demanding substituents ortho to the iodine can enhance reactivity by promoting favourable structural distortions, while other approaches have explored the introduction of diverse functional groups to modulate electronic properties. Building on these literature precedents, we constructed a comprehensive dataset of 284 substituted IBX derivatives, for which the relevant transition states were optimised using density functional theory (DFT). Machine learning analysis of this transition-state-level dataset revealed clear correlations between reactivity and key geometrical descriptors.
Guided by the combined DFT and machine learning insights, four iodine(V) structures were selected for experimental evaluation. Their reactivity was assessed using menthol oxidation as a stringent benchmark reaction to probe intrinsic oxidising power. Three of the investigated iodine(V) reagents are reported for the first time and exhibit substantially enhanced reactivity relative to IBX. Notably, 3-CF₃-IBX displays markedly faster conversion and higher overall efficiency, in excellent agreement with computational predictions.
Overall, this study demonstrates that literature-informed machine learning anchored to DFT data enables predictive control over hypervalent iodine(V) reactivity and provides a general strategy for the rational development of highly reactive iodine(V) reagents for organic synthesis.