Predicting solubility of ion pairs in aqueous inorganic chemistry

Polyoxometalates (POMs), ranging in size from several to 100’s of ångströms, represent building blocks of inorganic materials. Elucidating their complex solubility behavior with alkali countercations can inform and model natural and synthetic processes in aqueous media. In the study of POMs ([Nb 24 O 72 H 9 ] 15− , Nb 24 ) we discovered an unusual solubility trend (termed anomalous solubility) of alkali-POMs, in which Nb 24 is most soluble with the smallest (Li + ) and largests (Rb/Cs + ) alkalis, and least soluble with Na/K + . Via computation, we define a descriptor (σ-profile) and use an artificial neural network (ANN) to accurately predict all three described alkali-anion solubility trends: amphoteric, normal (Li + > Na + > K + > Rb + > Cs + ), and anomalous (Cs + >Rb + >K + >Na + >Li + ). Testing predicted amphoteric solubility along with X-ray scattering and single-crystal X-ray diffraction studies affirmed the accuracy of the desciptor, provided solution-phase snapshots of alkali-POM interactions, yielded a new POM formulated [Ti 6 Nb 14 O 54 ] 14- , and provided guidlines to exploit alkali-POM interactions for new POMs discovery.