Predictive modeling: Solubility of C 60 and C 70 fullerenes in diverse solvents.

Solubility of fullerenes imposes a major limitation to further advanced research and technological development using these novel materials. There have been continued efforts to discover better solvents and their properties that influence the solubility of fullerenes. Here, we have developed QSPR (quantitative structure-property relationship) models based on structural features of diverse solvents and large experimental data for predicting the solubility of C and C fullerenes. The developed models identified most relevant features of the solvents that encode the polarizability, polarity and lipophilicity properties which largely influence the solubilizing potential of the solvent for the fullerenes. We also established Inter-moieties solubility correlations (IMSC) based quantitative property-property relationship (QPPR) models for predicting solubility of C and C fullerenes. The QSPR and QPPR models were internally and externally validated deriving the most stringent statistical criteria and predicted C and C solubility values in different solvents were in close agreement with the experimental values. In test sets, the QSPR models yielded high correlations (R > 0.964) and low root mean squared error of prediction errors (RMSEP< 0.25). Results of comparison with other studies indicated that the proposed models could effectively improve the accuracy and ability for predicting solubility of C and C fullerenes in solvents with diverse structures and would be useful in development of more effective solvents.