Predict the Polarizability and Order of Magnitude of Second Hyperpolarizability of Molecules by Machine Learning

In order to determine the polarizability and hyperpolarizabilityof a molecule, several key parameters need to be known, includingthe excitation energy of the ground and excited states, the transitiondipole moment, and the difference of dipole moment between the groundand excited states. In this study, a machine-learning model was developedand trained to predict the molecular polarizability and second-orderhyperpolarizability on a subset of QM9 data set. The density of stateswas employed as input to the model. The results demonstrated thatthe machine-learning model effectively estimated both polarizabilityand the order of magnitude of second-order hyperpolarizability. However,the model was unable to predict the dipole moment and first-orderhyperpolarizability, suggesting limitations in its ability to predictthe difference of dipole moment between the ground and excited states.The computational efficiency of machine-learning models compared totraditional quantum mechanical calculations enables the possibilityof large-scale screening of molecules that satisfy specific requirementsusing existing databases. This work presents a potential solutionfor the efficient exploration and analysis of molecules on a largerscale.