We present a machine learning model to predict the electric dipole moment of diatomic molecules using only the atomic properties of the constituent atoms. The model can screen the entire periodic table to identify the molecules with the largest dipole moments for applications in cold molecular sciences or to find the molecule with the largest dipole moment that contains a given atom. Similarly, our model identifies useful trends that explain molecular dipole moments, improving our intuition in chemical physics beyond the paradigm of electronegativity differences. Finally, we condense our model into an analytical expression to predict the dipole moment in terms of atomic properties.
Elhalawani et al. (Mon,) studied this question.