Performance of plane-wave-based LDA+U and GGA+U approaches to describe magnetic coupling in molecular systems.

This work explores the performance of periodic plane wave density functional theory calculations with an on-site Coulomb correction to the standard LDA and GGA exchange-correlation potential--commonly used to describe strongly correlated solids--in describing the magnetic coupling constant of a series of molecular compounds representative of dinuclear Cu complexes and of organic diradicals. The resulting LDA+U or GGA+U formalisms, lead to results comparable to experiment and to those obtained by means of standard hybrid functionals provided that the value of the U parameter is adequately chosen. Hence, these methods offer an alternative efficient computational scheme to correct LDA and GGA approaches to adequately describe the electronic structure and magnetic coupling in large molecular magnetic systems, although at the expenses of introducing an empirical (U) parameter. For all investigated copper dinuclear systems, the LDA+U and GGA+U approaches lead to an improvement in the description of magnetic properties over the original LDA and GGA schemes with an accuracy similar to that arising from the hybrid B3LYP functional, by increasing the on-site Coulomb repulsion with a moderate U value. Nevertheless, the introduction of an arbitrary U value in the 0-10 eV range most often provides the correct ground-state spin distribution and the correct sign of the magnetic coupling constant.