Computationally using Density Functional Theory (DFT), is there any examples where Local Density Approximation (LDA) would be preferred over using Gradient Generalized Approximation (GGA) methods for the exchange-correlation term of the Kohn-Sham equation? By examples I mean some sort of collection of atoms like a crystal, molecule, etc.. If there are any obvious reasons, I would much like to hear them.
In earlier DFT studies of ferroelectric materials, GGAs such as PBE were avoided as they tended to exaggerate the ferroelectric distortion. Instead, LDA calculations were performed and an artificial (offset) pressure was applied to compensate for LDA otherwise overestimating lattice constants Philippe Ghosez, Javier Junquera: cond-mat/0605299 "First-Principles Modeling of Ferroelectric Oxide Nanostructures". With newer, solid-state optimized GGAs such as PBEsol, which not only reproduces LDA at zero density gradient, but also recovers low order perturbation of the homogeneous electron gas, lattice constants and distortions in such crystals are reproduced much better. If it wasn't for these newer solid-state type GGAs, LDA at offset pressure would be 'the best functional' for studying ferroelectric phase transitions.