One of the still-unsolved questions is an appropriate selection of experimental data that would allow for a more-or-less unique determination of the coupling constants defining the energy functional. To this end, one usually uses certain constraints obtained by extrapolating nuclear data to an infinite system and selected data for finite nuclei. The sensitivity of the final fit to the choice of this data set leads to a plethora of parameterizations currently available in the literature.
Most of the currently used density functionals correctly reproduce generic trends in nuclear masses - as selected masses are usually considered in the data set - but their descriptions of other quantities vary. Moreover, they often significantly differ in parameters or coupling constants [8]. This suggests that yet-unresolved correlations may exist between these parameters, and only certain combinations thereof are important [76,51]. Such correlations would explain the fact that widely different parameterizations lead to fairly similar results.
The present stage of theory requires constructing new energy density functionals supplemented by a complete error and covariance analysis. It is not sufficient to ``predict'' properties of exotic nuclei by extrapolating properties of those measured in experiment. One must also quantitatively determine errors related to such an extrapolation. Moreover, for experimental work it is essential that an improvement gained by measuring one or two more isotopes be quantitatively known. From a theoretical perspective, one must also know the confidence level with which the parameters of the functional are determined. An analysis of this type constitutes a standard approach in other domains of physics, but they are seldom performed in theoretical nuclear structure research.