Some comments on the DFT + D method

Although current implementations of density functional theory (DFT) provide a fast and often accurate way of calculating interaction energies, U most are incapable of describing the dispersion interactions between molecules. To overcome this problem an estimate of dispersion energy is often added to U calculated by DFT (DFT + D methods). As these methods involve empirical parameterization they have limited predictive power, unknown applicability and an uncertain theoretical basis. We present an analysis to show what assumptions are necessary to derive DFT + D and show that the main theoretical problem remaining lies with the choice of the approximate DFT method.