Quantum Monte Carlo Benchmark of Exchange-Correlation Functionals for Bulk Water.

The accurate description of the thermodynamic and dynamical properties of liquid water from first-principles is a very important challenge to the theoretical community. This represents not only a critical test of the predictive capabilities of first-principles methods, but it will also shed light into the microscopic properties of such an important substance. Density Functional Theory, the main workhorse in the field of first-principles methods, has been so far unable to properly describe water and its unusual properties in the liquid state. With the recent introduction of exact exchange and an improved description of dispersion interaction, the possibility of an accurate description of the liquid is finally within reach. Unfortunately, there is still no way to systematically improve exchange-correlation functionals, and the number of available functionals is very large. In this article we use highly accurate quantum Monte Carlo calculations to benchmark a selection of exchange-correlation functionals typically used in Density Functional Theory simulations of bulk water. This allows us to test the predictive capabilities of these functionals in water, giving us a way to choose optimal functionals for first-principles simulations. We compare and contrast the importance of different features of functionals, including the hybrid component, the vdW component, and their importance within different aspects of the PES. In addition, in order to correct the inaccuracies in the description of short-range interactions in the liquid, we test a recently introduced scheme that combines Density Functional Theory with Coupled Cluster calculations through a Many-Body expansion of the energy.