Towards an exact treatment of exchange and correlation in materials: application to the "CO adsorption puzzle" and other systems.

It is shown that the errors of present-day exchange-correlation (XC) functionals are rather short ranged. For extended systems, the correction can therefore be evaluated by analyzing properly chosen clusters and employing highest-quality quantum chemistry methods. The XC correction rapidly approaches a universal dependence with cluster size. The method is applicable to bulk systems as well as to defects in the bulk and at surfaces. It is demonstrated here for CO adsorption at transition-metal surfaces, where present-day XC functionals dramatically fail to predict the correct adsorption site, and for the crystal bulk cohesive energy.