Fighting dogma
by tzvicky
In the last paper of Seabra, Walker and Roitberg (DOI:10.1021/jp903474v) an interesting point is made: “It is a general dogma in the field of Computational Chemistry that the ability of theoretical methods to reproduce experimental properties is roughly proportional to computational cost.” Authors show that semiempirical Hamiltonians not always provide better results that their classical force-fields counterparts. In particular, in some QM/MM calculations of biological molecules they are defeated by classical force-fields in prediction of experimental thermodynamic properties. Btw, today I managed to (almost) perfectly describe the hydrogen bonding of dimethyl ether with water using empirical force-field. Two hours later it occurred that the quantum DFT+D in cp2k does it better, but still my classical force-field is muuuuch faster.
