Time scales and other problems in linking simulations of simple chemical systems to more complex ones

We shall start with very small systems like H₂ and H₃, computed with very accurate methods (Hylleraas–CI) or atomic systems up to Zn with accurate methods (CI), then move to more complex ones, like C₆₀, but now with somewhat less accurate methods, specifically Hartree–Fock with density functionals, the latter for the correlation energy but not for the exchange energy. For even more complex tasks like geometry optimization of C₆₀, we have resorted to even simpler and parametrized methods, like local density functionals. Then, we could use quantum mechanics either to provide interaction potentials for classical molecular dynamics or to directly solve dynamical systems, in a quantum molecular dynamics approximation. Having demonstrated that we can use the computational output from small systems as input to larger ones, we discuss in detail a new model for liquid water, which is borne out entirely from ab initio methods and nicely links spectroscopic, thermodynamics, and other physicochemical data. Concerning time scales, we use classical molecular dynamics to determine friction coefficients, and with these we perform stochastic dynamic simulations. The use of simulation results from smaller systems to provide inputs for larger system simulations is the “global simulation” approach, which, today, with the easily available computers, is becoming more and more feasible. Projections on simulations in the 1996–1998 period are discussed, new computational areas are outlined, and a N⁴ complexity algorithm is compared to density functional approaches. © 1993 John Wiley & Sons, Inc.