Towards linear scaling in continuum solvent models.: A new iterative procedure for energies and geometry optimizations

We present a new implementation of the polarizable continuum model, which allows the computation of solvation energies and their gradients for large systems. This is obtained by a balanced two-term iterative procedure free from any N3 computational bottleneck and from any N2 storage requirement, N being the number of finite elements used to describe the cavity occupied by the solute. Comparison with the results obtained by the standard matrix-inversion algorithm for small and medium-sized systems shows that the new algorithm couples a fully satisfactory numerical accuracy with remarkable computational efficiency.