Characterization of Dynamics and Reactivities of Solvated Ions by Ab Initio Simulations THOMAS S. HOFER, HUNG T. TRAN, CHRISTIAN F. SCHWENK, BERND M. RODE Department of Theoretical Chemistry, Institute of General, Inorganic and Theoretical Chemistry, University of Innsbruck Innrain 52a, A-6020 Innsbruck, Austria Received 22 May 2003; Accepted 2 July 2003 Abstract: Based on a systematic investigation of trajectories of ab initio quantum mechanical/molecular mechanical simulations of numerous cations in water a standardized procedure for the evaluation of mean ligand residence times is proposed. For the characterization of reactivity and structure-breaking/structure-forming properties of the ions a measure is derived from the mean residence times calculated with different time limits. It is shown that ab initio simulations can provide much insight into ultrafast dynamics that are presently not easily accessible by experiment. © 2003 Wiley Periodicals, Inc. J Comput Chem 25: 211–217, 2004 Key words: solvated ions; ab initio simulations; dynamics Introduction As one of the fundamental factors determining chemical reactivity, dynamics have always been the subject of intensive research. This is true, in particular for the dynamics of complex formation and dissociation reactions involving metal ions, solvent molecules, and other ligands, as such reactions determine not only industrial processes but also numerous essential biological systems. 1–3 Experimental methods have been successively developed to get access to ever faster exchange reactions by physicochemical mea- surements such as stopped-flow kinetics or ultrasonic techniques as well as spectroscopic methods such as nuclear magnetic reso- nance. The time scale of these methods also sets the limits for the measurability of characteristic data for very fast reactions. NMR methods allow accurate direct measurements of velocity constants up to 10 5 s -1 , and estimations via line widths to 10 9 s -1 . However, numerous ligand exchange processes as well as other dynamic processes within the complexes such as the Jahn–Teller distortions, occur on a much faster time scale than presently accessible by experimental tools. 4 Applying statistical mechanics to the processes in condensed systems at the molecular level has created the very successful tools of Monte Carlo (MC) and Molecular Dynamics (MD) simula- tions. 5 The latter method describes the time-dependent events in very small time steps (femtoseconds or fractions of femtoseconds) and thus supplies trajectories suitable to “measure” dynamics on the femtosecond and picosecond scale. This approach should be, therefore, the ideal tool to analyze ligand exchange reactions occurring so fast that experimental investigations are not yet feasible. However, methodical problems associated with the sim- ulation technique have to be clarified and the evaluation of trajec- tories to be standardized before using the simulation-based kinetic data as reliable and realistic measures, in particular when they are lying beyond the experimentally accessible time scale. In several previous works we have been able to demonstrate that classical, even three-body corrected MD simulations are ca- pable of supplying fairly accurate data for structure and composi- tion of the first coordination shell of main group 6–8 and transition metal ions, 9 –12 but that they fail to predict correct data for libra- tional/vibrational motions and for the composition of the second coordination shell. These data are of great importance, however, to describe dynamical processes, in particular ligand exchange reac- tions between both coordination spheres and with the bulk. Hybrid ab initio Hartree–Fock quantum mechanical/molecular mechanical simulations 13,14 proved to achieve the necessary accuracy, al- though at enormous costs in terms of computational effort, which increases by a factor of 100. On the other hand, the use of simple density functionals as in the RIDFT method 15,16 or in Car–Par- rinello simulations 17 can lead to first-shell coordination numbers that are lower than almost all experimental and other theoretical data, for example, six for Ca(II) and five for Cu(II). 6,18 –20 The more sophisticated B3LYP-DFT 21 formalism for the quantum mechanically treated region in QM/MM simulations leads to a similar degree of accuracy as the ab initio Hartree–Fock method, but without saving computer time. 6 The limits of the presently Correspondence to: B. M. Rode; e-mail: bernd.m.rode@uibk.ac.at Contract/grant sponsor: Austrian Science Foundation; Contract/grant number: P10221-N08 © 2003 Wiley Periodicals, Inc.