Multi-State VALBOND for Atomistic Simulations of Hypervalent Molecules, Metal Complexes, and Reactions.

The implementation, validation, and application of the multi-state VALBOND method for transition-metal-containing and hypervalent molecules are presented. This approach is particularly suited for molecules with unusual shapes and systems that need to be described by a superposition of resonance structures, each of which satisfies the octet rule. The implementation is based on the original VALBOND force field and allows us to smoothly switch between resonance structures, each of which can be characterized by its own force field, including varying charge distributions and coupling terms between the states. The implementation conserves total energy for simulations in the gas phase and in solution and is applied to a number of topical systems. For the small hypervalent molecule ClF3, the barrier for pseudorotation is found to be 4.3 kcal/mol, which compares favorably with the experimentally measured value of 4.8 kcal/mol. A transition-metal-containing complex, cisplatin, is characterized by six resonance states, for which the vibrational spectrum is found to be in good agreement with experiment. Finally, umbrella sampling simulations of the SN2 reaction BrMe + Cl- → Br- + MeCl in solution yield a barrier height of 24.6 kcal/mol, in good agreement with experiment (24.7 kcal/mol).

Toward a Broadly Applicable Force Field for d(6)-Piano Stool Complexes.

Three-legged piano stool complexes are prototypical organometallic complexes relevant to a wide range of chemically relevant questions. Force field parametrization of transition-metal complexes is difficult and underdeveloped, and metal-specific force fields and software are required. Here we report our efforts to derive parameters for the conventional CHARMM and the Valbond-CHARMM force fields for d(6)-piano stool complexes. In Valbond-CHARMM, the usual angular term is replaced with hybrid orbital strength functions. These functions describe the energy not only of distorted bond angles around the minimum but also at very large distortions. Structure optimizations led to a good agreement between the calculated force field and the X-ray structures. They were comparable to RMSDs obtained between X-ray and DFT structures. In addition, and contrary to treating the systems with DFT, molecular dynamics simulations on the multiple nanosecond time scale are possible and allow to compute meaningful structural and energetic observables. Explicit solvent simulations of the complexes in methanol and water allow to determine the solvent distribution around the complexes. The parametrization presented here will be a useful starting point for dynamics investigations of catalysts in structurally more demanding environments.