The reaction coordinate is of central importance in reaction rate theory. Accurate reaction coordinates facilitate calculations of activation barriers and rates, while also providing essentially complete mechanistic understanding. For bond making/breaking chemistry, accurate reaction coordinates can be constructed from the unstable mode at the saddle point between reactants and products. However for the rugged potential energy landscapes of nucleation processes or protein folding, identifying accurate and physically meaningful reaction coordinates has been a major challenge. I will show how likelihood maximization can systematically identify accurate coordinates that are consistent with the splitting probability by using transition path sampling data. As an example, I will present an accurate reaction coordinate for ion-pair dissocation - a seemingly simple problem but one that has eluded all previous efforts. Our findings have implications for the Grote-Hynes theory of solute and solvent dynamics. I will also show that a physical coordinate which predicts its own dynamical evolution toward two separate states (at short times) also predicts the splitting probability. This motivates a Kullback-Leibler test for coordinates that give dynamically self-consistent projections. The new method identifies reaction coordinates from the short time dynamics, i.e. without transition paths. We demonstrate that the coordinates from the new method are correct with examples from nucleation.
 B. Peters, J. Chem. Phys. (2006)
 B. Peters and B. Trout, J. Chem. Phys. (2006)
 B. Peters, G. Beckham, and B. Trout, J. Chem. Phys. (2007)
 B. Peters, Molec. Sim. (2010)
 R. Mullen, J.-E. Shea, B. Peters, in preparation
 G. Beckham and B. Peters, J. Phys. Chem. Lett. (2011)
 B. Peters and P. Bolhuis, in preparation