Geometry Optimization Speedup Through a Geodesic Approach to Internal Coordinate Optimization
We present a new geodesic-based method for geometry optimization in a basis of redundant internal coordinates.<br>This method realizes displacements along internal coordinates by following the geodesic generated by the displacement vector on the internal coordinate manifold.<br>Compared to the traditional Newton method approach to taking displacements in internal coordinates, this geodesic approach substantially reduces the number of steps required to reach convergence on a molecular structure minimization benchmark.<br>This new geodesic method can in principle be implemented in any existing optimization code, and only requires the implementation of derivatives of the Wilson B-matrix and the ability to solve a relatively inexpensive ordinary differential equation.