To guess or not to guess excited state amplitudes during optimization and dynamics

We report robust initial guesses for the amplitudes and z-vectors in a configuration interaction singles or Tamm-Dancoff approximation calculation that consistently reduce the total number of iterations required for an excited state calculation often by over 50%. The end result of these guesses is that the practicing chemist can expect to generate excited state optimized structures with a total wall time reduced by as much as 30% in the future without any approximations—simply by using information gathered at one geometry and applying it to another geometry.