| Previous work on slowing down the time evolution of vibrational wavepackets in polyatomic molecules---"freezing" intramolecular vibration redistribution (IVR)---required knowledge of the eigenstate-resolved phase and amplitude structure of the wavepacket. Experiments, however, can not provide such informa tion except at very high cost so such approaches are only useful for computational modeling of the control process. A reliable time-resolved, but still averaged, quantity readily available from pump-probe experiments is the survival probability, P(t) = 0&vbm0;t 2 , of the wavepacket. In order to design an experimentally-viable scheme for controlling the dynamics of vibrational dephasing by interaction with tailored laser pulses, a simple fitness function for control, based only on knowledge of P(t), is developed. Variations of this fitness function are tested in several scenarios to check for robustness and learn about the mechanism of control. Additionally, key features of the requisite feedback loop are developed. These include new versions of the shifted-update-rotation (SUR) wavepacket propagator able to handle phase-controlled electric fields interacting with a molecule, and a novel basis of Haar wavelets to build a physical model of the pulse shaper. A number of technical problems were solved in implementing the genetic algorithm in a wavelet basis, mainly resulting from the need to set bounds on the pulse shaper attenuation and phase shifts. The wavelet approach dramatically improves performance of the adaptive algorithm due to the ability of the wavelet functions to directly represent the relationships so vital to successful control. Finally, setup of a dual tunable femtosecond laser system, liquid crystal pulse shaper and LabWindows control software capable of interchanging experimental and computational P(t) are discussed in detail. This exchangeability allows testing of the genetic algorithm and fitness function by computation simulation in an environment identical to the proposed experiment. |
