Decoherence Allows Model Reduction in Nonadiabatic Dynamics Simulations

A nonadiabatic (NA) molecular dynamics (MD) simulation requires calculation of NA coupling matrix elements, the number of which scales as a square of the number of basis states. The basis size can be huge in studies of nanoscale materials, and calculation of the NA couplings can present a significant bottleneck. A quantum-classical approximation, NAMD overestimates coherence in the quantum, electronic subsystem, requiring decoherence correction. Generally, decoherence times decrease with increasing energy separation between pairs of states forming coherent super-positions. Since rapid decoherence stops quantum dynamics, one expects that decoherence-corrected NAMD can eliminate the need for calculation of NA couplings between energetically distant states, notably reducing the computational cost. Considering several types of dynamics in a semiconductor quantum dot, we demonstrate that indeed, decoherence allows one to reduce the number of needed NA coupling matrix elements. If the energy levels are spaced closer than 0.1 eV, one obtains good results while including only three nearest-neighbor couplings, and in some cases even with just the first nearest-neighbor coupling scheme. If the energy levels are spaced by about 0.4 eV, the nearest-neighbor model fails, while three or more nearest-neighbor schemes also provide good results. In comparison, the results of NAMD simulation without decoherence vary continuously with changes in the number of NA couplings. Thus, decoherence effects induced by coupling to a quantum-mechanical environment not only provide the physical mechanism for NAMD trajectory branding and improve the accuracy of NAMD simulations,but also afford significant computational savings.