Department of Chemistry, Northwestern University, 2145 Sheridan Road, Evanston, Illinois 60208, USA.
J Chem Phys. 2010 Apr 7;132(13):134112. doi: 10.1063/1.3314248.
We present a new algorithm for mixed quantum-classical dynamics that helps bridge the gap between mean-field (Ehrenfest) and surface-hopping dynamics by defining a natural rate of decoherence. In order to derive this decoherence result, we have expanded the number of independent variables in the usual Ehrenfest routine so that mixed quantum-classical derivatives are now propagated in time alongside the usual Ehrenfest variables. Having done so, we compute a unique rate of decoherence using two independent approaches: (i) by comparing the equations of motion for the joint nuclear-electronic probability density in phase space according to Ehrenfest dynamics versus partial Wigner transform dynamics and (ii) by introducing a frozen Gaussian interpretation of Ehrenfest dynamics which allows nuclear wave packets to separate. The first consequence of this work is a means to rigorously check the accuracy of standard Ehrenfest dynamics. Second, this paper suggests a nonadiabatic dynamics algorithm, whereby the nuclei are propagated on the mean-field (Ehrenfest) potential energy surface and undergo stochastic decoherence events. Our work resembles the surface-hopping algorithm of Schwartz and co-workers [J. Chem. Phys. 123, 234106 (2005)]--only now without any adjustable parameters. For the case of two electronic states, we present numerical results on the so-called "Tully problems" and emphasize that future numerical benchmarking is still needed. Future work will also treat the problem of three or more electronic states.
我们提出了一种新的混合量子经典动力学算法,通过定义自然的退相干速率来帮助弥合平均场( Ehrenfest )和表面跳跃动力学之间的差距。为了得出这个退相干结果,我们扩展了通常的 Ehrenfest 例程中的独立变量数量,以便混合量子经典导数现在与通常的 Ehrenfest 变量一起随时间传播。这样做之后,我们使用两种独立的方法计算了独特的退相干速率:(i)通过根据 Ehrenfest 动力学与部分维格纳变换动力学比较相空间中核电子概率密度的运动方程,以及(ii)通过引入 Ehrenfest 动力学的冻结高斯解释,允许核波包分离。这项工作的第一个结果是一种严格检查标准 Ehrenfest 动力学准确性的方法。其次,本文提出了一种非绝热动力学算法,其中原子核在平均场( Ehrenfest )势能面上传播,并经历随机退相干事件。我们的工作类似于 Schwartz 及其同事的表面跳跃算法[J. Chem. Phys. 123, 234106 (2005)] - 只是现在没有任何可调参数。对于两个电子态的情况,我们给出了所谓的“Tully 问题”的数值结果,并强调未来仍需要进行数值基准测试。未来的工作还将处理三个或更多电子态的问题。