Department of Chemistry and Supercomputing Institute, 207 Pleasant Street S.E., University of Minnesota, Minneapolis, Minnesota 55455-0431.
J Chem Theory Comput. 2005 Jul;1(4):527-40. doi: 10.1021/ct050021p.
Electronic energy flow in an isolated molecular system involves coupling between the electronic and nuclear subsystems, and the coupled system evolves to a statistical mixture of pure states. In semiclassical theories, nuclear motion is treated using classical mechanics, and electronic motion is treated as an open quantal system coupled to a "bath" of nuclear coordinates. We have previously shown how this can be simulated by a time-dependent Schrödinger equation with coherent switching and decay of mixing, where the decay of mixing terms model the dissipative effect of the environment on the electronic subdynamics (i.e., on the reduced dynamics of the electronic subsystem). In the present paper we reformulate the problem as a Liouville-von Neumann equation of motion (i.e., we propagate the reduced density matrix of the electronic subsystem), and we introduce the assumption of first-order linear decay. We specifically examine the cases of equal relaxation times for both longitudinal (i.e., population) decay and transverse decay (i.e., dephasing) and of longitudinal relaxation only, yielding the linear decay of mixing (LDM) and the population-driven decay of mixing (PDDM) schemes, respectively. Because we do not generally know the basis in which coherence decays, that is, the pointer basis, we judge the semiclassical methods in part by their ability to give good results in both the adiabatic and diabatic bases. The accuracy in the prediction of physical observables is shown to be robust not only with respect to basis but also with respect to the way in which demixing is incorporated into the master equation for the density matrix. The success of the PDDM scheme is particularly interesting because it incorporates the least amount of decoherence (i.e., the PDDM scheme is the most similar of the methods discussed to the fully coherent semiclassical Ehrenfest method). For both the new and previous decay of mixing schemes, four kinds of decoherent state switching algorithms are analyzed and compared to one another: natural switching (NS), self-consistent switching (SCS), coherent switching (CS), and globally coherent switching (GCS). The CS formulations are examples of a non-Markovian method, in which the system retains some memory of its history, whereas the GCS, SCS, and NS schemes are Markovian (time local). These methods are tested against accurate quantum mechanical results using 17 multidimensional atom-diatom test cases. The test cases include avoided crossings, conical interactions, and systems with noncrossing diabatic potential energy surfaces. The CS switching algorithm, in which the state populations are controlled by a coherent stochastic algorithm for each complete passage through a strong interaction region, but successive strong-interaction regions are not mutually coherent, is shown to be the most accurate of the switching algorithms tested for the LDM and PDDM methods as well as for the previous decay of mixing methods, which are reformulated here as Liouville-von Neumann equations with nonlinear decay of mixing (NLDM). We also demonstrate that one variant of the PDDM method with CS performs almost equally well in the adiabatic and diabatic representations, which is a difficult objective for semiclassical methods. Thus decay of mixing methods provides powerful mixed quantum-classical methods for modeling non-Born-Oppenheimer polyatomic dynamics including photochemistry, charge-transfer, and other electronically nonadiabatic processes.
在孤立分子体系中,电子能量流涉及电子和核子体系之间的耦合,并且耦合体系演化为纯态的统计混合。在半经典理论中,核运动采用经典力学处理,而电子运动则被视为与“浴”(核坐标的集合)耦合的开放量子系统。我们之前已经展示了如何通过具有相干切换和混合衰减的时变薛定谔方程来模拟这一点,其中混合衰减项模型了环境对电子子动力学(即电子子体系的约化动力学)的耗散效应。在本文中,我们将问题重新表述为刘维尔-冯诺依曼运动方程(即,我们传播电子子体系的约化密度矩阵),并引入一阶线性衰减的假设。我们特别考察了纵向(即种群)衰减和横向衰减(即退相)的弛豫时间相等以及仅存在纵向弛豫的情况,分别得到了混合线性衰减(LDM)和混合种群驱动衰减(PDDM)方案。由于我们通常不知道相干衰减的基,即指针基,因此我们部分通过它们在绝热基和非绝热基中都能给出良好结果的能力来判断半经典方法。预测物理可观测量的准确性不仅对基而且对混合矩阵主方程中混合分离的方式具有稳健性。PDDM 方案的成功特别有趣,因为它包含的去相干(即 PDDM 方案是讨论的方法中与完全相干半经典 Ehrenfest 方法最相似的方法)最少。对于新的和以前的混合衰减方案,分析并比较了四种去相干状态切换算法:自然切换(NS)、自洽切换(SCS)、相干切换(CS)和全局相干切换(GCS)。CS 公式是一种非马尔可夫方法的示例,其中系统保留了其历史的一些记忆,而 GCS、SCS 和 NS 方案是马尔可夫的(时间局部的)。使用 17 个多维原子-双原子测试案例,将这些方法与准确的量子力学结果进行了比较。测试案例包括避免交叉、锥形相互作用以及具有非交叉非绝热势能面的系统。CS 切换算法,其中状态种群由用于每个完整穿过强相互作用区域的相干随机算法控制,但连续的强相互作用区域彼此之间没有相干,被证明是用于 LDM 和 PDDM 方法以及重新表述为具有非线性混合衰减(NLDM)的刘维尔-冯诺依曼方程的测试切换算法中最准确的一种。我们还证明了 CS 与 PDDM 方法的一种变体在绝热和非绝热表示中表现几乎一样好,这对半经典方法来说是一个困难的目标。因此,混合衰减方法为包括光化学、电荷转移和其他非绝热电子过程在内的非 Born-Oppenheimer 多原子动力学提供了强大的混合量子经典方法。