Emergence of negative viscosities and colored noise under current-driven Ehrenfest molecular dynamics

arxiv(2023)

引用 3|浏览0
暂无评分
摘要
Molecules in molecular junctions are subject to current-induced forces that can break chemical bonds, induce reactions, destabilize molecular geometry, and halt the operation of the junction. Theories behind current-driven molecular dynamics simulations rely on a perturbative time-scale separation within the system with subsequent use of nonequilibrium Green's functions (NEGF) to compute conservative, non-conservative, and stochastic forces exerted by electrons on nuclear degrees of freedom. We analyze the effectiveness of this approximation, paying particular attention to the phenomenon of negative viscosities. The perturbative approximation is directly compared to the nonequilibrium Ehrenfest approach. We introduce a novel time-stepping approach to calculate the forces present in the Ehrenfest method via exact integration of the equations of motion for the nonequilibrium Green's functions, which does not necessitate a time-scale separation within the system and provides an exact description for the corresponding classical dynamics. We observe that negative viscosities are not artifacts of a perturbative treatment but also emerge in Ehrenfest dynamics. However, the effects of negative viscosity have the possibility of being overwhelmed by the predominantly positive dissipation due to the higher-order forces unaccounted for by the perturbative approach. Additionally, we assess the validity of the white-noise approximation for the stochastic forces, finding that it is justifiable in the presence of a clear time-scale separation and is more applicable when the current-carrying molecular orbital is moved outside of the voltage window. Finally, we demonstrate the method for molecular junction models consisting of one and two classical degrees of freedom.
更多
查看译文
关键词
molecular dynamics,negative viscosities,colored noise,ehrenfest,current-driven
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要