论文部分内容阅读
提出了一个固定节面量子Monte Carlo的新算法,与前人的算法相比,其“指导函数”的优化不是在扩散前,而是在扩散过程中同步进行;这不仅在机时上是节省的,更重要的是优化与扩散两者按相同的取样方式进行,达到互相改善的目的;这一优化方案是超线性收敛的,它能加快粒子的扩散;在扩散过程中,“指导函数”的节面不断发生改变,这有利于减小“固定节面误差”.这一新算法已被运用到CH_2的X~3B_1态和a~1A_1态,以及NH_2的π-X~2B_1态和σ-A~2A_1态总能量的计算,由此算出了CH_2单一三重态的“劈开”能△E_(s-T)=(45.542±1.840)kJ/mol和NH_2的σ-π“劈开”能△E_(σ-π)=(141.644±1.589)kJ/mol.计算结果表明这一新算法在精度、统计误差和计算量方面比一般固定节面量子Monte Carlo方法都要优越得多.
A new algorithm of fixed nodal quantum Monte Carlo is proposed. Compared with the previous algorithm, the optimization of “guiding function” is not before diffusion but also during diffusion. This not only saves time and energy , And more importantly, both optimization and diffusion are carried out according to the same sampling method so as to achieve the purpose of improving each other. This optimization solution is superlinearly convergent and can accelerate the diffusion of particles. In the diffusion process, the “guiding function” Of the nodal surface continuously changing, which is helpful to reduce the “fixed nodal error.” This new algorithm has been applied to the X ~ 3B_1 state and the ~ 1A_1 state of CH_2, as well as the π-X ~ 2B_1 state and σ -A ~ 2A_1 state, we calculated the “cleavage energy △ E sT = (45.542 ± 1.840) kJ / mol for CH 2 singlet triplet and σ-π” cleavage "△ E_ (σ-π) = (141.644 ± 1.589) kJ / mol. The calculation results show that this new algorithm is superior to the general fixed-nodal quantum Monte Carlo method in accuracy, statistical error and computational complexity.