量子多体系统张量网络态的交换分子动力学优化
Replica exchange molecular dynamics optimization of tensor network states for quantum many-body systems
摘要:
张量网络态(TNS)方法结合蒙特卡罗(MC)技术已被证明是模拟量子多体系统的一种强有力的算法。然而,由于基态能量是张量的高度非线性函数,因此在优化模拟物理系统的TNS时容易陷入局部极小值。为了克服这个困难,我们引入了一个复制交换分子动力学优化算法来获得TNS的基态,基于MC采样技术,通过映射的能量函数的TNS的经典动力学系统。该方法有望有效地避免局部极小值。我们对基于矩阵乘积态(MPS)的一维哈伯德模型和基于弦键态(SBS)的正方格子Heisenberg $J_1$ - $J_2$模型进行了基准测试。结果表明,与已有的结果相比,该优化方法是鲁棒的、有效的.
英文摘要:
The tensor network states (TNS) methods combined with Monte Carlo (MC) techniques have been proved a powerful algorithm for simulating quantum many-body systems. However, because the ground state energy is a highly non-linear function of the tensors, it is easy to get stuck in local minima when optimizing the TNS of the simulated physical systems. To overcome this difficulty, we introduce a replica-exchange molecular dynamics optimization algorithm to obtain the TNS ground state, based on the MC sampling techniques, by mapping the energy function of the TNS to that of a classical dynamical system. The method is expected to effectively avoid local minima. We make benchmark tests on a 1D Hubbard model based on matrix product states (MPS) and a Heisenberg $J_1$-$J_2$ model on square lattice based on string bond states (SBS). The results show that the optimization method is robust and efficient compared to the existing results.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。