For spin rotational symmetric models with a positive-definite high-temperature expansion of the partition function, a stochastic sampling of the series expansion upon partial resummation becomes logically equivalent to sampling an uncoloured closely-packed loop-gas model in one higher dimension. Based on this, we devise quantum Monte Carlo updates that importance-sample loop configurations for general $SU(N)$ in fundamental and higher-symmetric representations. The algorithmic performance systematically improves with increase in (continuous) $N$ allowing efficient simulation of quantum paramagnets. The underlying reason for the increased efficacy is the correspondence of quantum paramagnetic phases like valence bond solids to short-loop phases on the loop-gas side rather than the particular value of $N$. This also gives a connection between Sandvik's $JQ$ model class and classical loop-gas models in the deconfined universality class.
Efficiency of Continuous-Time Quantum Monte Carlo Updates in the Ferromagnetic State of the Doped SU(3) Fermi-Hubbard Model
Institute of Science Tokyo(东京科学大学);Osaka Metropolitan University(大阪公立大学);Nambu Yoichiro Institute of Theoretical and Experimental Physics (NITEP)(南部阳一郎理论与实验物理研究所);Chuo University(中央大学)
Accelerating ground-state auxiliary-field quantum Monte Carlo simulations by delayed update and block force-bias update
Institute of Modern Physics, Northwest University(西北大学现代物理研究所);Shaanxi Key Laboratory for Theoretical Physics Frontiers(陕西理论物理前沿重点实验室);Peng Huanwu Center for Fundamental Theory(彭桓武fundamental theory中心);Hefei National Laboratory(合肥国家实验室)
Solving quantum rotor model with different Monte Carlo techniques
系统地测试了量子转子模型$(2+1)$ d XY相变的几种Monte Carlo更新方案的性能。通过比较局域Metropolis(LM)、LM+超松弛(OR)、Wolff簇(WC)、混合Monte Carlo(HM)、混合Monte Carlo + Fourier加速(FA)方案,可以清楚地看出,在五种不同的更新方案中,在量子临界点处,WC和FA方案在达到相同水平的相对误差时获得最小的自相关时间和花费最少的CPU时间,FA还具有易于实现更复杂的交互(例如远程交互)的优点。这些结果赋予一个扩展的量子转子模型,它扮演的铁磁/反铁磁临界玻色子或Z $_2$拓扑顺序的作用,更现实的,但具有挑战性的模型,如费米表面Yukawa耦合到量子转子模型的必要知识。
Neural network quantum state with proximal optimization: a ground-state searching scheme based on variational Monte Carlo
Tsinghua University(清华大学);Nanjing University of Aeronautics and Astronautics(南京航空航天大学)
神经网络量子态(NQS)与变分蒙特卡罗(VMC)方法相结合,是研究量子多体物理的一种很有前途的方法。虽然普通VMC方法对每个样本执行一次梯度更新,但我们引入了一种具有近端优化(PO)的新型目标函数,可以通过重用不匹配的样本来实现多次更新。本文的VMC-PO方法保持了重要抽样梯度优化算法[L. Yang,{ \it et al},Phys. Rev. Research { \bf 2},012039(R)(2020)],其有效地使用采样状态。PO缓解了网络更新期间的数值不稳定性,这类似于随机重构(SR)方法,但实现了一种替代的、更简单的实现,具有更低的计算复杂度。我们研究了我们的VMC-PO算法的性能与一维横向场伊辛模型和二维海森堡反铁磁体在一个正方形晶格的基态搜索,并证明,达到基态能量是最先进的结果相媲美。