Generalized Lanczos method for systematic optimization of tensor network states<xref rid="cpb_27_7_070501fn1" ref-type="fn">*</xref><fn id="cpb_27_7_070501fn1"> <label>*</label><p>Project supported by the National Natural Science Foundation of China (Grant Nos. 11190024 and 11474331).</p> </fn>

Generalized Lanczos method for systematic optimization of tensor network states* *

Project supported by the National Natural Science Foundation of China (Grant Nos. 11190024 and 11474331).

Huang Rui-Zhen^{1, 2}, Liao Hai-Jun¹, Liu Zhi-Yuan^{2, 3}, Xie Hai-Dong^{1, 2}, Xie Zhi-Yuan⁴, Zhao Hui-Hai^{5, 6, 7}, Chen Jing^{1, 2}, Xiang Tao^{1, 2, 8, †}

(color online) Ground-state energy per site for the two-dimensional antiferromagnetic Heisenberg model on the L × L lattice obtained using the PEPS-based Lanczos method: (a) L = 6, (b) L = 8, (c) L = 10, (d) L = 12, (e) L = 14, (f) L = 16. The boundary MPS method is used to contract the PEPS. For the PEPS with bond dimension D = 2 and 3, the number of states retained in the boundary MPS calculation is 50 and 80, respectively. The results obtained by the quantum Monte Carlo (QMC) simulation are also shown for comparison.^[16]