arXiv:2609.30101
Leg-Tied Tensor Network States: Entanglement Beyond Virtual Bonds
Shuoyi Hu, Bing Gu
Abstract
We introduce a class of tensor-network states in which physical legs are shared among local tensors, termed leg-tied tensor ansätze (LETTA). Physical leg ties encode long-range correlations directly, while a virtual matrix product state (MPS) backbone retains short-range multipartite entanglement. The linear virtual backbone allows us to develop a deterministic density matrix renormalization group-like variational optimization algorithm using exact contractions over the active tie-boundary sets and local minimization. We demonstrate the advantages of LETTA for the two-dimensional frustrated $J_1$--$J_2$ Heisenberg model and the three-dimensional transverse-field Ising model. Our results show that LETTA is substantially more accurate than same-bond-dimension MPS calculations and can typically reach the accuracy of much larger MPS calculations using one order of magnitude fewer variational parameters. LETTA thus opens the door for explicitly correlated tensor-network states that can encode long-range correlation beyond virtual bonds.
AI Summary
The paper asks whether tensor-network states can encode correlations between sites far apart in an MPS ordering without increasing the virtual bond dimension. It introduces leg-tied tensor ansätze (LETTA), whose local tensors depend on their own physical variable and selected future-site variables, while retaining a linear virtual backbone. Ordinary MPS form a submanifold; at virtual dimension $D=1$, LETTA reduces to a correlator-product state rather than a product state. Across a chain cut after site $m$, let $\mathcal F_m$ contain the physical sites to the right tied to tensors on the left. A boundary factorization establishes the Schmidt-rank bound $r\leq D_m\prod_{j\in\mathcal F_m}d_j$, also capped by subsystem dimensions. For uniform local dimension $d$, this gives $S\leq\log D_m+|\mathcal F_m|\log d$. Optimization uses deterministic one-site sweeps, exact environment contractions, and generalized eigenvalue problems with norm-matrix whitening; conditional canonical gauges and parity-preserving updates are also developed. Numerical benchmarks cover the $6\times6$ frustrated $J_1$–$J_2$ Heisenberg model and open $N_x\times3\times3$ transverse-field Ising clusters with $N_x=3,6,9$. LETTA at $D=4$ achieves lower variational energies than MPS at $D=32$, using approximately 7% and 10% of their stored parameters, respectively. Ising entanglement spectra show additional Schmidt levels beyond the MPS rank limit. Exact contraction remains exponential in the maximum tie-boundary width, with Hamiltonian environments scaling as $O(D_i^2\chi_i d^{2|\mathcal F_i|})$. Parameter savings therefore do not establish computational-time savings. The general tie patterns require nontrivial norm metrics, and more robust optimization is identified as future work.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 14 pages, 5 figures, including Supplemental Material