Statistics-encoded tensor network approach in disordered quantum many-body spin chains
Hao Zhu, Ding-Zu Wang, Shi-Ju Ran, Guo-Feng Zhang · 2025
Read Statistics-encoded tensor network approach in disordered quantum many-body spin chains by Hao Zhu, Ding-Zu Wang, Shi-Ju Ran, Guo-Feng Zhang (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-16286.txt.
Introduction
Hao Zhu, Ding-Zu Wang, Shi-Ju Ran, and Guo-Feng Zhang fold disorder into an extra layer so quantum spin chains become tractable. Simulating many-body dynamics with random disorder is a standing challenge. Their statistics-encoded tensor network stores the disorder distribution in an auxiliary layer. Averaging then happens separately, restoring translational invariance and a clean transfer matrix.
A universal criterion ties the needed discretization to disorder strength and evolution time. Encoding pays off most in the weak-disorder, typically chaotic regime. Applied to the disordered transverse-field Ising chain, the method shows the spectral form factor is ruled by the leading transfer-matrix eigenvalue. That contrasts with the kicked Ising model's behavior. The framework opens disorder-driven dynamics to tensor network tools. It gives chaos in dirty systems a new computational handle.
Catalog record
| CREATOR | Hao Zhu, Ding-Zu Wang, Shi-Ju Ran, Guo-Feng Zhang |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 55 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-16286
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Statistics-encoded tensor network approach in disordered quantum many-body spin chains Hao Zhu, Ding-Zu Wang, Shi-Ju Ran, Guo-Feng Zhang Abstract Simulating the dynamics of quantum many-body systems with disorder is a fundamental challenge. In this work, we propose a general approach—the statistics-encoded tensor network (SeTN)—to study such systems. By encoding disorder into an auxiliary layer and averaging separately, SeTN restores translational invariance, enabling a well-defined transfer matrix formulation. We derive a universal criterion, n ≫ α 2 t 2 n\gg\alpha^{2}t^{2} , linking discretization n n , disorder strength α \alpha , and evolution duration t t . This sets the resolution required for faithful disorder averaging and shows that encoding is most efficient in the weak-disorder, typically chaotic regime. Applied to the disordered transverse-field Ising model, SeTN shows that the spectral form factor is governed by the leading transfer-matrix eigenvalue, in contrast to the kicked Ising model. SeTN thus provides a novel framework for probing the disorder-driven dynamical phenomena in many-body quantum systems. Introduction .—Simulating quantum many-body systems with disorder is crucial for understanding a wide range of phenomena, including quantum chaos Gutzwiller 1990 ; D’Alessio et al. 2016 ; Haake et al. 2018 , many-body localization (MBL) Nandkishore and Huse 2015 ; Abanin et al. 2019 , quantum spin glass Bray and Moore 1980 ; Sachdev and Ye 1993 , and criticality Fisher 1995 ; Pich et al. 1998 . Recently, the chaotic dynamics of disordered systems predicted by random matrix theory (RMT) has attracted growing interest, driven by advances in experimental techniques Li et al. 2024 ; Dong et al. 2025 . Analytical approaches to the disorder-averaged dynamics typically map them onto classical statistical models Chan et al. 2018 ; Bertini et al. 2018 ; Von Keyserlingk et al. 2018 ; Zhou and Nahum 2019 . This mapping relies on Haar averaging Von Keyserlingk et al. 2018 ; Zhou and Nahum 2019 or self-duality Chan et al. 2018 ; Bertini et al. 2018 techniques and is primarily applicable to random circuits and Floquet systems. However, it cannot be directly applied to generic time-independent disordered Hamiltonians that usually lack self-duality. Numerical methods for studying disordered dynamics mainly include exact diagonalization (ED) and tensor network (TN) techniques. ED is limited to modest system sizes, e.g., L\lesssim 24 for the disordered XXZ chain Pietracaprina et al. 2018 ; Sierant et al. 2020a . While TN methods can handle larger systems, they are mostly limited to the MBL regime Doggen et al. 2018 ; Chanda et al. 2020 , as extending them to the ergodic phase remains challenging due to volume-law entanglement growth. In this letter, we propose the statistics-encoded tensor network (SeTN), a general approach for studying disordered many-body systems with time-independent Hamiltonians. In SeTN, we encode disorder into an auxiliary layer and perform averaging separately (see Fig. 1 (b)). By analyzing the decay rate of the singular values, we derive a criterion for the temporal resolution required for efficient disorder encoding: n\gg\alpha^{2}t^{2} , where n is the number of time steps, \alpha the disorder strength, and t the evolution duration. Our approach offers two key advantages: encoding is more efficient for weak disorder, a regime challenging for conventional tensor networks, and it restores spatial translation invariance, enabling a well-defined transfer matrix that allows studying dynamics under the thermodynamic limit and naturally integrates with Krylov iteration Stewart 2002 and density matrix renormalization group (DMRG) methods White 1992 ; White 1993 . As a demonstration, we investigate the spectral form factor (SFF) of the disordered transverse-field Ising model in its chaotic regime, revealing a prolonged pre-RMT transient dominated by a single, non-degenerate leading eigenvalue of the transfer matrix. This pathway to RMT behavior contrasts with Floquet models such as the kicked Ising model, where chaotic spectral statistics and eigenvalue degeneracies appear from the outset Chan et al. 2018 . Statistics-encoded tensor network approach .—We consider a quantum many-body system with local disorder, described by the Hamiltonian \mathcal{H}[\boldsymbol{h}]=H+\sum_{i=1}^{L}h_{i}H_{i} , where H is the disorder-free part, \{H_{i}\} are local operators acting on site i , \boldsymbol{h}=\{h_{i}\} is a disorder realization, and L is the system size. Throughout this work we set \hbar=1 . To simulate the real-time dynamics, we apply a first-order Trotter decomposition with a small time step \tau\ll 1 , which gives the evolution operator U(\tau)=\mathrm{e}^{-\mathrm{i}\tau H}\mathrm{e}^{-\mathrm{i}\tau\sum_{i}h_{i}H_{i}}+O(\tau^{2}) . This decomposition enables a TN representation of the full evolution operator (see Fig. 1 (a)). For simplicity, we focus on the widely studied case where H=\sum_{i}H_{i,i+1} includes only nearest-neighbor interactions and H_{i}=\sigma^{z}_{i} is the Pauli- z operator on site i . Generalizations to longer-range interactions or different disorder couplings are straightforward sup . To proceed, as shown in Fig. 1 (a), we first decompose the two-body unitary gates into two local tensors, \displaystyle\langle j_{n}^{[i]}j_{n+1}^{[i]}|\mathrm{e}^{-\mathrm{i}\tau H_{i,i+1}}| \displaystyle j_{n}^{[i+1]}j_{n+1}^{[i+1]}\rangle=V^{(L)}_{j_{n}^{[i]}j_{n+1}^{[i]},k_{n}^{[i]}}V^{(R)}_{k_{n}^{[i]},j_{n}^{[i+1]}j_{n+1}^{[i+1]}}, (1) where j_{n}^{[i]} refers to site i and time step n , with repeated indices summed over implicitly. This can be achieved via singular value decomposition (SVD) or spectral decomposition. At the same time, we factor out the diagonal disorder term using a delta tensor: \langle j_{n}^{[i]}|\mathrm{e}^{-\mathrm{i}\tau\sigma^{z}_{i}}|j_{n+1}^{[i]}\rangle=v_{l_{n}^{[i]}}\delta_{l_{n}^{[i]}j_{n}^{[i]}j_{n+1}^{[i]}} , where \delta_{ijk} is the three-order Kronecker tensor (equal to 1 when i=j=k and 0 otherwise), and \boldsymbol{v}=[\mathrm{e}^{-\mathrm{i}\tau h_{i}},\mathrm{e}^{\mathrm{i}\tau h_{i}}] encodes the effect of local disorder. Throughout this work, we use the indices j,k and l to denote the time , space , and layer directions, respectively, as indicated by the coordinate system in Fig. 1 . By combining these decompositions, we define a local tensor W as W_{j_{n}^{[i]}j_{n+1}^{[i]},k_{n}^{[i-1]}k_{n}^{[i]},l_{n}^{[i]}}=\delta_{l_{n}^{[i]}j_{n}^{[i]}p_{n}^{[i]}}V^{(R)}_{k_{n}^{[i-1]},p_{n}^{[i]}q_{n}^{[i]}}V^{(L)}_{q_{n}^{[i]}j_{n+1}^{[i]},k_{n}^{[i]}}, (2) where p and q are auxiliary indices used for internal contractions. This construction is illustrated in the third column of Fig. 1 (a), where the spatial slice at site i corresponds to \prod_{n}v_{l_{n}^{[i]}}W_{j_{n}^{[i]}j_{n+1}^{[i]},k_{n}^{[i-1]}k_{n}^{[i]},l_{n}^{[i]}} . Figure 1: (Color online) (a) Tensor network (TN) representation of the Trotterized evolvation operator. The tensor encoding the disorder is decomposed into a vector (blue circle) and a Kronecker delta tensor (black point). (b) A general two-layer TN representation of the disorder-averaged transfer matrix \mathrm{T} . The orange-shaded region indicates the disorder average. (c) Diagrammatic illustration of the SeTN compression. S^{(n)} is the singular values obtained via corresponding singular value decomposition. We study disorder-averaged dynamics, where many observables such as expectation values, the Rényi entropy, and the SFF admit a two-layer tensor-network representation, obtained by contracting the forward and backward evolutions U and U^{\dagger} (dark/light blue in Fig. 1 (b)). The effect of disorder is encoded in a statistics layer (orange region), which after averaging renders all spatial sites equivalent. This naturally motivates the definition of a transfer matrix \mathrm{T} that captures the combined action of unitary dynamics and disorder averaging across a single spatial slice, \displaystyle\mathrm{T}_{k_{1}^{\prime[i-1]},...,k_{n}^{\prime[i-1]},k_{1}^{[i-1]},...,k_{n}^{[i-1]};k_{1}^{\prime[i]},...,k_{n}^{\prime[i]},k_{1}^{[i]},...,k_{n}^{[i]}} (3) \displaystyle=\Big\langle\prod_{p=1}^{n}W^{*}_{j_{p}^{\prime[i]}j_{p+1}^{\prime[i]},k_{p}^{\prime[i-1]}k_{p}^{\prime[i]},l_{p}^{\prime[i]}}v^{*}_{l_{p}^{\prime[i]}}v_{l_{p}^{[i]}}W_{j_{p}^{[i]}j_{p+1}^{[i]},k_{p}^{[i-1]}k_{p}^{[i]},l_{p}^{[i]}}\Big\rangle \displaystyle=\mathrm{O}_{l^{\prime}_{1},...,l^{\prime}_{n};l_{1},...,l_{n}}\prod W^{*}W, where the indices on the conjugated tensors are distinguished by primes and \langle{\cdot}\rangle denotes the disorder average. In the last line, We separate the disorder-averaged tensor \mathrm{O} from the unitary layers (omitting site superscripts and tensor subscripts for clarity), which can be obtained explicitly by integrating over the distribution of h : \displaystyle\mathrm{O}_{l^{\prime}_{1},...,l^{\prime}_{n};l_{1},...,l_{n}}\equiv\Big\langle{\prod_{p=1}^{n}v_{l^{\prime}_{p}}^{*}v_{l_{p}}}\Big\rangle=\int\mathrm{d}h\;\mathrm{e}^{\mathrm{i}h\tau\sum_{p}(l^{\prime}_{p}-l_{p})}P(h) (4) \displaystyle=\begin{cases}\operatorname{sinc}\left[\tau\alpha\sum_{p}(l^{\prime}_{p}-l_{p})\right],&h\sim\mathcal{U}(-\alpha,\alpha),\\[4.0pt] \exp\left[-\frac{1}{2}\sigma^{2}\tau^{2}\left(\sum_{p}(l^{\prime}_{p}-l_{p})\right)^{2}\right],&h\sim\mathcal{N}(0,\sigma^{2}).\end{cases} where P(h) denotes the probability density function of the disorder variable h . We give explicit formulas for both uniform \mathcal{U}(-\alpha,\alpha) and Gaussian \mathcal{N}(0,\sigma^{2}) disorder. While a similar expression for the Gaussian case was derived for KIM Bertini et