Local integrals of motion encoded in a few eigenstates
J. Pawłowski, P. Łydżba, M. Mierzejewski · 2026
Read Local integrals of motion encoded in a few eigenstates by J. Pawłowski, P. Łydżba, M. Mierzejewski (2026) 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-2603-01859.txt.
Introduction
J. Pawlowski, P. Lydzba, and M. Mierzejewski extract conserved quantities of a quantum system from just a few of its states. Their 2026 work focuses on the XXZ model, a standard chain of interacting quantum spins. Local integrals of motion, quantities that never change, define whether such a system counts as integrable.
A single energy state can already reveal if a system is integrable or chaotic, past work showed. The team shows those integrals can also be estimated from a small set of states. As system size grows, the needed share of states shrinks toward zero. A cousin model with fragmented Hilbert space breaks the pattern, since most states are needed there. That split marks a rare deep difference between true integrability and mere fragmentation of the state space.
Catalog record
| CREATOR | J. Pawłowski, P. Łydżba, M. Mierzejewski |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 44625 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.01859
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Local integrals of motion encoded in a few eigenstates J. Pawłowski, P. Łydżba, M. Mierzejewski Abstract Many properties of a quantum system can be obtained from just a single eigenstate of its Hamiltonian. For example, a single eigenstate can be used to determine whether a system is integrable or chaotic and, in the latter case, to establish its thermal properties. Focusing on the XXZ model, we show that the local integrals of motion, which lie at the heart of integrability, can also be estimated from a small number of eigenstates. Moreover, as the system size increases, fewer eigenstates are required, so that in the thermodynamic limit, the integrals of motion can be obtained from a vanishingly small fraction of all eigenstates. Interestingly, this property does not extend to integrals of motion arising solely from Hilbert space fragmentation, as found in the folded XXZ model, where the majority of eigenstates has to be used. This represents one of the few fundamental differences known between integrability and Hilbert space fragmentation. I Introduction Over the past two decades, significant effort has been devoted to understanding the non-equilibrium dynamics of closed quantum systems 63 ; 57 ; 21 ; 48 . In particular, whether and, if yes, why the expectation values of local observables at long times agree with the predictions of standard Gibbs ensembles. This behavior is known as thermalization, while systems that exhibit it are commonly referred to as ergodic. These studies have shown that a great deal of information can be extracted from just a single eigenstate in the thermodynamic limit, or from a small number of eigenstates in a finite system 26 . For example, it can be used to identify ergodic behavior. The bipartite entanglement entropy of a single eigenstate, which quantifies correlations between subsystems, is universal in ergodic systems. Specifically, its leading term matches that of a random state from the Hilbert space 53 and is maximal in the thermodynamic limit 8 ; 76 ; 30 ; 10 ; 38 . The only restriction is that the eigenstate cannot be too close to the edges of the spectrum 12 . For this reason, it has been used in studies of ergodicity-breaking transitions to pinpoint the transition point 6 ; 68 ; 55 ; 73 ; 69 . Moreover, thermalization is currently understood through the Eigenstate Thermalization Hypothesis (ETH), which asserts that individual eigenstates are essentially thermal 71 ; 16 ; 72 . This means that the expectation values of local observables in a single eigenstate agree with the predictions of standard Gibbs ensembles. Consequently, a single eigenstate is not only sufficient to identify ergodicity, but also encodes the equilibrium properties of ergodic systems. The most studied exceptions to thermalization are integrable models. Here, we focus on a paradigmatic example, the XXZ spin chain 78 ; 79 ; 80 . Integrable models support an extensive number of local integrals of motion (LIOMs), which are local operators that commute with the Hamiltonian, ensuring that their expectation values are conserved during time evolution. Integrable models do not satisfy the ETH, so their eigenstates are not thermal 9 ; 3 ; 40 ; 47 . Nevertheless, in this work, we demonstrate that a small number of eigenstates suffices to determine LIOMs and, in principle, to construct the Generalized Gibbs Ensemble 64 ; 59 ; 32 ; 75 ; 39 ; 13 ; 24 . Importantly, this number does not increase with system size, meaning that it represents a vanishing fraction of the Hilbert space dimension. This leads to the surprising conclusion that a small number of eigenstates can provide an accurate description of the stationary state also in an integrable model. The identification of LIOMs also provides the starting point for constructing hydrodynamics or generalized hydrodynamics 7 ; 14 ; 33 ; 19 ; 27 ; 15 ; 2 ; 23 ; 5 ; 28 ; 31 ; 18 . Finally, we verify whether this observation holds in systems exhibiting Hilbert space fragmentation 36 ; 65 ; 51 ; 49 ; 22 ; 11 ; 1 ; 25 . In this ergodicity-breaking phenomenon, the Hamiltonian, when written in a suitable local basis, decomposes into exponentially many disconnected blocks. Each block contains an exponentially small fraction of all eigenstates. This phenomenon is also believed to be distinct from integrability. Although some fragmented systems are known to support LIOMs 84 , it remains unclear whether all do. Furthermore, a seminal work connected Hilbert space fragmentation to a different class of conserved operators, which are not strictly local but behave as local in typical states from the Hilbert space 61 ; 50 . Despite this, there are not many reported differences between integrability and Hilbert space fragmentation. For this reason, we focus on the folded XXZ model, which can be obtained from the XXZ model in the limit of infinite anisotropy 58 ; 82 ; 81 . We demonstrate that, apart from LIOMs stemming from Bethe-ansatz integrability (which vanish when a perturbation breaking Bethe-ansatz integrability is introduced), the model also hosts LIOMs originating from fragmentation (which do not vanish under such perturbations). We observe that the former can still be numerically generated from a small number of eigenstates, while the latter require almost all eigenstates. This leads to the conclusion that the information about LIOMs originating from fragmentation is encoded differently in eigenstates. The paper is organized as follows. In Sec. II , we introduce the numerical method used to estimate LIOMs from an incomplete set of eigenstates. In Sec. III , we demonstrate that this estimation is accurate for a small number of eigenstates in the XXZ spin chain. We consider different symmetry sectors and extend our study to quasilocal integrals of motion. In Sec. IV , we examine the effectiveness of the estimation in the folded XXZ model. Finally, we summarize our findings and conclude in Sec. V . II Finding LIOMs from incomplete spectrum We first discuss the numerical method for estimating LIOMs, denoted hereafter as \hat{\cal Q}^{\alpha} , from an incomplete set of eigenstates of the Hamiltonian, H|n\rangle=E_{n}|n\rangle . Here, the index \alpha labels LIOMs. We focus on Hermitian operators, for which we use the Hilbert-Schmidt (HS) product \langle AB\rangle={\rm Tr}(AB)/Z and the HS norm \lVert A\lVert^{2}=\langle AA\rangle , where the trace is calculated over the entire Hilbert space with dimension Z . We require that ||A||^{2}=1 . For simplicity, we first consider a nondegenerate spectrum. This assumption is not essential, and minor modifications allow to study systems with degenerate eigenvalues. This is discussed in sections III and IV . Note that, in a case of nondegenerate spectrum, conserved quantities have nonvanishing matrix elements only along the diagonal, i.e., \langle m|{\cal Q}^{\alpha}|n\rangle\neq 0 only for m=n . Therefore, in order to confirm that \mathcal{Q}_{\alpha} is conserved, one needs to calculate at least all diagonal matrix elements and verify that 1/Z\sum_{n}\langle n|{\cal Q}^{\alpha}|n\rangle^{2}=\lVert{\cal Q}^{\alpha}\lVert^{2} . This task requires full diagonalization of the Hamiltonian, posing a bottleneck for numerical studies. In order to estimate LIOMs from selected eigenstates, it is necessary to implement a method that does not rely on the HS product or HS norm (unlike the one employed in 45 ; 46 ; 44 ). An appropriate approach has recently been