REVIEW 4 major objections 4 minor 33 references
Single-eigenstate test of eigenstate thermalization hypothesis via perturbed eigenstate quench
T0 review · 4 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash
Pith's one-line read A weak perturbation of a single energy eigenstate reveals whether it thermalizes: ETH gives an S-shaped subsystem-speed curve with an inflection near half the system, while ETH violation gives a J-shaped curve.
desk verdict A promising single-eigenstate ETH probe, but the S/J shape is only visually classified and may be a Page-curve artifact of the Haar perturbation; needs a random-state control before claims of robustness. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The central object is the perturbed eigenstate quench: take an energy eigenstate $|\psi_i\rangle$, mix in a small Haar-random component with strength $\epsilon = 10^{-3}$ to make it non-stationary, evolve under the original Hamiltonian, and measure the time-averaged subsystem evolution speed $v_A(t)$, defined as the trace-distance rate of change of the subsystem's reduced density matrix and rescaled by the full-system speed. The diagnostic signature is the curvature shape of the resulting curve as a function of $x = \ell/L$: an S-shaped profile with an inflection near $x = 1/2$ for ETH-obeying eigenstates versus a J-shaped, everywhere-convex profile for ETH-violating ones. The mechanism behind the inflection is that small thermalizing subsystems respond slowly and robustly to the perturbation, while large non-thermalizing subsystems respond quickly, producing the curvature change at the half-system mark.
What would settle it
Observe an ETH-satisfying eigenstate of a chaotic spin chain at larger sizes (for example, $L \geq 20$ via tensor-network time evolution) and compute the same rescaled speed curve: if the inflection point moves away from $x = 1/2$ or the S-shape disappears, the claimed signature does not survive the thermodynamic limit.
Extended reading notes
Core claim
The central claim is that the rescaled time-averaged subsystem evolution speed after a perturbed eigenstate quench is a reliable single-eigenstate ETH diagnostic. For an eigenstate that satisfies ETH, the curve of this speed versus subsystem-to-total size ratio $x$ is S-shaped: convex for $x < 1/2$, concave for $x > 1/2$, with an inflection point near half the system size. For an eigenstate that violates ETH, the curve is convex throughout the entire range $0 < x < 1$, giving it a J shape with no inflection. The paper verifies this qualitative distinction in exact-diagonalization studies of chaotic and integrable transverse-field Ising chains, a disordered XXZ chain in both chaotic and many-body-localized phases, and a spin-1 XX chain containing quantum many-body scar states, where the criterion correctly labels which eigenstates thermalize.
Load-bearing premise
The diagnostic assumes that an eigenstate's thermal versus non-thermal character directly controls how a weak random perturbation moves its subsystems: small thermalizing subsystems evolve slowly and large non-thermalizing subsystems evolve quickly, so the speed curve changes curvature at $x = 1/2$.
Editorial extensions
If this is right
- Provides a single-eigenstate ETH test that avoids constructing thermal ensembles or generalized Gibbs ensembles, removing the main computational bottleneck of trace-distance diagnostics.
- Correctly labels thermalizing versus non-thermalizing eigenstates across chaotic, integrable, MBL, and quantum-many-body-scar regimes, so it can serve as a general ergodicity-breaking detector.
- The protocol requires only time evolution from a slightly perturbed eigenstate and measurement of subsystem dynamics, making it feasible on cold-atom and superconducting-circuit quantum simulators.
- The same scheme can be extended to Floquet systems and Hilbert-space-fragmented models, where conventional ETH diagnostics are difficult or ill-defined.
Reading between the lines
- The location and sharpness of the inflection point could be read as a continuous measure of how strongly an eigenstate adheres to ETH, turning the binary S/J classifier into a graded probe.
- If the random perturbation is replaced by a physically motivated local perturbation, the shape of the speed curve may encode the local relaxation rate or transport timescale, a connection the paper does not explore.
- The diagnostic could in principle be applied to a single eigenstate of a larger system without full diagonalization, using variational or DMRG-targeted eigenstates plus quench dynamics, which would extend it beyond $L \approx 14$.
- A natural experimental test is to prepare a scar state in the spin-1 XX chain and verify its J-shaped curve on a quantum simulator, confirming the mechanism in a setting independent of exact diagonalization.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The manuscript proposes a single-eigenstate ETH diagnostic based on a perturbed eigenstate quench. It constructs an initial state |ψ(0)⟩ ∝ |ψ_i⟩ + ε|ψ_random⟩, evolves it with the Hamiltonian, and computes the time-averaged subsystem evolution speed ⟨v_A⟩/v_whole as a function of subsystem fraction x = ℓ/L. The central claim is that ETH-satisfying eigenstates produce an S-shaped curve with an inflection point near x = 1/2, whereas ETH-violating eigenstates produce a convex J-shaped curve. The criterion is demonstrated on chaotic and integrable Ising chains, a disordered XXZ chain in chaotic and MBL regimes, and a spin-1 XX chain with quantum many-body scars. The paper concludes that the diagnostic is a robust, visually unambiguous, and experimentally feasible single-eigenstate probe that avoids constructing thermal ensembles.
Significance. If the claimed qualitative distinction were established, the protocol would be a useful single-eigenstate ETH probe, since it avoids explicit thermal-state construction and uses a simple, potentially measurable quantity. The paper's strengths are its breadth of benchmark models (chaotic, integrable, MBL, and scarred systems), the use of exact diagonalization, and the clear presentation of the protocol. However, the current evidence is anecdotal: each classification rests on visual inspection of single curves with no error bars, no ensemble averaging, no quantitative inflection-point criterion, and no finite-size scaling. Most importantly, no random-state control is provided, leaving open the possibility that the S-shape is a generic property of the Haar-random perturbation rather than a fingerprint of the target eigenstate. The central claim is therefore not yet supported at the level required by the abstract.
major comments (4)
- [§II, Eqs. (1)–(5)] The heuristic in Section II does not connect ETH to the curve shape, and a leading-order calculation shows why this matters. Writing |ψ(0)⟩ = N^{-1}(|ψ_i⟩ + ε|ψ_random⟩), evolving with H, and using Eq. (4), one finds v_A(t) = (1/2)||dρ_A/dt||_1 = (ε/2)|| tr_B([H, |ψ_i⟩⟨r(t)| + |r(t)⟩⟨ψ_i|]) ||_1 + O(ε²), where |r(t)⟩ = e^{-iHt}|ψ_random⟩. The observable is therefore a cross-correlation between the target eigenstate and the time-evolved random component, not a property of the unperturbed reduced density matrix ρ_A^{ii} that the 'small thermalizing subsystems are more robust' argument refers to. A control in which |ψ_i⟩ is replaced by an independent Haar-random state (or omitted) must be shown to give a different curve shape; otherwise the S-shape can be attributed to the random component's generic entanglement structure (a Page-curve artifact).
- [§III–V, Figs. 1–3] Every conclusion in the paper is based on single curves: no error bars are shown, no averaging over Haar-random perturbations is reported, and Fig. 2 appears to use a single disorder realization. The number N of time points in Eq. (5) is never stated. To support the adjective 'robust,' the authors should report the mean and standard deviation of ⟨v_A⟩/v_whole over many random perturbations (and disorder samples), or at minimum show several representative realizations so the reader can assess the variability.
- [§III, §VI] The distinction between S-shaped and J-shaped curves is never made quantitative. The paper does not define an operational rule for 'inflection point near x = 1/2'; it is judged by eye. A testable criterion is needed, for example the sign of the discrete second difference of ⟨v_A⟩/v_whole at x = 1/2, together with its L-dependence. The Discussion acknowledges the absence of systematic finite-size scaling, but since the claim is about a structural transition at half system size in the thermodynamic limit, that absence is a load-bearing gap rather than a routine caveat.
- [§III–V] The benchmark samples are very small: two eigenstates per model in Fig. 1, one per regime in Fig. 2, and two states total in Fig. 3. A single-eigenstate ETH diagnostic should be validated on a larger set of eigenstates, including several bulk states and edge states, with the classification criterion applied uniformly; otherwise 'full agreement with established phenomenology' is an overstatement.
minor comments (4)
- [Abstract, Fig. 2 caption, affiliations] There are several typographical errors: 'spinchainscovering' in the abstract, 'eigenstates quench' in the Fig. 2 caption, and 'tju.eud.cn' in the author affiliation.
- [Eq. (7)] The prefactor before the Heisenberg term appears as 'h 1/4', which is likely a rendering error; please check the intended coefficient.
- [§III] Please specify the precise definition of the '1/4–spectrum state' when symmetries or conserved quantum numbers split the spectrum into sectors, in particular for the disordered XXZ chain.
- [§II] The sentence claiming that the trace distance is 'monotonically increasing with the dimension of the subsystem' is not a standard result in the cited reference; please provide a proof or a more precise citation.
Circularity Check
No circular derivation: the S/J criterion is empirically benchmarked against independent ETH/non-ETH cases, not derived from fitted parameters or self-citation.
full rationale
The paper's derivation chain starts from a kinematic definition of subsystem evolution speed (Eq. 4) and time-averaging (Eq. 5), with the perturbation strength and time window fixed rather than fitted to the target labels. The S-versus-J claim is presented as a heuristic hypothesis based on the known subsystem-ETH entanglement structure from an external reference [16], and it is then tested on eigenstates whose ETH status is fixed by independent means: integrable versus chaotic Ising chains, disordered XXZ phases, and exact bimagnon scar towers. The diagnostic is not used to assign the labels that define the criterion; instead it is compared with pre-existing labels, so there is no fitted-input-called-prediction or self-definitional reduction. Same-author citations [18,19] appear only for background and for the definition of v_A, and the paper explicitly notes that [19] cannot test single eigenstates, so these citations are not load-bearing. The skeptic's concern that the S-shape may be a generic artifact of the Haar-random perturbation is a control/validity issue rather than a circularity claim: lacking a random-state ensemble average does not make the observable equal to its input by construction. The Discussion honestly acknowledges finite-size and one-dimensional limitations, which are external-validity limitations, not circularity. Overall, no significant circularity is found.
Assumptions & free parameters
free parameters (3)
- perturbation strength epsilon =
0.001
- time step delta_t =
1e-6
- time averaging window =
t in [2L, 4L]
assumptions (4)
- domain assumption In an ETH eigenstate, subsystems with x<1/2 are thermal and with x>1/2 are not, following the entanglement structure of eigenstates versus thermal states.
- domain assumption Subsystem evolution speed is monotonically increasing with subsystem fraction x.
- ad hoc to paper The weak perturbation epsilon=1e-3 preserves the qualitative structure of the eigenstate, so the dynamics reflects the eigenstate's properties.
- domain assumption The specific eigenstates selected (ground state, 1/4-spectrum, one scar) are representative of their thermalization classes at the accessible sizes.
Cite this review
Pith. "Pith review of Single-eigenstate test of eigenstate thermalization hypothesis via perturbed eigenstate quench." pith.science (2026). https://pith.science/paper/MYUFQBA2
@misc{pith2026260804696,
author = {Pith},
title = {Pith review of: Single-eigenstate test of eigenstate thermalization hypothesis via perturbed eigenstate quench},
year = {2026},
howpublished = {\url{https://pith.science/paper/MYUFQBA2}},
note = {Machine review of arXiv:2608.04696}
}
read the original abstract
We propose and numerically validate an efficient single-eigenstate diagnostic for the eigenstate thermalization hypothesis (ETH) based on a perturbed eigenstate quench protocol. By introducing a weak random perturbation to an energy eigenstate to break its stationarity, we characterize the time-averaged subsystem evolution speed as a function of the subsystem-to-total system size ratio. The diagnostic relies on a robust qualitative distinction: eigenstates satisfying ETH exhibit an S-shaped curve with a clear inflection point near half the system size, while ETH-violating eigenstates display a convex J-shaped profile. We benchmark the criterion across paradigmatic one-dimensional spin chains covering chaotic, integrable, many-body localized, and quantum many-body scar regimes, obtaining full agreement with established thermalization phenomenology. Our method circumvents the need for explicit thermal ensemble construction, providing a robust, experimentally feasible probe of eigenstate thermalization at the single-eigenstate level.
Figures
Reference graph
Works this paper leans on
-
[19]
Subsystem Evolution Speed as Indicator of Relaxation
J. Zhang, M. A. Rajabpour, M. Heyl and R. Khasseh,Subsystem evolution speed as indicator of relaxation, Phys. Rev. B111, L140410 (2025), [arXiv:2410.17798]
work page Pith review arXiv 2025
-
[1]
J. M. Deutsch,Quantum statistical mechanics in a closed system, Phys. Rev. A43, 2046–2049 (1991)
1991
-
[2]
S+ j =S x j + iSy j being the standard local spin-1raising operator
The time average is taken over the time windowt∈[2L,4L]. S+ j =S x j + iSy j being the standard local spin-1raising operator. Each bimagnon state|Sn⟩is an exact eigen- state of the Hamiltonian with well-defined quantum numbers: energyE n =h(2n−L), total magnetiza- tionm n = 2n−L, and crystal momentumK n = nπL 2 (modL). The second family, known as the bond...
- [3]
-
[4]
Srednicki,Chaos and Quantum Thermalization, Phys
M. Srednicki,Chaos and Quantum Thermalization, Phys. Rev. E50, 888–901 (1994), [arXiv:cond-mat/9403051]
arXiv 1994
-
[5]
I. V. Gornyi, A. D. Mirlin and D. G. Polyakov, Interacting Electrons in Disordered Wires: Anderson Localization and Low-T Transport, Phys. Rev. Lett. 95, 206603 (Nov., 2005), [arXiv:cond-mat/0506411]. 6
work page Pith review arXiv 2005
- [6]
-
[7]
R. Nandkishore and D. A. Huse,Many body localization and thermalization in quantum statistical mechanics, Ann. Rev. Condensed Matter Phys.6, 15–38 (2015), [arXiv:1404.0686]
arXiv 2015
Show all 33 references
-
[8]
D. M. Basko, I. L. Aleiner and B. L. Altshuler,Metal insulator transition in a weakly interacting many-electron system with localized single-particle states, Ann. Phys.321, 1126–1205 (May, 2006), [arXiv:cond-mat/0506617]
2006 arXiv
-
[9]
D. A. Abanin, E. Altman, I. Bloch and M. Serbyn, Colloquium: Many-body localization, thermalization, and entanglement, Rev. Mod. Phys.91, 021001 (2019), [arXiv:1804.11065]
2019 arXiv
-
[10]
Alet and N
F. Alet and N. Laflorencie,Many-body localization: An introduction and selected topics, C. R. Phys.19, 498–525 (2018), [arXiv:1711.03145]
2018 arXiv
-
[11]
Bernien et al.,Probing many-body dynamics on a 51-atom quantum simulator, Nature551, 579–584 (2017), [arXiv:1707.04344]
H. Bernien et al.,Probing many-body dynamics on a 51-atom quantum simulator, Nature551, 579–584 (2017), [arXiv:1707.04344]
2017 arXiv
-
[12]
Sierant, M
P. Sierant, M. Lewenstein, A. Scardicchio, L. Vidmar and J. Zakrzewski,Many-Body Localization in the Age of Classical Computing,arXiv:2403.07111
-
[13]
Serbyn, D
M. Serbyn, D. A. Abanin and Z. Papić,Quantum many-body scars and weak breaking of ergodicity, Nature Phys.17, 675–685 (2021), [arXiv:2011.09486]
2021 arXiv
-
[14]
C. J. Turner, A. A. Michailidis, D. A. Abanin, M. Serbyn and Z. Papic,Weak ergodicity breaking from quantum many-body scars, Nature Phys.14, 745–749 (2018), [arXiv:1711.03528]
2018 arXiv
-
[15]
Chandran, T
A. Chandran, T. Iadecola, V. Khemani and R. Moessner,Quantum Many-Body Scars: A Quasiparticle Perspective, Ann. Rev. Condensed Matter Phys.14, 443–469 (2023), [arXiv:2206.11528]
2023 arXiv
-
[16]
Moudgalya, B
S. Moudgalya, B. A. Bernevig and N. Regnault, Quantum many-body scars and Hilbert space fragmentation: a review of exact results, Rept. Prog. Phys.85, 086501 (2022), [arXiv:2109.00548]
2022 arXiv
-
[17]
Dymarsky, N
A. Dymarsky, N. Lashkari and H. Liu,Subsystem eigenstate thermalization hypothesis, Phys. Rev. E 97, 012140 (2018), [arXiv:1611.08764]
2018 arXiv
-
[18]
J. R. Garrison and T. Grover,Does a single eigenstate encode the full Hamiltonian?, Phys. Rev. X8, 021026 (2018), [arXiv:1503.00729]
2018 arXiv
-
[20]
Khasseh, J
R. Khasseh, J. Zhang, M. Heyl and M. A. Rajabpour,Identifying Quantum Many-Body Integrability and Chaos Using Eigenstate Trace Distances, Phys. Rev. Lett.131, 216701 (2023), [arXiv:2301.13218]
2023 arXiv
-
[21]
E. H. Lieb, T. Schultz and D. Mattis,Two soluble models of an antiferromagnetic chain, Annals Phys. 16, 407 (1961)
1961
-
[22]
M. A. Nielsen and I. L. Chuang,Quantum Computation and Quantum Information. Cambridge University Press, Cambridge, UK, 10th anniversary ed., 2010, 10.1017/CBO9780511976667
2010 doi
-
[23]
Pfeuty,The one-dimensional Ising model with a transverse field, Annals Phys.57, 79 (1970)
P. Pfeuty,The one-dimensional Ising model with a transverse field, Annals Phys.57, 79 (1970)
1970
-
[24]
Katsura,Statistical mechanics of the anisotropic linear Heisenberg model, Phys
S. Katsura,Statistical mechanics of the anisotropic linear Heisenberg model, Phys. Rev.127, 1508 (1962)
1962
-
[25]
A. C. Cassidy, C. W. Clark and M. Rigol, Generalized Thermalization in an Integrable Lattice System, Phys. Rev. Lett.106, 140405 (2011), [arXiv:1008.4794]
2011 arXiv
-
[26]
Calabrese, F
P. Calabrese, F. H. L. Essler and M. Fagotti, Quantum Quench in the Transverse Field Ising Chain, Phys. Rev. Lett.106, 227203 (2011), [arXiv:1104.0154]
2011 arXiv
-
[27]
H. Kim, T. N. Ikeda and D. A. Huse,Testing whether all eigenstates obey the eigenstate thermalization hypothesis, Phys. Rev. E90, 052105 (2014), [arXiv:1408.0535]
2014 arXiv
-
[28]
M. C. Bañuls, J. I. Cirac and M. B. Hastings,Strong and Weak Thermalization of Infinite Nonintegrable Quantum Systems, Phys. Rev. Lett.106, 050405 (2011), [arXiv:1007.3957]
2011 arXiv
-
[29]
Pal and D
A. Pal and D. A. Huse,Many-body localization phase transition, Phys. Rev. B82, 174411 (2010), [arXiv:1010.1992]
2010 arXiv
-
[30]
A. W. Sandvik,Computational Studies of Quantum Spin Systems, AIP Conf. Proc.1297, 135 (2010), [arXiv:1101.3281]
2010 arXiv
-
[31]
Schecter and T
M. Schecter and T. Iadecola,Weak Ergodicity Breaking and Quantum Many-Body Scars in Spin-1 XY Magnets, Phys. Rev. Lett.123, 147201 (2019), [arXiv:1906.10131]
2019 arXiv
-
[32]
D. J. Luitz, N. Laflorencie and F. Alet,Many-body localization edge in the random-field Heisenberg chain, Phys. Rev. B91, 081103 (2015), [arXiv:1411.0660]
2015 arXiv
-
[43]
The time average is com- puted over the time windowt∈[2L,4L]
We have taken∆ = 1for both panels. The time average is com- puted over the time windowt∈[2L,4L]. turbed eigenstates, for the disordered XXZ chain in both the chaotic and MBL regimes. From the pro- file of the curves, we conclude that the 1/4-spectrum eigenstate satisfies ETH i...
Reviewed August 6, 2026 · model on record in the stance chip above.
Discussion (0). Sign in to comment.