A Multi-Resolvent Hierarchy for the ETH Smooth Function

Zhiqiang Huang

REVIEW 4 major objections 4 minor 23 references

A Multi-Resolvent Hierarchy for the ETH Smooth Function

T0 review · 4 major / 4 minor · reviewed 2026-08-01 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The ETH smooth function is not a free fit: it decomposes into a diagonal baseline plus a multi-resolvent series whose third-order term is generically odd under ω→−ω.

desk verdict The exact core — projector sum rule, two-frequency kernel, Appendix A moment preservation — is solid; the closed hierarchy and the r=3 odd-parity signature are conditional on a closure imported from an unpublished same-author companion. read the letter →

arxiv 2607.19861 v1 pith:TG44D36Z submitted 2026-07-22 quant-ph

Zhiqiang Huang This is my paper · ORCID

classification quant-ph

keywords eigenstatethermalizationhypothesisETHsmoothfunctionmulti-resolventhierarchydiagonalclosureapproximationcorrelationskewnessprojectorsumruleoff-diagonalmatrixelementsparityunderfrequencyreversal

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper aims to turn the smooth function in the eigenstate thermalization hypothesis from a phenomenological fitting object into a computable microscopic quantity. It proves an exact split of the off-diagonal ETH variance into a diagonal overlap baseline D_ji and a correlation correction g_ji, then expresses g_ji as a series Σ_{r≥2} g_ji^{(r)} whose terms are built only from diagonal spectral functions and interaction matrix elements. The r=2 term is even under ω→−ω; the r=3 term generically carries the first odd-parity, skewness component, something no single-resolvent closure can produce. A rigorous projector sum rule forces the integrated off-diagonal correlation to be negative of order unity, and the framework explains that negativity as the signature of cavity subtraction restoring projector idempotency. If the derivation holds, the ETH smooth function becomes a closed, systematically improvable series with a testable parity signature.

What carries the argument

The central object is the two-frequency correlation kernel K^{ji}_{μν}(z_1,z_2) = R_{μj,νj}(z_1) R_{νi,μi}(z_2), a product of off-diagonal resolvents that connect different bath labels at the same system label. Its double spectral representation exactly reproduces the ETH correlation term C^{jij}_{nmn}, which is inaccessible to diagonal overlaps. Each off-diagonal resolvent is expanded through a recursive projection identity into levels containing increasing numbers of diagonal resolvents; the diagonal closure approximation replaces projected diagonal resolvents by full ones, turning the kernel into an explicit series ordered by interaction-vertex count r. The parity of each level is governe

What would settle it

Exact-diagonalize a nonintegrable spin or fermion chain, construct σ^{ji}_{nm} = ⟨φ_S^j|Tr_B |ψ_n⟩⟨ψ_m| |φ_S^i⟩, bin the squared off-diagonal elements by E_+ and ω, and form the odd-parity projection C_odd(ω) = [C(ω)-C(-ω)]/2. The hierarchy predicts C_odd is generically nonzero at third order, with magnitude set by the three-channel coupling and spectral overlap; observing C_odd consistent with zero at that scale, or finding a nonzero odd part in a regime where the DCA is known to fail, would falsify the central claim.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that f_{ji}^2(E_+,ω) = D_{ji}(E_+,ω) + Σ_{r≥2} g_{ji}^{(r)}(E_+,ω), where the exact first term comes from diagonal overlaps and every correlation level g^{(r)} is determined by diagonal resolvents and fixed couplings. The leading level g^{(2)} is parity-even, while g^{(3)} generically contains an odd-parity component, so the correlation C^{jij}_{nmn} is asymmetric in ω; this odd component is the defining signature of multi-channel resolvent interference and is rigorously absent from all single-resolvent closures. The derivation is built on an exact two-frequency resolvent kernel and a recursive projection expansion, closed by a diagonal clos

Load-bearing premise

The load-bearing step is the diagonal closure approximation of Eq. (63), which replaces every projected diagonal resolvent by the full diagonal resolvent; the explicit closed forms for g^{(2)} and g^{(3)} and the parity classification hold only under this replacement, and the paper's validation of it is imported from an unpublished companion preprint by the same author. If the approximation error is not exponentially suppressed, the microscopic theory of f_ji is unjustified,

Editorial extensions

If this is right

The ETH smooth function f_ji can be computed, not fitted: its even part is a convolution of diagonal spectral functions and its odd part arises from three-channel interference, with all couplings fixed by the Hamiltonian.
A nonzero antisymmetric component of the correlation under ω→−ω becomes a direct numerical or experimental diagnostic for multi-channel interference; any observation of such skewness rules out all single-resolvent closures.
Truncating at r=2 recovers the even, Gaussian-like sector, while retaining r=3 adds the first non-Gaussian skewness; the entropy scaling g^{(r)} ~ e^{-(r-1)S/2} makes the series systematically improvable.
The projector sum rule gives an exact, model-independent constraint on off-diagonal correlations: the integrated off-diagonal correlation is negative of order unity, tying ETH correlations to pure-state idempotency.
The same hierarchy extends to OTOC four-point functions, Krylov Lanczos coefficients, work distributions, and open-system memory kernels, each inheriting the parity signature.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the r=3 prediction is robust, exact-diagonalization studies could isolate the odd component C_odd(ω) and use its sign and shape to infer the three-channel coupling V^{(3)}, effectively imaging the interference network.
The framework's validity hinges on entropy dilution; in small or moderately sized systems where e^{-S} is not tiny, the DCA error could dominate, so the hierarchy's predictive regime may be narrower than the formal S→∞ argument suggests.
The even-r/odd-r parity pattern invites the conjecture that all odd higher-order cumulants appear first at corresponding odd r levels; a connected multi-frequency generating functional would make this precise.
The OTOC connection implies a testable time-reversal asymmetry in four-point spectral functions of chaotic systems, a quantitative prediction worth extracting from the general spectral formula.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Two things to know. The rigorous core is solid: Theorem 1, the exact decomposition of |sigma|^2 into the diagonal product plus a correlation term, the two-frequency resolvent kernel, and the moment-preservation results in Appendix A are correct and self-contained. Everything the paper advertises as new and predictive — the closed hierarchy g = sum_r g^(r), the parity classification, the r=3 skewness signature — is conditional on a diagonal closure approximation (DCA) whose validity is imported from Ref. [11], an unpublished companion by the same author.

Credit where due. The projector sum rule is a genuinely nice result: sum_m C^{jij}_{nmn} = 0, with a negative O(1) integrated bias in the diagonal sector, follows purely from projector idempotency — no ETH input. The reformulation of the mu-neq-nu correlation through K = R_{mu j,nu j} R_{nu i,mu i} is exact and clean, and it makes clear why diagonal overlaps alone cannot see this sector. Appendix A is the other real contribution: it proves the DCA preserves the first three spectral moments of the off-diagonal resolvent and fixes the leading deviation at V_{alpha beta}|V_{alpha beta}|^2/z^4. Parameter-free and checkable. The paper is also honest about its own scaffolding; it labels Assumption A and states plainly that the combinatorial suppression proof is out of scope.

Soft spots, in proportion. The DCA is load-bearing, and the paper's own justification is thin: the entropy-dilution argument in Sec. IV A assumes the very delocalization the construction is supposed to ground, and Appendix A's moment matching constrains only the high-frequency Laurent regime. The r=3 odd-parity term is a finite-frequency interference effect; high-frequency moment preservation says nothing about its sign or magnitude. The reader's central worry is correct. The parity classification also rests on unproven inputs — approximate spectral symmetry about renormalized centers, near-degenerate centers (footnote 1), and generic absence of accidental cancellation — all acknowledged, none established.

One gap the stress-test missed: the entropy-scaling estimate, Eq. (102), does not close. The displayed counting e^S (e^{S_B} e^{-S})^r f^{2r+3} does not produce e^{-(r-1)S/2} as written, and the paper quotes e^{-(r-2)S/2} and e^{-(r-1)S/2} for the same equation in different places (Sec. V.C vs. Sec. V.E/VII). So the "controlled expansion" claim needs an actual derivation before truncation errors can be trusted. No numerics are reported; fine for a theory paper, but the falsifiable content is untested.

Bottom line: the exact core is worth taking seriously; the constructive hierarchy is a program, not yet a result. I would send it to referees — the DCA dependence and the scaling are exactly what a referee should pressure-test. I would cite the sum rule and Appendix A. For a reading group, only if someone has actually read Ref. [11]; otherwise the conditional claims can't be evaluated in the room.

Referee Report

4 major / 4 minor

Summary. The paper develops a multi-resolvent hierarchy to give a microscopic description of the ETH smooth function f_ji(E+,ω). It starts from an exact decomposition of the off-diagonal ETH variance into a diagonal overlap baseline D_ji and a correlation term C, proves a projector sum rule, and represents C exactly through a two-frequency resolvent kernel. Under a diagonal closure approximation (DCA), it obtains a series f_ji² = D_ji + Σ_{r≥2} g_ji^(r), with g^(2) even and g^(3) odd under ω↔−ω. The odd-parity term is presented as the first signature of multi-channel resolvent interference, absent from single-resolvent closures. The exact decomposition, sum rule, and resolvent representation are self-contained and appear correct; the constructive hierarchy and the parity prediction are conditional on the DCA, which is imported mainly from an unpublished companion paper.

Significance. If the DCA can be justified, the paper would provide a genuinely new microscopic derivation of the ETH smooth function and a testable odd-parity skewness prediction, beyond the phenomenological Gaussian ansatz. The exact projector sum rule (Theorem 1) and the resolvent kernel representation are valuable in themselves and do not depend on the DCA. The paper also contains a useful high-frequency moment-preservation analysis (Appendix A) showing that the DCA reproduces the first three spectral moments of the off-diagonal resolvent. However, the central constructive claim—closed formulas for g^(2), g^(3), and especially the r=3 odd-parity component—rests on an imported closure whose validity is not established in this manuscript. The entropy-suppression estimate used to justify truncation is also not derived and appears algebraically questionable.

major comments (4)

[Sec. IV A, Eq. (63)] The central constructive claim—the closed forms for g^(2), g^(3) and the odd-parity prediction—rests entirely on the DCA replacement R_α^(S) ≈ R_α. The paper says the DCA is "established" in Ref. [11], but the only self-contained support here is the heuristic entropy-dilution argument in Sec. IV A, which assumes a random-phase cancellation that is essentially the ETH property the hierarchy is supposed to organize, and Appendix A, which proves moment preservation only in the high-frequency Laurent regime (Eq. A5), with first deviation at O(z^-4). High-frequency moment matching does not constrain the finite-frequency r=3 odd component. Unless the DCA is validated numerically or an error bound is supplied, Eqs. (98)–(100) are not an established microscopic theory but a proposal conditional on an imported closure.
[Sec. V D, Eq. (102)] The entropy-scaling estimate g^(r) ~ e^{-(r-1)S/2} is asserted without derivation and is algebraically suspect as written: e^S (e^{S_B} e^{-S})^r f^{2r+3}, with S_B = S + const and f O(1), gives e^S, not e^{-(r-1)S/2}. Moreover, the text takes |V|² ~ e^{-S}, which is ETH eigenstate-matrix-element scaling, whereas V in Eqs. (44)–(45) is the interaction in the fixed unperturbed product basis; local few-body couplings are not exponentially small. This undermines the claimed systematic improvability and the truncation control stated in Eq. (102).
[Sec. III E, Eqs. (61)/(99), footnote 1] The odd-parity conclusion is stated as "generically" non-zero, but depends on three unproven assumptions: (i) each f_α is even about its renormalized centre, (ii) the centres satisfy ã_µj ≈ ã_νj for dominant pairs, and (iii) no accidental cancellation among bath triples after summation. These are acknowledged in footnote 1 and the surrounding text, but they should be elevated to explicit hypotheses of a theorem, with the parameter regime stated. As it stands, Eq. (99) is not a rigorous prediction but a plausible heuristic.
[Sec. IV C, Eq. (79) and Corollary (82)] The restoration argument leading to Eq. (82) is explicitly conditional on Assumption A and the positivity condition Eq. (81), both unproven. The logical-chain diagram on p. 14 marks these as conditional, which is commendable, but the abstract and conclusion state the negative-bias result as a rigorous consequence of projector idempotency. The exact Theorem 1 is rigorous; the DCA-level mechanistic explanation is not. Please separate these sharply in the presentation.

minor comments (4)

[Sec. III E] The sentence "the [2,3] and [3,2] contributions transform with opposite relative signs because the intermediate resolvent R_ξk appears in different frequency slots" appears twice verbatim. Please remove the duplication.
[Table I and Sec. IV G] Table I lists definite dominant parities for r=4 and r=5, but Sec. IV G states that a complete parity classification of higher levels remains open. Please clarify that the table entries for r≥4 are speculative or define the precise sense in which they are claimed.
[Sec. V F] The statement that "no free fitting functions remain" is too strong. The spectral functions f_µi are taken from Ref. [11]; in the present paper they are inputs. Please rephrase to "no free parameters are introduced in this work beyond the spectral functions determined in the companion paper."
[Appendix A, Eq. (A5)] The leading DCA deviation V_αβ |V_αβ|² / z^4 is derived for off-diagonal resolvents. It would be helpful to state explicitly whether this bound applies uniformly for all α,β or only for typical pairs, and how it translates to the binned correlation C in the thermodynamic limit.

Circularity Check

2 steps flagged · score 4.0 of 10

The exact decomposition and sum rule are self-contained, but the central predictive content — the closed DCA formulas and the r=3 odd-parity signature — is loaded onto the same-author unpublished Ref. [11] for DCA validity and for the symmetric spectral functions, with no in-paper finite-frequency verification.

self citation load bearing [Sec. III A around Eq. (43) and Sec. IV A, Eq. (63)]
"Under the DCA, whose validity in the ETH regime is established in Ref. [11] and section IV, the projected diagonal resolvents are replaced by their full counterparts: R^{(S)}_α ≈ R_α. ... The DCA, introduced and justified in Ref. [11], replaces every projected diagonal resolvent by its full counterpart: R^{(µi,α1,...,αm)}_α (z)≈ R_α(z). (63)"

All computed correlation levels (Eqs. (52)-(53), (66)-(70), (97)-(98)) and the resulting parity prediction are DCA-level statements. DCA validity is the load-bearing input, yet it is not established in this paper: it is deferred to Ref. [11], an unpublished preprint by the same first author. The in-paper justification is an entropy-dilution heuristic that already assumes the ETH e^{-S/2} off-diagonal scaling, plus Appendix A's high-frequency moment matching, which does not constrain the finite-frequency r=3 interference. The r=3 odd-parity 'prediction' therefore inherits its validity from the authors' own prior framework rather than from an independent derivation.
ansatz smuggled in via citation [Sec. III D-E, Eqs. (56)-(61), and footnote 1]
"adopting the standard approximation—inherited from the resolvent framework of Ref. [11]—that each smooth spectral function f_α is approximately symmetric about its renormalised centre, each spectral function transforms approximately as f≈f (even), while its Hilbert transform transforms approximately as H[f]≈−H[f] (odd). ... The parity analysis therefore additionally requires ã_{µj}≈ã_{νj} for the dominant bath pairs."

The odd-parity 'prediction' at r=3 is obtained by counting Hilbert transforms under symmetry assumptions imported from Ref. [11], not derived here. The paper's own footnote admits the parity result requires the near-degeneracy condition ã_{µj}≈ã_{νj}; and the evenness of f_α is inherited from the same prior framework. If the diagonal spectral functions are not even about the renormalized centers, Eqs. (61) and (99) are not guaranteed. Thus the headline parity signature is a consequence of assumed symmetry properties of the very diagonal spectral functions the framework claims to determine, i.e. an ansatz inherited by citation rather than a model-independent prediction.

full rationale

The paper contains a genuinely self-contained exact core: Eq. (11) is an algebraic identity; Theorem 1 and Eq. (15)/(73) follow from projector idempotency; Eq. (26) is a definitional decomposition; Eq. (41) is an exact projection expansion; Appendix A proves the high-frequency Laurent moment preservation M0,M1,M2 and the leading O(z^{-4}) deviation. These parts are not circular and would survive even if the DCA failed. The circularity burden is concentrated in the claim that the hierarchy gives a closed, microscopic, 'no free fitting functions' theory of f_ji with a robust odd-parity r=3 signature. That claim depends on (i) the DCA closure Eq. (63), whose validity is deferred to the same-author unpublished Ref. [11]; (ii) the approximate evenness of f_α about renormalized centers, also inherited from Ref. [11]; and (iii) the near-degeneracy condition in footnote 1. The in-paper entropy-dilution argument assumes the ETH e^{-S/2} scaling that the framework is meant to make precise for f_ji, and Appendix A's moment preservation is high-frequency and does not constrain finite-frequency r=3 interference. Hence the predictive part is not self-contained: it is loaded onto a self-citation chain, and the parity conclusion is conditional on imported symmetry assumptions. This is moderate circularity — score 4 — because the exact identities and algebraic hierarchy are independent content and the reduction is conditional rather than definitional; a score of 6+ would be warranted only if the DCA and spectral symmetries were redefined as the target result, which is not the case here.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The central derivation rests on exact resolvent/projection identities (standard math) plus several domain assumptions. No data-fit free parameters appear in this paper itself; the spectral functions f_α are taken from the companion self-consistent framework. The most fragile assumptions are the DCA, the symmetric-spectral-function parity input, Assumption A on r=2 dominance, and the random-phase no-cancellation condition for the odd term.

assumptions (5)

domain assumption DCA: every projected diagonal resolvent R_α^{(S)} is replaced by the full diagonal resolvent R_α.
Eq. (63); all explicit closed forms for g^(r) and the parity statements depend on this replacement; validity is argued by entropy dilution in Ref [11].
domain assumption Smooth spectral functions f_α are approximately symmetric about their renormalised centres ~a_α, with ~a_{μj} ≈ ~a_{νj} for dominant bath pairs.
Sec. III D footnote and Sec. IV B; the parity classification turns on f behaving even and H[f] odd under reflection, and on near-degenerate renormalized centers.
ad hoc to paper Assumption A: the r=2 term dominates the DCA hierarchy in magnitude.
Eq. (79); explicitly labelled as an unproved working assumption, needed for the claimed sign of the integrated cavity correction.
domain assumption ETH scaling: |V|^2 ~ e^{-S}, S ≈ S_B, and random phases suppress accidental cancellations.
Eqs. (102) and Sec. III E; used for the 'systematically improvable' claim and for survival of the odd-parity term after bath summation.
standard math Completeness of exact eigenstates, orthonormality of the unperturbed basis, and the Feshbach projection identity.
Used in Secs. II B-II D to derive the projector sum rule, the off-diagonal resolvent sum rule, and the recursive projection expansion.

how reviews work

0 comments

Cite this review

Pith. "Pith review of A Multi-Resolvent Hierarchy for the ETH Smooth Function." pith.science (2026). https://pith.science/paper/TG44D36Z

@misc{pith2026260719861,
  author       = {Pith},
  title        = {Pith review of: A Multi-Resolvent Hierarchy for the ETH Smooth Function},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TG44D36Z}},
  note         = {Machine review of arXiv:2607.19861}
}

abstract

The eigenstate thermalization hypothesis (ETH) provides a statistical description of thermalization in isolated quantum many-body systems, yet the phenomenological smooth function $f_O(\bar{E},\omega)$ -- which controls the energy dependence of off-diagonal matrix elements -- lacks a systematic microscopic foundation. We develop a multi-resolvent hierarchy for the correlation corrections entering the ETH smooth function. Using recursive projection identities together with a diagonal closure approximation (DCA), the hierarchy organizes multi-channel interference processes by the number of interacting bath channels, replacing the uncontrolled neglect of higher-order correlations with a systematically improvable expansion. The ETH smooth function is thereby obtained as $f_{ji}^2 = D_{ji} + \sum_{r\ge 2} g_{ji}^{(r)}$, where the diagonal baseline $D_{ji}$ and each correlation level $g_{ji}^{(r)}$ are expressed entirely through diagonal spectral functions and microscopic interaction couplings, providing a unified, closed, and systematically improvable microscopic theory. A rigorous projector sum rule constrains the entire hierarchy: the integrated off-diagonal correlation carries a negative bias of order unity, a consequence of projector idempotency. The hierarchy further reveals a parity structure in which even-$r$ sectors carry even parity under $\omega\to-\omega$ while the $r=3$ sector generates the first odd-parity (skewness) contribution -- absent from all single-resolvent closures -- suggesting experimentally testable signatures in quantum many-body systems.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

23 extracted references · 3 linked inside Pith

[11]

D’Alessio, Y

L. D’Alessio, Y. Kafri, A. Polkovnikov, M. Rigol, From quantum chaos and eigenstate thermalization to statisti- cal mechanics and thermodynamics, Advances in Physics, 65 (2016) 239-362

2016
[1]

Expanding at large|z|>∥H∥yields Rαβ(z) = ∞X k=0 M αβ k zk+1 , M αβ k := X n λk n qαβ n =⟨ϕ α|H k |ϕβ⟩

Laurent expansion and spectral moments The off-diagonal resolvent admits the spectral repre- sentation Rαβ(z) = X n qαβ n z−λ n , q αβ n :=⟨ϕ α|ψn⟩⟨ψn|ϕβ⟩, (A1) where|ψ n⟩are the exact eigenstates ofHwith eigen- valuesλ n, and|ϕ α⟩are the unperturbed basis states. Expanding at large|z|>∥H∥yields Rαβ(z) = ∞X k=0 M αβ k zk+1 , M αβ k := X n λk n qαβ n =⟨ϕ α...
[2]

Consequently, any truncation withℓ≥2automatically satisfiesR [≤L] αβ (z) = O(z −2), guaranteeing orthogonality without inter-level cancellation

Scaling of the hierarchy levels Lemma 1.Forα̸=βandℓ≥2, each level of the exact hierarchy (41) and of the DCA hierarchy (43) sat- isfiesR (ℓ) αβ(z) =O(|z| −ℓ)as|z| → ∞. Consequently, any truncation withℓ≥2automatically satisfiesR [≤L] αβ (z) = O(z −2), guaranteeing orthogonality without inter-level cancellation. Proof.For|z|>∥H∥, the full resolvent admits ...
[3]

Forα /∈ S, the projected diagonal resolvent R(S) α (z) :=⟨ϕ α|(z−Φ S HΦ S )−1 |ϕα⟩admits the large- zexpansion R(S) α (z) = 1 z + Hαα z2 + (H 2)αα − P γ∈S |Vαγ|2 z3 +O(z −4)

Projected diagonal resolvent expansion Proposition 1.LetSbe a set of unperturbed basis-state indices. Forα /∈ S, the projected diagonal resolvent R(S) α (z) :=⟨ϕ α|(z−Φ S HΦ S )−1 |ϕα⟩admits the large- zexpansion R(S) α (z) = 1 z + Hαα z2 + (H 2)αα − P γ∈S |Vαγ|2 z3 +O(z −4). (A3) Proof.SetP= P γ∈S |ϕγ⟩⟨ϕγ|, so Φ S =I−P. Since α /∈ S, ΦS |ϕα⟩=|ϕ α⟩and⟨ϕ α...
[4]

Proof.By Lemma 1, onlyℓ= 2 contributes atz −2

Preservation of the first two moments Corollary 1(First moment).Forα̸=β, the DCA hierarchy exactly reproduces the first spectral moment: M αβ,DCA 1 =V αβ =M αβ,exact 1 . Proof.By Lemma 1, onlyℓ= 2 contributes atz −2. The z−2 coefficient ofR (2) αβ uses only theO(z −1) terms of the two diagonal resolvents, which are identical (= 1/z) for projected and full...
[5]

Leading DCA deviation Proposition 3(Leading DCA deviation).For the hi- erarchy defined in Eqs. (41)–(43), and forα̸=β, the leading large-zdeviation between the DCA and exact off- diagonal resolvents is, at orderz −4, RDCA αβ (z)− Rexact αβ (z) = Vαβ |Vαβ|2 z4 +O(z −5) .(A5) Proof.Levelsℓ= 2,3,4 contribute atz −4. We examine each. 27 Levelℓ= 2:Thez −4 coef...
[6]

matching moments

Hilbert-transform interpretation of moment preservation The moment-preservation results established above ad- mit an equivalent spectral interpretation through the Hilbert-transform orthogonality condition of Sec. III B. The zeroth momentM αβ 0 = 0 forα̸=βis the state- ment R dλℑR αβ(λ−i0 +) = 0, i.e., the integrated orthog- onality of the off-diagonal sp...
[7]

Path–moment correspondence (conjecture) The explicit verification ofM 1 andM 2 reveals a struc- tural pattern: different hierarchy levels correspond to dif- ferent numbers of interaction vertices along the projected propagation path. We conjecture that theℓ-th level of the exact projection hierarchy collects all contributions to the resolvent with precise...

Show all 23 references

[8]

J. M. Deutsch, Quantum statistical mechanics in a closed system, Phys. Rev. A43, 2046 (1991)
[9]

Srednicki, Chaos and quantum thermalization, Phys

M. Srednicki, Chaos and quantum thermalization, Phys. Rev. E50, 888 (1994)

1994
[10]

Rigol, V

M. Rigol, V. Dunjko, and M. Olshanii, Thermalization and its mechanism for generic isolated quantum systems, Nature452, 854 (2008). 28

2008
[12]

Foini, J

L. Foini, J. Kurchan, Eigenstate thermalization hypoth- esis and out of time order correlators, Physical Review E, 99 (2019) 042139

2019
[13]

E. N. Economou, Green’s Functions in Quantum Physics, Springer, Berlin (2006)

2006
[14]

G. D. Mahan, Many-Particle Physics, Springer, New York (2000)

2000
[15]

Feshbach, Unified theory of nuclear reactions, Ann

H. Feshbach, Unified theory of nuclear reactions, Ann. Phys. (N.Y.)5, 357 (1958)

1958
[16]

Helbig, T

T. Helbig, T. Hofmann, R. Thomale, M. Greiter, Theory of Eigenstate Thermalisation, arXiv preprint arXiv:2406.01448 (2024)

2024 arXiv
[17]

Huang, Q.-y

Z. Huang, Q.-y. Cai, On the generic increase of entropy in isolated systems, arXiv preprint arXiv:2505.23041 (2025)

2025
[18]

Huang, Q.-y

Z. Huang, Q.-y. Cai, Resolvent-Based Self-Consistent Framework with Hierarchical Correlation Expansion for Strongly Correlated Many-Body Systems, arXiv preprint arXiv:2604.00606 (2026)

2026 arXiv
[19]

Murthy, M

C. Murthy, M. Srednicki, Bounds on Chaos from the Eigenstate Thermalization Hypothesis, Physical Review Letters, 123 (2019) 230606

2019
[20]

Barb´ on, E

J.L.F. Barb´ on, E. Rabinovici, R. Shir, R. Sinha, On the evolution of operator complexity beyond scrambling, Journal of High Energy Physics, 2019 (2019) 264

2019
[21]

Huang, Integral fluctuation theorems and trace- preserving map, Physical Review E, 109 (2024) 064111

Z. Huang, Integral fluctuation theorems and trace- preserving map, Physical Review E, 109 (2024) 064111

2024
[22]

O’Donovan, P

P. O’Donovan, P. Strasberg, K. Modi, J. Goold, M.T. Mitchison, Quantum master equation from the eigenstate thermalization hypothesis, arXiv preprint arXiv:2411.07706 (2024)

2024 arXiv
[23]

D’Alessio, Y

L. D’Alessio, Y. Kafri, A. Polkovnikov, and M. Rigol, From quantum chaos and eigenstate thermalization to statistical mechanics and thermodynamics, Adv. Phys. 65, 239 (2016)

2016

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.

[1] [11]

D’Alessio, Y

L. D’Alessio, Y. Kafri, A. Polkovnikov, M. Rigol, From quantum chaos and eigenstate thermalization to statisti- cal mechanics and thermodynamics, Advances in Physics, 65 (2016) 239-362

2016

[2] [1]

Expanding at large|z|>∥H∥yields Rαβ(z) = ∞X k=0 M αβ k zk+1 , M αβ k := X n λk n qαβ n =⟨ϕ α|H k |ϕβ⟩

Laurent expansion and spectral moments The off-diagonal resolvent admits the spectral repre- sentation Rαβ(z) = X n qαβ n z−λ n , q αβ n :=⟨ϕ α|ψn⟩⟨ψn|ϕβ⟩, (A1) where|ψ n⟩are the exact eigenstates ofHwith eigen- valuesλ n, and|ϕ α⟩are the unperturbed basis states. Expanding at large|z|>∥H∥yields Rαβ(z) = ∞X k=0 M αβ k zk+1 , M αβ k := X n λk n qαβ n =⟨ϕ α...

[3] [2]

Consequently, any truncation withℓ≥2automatically satisfiesR [≤L] αβ (z) = O(z −2), guaranteeing orthogonality without inter-level cancellation

Scaling of the hierarchy levels Lemma 1.Forα̸=βandℓ≥2, each level of the exact hierarchy (41) and of the DCA hierarchy (43) sat- isfiesR (ℓ) αβ(z) =O(|z| −ℓ)as|z| → ∞. Consequently, any truncation withℓ≥2automatically satisfiesR [≤L] αβ (z) = O(z −2), guaranteeing orthogonality without inter-level cancellation. Proof.For|z|>∥H∥, the full resolvent admits ...

[4] [3]

Forα /∈ S, the projected diagonal resolvent R(S) α (z) :=⟨ϕ α|(z−Φ S HΦ S )−1 |ϕα⟩admits the large- zexpansion R(S) α (z) = 1 z + Hαα z2 + (H 2)αα − P γ∈S |Vαγ|2 z3 +O(z −4)

Projected diagonal resolvent expansion Proposition 1.LetSbe a set of unperturbed basis-state indices. Forα /∈ S, the projected diagonal resolvent R(S) α (z) :=⟨ϕ α|(z−Φ S HΦ S )−1 |ϕα⟩admits the large- zexpansion R(S) α (z) = 1 z + Hαα z2 + (H 2)αα − P γ∈S |Vαγ|2 z3 +O(z −4). (A3) Proof.SetP= P γ∈S |ϕγ⟩⟨ϕγ|, so Φ S =I−P. Since α /∈ S, ΦS |ϕα⟩=|ϕ α⟩and⟨ϕ α...

[5] [4]

Proof.By Lemma 1, onlyℓ= 2 contributes atz −2

Preservation of the first two moments Corollary 1(First moment).Forα̸=β, the DCA hierarchy exactly reproduces the first spectral moment: M αβ,DCA 1 =V αβ =M αβ,exact 1 . Proof.By Lemma 1, onlyℓ= 2 contributes atz −2. The z−2 coefficient ofR (2) αβ uses only theO(z −1) terms of the two diagonal resolvents, which are identical (= 1/z) for projected and full...

[6] [5]

Leading DCA deviation Proposition 3(Leading DCA deviation).For the hi- erarchy defined in Eqs. (41)–(43), and forα̸=β, the leading large-zdeviation between the DCA and exact off- diagonal resolvents is, at orderz −4, RDCA αβ (z)− Rexact αβ (z) = Vαβ |Vαβ|2 z4 +O(z −5) .(A5) Proof.Levelsℓ= 2,3,4 contribute atz −4. We examine each. 27 Levelℓ= 2:Thez −4 coef...

[7] [6]

matching moments

Hilbert-transform interpretation of moment preservation The moment-preservation results established above ad- mit an equivalent spectral interpretation through the Hilbert-transform orthogonality condition of Sec. III B. The zeroth momentM αβ 0 = 0 forα̸=βis the state- ment R dλℑR αβ(λ−i0 +) = 0, i.e., the integrated orthog- onality of the off-diagonal sp...

[8] [7]

Path–moment correspondence (conjecture) The explicit verification ofM 1 andM 2 reveals a struc- tural pattern: different hierarchy levels correspond to dif- ferent numbers of interaction vertices along the projected propagation path. We conjecture that theℓ-th level of the exact projection hierarchy collects all contributions to the resolvent with precise...