{"id":"01d6a92a-402d-4071-823e-6370524e0849","arxiv_id":"2601.10409","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For any finite-dimensional Hamiltonian and initial pure state, the first recurrence time obeys t_rec ≤ t_exit (4π/ε)^{d-1}, with an explicit inverse quantum speed limit bound on t_exit.","lead":"This paper proves rigorous upper bounds on the time a closed quantum system takes to return near its initial state, for any Hamiltonian and any initial pure state. It also provides a partial inverse quantum speed limit and a random-Hamiltonian lower bound, sharpening and making rigorous earlier heuristic arguments.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The generic-saturation claim is not supported: the proof of Theorem S2 yields a lower bound with a weaker ε-exponent than the upper bound from Theorems 1+2.","rationale":"The upper-bound theorems (Theorem 1 and Theorem 2) appear internally correct and self-contained; the main deterministic bound is not threatened. The reader's CONDITIONAL verdict is appropriate because the auxiliary tightness/saturation claims are not fully supported. However, the reader's weakest_assumption focused on representativeness of random diagonal Hamiltonians and a missing 2^d union-bound factor. The more load-bearing issue is that even if those factors are fixed, the lower-bound proof yields a different ε-exponent than the upper bound, so the central 'tight bounds' claim is not valid as stated. This does not require rejecting the paper—the deterministic upper bound is still a genuine contribution—but it does require either a corrected lower-bound analysis or a revised claim about saturation. Hence the verdict stays CONDITIONAL.","tokens_in":19326,"tokens_out":42085,"duration_ms":353380,"concrete_test":"Re-derive T0 in SM Eq. (88) step by step, applying Prop. S5 with ε replaced by 2ε in Eq. (82) and keeping all factors. Then compare the ε-exponent of the resulting lower bound with the upper bound U = Θ(ε^{-(d-2)}) obtained from Theorems 1 and 2 for the same random model. If the lower bound scales as ε^{-(d-3)} (up to d-dependent constants) while the upper bound scales as ε^{-(d-2)}, the claimed generic saturation fails and the paper should either weaken the saturation claim or supply a genuinely matching lower bound.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's title and abstract assert that the recurrence-time upper bound is 'generically saturated' by random Hamiltonians. This is not established by the SM proof. In the random-diagonal model of Theorem 3, Theorem 1 (Eq. 3) together with Theorem 2 (Eq. 4) and Proposition S2 gives t_exit = Θ(ε) (specifically √2 ε ≤ t_exit ≤ 2√5 ε), so the upper bound is U = O(ε^{-(d-2)}) after multiplying by (4π/ε)^{d-1}. The lower-bound proof of Theorem S2, however, only establishes t_rec ≥ T0 with T0 = (ε^3/2π)(3/(4C))^d ε^{-d} = O(ε^{-(d-3)}) (SM Eq. 88). Thus the proven lower bound is smaller than the upper bound by a factor ~1/ε; in the theorem's allowed range ε > (40π)^{1/3}(1/2)^{d/3}, this gap grows like 2^{d/3}, so the bounds are not tight. The main-text statement t_rec ≳ (1/ε)^d is not what the proof yields. The union bound in SM Eqs. (83)–(88) also drops a 2^d factor when Prop. S5 is applied with ε→2ε, which further weakens the constants, but the exponent mismatch is the structural problem.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies recurrence times for unitary evolutions of pure states in finite-dimensional Hilbert spaces. It proves a deterministic upper bound, t_rec(ε) ≤ t_exit(ε)·(4π/ε)^{d−1} (Theorem 1), where t_exit is the first time the evolution leaves the ε-neighborhood of the initial state. It then provides an upper bound on t_exit under a moment condition: t_exit(ε) ≤ ε/√Δ(H²) · (1−ε/ε*)^{-1/2} for ε<ε* (Theorem 2). Combining these gives an explicit bound t_rec(ε) = O(ε^{−(d−2)}) up to constants. The paper also claims a matching generic lower bound for random diagonal Hamiltonians (Theorem 3, Theorem S2): t_rec ≳ (1/ε)^d with high probability. Additional results concern effective support, non-interacting dynamics, unitary channels, and the role of effective dimension.","tokens_in":19706,"tokens_out":30078,"duration_ms":239820,"significance":"Theorems 1 and 2 are clean, self-contained, and appear correct; they give a rigorous deterministic upper bound on recurrence time that improves on Peres's heuristic argument and applies to arbitrary initial states and Hamiltonians. This is a valuable contribution. The claimed generic saturation, however, is a central part of the paper's advertised message ('tight bounds'), and the proof of that claim contains a concrete union-bound error and does not, as written, establish the stated lower bound. If the saturation claim can be repaired or appropriately weakened, the upper-bound part alone would still be a solid contribution; the present version overstates what is proven.","major_comments":[{"comment":"The union bound over grid points applies Proposition S5 with ε replaced by 2ε, but the factor 2^d is dropped. Writing (68) with ε→2ε gives P(D<2ε) ≤ (2π/ε²)(2Cε)^d, not (2π/ε²)(Cε)^d. Consequently the admissible T0 should be reduced by a factor 2^d. With the stated choice of T0 in Eq. (88), the claimed inequality T0 > 20(1/(10000ε))^d fails for the allowed parameter range (e.g., d=50, ε=5×10^{-5}). Theorem 3 and the abstract's 'generically saturated' claim are therefore not established by the proof as written.","section":"Supplemental Material, Eqs. (83)–(88), Theorem S2"},{"comment":"The effective-support reduction uses δ=ε²/4, but the displayed inequality D_tr(ψ0,ψt) ≤ D_tr(ψS0,ψSt)+√(2δ) is too weak: if D_tr(ψS0,ψSt)≤ε/2, it only gives D_tr(ψ0,ψt)≤(1/2+1/√2)ε > ε. With this displayed inequality one needs δ≤ε²/8. The argument can be rescued by instead using the tighter bound D_tr(ψ0,ψt) ≤ √(D_tr² + 2δ), but as written the derivation of (18) is incorrect.","section":"§V, Eq. (17)–(18)"},{"comment":"Even apart from the missing 2^d, the proof's union bound yields a lower-bound horizon T0 of order ε^{3−d} up to constants, i.e. ε^{−(d−3)} for fixed d, whereas the upper bound from Theorems 1+2 is of order ε^{−(d−2)} (or ε^{−(d−1)} before using t_exit≈ε). The lower bound is therefore far smaller than the upper bound in the parameter range covered by Theorem S2, so the phrase 'generically saturated' is not supported by the quantitative estimates. The authors should either strengthen the lower-bound argument or weaken the saturation claim.","section":"Theorem 3 / SM Theorem S2, general tightness claim"}],"minor_comments":[{"comment":"Typos and language issues: 'lack of of rigorous', 'arbirtary', 'theorem' for 'Theorem', and a few other grammatical errors should be corrected.","section":"General"},{"comment":"There is a sign inconsistency in the definition of X_H: Theorem 2 defines X_H := −[H,[H,ψ0]], while Lemma 2 writes X_H = H²ψ0 − 2Hψ0H + ψ0H², which is the positive commutator. The norm bound is unaffected, but the notation should be made consistent.","section":"Eq. (8) and Lemma 2"},{"comment":"The statement 'for ε small enough' is imprecise. The SM theorem requires ε > (40π)^{1/3}(1/2)^{d/3}, so for fixed d, ε cannot be taken arbitrarily small; the parameter regime should be stated clearly in the main text.","section":"Main text, Theorem 3"},{"comment":"The grid spacing is ε and the Lipschitz step uses a threshold 2ε; this is fine, but the choice of T0 in Eq. (88) should be derived after correctly including the 2^d factor from the ε→2ε substitution. The numerical constants in Theorem S2 should be rechecked.","section":"Supplemental Material, proof of Theorem S2"}],"recommendation":"major_revision","confidential_remarks":"The upper-bound part of the paper is sound and publishable. The main obstacle is the lower-bound/tightness claim: the union-bound gap is clear and load-bearing, and the claimed saturation is not currently proven. These issues are likely fixable by correcting constants and/or weakening the claims, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The core of the paper is Theorem 1, and it looks right. The packing-number argument is clean: discretize the torus of relative phases, use metric invariance, and after at most (4π/ε)^{d-1} exit-time steps you get a recurrence. Theorem 2's inverse speed limit is also self-contained; the bound on ∥X_H∥∞ and the fidelity expansion check out, and the ε < ε* condition is stated honestly. This is a real improvement over Peres's heuristic — deterministic, explicit, valid for all Hamiltonians and all pure states — and the extension to subsequent recurrences and to unitary channels is a nice bonus.\n\nThe soft spots are in the advertised tightness claim, and they are not cosmetic. The effective-support improvement sets δ = ε²/4 where the inequality needs δ ≤ ε²/8; that's a one-line fix (take ε²/8) and doesn't touch the main theorem. The lower-bound proof in the Supplemental Material is the real problem. Theorem S2 claims t_rec ≥ 20(1/(10000ε))^d, but the calculation just before gives T0 = (ε^3/2π)(3/(4C))^d ε^{-d} = O(ε^{-(d-3)}). The inequality T0 > 20(1/(10000ε))^d does not follow from the stated assumption ε > (40π)^{1/3}(1/2)^{d/3}; plugging in the constants shows the ratio is tiny for allowed ε. So the proof only establishes a lower bound with a weaker ε-exponent than the upper bound, and the main-text sentence that the bound is 'generically saturated' is unsupported. There's also a dropped 2^d factor in the union bound when Prop. S5 is applied with ε → 2ε; that's a constant, but the exponent mismatch is structural.\n\nI agree with the stress-test note and with the conditional verdict. The upper bound doesn't depend on the lower-bound arguments, so the paper's central contribution stands. But the title and abstract promise tightness, and that part either needs a repaired proof or a weakened claim. It's possible the saturation result is true and just needs a different T0 choice or sharper constants — the general strategy is plausible — but as written, the claim outruns the proof.\n\nWho gets value: people working on quantum recurrence, equilibration times, or quantum speed limits. The upper bound is a clean result worth having. I'd send it to a serious referee with a request to check the SM lower bound carefully; if I were the editor, I'd invite revision rather than accept.","headline":"Theorems 1 and 2 give a clean, deterministic upper bound on recurrence time and are worth publishing; the claimed generic saturation is not established by the Supplemental proof, which yields only ε^{-(d-3)} rather than ε^{-d}.","tokens_in":20098,"tokens_out":6585,"would_cite":true,"duration_ms":407485,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"For any closed quantum system, the first recurrence time is bounded by the escape time times (4π/ε)^{d-1}, and random Hamiltonians saturate it.","keywords":["quantum recurrence","Poincaré recurrence","quantum speed limit","packing number","escape time","effective dimension","random Hamiltonian","unitary dynamics"],"falsifier":"Search numerically for a violation of Eq. (3): for d=4 or 5, take a Hamiltonian with rationally independent eigenvalues and a generic initial state, and compute t_rec(ε) by dense time sampling; if for any ε < ε*/2 the observed first recurrence time exceeds t_exit(ε)(4π/ε)^{d-1}, the upper bound is false. To test the saturation claim, simulate many random diagonal Hamiltonians with uniform eigenvalues and uniform initial state for d=6 and ε=10^{-3}; if the typical t_rec falls below the theorem's claimed (1/ε)^d lower bound (with its pre-factors), the high-probability lower bound is false.","tokens_in":19269,"feed_emoji":"⏳","tokens_out":10086,"duration_ms":82818,"temperature":0.7,"pith_summary":"The paper proves a quantitative, deterministic upper bound on how long a closed quantum system takes to return near its initial state. The first recurrence time t_rec(ε) never exceeds t_exit(ε)(4π/ε)^{d-1}, where d is the Hilbert-space dimension, ε the size of the return neighborhood, and t_exit the time the system first leaves that neighborhood. The authors turn bounding t_exit into an 'inverse quantum speed limit' and show that, below a threshold set by the Hamiltonian's variance and fourth moment, t_exit ≤ ε/√Δ(H²) up to a correction factor. They also show that for random diagonal Hamiltonians with a uniform initial superposition the (1/ε)^d growth is saturated, so the general bound cannot be improved in its dependence on d and ε. Versions with effective support and with non-interacting particle Hamiltonians improve the exponent.","feed_headline":"Recurrence time bounded by escape time times (1/ε)^{d-1}","feed_subtitle":"Rigorous for every Hamiltonian and state; random cases saturate it, so returns are exponentially slow in dimension.","key_machinery":"The load-bearing geometric fact is that unitary evolution is an isometry on the state space. Proposition 1 shows that in any metric space with an isometric flow, the recurrence time is at most the exit time times the packing number N_pack(ε) — the largest number of points mutually ε-apart. Lemma 1 bounds this packing number for the orbit-reachable set T_{ψ0} (the torus of relative phases of the initial state) by (4π/ε)^{d-1}, using an explicit ε-net on the phase coordinates. The escape-time estimate follows from the identity F''(t)=tr(X_H ψ_t) with X_H=−[H,[H,ψ0]], together with the operator bound ∥X_H∥_∞ ≤ Δ(H²)+√Δ(H⁴); integrating F'' over the interval before escape yields the inverse quan","core_discovery":"For any Hamiltonian H and any initial pure state ψ0, the first time the evolved state returns to within trace distance ε of ψ0 (after first leaving that neighborhood) obeys t_rec(ε) ≤ t_exit(ε)·(4π/ε)^{d-1}, with d the Hilbert-space dimension and t_exit the first exit time. Treating t_exit as an inverse quantum speed limit, the paper shows that for ε < ε* = Δ(H²)/(Δ(H²)+√Δ(H⁴)), t_exit(ε) ≤ (ε/√Δ(H²))·(1−ε/ε*)^{-1/2}; hence for ε < ε*/2, t_rec(ε) ≤ (ε/(2√Δ(H²)))·(4π/ε)^{d-1}. For random diagonal Hamiltonians with independent uniform eigenvalues and a uniform superposition initial state, t_rec grows as (1/ε)^d with high probability, so the exponent cannot be improved in general.","pith_inferences":["The packing-number argument is purely metric, so the same t_rec ≤ t_exit·N_pack bound should apply to any distance-invariant dynamics, including unitary channels under the diamond norm (which the paper treats) and possibly classical Hamiltonian flows on compact phase spaces.","For physical, local Hamiltonians the random-diagonal model is not a faithful proxy; the paper leaves open whether local non-commuting Hamiltonians recur faster, and the authors list this as an open problem. A numerical study of small spin chains could probe this gap.","The recurrence-time exponent (1/ε)^d is astronomical for macroscopic d; an immediate consequence is that observing recurrence in a many-body experiment requires engineering an effective low-dimensional subspace, which the effective-support version makes quantitative.","The threshold ε* depends on the initial state's variance and fourth moment; for states close to an eigenstate ε* becomes small, so the proven bound is vacuous for tiny ε — a refined treatment of that regime would be a natural next step."],"forward_implications":["First recurrences occur no later than (exit time)·(4π/ε)^{d-1}; with Theorem 2 this yields the explicit bound t_rec ≤ (ε/(2√Δ(H²)))(4π/ε)^{d-1} for ε < ε*/2.","If ψ0 has δ-effective support d_supp(δ), the exponent improves to d_supp−1, giving t_rec ≤ t_exit·(8π/ε)^{d_supp(ε²/4)−1}; the same argument works for subsequent (kth) recurrences.","For non-interacting n-particle Hamiltonians, the single-particle dimension d replaces the full Hilbert-space dimension: t_rec ≤ t_exit·(4πn/ε)^{d−1}.","Random diagonal Hamiltonians with uniform initial superpositions saturate the (1/ε)^d growth with probability 1−o(1), so the general bound is tight in its dependence on d and ε.","Effective dimension is not the right measure: states with small effective dimension (∼log d) but large effective support (∼d) still have recurrence times exponential in d, so it is the effective support that controls the lower bound."],"fun_headline_variants":["Recurrence time bound: exponential in dimension, saturated randomly","Quantum recurrences: rigorous exponential bound in Hilbert-space dimension","Exponential recurrence bound: t_rec ≤ t_exit (1/ε)^(d−1)","Recurrence time: exponentially slow in Hilbert-space dimension"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The upper bound requires only that the exit time be finite and that ε stay below ε* = Δ(H²)/(Δ(H²)+√Δ(H⁴)), which can fail for states nearly supported on one eigenstate; the tightness claim further assumes that random diagonal Hamiltonians with iid uniform eigenvalues and a uniform eigenbasis superposition represent 'generic' Hamiltonians — a transfer not proven for local or structured Hamiltonians.","fun_headline_variants_meta":{"raw":{"variants":["Recurrence time bound: exponential in dimension, saturated randomly","Quantum recurrences: rigorous exponential bound in Hilbert-space dimension","Exponential recurrence bound: t_rec ≤ t_exit (1/ε)^(d−1)","Recurrence time: exponentially slow in Hilbert-space dimension"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001126,"raw_usage":{"total_tokens":4566,"prompt_tokens":836,"completion_tokens":3730,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":580,"completion_tokens_details":{"reasoning_tokens":3653}},"tokens_in":580,"tokens_out":3730,"duration_ms":26066,"temperature":1.0,"reasoning_tokens":3653,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T06:23:49.273573+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Search numerically for a violation of Eq. (3): for d=4 or 5, take a Hamiltonian with rationally independent eigenvalues and a generic initial state, and compute t_rec(ε) by dense time sampling; if for any ε < ε*/2 the observed first recurrence time exceeds t_exit(ε)(4π/ε)^{d-1}, the upper bound is false. To test the saturation claim, simulate many random diagonal Hamiltonians with uniform eigenvalues and uniform initial state for d=6 and ε=10^{-3}; if the typical t_rec falls below the theorem's claimed (1/ε)^d lower bound (with its pre-factors), the high-probability lower bound is false.","supporting_citations":[],"review_version":2}