{"id":"bb8c0201-7a8c-45ee-8766-30e21da0dd3b","arxiv_id":"2505.00528","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The log-spectrum SFF equals |ζ(β+it)|^2; after removing the dip the β=0 ramp is exactly linear with slope 1 and time-average (π/24)t, giving an O(1) Thouless time.","lead":"This paper shows that the Riemann zeta function, viewed as the spectral form factor of the energy levels E_n = log n, has a linear ramp of slope exactly 1, and that the associated Thouless time is of order 1 as expected for black holes. It also derives a general ramp-exponent formula for L-functions and for power-law spectra such as E_n = n^α.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The O(1) Thouless-time claim is read off from ζ(it)ζ(−it) alone, not from the finite-N truncated log spectrum; cross terms in Eq. (3.9) are never bounded, so the identification of the RZF ramp onset with the physical ramp onset is unproven.","rationale":"The paper's main analytic result, the slope-1 ramp of the time-averaged |ζ(it)|^2, is derived correctly using the functional equation and standard mean-value theorems. The reader correctly identifies the weakest step: the leap from this infinite-zeta quantity to the finite-N log-spectrum SFF and its Thouless time. My stress-test sharpens this concern in two ways. First, the Euler–Maclaurin decomposition (3.9) contains a cross term between the classical term N^{1−s}/(1−s) and ζ(s); its envelope decays only as N/t^{1/2}, so it dominates |ζ|^2 until t ~ N^{2/3}. The paper does not estimate this term, and Figure 3 is not a substitute for an analytic bound. Second, the O(1) crossing in Figure 8 is the intersection of ζ(it)ζ(−it) with an extrapolated asymptotic line; the asymptotics used to derive the ramp are not valid at the intersection point, so the 'onset' is a feature of the infinite-zeta function rather than a measured property of the truncated spectrum. These points do not invalidate the slope-1 result, but they make the headline O(1) Thouless-time claim conditional on an unproven identification. The requested test—extracting the ramp onset from the actual truncated SFF for several N and checking the cross-term average—would settle the matter directly. I therefore keep the reader's CONDITIONAL verdict unchanged.","tokens_in":21442,"tokens_out":19252,"duration_ms":201078,"concrete_test":"For N = 10^3, 10^4, 10^5, compute the moving-averaged full SFF g_N(t) = |∑_{n≤N} n^{−it}|^2 and its dip-removed part g_N(t) − N^2/(1+t^2). Extract the first time t_*(N) after the local minimum at which the dip-removed part crosses and stays within a 10% band of the linear ramp (π/24)t. Also compute the time average of the cross term 2 Re[N^{1−it}ζ(−it)/(1−it)] over [1,T] for T = 10 and T = 100 and compare it with (π/24)T. If t_*(N) scales as N^{2/3}, or if the cross-term average is comparable to the ramp, the O(1) Thouless time is an artifact of the RZF-only definition.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The derivation of ⟨|ζ(it)|^2⟩ = (π/24)t (Eq. 3.34) is mathematically sound; the load-bearing step is the identification of this infinite-zeta ramp with the ramp of the finite-N truncated log spectrum and the extraction of the Thouless time from Figure 8. In the Euler–Maclaurin decomposition (3.9), Z_N(it) = N^{1−it}/(1−it) + ζ(it) + ..., so the full SFF contains a cross term 2 Re[N^{1−it}ζ(−it)/(1−it)] whose envelope is ~ N |ζ(it)| / t ~ const·N t^{−1/2}. This exceeds |ζ(it)|^2 ~ (π/24)t until t ≳ N^{2/3}. Thus the |ζ|^2 ramp dominates the actual truncated SFF only at an N-dependent time, whereas Figure 8 defines the Thouless time as the first intersection of ζ(it)ζ(−it) with the line (π/12)t, which occurs at O(1) and is independent of N. Moreover, that intersection lies in the transient regime where the asymptotic approximations (3.25) and the mean-value theorem used to derive the linear ramp are not controlled. The paper asserts, rather than proves, that the RZF ramp onset equals the truncated-spectrum ramp onset, and no estimate of the cross terms is given; this is the gap on which the O(1) black-hole claim rests.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the spectral form factor (SFF) of the logarithmic spectrum E_n = log n. It uses the Euler–Maclaurin decomposition Z_N(s) = N^{1-s}/(1-s) + ζ(s) + ... to separate a classical dip from quantum corrections, derives via the functional equation and mean-value theorems that the time-averaged quantity ⟨|ζ(it)|^2⟩ behaves as (π/24)t, and from this claims that the Riemann zeta function ramp has slope exactly 1. The authors then identify the RZF as the 'full ramp after removal of the dip', read an O(1) Thouless time from the first intersection of ζ(it)ζ(-it) with the ramp line, and argue that this matches black-hole expectations. They generalize the ramp exponent to L-functions, obtaining exponent 2d(w+1)/2 at β=0, and use Poisson resummation to derive the slope 1/(1-α) for power-law spectra E_n = n^α.","tokens_in":21724,"tokens_out":20221,"duration_ms":203242,"significance":"The derivation of the RZF ramp, Eqs. (3.23)-(3.34), is a clean and mostly sound application of standard mean-value theorems and Stirling approximation, and the numerical fits in Figure 6 support the constant π/24. The L-function scaling formula and the Poisson-resummation treatment of n^α spectra are concrete, falsifiable statements and give useful analytic control where previous work was numerical. If the identification of the RZF ramp with the finite-N truncated log-spectrum ramp were established, the O(1) Thouless time would be a significant and surprising result for black-hole toy models. At present, however, that identification is asserted rather than proved, and several of the paper's strongest claims depend on it.","major_comments":[{"comment":"The identification of ζ(it)ζ(-it) with the ramp of the finite-N truncated log spectrum is not established. Expanding |Z_N(it)|^2 via Eq. (3.9) produces, in addition to |ζ(it)|^2, the cross term 2 Re[N^{1-it} ζ(-it)/(1-it)]; its envelope is of order N |ζ(it)|/t ~ N t^{-1/2}, which exceeds |ζ(it)|^2 ~ t until t ≳ N^{2/3}. No estimate or cancellation argument for these cross terms is given, and Figure 3, which compares raw curves, does not prove that the RZF ramp is the ramp of the truncated SFF. Because the O(1), N-independent Thouless time in Section 4 is read directly from ζ(it)ζ(-it), the central physical claim rests on this unproven identification.","section":"3.3, 4; Eq. (3.9)"},{"comment":"The O(1) Thouless time is read in a regime where the asymptotic approximations used to derive the ramp are not controlled. Eq. (3.25) is a large-|t| Stirling approximation and Eq. (3.30) is a large-T mean-value estimate; neither controls the pointwise behavior of ζ(it)ζ(-it) at t = O(1). The first intersection in Figure 8 therefore lies in the transient regime and is not a derived onset time. In addition, the yellow line in Figure 8 is (π/12)t, whereas Eq. (3.34) gives the cumulative time average (π/24)T; the factor-of-two discrepancy and the distinction between a local moving average and a cumulative average need to be clarified, since the first-intersection time depends on which curve is used.","section":"4; Fig. 8"},{"comment":"The definition of the RZF as the 'full ramp after removal of the dip' makes the N-independence of the ramp, and hence the O(1) onset, partly true by construction. A statement about the actual truncated spectrum requires an independent check of the subtracted quantity |Z_N - Z_cl|^2 or of the finite-N SFF after dip removal; Figures 1-3 compare raw curves and do not isolate the claimed quantum contribution. Without such a check, the O(1) Thouless time is a property of the definition rather than a demonstrated property of the log spectrum.","section":"3.3, 4"}],"minor_comments":[{"comment":"Equation (3.33) writes I(T) with a lim_{T→∞} symbol in front of an expression that still depends on T; this should read I(T) = ∫_1^T ζ(it)ζ(-it) dt, and Eq. (3.34) should distinguish the cumulative average ⟨g⟩_T from the local moving-average ramp.","section":"3.5.4; Eq. (3.33)"},{"comment":"The expression 't^{2d w+1/2}' following Eq. (5.14) is ambiguous; it should read t^{2d(w+1)/2}.","section":"5; Eq. (5.14)"},{"comment":"For 1-β > 1/2 the error term in the mean-value estimate is not O(t^{2β}); a more careful statement would be O(t^{2β-1} log t) (or O(1) at β=0). The leading term is unaffected, but the formula as written is inaccurate.","section":"3.5.4; Eq. (3.30)"},{"comment":"The heading of the acknowledgments section contains a typo: 'Acknolwedgments' should be 'Acknowledgments'.","section":"8"},{"comment":"The caption states that for β < 1/2 the function displays a persistent ramp proportional to t^{1-2β}; this is the local-average behavior, whereas Eq. (3.32) is a cumulative integral. The caption should state this distinction to avoid confusion.","section":"Figure 4 caption"}],"recommendation":"major_revision","confidential_remarks":"The mathematical core of the paper — the mean-value derivation of the RZF ramp and the scaling formulas for L-functions and n^α spectra — is sound and likely publishable after revision. The advertised black-hole conclusion, however, is a load-bearing interpretive claim that is not supported by the present arguments. The authors should either supply a treatment of the cross terms in Eq. (3.9) and a valid finite-N check of the ramp onset, or explicitly withdraw the O(1) Thouless-time claim from the abstract and Section 4."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick read on 2505.00528. The paper earns its keep on the mathematics: using the functional equation and standard mean-value theorems, it shows the time average of |ζ(it)|^2 grows as (π/24)t, so the untruncated zeta SFF has a slope-one ramp. The constant matches their numerics, and the L-function generalization to ramp exponent 2d × Re(central line) is a clean, testable statement that appears to hold in their plots. I also give them the Poisson-resummation calculation for n^α spectra; that's a nice bit that explains a numerical guess from their earlier paper.\n\nThe soft spot is the Thouless time. The headline claim—that the log spectrum has O(1) Thouless time—is read off from ζ(it)ζ(−it) alone, not from the finite-N truncated SFF. In Eq. (3.9) the SFF is split as N^{1−s}/(1−s) + ζ(s) + ..., and the paper simply asserts that after the classical dip decays, ζ(s) is the ramp. But the cross term 2 Re[N^{1−it}ζ(−it)/(1−it)] is never bounded. A quick estimate gives an envelope ~ N t^{−1/2}, which dominates |ζ(it)|^2 ~ (π/24)t until t ≳ N^{2/3}. If that is right, the ramp of the actual truncated spectrum sets in at an N-dependent time, not at the O(1) intersection of ζ(it)ζ(−it) with (π/12)t plotted in Figure 8. The paper's own 'definition' of the zeta SFF as the full ramp after dip removal bypasses the question rather than answering it. So the O(1) statement is true for the zeta function by construction, and unproven for the log spectrum.\n\nI don't think this kills the paper. The ramp derivation stands on its own and is worth having. But the 'first model with demonstrably O(1) Thouless time' framing is too strong. The authors should be asked to either bound the cross terms or extract the ramp onset from the truncated SFF for several N; that would turn a suggestive observation into a result. The L-function section is more speculative but reasonable.\n\nWho is this for? People working on spectral form factors of deterministic spectra and black-hole chaos. It deserves a serious referee—the math is reproducible and the claims are concrete—but I would not cite the O(1) part as established. My recommendation: send to peer review, with the Thouless-time identification as the central requested revision.","headline":"Solid analytic derivation of the zeta ramp, but the O(1) Thouless time rests on an unproven identification of the infinite-zeta ramp with the finite-N truncated spectrum.","tokens_in":22320,"tokens_out":2356,"would_cite":true,"duration_ms":23428,"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":"The log-spectrum SFF is exactly the Riemann zeta ramp, with slope one and an O(1) Thouless time.","keywords":["spectral form factor","Riemann zeta function","logarithmic spectrum","Thouless time","black hole normal modes","L-functions","dip-ramp-plateau","Poisson resummation"],"falsifier":"Compute $I(T)=\\int_1^T|\\zeta(it)|^2\\,dt$ at very large $T$ with controlled numerical error; if $I(T)/T^2$ does not approach $\\pi/24\\approx0.1309$, or if the ramp onset extracted from the truncated SFF after subtracting the classical dip visibly moves with $N$, the central claim fails.","tokens_in":21201,"feed_emoji":"🕳️","tokens_out":12537,"duration_ms":118231,"temperature":0.7,"pith_summary":"This paper proves that the spectral form factor (SFF) of the deterministic spectrum $E_n = \\log n$, with the early-time classical dip removed, is exactly the Riemann zeta function, and that its time-averaged ramp is linear with slope one: $\\langle |\\zeta(it)|^2 \\rangle = \\frac{\\pi}{24} t$. The proof runs through the Euler-Maclaurin decomposition of the truncated partition function together with the functional equation $\\zeta(s) = \\chi(s)\\zeta(1-s)$, whose gamma factors supply the growth $t^{1-2\\beta}$ for $0 \\le \\beta < 1/2$. The same reasoning fixes the onset of the ramp, giving a Thouless time of $\\mathcal{O}(1)$ that is independent of the truncation size $N$, the behaviour expected of black-hole microstates. The authors extend the mechanism to $L$-functions, where the ramp exponent is twice the degree times the real part of the critical line, and they derive the ramp slope $1/(1-\\alpha)$ for $E_n = n^\\alpha$ via Poisson resummation. A reader should care because it shows that deterministic spectra, not only random matrices, can reproduce the black-hole spectral signatures from a first-principles calculation.","feed_headline":"Zeta ramp is exactly linear, with O(1) black-hole Thouless time","feed_subtitle":"Riemann's functional equation fixes the zeta ramp's slope, constant, and O(1) onset.","key_machinery":"The load-bearing object is the Euler-Maclaurin split $Z_N(s) \\approx N^{1-s}/(1-s) + \\zeta(s) + \\frac12 N^{-s} - \\frac{s}{12}N^{-1-s}$ (Eq. (3.9)), which separates the classical dip contribution $N^{1-s}/(1-s)$ from the quantum ramp $\\zeta(s)$. The second engine is the reflection property of the functional equation $\\zeta(s)=\\chi(s)\\zeta(1-s)$: Stirling's approximation gives $|\\chi(\\beta+it)|\\approx (t/2\\pi)^{1/2-\\beta}$, converting a mean-value theorem for $\\Re(s)>1/2$ into the ramp $t^{1-2\\beta}$ for $\\beta<1/2$. For $L$-functions, the same gamma-ratio reflection yields the general ramp exponent $2d\\,\\mathrm{Re}(\\text{critical line})$, where $d$ is the degree of the $L$-function and the critical line is the symmetry axis of the functional equation, located at real part $(w+1)/2$. Poisson resummation with a saddle-point evaluation supplies the slope $1/(1-\\alpha)$ for power-law spectra $E_n=n^\\alpha$.","core_discovery":"On its own terms, the central discovery is that Riemann's analytic continuation is the quantum correction to a truncated log spectrum, so the zeta function is the full ramp after removal of the dip. Concretely, the paper establishes Eq. (3.34): for $\\beta=0$, $\\langle |\\zeta(it)|^2\\rangle = (\\pi/24)t$, so the SFF ramp has slope exactly 1 on a log-log plot, with the coefficient fixed rather than fitted. For $0\\le\\beta<1/2$ the ramp grows as $t^{1-2\\beta}$ up to constants; at $\\beta=1/2$ it grows logarithmically; and for $\\beta>1/2$ it saturates to the plateau $\\zeta(2\\beta)$. The $s=1$ pole is read as a Hagedorn transition, and the onset of the zeta ramp, read off from its intersection with the classical dip, gives a Thouless time that is $\\mathcal{O}(1)$ and $N$-independent, matching the black-hole expectation and distinguishing the log spectrum from other black-hole toy models.","pith_inferences":["One can test the $\\mathcal{O}(1)$ onset independently by constructing the connected, unfolded SFF of the truncated log spectrum with a standard filtering method rather than reading the intersection with the classical dip; if the two onsets disagree, the claimed Thouless time is definition-dependent.","The gamma-reflection mechanism suggests a sharper criterion than the paper proves: any Selberg-class Dirichlet series with real spectral parameters should show a ramp with exponent fixed by degree and critical line, so checking a non-self-dual or higher-degree $L$-function with known data would either extend or test the claimed universality.","Because the ramp is deterministic, the fluctuations around $\\pi t/24$ are arithmetic rather than stochastic; one could ask whether their time-averaged moments obey random-matrix universality, which the paper leaves open."],"forward_implications":["The SFF of the $\\log n$ spectrum has an exact linear ramp of slope 1 with coefficient $\\pi/24$; no ensemble averaging is needed because the ramp is a deterministic property of the arithmetic sequence.","The log spectrum has an $\\mathcal{O}(1)$, $N$-independent Thouless time; this is the first black-hole-inspired toy model, apart from random matrices, for which that has been demonstrated (the SYK model scales as $\\mathcal{O}(\\log N)$ and fuzzball constructions as $\\mathcal{O}(N^{\\#})$).","For $L$-functions, the nontruncated SFF at $\\beta=0$ ramps forever as $t^{2d\\,\\mathrm{Re}(\\text{critical line})}$, and truncating the Dirichlet series restores a plateau; the functional equation, not the zeros, controls the ramp.","For power-law spectra $E_n=n^\\alpha$ with $0<\\alpha<1$, the ramp slope $1/(1-\\alpha)$ follows analytically from a saddle-point estimate of Poisson-resummed terms, confirming the earlier numerical guess."],"supporting_citations":[{"why":"Earlier paper establishing a linear ramp in the SFF of black hole normal modes; supplies the black-hole-motivated baseline the log spectrum is meant to reproduce.","marker":"[3]"},{"why":"Reported numerical evidence that the logn spectrum has a slope-one ramp; the present paper's analytic proof targets that observation.","marker":"[5]"},{"why":"Defines the RMT dip-ramp-plateau structure and the linear ramp expected of black holes; the comparison standard for the log-spectrum SFF.","marker":"[16]"},{"why":"Argues that black hole microstates should have an O(1) Thouless time; the target the paper's zeta-ramp onset is compared against.","marker":"[17]"},{"why":"Introduces the primon gas picture, mapping the log spectrum onto the Riemann zeta function; the physical map used throughout.","marker":"[20]"},{"why":"Standard treatise providing the Euler-Maclaurin formula, functional equation, and mean-value theorems from which Eq. (3.34) is derived.","marker":"[25]"},{"why":"Supplies the second-moment results and functional-equation conventions used to derive the general L-function ramp exponent.","marker":"[30]"},{"why":"Provides the mean-value bound used to treat L(1-w/2-β-it) as O(1) in the L-function ramp derivation.","marker":"[31]"},{"why":"The standard database of concrete L-functions whose numerical values are plotted for Dirichlet, Dedekind, and elliptic-curve examples.","marker":"[32]"}],"fun_headline_variants":["Exact linear ramp from Riemann zeta, O(1) black-hole time","Zeta ramp slope exactly 1, black-hole Thouless time O(1)","Riemann zeta ramp: slope 1, black-hole time O(1)","Zeta function gives exact ramp, black-hole Thouless time O(1)","Black-hole zeta ramp: exact slope 1, O(1) time"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes that the zeta term in Eq. (3.9) is the full ramp after the dip is removed, so cross terms between $N^{1-s}/(1-s)$ and $\\zeta(s)$ must be negligible in the ramp window and the onset time read from $\\zeta(it)\\zeta(-it)$ must equal the onset for the actual truncated spectrum; neither is proved.","fun_headline_variants_meta":{"raw":{"variants":["Exact linear ramp from Riemann zeta, O(1) black-hole time","Zeta ramp slope exactly 1, black-hole Thouless time O(1)","Riemann zeta ramp: slope 1, black-hole time O(1)","Zeta function gives exact ramp, black-hole Thouless time O(1)","Black-hole zeta ramp: exact slope 1, O(1) time"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000321,"raw_usage":{"total_tokens":1888,"prompt_tokens":1104,"completion_tokens":784,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":720,"completion_tokens_details":{"reasoning_tokens":679}},"tokens_in":720,"tokens_out":784,"duration_ms":6989,"temperature":1.0,"reasoning_tokens":679,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T04:42:14.242875+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute $I(T)=\\int_1^T|\\zeta(it)|^2\\,dt$ at very large $T$ with controlled numerical error; if $I(T)/T^2$ does not approach $\\pi/24\\approx0.1309$, or if the ramp onset extracted from the truncated SFF after subtracting the classical dip visibly moves with $N$, the central claim fails.","supporting_citations":[{"cited_title":"Julia,STATISTICAL THEORY OF NUMBERS, inLes Houches School of Theoretical Physics: Number Theory and Physics, 9, 1989","cited_arxiv_id":null,"evidence_quote":"Introduces the primon gas picture, mapping the log spectrum onto the Riemann zeta function; the physical map used throughout."},{"cited_title":"Titchmarsh,The Theory of the Riemann Zeta-Function, Oxford University Press, second edition, revised by d","cited_arxiv_id":null,"evidence_quote":"Standard treatise providing the Euler-Maclaurin formula, functional equation, and mean-value theorems from which Eq. (3.34) is derived."},{"cited_title":"Iwaniec and E","cited_arxiv_id":null,"evidence_quote":"Supplies the second-moment results and functional-equation conventions used to derive the general L-function ramp exponent."},{"cited_title":"Montgomery and R.C","cited_arxiv_id":null,"evidence_quote":"Provides the mean-value bound used to treat L(1-w/2-β-it) as O(1) in the L-function ramp derivation."},{"cited_title":"The L-functions and modular forms database","cited_arxiv_id":null,"evidence_quote":"The standard database of concrete L-functions whose numerical values are plotted for Dirichlet, Dedekind, and elliptic-curve examples."}],"review_version":1}