{"id":"6cb1f6bc-c0b3-4ff0-920e-cebb1316cb39","arxiv_id":"1908.06920","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"The reflection time difference, the energy derivative of the phase difference between two reflection amplitudes, is a signed sum of Lorentzians centered at the S-matrix zeroes and can be read off from the unitary deficit under weak absorption.","lead":"This paper defines a new observable, the reflection time difference, whose energy dependence exposes the complex zeros of the scattering matrix. It gives a concrete recipe to extract those zeros from reflection measurements in weakly absorbing chaotic cavities.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. (15) is nonuniform in the zeroes: for |Im Z_n| less than or comparable to epsilon the exact unitary deficit differs from -2 epsilon delta T by a factor of order (Im Z_n/epsilon)^2, so near-real zeroes are extracted with Im Z about epsilon; epsilon much less than Delta is not sufficient.","rationale":"The reader's verdict is CONDITIONAL with high confidence, and I agree with that overall judgement. The central mathematical claim, Eq. (13), is sound: the ratio R1/R2 is a Blaschke-type product over the eigenvalues of H + i(Gamma_1 - Gamma_2), and the log derivative gives the sign-weighted sum of Lorentzians. My stress-test found no error in that derivation. The load-bearing concern is in the proposed experimental extraction via Eq. (15). The reader identified the weak-loss approximation as the weakest assumption, mentioning nonuniformity and unknown epsilon. My check sharpens this: even for perfectly uniform losses with epsilon much less than Delta, the first-order expansion is nonuniform in the zero positions. Zeroes with |Im z| less than or comparable to epsilon are not faithfully represented by the unitary deficit; their apparent widths and amplitudes are set by epsilon rather than by their true imaginary parts. This is especially relevant for the CPA motivation, where zeroes approach the real axis. Because the paper is a short proposal and the central identity is correct, the appropriate verdict remains CONDITIONAL rather than REJECT; the paper should either state the additional resolvability condition epsilon much less than |Im z_n| or quantify the bias for near-real zeroes. Therefore I do not change the reader's verdict, but I would strengthen the condition attached to Eq. (15).","tokens_in":8800,"tokens_out":26063,"duration_ms":262989,"concrete_test":"Using the same N=400 model and epsilon=1e-5 as in Fig. 1, construct the exact ratio R1/R2 from its product representation, but replace one zero by z_j = E_j + i*1e-6 while keeping all other zeroes fixed. Compute D(lambda) = -(1/(2 epsilon)) log|R1/R2| at lambda + i epsilon, then run the same harmonic inversion routine used for the red x's in Fig. 1. If the recovered imaginary part is approximately 1e-5 instead of 1e-6, the nonuniformity of the epsilon expansion is confirmed and the extraction bias for near-real zeroes is real; if it recovers 1e-6, the objection fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The algebraic identity Eq. (13) is correct: R1/R2 = prod_n (lambda - z_n)/(lambda - z_n*), so delta T = sum Im z_n / |lambda - z_n|^2. The problem is the weak-loss extraction relation, Eq. (15). From the product representation, the exact magnitude at lambda + i epsilon is log|R1/R2| = (1/2) sum_n log[(a_n^2 + (epsilon - b_n)^2)/(a_n^2 + (epsilon + b_n)^2)], with a_n = lambda - Re z_n and b_n = Im z_n. Expanding this in epsilon gives -2 epsilon sum_n b_n/(a_n^2 + b_n^2) only when epsilon is small compared with |lambda - z_n| for every n. The paper states the condition epsilon much less than Delta, but this is not enough: a zero can have b_n = Im z_n much smaller than epsilon even when epsilon is much smaller than Delta. For such a zero, at a_n = 0 the exact log magnitude is log[(epsilon - b)/(epsilon + b)] approximately -2b/epsilon, whereas the first-order expression used for Eq. (15) would give -2 epsilon/b; the exact contribution is smaller by a factor (b/epsilon)^2. Thus the signal D(lambda) = -(1/(2 epsilon)) log|R1/R2| contains, for a zero with b << epsilon, a Lorentzian of width about epsilon and amplitude b/epsilon^2, not the true Lorentzian b/(a^2 + b^2). Harmonic inversion of D will therefore return Im z approximately epsilon rather than the true b for every zero with |Im z| less than or comparable to epsilon. This is not a peripheral concern: the paper motivates the construction with coherent perfect absorption, where zeroes cross the real axis, and those are exactly the zeroes with small Im z. The numerical illustration uses epsilon = 1e-5 with simulated zeroes whose typical |Im z| is much larger, so Fig. 1 does not probe this regime. The claim that Eq. (15) provides a route to extract zero positions is therefore incomplete without an additional condition epsilon much less than the smallest |Im z_n| to be resolved.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"Fyodorov considers the two-channel Heidelberg model of chaotic resonance scattering and defines the Reflection Time Difference δT(λ) = (1/2)∂(φ1−φ2)/∂λ, where φ1, φ2 are the phases of the two diagonal reflection amplitudes. Using the determinant identity Eq. (12), he shows that δT equals a sum over the complex zeroes Z_n of R1 of Im Z_n / |λ−Z_n|^2, making δT the zero-analogue of the Wigner time delay. He further proposes that under weak uniform absorption the ratio |R1/R2| at λ+iε obeys |R1/R2| ≈ exp(−2ε δT(λ)), enabling extraction of the zeroes from the unitary deficit via harmonic inversion, and he presents a numerical simulation with N=400 zeroes.","tokens_in":9209,"tokens_out":12868,"duration_ms":138061,"significance":"The central algebraic content is correct and valuable: Eq. (12) is a parameter-free identity within the model, and Eq. (13) provides an exact, sign-weighted Lorentzian representation of δT analogous to the Wigner delay, with no fitted constants. The integral identity Eq. (14) is a clean topological consequence. These identifications are the paper's main strength and justify publication of the theoretical construction. The proposed experimental extraction via Eq. (15), however, is not uniformly valid and needs substantial qualification; the conceptual advance survives, but the practical protocol as stated is not reliable for the near-real zeroes that motivate the coherent-perfect-absorption discussion.","major_comments":[{"comment":"The weak-loss relation (15) is not uniformly valid, and the stated condition ε ≪ Δ is insufficient. For a zero Z_n = x + ib with 0 < b ≪ ε, the exact product representation gives, at λ = x + iε, |R1/R2| = (ε−b)/(ε+b) ≈ 1 − 2b/ε, so the quantity D(λ) = −(2ε)^{-1} ln|R1/R2| equals ≈ b/ε^2, whereas the first-order formula (15) would give δT(x) ≈ 1/b; the true signal is smaller by a factor (b/ε)^2. Equivalently, D(λ) near such a zero is a Lorentzian of width ε and amplitude 2b/ε^2 rather than the true Lorentzian b/(a^2+b^2), so harmonic inversion will return Im Z ≈ ε instead of the true b. For b = 0, which is exactly the CPA case motivating the paper, D vanishes identically at the real-axis point and the method is blind. The condition needed for (15) is ε ≪ min_n |Im Z_n|, not ε ≪ Δ; since Im Z_n can be arbitrarily small, no single ε works for all zeroes. This point is load-bearing for the proposed extraction protocol and must be corrected or qualified.","section":"Eq. (15)"},{"comment":"The numerical demonstration does not test the regime where the approximation fails. With ε = 10^{-5} and the stated couplings γ1 = 0.1, γ2 = 0.05, the imaginary parts of the zeroes are likely of order 10^{-2}–10^{-1}, so the figure only verifies the first-order regime |Im Z_n| ≫ ε. Since Eq. (15) fails precisely when |Im Z_n| ≲ ε, the figure cannot support the general claim that harmonic inversion extracts zeroes from the unitary deficit. The simulation should be repeated with zeroes deliberately close to the real axis (for example, by choosing γ1 ≈ γ2 or by post-selecting realizations), and the recovered imaginary parts compared with the true values; the b = 0 case should be discussed explicitly. The paper should also state that the test is a self-consistency check within the same model, not an independent test of the extraction procedure.","section":"Fig. 1 and numerical test"}],"minor_comments":[{"comment":"The term 'S-matrix zeroes' is used for zeroes of the reflection amplitude R1; in the two-channel unitary case R1 = 0 does not make the full S-matrix singular. Please define the terminology precisely at the start.","section":"Abstract and Introduction"},{"comment":"The integral identity assumes that no zero of R1 lies exactly on the real axis within [λ1, λ2]; at a CPA crossing the numbers N± are not well-defined. This should be stated.","section":"Eq. (14)"},{"comment":"The axes are unlabeled, and the caption does not specify the random-matrix ensemble (GUE or GOE), the distribution of the coupling vectors, or the mean level spacing Δ used to set ε. Please add these details so the numerical test is reproducible.","section":"Fig. 1"},{"comment":"The sentence reporting that 'poles were extracted' should say 'zeroes were extracted', since the objects being recovered are the zeroes of R1.","section":"Numerical section"},{"comment":"Ref. [50] contains a typo in the author name ('M. K´’uhmayer1' should be 'M. K¨uhmayer'); please correct.","section":"References"},{"comment":"The central determinant identity is only justified as 'straightforward algebraic manipulations'; a short derivation in an appendix would make the paper self-contained, as this identity underlies the entire construction.","section":"Eq. (12)"}],"recommendation":"major_revision","confidential_remarks":"The paper is a short proceedings-style contribution. The main identity, Eq. (13), is correct and worthwhile, but the proposed loss-based extraction via Eq. (15) needs revision: the nonuniformity near zeroes with |Im Z_n| ≲ ε is a genuine gap, and the current numerical test does not probe it. I recommend major revision rather than rejection, because the issue is local and should be fixable with a corrected condition and additional numerical checks."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The short version: the paper's central identity is right, and the notion of reflection time difference is genuinely new; but the loss-based extraction formula (15) has a uniformity problem that the paper understates, and it happens exactly in the CPA regime that motivates the work.\n\nThe algebraic result is clean. For a two-channel unitary S-matrix in the Heidelberg model, R1/R2 equals a ratio of determinants whose poles and zeros are the zeros of R1 and R2 (conjugates). Differentiating the phase gives δT as a sum of signed Lorentzians with weights Im Z_n. That is a faithful analogue of Wigner time delay for poles, with signs reflecting that zeros live on both sides of the real axis. I checked the determinant identity by hand; it is correct. The integrated result (14) is also nice.\n\nWhere the paper gets soft is Eq. (15). The expansion of log|R1/R2| at λ+iε to first order in ε requires ε to be small compared with |λ-Z_n| for every zero, not just ε≪Δ. If some zero has |Im Z_n|≪ε, its exact contribution to the unitary deficit is of order b/ε (in log magnitude), not 2εb/((λ-Re Z)^2+b^2). Concretely, the signal D(λ)=-(1/(2ε)) log|R1/R2| replaces a true Lorentzian of width b and area π with a Lorentzian of width ε and area πb/ε. Harmonic inversion will therefore report Im Z≈ε for every zero with b<ε, or miss it entirely. The paper's own motivation is coherent perfect absorption, where zeros cross the real axis and b→0, so this is not a peripheral corner case. A correct statement would add the condition ε≪min|Im Z_n| over the zeros one wants to resolve, with the obvious caveat that experimental noise may prevent taking ε that small.\n\nThe other weakness is the numerical illustration: Fig.1 has no code, no data, and no quantitative error metric, and it uses the same model to generate and to extract, so it is a self-consistency check, not an independent test. A quick simulation with a zero placed at b=10^-7 and ε=10^-5 would make the point.\n\nNone of this kills the paper. The exact identity (13) stands on its own, and the signed-Lorentzian representation is likely to be useful. The extraction proposal needs a sharper statement of its validity range. I'd send it to a competent referee, and I'd want the referee to check the uniformity condition on Eq. (15) and ask for an honest numerical test in the b<ε regime.","headline":"The reflection-time-difference identity is correct and new, but the loss-based extraction formula (15) fails for near-real zeros, which is exactly the CPA-motivated regime.","tokens_in":9808,"tokens_out":4537,"would_cite":true,"duration_ms":45360,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81Q50","82B44"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper introduces the reflection time difference, a measurable quantity that plays for S-matrix zeroes the role the Wigner time delay plays for poles, and shows how to extract complex zero positions from weak absorption data.","keywords":["S-matrix zeroes","reflection time difference","Wigner time delay","chaotic scattering","Heidelberg model","coherent perfect absorption","microwave billiards","random matrix theory"],"falsifier":"Measure the two reflection coefficients of a real two-channel microwave billiard with a small tunable absorber; extract δT independently from the phase derivative via Eq. (13) and from the modulus ratio via Eq. (15), and compare recovered zero positions with direct numerical diagonalization of the effective Hamiltonian for the same geometry. Agreement would confirm the relations; a systematic mismatch between the two extractions at absorber strengths below the CPA threshold would falsify them.","tokens_in":8592,"feed_emoji":"⚛️","tokens_out":5905,"duration_ms":56792,"temperature":0.7,"pith_summary":"The paper introduces the reflection time difference δT(λ), half the energy derivative of the phase difference between the two reflection coefficients of a two-channel scattering system, and shows it is a sum of sign-weighted Lorentzians centered at the complex zeroes of the reflection amplitude. This makes δT the zero analogue of the Wigner time delay, which is a sum over resonance poles. The paper further shows that in the presence of weak uniform losses, the ratio of reflection moduli satisfies |R1/R2| ≈ exp(−2εδT), so δT can be extracted directly from measured unitary deficit. A numerical test with a 400-resonance Heidelberg model recovers zero positions from this relation.","feed_headline":"Reflection time difference exposes S-matrix zeroes","feed_subtitle":"A measurable quantity reads complex zeroes as Lorentzian peaks, mirroring how time delay reveals resonances.","key_machinery":"The key object is the ratio of reflection amplitudes R1/R2 expressed via the Heidelberg model as a ratio of characteristic polynomials, det(λ1N − H_N − iΓ1 + iΓ2)/det(λ1N − H_N + iΓ1 − iΓ2). The log-derivative of this ratio is the reflection time difference δT, and the weak-absorption identity |R1/R2|_{λ+iε} ≈ $e^{{−2εδT}}$ converts δT into an experimentally measurable quantity. The spectral representation of δT as Σ_n Im Z_n/((λ−Re Z_n)^2 + (Im Z_n)^2) is the mechanism that makes zero positions directly visible as Lorentzian peaks.","core_discovery":"In a two-channel flux-conserving system, the ratio R1/R2 is unimodular on the real axis and equal to a ratio of two spectral determinants; its complex-energy dependence is governed by the complex zeroes of R1. The reflection time difference δT(λ) := (1/2) ∂(φ1−φ2)/∂λ is the log-derivative of that ratio and therefore decomposes into a sum of sign-weighted Lorentzians, one for each zero, with positive or negative weights depending on whether the zero lies above or below the real axis. Because a small uniform absorption shifts λ to λ+iε and makes the modulus ratio |R1(λ+iε)/R2(λ+iε)| ≈ $e^{{−2εδT(λ)}}$, the same quantity can be read from reflection amplitudes in realistic lossy cavities. The paper demonstrates numerically that harmonic inversion of δT yields the true zero positions.","pith_inferences":["If the relation survives nonuniform losses, δT could become a general tool for imaging complex spectral points of non-Hermitian effective Hamiltonians in any two-port wave experiment.","The same construction may apply to transmission or other off-diagonal S-matrix entries, providing new spectral functions sensitive to zeros rather than poles.","One could test the prediction by comparing measured δT against direct numerical solution of the wave equation in a microwave billiard with known geometry, without any random-matrix assumption.","Since coherent perfect absorption occurs when a zero crosses the real axis, δT could serve as a pre-threshold diagnostic for anti-lasing."],"forward_implications":["Positions of S-matrix zeroes can be extracted from scattering measurements in weakly absorbing two-channel cavities, without tuning to the CPA condition.","Harmonic inversion of δT(λ) recovers the complex zero set; the paper's numerical simulation with N = 400 confirms the two complementary routes agree.","The energy integral of δT over a wide interval equals π(N+−N−), giving a direct count difference between zeroes above and below the real axis.","For more than two channels, the construction generalizes by replacing R1,R2 with determinants of reflection sub-blocks.","The identity gives a practical way to test random-matrix predictions for zero distributions in the complex plane."],"supporting_citations":[{"why":"Supplies the Heidelberg random-matrix model of resonance scattering used for the entire derivation.","marker":"[6]"},{"why":"Establishes the random-matrix description of S-matrix zeroes under weak localized losses, which motivates probing zeroes.","marker":"[48]"},{"why":"Provides non-perturbative results for zero distributions that the proposed probe could test experimentally.","marker":"[49]"},{"why":"Shows the unitary-deficit method for extracting Wigner time delay, the precedent the paper adapts to δT.","marker":"[65]"},{"why":"Supplies the harmonic inversion method used to extract zero positions from δT in the numerical test.","marker":"[39]"},{"why":"Derives the rank-one absorber representation that underlies the effective non-Hermitian Hamiltonians for reflection coefficients.","marker":"[51]"},{"why":"Defines the Wigner time delay, the pole-side analogue that δT mirrors.","marker":"[52]"},{"why":"Reports the experimental realization of random anti-lasing, the physical context that makes zeroes observable.","marker":"[50]"}],"fun_headline_variants":["Reflection time difference maps S-matrix zeroes","Zeroes of S-matrix read via reflection time difference","Probing S-matrix zeroes via reflection time difference","Reflection time difference: reading S-matrix zeroes","S-matrix zeroes from reflection time difference"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The experimental extraction assumes losses are weak and uniform, so that the spectral parameter shift λ→λ+iε with ε ≪ Δ is valid and |R1/R2| is well approximated by exp(−2εδT); if losses are strong or nonuniform the simple exponential relation fails.","fun_headline_variants_meta":{"raw":{"variants":["Reflection time difference maps S-matrix zeroes","Zeroes of S-matrix read via reflection time difference","Probing S-matrix zeroes via reflection time difference","Reflection time difference: reading S-matrix zeroes","S-matrix zeroes from reflection time difference"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000939,"raw_usage":{"total_tokens":3929,"prompt_tokens":772,"completion_tokens":3157,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":388,"completion_tokens_details":{"reasoning_tokens":3079}},"tokens_in":388,"tokens_out":3157,"duration_ms":19447,"temperature":1.0,"reasoning_tokens":3079,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:31:49.773843+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the two reflection coefficients of a real two-channel microwave billiard with a small tunable absorber; extract δT independently from the phase derivative via Eq. (13) and from the modulus ratio via Eq. (15), and compare recovered zero positions with direct numerical diagonalization of the effective Hamiltonian for the same geometry. Agreement would confirm the relations; a systematic mismatch between the two extractions at absorber strengths below the CPA threshold would falsify them.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Heidelberg random-matrix model of resonance scattering used for the entire derivation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the random-matrix description of S-matrix zeroes under weak localized losses, which motivates probing zeroes."},{"cited_title":"Fyodorov, S","cited_arxiv_id":null,"evidence_quote":"Provides non-perturbative results for zero distributions that the proposed probe could test experimentally."},{"cited_title":"Doron, U","cited_arxiv_id":null,"evidence_quote":"Shows the unitary-deficit method for extracting Wigner time delay, the precedent the paper adapts to δT."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the harmonic inversion method used to extract zero positions from δT in the numerical test."},{"cited_title":"Fyodorov","cited_arxiv_id":null,"evidence_quote":"Derives the rank-one absorber representation that underlies the effective non-Hermitian Hamiltonians for reflection coefficients."},{"cited_title":"Wigner, Lower limit for the energy derivative of the scattering phase shift","cited_arxiv_id":null,"evidence_quote":"Defines the Wigner time delay, the pole-side analogue that δT mirrors."},{"cited_title":"Pichler, M","cited_arxiv_id":null,"evidence_quote":"Reports the experimental realization of random anti-lasing, the physical context that makes zeroes observable."}],"review_version":1}