{"id":"675218df-9cb8-419f-b34b-f016ed29b319","arxiv_id":"2501.15329","paper_version":3,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":5,"one_line_summary":"A closed-form error-propagation formula, verified with GEANT4 and toy simulations, predicts the resolution of dual readout calorimeters and explains when the correction degrades performance.","lead":"Researchers derive a simple formula that predicts the energy resolution of dual readout calorimeters from the fluctuations in their two signals. The formula, tested against detailed simulations, shows when the dual readout correction improves measurements and when it makes them worse.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The covariance term in Eq. 11 assumes all S-C correlation comes from σf; correlated loss and sampling fluctuations (Figs. 11, 18) add terms not included, and Table 6 validates only the final σD, leaving the key assumption indirectly tested.","rationale":"The paper's central claim is Eq. 11 as a predictive formula for dual-readout resolution. The formula is a linear combination of S and C, so the only non-trivial ingredient beyond measured single-signal resolutions is the covariance model. The load-bearing assumption is therefore that all S-C correlation arises from the common dependence on f, with constant hS, hC and f-independent noise. The simulation itself shows conditions where this assumption is violated: sampling fractions depend on f (Fig. 11), and the loss fractions fNC and fES are correlated (Fig. 18). The final comparison in Table 6 tests the aggregate prediction, but it cannot separate a failure of the covariance term from compensating deviations in σS and σC. A direct covariance comparison is the cleanest check of the formula's validity. The same concern was identified by the reader as the weakest assumption; this stress-test sharpens it into a specific, testable prediction and notes that the paper's own qualitative caveat in Sec. 4.3 does not quantify the impact on Eq. 11. Because the existing evidence supports the formula and the paper is appropriately hedged, the verdict does not need to change, but the proposed check would materially strengthen the verification claim without requiring new simulations beyond the existing samples.","tokens_in":12589,"tokens_out":11982,"duration_ms":108326,"concrete_test":"Using the existing GEANT4 samples, compute cov(S,C) directly from per-event S and C for each of the five calorimeters, and compare it with (1−hS)(1−hC)σf² using hS, hC, σf from Tables 4–5. Then recompute the predicted σD by inserting the directly measured covariance into Eq. 11 in place of the model covariance term. If the residuals in Table 6 shift by more than the quoted statistical uncertainties, the covariance assumption behind Eq. 11 is falsified; if they remain within uncertainties, the formula's key input is validated directly.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Equation 11 follows from the error-propagation formula (Eq. 6) only when the covariance is exactly cov(S,C) = (1−hS)(1−hC)σf² (Eq. 10, Appendix A). This derivation treats hS and hC as constants and the noise terms as independent of f and of each other. The full simulation, however, shows that sampling fractions vary with f (Fig. 11) and that fNC and fES are correlated event-by-event (Fig. 18). These shared fluctuations add contributions to cov(S,C) that are not contained in Hσf². Table 6 compares the full prediction of Eq. 11 with the measured σD, but because σD is built from a²σS² + b²σC² + 2ab·cov(S,C), agreement in σD is only an indirect and possibly insensitive test of the covariance model. A direct measurement of cov(S,C) from the simulated events is not reported, so the central quantitative assumption of the paper remains unvalidated in the regime (sampling calorimeters, SampS-like geometries) where it is most likely to fail. The paper honestly acknowledges the assumptions in Sec. 4.3, but the verification claim in Sec. 4.4 would be strengthened decisively by a direct covariance check.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper derives a closed-form expression for the energy resolution of a dual-readout calorimeter after the dual-readout correction, Eq. (11), starting from the ansatz that the scintillation and Cherenkov signals share a common dependence on the electromagnetic fraction f, plus independent noise terms. The covariance between S and C is computed in Appendix A, giving cov(S,C) = (1-hS)(1-hC)σf². The formula is tested against a toy Monte Carlo and against GEANT4 simulations of five calorimeter geometries (PbWO, fiber, and sampling calorimeters), with the predicted corrected resolution compared in Table 6. The paper also presents approximate formulae for hS and σS in terms of binding-energy loss, escaping energy, and sampling-fraction fluctuations. The stated purpose is to explain when the dual-readout correction improves or worsens the resolution.","tokens_in":12946,"tokens_out":18592,"duration_ms":141570,"significance":"If the formula is reliable, it provides a simple, design-oriented tool for estimating the performance of dual-readout calorimeters without full simulation, and it clarifies the counterintuitive result reported in Ref. [4] that the correction can degrade resolution. The paper is explicitly pedagogical and is likely to be useful to newcomers to the field. The full GEANT4 comparison across five geometries, including one (SampS) where the correction worsens resolution, is a valuable stress test. The toy simulation code is made publicly available on GitHub, and the derivation in Appendix A is algebraically correct. The main weaknesses are an incorrect inequality in one of the derived conditions, and a verification that tests the covariance model only indirectly through the final σD.","major_comments":[{"comment":"The inequality in Eq. (14) is reversed. Starting from Eq. (11), σD > σS is equivalent to σf² < [(2−χ)σS²] / [2(1−hC)(1−hS)] + σC² / [2(1−hC)²], not the '>' condition printed. For a concrete counterexample with hS=0.9, hC=0.6, σS=σC=0.05, and σf=0.3, the RHS of Eq. (14) is 0.0625 while σf²=0.09, so the printed condition says σD > σS, but Eq. (11) gives σD≈0.039 < σS=0.05. The accompanying discussion near Eq. (13) should be re-examined as well, since the condition for the sum of the second and third terms in Eq. (12) to be positive depends on the same inequality direction.","section":"Sec. 2, Eq. (14)"},{"comment":"The verification of Eq. (11) against the full simulation is indirect: hS, hC, and σf are all extracted from the same simulated events used to measure σD, and the comparison only checks the final σD. Since the only non-trivial input of Eq. (11) beyond the measured σS and σC is the covariance model of Eq. (10), a direct comparison between the model covariance (1−hS)(1−hC)σf² and the covariance computed from the simulated (S,C) pairs would provide a much more sensitive test of the central assumption. The statement in Sec. 4.3 that correlations between fNC and fES 'do not affect the prediction' of the corrected resolution is not demonstrated; reporting a direct covariance comparison, or at least a quantitative estimate of the neglected contributions, would substantially strengthen the verification claim.","section":"Sec. 4.4, Table 6"}],"minor_comments":[{"comment":"The partial derivatives in Eq. (6) are written as ∂E/∂S and ∂E/∂C, but the quantity being propagated is D; this should be ∂D/∂S and ∂D/∂C.","section":"Sec. 2, Eq. (6)"},{"comment":"Equation (31) contains a typographical error: the term (1−<S)2 is missing the closing angle bracket, and the overall layout makes the placement of the denominator ambiguous; please reformat.","section":"Sec. 4.4, Eq. (31)"},{"comment":"The definition fNC = (EB − EES − EI)/EB normalizes the binding-energy loss to the total beam energy, but in Eqs. (19), (22), and (24) the same symbol is used as a fractional loss of the non-EM component. Please clarify the intended normalization and reconcile it with the caption of Fig. 15, which refers to division by the non-EM energy.","section":"Sec. 4.2, Eq. (20)"},{"comment":"In the Fiber2 row, σf is listed as 0.12 ± 0.1; the uncertainty appears to be a factor of 10 too large and is inconsistent with the other entries; likely ±0.01 is intended.","section":"Table 5"},{"comment":"There are several typographical errors, including 'simulatate' in Sec. 4.1, 'the the D resolution' in Sec. 3, 'ansantz' in Sec. 4.2, and the use of 'fN C' and 'fESC' with inconsistent spacing; a careful proofread is recommended.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The paper is a good fit for an instrumentation journal and the central formula is likely to be useful. The incorrect inequality in Eq. (14) is a clear technical error that must be corrected. The covariance-verification concern is less severe but worth addressing, as it bears on the strength of the central verification claim. I do not see grounds for rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe short version: this is a useful paper. It derives a closed-form expression for the resolution of the dual-readout-corrected energy, Eq. 11, and shows it holds across five GEANT4 calorimeter geometries, including one where the correction makes the resolution worse. That last point is the one that will grab people: the formula gives a quick way to know when dual readout helps and when it hurts.\n\nWhat's actually new: the closed-form formula itself, the explicit role of σf in the covariance, and the demonstration that escaping energy is corrected by the same mechanism as binding-energy loss. The auxiliary expressions for hS, σS etc. are also useful for design studies. The writing is clear and the limitations are stated honestly. The toy MC and GEANT4 comparisons are appropriate, and Table 6 shows agreement within uncertainties for all five geometries; Table 7's prediction of hS from loss fractions is a nice independent-ish cross-check.\n\nThe soft spots are real but not fatal. The covariance in Eq. 10 assumes hS and hC are constant and that the noise and loss terms are independent of f and each other. The full simulation shows those assumptions are violated, especially for SampS (correlations between fNC and fES, sampling fractions varying with f). The paper acknowledges this in Sec. 4.3, and the formula still lands within 0.4σ on SampS, so the indirect test is not blind. Still, the stress-test note is right that a direct measurement of cov(S,C) from the simulated events would be the decisive check; comparing only σD leaves the covariance model one step removed.\n\nThere is also mild circularity in using the same simulated distributions to extract hS, hC, σf and to evaluate σD. That does not sink the paper because Eq. 11 is derived, not fitted, and the auxiliary predictions in Tables 7 and 8 give some leverage. I'd call it a moderate caveat, not a flaw.\n\nBottom line: for anyone designing or evaluating dual-readout calorimeters, this formula will become a standard rule of thumb. It deserves a serious referee. My recommendation: send it out; ask the authors to add a direct covariance comparison if they have the events handy. That would close the last gap.","headline":"A clean, honest derivation of a dual-readout resolution formula, validated broadly; the covariance check the stress-test wants would be nice but the paper already earns a serious review.","tokens_in":13437,"tokens_out":2090,"would_cite":true,"duration_ms":18240,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["29.40.Vj"],"model":"deepseek-v4-flash","headline":"One resolution formula predicts when the dual-readout correction improves hadronic calorimeters and when it hurts them.","keywords":["dual readout calorimetry","hadronic energy resolution","electromagnetic fraction","Cherenkov readout","scintillation readout","energy resolution formula","escaping energy","nuclear binding energy loss"],"falsifier":"Run a full shower simulation, compute the energy estimates $S$ and $C$ shower by shower, and measure their covariance directly; the formula requires $\\mathrm{cov}(S,C) = (1-h_S)(1-h_C)\\sigma_f^2$, so a disagreement larger than the statistical uncertainties would invalidate the error-propagation chain that produces Eq. (11).","tokens_in":12388,"feed_emoji":"⚛️","tokens_out":10785,"duration_ms":87011,"temperature":0.7,"pith_summary":"This paper derives a closed-form formula for the resolution of a dual-readout calorimeter's corrected energy estimate $D$. The formula combines the scintillation resolution $\\sigma_S$, the Cherenkov resolution $\\sigma_C$, and the shower-to-shower fluctuation $\\sigma_f$ of the electromagnetic fraction, explicitly subtracting the part of the two signals that is correlated through $f$. The authors verify the formula with a toy Monte Carlo and with full simulations of five calorimeter geometries, including a poorly contained sampling calorimeter in which the dual-readout correction makes the resolution worse. The result matters because dual-readout calorimeters are leading candidates for future collider calorimetry, and the formula lets designers know in advance whether the correction will help a given geometry or be defeated by uncorrelated noise and escaping energy.","feed_headline":"One formula predicts when dual-readout calorimeters help","feed_subtitle":"The dual-readout correction helps only if electromagnetic-fraction fluctuation beats Cherenkov noise","key_machinery":"The load-bearing object is the corrected estimator $D$, a linear projection of the $(C,S)$ pair along their correlation line onto the electron energy scale. The argument uses simple error propagation: because the scintillation and Cherenkov signals both depend on the same electromagnetic fraction $f$, their covariance is $(1-h_S)(1-h_C)\\sigma_f^2$; inserting this covariance into the standard error-propagation formula gives Eq. (11). The secondary machinery is the decomposition of the shower into an electromagnetic part, a non-electromagnetic part, and the loss fractions $f_{NC}$ (nuclear binding) and $f_{ES}$ (escaping energy), which yields approximate formulas for $h_S$ and $\\sigma_S$ and makes the meaning of the correction concrete.","core_discovery":"The paper's central claim is Eq. (11): $$\\sigma_D = \\frac{1}{h_S-h_C}\\sqrt{(1-h_C)^2\\$sigma_S^{2}$ + (1-h_S)^2\\$sigma_C^{2}$ - 2(1-h_S)^2(1-h_C)^2\\$sigma_f^{2}$},$$ which predicts the resolution of the dual-readout-corrected energy estimate $D=\\frac{(1-h_C)S-(1-h_S)C}{h_S-h_C}$ for hadrons of fixed energy. Here $h_S$ and $h_C$ are the average responses of the scintillation and Cherenkov readouts to the non-electromagnetic part of the shower, and $\\sigma_f$ is the rms fluctuation of the electromagnetic fraction. The paper tests this expression against simulated showers in a homogeneous crystal calorimeter, two fiber calorimeters, and two sampling-tile calorimeters, finding predicted corrected resolutions close to the simulated values for all five geometries, including the SampS case where the correction degrades resolution because escaping energy and large correlated fluctuations dominate. The same derivation shows the correction compensates not only for nuclear binding-energy loss but also for energy escaping the calorimeter and energy missed by the clustering algorithm, and it yields approximate formulas for $h_S$ and $\\sigma_S$ in terms of sampling fractions and the loss fractions $f_{NC}$ and $f_{ES}$.","pith_inferences":["A practical consequence not spelled out in the paper: Eq. (13) can be used as a go/no-go test during design, comparing expected Cherenkov noise to the electromagnetic-fraction fluctuation before investing in the second readout.","The same covariance argument transfers to any two-readout estimator that shares a common latent variable: whenever the correlated fluctuation term is small compared with the uncorrelated noises, the corrective linear combination will degrade rather than improve resolution.","The formula suggests a direct simulation diagnostic: measure the empirical shower-by-shower covariance of $S$ and $C$ and compare it with $(1-h_S)(1-h_C)\\sigma_f^2$; a large discrepancy would identify correlations beyond the simple Gaussian-$f$ model.","For future collider calorimeters, the explicit appearance of $f_{ES}$ in $h_S$ turns containment into a quantitative resolution input, so optimizing length and transverse size becomes an optimization of the resolution formula itself."],"forward_implications":["With measured or estimated $h_S$, $h_C$, $\\sigma_S$, $\\sigma_C$, and $\\sigma_f$, Eq. (11) predicts before full construction whether the dual-readout correction will improve a proposed calorimeter's hadronic resolution.","The correction is expected to hurt when the Cherenkov readout is noisy: Eq. (13) states that the dual-readout estimate is worse than the scintillation-only estimate unless $\\sigma_f > \\sigma_C/(\\sqrt{2}(1-h_C))$.","Dual-readout compensation applies to scale fluctuations from escaping energy and clustering losses as well as from nuclear binding energy, so containment and algorithm efficiency enter the resolution budget explicitly.","For nearly compensating geometries such as Fiber2, the dual-readout correction gives only a modest improvement because the scintillation response already has little dependence on $f$.","The formula explains the counterintuitive SampS result from an earlier simulation, predicting the observed worsening of resolution for a small, poorly contained calorimeter."],"supporting_citations":[{"why":"Establishes the two-signal model (S and C as linear functions of the electromagnetic fraction) and the standard dual-readout correction this paper builds on.","marker":"[1]"},{"why":"A previous full simulation showing the dual-readout correction can worsen resolution, the counterintuitive result that motivates and is explained by Eq. (11).","marker":"[4]"},{"why":"Supplies the fiber-calorimeter geometry and simulated resolution used as an independent check of the predicted corrected resolution.","marker":"[3]"},{"why":"The toy Monte Carlo implementation used to test Eq. (11) before the full simulation comparison.","marker":"[5]"},{"why":"The detector simulation toolkit used to generate the full shower results for the five calorimeter geometries.","marker":"[6]"}],"fun_headline_variants":["A single formula tells when dual-readout calorimetry pays off","Dual-readout correction's benefit hinges on EM fraction noise","New formula predicts dual-readout calorimeter resolution","Dual-readout works when EM fluctuation beats Cherenkov noise"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The formula assumes that the electromagnetic fraction $f$ of the shower is roughly Gaussian with moderate width and that $h_S$ and $h_C$ are constants independent of $f$ and of the noise terms; where those assumptions break down, as in the small SampS calorimeter, the quantitative agreement loosens even though the formula still describes the trend.","fun_headline_variants_meta":{"raw":{"variants":["A single formula tells when dual-readout calorimetry pays off","Dual-readout correction's benefit hinges on EM fraction noise","New formula predicts dual-readout calorimeter resolution","Dual-readout works when EM fluctuation beats Cherenkov noise"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000783,"raw_usage":{"total_tokens":3448,"prompt_tokens":926,"completion_tokens":2522,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":542,"completion_tokens_details":{"reasoning_tokens":2453}},"tokens_in":542,"tokens_out":2522,"duration_ms":15920,"temperature":1.0,"reasoning_tokens":2453,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T14:23:35.225602+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a full shower simulation, compute the energy estimates $S$ and $C$ shower by shower, and measure their covariance directly; the formula requires $\\mathrm{cov}(S,C) = (1-h_S)(1-h_C)\\sigma_f^2$, so a disagreement larger than the statistical uncertainties would invalidate the error-propagation chain that produces Eq. (11).","supporting_citations":[{"cited_title":"Chekanov, S","cited_arxiv_id":null,"evidence_quote":"A previous full simulation showing the dual-readout correction can worsen resolution, the counterintuitive result that motivates and is explained by Eq. (11)."},{"cited_title":"URL https://github.com/saraheno/DualReadoutToy","cited_arxiv_id":null,"evidence_quote":"The toy Monte Carlo implementation used to test Eq. (11) before the full simulation comparison."}],"review_version":1}