{"id":"a5a64bba-1900-433b-ad6c-62f66cb1c192","arxiv_id":"2412.10598","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A data-only continued-fraction analysis of 29 proton form factor ratio measurements predicts a zero at Q^2 = 10.37 +0.87/-0.68 GeV^2, with high stated confidence.","lead":"Using 29 polarization-transfer measurements and a model-free interpolation method, this paper predicts that the proton's electric-to-magnetic form factor ratio crosses zero at momentum transfer Q^2 = 10.37 GeV^2. If correct, the result would settle a long-running debate about proton structure and challenge many QCD-based models.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 99.9% and 1/1-million confidence statements are calibrated inside a shape-filtered SPM ensemble plus a Gaussian tail model, not against the data under the no-zero hypothesis; a closure test is needed before these claims can be accepted.","rationale":"The paper makes a real and useful contribution: the SPM is a sensible interpolation tool, the internal checks with two replica sets and varying numbers of interpolators are credible, and the result is consistent with several earlier predictions of a zero near Q^2 = 10 GeV^2. The M = 4k+2 exclusion is stated openly rather than hidden, and the analysis is internally coherent. However, the abstract's claims of model independence and the quantitative confidence statements go beyond what the method actually establishes. The zero-location estimate is stable across the accepted M values, but the ensemble is defined by shape constraints that are not derived from QCD or from the data, and the final probabilities are obtained by fitting a variable-width Gaussian to ten combined asymmetric measurements. The single most load-bearing uncertainty is therefore whether the accepted function space and the Gaussian tail model faithfully represent the data under the no-zero hypothesis. This is directly testable by a closure experiment: if synthetic positive-definite monotone data are fed through the same pipeline and produce 99.9% or 1/1-million false evidence for a zero, then the prior, not the data, is driving the headline. Until that calibration is demonstrated, the original CONDITIONAL verdict is appropriate; the result should not be presented as a model-independent 1/1-million exclusion.","tokens_in":10704,"tokens_out":9353,"duration_ms":100361,"concrete_test":"Run a closure test on synthetic data: generate many replica data sets from a positive-definite, monotone-decreasing input (e.g., R(Q^2) = 0.2 + 0.8/(1 + Q^2/4 GeV^2)) at the same 29 Q^2 points with the reported uncertainties, then apply Steps 1–4 exactly, including the monotone/C^1 acceptance filter and the M = 4k+2 exclusion. Record the fraction of replica-ensemble averages that cross zero before Q^2 = 15 GeV^2 and the resulting z-score. If a positive-definite input produces false evidence at 99.9% or 1/1-million significance, the shape prior is the load-bearing assumption and the headline confidence claims are not calibrated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—a zero at Q_z^2 = 10.37 GeV^2 and a 1/1-million exclusion of no-zero hypotheses—depends on treating the SPM interpolator ensemble as an objective likelihood, and that equivalence is the least secure step. Step 2 accepts only interpolators that are C^1 and monotonically decreasing on 0 < Q^2/GeV^2 < 20, and the text explicitly discards all M = 4k+2 orders because such interpolators cannot be monotone and have a single zero. These are shape priors, not consequences of QCD or of the data: they filter out positive-definite non-monotone curves and the entire monotone positive-definite branch of the M = 4k+2 family, and replicas that cannot supply 25 such functions are discarded. The later confidence statements (Sec. 3, Fig. 6) then convert the asymmetric bootstrap errors into a Gaussian ln L via Eqs. (4)–(6) and read P(H2) = 1/1-million off a Gaussian tail. This is a parametric extrapolation of an empirical distribution of zero locations, not a likelihood of the data under H2, and no systematic uncertainties or data correlations enter the quoted numbers. If the acceptance filter is doing the work, or if the zero-location distribution has heavy tails, the 99.9% and 1/1-million statements can be badly miscalibrated even if the central estimate is stable.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper applies the Schlessinger point method (SPM) to the 29 available polarization-transfer data points for the proton form factor ratio Rp(Q^2) = mu_p G_E^p(Q^2)/G_M^p(Q^2). Gaussian replicas of the data are generated, and for each replica the authors select 25 SPM interpolators that are C^1 and monotonically decreasing on 0 < Q^2/GeV^2 < 20; replicas that cannot supply 25 such interpolators are discarded. The analysis is repeated for M = 9, 8, 7, 5, 4, and the resulting zero locations are combined using an asymmetric Gaussian likelihood. The paper's central claims are a zero at Q_z^2 = 10.37^{+0.87}_{-0.68} GeV^2, a 99.9% confidence statement that the data are consistent with a zero on Q^2 <= 13.06 GeV^2, and a 1/1-million likelihood that the data are consistent with no zero on Q^2 <= 14.49 GeV^2.","tokens_in":10978,"tokens_out":7154,"duration_ms":68026,"significance":"If the central result is correct, the paper resolves a long-standing question about proton structure and provides a benchmark for QCD-based calculations. The analysis has several genuine strengths: it uses public data, it is transparent about the interpolation machinery, it checks stability across several M values and independent replica sets, and its central zero location agrees with the authors' Faddeev-equation prediction. However, the headline confidence statements are not supported by the present evidence because they are conditioned on a shape filter and on Gaussian tail extrapolations that are neither derived from the data nor validated by closure tests. The paper would be significant if the authors can show that the existence and location of the zero are robust to relaxing those assumptions, but as written the 99.9% and 1/1-million claims outrun the statistical methodology.","major_comments":[{"comment":"The acceptance criterion that interpolators must be C^1 and monotonically decreasing on 0 < Q^2/GeV^2 < 20 is a shape prior, not a consequence of analyticity, QCD, or the data. The paper's claim that the SPM is 'blind to any and all prejudice' is therefore not supported by the procedure actually implemented. The exclusion of all M=4k+2 orders, because those interpolators cannot be monotone and have a single real zero, removes an entire family that could describe a positive-definite monotone ratio; the reported fractions of zero-crossing interpolators (e.g., 92% for M=9) are conditional on this filter. To support the existence claim, the authors should report the fraction of initial 15,000 replicas and of generated interpolators rejected by the monotonicity condition, and should repeat the analysis with relaxed filters that include non-monotone C^1 interpolators and the M=4k+2 family without the monotonicity requirement.","section":"Sec. 2, Step 2(i)-(ii) and the M=4k+2 discussion"},{"comment":"The 99.9% and 1/1-million statements are Gaussian tail probabilities of the bootstrap distribution of the zero location, not likelihoods of the data under H1 or H2. The H2 tail is evaluated at 14.49 GeV^2 as (14.49 - 10.37)/0.87, which is about 4.7 sigma, and the quoted probability is read off a Gaussian tail. This assumes that the zero-location estimator is Gaussian with the fitted sigma and that the accepted SPM ensemble is an unbiased likelihood for the data; neither assumption is tested. Under H2, the acceptance filter would be applied to positive-definite curves, and the composition of the accepted ensemble, and hence the tail probability, would be different. The authors should provide a closure test: generate synthetic data from positive-definite functions (e.g., dipole or no-zero Faddeev-like forms) with the same noise, run Steps 1-4, and report how often a zero is claimed and at what confidence. Without this calibration, the '1/1-million' claim is not statistically meaningful.","section":"Sec. 3, Eqs. (4)-(6) and Fig. 6"},{"comment":"The procedure discards any replica for which 25 acceptable interpolators cannot be obtained, but the manuscript does not report how many of the 15,000 initial replicas are discarded. If the acceptance rate is low, the 5,000 retained replicas are a strongly selected subsample, and the uncertainty estimates derived from their zero locations will understate the variability implied by the data. The authors already check the effect of using 10 versus 50 interpolators per replica, but the selection rate itself is an essential diagnostic and should be reported for each M value. The absence of this information makes it difficult to assess whether the quoted uncertainties, and the subsequent confidence statements, reflect the data or the selection procedure.","section":"Sec. 2, Steps 3-4"}],"minor_comments":[{"comment":"The statement that 'systematic errors are small' should be justified with the systematic uncertainty estimates from Refs. [21-25], and a sensitivity test with inflated uncertainties would be useful given that the final confidence claims are at the 1e-6 level.","section":"Sec. 2, Step 1"},{"comment":"The text says 'the 5,000 M = 9 interpolators that represent the N = 29 available data', but Step 2 describes 25 interpolators per replica; please clarify whether the plotted purple curves are the 25 individual interpolators or the replica-averaged representatives.","section":"Sec. 2, after Step 4"},{"comment":"The continued-fraction notation with an ellipsis is ambiguous; please display the nested-fraction form or define the recursion for the coefficients a_i.","section":"Eq. (3)"},{"comment":"The y-axis label 'Likelihood H1' is misleading because the plotted quantity is a tail probability or posterior probability under the SPM distribution, not a likelihood of the data; please relabel accordingly.","section":"Fig. 6"},{"comment":"Ref. [31] is cited as an arXiv preprint; if a journal version now exists, please update the reference.","section":"References"},{"comment":"The phrase 'C1 function' should be written as 'C^1 function' for consistency with standard notation.","section":"Notation"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of a particle-physics journal, and the underlying question is of genuine interest. My main concern is that the eye-catching '1/1-million' statement is likely to be quoted out of context unless the authors add closure tests and soften the statistical language. I would encourage the editor to request a robustness analysis that does not presuppose monotonicity, and to require that the acceptance rate of replicas be reported."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe headline is a competent SPM analysis predicting a zero at Qz^2 = 10.37 +0.87/-0.68 GeV^2. The central value is plausible and consistent with earlier Faddeev predictions. The new thing here is the quantitative confidence claim: 99.9% for a zero at Q^2 <= 13.06 and 1/1-million against a positive-definite ratio below 14.49. That is a stronger statement than anything in the prior SPM literature, and it is the paper's real contribution.\n\nWhat is done well: the SPM machinery is standard and clearly described, the sensitivity to the number of interpolators is checked, two independent replica sets give consistent answers, and the comparison with the Faddeev curve is useful. The analysis is transparent, which is more than most phenomenological papers can say.\n\nThe soft spots are in the confidence statements. First, the acceptance filter (C^1, monotonically decreasing on 0 < Q^2 < 20) is a shape prior, not a consequence of QCD or of the data. It can push the ensemble toward functions that either stay positive or cross zero, but it excludes non-monotone positive-definite functions that might also fit the data. Second, the exclusion of M = 4k+2 orders is based on a claim I find shaky: a rational interpolator with even numerator degree can have a single positive zero if the other roots are negative or complex. Even if that claim is correct, it removes a family of monotone positive-definite functions. Third, the 'statistically independent' statement for the results in Table 1 is wrong: the same 5000 replicas are used for every M, so the M = 9, 8, 7, 5, 4 results are correlated. Combining them as independent artificially shrinks the error on Eq. (8). Fourth, the 1/1-million and 99.9% figures come from converting asymmetric errors to a variable Gaussian (Barlow's method) and reading off a tail. That is a parametric extrapolation from an empirical distribution of zero locations, not a likelihood of the data under the no-zero hypothesis. No closure test is presented.\n\nNone of this kills the central prediction. I think the zero is probably there, somewhere in the 9-12 GeV^2 range. But the current confidence claims are not supported by the analysis as written. For a serious referee, I'd ask for a closure test on synthetic positive-definite data, a treatment that accounts for the correlation across M, and replacement of the 1/1-million language with something calibrated. Send it to review.","headline":"A plausible data-driven estimate of the proton form factor zero, but the headline confidence claims rest on shape priors and on treating correlated results as independent.","tokens_in":11541,"tokens_out":4918,"would_cite":true,"duration_ms":44073,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["13.40.Gp","14.20.Dh"],"model":"deepseek-v4-flash","headline":"A model-free analysis of 29 proton form factor ratios predicts a zero in the electric form factor at Q^2 = 10.37 GeV^2, with data inconsistent with no zero at the one-in-a-million level.","keywords":["Schlessinger point method","proton electric form factor","form factor ratio","zero crossing","polarization transfer","continued-fraction interpolation","emergent hadron mass","QCD"],"falsifier":"Take polarization-transfer data on mu_p G_E^p/G_M^p at $Q^{2}$ = 10, 11, 12, and 13 $GeV^{2}$; if any measured value at or above 13 $GeV^{2}$ is positive and away from zero by more than the quoted uncertainty, the predicted zero below 13.06 $GeV^{2}$ is falsified.","tokens_in":10458,"feed_emoji":"⚛️","tokens_out":5365,"duration_ms":46954,"temperature":0.7,"pith_summary":"This paper asks a concrete question about the proton: does the ratio of its electric to magnetic form factor, mu_p G_E^p/G_M^p, actually cross zero at some momentum transfer? Using only 29 polarization-transfer data points and no model of proton structure, the authors build thousands of continued-fraction interpolations of the data and extrapolate each one. The ensemble predicts a zero at $Q^{2}$ = 10.$37^{{+0.87}}$_{-0.68} $GeV^{2}$, puts 99.9% confidence on the zero lying below 13.06 $GeV^{2}$, and estimates a one-in-a-million likelihood that the ratio stays positive through 14.49 $GeV^{2}$. If correct, the proton's electric form factor changes sign, a distinctive and testable feature of how charge is distributed inside the proton.","feed_headline":"Proton electric form factor zero predicted near Q2 = 10 GeV^2","feed_subtitle":"A model-free fit to 29 scattering data points says the ratio crosses zero with 99.9% confidence below 13 GeV^2.","key_machinery":"The Schlessinger point method (SPM), a continued-fraction interpolation also known as a multipoint Pade approximant, is the engine of the analysis. For each of 15,000 Gaussian replicas of the 29 data points, random M-point subsets are used to build interpolating rational functions, which are accepted only if they are $C^{1}$ and monotonically decreasing on 0 < $Q^{2}$/$GeV^{2}$ < 20; 5,000 replicas with 25 accepted interpolators each define the ensemble, and the zero location is read from each curve. The paper excludes M = 4k+2 interpolator orders because those cannot be monotone-decreasing and cross zero at real $Q^{2}$, a hidden constraint the authors identify and remove.","core_discovery":"The paper's central claim is that existing data, extrapolated without any hadron model, already imply a zero in the proton electric form factor ratio. The Schlessinger point method, applied to 29 measured values of mu_p G_E^p/G_M^p, yields a predicted zero at $Q_z^{2}$ = 10.$37^{{+0.87}}$_{-0.68} $GeV^{2}$ (Eq. 8). With 99.9% confidence the data are consistent with a zero on $Q^{2}$ <= 13.06 $GeV^{2}$, and the likelihood that the data are consistent with a positive-definite ratio on $Q^{2}$ <= 14.49 $GeV^{2}$ is 1/1-million. The authors present this as an objective, function-form unbiased statement about what the data themselves imply.","pith_inferences":["The same bootstrap-plus-continued-fraction recipe could be applied to neutron or hyperon form factor ratios; a zero there would indicate whether the sign change is a generic feature of baryons or specific to the proton.","The 1/1-million figure is a likelihood within the SPM ensemble, not a classical p-value against a particular alternative model, so a single high-precision measurement above 12 GeV^2 could carry more evidential weight than the ensemble tail.","If future data confirm the zero, the crossing location and its asymmetric uncertainty become a precise target for lattice QCD and for relativistic quark models of the nucleon."],"forward_implications":["A zero near Q^2 = 10.4 GeV^2 means the proton electric form factor is not positive definite, implying a diffraction-like sign change in the electric charge distribution.","The prediction is directly testable with planned polarization-transfer measurements reaching Q^2 near 12 GeV^2; the 90% confidence band already extends to 11.49 GeV^2.","The SPM result agrees with parameter-free Faddeev equation predictions of the zero location, strengthening the case that emergent hadron mass controls the falloff of the ratio.","Phenomenological fits that force the proton electric form factor to remain positive are inconsistent with the data at the one-in-a-million likelihood level over the range up to 14.49 GeV^2."],"supporting_citations":[{"why":"Supplies the first high-Q^2 polarization-transfer measurements that established the ratio falls below unity.","marker":"[21]"},{"why":"Extends the ratio measurement to Q^2 = 5.6 GeV^2, anchoring the downward trend used in the extrapolation.","marker":"[22]"},{"why":"Provides polarization-transfer data to Q^2 = 3.5 GeV^2 included in the 29-point set.","marker":"[23]"},{"why":"Adds the Q^2 = 8.5 GeV^2 measurement, the point closest to the predicted zero.","marker":"[24]"},{"why":"Supplies the precise recoil polarization observables and uncertainties that dominate the high-Q^2 tail of the data set.","marker":"[25]"},{"why":"Introduces the Schlessinger point method as the analytic-continuation tool on which the whole analysis rests.","marker":"[41]"},{"why":"Establishes continued-fraction interpolation of scattering data, the core numerical procedure of the SPM.","marker":"[42]"},{"why":"Provides the method for combining asymmetric statistical errors used to merge the independent M-value results into Eq. (8).","marker":"[49]"},{"why":"Gives a parameter-free Faddeev equation prediction of the zero location used as a benchmark comparison.","marker":"[31]"}],"fun_headline_variants":["Proton form factor zero at 10 GeV2, no model needed","99.9% confident: proton electric form factor has a zero","1 in a million: proton form factor has a zero","Model-free: proton form factor zero predicted at Q2=10","Proton form factor zero: data alone says 99.9%"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The analysis keeps only interpolating curves that are smooth and steadily decreasing, and it discards a whole family of curves that could not cross zero while staying decreasing; if the true ratio is positive but wiggles, this filter alone could manufacture the predicted zero.","fun_headline_variants_meta":{"raw":{"variants":["Proton form factor zero at 10 GeV2, no model needed","99.9% confident: proton electric form factor has a zero","1 in a million: proton form factor has a zero","Model-free: proton form factor zero predicted at Q2=10","Proton form factor zero: data alone says 99.9%"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001289,"raw_usage":{"total_tokens":5235,"prompt_tokens":885,"completion_tokens":4350,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":501,"completion_tokens_details":{"reasoning_tokens":4259}},"tokens_in":501,"tokens_out":4350,"duration_ms":24362,"temperature":1.0,"reasoning_tokens":4259,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T15:48:16.022393+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take polarization-transfer data on mu_p G_E^p/G_M^p at $Q^{2}$ = 10, 11, 12, and 13 $GeV^{2}$; if any measured value at or above 13 $GeV^{2}$ is positive and away from zero by more than the quoted uncertainty, the predicted zero below 13.06 $GeV^{2}$ is falsified.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the first high-Q^2 polarization-transfer measurements that established the ratio falls below unity."},{"cited_title":"Gayou, et al., Measurement of G(E(p)) /G(M(p)) in⃗ep→ e⃗p to Q2 = 5.6 GeV2, Phys","cited_arxiv_id":null,"evidence_quote":"Extends the ratio measurement to Q^2 = 5.6 GeV^2, anchoring the downward trend used in the extrapolation."},{"cited_title":"Punjabi, et al., Proton elastic form factor ratios to Q2 = 3.5 GeV2 by polarization transfer, Phys","cited_arxiv_id":null,"evidence_quote":"Provides polarization-transfer data to Q^2 = 3.5 GeV^2 included in the 29-point set."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the precise recoil polarization observables and uncertainties that dominate the high-Q^2 tail of the data set."},{"cited_title":"Schlessinger, C","cited_arxiv_id":null,"evidence_quote":"Introduces the Schlessinger point method as the analytic-continuation tool on which the whole analysis rests."},{"cited_title":"Barlow, Asymmetric statistical errors, in: PHYSTAT (2005): Statis- tical Problems in Particle Physics, Astrophysics and Cosmology, 56–59, 2004","cited_arxiv_id":null,"evidence_quote":"Provides the method for combining asymmetric statistical errors used to merge the independent M-value results into Eq. (8)."}],"review_version":1}