Pith. sign in

REVIEW 4 major objections 5 minor 28 references

This paper claims a Bayesian smooth fit pins the global 12C+12C S-factor at 1.5 MeV to (2.13±0.01)×10^16 keV b in the body and 1.20×10^16 keV b in the abstract—both below the traditional 1975 normalization.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-03 15:36 UTC pith:VD2Q436E

load-bearing objection A well-intentioned re-analysis undermined by an error in the uncertainty weighting and an abstract that disagrees with the body; the sub-percent error bar is an artifact. the 4 major comments →

arxiv 2512.16169 v2 pith:VD2Q436E submitted 2025-12-18 astro-ph.SR

Bayesian Smooth-Fit Extrapolation of the ¹²C+{}¹²C Astrophysical S Factor

A. M. Mukhamedzhanov This is my paper
classification astro-ph.SR
keywords 12C+12C fusionastrophysical S factorBayesian fitcarbon burningstellar reaction rateTrojan Horse Methodinverse kinematicslow-energy extrapolation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper tries to establish a smooth, statistically defensible value for the astrophysical S factor of 12C+12C fusion in the energy range that controls carbon burning in stars, using a Bayesian quadratic fit to log10 S*(E) over 1.5–3.5 MeV. The body of the paper quotes the global value at 1.5 MeV as (2.13+0.01−0.01)×10^16 keV b, a 68% interval of only ±0.45%, while the abstract states a median of 1.20×10^16 keV b; both are below the classic 1975 reference value of 3.0×10^16 keV b. If the lower trend is right, the thermonuclear 12C+12C rate is only about a third to a half of the widely used analytic rate at temperatures T9 = 0.3–1.2, which would slow carbon burning in stellar evolution and supernova scenarios. The paper explicitly does not model resonances; it claims only to constrain the smooth background component, with uncertainties propagated from the assigned data errors.

Core claim

On its own terms, the paper's discovery is that when heterogeneous low-energy datasets—direct measurements, Coulomb-corrected Trojan Horse points, and inverse-kinematics results—are forced to agree with a single smooth quadratic in log10 S*(E), the allowed band collapses to a very narrow envelope. The central number is quoted in the body as S*(1.5 MeV) = (2.13+0.01−0.01)×10^16 keV b, corresponding to 0.45% uncertainty; the abstract quotes S*(1.5 MeV) = 1.20×10^16 keV b. Both values are below the 1975 reference normalization of 3.0×10^16 keV b. The author takes this as evidence against a sharp rise of S*(E) at low energies and against a low-energy fusion-hindrance pattern, and in favor of a r

What carries the argument

The carrying object is the modified astrophysical factor S*(E) = S(E)e^{0.46E}, whose base-10 logarithm is fitted as a quadratic a0 + a1(E−E0) + a2(E−E0)^2 over 1.5–3.5 MeV with pivot E0 = 1.5 MeV. The fit is done in log space so errors become approximately Gaussian; a flat prior yields a Gaussian posterior for the coefficients, and Monte Carlo draws from the covariance matrix propagate the band to S*(E). To keep point clusters from dominating, the analysis replaces each uncertainty by σ_i/√N when N measurements occupy a close energy region, and assigns a flat 25% relative error to all datasets except the inverse-kinematics one. The machinery's job is to turn many inconsistent data points in

Load-bearing premise

The load-bearing premise is that, after assigning a flat 25% uncertainty and applying the √N reduction, the remaining scatter among the 614 points can be treated as independent Gaussian noise; if the real systematic differences among experiments are larger or correlated, the quoted 0.45% band and the central value itself do not follow.

What would settle it

Recalculate the same posterior exactly as specified in the paper: if the pipeline reproduces 1.20×10^16 keV b, the body's 2.13 is wrong; if it reproduces 2.13, the abstract is wrong. Independently, repeat the fit with per-dataset normalization nuisance parameters and no √N grouping; if the 68% interval at 1.5 MeV widens beyond roughly 20% or the median moves outside 1.2–2.2 ×10^16 keV b, the claimed narrow constraint does not survive a more honest error model.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • If the smooth posterior is correct, the 12C+12C fusion rate at T9 = 0.3–1.2 is only 33–46% of the commonly used analytic rate, materially slowing carbon burning in stellar models.
  • The 68% credible interval quoted at 1.5 MeV is so narrow (0.45%) that it would make S*(1.5 MeV) a sharp anchor for any future R-matrix or reaction-rate evaluation.
  • The absence of a steep low-energy rise in the smooth fit means the low-energy enhancement suggested by a plane-wave Trojan Horse analysis is not supported once Coulomb corrections and inverse-kinematics data are included.
  • The fit is deliberately restricted to the smooth component, so narrow resonances remain to be added; the quoted rate should be regarded as a truncated rate above T9 ≈ 1.2.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: the 0.45% credible interval is almost certainly an artifact of the error model, because assigning every non-inverse-kinematics dataset the same 25% uncertainty and then dividing by √N removes the between-dataset scatter that carries real systematic information.
  • Editorial inference: the abstract/body disagreement (1.20 vs 2.13 ×10^16 keV b) is large enough that no downstream comparison—rate ratios, stellar yields—should be trusted until the published number is reconciled in a single consistent workflow.
  • Editorial inference: a straightforward calibration test would be to generate synthetic 12C+12C-like datasets from a known smooth S*(E) plus realistic correlated systematics, then apply this pipeline and check whether the 68% interval covers the truth; one would expect it to badly undercover.
  • Editorial inference: the smooth-component posterior could be used as a prior for the background term of a full R-matrix fit, which would yield a complete stellar rate and test whether residual resonance structure changes the 1.5 MeV value.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes a Bayesian smooth-fit of log10 S*(E) using a quadratic polynomial over the energy range 1.5–3.5 MeV, combining direct measurements, Coulomb-renormalized Trojan Horse Method data, and the inverse-kinematics results of Nan et al. The central claims are a tightly constrained value S*(1.5 MeV) = (2.13^{+0.01}_{-0.01}) × 10^16 keV b (a 68% credible interval of ±0.45%) and a reduced stellar carbon-burning rate relative to the CF88 reference. The linear algebra of the Gaussian posterior and the Monte Carlo propagation are standard. However, the abstract quotes a different central value (1.20 × 10^16 keV b), the claimed reaction-rate reduction is not derived anywhere in the body, and the quoted sub-percent uncertainty is controlled by ad hoc uncertainty assignments, including a σ/√N rescaling that, as implemented, increases rather than decreases the weight of clustered data.

Significance. If the analysis were sound, a smooth global extrapolation anchored by inverse-kinematics and THM constraints would be a useful input for nuclear astrophysics, and the paper is genuinely transparent in presenting its equations and pipeline (data → likelihood → posterior → S*(E)). It also claims robustness checks against cubic/quartic extensions. But the central numerical result is internally inconsistent, the uncertainty model has a load-bearing flaw, and the astrophysical rate conclusion is not substantiated. Until these are resolved, the paper's contribution cannot be assessed; the sub-percent credible interval, in particular, appears to be an artifact of the uncertainty model rather than a robust data-driven constraint.

major comments (4)
  1. [Abstract vs. Sec. V B and Sec. VI] The abstract states S*(1.5 MeV) = 1.20 × 10^16 keV b, whereas Eq. (34) and Eq. (35) quote S*(1.5 MeV) = 2.13 × 10^16 keV b. These differ by a factor of 1.8, which is large enough to alter the astrophysical conclusion. The abstract also claims R_present/R_CF88 ≈ 0.33–0.46, but no reaction-rate calculation, temperature integration, or comparison with CF88 appears in the body; this is an unsupported central claim.
  2. [Sec. V D, Eq. (8)] The σ_eff = σ_i/√N prescription is introduced to prevent clusters of closely spaced points from dominating the fit. In the likelihood of Eq. (8), however, replacing σ_i by σ_i/√N multiplies each individual point's weight by N. A cluster of N points therefore contributes N^2/σ_i^2 rather than N/σ_i^2 to the chi-square, so clusters dominate more, not less. This directly contradicts the stated purpose and inflates the information content of the 614-point dataset. Since the 0.45% credible interval in Eq. (34) depends on this weighting, the headline uncertainty is not meaningful as presented.
  3. [Eqs. (1), (2), (31), (34)] The central value and its uncertainty are not independent measurements: because the pivot is E0 = 1.5 MeV, Eq. (31) gives y0 = a0, so S*(1.5 MeV) = 10^{a0} and the ±0.45% interval are direct re-statements of the fitted polynomial coefficient and its covariance. That covariance is governed by the uniform 25% uncertainty (Eq. 1), the 5% floor for the Nan et al. data (Eq. 2), and the σ/√N rescaling. These are ad hoc choices, and no sensitivity analysis is provided. The claim in Sec. V D that the narrow posterior arises naturally from N = 614 rather than from the uncertainty assignments is therefore not supported.
  4. [Sec. II D, Eq. (3)] The adequacy of the quadratic model is asserted rather than demonstrated. Item 4 says higher-order terms do not improve the Bayesian evidence and item 5 says robustness checks confirm stability, but no evidence values, chi-square comparisons, or robustness plots are shown. Without these, the reader cannot check whether the smooth-background assumption is consistent with the data or whether the narrow posterior is an artifact of the restricted model space.
minor comments (5)
  1. [Abstract] The abstract says data are used at 'E < 3.5 MeV', but the fit is restricted to 1.5 ≤ E ≤ 3.5 MeV. The lower bound should be stated explicitly.
  2. [Eq. (33)] Equation (33) contains (1.5 − E0), which is zero by the definition E0 = 1.5 MeV. This is needlessly confusing and should be simplified to S*(1.5 MeV) = 10^{a0}.
  3. [Figs. 1 and 2] Both captions state that the median curve is visually indistinguishable from the band boundaries. As a result, the figures do not convey the fit quality or the energy dependence of the uncertainty and should be replotted with a scale where the median is visible.
  4. [References] Reference [8] contains a duplicated '192702)', and reference [21] contains garbled text ('tieacce♪tlowercase'). The reference list needs a careful editorial pass.
  5. [Notation] S and S^* are used interchangeably; Eq. (7) defines S^* = S e^{0.46E}, but Eqs. (29) and (34) sometimes quote Sglobal and sometimes S^*. The distinction should be made consistently.

Circularity Check

0 steps flagged

No significant circularity: the reported S*(1.5 MeV) is the fitted quadratic coefficient at the pivot energy, and the band is standard covariance propagation.

full rationale

The paper is transparently an empirical fit rather than a first-principles derivation. Equation (3) posits log10 S*(E) as a quadratic polynomial in E, and the pivot is chosen at E0 = 1.5 MeV. Equation (31) explicitly identifies the mean of the posterior for y(E0) with the MAP value a0,MAP = 16.32, and Eq. (33) reduces S*(1.5 MeV) to 10^{a0}. Thus the headline value is a fitted coefficient, not an independent prediction. This is what a parameter-estimation paper reports; it is not circular unless the paper claims to predict a quantity that was used as input, which it does not. The LO/MED/HI band is obtained by propagating the covariance matrix of the same linear model (Eqs. 23, 28), again standard uncertainty propagation, not a separate derived result. The datasets enter only through the likelihood; no low-energy S*(1.5 MeV) value is imposed as a prior or constraint. The self-citations (Refs. [13], [18]) are not load-bearing in a circular way: the Bayesian formalism is standard, and the low-energy conclusion is also anchored by the independent inverse-kinematics data of Ref. [12]. The σ_eff = σ_i/√N weighting and uniform 25% uncertainties are ad hoc modeling choices that affect the width of the quoted interval, and the abstract/body discrepancy (1.20 vs 2.13 ×10^16 keV b) is a serious internal-consistency problem, but these are statistical/reproducibility concerns, not circularity under the specified definitions. The derivation chain does not reduce to its own inputs by construction.

Axiom & Free-Parameter Ledger

7 free parameters · 6 axioms · 0 invented entities

The paper introduces no new physical entities, but its central result depends on several hand-chosen statistical ingredients: the fitted polynomial coefficients, uniform 25% uncertainties, a 5% floor, and an ad hoc σ/√N variance reduction. These choices, rather than data alone, shape the posterior and the quoted 0.45% credible interval.

free parameters (7)
  • a0 (log10 S* at E0=1.5 MeV) = 16.32
    Intercept of the quadratic fit in log10 S*(E); the headline S*(1.5 MeV)=10^a0 is this fitted coefficient.
  • a1 (linear coefficient)
    Linear term of the quadratic polynomial fit; value not quoted in the text.
  • a2 (quadratic coefficient)
    Quadratic term of the polynomial fit; value not quoted in the text.
  • Uniform fractional uncertainty δ_i=0.25 = 0.25
    Assigned to all datasets except Nan et al. (Sec. II A, Eq. 1). This choice directly controls the posterior width and is not derived from experimental quoted errors.
  • Minimum 5% uncertainty for Nan et al. data = 0.05
    Equation (2) imposes a floor on the inverse-kinematics data uncertainty to avoid an artificially narrow constraint.
  • Effective uncertainty σ_eff = σ_i/√N
    Ad hoc variance reduction introduced in Sec. V D to prevent clusters of points from dominating the fit. It strongly narrows the final credible interval.
  • Pivot energy E0=1.5 MeV = 1.5 MeV
    Chosen pivot for the polynomial. The paper states the result is insensitive to pivot choice within 1.5–2.5 MeV, but the headline value is evaluated at the pivot.
axioms (6)
  • domain assumption The experimental data from Refs. [3–10,12,13] are accurate, representative, and their systematic differences are fully captured by the assigned uncertainties.
    The entire fit takes the data as ground truth; no validation of cross-dataset consistency is shown beyond the smoothness of the fit.
  • ad hoc to paper log10 S*(E) is well approximated by a quadratic polynomial over 1.5–3.5 MeV.
    Section II D argues the model is a local regression, not a physical model. The robustness checks are mentioned but not shown in the text.
  • domain assumption After log transformation, experimental uncertainties are Gaussian and statistically independent.
    Used in constructing the likelihood (Eq. 8). No tests of Gaussianity or independence are provided.
  • domain assumption Flat priors on a0, a1, a2 are sufficient; the Gaussian approximation to the posterior is valid.
    Section III B justifies the Gaussian posterior by the large number of points, but no prior sensitivity or convergence diagnostics are shown.
  • ad hoc to paper The effective uncertainty reduction σ_eff = σ_i/√N is a valid way to combine clustered measurements.
    Introduced in Sec. V D without derivation; it is not standard for regression with correlated model error and directly affects the reported uncertainty.
  • domain assumption S*(E) = S(E)e^{0.46E} is the appropriate modified S-factor for combining the datasets.
    Equation (7) adopts this conventional renormalization; the constant 0.46 MeV^-1 is taken from the literature, not derived here.

pith-pipeline@v1.3.0-alltime-deepseek · 10027 in / 11304 out tokens · 103679 ms · 2026-08-03T15:36:42.803724+00:00 · methodology

0 comments
read the original abstract

A Bayesian analysis of the astrophysical \(S\) factor for the \(^{12}{\rm C}+{}^{12}{\rm C}\) fusion reaction is presented using available low-energy information at carbon--carbon relative energies \(E <3.5~{\rm MeV}\), including direct measurements and recent inverse-kinematics data. The goal of the global Bayesian fit is not to reproduce the local resonance-by-resonance structure of the \(^{12}{\rm C}+{}^{12}{\rm C}\) system, but rather to constrain the smooth global component of \(S^{*}(E)\) using the direct and inverse-kinematics constraints. The resulting posterior \(S^{*}(E)\) lies systematically below the traditional FCZ75 reference normalization over the energy interval considered. For example, at \(E=1.5~{\rm MeV}\) the median posterior value is \(S^{*}(1.5~{\rm MeV})=1.20\times10^{16}~{\rm keV\,b}\), whereas the Fowler--Caughlan--Zimmerman reference value is \(S^{*}_{\rm FCZ75}=3.0\times10^{16}~{\rm keV\,b}\) [W.~A. Fowler, G.~R. Caughlan, and B.~A. Zimmerman, Annu. Rev. Astron. Astrophys. \textbf{13}, 69 (1975)]. Thus, the updated posterior favors a lower smooth low-energy trend and does not support a sharp increase of \(S^{*}(E)\) as the energy decreases. The astrophysical consequence of the analysis is assessed through the thermonuclear reaction rate \(N_A\langle\sigma v\rangle\). The resulting median rate is lower than the CF88 analytic rate [G.~R. Caughlan and W.~A. Fowler, At. Data Nucl. Data Tables \textbf{40}, 283 (1988)] over the temperature interval considered, with \(R_{\rm present}/R_{\rm CF88}\simeq0.33\)--\(0.46\) for \(0.3\leq T_9\leq1.2\). At higher temperatures, where the Gamow window extends above the upper end of the adopted integration interval, the calculated rate should be regarded as a truncated rate rather than a complete stellar rate.

Figures

Figures reproduced from arXiv: 2512.16169 by A. M. Mukhamedzhanov.

Figure 1
Figure 1. Figure 1: FIG. 1. Global posterior distribution of [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

28 extracted references · 5 canonical work pages

  1. [1]

    [12] are assigned a uni- form fractional uncertainty δi = 0.25,(1) corresponding to a25%relative uncertainty on S(Ei)

    All datasets except for Ref. [12] are assigned a uni- form fractional uncertainty δi = 0.25,(1) corresponding to a25%relative uncertainty on S(Ei). This choice reflects the typical normaliza- tion uncertainties quoted for the majority of direct and indirect measurements and avoids the intro- duction of dataset-specific tuning parameters

  2. [2]

    [12] are pro- vided in the form of adopted central valuesSad i (Ei) together with an energy-dependentS-factor band defined by lower and upper boundsS lo i (Ei)and Shi i (Ei)

    The inverse-kinematics data of Ref. [12] are pro- vided in the form of adopted central valuesSad i (Ei) together with an energy-dependentS-factor band defined by lower and upper boundsS lo i (Ei)and Shi i (Ei). A relative uncertainty is assigned as δi = max Shi i (Ei)−S lo i (Ei) 2S ad i (Ei) ,0.05 ,(2) imposing a minimum5%uncertainty in order to avoid an...

  3. [3]

    Narrow resonances appear as localized deviations from this background and therefore do not bias the extracted global trend

  4. [4]

    The quadratic model is applied over arestricted energy interval (1.5–3.5 MeV), within which the smooth envelope oflog10 S∗(E)varies slowly

  5. [5]

    The polynomial parametrization is not intended as a physical model of the reaction mechanism; it serves as a local regression that approximates the smooth background trend

  6. [7]

    The inclusion of higher-order polynomial terms does not improve the Bayesian evidence and leads to additional coefficients that are statistically in- significant

  7. [8]

    Robustness checks confirm that the extracted value ofS(1.5 MeV)is stable under cubic or quartic ex- tensions of the polynomial, with deviations well be- low the quoted credible intervals. III. BAYESIAN FORMALISM A. Bayes’ theorem The statistical inference problem addressed in this work is formulated within a Bayesian framework. Given a set of experimental...

  8. [9]

    Moriet al., MNRAS Lett.482, L70 (2019) arXiv:1810.01025 (2018)

    K. Moriet al., MNRAS Lett.482, L70 (2019) arXiv:1810.01025 (2018)

    Pith/arXiv arXiv 2019

  9. [10]

    Chieffi, S

    A. Chieffi, S. Courtin, R. James deBoeret al., Eur. Phys. J. A (In Press) (2025)

    2025

  10. [11]

    M. G. Mazarakis, W. E. Stephens, Phys. Rev. C7, 1280 (1973). (doi.org/10.1103/PhysRevC.7.1280)

    work page doi:10.1103/physrevc.7.1280 1973

  11. [12]

    Becker, K.U

    H.W. Becker, K.U. Kettner, C. Rolfset al., Z. Phys. A 303, 305 (1981), (doi.org/10.1007/BF01421528)

    work page doi:10.1007/bf01421528 1981

  12. [13]

    Spillane, F

    T. Spillane, F. Raiola, C. Rolfs, Phys. Rev. Lett.98, 122501, (2007) (doi:10.1103/PhysRevLett.98.122501)

    doi:10.1103/physrevlett.98.122501 2007

  13. [14]

    Taniguchi, M

    Y. Taniguchi, M. Kimura, Phys. Lett. B823, 136790 (2021), (doi.org/10.1016/j.physletb.2021.136790)

    arXiv 2021

  14. [15]

    W. P. Tan, A. Boeltzig, C. Dulalet al., Phys. Rev. Lett. 124, 192702 (2020) (doi:10.1103/PhysRevLett.124)

    doi:10.1103/physrevlett.124 2020

  15. [16]

    W. P. Tan, A. Gula, K. Lee, Phys. Rev.C,110, 035808 (2024) (doi:10.1103/PhysRevC.110.035808).192702)

    work page doi:10.1103/physrevc.110.035808).192702 2024

  16. [17]

    has also found no compelling evidence for a fusion hindrance pattern at low energies, further underscoring the uncertainties associated with extrapolations based on specific hindrance assumptions. The central challenge is therefore to combine these het- erogeneous datasets in a statistically consistent manner, accounting for their different energy coverag...

    Pith/arXiv arXiv 2025

  17. [18]

    Fruet, S

    G. Fruet, S. Courtin, M. Heineet al., Phys. Rev. Lett.124, 192701 (2020) (doi:10.1103/PhysRevLett.124.192701)

    doi:10.1103/physrevlett.124.192701 2020

  18. [19]

    Nippert, S

    J. Nippert, S. Courtin, M. Heine,et al., arXiv: 2509.14717 (2025)

    arXiv 2025

  19. [20]

    Tumino ,C

    A. Tumino ,C. Spitaleri, M.LaCognata,et al., Nature, 557, 687 (2018) (doi:10.1038/s41586-018-0149-4)

    doi:10.1038/s41586-018-0149-4 2018

  20. [21]

    W. Nan, Y. Wang, J. Suet al., Phys. Lett. B862, 139341 (2025) (doi: 10.1016/j.physletb.2025.139341)

    arXiv 2025

  21. [22]

    Mukhamedzhanov, D.Y

    A.M. Mukhamedzhanov, D.Y. Pang, A.S. Kady- rovet al., Phys. Rev. C99, 064618 (2019) (doi.org/10.1103/PhysRevC.99.064618)

    work page doi:10.1103/physrevc.99.064618 2019

  22. [23]

    Taniguchi, M

    Y. Taniguchi, M. Kimura, Phys. Lett. B849, 138434 (2024) (doi.org/10.1016/j.physletb.2023.138434)

    arXiv 2024

  23. [24]

    C. L. Jiang, H. Esbensen, K. E. Rehm,et al.Phys. Rev. Lett.89, 052701 (2002). (doi: 10.1103/Phys- RevLett.89.052701)

    doi:10.1103/phys- 2002

  24. [25]

    Uzawa and K

    K. Uzawa and K. Hagino, Phys. Rev. C112, L061601 (2025). (doi: 10.1103/PhysRevC.112.L061601)

    work page doi:10.1103/physrevc.112.l061601 2025

  25. [26]

    A. M. Mukhamedzhanov Phys Rev C (in Press) (2025); arXiv

    2025

  26. [27]

    J. R. Patterson, H. Winkler, C. S. Zaidins, Astrophys. J. 157, 367 (1969)

    1969

  27. [28]

    Caughlan, W.A

    G.R. Caughlan, W.A. Fowler, At. Data Nucl. Data Ta- bles40, 283 (1988) (doi: 10.1016/0092-640X(88)90009-5)

    doi:10.1016/0092-640x(88)90009-5 1988

  28. [29]

    Gasques⁄tieacce♪tlowercaseet al., Phys

    L.R. Gasques⁄tieacce♪tlowercaseet al., Phys. Rev. C76, 035802 (2007) (doi: 10.1103/PhysRevC.76.035802)

    doi:10.1103/physrevc.76.035802 2007