Pith. sign in

REVIEW 4 major objections 5 minor 50 references

Search for a Dark Gauge Boson Within Einstein-Cartan Theory at the ILC Using Multivariate Analysis

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Machine-learning analysis finds 8 of 10 dark-boson points discoverable

desk verdict A solid MVA sensitivity projection for an Einstein-Cartan dark gauge boson at the ILC, but the quoted significances rest on an undefined A'->mu+mu- coupling; worth refereeing with a required revision. read the letter →

arxiv 2507.08678 v1 pith:KU5SAQIU submitted 2025-07-11 hep-ex

Hossam Taha , El-sayed A. El-dahshan , S. Elgammal This is my paper · ORCID
classification hep-ex
keywords darkgaugebosonEinstein-CartantheorymatterInternationalLinearCollidermultivariateanalysisboosteddecisiontreemuonicdecaychannelmissingtransverseenergy
topics Dark Matter
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper argues that a dark gauge boson called $A'$, predicted by an Einstein-Cartan extension of the Standard Model, can be discovered at a 500 GeV electron-positron collider through its decay into a muon pair. After training a boosted decision tree on nine kinematic variables, eight of ten benchmark mass points reach statistical significances between 5.87 and 19.33 $\sigma$ with 500 fb$^{-1}$ of data. The reachable region covers $A'$ masses from 50 to 80 GeV with the heavy torsion mediator at 1.5 or 2 TeV. For the most favorable point a 5-$\sigma$ discovery would need only about 33 fb$^{-1}$, so the search could succeed early in a collider run.

What carries the argument

The carrying object is the signal topology $e^+e^- \to ST \to \chi\bar{\chi} \to \chi\bar{\chi} A' \to \chi\bar{\chi} \mu^+\mu^-$, in which a heavy scalar torsion field $ST$ produces two dark fermions $\chi$, one of which radiates $A'$ that decays to a muon pair. The discriminating engine is a boosted decision tree trained on nine observables: missing transverse energy, di-muon invariant mass, di-muon angular separation, azimuthal angle between the di-muon system and missing momentum, pseudorapidities and transverse momenta of the two muons, and the cosine of the 3D angle between the di-muon vector and missing momentum. The di-muon invariant mass has the largest separating power of any single variable.

What would settle it

Search for a di-muon resonance plus missing transverse energy in 500 fb$^{-1}$ of 500 GeV electron-positron collisions; if no excess appears at invariant masses from 50 to 80 GeV, the paper's discovery claim for the benchmark points is contradicted.

Watch

Extended reading notes

Core claim

Within the Einstein-Cartan model, the paper's central claim is that the dark gauge boson $A'$ is visible above Standard Model backgrounds in the muon-pair-plus-missing-energy final state. Using a boosted decision tree, the analysis quotes discovery significances at 500 fb$^{-1}$ of 19.33, 17.07, 14.42, 11.20, 8.72, 7.01, and 5.87 $\sigma$ for the benchmark points with $A'$ masses of 50, 60, 70, 80 GeV at mediator mass 1.5 TeV and 50, 60, 70 GeV at mediator mass 2 TeV. The two points with $A'$ at 90 GeV, where the di-muon invariant mass sits close to the $Z$ pole, fall below 5 $\sigma$, and the heaviest point (90 GeV $A'$, 2 TeV mediator) is effectively unreachable.

Load-bearing premise

The entire signal rate rests on unpublished model implementation files and on the assumption that $A'$ decays to muon pairs with the implied branching fraction; if either is wrong, every quoted cross-section and significance changes.

Editorial extensions

If this is right

  • If the central claim is correct, a 500 GeV electron-positron collider can probe Einstein-Cartan dark matter in a mass range that hadron colliders struggle to reach.
  • The required luminosity for 5-sigma discovery ranges from about 33 to 363 fb$^{-1}$ across the eight discoverable points, so most of the parameter space opens up before the full 500 fb$^{-1}$ dataset is collected.
  • $A'$ masses near the $Z$ mass are the difficult corner of the model, needing several times more luminosity or new discriminating variables to reach 5 sigma.
  • Before the boosted decision tree cut the signal is entirely hidden; the multivariate step is what actually makes the search sensitive.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper leaves the $A'$-muon kinetic-mixing term unspecified and the model implementation files unavailable, so re-deriving the cross-sections from the stated Lagrangian would be a direct check of whether the quoted significances survive.
  • The same classifier pipeline could be exported to other proposed lepton colliders, and the accessible $A'$ mass window would likely shift with center-of-mass energy while the analysis logic stays unchanged.
  • If the dark fermion mass were lowered so that $A'$ could decay invisibly, the clean muon-pair tag would disappear and the search would have to rely on missing-energy-only signatures such as mono-photon events.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper presents a Monte Carlo study of the process e+e− → χχA′, A′ → μ+μ− at the ILC (√s = 500 GeV, L = 500 fb−1) within an Einstein-Cartan theory model with a torsion mediator. The authors generate SM backgrounds (Drell-Yan, WW, ZZ, t-tbar) and ten signal benchmark points with M_ST = 1500/2000 GeV and M_A′ = 50–90 GeV, apply Fisher, DNN, and BDT classifiers using nine kinematic variables, and report statistical significances between 1.22σ and 19.33σ after an optimized BDT cut. They conclude that most benchmark points are discoverable within 500 fb−1 and provide required luminosities for 5σ discovery.

Significance. If the model input and significance calculation were self-contained, this would be a useful projection for a low-mass dark gauge boson at a lepton collider. The MVA pipeline is conventional and reasonably documented: the paper gives a hyperparameter table, k-fold cross-validation, KS overtraining checks, ROC curves, and variable-ranking information. However, the central numerical claim currently rests on an unspecified A′→μμ branching ratio, on a statistical-only significance with an in-sample or unspecified cut-optimization procedure, and on an incomplete SM background list. These issues must be resolved before the quoted significances can be considered reliable; with the present text the discovery claim is not independently checkable.

major comments (4)
  1. [Section II, Eqs. (1)–(3) and Table II] No coupling of A′ to muons is defined. Equation (2) couples A′ only to the dark fermion χ through g_D, and Eq. (1) couples the torsion field to SM fermions; the kinetic mixing mentioned in the Introduction is never given a parameter, mass-mixing term, or numerical value. Table II nonetheless quotes σ×BR for A′→μ+μ−, and Table VI converts these numbers into discovery significances. Unless a kinetic-mixing coupling and the resulting BR(A′→μ+μ−) are specified, computed, and constrained, the signal rates and every significance in Table VI are not determined by the model as presented.
  2. [Section V, Eq. (6), and Table VI] The significance is computed as S = N_S / sqrt(N_S + N_B) using numbers of events after a BDT 'optimized cut,' but the paper does not describe how BDT_opt is selected or whether it is evaluated on an independent sample. If the cut is chosen to maximize the same quantity on the events that are then used for counting, the quoted significances are in-sample estimates and are biased upward. In addition, no systematic uncertainties are propagated into Eq. (6), despite the statement in Section III that systematic uncertainties were considered; the background counts after the BDT cut are as low as 70 events, so a modest systematic uncertainty or a background fluctuation would change several points by more than 1σ. The authors should use a likelihood-based or profile-likelihood significance, quantify all relevant systematic uncertainties, and validate the cut choice on a held-out sample.
  3. [Section III, Table I] The SM background list omits e+e− → τ+τ− with both taus decaying leptonically to muons (τ → μνν). This process has the same μ+μ− + missing transverse energy signature as the signal and is not included in the total background. At √s = 500 GeV its contribution after the muonic branching fractions is expected to be comparable to or larger than the listed ZZ background and non-negligible relative to the signal rates in Table II. The authors should add this process to the background samples or provide a quantitative argument for its negligibility.
  4. [Section III and Data Availability Statement] The ECT signal is generated with UFO files 'requested from the author of [13]' that are not included in the paper, and the Data Availability Statement says the manuscript has no associated data. Consequently the σ×BR values in Table II cannot be reproduced or cross-checked by a reader. The authors should provide the model files or, at minimum, a complete Lagrangian with all parameters, the derived BR(A′→μ+μ−), and a validation of the generator output against an independent calculation of the production cross-section.
minor comments (5)
  1. [Section II] The text calls S_μ the 'scalar torsion field (ST)' while Eq. (1) treats it as an axial-vector coupling (γ^μ γ^5 S_μ); the terminology is inconsistent and should be corrected.
  2. [Throughout] There are several typos and spacing issues, e.g. 'MV A' for MVA, 'psuedo-rapidity', 'proved discoverable' (should be 'shown to be discoverable'), and 'Finally' in the Table VI caption.
  3. [Figure 3] The y-axis labels such as '6.22 GeV / (1/N) dN' appear to be misplaced bin-width annotations; the axes should be labeled as normalized event counts per bin.
  4. [Figure 9] The legend entries 'S+B' and 'S/0' are not defined in the caption or text; presumably one is S/√(S+B), but this should be stated explicitly.
  5. [Section V] The statement that BMPs 5 and 9 may be reachable at √s = 1000 GeV and L = 1000 fb−1 is an extrapolation, since no simulation at that energy is presented; this should be phrased as a conjecture or supported by a dedicated study.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: the discovery projections follow from simulated event counts, with only a non-load-bearing self-citation.

full rationale

The paper's central claim is an expected discovery projection, not a derived first-principles result. The signal cross-sections in Table II are generated by WHIZARD from the ECT UFO implementation supplied by the author of [13]; the background rates in Table I are generated from SM processes; and the significances in Table VI are computed from Eq. (6) using the simulated event counts N_S and N_B that survive the BDT cut. No significance value is used to define g_eta, g_D, M_chi, M_ST, or M_A', and no fitted parameter is renamed as a prediction. The only self-citation is [32], the authors' previous fixed-mass A' = 10 GeV ILC study, which is used only to motivate the 50-90 GeV mass window; the numerical projection for the new benchmark points does not depend on [32] for its values. The absence of an explicit kinetic-mixing parameter and branching-ratio formula is a genuine reproducibility gap, because the A' -> mu+mu- rate is inherited from the UFO files rather than derived from the Lagrangian shown in Eqs. (1)-(2), but that is an unstated input, not a circular re-derivation. The BDT 'optimized cut' is selected on the simulated samples, so the quoted significances are in-sample optima; this is a statistical overfitting concern, not a circular equivalence. For these reasons, no circular step is identified; the non-load-bearing self-citation is the only reason the score is not zero.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central projection rests on model parameters inherited from Ref. [13] and the authors' own prior study, plus assumptions about background completeness and statistical treatment. No new entities are introduced by this paper; the A' and torsion field are carried over from prior work.

free parameters (6)
  • g_eta (lepton-ST coupling) = 0.125 for leptons; 0 for quarks
    Adopted from Ref. [13] for leptons and set to zero for quarks to sidestep LHC constraints; controls ST production in e+e- collisions (Eq. 1).
  • g_D (A'-DM fermion coupling) = 1.2
    Taken from Ref. [13]; controls A' bremsstrahlung from DM fermions (Eq. 2).
  • M_chi (DM fermion mass) = 90 GeV
    Chosen to satisfy M_A'<2 M_chi and to maximize the signal cross-section (Section II).
  • M_ST (torsion field mass) = 1500 GeV and 2000 GeV
    Benchmark values scanned to set the mediator mass; they determine the production cross-section scale.
  • M_A' (dark gauge boson mass) = 50 to 90 GeV in 10 GeV steps
    Benchmark points chosen above the CMS low-mass dilepton exclusion (11.5-45 GeV) and mostly below the Z mass to reduce overlap.
  • A' to mu+mu- branching ratio (kinetic mixing coupling) = not stated
    Table II quotes sigma x BR but no kinetic-mixing parameter or branching ratio value is given; the signal normalization depends on it.
assumptions (4)
  • domain assumption The UFO files implementing the ECT model, obtained from the author of Ref. [13], correctly encode the model interactions used in WHIZARD.
    Section III states the files were requested from the author; no public version or validation is provided.
  • domain assumption A' decays to mu+mu- with the branching ratio assumed in Ref. [13], mediated by an unspecified kinetic-mixing coupling.
    Section II lists only g_eta and g_D; no kinetic-mixing term or epsilon appears, yet the muonic signature is central.
  • domain assumption The SM background processes in Table I are complete for the mu+mu- plus missing energy signature.
    Tau+tau- production with leptonic tau decays is not listed and would produce the same final state.
  • ad hoc to paper Statistical uncertainty alone is sufficient to characterize the significance; systematic uncertainties are negligible.
    Section V uses S=NS/sqrt(NS+NB) without applying any systematic term, although the text says systematics were considered.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Search for a Dark Gauge Boson Within Einstein-Cartan Theory at the ILC Using Multivariate Analysis." pith.science (2026). https://pith.science/paper/KU5SAQIU

@misc{pith2026250708678,
  author       = {Pith},
  title        = {Pith review of: Search for a Dark Gauge Boson Within Einstein-Cartan Theory at the ILC Using Multivariate Analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KU5SAQIU}},
  note         = {Machine review of arXiv:2507.08678}
}
abstract

Multivariate analysis (MVA) is employed to probe the dark matter candidate A$^{\prime}$, a gauge boson of a model rooted within Einstein-Cartan Theory, at the International Linear Collider (ILC). \texttt{WHIZARD} package is used as the event generator to simulate electron-positron interactions at the ILC, at a 500 GeV center-of-mass energy ($\sqrt s$) and a 500 fb$^{-1}$ detector's integrated luminosity ($\mathcal{L}$), to produce the A$^{\prime}$ signal and the expected standard model background. The study focuses on the muonic decay channel of A$^{\prime}$, utilizing several MVA classifiers such as Fisher, Deep Neural Network (DNN), and the Boosted Decision Tree (BDT), aiming to discriminate between the signal of several benchmark points, within theoretical and experimental limits, and the standard model background, and to explore their discovery potential at the ILC. Most benchmark points were proved discoverable at the ILC within $\mathcal{L}= 500$ fb$^{-1}$.

Figures

Figures reproduced from arXiv: 2507.08678 by the authors.

Figure 1
Figure 1. The topology of the ECT signal - mediated by [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The missing transverse energy distributions [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The distributions of the nine input variables employed in the MVA testing processes for the SM [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: The separating power of each of the input [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Correlation matrices calculated via the TMVA for signal BMP6 and total SM background at √ s = 500 GeV. Positive and negative values represent correlation and anti-correlation, respectively. each process (background and signal). Though minimally correlated variables sho…
Figure 6
Figure 6. Figure 6: The ROC curves of the three classifiers for [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 8
Figure 8. Figure 8: The relative importance of each of the BDT [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 7
Figure 7. Figure 7: TMVA overtraining check for the three classifiers BDT in 7a, DNN in 7b and Fisher in 7c for SM background and signal BMP6, with MA′ set at 50 GeV, MST = 2000 GeV, Mχ = 90 GeV, with coupling constants of gD = 1.2 and gη = 0.125. formance of all three classifiers. Quanti…
Figure 9
Figure 9. Figure 9: The BDT cut efficiency for BMP6 versus total [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]
Figure 10
Figure 10. Figure 10: The discovery potential for all signal benchmark points used in this study before (in 10a) and after (in 10b) applying the BDT optimal cut. The dotted horizontal blue line refers to the statistical 5-σ value. of ST were not added as they also have L req 5σ > 1000 fb−1…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 46 canonical work pages

  1. [13]

    R. W. Tucker and C. Wang, Class. Quantum Grav. 15, 933 (1998)

    work page 1998

  2. [1]

    Emiss T , the missing energy in the transverse direc- tion

    work page

  3. [2]

    Mµ+µ− , The di-muon’s invariant mass

    work page

  4. [3]

    ∆ ϕµ+µ−, ⃗Emiss T , The azimuthal angle difference be- tween the di-muon system and the missing energy in the transverse direction

    work page

  5. [4]

    ∆ Rµ+µ− , the angular separation of the two muons

    work page

  6. [5]

    ηµ1 , the psuedo-rapidity of the first muon

    work page

  7. [6]

    ηµ2 , the psuedo-rapidity of the second muon

    work page

  8. [7]

    pµ1 T , the momentum of the first muon in the trans- verse direction

    work page

Show all 50 references
  1. [8]

    pµ2 T , the momentum of the second muon in the transverse direction

  2. [9]

    Figure 3 shows the normalized distributions of the in- put variables for the signal BMP6 versus all of the SM background, where the Y-axis represents the bin size

    cos(θ3D), the cosine of the 3D angle between the di-muon system vector and the missing transverse energy vector. Figure 3 shows the normalized distributions of the in- put variables for the signal BMP6 versus all of the SM background, where the Y-axis represents the bin size. ...

    2000

  3. [10]

    M. K. Gaillard, P. D. Grannis, and F. J. Sciulli, Rev. Mod. Phys. 71, S96 (1999)

    1999

  4. [11]

    Crivellin and B

    A. Crivellin and B. Mellado, Nat. Rev. Phys. 6, 294 (2024)

    2024

  5. [12]

    J. A. Frieman, M. S. Turner, and D. Huterer, Annu. Rev. Astron. Astrophys. 46, 385 (2008)

    2008

  6. [14]

    Clowe, M

    D. Clowe, M. Bradaˇ c, A. H. Gonzalez, et al., Astrophys. J. 648, L109 (2006)

    2006

  7. [15]

    Lage and G

    C. Lage and G. R. Farrar, J. Cosmol. Astropart. Phys. 02, 038 (2015)

    2015

  8. [16]

    Langacker, The Standard Model and Beyond (CRC Press, Boca Raton, FL, 2010)

    P. Langacker, The Standard Model and Beyond (CRC Press, Boca Raton, FL, 2010)

    2010

  9. [17]

    Holdom, Phys

    B. Holdom, Phys. Lett. 166B, 196 (1986)

    1986

  10. [18]

    Galison and A

    P. Galison and A. Manohar, Phys. Lett. 136B, 279 (1984)

    1984

  11. [19]

    T. W. B. Kibble, J. Math. Phys. 2, 212 (1961)

    1961

  12. [20]

    F. W. Hehl, P. von der Heyde, G. D. Kerlick, et al., Rev. Mod. Phys. 48, 393 (1976)

    1976

  13. [21]

    Shaposhnikov, A

    M. Shaposhnikov, A. Shkerin, I. Timiryasov, et al., Phys. Rev. Lett. 127, 169901 (2021)

    2021

  14. [22]

    C. H. Nam, Phys. Rev. D 105, 075015 (2022)

    2022

  15. [23]

    CMS Collaboration, Phys. Rev. D 97, 092005 (2018)

    2018

  16. [24]

    High Energy Phys

    ATLAS Collaboration, J. High Energy Phys. 10, 180 (2018)

    2018

  17. [25]

    CMS Collaboration, Eur. Phys. J. C 81, 13 (2021); Eur. Phys. J. C 81, 333 (2021)

    2021

  18. [26]

    High Energy Phys

    CMS Collaboration, J. High Energy Phys. 10, 073 (2017)

    2017

  19. [27]

    High Energy Phys

    ATLAS Collaboration, J. High Energy Phys. 02, 226 (2021)

    2021

  20. [28]

    High Energy Phys

    CMS Collaboration, J. High Energy Phys. 03, 025 (2020)

    2020

  21. [29]

    High Energy Phys

    ATLAS Collaboration, J. High Energy Phys. 11, 209 (2021)

    2021

  22. [30]

    High Energy Phys

    ATLAS Collaboration, J. High Energy Phys. 10, 013 (2021)

    2021

  23. [31]

    Cvetiˇ c and S

    M. Cvetiˇ c and S. Godfrey, in Electroweak Symmetry Breaking and New Physics at the TeV Scale, edited by T. Barklow, S. Dawson, H. E. Haber, and J. L. Siegrist (World Scientific, Singapore, 1996), p. 383

    1996

  24. [32]

    Leike, Phys

    A. Leike, Phys. Rep. 317, 143 (1999)

    1999

  25. [33]

    Cvetiˇ c, P

    M. Cvetiˇ c, P. Langacker, and B. Kayser, Phys. Rev. Lett. 68, 2871 (1992)

    1992

  26. [34]

    Dimopoulos and H

    S. Dimopoulos and H. Georgi, Nucl. Phys. B 193, 150 (1981)

    1981

  27. [35]

    High Energy Phys

    CMS Collaboration, J. High Energy Phys. 07, 208 (2021)

    2021

  28. [36]

    ATLAS Collaboration, Phys. Lett. B 796, 68 (2019)

    2019

  29. [37]

    Bambade, T

    P. Bambade, T. Barklow, T. Behnke, et al., arXiv:1903.01629

    1903 arXiv

  30. [38]

    Behnke, J

    T. Behnke, J. Brau, B. Foster, et al., arXiv:1306.6327

    arXiv

  31. [39]

    H. Baer, T. Barklow, K. Fujii, et al., arXiv:1306.6352

    arXiv

  32. [40]

    CMS Collaboration, Phys. Rev. Lett. 124, 131802 (2020)

    2020

  33. [41]

    H. Taha, E. A. El-Dahshan, and S. Elgammal, Nucl. 11 Phys. B 1015, 116898 (2025)

    2025

  34. [42]

    Kilian, T

    W. Kilian, T. Ohl, and J. Reuter, Eur. Phys. J. C 71, 1742 (2011)

    2011

  35. [43]

    Sj¨ ostrand, S

    T. Sj¨ ostrand, S. Mrenna, and P. Skands, J. High Energy Phys. 05, 026 (2006)

    2006

  36. [44]

    de Favereau, C

    J. de Favereau, C. Delaere, P. Demin, et al., J. High Energy Phys. 02, 057 (2014)

    2014

  37. [45]

    H¨ ocker, et al., arXiv:physics/0703039

    A. H¨ ocker, et al., arXiv:physics/0703039

    arXiv

  38. [46]

    Brun and F

    R. Brun and F. Rademakers, Nucl. Instrum. Methods Phys. Res., Sect. A 389, 81 (1997)

    1997

  39. [47]

    Bhardwaj, J

    A. Bhardwaj, J. Dutta, P. Konar, et al., J. High Energy Phys. 10, 083 (2020)

    2020

  40. [48]

    Refaeilzadeh, L

    P. Refaeilzadeh, L. Tang, and H. Liu, in Encyclopedia of Database Systems, edited by L. Liu and M. T. ¨Ozsu (Springer, Boston, MA, 2009)

    2009

  41. [49]

    Arlot and A

    S. Arlot and A. Celisse, Stat. Surv. 4, 40 (2010)

    2010

  42. [50]

    F. J. Massey, J. Am. Stat. Assoc. 46, 68 (1951)

    1951

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.