A Window onto New Invisible Particles via Semi-Visible Higgs Decays

Arnab Roy; German Valencia; Sally Dawson

REVIEW 2 major objections 5 minor 16 references

Semi-visible Higgs decays could expose new invisible particles with masses below 50 GeV at the HL-LHC.

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

2026-08-02 07:35 UTC pith:NFGDR32S

load-bearing objection Useful projection for semi-visible Higgs decays, but the reach numbers are optimistic: the dominant ttbar background is treated as known to 1.5%, and the BDT improvement is delegated to a companion paper. the 2 major comments →

arxiv 2607.09344 v2 pith:NFGDR32S submitted 2026-07-10 hep-ph

A Window onto New Invisible Particles via Semi-Visible Higgs Decays

Sally Dawson , Arnab Roy , German Valencia This is my paper

classification hep-ph

keywords semi-visible Higgs decayinvisible particleseffective field theorydimension-six operatorsmissing transverse energyHL-LHCHiggs decaysdark matter

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 aims to show that the semi-visible Higgs decay modes H→ℓ+ℓ−+missing energy and H→jj+missing energy can serve as a window to new invisible particles with masses below about 50 GeV. Working in a dimension-six effective field theory that couples the Higgs to invisible scalars or fermions, the authors compute the signal and background for ZH production at the 14 TeV LHC. They find that with 3000 fb^-1 of HL-LHC data, a cut-based analysis reaches 3σ sensitivity to Wilson coefficients C_DHχχ ≲ 41 TeV^-2, C_DH∂φ ≲ 83 TeV^-2, and C_eφ ≲ 152 TeV^-2, and that a boosted decision tree significantly improves these limits. They further argue that the Standard Model 'Higgs neutrino floor' — the irreducible background from H→ZZ*→ℓ+ℓ−νν — is not an absolute barrier, because kinematic correlations allow it to be reduced. If this projection holds, the HL-LHC would probe a previously difficult parameter region and complement invisible-Higgs, monojet, and dark-matter searches.

Core claim

The paper's central claim is that the final states H→ℓ+ℓ−+E_T^miss and H→jj+E_T^miss, produced in association with a Z boson, are observable channels for new invisible scalars and fermions in the mass window m_φ,χ ≲ 50 GeV. Through a dimension-six EFT with operators coupling the Higgs to invisible scalar or fermion currents, and using a cut-based analysis followed by a boosted decision tree, the paper projects 3σ reaches at HL-LHC (14 TeV, 3000 fb^-1) of C_DHχχ ≲ 41 TeV^-2, C_DH∂φ ≲ 83 TeV^-2, and C_eφ ≲ 152 TeV^-2. It also demonstrates that the SM background H→ZZ*→ℓ+ℓ−νν, the so-called 'Higgs neutrino floor', is reducible through kinematic correlations, rather than an irreducible limit. The

What carries the argument

The central mechanism is a set of dimension-six effective operators: O_DHχχ=(iH†D↔μH)(χγμχ), O_DH∂φ=(H†D↔μH)(φ†∂↔μφ), and O_eφ=(ℓie_jH)φ†φ, which give the Higgs an off-shell coupling to a new invisible Dirac fermion χ or complex scalar φ. These generate the semi-visible decays in the ZH production mode, where the accompanying Z tags the event via its own decay. The key to signal extraction is the kinematic separation: the visible lepton pair or jet pair from the Higgs lies well below the Z mass, while the tagging Z pair sits near it, and a multivariate classifier adds angular variables. The paper's reach estimates rest on the ability of this machinery to suppress a background list that inclu

Load-bearing premise

The load-bearing premise is that the fast detector simulation, leading-order background cross-sections, and the unpublished BDT performance reproduce the actual HL-LHC response; if jet-energy-scale uncertainties, pile-up, or higher-order QCD corrections raise the post-cut background beyond the simulated 1.52 fb, the quoted 3σ reaches would not hold.

What would settle it

A decisive check is to measure in HL-LHC data, in a background-dominated control region defined by the paper's final cuts, the event rate for H→ℓ+ℓ−+MET; if the observed rate exceeds the simulated 1.52 fb by more than ~40%, the claimed 3σ reach to C_DHχχ ≲ 41 TeV^-2 would not stand. Alternatively, if the boosted decision tree's separation on the irreducible H→ZZ*→ℓ+ℓ−νν background is no better than the cut-based analysis, the claim that the 'Higgs neutrino floor' is reducible would be falsified.

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

If this is right

The semi-visible decays complement invisible-Higgs searches for new scalars and fermions with masses below about 50 GeV.
The Standard Model 'Higgs neutrino floor' is not an absolute barrier; kinematic selections and a multivariate classifier reduce it, so sensitivities below the SM rate are achievable.
For quark-type Yukawa operators, monojet searches remain stronger, but in multi-operator scenarios the two channels are complementary.
For thermal dark matter, the semi-visible reach is weaker than direct and indirect detection, yet it covers more general dark-matter models and parameter regions where several operators contribute.

Where Pith is reading between the lines

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

The same kinematic-discrimination logic would transfer to other Higgs production modes, such as vector-boson fusion, where the background composition differs and the reach for the hadronic operators might improve.
If the BDT gain is stable once systematic uncertainties are included, the effective coupling limits could approach the level where semi-visible decays become competitive with direct dark-matter detection in some operator combinations—an inference the paper does not draw.
The semi-visible framework generalizes to non-Standard Higgs-like resonances: any narrow scalar decaying to a visible pair plus invisible particles could be searched with the same invariant-mass windows, which the paper leaves implicit.
The paper treats each operator independently; a global fit combining semi-visible Higgs decays with Z-invisible-width and monojet bounds could exploit correlations between operators, but no such fit is performed here.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

Useful projection for semi-visible Higgs decays, but the reach numbers are optimistic: the dominant ttbar background is treated as known to 1.5%, and the BDT improvement is delegated to a companion paper.

read the letter

Colleague,

This is a conference-style proceedings paper on using H->ll+MET and H->jj+MET to look for light invisible scalars and fermions at the HL-LHC, in a DSMEFT framework. The genuinely new piece is the systematic extension of the operator set to the semi-visible channels, with two tagging modes, and the observation that the SM 'Higgs neutrino floor' in these modes is reducible rather than irreducible. The cut-based numbers are internally consistent: Eq. (4) follows from Table 2, and the paper is honest about the small-signal/large-background nature of the search. It even includes a footnote correcting a table from the original presentation.

The soft spots are real and not minor. The dominant background after cuts is ttbar-inclusive, 1.4 fb out of 1.52 fb total, and the significance is evaluated as S/sqrt(B) with B = 4560 events at 3000 fb^-1. No systematic uncertainty is attached. A 20% normalization error on ttbar alone makes the background error about 840 events, so the 207 signal events quoted for a 3sigma reach correspond to a significance below 1sigma. This is a known issue with pure Poisson projections, but it matters here because Eq. (4) carries the paper's quantitative claim. There is no control-region or profile-likelihood discussion, so the reader cannot check whether the ttbar rate can be pinned down. The BDT improvement is also only cited to ref. [11], not shown; a KS test mention is not enough to reproduce the claimed gain. The simulation is standard MadGraph/Pythia/Delphes with the CMS card, at LO, so additional 10-20% normalization uncertainties are plausible. The paper does not overclaim; it explicitly notes that the sensitivities are weaker than direct detection for thermal dark matter.

For whom is this useful? People working on Higgs portal EFT and LHC dark-sector searches. It is a reasonable starting point, but as it stands the reach numbers should not be taken at face value. A serious referee should ask for at least one of: a profile-likelihood treatment with a nuisance parameter for ttbar, a control-region estimate, or a clear statement that systematic uncertainties are omitted. If the BDT plot is in the companion paper, the authors should either reproduce it or give enough detail to evaluate it.

I would send it to peer review, but conditional on a revision that addresses the background systematic. The underlying idea is sound; the execution is incomplete.

Referee Report

2 major / 5 minor

Summary. This proceedings-style paper proposes a search for semi-visible Higgs decays H→ℓ⁺ℓ⁻+E/T and H→jj+E/T at the HL-LHC (√s=14 TeV, 3000 fb⁻¹), where the missing energy is carried by a new light invisible complex scalar or Dirac fermion described by a set of dimension-six DSMEFT operators. The authors use ZH production for tagging, simulate signal and backgrounds with MadGraph5/Pythia8/Delphes, and present a cut-based analysis yielding a total background σ_bkg=1.52 fb after all cuts. From this they quote a 3σ reach of C_DHχχ ≲ 41 TeV⁻², C_DH∂φ ≲ 83 TeV⁻², C_eφ ≲ 152 TeV⁻² (Eq. (4)). They further claim that a BDT classifier, whose details are deferred to the companion paper ref. [11], gives a significant improvement and makes the 'Higgs neutrino floor' reducible.

Significance. If the quoted reaches are robust, the paper addresses a genuinely useful gap: the kinematic window m_φ,χ ≲ 50 GeV for invisible particles produced in semi-visible Higgs decays is difficult to access through standard invisible-Higgs searches, and the explicit treatment of the SM neutrino background as a reducible rather than irreducible floor is an important conceptual point. The paper benefits from a clear EFT operator list, a simple and reproducible simulation workflow, and a transparent cut flow with explicit cross-sections in Table 2. Its main quantitative claims, however, rest on a statistical-only significance estimate and on BDT results imported from an accompanying paper. The central physics message is plausible, but the numerical centerpiece is not yet supported to the standard required for a journal publication.

major comments (2)

[Cut-based analysis, Table 2, Eq. (4)] The quoted reach is computed as S/√B with the background treated as known to the statistical accuracy of the Monte Carlo sample. Table 2 shows that after all cuts the t-tbar-inclusive background contributes 1.4 fb of the total 1.52 fb, i.e. 92% of σ_bkg. With 3000 fb⁻¹ this is B=4560 events, so the 207 signal events required for 3σ correspond to S/√B=3.07, which assumes the t-tbar background is known to about 1.5%. No systematic uncertainty, control-region constraint, or profile-likelihood treatment is described anywhere in the paper. A 20% normalization uncertainty on the t-tbar contribution alone gives a background uncertainty sqrt(4560+(0.2×4200)²)≈840 events, reducing the 207 signal events to about 0.25σ; even a 5% uncertainty gives less than 1σ. Since Eq. (4) is the only quantitative support for the abstract's central claim, the reach numbers must be recomputed with a realistic trea
[BDT section and Figure 3] The abstract and summary present the BDT analysis as an integral part of the result, but the BDT itself is not shown in this paper. The text states that the variable ranking and the signal significance as a function of the classifier output 'can be found in 11', and then quotes a final classifier cut at 0.997 with a 'significant improvement'. Figure 3, which is supposed to display the final BDT exclusions, is not self-contained: the axes are not labeled, the meaning of the shaded regions is not defined in the caption, and no numerical significance or systematic uncertainty is given. As a result, the BDT-based improvement cannot be checked from the manuscript alone. The authors should either include the essential BDT results (for example, the significance versus cut curve and the final reach table) or explicitly state that this is a proceedings summary and clearly mark all BDT-based numbe

minor comments (5)

[Table 2 footnote] The footnote 'This table has been updated from the original presentation to correct erroneous entries' is not explained. Please specify which entries were corrected and why, so that readers can assess the reliability of the table.
[Eqs. (1)-(3)] The operator notation is compact but omits flavor and gauge indices. Defining the fermion fields and the shorthand (e.g. l_i, q_i, u_j, d_j) would improve reproducibility, especially since the operators are dimension-six EFT operators.
[Table 2] The line 'N(/ b−jet) = 2 and b-jet vetoing' is garbled. It should read something like 'N(b-jet)=0 (b-jet veto)' or indicate the b-jet multiplicity requirement clearly.
[Figure 3 and Figure 4] Both figures lack full axis labels and numerical scales. In particular, Figure 3 needs labeled axes and a clear legend for the brown/blue regions, and Figure 4 should identify the curves corresponding to the different operators.
[General writing] There are several typographical issues, including 'TMV A. (Toolkit for Multivariate Analysis)' and inconsistent formatting of Wilson coefficients (e.g. C_eφ vs C_eφ). A final proofreading pass is needed.

Circularity Check

0 steps flagged

No construction-level circularity; the quoted reach is an independent MC-derived sensitivity, with only minor self-citations to a companion paper for BDT details.

full rationale

The central numerical claim, Eq. (4), is obtained by counting simulated signal and background events after cuts: Table 2 lists background cross-sections that sum to sigma_bkg = 1.52 fb, so B = 4560 events at 3000 fb^-1, and S = 207 events gives S/sqrt(B) ~ 3. The Wilson-coefficient reach is then the simple scaling C_reach = 10*sqrt(207/(sigma_sig(C=10)*3000)) using the signal rows of the same table. Nothing is fitted to the target limit; the coefficient is an input to the LO FeynRules/MadGraph/Delphes simulation, and the quoted numbers follow arithmetically. The absence of a systematic uncertainty on the dominant t-tbar background is a robustness/correctness concern, not a circular reduction, because the background is independently generated rather than tuned to make the signal appear. The paper does rely on its own companion paper [11] for the BDT variable ranking and significance-versus-output curves, and on ref. [10] (sharing an author) for DSMEFT definitions; however, the cut-based reach in Eq. (4) does not reduce to these citations, the resulting Figure 3 is shown in the paper, and the EFT formalism is also referenced to independent work [8,9]. I therefore find no step in which a 'prediction' is equivalent to its input by definition.

Axiom & Free-Parameter Ledger

3 free parameters · 5 axioms · 1 invented entities

The paper introduces no fitted numbers; the reach is computed from Monte Carlo simulation. The main bookkeeping is the operator set, the benchmark mass, and the unitarity-scale assumptions used in the comparison plots.

free parameters (3)

Mass benchmark m_φ,χ = 10 GeV = 10 GeV
The signal cross-sections in Table 2 and Figure 2 are computed for invisible masses of 10 GeV (with a few DM mass values for the BDT); the reach for other masses in the kinematic window m ≲ m_h/2 is not fully shown.
Classifier cut value for BDT = 0.997
The final BDT significance is obtained by a cut at 0.997 on the classifier output; this threshold is chosen post hoc to maximize significance.
Unitarity-limit new-physics scale = 3 TeV (or 1 TeV for Yukawa-like operators)
The comparison to unitarity limits assumes a new physics scale near 3 TeV (or 1 TeV for the Yukawa-like operators) and unitarity holding up to 1 TeV (or 300 GeV); these scales are assumptions, not derived.

axioms (5)

domain assumption The dimension-six DSMEFT operator basis in Eqs. (1)-(3) is the complete set of relevant SM-invisible interactions.
The paper selects these operators from refs. 8-10 and omits operators involving only the Higgs and the new invisible fields, arguing they are suppressed by light fermion masses and invisible-width constraints; completeness is assumed, not proven.
domain assumption The invisible new states are the only new particles below the electroweak scale.
This is the basis for the EFT validity; if other light states exist, the interpretation of missing energy and the operator mapping would change.
domain assumption The simulation chain (FeynRules/MadGraph/Pythia/Delphes) gives reliable signal and background normalizations at leading order.
The reach numbers are computed with LO matrix elements and fast detector simulation; no NLO k-factors or systematic uncertainties are applied.
domain assumption SM fermions in the final state can be treated as massless.
This is justified for e, mu, and light quarks, and it enters the derivation of the operator list and the decay kinematics.
domain assumption The invisible Z width constraints from refs. 8 and 11 are correct and apply to these operators.
For derivative operators these constraints are stronger than the projected semi-visible sensitivity, so the complementarity argument relies on them.

invented entities (1)

Light invisible complex scalar φ / Dirac fermion χ (with Z2-symmetric real/Majorana limits) no independent evidence
purpose: Source of missing transverse energy in H→ll+MET and H→jj+MET; the operators in Eqs. (1)-(3) couple them to the SM.
The paper does not propose a specific dark matter candidate with a new mass or coupling prediction; the states are generic EFT degrees of freedom, so there is no external falsifiable handle beyond the Higgs decay rates they predict.

pith-pipeline@v1.3.0-alltime-deepseek · 3866 in / 10637 out tokens · 109978 ms · 2026-08-02T07:35:46.229672+00:00 · methodology

0 comments

read the original abstract

Searches for new physics continue at the LHC in several forms, including new-particle searches, precision measurements of SM couplings, and searches for signals of new invisible particles. In this talk, we discuss the reach in parameter space of new invisible particles with the semi-visible Higgs decay modes $H\to \ell^+\ell^- + ~\rm E{\!\!\!/}_T$ and $H\to jj + ~\rm E{\!\!\!/}_T$. We first parametrise the new invisible particles and their interactions with the SM through an effective field theory at dimension six. We then study the respective signals and the corresponding background, finding small signals with large backgrounds. We find, however, that the kinematics of these processes are sufficiently rich to allow a signal extraction that we first quantify with a cut-based analysis and later with a multivariate BDT.

Review history (2 revisions) →

discussion (0)

Reference graph

Works this paper leans on

16 extracted references · 1 linked inside Pith

[1]

de Florian et al.CERN Yellow Rep

D. de Florian et al.CERN Yellow Rep. Monogr., 2:1–869, 2017

2017
[2]

David d’Enterria and Van Dung Le.J. Phys. G, 52(5):053001, 2025

2025
[3]

Navas et al.Phys

S. Navas et al.Phys. Rev. D, 110(3):030001, 2024

2024
[4]

Cepeda et al.CERN Yellow Rep

M. Cepeda et al.CERN Yellow Rep. Monogr., 7:221–584, 2019

2019
[5]

J. A. Aguilar-Saavedra, J. M. Cano, J. M. No, and D. G. Cerde˜ no.Phys. Rev. D, 106(11):115023, 2022

2022
[6]

Monoranjan Guchait and Arnab Roy.Phys. Rev. D, 102(7):075023, 2020

2020
[7]

A. Dey, V. Keus, S. Moretti, and C. Shepherd-Themistocleous.JHEP, 07:038, 2024

2024
[8]

J. C. Criado, A. Djouadi, M. Perez-Victoria, and J. Santiago.JHEP, 07:081, 2021

2021
[9]

Aebischer, W

J. Aebischer, W. Altmannshofer, E. E. Jenkins, and A. V. Manohar.JHEP, 06:086, 2022

2022
[10]

Xiao-Gang He, Xiao-Dong Ma, and German Valencia.JHEP, 03:037, 2023

2023
[11]

arXiv:2511.09778, 2025

Sally Dawson, Arnab Roy, and German Valencia. arXiv:2511.09778, 2025

Pith/arXiv arXiv 2025
[12]

Alloul, N

A. Alloul, N. D. Christensen, C. Degrande, C. Duhr, and B. Fuks.Comput. Phys. Commun., 185:2250–2300, 2014

2014
[13]

Alwall, R

J. Alwall, R. Frederix, S. Frixione, V. Hirschi, F. Maltoni, O. Mattelaer, H. S. Shao, T. Stelzer, P. Torrielli, and M. Zaro.JHEP, 07:079, 2014

2014
[14]

Skands.JHEP, 05:026, 2006

Torbjorn Sjostrand, Stephen Mrenna, and Peter Z. Skands.JHEP, 05:026, 2006

2006
[15]

de Favereau et al.JHEP, 02:057, 2014

J. de Favereau et al.JHEP, 02:057, 2014

2014
[16]

Schmidt, and German Valencia.JHEP, 02:167, 2026

Arnab Roy, Michael A. Schmidt, and German Valencia.JHEP, 02:167, 2026

2026

[1] [1]

de Florian et al.CERN Yellow Rep

D. de Florian et al.CERN Yellow Rep. Monogr., 2:1–869, 2017

2017

[2] [2]

David d’Enterria and Van Dung Le.J. Phys. G, 52(5):053001, 2025

2025

[3] [3]

Navas et al.Phys

S. Navas et al.Phys. Rev. D, 110(3):030001, 2024

2024

[4] [4]

Cepeda et al.CERN Yellow Rep

M. Cepeda et al.CERN Yellow Rep. Monogr., 7:221–584, 2019

2019

[5] [5]

J. A. Aguilar-Saavedra, J. M. Cano, J. M. No, and D. G. Cerde˜ no.Phys. Rev. D, 106(11):115023, 2022

2022

[6] [6]

Monoranjan Guchait and Arnab Roy.Phys. Rev. D, 102(7):075023, 2020

2020

[7] [7]

A. Dey, V. Keus, S. Moretti, and C. Shepherd-Themistocleous.JHEP, 07:038, 2024

2024

[8] [8]

J. C. Criado, A. Djouadi, M. Perez-Victoria, and J. Santiago.JHEP, 07:081, 2021

2021

[9] [9]

Aebischer, W

J. Aebischer, W. Altmannshofer, E. E. Jenkins, and A. V. Manohar.JHEP, 06:086, 2022

2022

[10] [10]

Xiao-Gang He, Xiao-Dong Ma, and German Valencia.JHEP, 03:037, 2023

2023

[11] [11]

arXiv:2511.09778, 2025

Sally Dawson, Arnab Roy, and German Valencia. arXiv:2511.09778, 2025

Pith/arXiv arXiv 2025

[12] [12]

Alloul, N

A. Alloul, N. D. Christensen, C. Degrande, C. Duhr, and B. Fuks.Comput. Phys. Commun., 185:2250–2300, 2014

2014

[13] [13]

Alwall, R

J. Alwall, R. Frederix, S. Frixione, V. Hirschi, F. Maltoni, O. Mattelaer, H. S. Shao, T. Stelzer, P. Torrielli, and M. Zaro.JHEP, 07:079, 2014

2014

[14] [14]

Skands.JHEP, 05:026, 2006

Torbjorn Sjostrand, Stephen Mrenna, and Peter Z. Skands.JHEP, 05:026, 2006

2006

[15] [15]

de Favereau et al.JHEP, 02:057, 2014

J. de Favereau et al.JHEP, 02:057, 2014

2014

[16] [16]

Schmidt, and German Valencia.JHEP, 02:167, 2026

Arnab Roy, Michael A. Schmidt, and German Valencia.JHEP, 02:167, 2026

2026