Search for heavy neutral leptons in top quark decays with same-sign, same-flavor dilepton final states in proton-proton collisions at sqrt s = 13 TeV

arXiv:2611.04904 · hep-ex · Submitted 2026-11-03 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: High Energy Physics: Experiment. The day in high-energy experiment: detectors, searches and limits.

Kai: We're looking at the paper: "Search for heavy neutral leptons in top quark decays with same-sign, same-flavor dilepton final states in proton-proton collisions at sqrt s = 13 TeV".

Mira: Here's the gist of the paper. A search was conducted for a Majorana-type heavy neutral lepton (HNL) produced in top quark decays using proton-proton collision data at sqrt s = 13 TeV, with an integrated luminosity of 138 fb-1 collected between 2016 and 2018 by the CMS detector at the LHC.

Kai: This search is motivated by the type-I seesaw model, which allows a Majorana HNL to mediate lepton-number-violating top quark decays.

Mira: The specific search targets top quark pair production events where one top quark decays via t to bW+ to b+ N, and the subsequent decay of the HNL into a W boson and a charged lepton with the same sign and flavor as the produced lepton, leading to an experimental signature characterized by same-sign, same-flavor lepton pairs accompanied by jets.

Kai: The search probes HNL masses in the range between 20 and 100 GeV.

Paper summary: Kai: This paper describes a search conducted using proton-proton collision data at ten tera electron volt, with an integrated luminosity of one hundred thirty-eight femtobarns, collected between two thousand sixteen and two thousand eighteen by the CMS detector at CERN. The thesis is that they are searching for this HNL signature.

Mira: The search targets those specific top quark pair production events we just talked about, looking for the same-sign, same-flavor lepton pairs accompanied by jets as the experimental signature.

Lev: They use a multivariate classifier to separate the actual signal events from all that background noise they expect to see in the data.

Kai: And what they found is that there was no significant excess over what their background prediction expected them to see at all. It didn't show any clear signal for this HNL.

Mira: Because of that, the paper sets some upper limits on the branching fraction for top quark decay into a bottom quark, a W boson, and a charged lepton.

Lev: Specifically, they set limits between one point one times ten to the power of minus five and four point eight times ten to the power of minus five in the dielectron channel.

Kai: And then they have similar limits for the dimuon channel, setting them between three point seven times ten to the power of minus six and two point four times ten to the power of minus five.

Mira: They also looked at limits on things like V mu N squared in the dimuon channel, which ranged from four point two times ten to the power of minus five up to three point nine times ten to the power of minus two.

Lev: These observed upper limits are consistent with what they expected from just the background alone, which is pretty standard for a search like this.

Kai: So it confirms that in this specific search channel, they didn't find any evidence for the HNL at these mass ranges between twenty and one hundred gigaelectron volts.

Mira: This gives us a sense of how constrained these searches are right now, especially considering the type-I seesaw model that predicts this physics.

Conclusion: Kai: So, looking at the title, "Search for heavy neutral leptons in top quark decays with same-sign, same-flavor dilepton final states in proton-proton collisions at sqrt s = thirteen TeV", it really tells you exactly what they were trying to find <ref:2611.04904#pg1,Search for heavy neutral leptons in top quark decays with same-sign>.

Mira: It frames the entire effort around that specific experimental setup: those top decays leading to same-sign, same-flavor lepton pairs.

Lev: The implication is that even with a large dataset like one hundred thirty-eight femtobarns of data collected over two years, they haven't found a direct signal for this heavy neutral lepton in the twenty to one hundred gigaelectron volt mass range.

Kai: This means we’ve set tighter constraints on the branching fraction for those specific top decays, which is a key piece of information for model building.

Mira: The results show upper limits on things like V e N squared and V mu N squared, which are the squared mixing amplitudes between the HNL and the standard model electron and muon neutrinos.

Lev: The observed limits on those mixing amplitudes are between sixteen point five times ten to the power of minus five up to eleven point two times ten to the power of minus two for electron neutrino mixing, and four point two times ten to the power of minus five up to three point nine times ten to the power of minus two for muon neutrino mixing.

Kai: So what this changes for us is that it puts strong limits on how strongly these heavy neutral leptons might mix with the standard model neutrinos.

Mira: It tells us that if the type-I seesaw model is correct, then the mixing parameters must be relatively small within these probed mass ranges.

Lev: The paper also points out something important about their search strategy, noting that for masses between twenty and seventy five gigaelectron volts, their limits on V e N squared and V mu N squared are about two times stronger than those from the ATLAS search for HNLs produced in top quark decays.

Kai: And they also noted that the expected results for this search channel are comparable to existing dilepton channel results but weaker by about a factor of ten compared to trilepton channel results when the HNL mass is smaller than the W boson mass.

Mira: So, while they didn't find anything, they did put these upper bounds on the parameters that describe how this new physics connects to our known neutrino sector.

CMS Collaboration

European Organization for Nuclear Research (CERN)

hep-ex

Submitted: 2026-11-03

Updated: 2026-11-03

Comments: Submitted to the Journal of High Energy Physics. All figures and tables can be found at http://cms-results.web.cern.ch/cms-results/public-results/publications/B2G-24-023 (CMS Public Pages)

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 72/100

The gist: A search was conducted for a Majorana-type heavy neutral lepton (HNL) produced in top quark decays using proton-proton collision data at sqrt s = 13 TeV…

Key concepts

Majorana Heavy Neutral Lepton (HNL)
A hypothetical heavy neutral lepton that is its own antiparticle. In this search, it is produced in top quark decays and then decays into a W boson and a charged lepton, creating the same-sign dilepton signature used to identify it.
Type-I Seesaw Model
A theoretical model suggesting that the existence of heavy neutral leptons is related to neutrino masses. This model allows for lepton-number violating processes, such as top quark decays into an HNL and a charged lepton, which is the specific mechanism being investigated here.
Same-Sign, Same-Flavor Dilepton Signature
An experimental signature where two leptons of the same charge (e.g., two electrons or two muons) and the same flavor are observed in the final state. This specific pattern is a key experimental handle used to distinguish potential signal events from Standard Model backgrounds.
Branching Fraction ($ ext{B}(t ightarrow b u N)$)
The probability that a top quark will decay into a bottom quark, a W boson, and the heavy neutral lepton (HNL) instead of the expected Standard Model decay. The search sets upper limits on this probability based on the observed data.

Terminology

Summary

A search was conducted for a Majorana-type heavy neutral lepton (HNL) produced in top quark decays using proton-proton collision data at sqrt s = 13 TeV, with an integrated luminosity of 138 fb-1 collected between 2016 and 2018 by the CMS detector at the LHC. p. 0

Where the paper says it

EUROPEAN ORGANIZATION FOR NUCLEAR RESEARCH (CERN) CERN-EP-2026-262 2026/11/04 CMS-B2G-24-023 Search for heavy neutral leptons in top quark decays with same-sign, same-flavor dilepton final states in proton-proton collisions at √ s = 13 TeV The CMS Collaboration* Abstract A search for a Majorana-type heavy neutral lepton (HNL) produced in top quark decays is reported. The search is based on proton-proton collision data at √ s = 13 TeV, corresponding to an integrated luminosity of 138 fb−1 , collected during 2016–2018 with the CMS detector at the CERN LHC.

This search is motivated by the type-I seesaw model, which allows a Majorana HNL to mediate lepton-number-violating top quark decays. p. 0
Where the paper says it

The search is based on proton-proton collision data at √ s = 13 TeV, corresponding to an integrated luminosity of 138 fb−1 , collected during 2016–2018 with the CMS detector at the CERN LHC. In the type-I seesaw model, a Majorana HNL can mediate lepton-number-violating top quark decays.

The specific search targets top quark pair production events where one top quark decays via t to bW+ to b+ N, and the subsequent decay of the HNL into a W boson and a charged lepton with the same sign and flavor as the produced lepton, leading to an experimental signature characterized by same-sign, same-flavor lepton pairs accompanied by jets. p. 0
Where the paper says it

The search targets top quark pair production events in which one top quark decays into a bottom quark and a W boson, which subsequently decays into a charged lepton (electron or muon) and an HNL. The Majorana HNL then decays into a W boson and a charged lepton with the same sign and flavor as the one produced with the HNL.

The search probes HNL masses in the range between 20 and 100 GeV. p. 3
Where the paper says it

In this search, mN values between 20 and 100 GeV are considered. For masses below mW, the HNL decay width scales approximately as m5 N for a fixed mixing parameter, so a smaller mN results in a longer lifetime of the HNL [16].

The analysis employed a multivariate classifier to distinguish signal events from background based on their unique characteristics. p. 3

Where the paper says it

The production rate decreases rapidly with increasing mN values, and this analysis therefore restricts the search to mN < 100 GeV. A machine-learning technique is used to classify events as background or signal, based on their distinguishing characteristics.

No significant excess over the background prediction was observed in this search channel. p. 0
Where the paper says it

The data agree with the standard model background prediction. Observed (expected) upper limits at 95% confidence level are set on the branching fraction of the top quark decay into an HNL, a charged lepton, and a bottom quark, varying between [1.

The expected upper limits on the branching fraction for the top quark decay into an HNL, a charged lepton, and a bottom quark (B(t to b N)) were set between [1. p. 0
Where the paper says it

The data agree with the standard model background prediction. Observed (expected) upper limits at 95% confidence level are set on the branching fraction of the top quark decay into an HNL, a charged lepton, and a bottom quark, varying between [1.

1 (0.71)] times 10-5 and [4. p. 8
Where the paper says it

The selected data and background event sets for the two mass regimes can overlap. After the baseline selection, the signal acceptance times efficiency ranges from 0.

8 (3.0)] times 10-5 in the dielectron channel, and between [3. p. 8
Where the paper says it

After the baseline selection, the signal acceptance times efficiency ranges from 0. 54)%, depending on mN, in the electron (muon) channel.

7 (3.4)] times 10-6 and [2. p. 8
Where the paper says it

The selected data and background event sets for the two mass regimes can overlap. After the baseline selection, the signal acceptance times efficiency ranges from 0.

4 (1.3)] times 10-5 in the dimuon channel. p. 8
Where the paper says it

After the baseline selection, the signal acceptance times efficiency ranges from 0. 54)%, depending on mN, in the electron (muon) channel.

The observed upper limits on B(t to b N) varied between [7. p. 0
Where the paper says it

The data agree with the standard model background prediction. Observed (expected) upper limits at 95% confidence level are set on the branching fraction of the top quark decay into an HNL, a charged lepton, and a bottom quark, varying between [1.

1 (3.4)] times 10-6 to [3. p. 8
Where the paper says it

The selected data and background event sets for the two mass regimes can overlap. After the baseline selection, the signal acceptance times efficiency ranges from 0.

0 (1.3)] times 10-5 in the dielectron channel, and from 4. p. 8
Where the paper says it

After the baseline selection, the signal acceptance times efficiency ranges from 0. 54)%, depending on mN, in the electron (muon) channel.

2 times 10-5 to 3. p. 8
Where the paper says it

The selected data and background event sets for the two mass regimes can overlap. After the baseline selection, the signal acceptance times efficiency ranges from 0.

p. 8
Where the paper says it

The selected data and background event sets for the two mass regimes can overlap. After the baseline selection, the signal acceptance times efficiency ranges from 0.

9 times 10-2 for V mu N squared in the dimuon channel. p. 8
Where the paper says it

After the baseline selection, the signal acceptance times efficiency ranges from 0. 54)%, depending on mN, in the electron (muon) channel.

These observed limits are consistent with the background-only hypothesis. p. 0
Where the paper says it

The data agree with the standard model background prediction. Observed (expected) upper limits at 95% confidence level are set on the branching fraction of the top quark decay into an HNL, a charged lepton, and a bottom quark, varying between [1.

The search also focused on lepton number violating (LNV) processes in tt production, considering decay chains such as t to bW+ to b+ N to b++ W- to b++ ff', alongside the Standard Model decay t to bW-. The expected results are comparable to existing dilepton channel results and weaker by approximately a factor of ten compared to trilepton channel results for m N < m W. p. 3

Where the paper says it

In this search, mN values between 20 and 100 GeV are considered. For masses below mW, the HNL decay width scales approximately as m5 N for a fixed mixing parameter, so a smaller mN results in a longer lifetime of the HNL [16].

For the mass range 20 at most m N at most 75 GeV, the expected limits on V e N squared and V mu N squared from this search are approximately a factor of two stronger than those from the ATLAS search for HNLs produced in top quark decays. p. 2
Where the paper says it

The ATLAS Collaboration recently searched for Majorana HNLs produced in top quark decays in the same-sign (SS), same-flavor (SF) dilepton final state, using 140 fb−1 of data at √ s = 13 TeV, and set limits for HNL masses between 15 and 75 GeV [20]. In this article, we present a search for HNLs in top quark decays using SS, SF dilepton events in the pp collision data at √ s = 13 TeV, collected using the CMS detector between 2016 and 2018, corresponding to an integrated luminosity of 138 fb−1 .

The upper limits on the branching fractions for electron neutrino mixing (V e N squared) and muon neutrino mixing (V mu N squared) vary between [12 (5. p. 2
Where the paper says it

1 1 Introduction The observation of neutrino oscillations has confirmed the nonzero masses for at least two mass eigenstates of the standard model (SM) neutrinos (νℓ ) [1]. The mass-squared differences measured from neutrino oscillation experiments have set lower limits on the mass values of the two heavier neutrinos at 8.

6)] times 10-5 and [9. p. 8
Where the paper says it

The selected data and background event sets for the two mass regimes can overlap. After the baseline selection, the signal acceptance times efficiency ranges from 0.

9 (4.3)] times 10-2, respectively, for the expected limits, while the observed limits range from 16 times 10-5 to 11 times 10-2 for V e N squared and from 4. p. 0
Where the paper says it

The data agree with the standard model background prediction. Observed (expected) upper limits at 95% confidence level are set on the branching fraction of the top quark decay into an HNL, a charged lepton, and a bottom quark, varying between [1.

2 times 10-5 to 3.9 times 10-2 for V mu N squared. p. 2
Where the paper says it

1 1 Introduction The observation of neutrino oscillations has confirmed the nonzero masses for at least two mass eigenstates of the standard model (SM) neutrinos (νℓ ) [1]. The mass-squared differences measured from neutrino oscillation experiments have set lower limits on the mass values of the two heavier neutrinos at 8.

These results are interpreted as upper limits on the squared mixing amplitudes between the HNL and the Standard Model electron and muon neutrinos, denoted V e N squared and V mu N squared, respectively. p. 2
Where the paper says it

1 1 Introduction The observation of neutrino oscillations has confirmed the nonzero masses for at least two mass eigenstates of the standard model (SM) neutrinos (νℓ ) [1]. The mass-squared differences measured from neutrino oscillation experiments have set lower limits on the mass values of the two heavier neutrinos at 8.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Enhanced HNL Search Sensitivity: The analysis uses a multivariate classifier based on input variables such as m(l l) and angular separation of the two leptons to distinguish signal from background, which allows for an improved discrimination capability over simple selection cuts alone.

  2. Improved Background Modeling for Nonprompt Leptons: A matrix method is employed to estimate nonprompt-lepton backgrounds, utilizing a connection matrix M that links observed lepton ID configurations to underlying event origins, providing a more nuanced background estimation than traditional extrapolation.

  3. Systematic Uncertainty Quantification: The analysis incorporates various systematic uncertainties as nuisance parameters in the statistical fit, including uncertainties related to nonprompt-lepton backgrounds, Lepton charge mismeasurement backgrounds, and Experimental systematic uncertainties.

  4. Refined Signal Acceptance Modeling: The study uses a gradient boosted decision tree (BDT) technique trained on signal samples, which allows the system to search for the signal by inspecting partial decay products rather than requiring all final-state partons within acceptance, leading to a more robust search strategy.

  5. Parameter Limit Derivation: The final results are interpreted as upper limits on the squared mixing amplitudes VeN squared and VµN squared, providing direct constraints on the type-I seesaw model parameters based on the observed data.

Abstract

A search for a Majorana-type heavy neutral lepton (HNL) produced in top quark decays is reported. The search is based on proton-proton collision data at sqrt s = 13 TeV, corresponding to an integrated luminosity of 138 fb-1, collected during 2016 - 2018 with the CMS detector at the CERN LHC. In the type-I seesaw model, a Majorana HNL can mediate lepton-number-violating top quark decays. The search targets top quark pair production events in which one top quark decays into a bottom quark and a W boson, which subsequently decays into a charged lepton (electron or muon) and an HNL. The Majorana HNL then decays into a W boson and a charged lepton with the same sign and flavor as the one produced with the HNL. The experimental signature features a same-sign, same-flavor lepton pair and jets. The HNL masses between 20 and 100 GeV are probed. The data agree with the standard model background prediction. Observed (expected) upper limits at 95% confidence level are set on the branching fraction of the top quark decay into an HNL, a charged lepton, and a bottom quark, varying between [1.1 (0.71)] times 10-5 and [4.8 (3.0)] times 10-5 in the dielectron channel, and between [3.7 (3.4)] times 10-6 and [2.4 (1.3)] times 10-5 in the dimuon channel.

Sources

Related papers