Heavy Neutrinos across the Electroweak-to-Multi-TeV Frontier via Novel ML-Enhanced Probes

arXiv:2601.09095 · hep-ph · Submitted 2026-01-14 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: High Energy Physics: Phenomenology. The day in particle phenomenology: models, signatures and constraints.

Kai: I'm Kai.

Mira: And I'm Mira, here with Lev.

Kai: We're looking at the paper: "Heavy Neutrinos across the Electroweak-to-Multi-TeV Frontier via Novel ML-Enhanced Probes".

Mira: Here's the gist of the paper. Heavy neutrinos in the mass range 50 GeV–10 TeV can be probed with sensitivity to the mixing parameter VlN 2 spanning from O10−5 to 1 <ref:2601.09095#pg4>.

Kai: This approach enhances the discovery potential for heavy neutrinos and provides a complementary pathway to existing search strategies <ref:2601.09095#pg4>.

Mira: Theoretical Framework The theoretical setup follows the spirit of the seesaw mechanism and gauge invariance with terms involving right-handed Majorana neutrino fields <ref:2601.09095#pg3>.

Kai: After spontaneous electroweak symmetry breaking, the Dirac mass matrix (MD) arises from Yukawa interactions, leading to light and heavy mass eigenstates <ref:2601.09095#pg3>.

Paper summary: Kai: To wrap up this paper, what is the real significance of this work called "Heavy Neutrinos across the Electroweak-to-Multi-TeV Frontier via Novel ML-Enhanced Probes"?

Mira: The core message is that by combining inclusive heavy neutrino production with a machine learning search, we can significantly extend the discovery reach at hadron colliders.

Lev: Specifically, this methodology improves sensitivity in both the lower and high mass regimes while also giving better coverage for third generation heavy neutrinos involving tau leptons.

Kai: So, to put it simply for someone who just listens to the show, it means we have a new set of tools that helps us search for these heavy neutrinos with much higher confidence at the LHC.

Mira: The projected sensitivity reaching O ten minus five for mN equals fifty GeV is a key result here, and they confirm this reach remains competitive even up to the multi-TeV scale <ref:2601.09095#pg2>.

Lev: They successfully explore scenarios where contributions from different charged lepton flavors are combined, which allows us to test for potential lepton universality violation.

Kai: It really shows that this kind of analysis strategy complements what we're already doing in the field and opens up promising opportunities to explore physics beyond the standard model in the neutrino sector.

Conclusion: Kai: So, we've been talking about this new paper on heavy neutrinos using machine learning to find them at the LHC.

Mira: Yeah, this one is called "Heavy Neutrinos across the Electroweak-to-Multi-TeV Frontier via Novel ML-Enhanced Probes."

Kai: It’s about using a new way to look for these heavy neutrinos that spans from electroweak scales all the way up to multi-TeV energies.

Lev: The authors are focusing on proton proton collisions at energy level thirteen point six TeV and looking at final states with a charged lepton, missing transverse energy, and two jets.

Mira: What they did was use gradient boosted decision trees to classify these events, which is a machine learning method that can capture complex relationships in the data.

Kai: They found that the most important features for separating signal from background were things like transverse masses involving missing transverse energy and charged leptons.

Lev: That’s interesting because it tells us where we should focus our search to get the best signal, which is crucial when you're dealing with real hardware limitations.

Mira: They projected sensitivities reaching a level of one part in ten to the power of five for heavy neutrino masses starting at fifty giga-electron volts.

Kai: That level of sensitivity is quite aggressive, and they show it’s still competitive all the way up to multi-TeV scales.

Lev: I gotta ask, how does this actually translate to running on real detectors? Because we're talking about these complex ML models needing a lot of training data.

Mira: The paper includes details on how they built those signal samples, using sixteen different heavy neutrino mass points from fifty giga-electron volts to ten tera electron volts.

Kai: They used a lot of events for tuning the hyperparameters, and then held back some for final validation on the models.

Lev: It sounds like a solid plan, but I wonder about the systematic uncertainties they mentioned—they said that all those factors together could add around twenty percent uncertainty.

Mira: That's a realistic estimate, given how much they had to account for things like lepton and hadron identification efficiencies being reduced.

Kai: So, what does this overall strategy mean for us right now? It suggests a new way to look for physics beyond the standard model in the neutrino sector at colliders.

Lev: It’s a complementary probe, which means it doesn't just replace what we're doing; it adds something different to our search toolkit.

Mira: And because they looked at different charged lepton flavors, they can actually test for potential lepton universality violation in this framework.

Kai: So the conclusion is that combining inclusive production with machine learning opens up a much wider discovery reach at hadron colliders.

Department of Physics and Astronomy, Vanderbilt University · Department of Physics, Universidad de los Andes

hep-ph

Submitted: 2026-01-14

Updated: 2026-11-11

DOI: 10.1140/epjc/s10052-026-16572-5

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

Importance score: 83/100

The gist: Heavy neutrinos in the mass range 50 GeV–10 TeV can be probed with sensitivity to the mixing parameter VlN 2 spanning from O10−5 to 1 <ref:2601.09095#pg4>.

Key concepts

Heavy Neutrinos
These are hypothetical right-handed Majorana neutrinos with masses between 50 GeV and 10 TeV. They arise in extensions of the Standard Model, such as the seesaw mechanism, and are sought at colliders by looking for specific decay signatures involving leptons and jets.
Machine Learning (BDTs)
Gradient-boosted decision trees are used to classify events into signal or background. These models analyze 31 kinematic features of the event, such as jet momenta and lepton properties, to capture complex, non-linear relationships that distinguish heavy neutrino signals from Standard Model backgrounds.
Mixing Parameter |VℓN|2
This parameter quantifies the coupling strength between a light neutrino and a heavy right-handed neutrino. The sensitivity of the search is directly related to how precisely this mixing parameter can be constrained, with projected limits reaching O10−5 for certain heavy neutrino masses.
Production Mechanisms (s/t channels)
The heavy neutrinos are produced via different processes at the LHC, including s-channel resonance production and vector boson fusion (VBF) in the t-channel. The analysis accounts for these distinct production modes to cover a wide range of heavy neutrino masses.

Terminology

Summary

Heavy neutrinos in the mass range 50 GeV–10 TeV can be probed with sensitivity to the mixing parameter VlN 2 spanning from O10−5 to 1 <ref:2601. p. 0

Where the paper says it

We simulate both signal and Standard Model background events and employ gradient-boosted decision trees to optimize event classification. Assuming an integrated luminosity of 3000 fb−1 , expected for the high-luminosity, and considering realistic statistical and systematic uncertainties, we find that heavy neutrinos in the mass range 50 GeV–10 TeV can be probed with sensitivity to the mixing parameter |VℓN | 2 spanning from O 10−5 to 1.

09095#pg4>. This approach enhances the discovery potential for heavy neutrinos and provides a complementary pathway to existing search strategies <ref:2601. p. 0
Where the paper says it

Assuming an integrated luminosity of 3000 fb−1 , expected for the high-luminosity, and considering realistic statistical and systematic uncertainties, we find that heavy neutrinos in the mass range 50 GeV–10 TeV can be probed with sensitivity to the mixing parameter |VℓN | 2 spanning from O 10−5 to 1. This approach enhances the discovery potential for heavy neutrinos and provides a complementary pathway to existing search strategies.

09095#pg4>.

Theoretical Framework

The theoretical setup follows the spirit of the seesaw mechanism and gauge invariance with terms involving right-handed Majorana neutrino fields <ref:2601.09095#pg3>. After spontaneous electroweak symmetry breaking, the Dirac mass matrix (MD) arises from Yukawa interactions, leading to light and heavy mass eigenstates <ref:2601. p. 1

Where the paper says it

The number of heavy right-handed neutrinos depends on whether the lightest neutrino is massless or not [59, 60]. After spontaneous electroweak symmetry breaking, the Dirac mass matrix (MD)ij for neutrinos arises from the Yukawa interactions.

09095#pg3>. The weak interaction terms are given by expressions involving mixing parameters Uli and Vlj <ref:2601. p. 1
Where the paper says it

The corresponding weak interaction terms are given by L ∋ − g √ 2 (U ∗ ℓiν¯i + V ∗ ℓjN¯ j )W/ + PLℓ + h. − g 2 cos θW (U ∗ ℓiν¯i + V ∗ ℓjN¯ j )ZP/ L(Uℓmνm + VℓnNn), (3) where g is the gauge coupling constant of the SU(2)L group, Uℓi (Vℓj ) are the mixing parameters between the weak interaction eigenstates and the light (heavy) mass eigenstates, θW is the weak-mixing angle, and PL is the left-handed projection operator.

09095#pg3>. Scenarios with heavy Majorana neutrinos can also provide viable explanations for the persistent R(D(∗)) anomalies in semileptonic B-meson decays <ref:2601. p. 1
Where the paper says it

Since our projected sensitivity can reach |Vℓj | ∼ 1 in specific mass ranges, we perform a comprehensive analysis of the full |Vℓj | phase space, extending up to unity. We note that scenarios with heavy Majorana neutrinos can also provide viable explanations for the persistent R(D(∗) ) anomalies in semileptonic B-meson decays [65–67].

09095#pg4>.

Signal and Background Simulation

Signal and background events are simulated using FeynRules, MadGraph5 aMC@NLO 3.5.6, and NNPDF2. p. 2

Where the paper says it

6 [74, 75], and MadWidth [76]. The NNPDF2.

3 LO parton distribution function set for proton–proton collision energy fixed at √s = 13. p. 2
Where the paper says it

The NNPDF2. 3 LO parton distribution function set [77] is employed for event generation, with the proton–proton collision energy fixed at √ s = 13.

6 TeV <ref:2601. p. 2
Where the paper says it

The three curves converge near mN ≃ 1 TeV, beyond which the VBF contribution becomes dominant. Our numerical results are consistent with Ref.

09095#pg4>. The signal samples are generated via pure electroweak processes pp → lνjj at order O g4 <ref:2601. p. 2
Where the paper says it

The parameter VµN denotes the corresponding mixing parameter. Signal samples are generated via pure electroweak processes pp → ℓνjj at order O g 4 .

09095#pg4>. The SM background cross section for the inclusive process pp → lν + ≥2j is calculated to be 1888 pb <ref:2601. p. 3
Where the paper says it

We therefore simulate the inclusive SM process pp → ℓν + ≥ 2j, including QCD contributions, which effectively captures both the V + jets and diboson components. The resulting background cross section is 1888 pb.

09095#pg4>.

Machine Learning Application

Gradient-boosted decision trees (BDTs) are employed to optimize event classification, as they can capture complex nonlinear relationships in the input data <ref:2601.09095#pg5>. The analysis pipeline relies on a dedicated MadAnalysis expert-mode C++ framework to process simulated event samples and extract kinematic observables <ref:2601. p. 4

Where the paper says it

Specifically, we used the XGBClassifier with the objective function set to binary: logistic, which models the probability of a binary outcome using logistic regression. Our analysis pipeline relies on a dedicated MadAnalysis expert-mode C++ framework designed to process the simulated event samples and extract the relevant kinematic and event-level observables.

09095#pg5>. Each model was trained with 31 kinematic features, including the transverse momentum (pT), longitudinal momentum (pz), azimuthal angle (ϕ), and pseudorapidity (η) of the two jets and the charged lepton <ref:2601. p. 4
Where the paper says it

All jet pairs satisfying these criteria are subsequently combined to construct dijet candidates. Each model was trained with 31 kinematic features: the transverse momentum (pT), longitudinal momentum (pz), azimuthal angle (ϕ), and pseudorapidity (η) of the two jets and the charged lepton; the missing transverse energy (Emiss T ); the azimuthal angle of the transverse missing energy; the azimuthal angle separation (∆ϕ), pseudorapidity separation (∆η), and angular separation (∆R) between the two jets and between the missing transverse energy and the charged lepton; the pseudorapidity separation between the leading jet and the charged lepton; the transverse mass MT of each visible object and missing transverse energy; visible transverse energy ET; hadronic transverse energy HT; the invariant mass (M) of the dijet system and charged lepton; and the relative transverse momentum between the leading jet and the charged lepton, defined as (pT[j1] − pT[ℓ])/pT[j1].

09095#pg5>.

Sensitivity and Results

The relative feature importance analysis shows that transverse masses involving missing transverse energy and charged leptons provide key discriminating power <ref:2601.09095#pg8>. For the case of l = τ, the BDT model efficiency in the τ channel is reduced compared to light lepton channels due to less precise kinematic reconstruction <ref:2601. p. 5

Where the paper says it

However, because hadronic τ decays involve additional neutrinos in the final state, the kinematic reconstruction is less precise, which degrades the overall discriminating power of the input variables. As a result, the BDT model efficiency in the τ channel is reduced compared to the light lepton channels.

09095#pg9>. The expected limits on the mixing parameter VlN 2 for mN = 50 GeV reach O10−5, extending beyond previous searches. p. 8
Where the paper says it

10 through FIG. 12, the sensitivity to the mixing parameter |VℓN | 2 reaches values as low as O 10−5 for a heavy neutrino mass of mN = 50 GeV, extending beyond the reach of previous searches such as those performed by DELPHI [27].

Conclusion

An analysis strategy that combines inclusive heavy neutrino production—encompassing both s channel and vector boson fusion mechanisms—with a machine-learning-based search significantly extends the discovery reach at hadron colliders <ref:2601.09095#pg11>. This methodology complements existing search strategies and highlights promising opportunities to explore physics beyond the SM in the neutrino sector. p. 9

Where the paper says it

10 and high-mass regimes and markedly improves coverage of third-generation heavy neutrinos involving τ leptons, an experimentally challenging sector where existing constraints remain comparatively weak. This methodology complements existing search strategies and highlights promising opportunities to explore physics beyond the SM in the neutrino sector.

The projected sensitivity reaches O10−5 for mN = 50 GeV and remains competitive up to the multi-TeV scale. p. 8
Where the paper says it

Using the CMS Combine tool, we obtained expected upper limits on the mixing parameter |VℓN | 2 at each mass point, assuming an integrated luminosity of 3000 fb−1 at the HL-LHC. The projected sensitivity reaches O 10−5 for mN = 50 GeV and remains competitive up to the multi-TeV scale.

The analysis successfully explores scenarios where contributions from different charged-lepton flavors are combined, enabling tests of potential lepton universality violation <ref:2601. p. 8
Where the paper says it

We further explored scenarios where contributions from different charged-lepton flavors are combined, enabling tests of potential lepton-universality violation and providing a more inclusive framework for heavy neutrino searches. In conclusion, an analysis strategy that combines inclusive heavy neutrino production—encompassing both s channel and vector boson fusion mechanisms—with a machine-learning-based search significantly extends the discovery reach at hadron colliders.

p. 8
Where the paper says it

We further explored scenarios where contributions from different charged-lepton flavors are combined, enabling tests of potential lepton-universality violation and providing a more inclusive framework for heavy neutrino searches. In conclusion, an analysis strategy that combines inclusive heavy neutrino production—encompassing both s channel and vector boson fusion mechanisms—with a machine-learning-based search significantly extends the discovery reach at hadron colliders.

09095#pg4>.

Pruning redundant reasoning tokens

The paper proposes a new strategy to probe heavy neutrinos with non-universal fermion couplings at the LHC using a novel production mechanism and machine-learning algorithms <ref:2601.09095#pg4>. Focusing on proton–proton collisions at √s = 13. p. 0

Where the paper says it

Received: / Accepted: Abstract We propose a new strategy to probe heavy neutrinos with non-universal fermion couplings at the Large Hadron Collider (LHC) using a novel production mechanism and machine-learning algorithms. Focusing on proton–proton collisions at √ s = 13.

p. 0
Where the paper says it

Received: / Accepted: Abstract We propose a new strategy to probe heavy neutrinos with non-universal fermion couplings at the Large Hadron Collider (LHC) using a novel production mechanism and machine-learning algorithms. Focusing on proton–proton collisions at √ s = 13.

6 TeV, they investigate final states containing a charged lepton, missing transverse energy, and two jets <ref:2601. p. 0
Where the paper says it

Focusing on proton–proton collisions at √ s = 13. 6 TeV, we investigate final states containing a charged lepton, missing transverse energy, and two jets.

p. 0
Where the paper says it

Focusing on proton–proton collisions at √ s = 13. 6 TeV, we investigate final states containing a charged lepton, missing transverse energy, and two jets.

09095#pg4>.

How the training signal is built

Signal samples are generated for 16 heavy neutrino mass points, ranging from 50 GeV to 10 TeV, with varying step sizes <ref:2601.09095#pg5>. For each mass point, an additional 0. p. 4

Where the paper says it

A separate model was trained for each heavy neutrino mass, using one million signal and background events per mass point. An additional 0.

5 million events were used for hyperparameter tuning via Optuna <ref:2601. p. 4
Where the paper says it

An additional 0. 5 million events were used for hyperparameter tuning via Optuna [97] with 200 trials, and a further 0.

09095#pg5>. A further 0.5 million events are reserved for final model validation <ref:2601. p. 4
Where the paper says it

5 million events were used for hyperparameter tuning via Optuna [97] with 200 trials, and a further 0. 5 million events are reserved for final model validation.

09095#pg5>.

Where the gains come from

The feature-importance analysis shows that transverse masses involving missing transverse energy and charged leptons provide key discriminating power <ref:2601. p. 8

Where the paper says it

By employing modern machine-learning methods, we trained BDT models on an extensive set of kinematic features, achieving strong discrimination between signal and background. The feature-importance analysis shows that transverse masses involving missing transverse energy and charged leptons provide key discriminating power.

09095#pg8>. For lower heavy neutrino masses, a sharp peak appears in the range M(j1j2) ∼ 100 GeV, which originates from resonant heavy neutrino production via the s channel <ref:2601. p. 4
Where the paper says it

6, for lower heavy neutrino masses, a sharp peak appears in the range M(j1j2) ∼ 100 GeV. This feature originates from resonant heavy neutrino production via the s channel, followed by its decay into an on-shell W or Z boson that subsequently decays into a dijet system.

09095#pg7>. For larger heavy neutrino masses, the t channel VBF process becomes dominant, leading to a smoother M(j1j2) distribution with larger invariant masses typical of VBF processes <ref:2601. p. 4
Where the paper says it

This feature originates from resonant heavy neutrino production via the s channel, followed by its decay into an on-shell W or Z boson that subsequently decays into a dijet system. For larger heavy neutrino masses, the t channel VBF process becomes dominant, leading to a smoother M(j1j2) distribution with larger invariant masses typical of VBF processes [54, 55, 57, 98–109].

09095#pg7>. The results for mN = 50 GeV reach O10−5, extending beyond the reach of previous searches such as those performed by DELPHI. p. 8
Where the paper says it

10 through FIG. 12, the sensitivity to the mixing parameter |VℓN | 2 reaches values as low as O 10−5 for a heavy neutrino mass of mN = 50 GeV, extending beyond the reach of previous searches such as those performed by DELPHI [27].

The remainder of this paper is organized as follows <ref:2601.09095#pg4>. In Section 2, they briefly review the seesaw mechanism and define the relevant model parameters <ref:2601. p. 1

Where the paper says it

The remainder of this paper is organized as follows. In Section 2, we briefly review the seesaw mechanism and define the relevant model parameters.

09095#pg4>. In Section 3, they describe the generation of signal samples for various heavy neutrino masses, together with the background samples, and present the associated cross sections <ref:2601. p. 1
Where the paper says it

In Section 2, we briefly review the seesaw mechanism and define the relevant model parameters. In Section 3, we describe the generation of signal samples for various heavy neutrino masses, together with the background samples, and present the associated cross sections.

09095#pg4>. In Section 4, they explain the application of BDTs to event classification and discuss the resulting performance <ref:2601. p. 1
Where the paper says it

In Section 3, we describe the generation of signal samples for various heavy neutrino masses, together with the background samples, and present the associated cross sections. In Section 4, we explain the application of BDTs to event classification and discuss the resulting performance.

09095#pg4>. In Section 5, they present the expected sensitivity contours and uncertainty bands for the mixing parameters derived from the BDT analysis under different scenarios <ref:2601. p. 1
Where the paper says it

In Section 4, we explain the application of BDTs to event classification and discuss the resulting performance. In Section 5, we present the expected sensitivity contours and uncertainty bands for the mixing parameters derived from the BDT analysis under different scenarios.

09095#pg4>.

The authors note that in principle, the signal process may interfere with the SM background, for example, when the heavy neutrino is replaced by a light neutrino <ref:2601.09095#pg4>. However, they have evaluated the interference effects on the kinematic distributions over a wide range of heavy neutrino masses and find them to be negligible <ref:2601. p. 2

Where the paper says it

In principle, the signal process may interfere with the SM background, for example, when the heavy neutrino is replaced by a light neutrino. However, we have evaluated the interference effects on the kinematic distributions over a wide range of heavy neutrino masses and find them to be negligible.

09095#pg4>. They also verify that variations in VlN do not affect the shapes of the kinematic distributions <ref:2601. p. 2
Where the paper says it

However, we have evaluated the interference effects on the kinematic distributions over a wide range of heavy neutrino masses and find them to be negligible. In addition, we verify that variations in VℓN do not affect the shapes of the kinematic distributions.

09095#pg4>.

The authors adopt conservative performance assumptions by allowing for a modest reduction in lepton and hadron identification efficiencies <ref:2601.09095#pg5>. The efficiency for reconstructing charged hadrons is taken to be approximately 97% in the pseudorapidity range η. p. 3

Where the paper says it

[83] and correspond to an average of 140 pileup interactions per bunch crossing. The efficiency for reconstructing charged hadrons, which directly impacts both jet/tau reconstruction and the determination of missing transverse momentum, is taken to be approximately 97% in the pseudorapidity range |η| < 1.

Statistical uncertainties associated with the finite size of the simulated signal and background samples are included on a bin-by-bin basis in the BDT output histograms <ref:2601.09095#pg9>. These uncertainties range from approximately 1% to 35%, depending on the process and the specific region of the BDT discriminant <ref:2601. p. 7

Where the paper says it

Finally, statistical uncertainties associated with the finite size of the simulated signal and background samples are included on a bin-by-bin basis in the BDT output histograms. These uncertainties range from approximately 1% to 35%, depending on the process and the specific region of the BDT discriminant, and are treated as uncorrelated across both processes and bins.

09095#pg9>. All systematic effects are combined, resulting in an overall impact conservatively estimated to be of order 20% <ref:2601. p. 6
Where the paper says it

This contribution is taken to be uncorrelated between signal and background processes and independent of the BDT output bin. When all systematic effects are combined, the resulting overall impact is conservatively estimated to be of order 20%.

09095#pg9>.

The relationship between the cross section and the mixing parameter varies substantially across the mass range mN ∈ [50 GeV, 10 TeV], so they generate additional samples for different mixing parameter values within VlN 2 ∈ [0, 1] <ref:2601.09095#pg9>. This fitted relation enables them to extract the sensitivity to the mixing parameter VlN 2 at the HL-LHC with an integrated luminosity of 3000 fb−1 <ref:2601. p. 7

Where the paper says it

Because the relationship between the cross section and the mixing parameter varies substantially across the mass range mN ∈ [50 GeV, 10 TeV], we generate additional samples for different mixing parameter values within |VℓN | 2 ∈ [0, 1] and fit the resulting cross sections using the parameterization σ = a|VℓN | 2 + b|VℓN | 4 , (4) where a and b are free parameters determined independently at each mass point. This fitted relation enables us to extract the sensitivity to the mixing parameter |VℓN | 2 at the HL-LHC with an integrated luminosity of 3000 fb−1 .

09095#pg9>. The results for l = τ and l = µ are shown in FIG. p. 7
Where the paper says it

This fitted relation enables us to extract the sensitivity to the mixing parameter |VℓN | 2 at the HL-LHC with an integrated luminosity of 3000 fb−1 . The results for ℓ = τ and ℓ = µ are shown in FIG.

10 and FIG. p. 1
Where the paper says it

We focus on the final state ℓνjj, where ℓ and ν denote a charged lepton and a (light) neutrino, respectively, and jj denotes two jets. Representative Feynman diagrams contributing to signal processes are shown in Fig.

11, respectively <ref:2601. p. 0
Where the paper says it

At even higher masses, the LHC remains the primary facility for probing heavy neutrinos: in the subTeV region, they can be resonantly produced and their decay products fully reconstructed, whereas at larger masses the production cross section becomes increasingly suppressed by phase space [52, 53]. In this regime, t channel mechanisms such as vector boson fusion (VBF) offer a more robust alternative, as their cross section dearXiv:2601.

09095#pg9>. For light-lepton final states, the proposed strategy can extend the discovery reach at the HL-LHC and offers a complementary probe to existing LHC search methodologies. p. 8
Where the paper says it

In particular, we demonstrate that a novel VBF production mechanism mediated by virtual t channel heavy neutrino exchange, when combined with advanced machine-learning–based analysis techniques, can lead to a substantial enhancement in sensitivity to third-generation heavy neutrino searches involving tau leptons–an experimentally challenging regime where current constraints remain weak. For light-lepton final states, the proposed strategy can extend the discovery reach at the HL-LHC and offers a complementary probe to existing LHC search methodologies.

In conclusion, an analysis strategy that combines inclusive heavy neutrino production—encompassing both s channel and vector boson fusion mechanisms—with a machine-learning-based search significantly extends the discovery reach at hadron colliders. This approach simultaneously improves sensitivity in both the lo and high-mass regimes and markedly improves coverage of third-generation heavy neutrinos involving τ leptons, an experimentally challenging sector where existing constraints remain comparatively weak. p. 9

Where the paper says it

10 and high-mass regimes and markedly improves coverage of third-generation heavy neutrinos involving τ leptons, an experimentally challenging sector where existing constraints remain comparatively weak. This methodology complements existing search strategies and highlights promising opportunities to explore physics beyond the SM in the neutrino sector.

The methodology complements existing search strategies and highlights promising opportunities to explore physics beyond the SM in the neutrino sector <ref:2601. p. 9
Where the paper says it

10 and high-mass regimes and markedly improves coverage of third-generation heavy neutrinos involving τ leptons, an experimentally challenging sector where existing constraints remain comparatively weak. This methodology complements existing search strategies and highlights promising opportunities to explore physics beyond the SM in the neutrino sector.

09095#pg4>. The approach extends the discovery reach at the HL-LHC and offers a complementary probe to existing LHC search methodologies. p. 8
Where the paper says it

In particular, we demonstrate that a novel VBF production mechanism mediated by virtual t channel heavy neutrino exchange, when combined with advanced machine-learning–based analysis techniques, can lead to a substantial enhancement in sensitivity to third-generation heavy neutrino searches involving tau leptons–an experimentally challenging regime where current constraints remain weak. For light-lepton final states, the proposed strategy can extend the discovery reach at the HL-LHC and offers a complementary probe to existing LHC search methodologies.

p. 8
Where the paper says it

In particular, we demonstrate that a novel VBF production mechanism mediated by virtual t channel heavy neutrino exchange, when combined with advanced machine-learning–based analysis techniques, can lead to a substantial enhancement in sensitivity to third-generation heavy neutrino searches involving tau leptons–an experimentally challenging regime where current constraints remain weak. For light-lepton final states, the proposed strategy can extend the discovery reach at the HL-LHC and offers a complementary probe to existing LHC search methodologies.

The results for mN = 50 GeV reach O10−5 and remain competitive up to the multi-TeV scale <ref:2601. p. 8
Where the paper says it

Using the CMS Combine tool, we obtained expected upper limits on the mixing parameter |VℓN | 2 at each mass point, assuming an integrated luminosity of 3000 fb−1 at the HL-LHC. The projected sensitivity reaches O 10−5 for mN = 50 GeV and remains competitive up to the multi-TeV scale.

09095#pg9>. The analysis successfully explores scenarios where contributions from different charged-lepton flavors are combined, enabling tests of potential lepton universality violation <ref:2601.09095#pg4>. The approach extends the discovery reach at the HL-LHC and offers a complementary probe to existing LHC search methodologies.

The authors thank Umar Sohail Qureshi for discussions on the application of machine learning techniques <ref:2601.09095#pg4>. They gratefully acknowledge support from Vanderbilt University and the U. p. 9

Where the paper says it

Acknowledgements We thank Umar Sohail Qureshi for discussions on the application of machine learning techniques. We gratefully acknowledge support from Vanderbilt University and the U.

S. National Science Foundation <ref:2601. p. 9
Where the paper says it

We gratefully acknowledge support from Vanderbilt University and the U. National Science Foundation.

09095#pg4>.

The authors have proposed a new strategy to probe heavy neutrinos with non-universal fermion couplings at the LHC using a novel production mechanism and machine-learning algorithms <ref:2601.09095#pg4>. Focusing on proton–proton collisions at √s = 13.6 TeV, they investigate final states containing a charged lepton, missing transverse energy, and two jets <ref:2601.09095#pg4>.

The cross sections of all samples are taken directly from the MadGraph5 aMC@NLO output, and calculated at LO in QCD <ref:2601.09095#pg4>. All remaining background processes listed above are found to be subdominant and do not contribute appreciably to their analysis <ref:2601. p. 3

Where the paper says it

The resulting background cross section is 1888 pb. All remaining background processes listed above are found to be subdominant and do not contribute appreciably to our analysis.

09095#pg5>. The resulting background cross section is 1888 pb <ref:2601. p. 3
Where the paper says it

We therefore simulate the inclusive SM process pp → ℓν + ≥ 2j, including QCD contributions, which effectively captures both the V + jets and diboson components. The resulting background cross section is 1888 pb.

09095#pg5>.

The BDT models were trained on an extensive set of kinematic features, achieving strong discrimination between signal and background <ref:2601.09095#pg5>. The feature-importance analysis shows that transverse masses involving missing transverse energy and charged leptons provide key discriminating power <ref:2601.09095#pg8>.

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. Improved AI system can perform novel heavy neutrino searches at hadron colliders by combining inclusive production (s channel) and vector boson fusion (VBF) mechanisms with machine learning classifiers. This allows the system to cover a broad mass range from 50 GeV to 10 TeV and maintain sensitivity in the multi-TeV regime, as shown by the cross sections dominating different mass regimes.

  2. The improved system can utilize gradient-boosted decision trees (BDTs) trained on 31 kinematic features to achieve strong discrimination between signal and background events. This capability is enhanced by exploiting correlations and higher-order dependencies among input variables, leading to a more robust classification boundary than traditional cut-based analyses.

  3. The system can be extended to search for heavy neutrinos coupling predominantly to third-generation leptons, such as tau leptons, using hadronic tau decays. This provides a complementary probe in an experimentally challenging sector where current constraints remain weak by focusing on the mixing parameter VtauN squared.

  4. The AI system can fit the cross section dependence on the mixing parameter using a parameterized relation sigma equals aVlN squared plus bVlN 4, determined independently at each mass point. This enables precise extraction of sensitivity to the mixing parameter at the high-luminosity LHC with an integrated luminosity of 3000 fb-1.

  5. The system can combine searches across different charged-lepton flavors by implementing a combination where VmuN squared is replaced by VeN squared plus VmuN squared and combining the channels when the heavy neutrino mass mN is within a certain range, allowing for tests of potential lepton universality violation.

Abstract

We propose a new strategy to probe heavy neutrinos with non-universal fermion couplings at the Large Hadron Collider (LHC) using a novel production mechanism and machine-learning algorithms. Focusing on proton--proton collisions at sqrt s = 13.6 TeV, we investigate final states containing a charged lepton, missing transverse energy, and two jets. For heavy neutrino masses below O(1 TeV), production is dominated by the s channel process. At higher masses, vector boson fusion becomes the dominant production mechanism, with cross sections that decrease slowly as the heavy neutrino mass increases. We simulate both signal and Standard Model background events and employ gradient-boosted decision trees to optimize event classification. Assuming an integrated luminosity of 3000 fb-1, expected for the high-luminosity, and considering realistic statistical and systematic uncertainties, we find that heavy neutrinos in the mass range 50 GeV -- 10 TeV can be probed with sensitivity to the mixing parameter V N squared spanning from O(10-5) to 1. This approach enhances the discovery potential for heavy neutrinos and provides a complementary pathway to existing search strategies.

Sources

Related papers