Machine Learning in the Hunt for Charged Higgs Bosons at Gamma-Gamma Colliders in the Type III Two Higgs Doublet Model
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: We're looking at the paper: "Machine Learning in the Hunt for Charged Higgs Bosons at Gamma-Gamma Colliders in the Type III Two Higgs Doublet Model".
Paper summary: Kai: So wrapping up this paper "Machine Learning in the Hunt for Charged Higgs Bosons at Gamma-Gamma Colliders in the Type III Two Higgs Doublet Model". The authors are using machine learning to hunt for charged Higgs bosons within a specific mass range of one hundred to one hundred ninety GeV at a three point zero tera electron volt gamma-gamma collider.
Mira: The implication is that these ML techniques can significantly improve the identification of these heavy charged Higgs bosons against Standard Model backgrounds, suggesting significant production prospects for the H plus particles.
Lev: It suggests that if we build a machine learning pipeline for real data from such a collider, we could see signals that are currently too buried in the noise to detect with traditional methods.
Kai: The work grounds itself in benchmark points that comply with experimental constraints, which lets them systematically account for uncertainties inherent in the analysis when they examine these decay channels.
Mira: It illuminates the interplay between the charged Higgs boson and the established Standard Model by looking at how it interacts through its decay channels, like top bottom and tau neutrino channels.
Lev: The paper highlights that while this is a theoretical exploration within Type III Two Higgs Doublet Model, it’s pointing toward specific experimental avenues for future searches at photon colliders.
Kai: Ultimately, the paper shows that using these advanced multivariate analysis techniques is key to uncovering what these charged Higgs bosons are and how they fit into particle physics beyond the Standard Model.
Conclusion: Kai: So we've been looking at how machine learning helps us find these charged Higgs bosons in photon collisions for this paper, "Machine Learning in the Hunt for Charged Higgs Bosons at Gamma-Gamma Colliders in the Type III Two Higgs Doublet Model."
Mira: Yeah, the authors are trying to use different machine learning tools like Boosted Decision Trees and Multilayer Perceptrons to separate those heavy charged Higgs signals from the messy Standard Model background.
Lev: From an error correction standpoint, they’re looking at how these classifiers perform when we feed them variables like jet mass or missing transverse energy. It’s about seeing if the signal stands out clearly enough for our actual hardware setup to handle.
Kai: The results show that these machine learning methods are quite effective at identifying the charged Higgs in this specific range, which is between one hundred and one hundred ninety gigaelectron volts at a three point zero tera electron volt photon collider.
Mira: What this really means is that even with the complexity of a Type III Two Higgs Doublet Model, we have a solid path forward for searching for these particles if we can build the right detection systems.
Lev: It tells us that the production prospects aren't just theoretical possibilities anymore; there’s a way to filter through the noise and get a clean sample of what we’re looking for.
Kai: The implication is that these searches could give us real clues about what these exotic Higgs bosons are actually like, beyond just their mass and how they interact.
Mira: It moves us from just postulating possibilities to having a data-driven way to probe the fundamental structure of our universe.
Lev: But we still have the challenge of building detectors capable of measuring those precise kinematic variables like jet pseudorapidity separation and azimuthal angle separation accurately enough for these models to hold up.
Kai: That’s right, so if we can get that precision, this machine learning approach gives us a really strong tool to make those measurements meaningful.
Mira: And that brings us to the next thing we need to think about—what kind of physical constraints this actually puts on the Higgs sector itself.
Federal Urdu University of Arts, Science and Technology, Islamabad Pakistan · Sang-Ho College, Konkuk University, Seoul South Korea · Facultad de Ciencias Físico-Matemáticas and Centro Interdisciplinario de Investigación y Enseěanza de la Ciencia (CIIEC), Benemérita Universidad Autonoma de Puebla
hep-ph
Submitted: 2025-04-09
Updated: 2026-11-17
Importance score: 83/100
The gist: Machine Learning in the Hunt for Heavy Charged Higgs Bosons at Gamma-Gamma Colliders in the Type III Two Higgs Doublet Model is a detailed exploration of charged Higgs boson masses within a specific…
Key concepts
- Type III Two Higgs Doublet Model (2HDM-III)
- This is an extension of the Standard Model that introduces two scalar Higgs doublets (H1 and H2) instead of one. This specific model has unique ways that these new particles interact with fermions, which is crucial for understanding how the charged Higgs boson behaves and how it can be found in collisions.
- Charged Higgs Boson (H±)
- This is a hypothetical heavy particle predicted by the 2HDM-III. The study looks for this particle by looking at specific decay channels, such as those involving tau leptons or top quarks. Detecting these particles would provide direct evidence for physics beyond the Standard Model.
- Boosted Decision Tree (BDT)
- A machine learning algorithm that creates a decision structure similar to a tree. It learns complex relationships between various input variables (like jet energy and momentum) to classify events as either signal or background, making it highly effective for separating rare signals from noise.
- Multilayer Perceptron (MLP)
- A neural network architecture that uses multiple layers of interconnected nodes to process data. In this analysis, the MLP is used as a classifier alongside BDT to improve the identification accuracy of the heavy charged Higgs boson by learning non-linear patterns in the collision data.
Terminology
Summary
Machine Learning in the Hunt for Heavy Charged Higgs Bosons at Gamma-Gamma Colliders in the Type III Two Higgs Doublet Model is a detailed exploration of charged Higgs boson masses within a specific range using machine learning methods to improve identification compared to Standard Model backgrounds at a 3. p. 00 TeV gamma-gamma collider Our results indicate significant production prospects for the charged Higgs, which could provide essential insights into the properties of H± bosons The research uses machine learning methods like Boosted Decision Trees (BDT) and Multilayer Perceptrons (MLP), as well as Likelihood and LikelihoodD, to improve the identification of heavy charged Higgs bosons compared to Standard Model backgrounds at a 3. p. 00 TeV γγ collider with an integrated luminosity of Lint = 3000 f b−1
Theoretical Framework
The investigation is grounded in the benchmark points that comply with experimental constraints, allowing for systematic accounting of uncertainties inherent in the analysis The paper focuses on a multivariate analysis of charged Higgs boson production at the photon collider of the International Linear Collider (ILC) The analysis is conducted within the Two Higgs Doublet Model Type III (2HDM-III) framework This model expands the Standard Model by incorporating two scalar Higgs doublets, labeled as H1 and H2 The Lagrangian of the 2HDM consists of kinetic terms, potential terms, and Yukawa interaction terms The scalar to fermion couplings are described by the most general expressions in the Yukawa sector A common approach to maintain control over flavor-changing neutral currents (FCNCs) while still generating flavor-violating Higgs signals is to impose a flavor symmetry that constrains the structure of the Yukawa matrices
Signal Definition and Backgrounds
The signal process under study is pair creation of charged Higgs bosons in photon-photon collisions, denoted as γ γ → H+ H− The decay of charged Higgs in top-bottom and tau-neutrino channels is used for analysis The final state involves two charged tau leptons, multiple jets (also b-jets), and significant missing transverse energy (ET) from neutrinos The background processes used for analysis arise from Standard Model production of electroweak and top-quark final states of photon collisions These backgrounds include γ γ → W Z, W → τ ντ, and γ γ → t t Z, t ¯ → b τ ντ.
Multivariate Analysis Techniques
The analysis utilizes an integrated ROOT framework for parallel running and computation work with different multivariate techniques called “Toolkit for Multivariate Analysis” Input variables include P jet, T Transverse momentum, ηjet Pseudorapidity, Njet Number of Jets, ET Missing transverse energy, Scalar HT Sum of PT, Mjet Jet mass, ∆ηjet Pseudorapidity separation and ∆φjet Azimuthal angle separation The classifiers employed include MLP, LikelihoodD (Decorrelation), Boosted Decision Tree (BDT), and Likelihood The BDT algorithm involves a structure resembling a tree that uses branching to show many outcomes of a decision The MLP is a reduced layout known as multilayer perception (MLP) which can also be used to speed up processing
Performance Evaluation
The performance of the classifiers is assessed using the Area Under (AUC) curve with cuts and without cuts values Table IV demonstrates that MLP and BDT are the best classifiers overall The application of optimal cuts slightly improves MLP and BDT classifiers because they can take nonlinear correlation among variables For 171 signal and 604 background events the maximum S/ S+B is 9.8630 when cutting at 0.
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
-
Improved signal-to-background discrimination in photon collisions by utilizing multivariate analysis techniques like Boosted Decision Trees (BDT) and Multilayer Perceptrons (MLP). The paper shows that
the application of optimal cuts slightly improves MLP and BDT classifiers because they can take nonlinear correlation among variables ∆ηjet, ∆φjet, scalar transverse momentum, HT and ET.
-
Enhanced ability to distinguish charged Higgs boson signals from Standard Model backgrounds by employing Likelihood with Decorrelation (LikelihoodD) when input variables are correlated. The paper notes that
Only linearly coupled and Gaussian-distributed variables have full decorrelation,
suggesting this method is crucial for accurately analyzing complex final states. -
Optimized event selection using kinematic cuts to suppress irreducible backgrounds like gamma gamma -> W Z and photon collisions with top-quark final states. These preselected cuts
improved the signal-to-background ratio in the phase space of charged Higgs production in 2HDM Type-III.
-
Automated parameter space exploration for charged Higgs boson masses within a specific range (100–190 GeV) using Monte Carlo generators like MadGraph5 v3.4.2 to separate signals and backgrounds at different benchmark points. This allows for systematic accounting of uncertainties inherent in the analysis based on
benchmark points that comply with experimental constraints.
-
Predictive capability for charged Higgs boson properties, such as decay branching ratios, based on specific model parameters (like tan beta) and mass (MH±). The system can predict decay modes by observing the transition:
the partial decay width indicates that top-quark channel is dominant for higher masses and leptonic mode dominant for lower masses.
Sources
- Hunting a charged Higgs boson pair in proton-proton collisions
- Prospects for observing a novel photon-induced $H^-H^+$ production at HL-LHC
- 2HDMC - a two Higgs Doublet Model Calculator
Related papers
- MadNIS at NLO
- The NANOGrav 15-year data set: A Limit on Continuous Spin Gravitons
- DIS-secting the universe with high-energy neutrinos
- Crowdsourcing Gravitational Waves from Superradiant Axions
- Classification of g-modes for neutron stars with a strong transition: Novel universal relation including slow stable hybrid stars
- Scalar axion field of toroidal electromagnetic pulses