Boosted Higgs Boson Searches

• Boosted Higgs boson searches are experimental strategies that focus on Higgs particles with high transverse momentum to enhance precision measurements of Standard Model couplings and probe new physics.
• They employ advanced jet grooming, substructure discriminants, and double-b tagging to identify merged decay products and suppress overwhelming QCD backgrounds.
• Data-driven background estimations and robust statistical fits are used to extract signals and control systematics, enabling sensitivity to TeV-scale resonances and extended Higgs sectors.

A "boosted Higgs boson" search refers to experimental strategies that exploit Higgs bosons produced with transverse momenta much larger than their mass, such that their decay products are highly collimated. This regime, increasingly accessible at the LHC and future hadron colliders, requires specialized reconstruction and analysis methods that differ from canonical searches. Boosted Higgs techniques are pivotal for precise measurements of Standard Model (SM) couplings at high energy, for extending sensitivity to new physics, for studying extended Higgs sectors, and for controlling backgrounds in challenging final states.

1. Boosted Higgs Production: Dynamical Origins and Motivation

Boosted Higgs bosons typically arise in processes where the Higgs is produced recoil against a hard object or in the decays of heavy new states. The principal mechanisms are:

• Standard Model processes: In gluon fusion (GF), the Higgs can recoil against energetic jets, leading to high-pₜ configurations. Vector boson fusion (VBF) produces Higgs bosons accompanied by forward, high-mass jets; at high Higgs transverse momentum (pₜ ≳ 450 GeV), VBF becomes a substantial contributor relative to GF (Collaboration, 10 Jul 2024). Associated production with W or Z bosons, top quark pairs, or additional jets can also yield boosted Higgses (Krohn et al., 2017, Collaboration, 2022).
• New physics scenarios: Cascade decays of heavy particles (e.g., squarks and gluinos in supersymmetry (1103.4138), Kaluza–Klein gravitons (Cooper et al., 2013), heavy scalars (Collaboration, 2022)) generically produce boosted Higgses as heavy states decay to lighter ones. Composite Higgs, 2HDM, NMSSM, left–right symmetric, and other extended frameworks predict resonances or decay chains with energetic Higgs emission (Backović et al., 2014, Chen, 2016, Bhattacharyya, 2023).

The boosted regime offers enhanced background suppression and unique sensitivity to anomalous couplings, dimension-6 operators, and rare decay topologies, motivating its centrality in both SM and BSM searches (Edezhath, 2015, Behr, 2014).

2. Jet Substructure Methods and Heavy Flavor Tagging

The haLLMark of the boosted regime is that the decay products of the Higgs—most notably the b and $\overline{b}$ quarks in $H \to b\overline{b}$—are emitted with a small angular separation, $\Delta R \sim 2m_H/p_{\rm T}$. For $p_{\rm T} \gtrsim 450$ GeV, the two $b$-jets merge into a single large-radius "fat jet" (Vernieri, 2017, Collaboration, 2017). Discriminating such jets from the overwhelmingly abundant QCD background requires advanced substructure and tagging techniques:

• Jet grooming: Techniques such as soft drop (SD) (Collaboration, 2020), trimming (1107.1808), and pruning remove soft, wide-angle radiation, revealing the true two-pronged structure and stabilizing the reconstructed jet mass at $m_H$.
• Substructure discriminants: Observables like $N_2^1$ (derived from energy correlation functions), $n$-subjettiness ratios $\tau_{2}/\tau_{1}$, and the mass-drop algorithm (BDRS) (Chen, 2016) quantify the presence of the two-pronged decay signature expected from $H \to b\overline{b}$. Decorrelated versions (DDT) ensure flat backgrounds, e.g., $$N_2^1,_{\rm DDT} = N_2^1 - N_{2,26\%}(p_{\rm T},\rho)$$ with $\rho = \ln m_{\rm SD}^2/p_{\rm T}^2$ (Collaboration, 2017, Collaboration, 2020).
• Heavy flavor tagging: Dedicated double-$b$ taggers (DeepDoubleB, ParticleNet-MD, Global Particle Transformer/GloParT) exploit secondary vertex, impact parameter, and deep learning approaches to distinguish merged $b$-jets from QCD or $c$-jets (Mokhtar, 16 Jul 2025, Collaboration, 2022). Modern taggers achieve $\sim$75% $H\to b\overline{b}$ efficiency with $\lesssim$1% QCD misidentification rate.

The combination of these methods is central to both background suppression and reliable signal extraction across all major analyses (Vernieri, 2017, Collaboration, 10 Jul 2024).

3. Signal Extraction, Control of Systematics, and Statistical Frameworks

Boosted Higgs searches employ data-driven techniques and maximum likelihood fits to robustly estimate backgrounds and extract signal strengths:

• Background estimation: QCD multijet background is estimated using control samples (e.g., events failing the double-$b$ tag or substructure requirement), with transfer factors parametrized as Bernstein polynomials in $p_{\rm T}$ and $\rho$ relating pass/fail event yields (Collaboration, 10 Jul 2024, Collaboration, 2020). The functional form

$$N_\mathrm{P}^i = N_\mathrm{F}^{\mathrm{data, }i} \cdot R_\mathrm{P/F}^{\mathrm{MC, }i} \cdot T_\mathrm{P/F}(p_{\rm T}^i, \rho^i),$$

is typical, where $N_\mathrm{F}^\mathrm{data}$ is the yield in the control region, and $T_\mathrm{P/F}$ absorbs systematic and kinematic dependencies.

• Simultaneous fits: Binned maximum likelihood fits are performed to the jet mass in multiple regions (tag pass/fail, VBF/GF categories, control samples), allowing simultaneous extraction of signals (Higgs, $Z$) and normalization of backgrounds (Collaboration, 10 Jul 2024, Mokhtar, 16 Jul 2025, Collaboration, 2020).
• Calibration and validation: The Lund jet plane reweighting technique is used to calibrate simulation-based taggers to data using substructure observables. Observing $Z \to b\overline{b}$ in boosted topologies validates tagging and background methods (Collaboration, 2017, Collaboration, 2020, Mokhtar, 16 Jul 2025).

A significant focus is placed on modeling systematics: uncertainties in jet energy/mass scale, tagging efficiency, theory, and MC modeling (e.g., through Herwig++ parameter scans (1207.0380)) are propagated through the fits, ensuring robust and reliable limits or measurements.

4. Search Channels, Results, and Physics Reach

Boosted Higgs analyses are deployed across a wide spectrum of SM and BSM searches:

• Inclusive and differential SM $H\to b\overline{b}$ measurements: Inclusive searches for high-$p_{\rm T}$ Higgs bosons yield the first observation of $Z\to b\overline{b}$ in a single-jet topology and measure Higgs cross sections consistent with the SM within uncertainties (e.g., $$\mu_H = 3.7 \pm 1.2~\text{(stat)}^{+0.8}_{-0.7}~\text{(syst)}^{+0.8}_{-0.5}~\text{(theo)}$$ for $p_{\rm T} > 450$ GeV (Collaboration, 2020)). Differential cross section measurements are presented unfolded in $p_{\rm T}$ bins (Collaboration, 10 Jul 2024, Collaboration, 2020).
• VBF and gluon fusion separation: Analyses define regions based on VBF topology (forward jets, high $m_{jj}$ and $|\Delta\eta|$), with simultaneous fits extracting the signal strengths of both VBF and GF production. For example, (Collaboration, 10 Jul 2024) reports $\mu_{\rm VBF} = 4.9^{+1.9}_{-1.6}$ and $\mu_{\rm GF} = 1.6^{+1.7}_{-1.5}$ using 138 fb$^{-1}$ at 13 TeV.
• Beyond SM resonance and cascade searches: Boosted topologies drastically suppress backgrounds in fully hadronic final states for $hh\rightarrow b\overline{b}b\overline{b}$, $X\to YH\to b\overline{b}b\overline{b}$, and heavy scalar searches. Cross-section upper limits are set at the level of $0.1$–$150$ fb depending on masses in (Collaboration, 2022). Advanced ML-based taggers and kinematic selections yield sensitivity to new particles at the TeV scale (Cooper et al., 2013, Backović et al., 2014).
• Associated production ($VH$, $t\overline{t}H$): Boosted techniques enable efficient signal selection in challenging backgrounds, and machine learning classifiers are crucial for separating $VH$ or $t\overline{t}H$ from $V$+jets and $t\overline{t}$ backgrounds (Collaboration, 2022, Krohn et al., 2017).
• Extended and exotic Higgs sector searches: The methodology generalizes to channels like $H^+\to W^+A$, $A\to hZ$, and $hh\to 4b$, including for heavy neutral or charged Higgses in UV-complete models. Multivariate techniques (e.g., BDTs) and boosted-object tagging are foundational for these analyses (Chen, 2016, Li et al., 2016, Bhattacharyya, 2023).

5. Machine Learning and Advanced Tagging Architectures

The sensitivity of boosted Higgs searches has improved markedly with the deployment of deep learning-based jet taggers:

• ParticleNet and variants: ParticleNet-MD, employing graph neural networks on particle constituents, achieves the highest signal efficiency at a fixed background rejection among current taggers for H→$b\bar{b}$, outperforming traditional double-b or DeepAK8 taggers (Mokhtar, 16 Jul 2025, Collaboration, 2022).
• Global Particle Transformer (GloParT): Attention-based architectures incorporating global event information, GloParT discriminates between a broad set of jet categories including Higgs, top, and vector boson decays (Mokhtar, 16 Jul 2025).
• Transfer learning: Fine-tuning pre-trained models (e.g., P(H₁ℓ) for H→WW in leptonic final states) on specific sub-datasets leads to significant performance gains, as demonstrated by a 70% increase in expected significance for the 1-lepton boosted Higgs analysis (Mokhtar, 16 Jul 2025).
• Calibration: Data-driven methods, such as reweighting in the Lund jet plane, correct potential mis-modeling in simulation-trained taggers, ensuring reliable application in data, especially for non-SM or multi-prong jets (Mokhtar, 16 Jul 2025).

These methods form the backbone of searches in all current and forthcoming analyses in the boosted regime.

6. Impact, Future Prospects, and Theoretical Implications

Boosted Higgs boson searches are delivering stringent cross-section measurements and setting competitive limits on new physics, with transformative impact on collider phenomenology:

• SM precision and EFT sensitivity: High-$p_\mathrm{T}$ measurements in the boosted regime probe higher orders in effective field theory (EFT) and anomalous couplings—especially in VBF and $t\overline{t}H$ channels—yielding tight constraints on Wilson coefficients for dimension-6 operators (Collaboration, 2022, Edezhath, 2015).
• Background suppression and mass reach: Boosted topologies exploit kinematic and subjet features for dramatic background reduction in multi-jet, fully hadronic, and cascade decay final states, enabling sensitivity to TeV-scale resonances and new scalar sectors (Cooper et al., 2013, Collaboration, 2022).
• Technological trajectory: As the LHC transitions to higher luminosities and dataset sizes increase, further advances are anticipated: improvements in ML taggers, substructure algorithms, pileup mitigation, and calibration methods will enhance discovery or exclusion sensitivity (Mokhtar, 16 Jul 2025, Collaboration, 10 Jul 2024). Multivariate and ML-based approaches will become standard tools for all searches in the boosted regime.
• Interpretation in extended models: The approach is broadly model-independent and underpins searches in supersymmetry (MSSM, NMSSM), composite Higgs, left–right symmetric, and other UV-motivated frameworks (1103.4138, Collaboration, 2022, Bhattacharyya, 2023).

The systematic use of boosted Higgs strategies is therefore central to Higgs coupling measurements, electroweak symmetry breaking studies, and the search for physics beyond the Standard Model at present and future hadron colliders.