Exotic Higgs Decays at a Muon Collider

This paper demonstrates that a future muon collider, particularly at 10 TeV, offers significantly enhanced sensitivity to exotic Higgs decays into light singlet scalars ($h \to SS$) compared to the LHC, achieving branching ratio limits of ${\cal O}(10^{-3})$ for the $4b$ final state and $10^{-5}$ for the $2b2\mu$ channel through advanced machine-learning techniques.

The Gist

Imagine the universe is a giant, bustling construction site. For decades, physicists have been trying to understand the blueprints of this site, specifically focusing on a very important, but mysterious, foreman called the Higgs boson. We know the foreman exists, but we want to know if he ever sneaks off to do secret, unauthorized work.

This paper is a proposal for a new, super-powered construction site inspection team: a Muon Collider. The authors are asking, "If we build this new, ultra-precise machine, can we catch the Higgs boson doing something weird that we've never seen before?"

Here is the story of their investigation, broken down into simple concepts.

1. The Suspect: The "Secret Agent" Higgs

In our current understanding (the Standard Model), the Higgs boson usually breaks down into familiar particles like bottom quarks (heavy cousins of electrons) or photons. But what if the Higgs is actually a "secret agent"? What if it occasionally decays into two invisible, lightweight "ghost" particles (called Singlet Scalars, or S) that don't belong in our current rulebook?

If these ghosts exist, they would quickly vanish again, turning into normal particles. The paper focuses on two specific ways this might happen:

• The "Four-Bottom" Case (4b): The Higgs splits into two ghosts, and each ghost turns into two heavy bottom quarks. Result: A messy pile of four bottom quarks.
• The "Two-Bottom, Two-Muon" Case (2b2µ): The Higgs splits into two ghosts. One ghost turns into two bottom quarks, and the other turns into two muons (heavy electrons). Result: A mix of heavy quarks and a clean pair of muons.

2. The New Machine: The Muon Collider

To catch these rare events, we need a better microscope than the ones we have now (like the Large Hadron Collider, or LHC). The authors propose a Muon Collider.

• The Analogy: Imagine the LHC is like trying to find a specific needle in a haystack by throwing two haystacks at each other at high speed. It's chaotic, full of dust, and hard to see anything clearly.
• The Muon Collider: This is like taking two perfectly clean, identical needles and smashing them together in a vacuum. Because muons are elementary particles (not made of smaller parts like protons), the collision is much cleaner. It's a high-precision scalpel rather than a sledgehammer. This allows physicists to see very faint signals that would be drowned out by noise in a regular collider.

3. The Challenge: Finding a Needle in a Haystack

Even with a clean machine, finding these exotic decays is hard.

• The 4b Problem (The Messy Pile): When the Higgs decays into four bottom quarks, it looks very similar to a lot of boring, common background noise. It's like trying to find a specific four-leaf clover in a field of four-leaf clovers that are all slightly different. The particles are so close together they blur into a single blob.
• The 2b2µ Problem (The Clean Pair): This is easier to spot because the two muons act like a "signature." If you see two muons with a specific energy, it's a strong hint that a ghost particle was there. However, the ghost particle rarely turns into muons, so this signal is very faint.

4. The Solution: AI as the Detective

This is where the paper gets really cool. The authors realized that human eyes and simple math aren't enough to separate the signal from the noise, especially in the messy "four-bottom" case.

• The Analogy: Imagine you are trying to find a specific person in a crowded stadium. You know what they look like, but there are thousands of people wearing similar clothes.
• The Machine Learning (ML) Trick: The authors trained a computer program (an AI called a "Boosted Decision Tree") to be the ultimate detective. They fed the AI thousands of examples of "fake" events (background noise) and "real" events (signals).
• The Result: The AI learned subtle patterns that humans miss. It could tell the difference between a messy pile of particles caused by a rare Higgs decay and a messy pile caused by common background noise. It was like the AI could see the "aura" of the event.

5. The Results: A New Era of Discovery

The paper crunches the numbers for two versions of this future machine:

• The "3 TeV" Machine: A powerful collider with 10 years of data.
• The "10 TeV" Machine: A super-powerful collider with 100 years of data.

What did they find?

• For the Messy Pile (4b): The 10 TeV machine could detect these exotic decays if they happen as often as 1 in 1,000 times. This is 100 times better than what we can do with current technology (the HL-LHC). It's a massive leap forward.
• For the Clean Pair (2b2µ): The machine could detect this even if it happens only 1 in 100,000 times. However, in the specific model they studied, the "ghost" particle rarely turns into muons, so this channel is less useful for this specific theory, though it would be amazing for other theories.

The Bottom Line

This paper is a roadmap for the future. It argues that if we build a Muon Collider, we will have a "super-vision" tool that can see the Higgs boson doing secret, exotic things that we are currently blind to.

By using AI to clean up the noise, a Muon Collider could prove that the Higgs boson is connected to a hidden world of new particles, potentially solving mysteries like Dark Matter or why the universe exists the way it does. It's not just about finding a new particle; it's about opening a door to a completely new chapter in our understanding of the universe.

Technical Summary

1. Problem Statement

The discovery of the Higgs boson at the LHC completed the Standard Model (SM), but the search for "exotic" Higgs decays (decays into new light particles beyond the SM) remains a primary frontier for new physics. These decays are motivated by deep theoretical questions such as naturalness, dark matter, and the nature of the Electroweak Phase Transition (EWPT).

While the High-Luminosity LHC (HL-LHC) will produce vast Higgs samples, it faces significant challenges in detecting exotic decays with fully hadronic final states (e.g., four bottom quarks) due to overwhelming QCD backgrounds and jet combinatorics. The paper investigates whether a future Muon Collider (MC), offering a unique combination of high energy and a clean collision environment (elementary particles vs. composite hadrons), can significantly improve sensitivity to these rare processes compared to the HL-LHC.

2. Methodology

2.1 Theoretical Model

The authors analyze a minimal extension of the SM where the Higgs boson ($h$) mixes with a light SM-gauge-singlet scalar ($S$).

• Interaction: The Higgs potential includes terms allowing mixing between the Higgs doublet $H$ and the singlet $S$.
• Decay Chain: The primary process studied is $h \to SS$, followed by the decay of the singlet scalars back to SM particles.
• Focus Channels: Two specific final states are analyzed:
  1. Fully Hadronic: $h \to SS \to 4b$ (four bottom quarks).
  2. Semi-Leptonic: $h \to SS \to 2b2\mu$ (two bottom quarks and two muons).
• Parameters: The study scans singlet masses ($m_S$) from 10 to 60 GeV. In the "Higgs-portal" scenario, $S$ decays to SM particles via mixing, with $S \to b\bar{b}$ being dominant ($\sim 80\%$) and $S \to \mu^+\mu^-$ being highly suppressed.

2.2 Simulation and Detector Setup

• Colliders: Two benchmark configurations are simulated:
  • $\sqrt{s} = 3$ TeV with $1 \text{ ab}^{-1}$ luminosity.
  • $\sqrt{s} = 10$ TeV with $10 \text{ ab}^{-1}$ luminosity.
• Production: The dominant Higgs production mechanism is Vector Boson Fusion (VBF), specifically $W$-boson fusion ($\mu^+\mu^- \to \nu\bar{\nu}h$).
• Tools:
  • Event Generation: MadGraph5 (matrix elements), Pythia8 (showering/hadronization), FeynRules (model implementation).
  • Detector Simulation: Delphes 3 using a muon collider template.
• Reconstruction:
  • Jets reconstructed with the Valencia (VLC) algorithm ($R=0.5$).
  • b-tagging: 70% efficiency with a reduced matching cone ($\Delta R < 0.3$) to mitigate mis-tagging in dense environments.
  • Kinematic Cuts: Preselection includes $p_T > 20$ GeV for jets, $p_T > 5$ GeV for muons, and strict angular separation requirements ($\Delta R > 0.4$) to resolve collinear jets.

2.3 Analysis Strategy

The analysis employs a two-stage approach:

1. Preselection: Standard kinematic cuts to remove soft radiation and ensure jet resolution. Crucially, an invariant mass window ($100 < m_{4j/2j2\mu} < 150$ GeV) is applied to isolate the Higgs resonance.
2. Machine Learning (ML) Selection:
   • Algorithm: XGBoost (Boosted Decision Trees).
   • Input Features: Jet/muon kinematics ($p_T, \eta, \phi$), invariant masses of pairs ($m_{jj}, m_{\mu\mu}$), and deviations from the benchmark mass $m_S$.
   • Strategy: For the $4b$ channel, ML is critical to resolve jet combinatorics and distinguish signal from the irreducible $h \to ZZ^* \to 4b$ background. For the $2b2\mu$ channel, the clean dimuon resonance is used, but ML helps optimize selection before applying the final mass window.
   • Optimization: Thresholds are chosen to maximize the statistical significance $S/\sqrt{S+B+\delta B^2}$ (assuming 5% background systematic uncertainty).

3. Key Contributions

• First Comprehensive MC Study: This is one of the first detailed studies quantifying the sensitivity of a multi-TeV muon collider to exotic Higgs decays in hadronic channels, a regime where HL-LHC struggles.
• ML-Driven Background Suppression: The paper demonstrates that Machine Learning (specifically XGBoost) is essential for the $4b$ channel, improving the signal-to-background ratio by a factor of $\sim 2$ and increasing statistical significance by $\sim 30\%$ compared to cut-based methods alone.
• Resolution of Jet Combinatorics: The study highlights how the clean environment of a muon collider, combined with ML, allows for the reconstruction of the intermediate $S$ resonance mass even in the complex $4b$ final state, which is notoriously difficult at hadron colliders.
• Model-Independent vs. Model-Dependent Limits: The authors provide limits both for the specific Higgs-portal model (where $S$ decays are fixed by mixing) and general model-independent limits (where $BR(S \to \text{final state})$ is treated as a free parameter).

4. Key Results

4.1 The $4b$ Channel (Fully Hadronic)

• Sensitivity:
  • 10 TeV MC: Can probe $BR(h \to SS \to 4b)$ down to $O(10^{-3})$.
  • 3 TeV MC: Can probe down to $O(10^{-2})$.
• Comparison: This represents an improvement of almost two orders of magnitude over HL-LHC projections (which are limited to $\sim O(10^{-1})$).
• Mass Dependence: Sensitivity is highest for $m_S \in [30, 40]$ GeV. For $m_S \lesssim 20$ GeV, sensitivity drops significantly because the $b$-jets become too collimated to satisfy the $\Delta R > 0.4$ resolution requirement.
• Backgrounds: The dominant background is $h \to ZZ^* \to 4b$. The ML classifier effectively separates the signal peak from this background.

4.2 The $2b2\mu$ Channel (Semi-Leptonic)

• Model-Independent Sensitivity: In a scenario where $S$ decays are not fixed, the 10 TeV MC can probe $BR(h \to SS \to 2b2\mu)$ down to $10^{-5}$, a factor of 2–4 better than HL-LHC, thanks to the clean dimuon resonance.
• Higgs-Portal Sensitivity: In the specific Higgs-portal model, $S \to \mu^+\mu^-$ is highly suppressed (branching ratio $\ll 1$). Consequently, the sensitivity to the total $BR(h \to SS)$ in this channel is much weaker than in the $4b$ channel.
• Conclusion: The $4b$ channel is the primary discovery mode for the Higgs-portal model, while $2b2\mu$ offers superior reach only if $S$ has enhanced couplings to leptons.

5. Significance

• Superiority over HL-LHC: The study conclusively shows that for exotic Higgs decays into hadronic final states, a muon collider offers a distinct advantage over the HL-LHC. This is not due to higher energy transfer (the process is electroweak scale) but due to the substantially lower hadronic background rates and the ability to resolve complex multi-jet topologies.
• Physics Reach: A 10 TeV muon collider could probe parameter spaces for strong first-order EWPT and other BSM scenarios that are currently inaccessible.
• Future Directions: The authors suggest that while Higgs factories (like FCC-ee) are excellent for precision, multi-TeV muon colliders are uniquely competitive for rare processes and associated production ($hS$) where high energy is beneficial. The study also identifies $2b2\tau$ and $4\tau$ channels as promising future targets.

In summary, the paper establishes that a future muon collider, utilizing advanced machine learning techniques to handle jet combinatorics, will be a powerful discovery machine for exotic Higgs decays, particularly in the challenging fully hadronic $4b$ channel.