⚛️ phenomenology

A Window onto New Invisible Particles via Semi-Visible Higgs Decays

This paper investigates the potential to discover new invisible particles at the LHC by analyzing semi-visible Higgs decays ( $H\to \ell^+\ell^- + \rm E{\!\!\!/}_T$ and $H\to jj + \rm E{\!\!\!/}_T$ ) using an effective field theory framework, demonstrating that despite large backgrounds, signal extraction is feasible through both cut-based and multivariate BDT analyses.

Original authors: Sally Dawson, Arnab Roy, German Valencia

Published 2026-07-13

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Sally Dawson, Arnab Roy, German Valencia

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the Higgs boson as a cosmic magician at the world's biggest particle party, the Large Hadron Collider (LHC). Usually, when this magician pulls a rabbit out of a hat, we see the rabbit. But sometimes, the rabbit is invisible, vanishing into thin air and leaving behind only a "missing energy" gap. Physicists have been hunting for these invisible rabbits for a long time, like counting how many types of neutrinos exist by seeing how much energy the Z-particle "hides."

But in this new study, authors Sally Dawson, Arnab Roy, and German Valencia are looking for a different kind of trick: the semi-visible rabbit.

The "Semi-Visible" Magic Trick

Instead of the Higgs disappearing entirely, the team is watching for a specific, sneaky decay where the Higgs splits into two things: a pair of visible particles (like two electrons or two jets of particles) and a pair of invisible particles that vanish into the darkness. They call this a "semi-visible" decay because you see half the act, but the other half is missing.

Think of it like watching a magician pull a red ball and a blue ball from a hat, but the blue ball instantly turns into a ghost. You see the red ball, you know a blue ball should be there, and you can feel the ghost tugging at the air (missing energy), but you can't see it.

The team focused on two specific versions of this trick:

The Leptonic Mode: The Higgs turns into two charged particles (like electrons) and a ghost.
The Hadronic Mode: The Higgs turns into two jets of particles and a ghost.

The Mountain of Noise

Here is the tricky part: finding these ghosts is like trying to hear a whisper in a hurricane. The paper explains that while the signal (the semi-visible Higgs) is tiny, the background noise (other particle collisions that look similar) is massive.

To make this manageable, the researchers decided to look for the Higgs when it's produced alongside a Z boson (a heavy particle that acts like a "tag"). It's like looking for a specific rare coin, but only when it's found inside a specific, heavy, golden box. This "Z tag" helps them ignore the chaos of the crowd.

They ran detailed computer simulations (using tools like Madgraph5 and Delphes) to see if they could spot this signal against the noise. They found that while the signal is small, the "kinematics" (the way the particles move and bounce) are rich enough to tell the difference.

The Two-Step Hunt

The team tried two methods to separate the whisper from the hurricane:

The Cut-Based Analysis: This is like using a sieve. They applied strict rules (cuts) to filter out the noise. For example, they looked for events where the missing energy was high and the mass of the visible particles was low. After applying all these rules to a massive amount of data (3000 inverse femtobarns, which is a huge amount of particle collisions), they calculated that they would need about 207 signal events to be 99.7% sure (3σ significance) that they found something new.
- The Result: This method could reach sensitivity limits of 41 TeV⁻², 83 TeV⁻², and 152 TeV⁻² for different types of invisible particles.
The Multivariate BDT (The Smart Filter): To do better, they used a "Boosted Decision Tree" (BDT), which is like a super-smart AI filter. Instead of just cutting out the noise, the AI learned to recognize the pattern of the ghost's movement. They fed the computer 60% of the data to learn and 40% to test.
- The Result: This smart filter was much better. It significantly improved their ability to spot the signal, pushing the boundaries of what they can exclude.

What They Found (and What They Didn't)

The study didn't find a new particle yet. Instead, it mapped out the "no-go zones."

The "Higgs Neutrino Floor": There is a natural background limit caused by the Standard Model itself (specifically, the Higgs decaying into a Z boson that turns into neutrinos). The authors show that their method can actually push below this floor, meaning they can find new particles even if they are hiding in the same energy range as standard neutrinos.
The Limits: For "derivative operators" (a specific mathematical way particles interact), the study suggests that looking at the "invisible Z width" (how much energy the Z particle hides) is actually a stricter rule than their semi-visible search. However, for other types of interactions, their method opens up new territory.
Dark Matter: If these invisible particles are dark matter, the sensitivity they achieved is weaker than what direct detection experiments (like underground tanks waiting for a dark matter hit) can do. However, the paper argues that for more complex scenarios where many different forces are at play, or when multiple types of operators contribute at once, this method is a crucial, complementary tool.

The Bottom Line

The paper concludes that searching for these semi-visible Higgs decays is a viable strategy for the High-Luminosity LHC (the future, super-powered version of the collider). While the signal is small and the background is huge, the "rich kinematics" allow for a meaningful extraction of the signal.

They haven't discovered a new particle, but they have drawn a new map of the territory, showing exactly where new invisible particles could be hiding and proving that even with a "neutrino floor" blocking the view, there is still room to dig deeper. It's a simulation-based roadmap, suggesting that with the right tools and enough data, we might just catch a glimpse of the invisible world.

Technical Summary: A Window onto New Invisible Particles via Semi-Visible Higgs Decays

Problem Statement
This study investigates the potential to discover new invisible particles through "semi-visible" Higgs decays at the High-Luminosity Large Hadron Collider (HL-LHC). While traditional searches focus on fully invisible decays ( $H \to \text{invisible}$ ), this work targets specific decay modes where the Higgs boson decays into Standard Model (SM) visible particles accompanied by missing transverse energy ( $E_T$ ). The specific channels analyzed are the leptonic mode $H \to \ell^+\ell^- + E_T$ and the hadronic mode $H \to jj + E_T$ . The motivation stems from the need to probe new weakly interacting particles, hidden sectors, and dark matter candidates that appear as missing energy, particularly in the kinematic window where the mass of the new invisible state ( $m_{\chi, \phi}$ ) is less than half the Higgs mass ( $m_h/2 \approx 50$ GeV).

Methodology
The authors employ an Effective Field Theory (EFT) framework at dimension six to parametrize interactions between the Higgs boson and new invisible states, modeled as complex scalars ( $\phi$ ) or Dirac fermions ( $\chi$ ). The analysis focuses on a specific set of operators:

Scalar interactions: $O_{DH\partial\phi}$ , $O_{e\phi}$ , $O_{u\phi}$ , and $O_{d\phi}$ .
Fermion interactions: $O_{DH\chi\chi}$ and $O_{DH\chi\chi^2}$ .

The study assumes the Higgs is produced via the $ZH$ channel, chosen for its cleaner signature and the ability to use the associated $Z$ boson as a tag. Two distinct signal topologies are considered (Figure 1):

Leptonic Mode: $Z \to jj$ (tagging) and $H \to \ell^+\ell^- + \text{invisible}$ .
Hadronic Mode: $Z \to \ell^+\ell^-$ (tagging) and $H \to jj + \text{invisible}$ .

Signal and background events are simulated using FeynRules, MadGraph5 (LO), Pythia8, and Delphes3 with CMS detector configurations. The background includes irreducible SM processes such as $ZH \to \ell^+\ell^- \nu\bar{\nu}$ (denoted $ZH$-lep), $ZZ$, $ZZZ$, $WWZ$, $t\bar{t}Z$ , and inclusive $t\bar{t}$ .

The analysis proceeds in two stages:

Cut-Based Analysis: A standard selection of kinematic cuts is applied, including requirements on lepton/jet transverse momentum ( $p_T$ ), pseudorapidity ( $\eta$ ), invariant masses ( $M_{\ell\ell} < 60$ GeV, $70 < M_{jj} < 110$ GeV), and missing energy ( $E_T > 20$ GeV).
Multivariate Analysis (BDT): To improve sensitivity against the large backgrounds, a Boosted Decision Tree (BDT) is implemented using TMVA. The classifier utilizes additional kinematic variables, specifically the azimuthal separations $\Delta\phi(\ell, \ell)$ , $\Delta\phi(\ell_1, E_T)$ , and $\Delta\phi(\ell_2, E_T)$ . The training and testing are performed on 60% and 40% of the event samples, respectively.

Key Contributions and Results
The paper quantifies the reach of the HL-LHC (14 TeV, 3000 fb $^{-1}$ ) in constraining the Wilson coefficients of the dimension-six operators.

Signal vs. Background: The study confirms that these processes are characterized by small signal cross-sections against large backgrounds. For the leptonic channel, after all selection cuts, the background cross-section is reduced to $\sigma_{bkg} = 1.52$ fb.
Cut-Based Sensitivity: To achieve a $3\sigma$ $3 σ$ significance with the cut-based analysis, approximately 207 signal events are required. This translates to the following exclusion limits on the Wilson coefficients:
- $C_{DH\chi\chi} \lesssim 41 \text{ TeV}^{-2}$
- $C_{DH\partial\phi} \lesssim 83 \text{ TeV}^{-2}$
- $C_{e\phi} \lesssim 152 \text{ TeV}^{-2}$
BDT Improvement: The multivariate BDT analysis yields a significant improvement in signal significance compared to the cut-based approach. By applying a cut on the classifier output at 0.997, the analysis achieves tighter constraints on the parameter space.
Comparison with Other Constraints:
- For derivative operators, the constraints from the invisible $Z$ width are found to be more stringent than those from semi-visible Higgs decays.
- For Yukawa-like operators, the semi-visible Higgs decay provides competitive sensitivity, though monojet searches offer stronger constraints for specific quark operators.
- The study demonstrates that the "Higgs neutrino floor" (the irreducible SM background from $H \to ZZ^* \to \ell^+\ell^- \nu\bar{\nu}$ ) is reducible in this context, allowing for sensitivity to new physics below the SM background level.

Significance and Claims
The authors conclude that semi-visible Higgs decay searches complement traditional invisible Higgs searches, specifically covering the kinematic window $m_{\phi, \chi} \lesssim 50$ GeV. While the signal is small and backgrounds are large, the richness of the kinematics allows for meaningful signal extraction at the HL-LHC.

The paper modestly claims that while the sensitivities achieved for thermal dark matter scenarios are generally weaker than those from direct or indirect detection experiments, these searches are crucial for:

Probing general dark matter scenarios where multiple operators contribute simultaneously.
Exploring parameter regions where other constraints (like monojet searches) may not apply or are less effective.
Providing a method to reduce the "Higgs neutrino floor," thereby opening a window to new invisible particles that would otherwise remain hidden by SM backgrounds.