Search for Higgs boson exotic decays into Lorentz-boosted light bosons in the four-$\tau$ final state at $\sqrt{s}=13$ TeV with the ATLAS detector

Using 140 fb$^{-1}$ of 13 TeV proton-proton collision data collected by the ATLAS detector, this paper presents a search for exotic Higgs boson decays into a pair of light scalars that subsequently decay into four $\tau$-leptons, finding no significant excess over Standard Model predictions and setting 95% confidence level upper limits on the signal strength between 0.03 and 0.10 for scalar masses between 4 and 15 GeV.

The Gist

The Big Picture: Hunting for a "Ghost" in the Higgs Boson's Shadow

Imagine the Higgs boson as a very famous, heavy celebrity at a massive party (the Large Hadron Collider). We know this celebrity exists, but we want to know exactly what they do when they leave the party. Usually, they break up into known, standard groups of friends (Standard Model particles).

However, physicists suspect that sometimes, the Higgs might be sneaking off with a secret, invisible friend (a new, light particle called an "a-boson") that we haven't seen yet. If this secret friend exists, the Higgs might decay into a pair of them, and those secret friends might then turn into a specific type of particle called a tau-lepton.

This paper is the ATLAS experiment's report on a search for this specific "secret handshake": Higgs → Two Secret Friends → Four Tau-Leptons.

The Challenge: The "Speeding Twins" Problem

Here is the tricky part: The secret friends (the a-bosons) are very light. Because they are so light, when the heavy Higgs splits into them, they zoom away incredibly fast.

Think of it like this: If you throw a heavy bowling ball (the Higgs) and it splits into two ping-pong balls (the a-bosons), those ping-pong balls will fly apart at lightning speed.

When these fast-moving ping-pong balls decay into tau-leptons, the two taus from each ball are born so close together and moving so fast that they look like a single, messy blob in the detector. It's like trying to see two fireflies buzzing inside a single jar; from a distance, they just look like one glowing spot.

Normally, detectors struggle to tell the difference between this "glowing blob" and a random piece of junk (a jet of particles) flying through the detector.

The Solution: The "Muonic Eraser"

To solve this, the ATLAS team invented a clever trick called the "muon-removal technique."

In this specific decay, one of the tau-leptons turns into a muon (a heavy cousin of the electron) and some invisible neutrinos. The other tau turns into a spray of hadrons (particles that interact with the detector's walls).

Usually, if a muon is born right next to a spray of hadrons, the detector gets confused. It thinks, "Is this one big messy particle, or two separate ones?" The muon messes up the measurement of the hadrons.

The Analogy: Imagine trying to count the number of people in a crowded room, but one person is wearing a giant, flashing neon sign (the muon) that blocks your view of the person standing right next to them.

• Old Method: You try to guess how many people are there, but the neon sign makes it hard.
• New Method (Muon Removal): The ATLAS team essentially says, "Okay, we see the neon sign. Let's digitally erase the neon sign from our photo." Once the sign is gone, we can clearly see the person standing next to it and count them correctly.

By digitally removing the muon's influence from the data, they could reconstruct the "blob" and realize, "Ah, this isn't a mess; it's actually two distinct tau-leptons!"

The Search Strategy

The team looked at 140 "years" of collision data (140 inverse femtobarns) collected between 2015 and 2018. They set up a filter to catch events that looked like:

1. Two "muon-neon-signs" (muons).
2. Two "blobs" that turned out to be pairs of tau-leptons once the neon signs were erased.

They divided their search into two groups:

• Same-Sign Group: Both muons have the same electrical charge (like two positive magnets). This is a very clean group because most background noise (random junk) usually comes in opposite pairs.
• Opposite-Sign Group: The muons have opposite charges. This group has more noise (background), so they had to be extra careful to filter out the "fake" signals.

The Results: The "Silent Room"

After running all the numbers and applying their "muon-removal" trick, what did they find?

Nothing.

They looked at the data and compared it to what the Standard Model (our current best theory of physics) predicts should happen. The number of events they saw matched the background noise perfectly. There was no "excess" of events that would indicate the existence of the secret a-boson.

The Verdict:

• No new physics found: They did not discover the Higgs decaying into these light, exotic particles.
• Setting Limits: Even though they didn't find it, they set a very strict boundary. They can say with 95% confidence that if this exotic decay does happen, it happens less than 3% to 10% of the time (depending on the mass of the secret particle).

Why This Matters (Without Speculating)

This paper is significant because it is the first time ATLAS has used this specific "muon-removal" technique to hunt for this type of decay. It proves that the method works and allows them to look for these "merged" particles with much higher precision than before.

While they didn't find the new particle, they effectively closed the door on a specific range of possibilities. If nature is hiding a light particle that the Higgs turns into, it's not hiding in the 4 to 15 GeV mass range in the way this specific model predicted. The search continues, but the "net" they cast this time was much finer and more effective than previous attempts.

Technical Summary

Technical Summary: Search for Higgs Boson Exotic Decays into Lorentz-Boosted Light Bosons in the Four-$\tau$ Final State

Problem and Motivation
Following the discovery of the Higgs boson, the ATLAS and CMS collaborations have focused on measuring its properties and searching for Beyond-the-Standard-Model (BSM) physics. Exotic decays of the Higgs boson into light BSM particles offer a pathway to detect new physics that is weakly coupled to the Standard Model (SM). In models where the SM is extended by a light scalar ($a$) that mixes with the Higgs boson, the $a$-boson inherits the Higgs's Yukawa coupling pattern, decaying preferentially to the heaviest kinematically allowed fermion pair. For light $a$-bosons in the mass range $2m_\tau < m_a < 2m_b$, the decay $a \to \tau^+\tau^-$ is favored.

A specific challenge arises in the mass range $4 \text{ GeV} < m_a < 15 \text{ GeV}$. In this regime, the $a$-boson is highly Lorentz-boosted, causing the two $\tau$-leptons from its decay to overlap significantly within the detector. This overlap complicates the reconstruction of hadronically decaying $\tau$-leptons ($\tau_{\text{had}}$) when they are in close proximity to muons from the other $\tau$ decay ($\tau \to \mu \nu_\mu \nu_\tau$). Previous searches by ATLAS and CMS in similar channels often relied on track-only identification or different decay topologies, potentially missing efficiency in these highly boosted, overlapping scenarios.

Methodology
This analysis utilizes the full Run 2 dataset of proton-proton collisions at $\sqrt{s} = 13 \text{ TeV}$ recorded by the ATLAS detector, corresponding to an integrated luminosity of $140.1 \pm 1.2 \text{ fb}^{-1}$. The search targets the exotic decay chain $H \to aa \to \tau^+\tau^-\tau^+\tau^-$, specifically focusing on the final state where each $a$-boson decays into one $\tau$ decaying to a muon and one $\tau$ decaying to hadrons ($\tau_{\text{had}}$).

• Object Reconstruction and Muon Removal: The core innovation of this analysis is the application of a dedicated "muon-removal" technique. In standard reconstruction, a muon overlapping with a $\tau_{\text{had}}$ seed jet degrades the reconstruction and identification efficiency of the $\tau_{\text{had}}$. The analysis removes the tracks and calorimeter energy clusters associated with the reconstructed muon from the inputs used for $\tau_{\text{had}}$ reconstruction and identification (specifically the Recurrent Neural Network and Boosted Decision Tree algorithms). This technique recovers the efficiency of merged $\tau$-pairs to levels comparable to isolated $\tau$-leptons.
• Event Selection: Events are required to contain exactly two $\mu\tau_{\text{had}}$ objects. The leading muon must have $p_T > 14 \text{ GeV}$, and the leading (subleading) $\tau_{\text{had}}$ must have uncorrected $p_T > 30 \text{ GeV}$ ($> 25 \text{ GeV}$). Both $\tau_{\text{had}}$ candidates must satisfy tight identification criteria.
• Signal Regions: Two orthogonal signal regions (SR) are defined based on the charge of the two muons: a same-sign ($SS_\mu$) region and an opposite-sign ($OS_\mu$) region. An additional requirement of $m_{\mu\mu} < 50 \text{ GeV}$ is applied in the $OS_\mu$ region to suppress $Z$+jets background.
• Background Estimation: The dominant background arises from processes where jets are misidentified as $\tau_{\text{had}}$ (fake-$\tau_{\text{had}}$). This is estimated using a data-driven "tight-to-loose" method with fake factors derived from $Z$+jets events. Processes with four real $\tau$-leptons (e.g., $ZZ$, $H \to ZZ^*$) are modeled via simulation but contribute less than 0.1% to the total background.
• Statistical Analysis: The average mass of the two $\mu\tau_{\text{had}}$ candidates is used as the primary discriminant. A modified profile likelihood ratio is employed to set upper limits on the signal strength $\mu = (\sigma(H)/\sigma_{\text{SM}}(H)) \times \mathcal{B}(H \to aa \to 4\tau)$.

Key Contributions

1. First ATLAS Search in this Channel: This is the first ATLAS search for the $H \to aa \to \tau^+\tau^-\tau^+\tau^-$ decay in the $4 \text{ GeV} < m_a < 15 \text{ GeV}$ mass range.
2. Muon-Removal Technique: The paper presents the first application of the muon-removal technique by ATLAS to reconstruct merged di-$\tau$ decays. This method significantly improves the reconstruction and identification efficiency of $\tau_{\text{had}}$ candidates when they are overlapped with muons, a critical capability for boosted light scalar searches.
3. Comprehensive Background Handling: The analysis provides a robust data-driven estimation of the fake-$\tau_{\text{had}}$ background, validated in dedicated validation regions, and demonstrates that simulation-based backgrounds are negligible.

Results
No significant excess of events over the Standard Model background prediction was observed in either the $SS_\mu$ or $OS_\mu$ signal regions. The observed data yields were consistent with the expected background:

• $SS_\mu$ Region: 121 observed events vs. $129 \pm 12$ expected background.
• $OS_\mu$ Region: 380 observed events vs. $350 \pm 31$ expected background.

Upper limits at the 95% confidence level (CL) were set on the product of the Higgs production cross-section ratio and the branching ratio, $(\sigma(H)/\sigma_{\text{SM}}(H)) \times \mathcal{B}(H \to aa \to 4\tau)$. The limits range from 0.03 to 0.10 depending on the $a$-boson mass ($m_a$).

• The best limit of 0.03 (expected 0.04) is achieved at $m_a = 10 \text{ GeV}$.
• The limit degrades to 0.10 (expected 0.13) at $m_a = 4 \text{ GeV}$ due to poorer signal-background separation.
• Limits worsen for $m_a > 10 \text{ GeV}$ due to reduced signal efficiency as the $\tau$-leptons become less merged.

When interpreted within the framework of the Type-III Two-Higgs-Doublet Model plus a singlet (2HDM+S) with $\tan\beta = 5$, these results provide strong constraints on the model, offering significant improvements over previous ATLAS results in the $\mu\mu\tau\tau$ and $\mu\mu\mu\mu$ final states.

Significance
The paper claims that this measurement significantly improves upon previous similar measurements by extending the search to the challenging low-mass, highly boosted regime using a novel reconstruction technique. The successful application of the muon-removal technique demonstrates its viability for future searches involving overlapping decay products. The absence of a signal sets stringent constraints on exotic Higgs decays into light scalars, narrowing the parameter space for models involving dark matter, electroweak baryogenesis, and neutral naturalness that predict such decays.