Search for heavy resonances decaying into two Higgs bosons in the $\mathrm{b\bar{b}}τ^+τ^-$ final state in proton-proton collisions at $\sqrt{s}$ = 13 TeV

Using 138 fb$^{-1}$ of proton-proton collision data at 13 TeV collected by the CMS detector, this study searches for heavy resonances decaying into two Higgs bosons in the $\mathrm{b\bar{b}}\tau^+\tau^-$ final state, finding no evidence of new physics and setting the most sensitive limits to date on such production for resonance masses between 1.4 and 4.5 TeV.

The Gist

The Big Picture: Hunting for Heavy Ghosts

Imagine the universe is like a giant, high-speed racetrack. At the CERN laboratory in Switzerland, scientists smash tiny particles (protons) together at nearly the speed of light. This creates a massive explosion of energy that briefly turns into new, heavy particles.

For years, we've known about the Higgs boson (the particle that gives other things mass), but we still have big questions about why the universe is the way it is. This paper is about a search for a "ghost" particle—a heavy, invisible resonance (let's call it X) that might exist but hasn't been seen yet.

The scientists are looking for a very specific "signature" left behind if this ghost particle X exists. They are looking for a scenario where X crashes into two Higgs bosons, and those two Higgs bosons immediately break apart into specific pieces:

1. Two heavy bottom quarks (which turn into a spray of particles called a "jet").
2. Two tau leptons (heavy cousins of electrons that decay quickly).

The Challenge: Finding a Needle in a Haystack

The problem is that these heavy particles are incredibly rare, and the "haystack" (the background noise from normal particle collisions) is enormous.

Think of it like trying to hear a specific whisper in a crowded stadium. The crowd is shouting (this is the Standard Model background—normal physics we already understand). The scientists are trying to hear a specific, faint whisper (the signal of the new particle X).

To make this harder, the particles they are looking for are moving so fast (they are "boosted") that their decay products get squished together.

• The Higgs to Bottom Quarks: Usually, a Higgs decaying into bottom quarks creates two separate sprays. But because this Higgs is moving so fast, the two sprays merge into one giant, messy spray. The scientists had to build a special "smart filter" (an AI called PARTICLENET) to recognize that this one giant spray is actually two bottom quarks stuck together.
• The Higgs to Tau Leptons: Similarly, the tau leptons are moving so fast that they overlap. The team used another advanced AI tool (called BOOSTEDDEEPTAU) to untangle these overlapping particles and identify them correctly.

The Search Strategy: The 2016–2018 Data

The team looked at data collected over three years (2016, 2017, and 2018) using the CMS detector. This is a massive, layered camera and sensor system the size of a building that records every detail of the collisions.

They analyzed 138 "inverse femtobarns" of data. To use an analogy: if a femtobarn is a single grain of sand, they looked at a beach the size of a small city to find their specific grain of sand.

They focused on a mass range between 1 and 4.5 TeV (Tera-electronvolts). To put that in perspective, a proton weighs about 1 GeV. So, they were looking for particles roughly 1,000 to 4,500 times heavier than a proton.

The Results: No Ghosts Found (Yet)

After running their complex algorithms and filtering out the noise, they compared what they saw in the data against what the Standard Model predicts should happen.

• The Outcome: The data matched the "crowd noise" perfectly. There was no whisper. No heavy resonance X was found.
• The Limits: Even though they didn't find the particle, they didn't come up empty-handed. They were able to say, "If this particle exists, it cannot be heavier than X or lighter than Y, and it cannot be produced more often than Z."

They set the strictest limits to date for this specific type of particle decay in the mass range of 1.4 to 4.5 TeV. This means that if a particle like this does exist, it is even more elusive than we thought, or it simply doesn't exist in the way these theories predicted.

Why This Matters

This paper is a "negative result," but in physics, that's a huge deal. It's like checking a map and confirming, "The treasure is definitely not buried here." By ruling out these possibilities, the scientists are narrowing down the search area for future experiments. They are telling the theoretical physicists: "Stop looking for the particle in this specific spot; it's not there."

In summary: The CMS team used a massive dataset and advanced AI to look for a heavy, invisible particle that breaks into two Higgs bosons. They didn't find it, but they successfully proved that if it exists, it's hiding in a way that is even harder to detect than previously thought, setting new boundaries for where physicists should look next.

Technical Summary

Technical Summary: Search for Heavy Resonances Decaying into Two Higgs Bosons in the $bb\tau^+\tau^-$ Final State

Problem and Motivation
The discovery of the Higgs boson ($H$) confirmed the Standard Model (SM) as a robust framework for high-energy interactions, yet unresolved theoretical issues, such as the hierarchy problem, necessitate the search for Beyond-the-Standard-Model (BSM) physics. Many BSM theories predict the existence of heavy resonances ($X$) with masses exceeding 1 TeV that decay into pairs of Higgs bosons ($HH$). While SM Higgs pair production is a rare process, resonant production via BSM mechanisms could significantly enhance the cross-section. Theoretical frameworks such as the Randall–Sundrum (RS) warped extra dimensions model predict new neutral spin-0 radions and spin-2 Kaluza–Klein gravitons that decay into $HH$. This paper presents a search for such narrow-width resonances in the mass range of 1 to 4.5 TeV, specifically targeting the $X \to HH \to bb\tau^+\tau^-$ final state.

Methodology
The analysis utilizes proton-proton collision data collected by the CMS detector at a center-of-mass energy of $\sqrt{s} = 13$ TeV during 2016–2018, corresponding to an integrated luminosity of 138 fb$^{-1}$.

• Event Reconstruction and Topology: The search targets events where one Higgs boson decays to a bottom quark pair ($H \to bb$) and the other to a tau lepton pair ($H \to \tau^+\tau^-$). In the high-mass regime (1–4.5 TeV), the Higgs bosons are highly Lorentz-boosted, causing their decay products to be collimated.

  • The $H \to bb$ decay is reconstructed using a single large-radius jet (AK8, $R=0.8$). The jet is identified using the PARTICLENET graph convolutional neural network tagger, which distinguishes $H \to bb$ decays from background jets based on jet substructure and secondary vertex information.
  • The $H \to \tau^+\tau^-$ decay is reconstructed in two channels: fully hadronic ($\tau_h\tau_h$) and mixed leptonic-hadronic ($\ell\tau_h$, where $\ell$ is an electron or muon). Due to the boost, tau decay products often overlap. The analysis employs a dedicated "boosted" reconstruction algorithm using Cambridge–Aachen jets (CA8) and a specialized deep neural network tagger, BOOSTEDDEEPTAU, optimized to distinguish genuine boosted tau leptons from background jets. This tagger demonstrates significantly improved discrimination power compared to previous methods, particularly for tau transverse momenta ($p_T$) above 100 GeV.
  • The di-tau system's four-momentum is reconstructed using the FASTMTT algorithm, which employs a dynamical likelihood approach and the collinear approximation to account for neutrinos.

• Event Selection: Events are triggered on missing transverse momentum ($p_T^{miss}$) or missing hadronic transverse momentum ($H_T^{miss}$). Offline selection requires $p_T^{miss} > 180$ GeV. The analysis selects events with one AK8 jet (candidate for $H \to bb$) and one $\tau^+\tau^-$ pair. Additional kinematic requirements include angular separation between the $\tau$ candidates ($\Delta R < 1.5$), separation between the $H \to bb$ jet and $p_T^{miss}$ ($\Delta \phi > 1$), and a soft-drop jet mass $> 30$ GeV for the $H \to bb$ candidate.

• Background Estimation and Statistical Analysis: The dominant backgrounds are $t\bar{t}$, $Z$+jets, and $W$+jets. The analysis defines a Signal Region (SR) based on the reconstructed $H \to bb$ jet mass ($100 < M_{H(bb)} < 150$ GeV) and a Sideband (SB) region ($M_{H(bb)} < 100$ GeV or $> 150$ GeV). The SB data is used to constrain the normalization and shape of the $t\bar{t}$ background in the SR via a simultaneous fit. The final discriminant is the reconstructed resonance mass ($M_X$), defined as the invariant mass of the $H \to bb$ jet and the FASTMTT-reconstructed di-tau system. A profile likelihood ratio test statistic is used to set limits.

Key Contributions

• Advanced Reconstruction Techniques: The paper introduces and validates the use of the BOOSTEDDEEPTAU tagger specifically for reconstructing highly boosted tau leptons in the $\tau_h\tau_h$ and $\ell\tau_h$ channels, achieving a factor of 2–4 better discrimination than previous methods for $p_T < 100$ GeV and over a factor of 10 better for $p_T > 100$ GeV.
• Model-Independent Search: The analysis strategy is designed to be model-independent, applying a single search strategy for both spin-0 and spin-2 resonances.
• Comprehensive Kinematic Coverage: The analysis covers the resonance mass range of 1 to 4.5 TeV, utilizing the full Run 2 dataset (2016–2018) of the CMS experiment.

Results
The observed data are consistent with Standard Model background expectations across the entire mass range. No significant excess of events is observed.

• Upper Limits: 95% confidence level (CL) upper limits are set on the production cross section for resonant $HH$ production.
  • For spin-0 resonances, the observed (expected) limits range from 62.2 fb (76.2 fb) at 1 TeV down to 3.8 fb (2 fb) at 4.5 TeV.
  • For spin-2 resonances, the observed (expected) limits range from 42.5 fb (51.8 fb) at 1 TeV down to 3.2 fb (1.7 fb) at 4.5 TeV.
• The observed limits are slightly higher than expected for masses above 2.5 TeV, primarily due to a small excess of events in the highest mass bin of the $M_X$ distribution.

Significance
This analysis establishes the most sensitive limits to date on $X \to HH \to bb\tau^+\tau^-$ decays in the mass range of 1.4 to 4.5 TeV. It improves upon previous CMS searches using smaller datasets and surpasses the sensitivity of ATLAS searches in the Lorentz-boosted regime (1–3 TeV) and the resolved topology search for masses above 1.4 TeV. The results provide stringent constraints on BSM scenarios predicting heavy resonances decaying into Higgs boson pairs, such as warped extra dimensions models.