Search for a boosted Higgs boson decaying to bottom quark pairs in association with a W or Z boson 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, a search for boosted Higgs bosons decaying to bottom quark pairs in association with hadronically decaying W or Z bosons yielded an observed signal strength of $\mu$ = 0.72 $^{+0.75}_{-0.71}$, consistent with the Standard Model expectation within uncertainties.

The Gist

The Big Picture: Hunting for a "Super-Heavy" Higgs

Imagine the Large Hadron Collider (LHC) as the world's most powerful particle smasher. It smashes protons together to create a chaotic explosion of new particles. Among these, scientists are looking for the Higgs boson, a particle that gives other particles their mass.

Usually, when the Higgs is created, it is like a slow-moving, sleepy turtle. It drifts away gently and decays (breaks apart) into smaller pieces. But sometimes, the Higgs gets a massive boost of energy and zooms away at nearly the speed of light. This is called a "boosted" Higgs.

This paper is a report from the CMS experiment at CERN. The team went on a treasure hunt to find these fast-moving Higgs bosons. Specifically, they were looking for Higgs bosons that were created alongside a W or Z boson (two other heavy particles), and where the Higgs itself broke apart into a pair of bottom quarks (heavy particles that are notoriously hard to spot because they look like a messy pile of debris).

The Challenge: Finding a Needle in a Haystack

Finding a boosted Higgs is incredibly difficult. It's like trying to find a specific, rare type of firework in a stadium full of people setting off thousands of cheap fireworks.

1. The Noise: The biggest problem is the "background noise." When protons smash, they create millions of ordinary jets of particles (like random sparks). These look very similar to the Higgs we are looking for.
2. The Signal: The Higgs we want is special because it is heavy and moves fast. When it breaks into two bottom quarks, those two quarks are so close together that they merge into a single, giant, fuzzy blob (a "large-radius jet").
3. The Accomplice: To make things even harder, the Higgs is often produced with a W or Z boson. In this specific search, the team looked for cases where both the Higgs and the W/Z boson broke apart into messy jets, rather than clean, easy-to-spot particles like electrons or muons.

The Detective Work: How They Solved the Case

The CMS team used a multi-step strategy to filter out the noise and find the signal.

1. The High-Speed Filter (Triggers)
First, they set up a "speed trap." They only kept data from collisions where particles were moving incredibly fast (transverse momentum > 450 GeV). This is like a bouncer at a club who only lets in people running faster than a certain speed, ignoring everyone else.

2. The "Smart" Eye (AI and Neural Networks)
Once they had the fast collisions, they needed to tell the difference between a "Higgs jet" and a "random junk jet."

• They used a sophisticated AI tool called PARTICLENET, which acts like a super-smart detective.
• This AI looks at the internal structure of the giant jet. A Higgs jet has a specific "fingerprint" (it looks like two distinct things merged together), whereas a random junk jet looks like a chaotic mess.
• The AI also checks for "heavy flavor," looking for signs of bottom quarks, which are the specific ingredients of the Higgs decay.

3. The Control Groups (Sidebands)
To be sure their AI wasn't just guessing, they used "control groups." They looked at regions of data where they knew the Higgs wasn't present (the "sidebands"). By studying these, they could accurately estimate how much "junk" was hiding in the real data and subtract it out.

The Results: A Near Miss, But a Success in Method

After analyzing data from 2016 to 2018 (a massive amount of information, equivalent to 138 "inverse femtobarns" of collisions), here is what they found:

• The Count: They found a signal that looked very much like the Standard Model Higgs.
• The Strength: They measured the "signal strength" (how often this happens compared to what theory predicts). They found a value of 0.72.
  • Analogy: If the theory predicted 100 Higgs bosons should appear, they found evidence for about 72.
  • The Catch: Because the data is noisy and the event is rare, the uncertainty is huge. The result is written as 0.72 ± 0.75. This means the true number could be anywhere from almost zero to nearly 1.5 times the prediction.
• The Significance: Statistically, this result is 1.0 standard deviation away from "nothing happened." In the world of particle physics, you usually need 5 standard deviations to claim a "discovery." So, this is not a discovery; it's a "hint" or a "nudge."

However, there is a silver lining:

• Validation: They also looked for a similar process involving the Z boson (VZ) to test their method. The fact that their method worked well enough to measure the Z boson confirms that their "detective tools" (the AI and the selection criteria) are working correctly.

The Conclusion

The paper concludes that while they didn't find a "smoking gun" discovery of a new physics phenomenon, they successfully proved the method works.

They showed that it is possible to hunt for these elusive, fast-moving Higgs bosons in the messy "hadronic" (all-jet) decay channels using advanced AI and large-radius jets. The sensitivity of the search was limited mostly by the amount of data available. It's like trying to hear a whisper in a hurricane; they have the right microphone (the detector and AI), but they need more time listening (more data) to be absolutely sure of what they are hearing.

In short: They built a high-tech net to catch fast, messy Higgs bosons. They caught a few that looked promising, but the net wasn't quite big enough yet to be 100% certain. They are ready to cast the net again with more data in the future.

Technical Summary

Technical Summary: Search for a Boosted Higgs Boson Decaying to Bottom Quark Pairs in Association with a W or Z Boson

Problem and Motivation
The observation of the Higgs boson (H) in 2012 and subsequent property measurements have solidified the Standard Model (SM) understanding of electroweak symmetry breaking. However, precise measurements of Higgs production in association with vector bosons (VH, where V = W or Z) are critical for uncovering Beyond-the-Standard-Model (BSM) phenomena, which may manifest more prominently at high transverse momentum ($p_T$). While the dominant gluon fusion (ggH) production mechanism decreases in relative contribution as $p_T$ increases, the fraction of associated production (VH) rises at high $p_T$.

The decay channel $H \to b\bar{b}$, with a branching fraction of 58%, is ideal for studying boosted Higgs production. At high $p_T$, the decay products of the Higgs and the associated vector boson are collimated, reconstructing as single large-radius jets. While searches for VH production with $H \to b\bar{b}$ have been performed in leptonic V decay channels, the hadronic V decay channel ($V \to q\bar{q}$) had previously been explored only by the ATLAS experiment. This work constitutes the first such search by the CMS Collaboration, aiming to extend high-$p_T$ Higgs measurements to the $V(qq)H(b\bar{b})$ 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 the 2016–2018 run periods, corresponding to an integrated luminosity of 138 fb$^{-1}$.

• Event Selection: Events are selected requiring exactly two large-radius (AK8) jets with $p_T > 450$ GeV and $|\eta| < 2.5$. One jet is designated as the Higgs candidate (Jet 1) and the other as the vector boson candidate (Jet 2).
• Jet Substructure and Flavor Tagging: The analysis employs the soft-drop algorithm to groom jets and calculate the groomed jet mass ($m_{SD}$). To identify the boosted Higgs and vector boson candidates, the analysis utilizes the PARTICLENET-MD algorithm, a mass-decorrelated version of a graph convolutional neural network. This tagger provides discriminants for $b$-quark, $c$-quark, light-quark, and generic QCD jet origins.
  • The Higgs candidate is selected based on a high $D(bb \text{ vs. } cc, qq)$ discriminant.
  • The vector boson candidate is required to have a low QCD misidentification probability ($p(QCD)$) and is categorized into three mass windows: 40–68 GeV, 68–110 GeV (signal region for V), and 110–201 GeV.
• Background Estimation:
  • QCD Multijets: The dominant background is estimated directly from data using a control region (CR) where events fail the $D(bb \text{ vs. } QCD)$ requirement. A parametric transfer factor, derived from a simultaneous fit, extrapolates the QCD shape and yield to the signal region (SR).
  • Resonant Backgrounds (W/Z+jets, Dibosons): Estimated using simulation, calibrated with a W-enriched CR containing single muons and a $b$-tagged jet to constrain jet mass scale, resolution, and tagging efficiency.
  • Top Quark Backgrounds: Estimated from simulation, with normalizations constrained using a CR targeting semi-leptonic $t\bar{t}$ events.
• Statistical Analysis: A binned maximum likelihood fit is performed on the Higgs candidate $m_{SD}$ distribution across the pass/fail regions of the flavor discriminant and the three V mass categories. The signal strength ($\mu$) is extracted relative to the SM expectation.

Key Contributions

• First CMS Search in Hadronic V Channel: This analysis presents the first CMS search for boosted $H \to b\bar{b}$ in association with a hadronically decaying W or Z boson.
• Validation Strategy: The analysis simultaneously measures the $V(qq)Z(b\bar{b})$ process. Since this process shares the same selection criteria as the signal but has a known SM cross-section, it serves as an in-situ validation of the analysis strategy and background modeling.
• Advanced Machine Learning: The implementation of the mass-decorrelated PARTICLENET-MD classifier allows for robust flavor tagging of boosted jets while minimizing dependence on the jet mass, crucial for separating the Higgs signal from QCD backgrounds.

Results
The observed signal strength for the $V(qq)H(b\bar{b})$ process is:
$$\mu_{VH} = 0.72^{+0.75}_{-0.71}$$
This result includes both statistical and systematic uncertainties. The observed significance with respect to the background-only hypothesis is 1.00 standard deviations ($\sigma$), compared to an expected significance of 1.64 $\sigma$.

For the validation channel $V(qq)Z(b\bar{b})$, the observed signal strength is $\mu_{VZ} = 0.09^{+0.63}_{-0.63}$, with an observed significance of 0.15 $\sigma$ (expected 1.76 $\sigma$).

The results are consistent with the Standard Model prediction within uncertainties. The precision of the measurement is primarily limited by the available dataset size, specifically the finite number of events in the sideband regions used to constrain the QCD background via the transfer factor.

Significance
The paper claims that this measurement extends the CMS high-$p_T$ Higgs program into the hadronic vector boson decay channel, providing a complementary probe to leptonic searches. While the current result does not show a significant excess over the SM, it establishes the methodology for future analyses with increased luminosity. The modest significance reflects the current data limitations rather than a failure of the method, as evidenced by the successful validation using the $VZ$ process. The analysis demonstrates the capability to reconstruct and identify boosted Higgs and vector boson candidates in a challenging all-hadronic final state, setting the stage for more sensitive searches as the LHC accumulates more data.