Search for $H\rightarrow c\bar{c}$ and measurement of $H\rightarrow b\bar{b}$ via $t\bar{t}H$ production

Using 136 fb$^{-1}$ of 13 TeV proton-proton collision data from the CMS detector, this study simultaneously measures the $t\bar{t}H(H\rightarrow b\bar{b})$ signal and searches for the $t\bar{t}H(H\rightarrow c\bar{c})$ process, finding the former consistent with Standard Model predictions while setting an observed upper limit on the charm quark Yukawa coupling modifier at 3.0 times its SM value.

The Gist

Imagine the universe is a giant, high-speed particle collider, like a cosmic pinball machine where we smash protons together at nearly the speed of light. When these particles collide, they sometimes create a very special, heavy, and short-lived ball called the Higgs boson.

This paper is about two main things:

1. Confirming a known trick: We already knew the Higgs boson sometimes turns into a pair of "bottom" quarks (heavy particles). We wanted to see if our detectors could catch this happening when the Higgs is produced alongside a pair of "top" quarks (the heaviest particles).
2. Hunting for a ghost: We wanted to see if the Higgs boson ever turns into a pair of "charm" quarks. This is much harder because charm quarks are lighter and look very similar to other common particles in the chaos of the collision.

Here is the breakdown of the story using some everyday analogies:

1. The Setup: The "Top" Party

Think of the collision as a wild party. Usually, we just see a mess of debris. But sometimes, a VIP guest (the Higgs boson) shows up, but they only arrive if they are accompanied by two very heavy bodyguards (the Top quarks). This is called $ttH$ production.

Once the Higgs arrives, it immediately disappears (decays). We need to figure out what it turned into:

• The Easy Case ($H \to bb$): It turns into two "bottom" quarks. These are like big, heavy suitcases. They are easy to spot because they are heavy and leave a distinct trail.
• The Hard Case ($H \to cc$): It turns into two "charm" quarks. These are like small, lightweight backpacks. In a room full of people carrying backpacks, it's incredibly hard to tell which ones came from the VIP guest and which ones were just there anyway.

2. The Challenge: The "Needle in a Haystack" Problem

The biggest problem is that the "background noise" is huge. Most of the time, the collision just creates a pile of random junk (jets) that happens to look like bottom or charm quarks.

• The Haystack: Millions of collisions where random particles look like charm quarks.
• The Needle: The rare event where the Higgs actually decays into charm quarks.

To find the needle, the scientists used AI detectives. They trained two super-smart computer brains:

• ParticleNet: This AI acts like a bouncer at a club. It looks at every particle (jet) and asks, "Are you a bottom quark? A charm quark? Or just a regular light particle?" It sorts them into different VIP sections based on how confident it is.
• Particle Transformer (ParT): This AI acts like a Sherlock Holmes. It doesn't just look at individual particles; it looks at the whole party scene. It analyzes how the jets, electrons, and missing energy are arranged to decide: "Is this a Higgs party, or just a random noise party?"

3. The Results: What Did They Find?

The "Bottom" Confirmation (The Success Story)
The scientists looked for the Higgs turning into bottom quarks ($H \to bb$).

• Result: They found it! The number of times they saw it matched the predictions of the Standard Model (our rulebook for physics) almost perfectly.
• Significance: They were 4.4 standard deviations sure this wasn't a fluke. In science, 5 is the gold standard for a "discovery," so 4.4 is a very strong "we definitely saw this." It's like flipping a coin 100 times and getting 90 heads; you're pretty sure the coin is weighted.

The "Charm" Hunt (The Limit)
Now, the hunt for the Higgs turning into charm quarks ($H \to cc$).

• Result: They didn't find a clear signal. They couldn't say, "Yes, the Higgs definitely turns into charm quarks here."
• The Limit: However, they set a very strict boundary. They said, "If the Higgs does turn into charm quarks, it happens less than 7.8 times more often than our theory predicts."
• The Yukawa Coupling: This is a fancy way of asking, "How strong is the Higgs' handshake with the charm quark?" The scientists measured this "handshake strength" (called $\kappa_c$). They found it is less than 3 times the strength predicted by the Standard Model.

4. Why Does This Matter?

Think of the Standard Model as a map of the universe. We know the Higgs talks to heavy particles (like the Top quark) and medium particles (like the Bottom quark). But the "Charm" quark is in the middle ground.

• If the Higgs talks to the Charm quark much stronger than we think, it could mean there is new physics hiding in the shadows—maybe a secret force or a new particle we haven't discovered yet.
• By proving that the "handshake" isn't wildly strong (it's less than 3x normal), this paper helps us rule out some wild theories and confirms that our current map is still mostly correct.

The Bottom Line

The CMS team used 138 "years" worth of data (collected over 2016–2018) and super-advanced AI to:

1. Confirm they can see the Higgs turning into bottom quarks when produced with top quarks.
2. Prove that if the Higgs turns into charm quarks, it's doing so at a rate that is still consistent with our current understanding of the universe, though we need more data to be absolutely sure.

It's a bit like saying, "We found the VIP guest's heavy luggage, but we haven't found their light backpack yet. However, we know for sure they aren't carrying too many light backpacks."

Technical Summary

Based on the provided paper, here is a detailed technical summary of the search for $H \to c\bar{c}$ and the measurement of $H \to b\bar{b}$ via $t\bar{t}H$ production by the CMS Collaboration.

1. Problem Statement

The Standard Model (SM) predicts the existence of Yukawa couplings between the Higgs boson and fermions. While the coupling to the third-generation top and bottom quarks has been well-established, the coupling to the second-generation charm quark ($H \to c\bar{c}$) remains elusive. Direct measurement is challenging due to:

• Small Branching Ratio: The SM branching ratio for $H \to c\bar{c}$ is approximately 2.9%, compared to ~58% for $H \to b\bar{b}$.
• Background Contamination: The dominant background, $t\bar{t} + \text{jets}$, frequently produces charm and bottom jets, making it difficult to distinguish the signal from background processes.
• Flavor Tagging: Distinguishing charm jets from bottom jets and light-flavor jets requires high-precision discrimination, which is complicated by the similar kinematic properties of heavy-flavor jets.

This analysis aims to probe the charm Yukawa coupling by searching for the associated production of a Higgs boson with a top-quark pair ($t\bar{t}H$) where the Higgs decays to charm quarks, while simultaneously measuring the $t\bar{t}H(H \to b\bar{b})$ process to validate the analysis framework.

2. Methodology

Data and Event Selection

• Dataset: Proton-proton collisions at $\sqrt{s} = 13$ TeV collected by the CMS detector (2016–2018), corresponding to an integrated luminosity of 138 fb$^{-1}$.
• Channels: The analysis considers all three decay modes of the $t\bar{t}$ system:
  • Fully hadronic (0L): 0 leptons.
  • Semileptonic (1L): 1 lepton.
  • Dileptonic (2L): 2 leptons.
• Kinematic Selection: Events are selected based on jet multiplicity (7, 5, and 4 jets for 0L, 1L, and 2L respectively), missing transverse momentum ($p_T^{\text{miss}} > 20$ GeV for 1L/2L), and scalar sum of jet $p_T$ ($> 500$ GeV for 0L).

Machine Learning Techniques

The analysis employs a two-stage machine learning approach:

1. Jet Flavor Tagging (ParticleNet):

   • A graph-neural-network-based algorithm, ParticleNet, is used to tag jets.
   • It outputs discriminant scores to categorize jets into light, charm ($c$), and bottom ($b$) flavors.
   • Categorization: Jets are binned into specific efficiency categories (e.g., $B1\text{--}B4$ for $b$-jets with ~70% efficiency; $C1\text{--}C4$ for $c$-jets with ~50% efficiency).
   • Selection: Events require at least 3 jets tagged as $b$ or $c$, with at least one specifically tagged as a $b$-jet.

2. Event Classification (Particle Transformer / ParT):

   • Selected events are processed by a transformer-based neural network, Particle Transformer (ParT).
   • Inputs: Properties and pairwise features of event objects (jets, leptons, $p_T^{\text{miss}}$).
   • Training: The network is trained separately for each channel to assign likelihood scores for:
     • Signals: $t\bar{t}H(H \to c\bar{c})$, $t\bar{t}H(H \to b\bar{b})$, $t\bar{t}Z(Z \to c\bar{c})$, and $t\bar{t}Z(Z \to b\bar{b})$.
     • Backgrounds: $t\bar{t} + \text{jets}$ is split into 5 classes based on the flavor of extra jets ($t\bar{t}+b$, $t\bar{t}+\ge2b$, $t\bar{t}+c$, $t\bar{t}+\ge2c$, $t\bar{t}+\text{LF}$). In the 0L channel, a QCD multijet class is added.
   • Region Definition:
     • Signal Regions (SR): Events with high signal scores ($D_{ttX} > 0.85$) are assigned to 4 SRs enriched in specific signal processes.
     • Control Regions (CR): Events with lower scores fall into 5 CRs to constrain the normalization of the $t\bar{t} + \text{jets}$ backgrounds.
     • QCD Suppression: In the 0L channel, events with $D_{\text{QCD}} < 10^{-4}$ are selected to reduce QCD multijet background to negligible levels.

Statistical Analysis

• A binned profile likelihood fit is performed on the ParT discriminant distributions in each category.
• Parameters: Signal strengths ($\mu$) for each process are fitted as ratios of observed yield to SM expectation.
• Constraints: The five $t\bar{t} + \text{jets}$ background normalizations are left floating. Experimental and theoretical uncertainties are included as nuisance parameters.
• Interpretation: Results are interpreted in the $\kappa$-framework, where $\kappa_c$ and $\kappa_b$ represent modifiers of the charm and bottom Yukawa couplings relative to the SM.

3. Key Contributions

• Novel Channel: This is a dedicated search for $t\bar{t}H(H \to c\bar{c})$, a channel previously unmeasured with high precision.
• Advanced ML Integration: The successful application of ParticleNet for jet flavor separation and Particle Transformer (ParT) for complex event classification in a high-multiplicity environment ($t\bar{t}H$).
• Validation: The use of $t\bar{t}Z(Z \to b\bar{b})$ and $t\bar{t}Z(Z \to c\bar{c})$ as control processes to validate the flavor tagging and background modeling.
• Simultaneous Measurement: A unified framework measuring $H \to b\bar{b}$ and searching for $H \to c\bar{c}$ within the same dataset and analysis strategy.

4. Results

$t\bar{t}H(H \to b\bar{b})$ Measurement

• Signal Strength: $\mu_{t\bar{t}H(H \to b\bar{b})} = 0.91^{+0.26}_{-0.22}$.
• Significance: The observation is compatible with the SM prediction, with a significance of 4.4$\sigma$ above the background-only hypothesis.
• Uncertainty: The dominant systematic uncertainty arises from the theoretical prediction of the $t\bar{t} + \ge2b$ background.

$t\bar{t}H(H \to c\bar{c})$ Search

• Upper Limit: No significant excess was observed. The observed (expected) upper limit on the signal strength is $\mu_{t\bar{t}H(H \to c\bar{c})} < 7.8$ (8.7) at 95% Confidence Level (CL).
• Charm Yukawa Coupling ($\kappa_c$):
  • Assuming SM-like couplings for all other Higgs interactions (fixing $\kappa_b$), the observed (expected) upper bound on the charm Yukawa coupling modifier is $\kappa_c < 3.0$ (3.3).
  • This improves upon previous limits set by CMS and ATLAS using vector-boson-associated Higgs production.
• Combined Limit: When combined with previous CMS results from this channel, the observed (expected) limit becomes $\kappa_c < 3.5$ (2.7).

Validation ($t\bar{t}Z$)

• Measured signal strengths for $t\bar{t}Z(Z \to c\bar{c})$ and $t\bar{t}Z(Z \to b\bar{b})$ were $1.02^{+0.79}_{-0.84}$ and $1.47^{+0.45}_{-0.41}$, respectively. Both are compatible with SM expectations and previous measurements, validating the analysis strategy.

5. Significance

This work represents a significant step forward in probing the second generation of fermions in the Higgs sector. By utilizing the $t\bar{t}H$ production mode and state-of-the-art deep learning techniques (ParticleNet and ParT), the CMS Collaboration has:

1. Established a robust measurement of $t\bar{t}H(H \to b\bar{b})$ with 4.4$\sigma$ significance, confirming the SM prediction in a complex final state.
2. Set the tightest constraints to date on the charm Yukawa coupling in the $t\bar{t}H$ channel, excluding values of $\kappa_c$ greater than 3.0 at 95% CL.
3. Demonstrated the feasibility of separating charm jets from bottom jets in high-multiplicity environments, a critical capability for future Higgs physics and potential New Physics searches involving heavy flavor.

The results are statistically dominated, suggesting that with the increased luminosity of the High-Luminosity LHC (HL-LHC), the sensitivity to the charm Yukawa coupling will improve significantly, potentially allowing for a direct observation of $H \to c\bar{c}$ in the $t\bar{t}H$ channel.