Search for flavour-changing neutral current couplings between the top quark and the Higgs boson in multilepton final states with the ATLAS detector

Using 140 fb$^{-1}$ of 13 TeV proton-proton collision data from the ATLAS detector, this study searches for flavour-changing neutral current interactions between the top quark and the Higgs boson in multilepton final states, setting observed upper limits on the branching ratios of $\mathcal{B}(t\to Hu)$ and $\mathcal{B}(t\to Hc)$ at $2.8 \times 10^{-4}$ and $3.3 \times 10^{-4}$, respectively.

The Gist

Imagine the universe as a massive, high-speed dance floor where particles are the dancers. In this dance, there are strict rules about who can partner with whom. The "Standard Model" is the rulebook that physicists have written down. According to this rulebook, a specific dancer called the top quark (the heaviest and most energetic dancer) is only allowed to partner with certain other dancers in very specific ways.

One of the strictest rules is that the top quark should never switch partners with a "Higgs boson" (the particle that gives others mass) and then immediately switch to a "charm" or "up" quark. This forbidden move is called a Flavour-Changing Neutral Current (FCNC). In the standard rulebook, this move is so rare it's practically impossible—like expecting a dancer to teleport across the room instead of walking.

However, some physicists suspect there might be a "secret rulebook" (New Physics) that allows this forbidden dance move to happen more often than we think.

The Investigation: A High-Stakes Search

The paper you provided describes a massive investigation by the ATLAS collaboration at the Large Hadron Collider (LHC). Think of the LHC as a giant particle smasher that collides protons together at nearly the speed of light, creating a chaotic explosion of particles. The ATLAS detector is like a giant, ultra-high-speed camera trying to capture every single move in that explosion.

The team looked at 140 trillion (140 fb⁻¹) of these collisions from 2015 to 2018. They were specifically hunting for two types of "dance routines" where the top quark might break the rules:

1. The "Decay" Routine: A pair of top quarks is created, and one of them spontaneously breaks the rules to turn into a Higgs boson and a lighter quark.
2. The "Production" Routine: A top quark and a Higgs boson are created together, but they are linked by a forbidden connection to an up or charm quark.

The Clues: Finding the Needle in the Haystack

The problem is that these rule-breaking events are incredibly rare, and the "haystack" (normal particle collisions) is enormous. To find the "needle," the scientists had to look for specific patterns, or "fingerprints," in the debris.

• The Signature: They looked for events where the debris included leptons (particles like electrons and muons) that had the same electric charge. In a normal dance, charges usually balance out. Finding two positive or two negative leptons together is a strong hint that something unusual happened.
• The Filter: They used computer algorithms (like a very smart bouncer) to filter out the millions of boring, normal events. They focused on events with specific energy levels and specific types of jets (sprays of particles) to ensure they were looking at the right kind of dance floor.
• The "Fake" Dancers: A major challenge was distinguishing real rule-breakers from "impostors." Sometimes, normal particles decay in a way that looks like a rule-breaker, or a detector mistake makes a particle look like it has the wrong charge. The team used statistical methods to estimate how many of these "fake" events were hiding in their data and subtracted them out.

The Verdict: No Rule-Breakers Found

After running their complex analysis, which involved training artificial intelligence (neural networks) to spot the subtle differences between a normal dance and a forbidden one, the results were clear:

They found zero evidence of the forbidden dance.

The number of suspicious events they saw was exactly what they expected to see if the Standard Model rules were being followed perfectly. There were no extra "rule-breakers" hiding in the data.

The Conclusion: Tightening the Rules

Because they didn't find the forbidden move, they didn't just say "we didn't find it." They calculated exactly how rare it must be.

• They set a new, stricter limit: If the top quark does break the rules to turn into a Higgs and an up quark, it happens less than 2.8 times out of 10,000.
• If it turns into a Higgs and a charm quark, it happens less than 3.3 times out of 10,000.

These numbers are the tightest constraints (the strictest rules) we have ever established. While they didn't discover "New Physics" in this specific search, they successfully closed the door on many theories that predicted this move would happen more often. It's like a detective searching a city for a specific criminal; even though they didn't catch the criminal, they proved that the criminal isn't hiding in the places they looked, forcing the criminal (or the theory) to be even more elusive than previously thought.

In short: The top quark is still following the rules, and the universe is still playing by the Standard Model's book—at least as far as this specific dance move is concerned.

Technical Summary

Technical Summary: Search for Flavour-Changing Neutral Current Couplings between the Top Quark and the Higgs Boson

Problem Statement
Flavour-changing neutral-current (FCNC) interactions involving the top quark and the Higgs boson ($t \to Hq$, where $q=u$ or $c$) are forbidden at tree-level in the Standard Model (SM) and are strongly suppressed by the Glashow-Iliopoulos-Maiani (GIM) mechanism, with predicted branching ratios of approximately $10^{-15}$. However, various Beyond the Standard Model (BSM) theories, such as two-Higgs-doublet models, predict significantly enhanced rates up to $10^{-3}$. Consequently, the observation of these rare interactions would provide direct evidence of new physics. This work addresses the search for such couplings using data from the Large Hadron Collider (LHC).

Methodology
The analysis utilizes 140 fb$^{-1}$ of proton–proton collision data recorded by the ATLAS detector at a centre-of-mass energy of 13 TeV during LHC Run 2 (2015–2018). The search targets two specific production mechanisms:

1. Decay Channel ('Dec'): Top-antitop ($t\bar{t}$) pair production where one top quark decays via the FCNC process $t \to Hq$.
2. Production Channel ('Prod'): Associated production of a top quark and a Higgs boson ($tH$), dominated by the $qg \to tH$ process.

Event Selection and Reconstruction
The analysis focuses on multilepton final states to enhance signal purity:

• 2 Same-Sign Leptons (2$\ell$SS): Events with two leptons ($e, \mu$) of the same charge.
• 3 Leptons (3$\ell$): Events with three leptons, requiring exactly two to have the same charge.

Selection criteria require at least one lepton with transverse momentum ($p_T$) $> 28$ GeV and additional leptons with $p_T > 10$ GeV. Events must contain at least one jet, with at least one $b$-tagged jet.

Background Estimation
A significant portion of the background arises from non-prompt leptons (originating mainly from $B$-hadron decays in $t\bar{t}$ production). These are estimated using a Template Fit Method, where normalization factors are determined via simultaneous maximum-likelihood fits in both the 2-lepton and 3-lepton final states. Control regions are defined to validate this method.

• Prompt Lepton Backgrounds: Processes such as $t\bar{t}W$ and $t\bar{t}Z$ are modeled with free-floating normalizations constrained by dedicated control regions.
• Charge Misidentification: In the 2$\ell$SS channel, backgrounds from prompt electrons with misidentified charges are modeled using a data-driven approach involving $Z$-boson mass regions.
• Other Backgrounds: Minor contributions are grouped into an "Others" category.

Signal Discrimination
To separate signal from background, two custom reconstruction algorithms are employed: Recursive Jigsaw Reconstruction and the Neutrino Independent Combinatorics Estimator. Variables derived from these algorithms are combined into a single discriminant using Artificial Neural Networks (DNN). The DNN output distribution serves as the input for a profile-likelihood fit to extract the signal normalization.

Key Results
The analysis found no significant excess of events over the SM background prediction. The best-fit signal normalization for both $tHu$ and $tHc$ signals was compatible with zero, with no discrepancies beyond $1\sigma$ for any nuisance parameters.

Upper exclusion limits at 95% confidence level (CL) were calculated using the $CL_s$ method:

• Branching Ratios:
  • $B(t \to Hu) < 2.8 \times 10^{-4}$ (Expected: $3.0 \times 10^{-4}$)
  • $B(t \to Hc) < 3.3 \times 10^{-4}$ (Expected: $3.8 \times 10^{-4}$)
• Wilson Coefficients: Limits were also set on the absolute value of the Wilson coefficient $|C_{u\phi}^{qt,tq}|$ for dimension-6 effective field theory (EFT) operators:
  • $tHu$:$0.71$ (Expected: $0.73$)
  • $tHc$:$0.76$ (Expected: $0.82$)

Significance and Combination
These results represent the strongest constraints to date on FCNC couplings between the top quark and the Higgs boson in the multilepton channel. The paper notes that these limits are further improved when combined with other ATLAS searches involving Higgs decays to $\tau\tau$, $b\bar{b}$, and $\gamma\gamma$. The combined analysis yields:

• $B(t \to Hu) < 2.6 \times 10^{-4}$
• $B(t \to Hc) < 3.4 \times 10^{-4}$

The work establishes new benchmarks for future studies of FCNC interactions, confirming the absence of such signals within the current sensitivity of the ATLAS detector.