Search for nonresonant new physics signals in high-mass dilepton events produced in association with b-tagged jets in proton-proton collisions at $\sqrt{s}$ = 13 TeV

Using 138 fb$^{-1}$ of proton-proton collision data at 13 TeV, the CMS experiment performed a search for nonresonant new physics in high-mass dilepton events associated with b-tagged jets, finding no significant deviations from the Standard Model and setting lower limits on the energy scale of effective field theory operators involving four-fermion contact interactions.

The Gist

Imagine the Large Hadron Collider (LHC) at CERN as the world's most powerful particle smashing machine. It takes two beams of protons (tiny subatomic particles) and crashes them together at nearly the speed of light. When they collide, they create a shower of new particles, much like smashing two watches together to see what gears and springs fly out.

Physicists usually look for specific, heavy "new" particles that pop out and then immediately decay. But this paper is about a different kind of search: looking for the "ghosts" of new physics.

Here is a simple breakdown of what the scientists did, using everyday analogies.

1. The Goal: Finding the Invisible Hand

The Standard Model of physics is like a very good, but incomplete, instruction manual for how the universe works. It explains almost everything we see, but it has holes (like dark matter or why gravity is so weak).

The scientists suspect that there might be a "new physics" force or particle that is too heavy to be created directly in the collision. Instead of seeing the heavy particle itself, they are looking for its footprints.

The Analogy: Imagine you are in a dark room and you hear a heavy door slam, but you can't see the door. You know someone is there because the air pressure changed. In this experiment, they aren't looking for the "door" (the new heavy particle); they are looking for the "air pressure change" (a subtle distortion in the energy of the particles that are created).

2. The Specific Search: The "B-Tagged" Clue

The scientists focused on a very specific type of crash. They looked for events where two heavy particles called b-quarks (bottom quarks) were produced alongside a pair of leptons (either two electrons or two muons).

• The Leptons: Think of these as the "messengers" flying out of the crash.
• The b-quarks: Think of these as the "heavy luggage" left behind.

They used a special detector called DeepJet (part of the CMS experiment) to identify these "luggage" items. It's like having a super-smart security scanner that can tell, "That suitcase definitely contains a bottom quark," with high accuracy.

3. The Two Theories They Tested

The team tested two different "stories" about what might be causing these footprints:

• Story A (The bbℓℓ Model): Imagine two people (leptons) shaking hands with two heavy suitcases (b-quarks) at the same time. In the Standard Model, this happens rarely and in a predictable way. The scientists asked: "Is this handshake happening more often or harder than our manual predicts?" They looked for a "contact interaction," which is like a magic handshake that happens instantly without a middleman.
• Story B (The bsℓℓ Model): This is a "flavor-changing" story. Imagine a bottom-quark suitcase magically turning into a strange-quark suitcase while shaking hands with the leptons. In our current physics manual, this is almost impossible (it's like a chameleon changing color instantly). If they saw this happening, it would be a huge sign of new physics.

4. The Method: The "Filter" and the "Scale"

To find these rare events, they had to filter through 138 "femtobarns" of data.

• The Analogy: Imagine trying to find a single specific grain of sand on all the beaches in the world. They collected data from 2016 to 2018 (about 138 fb⁻¹ of "beach sand").

They used a Deep Neural Network (DNN) as a super-smart filter.

• How it works: They trained a computer program (the DNN) on millions of simulated crashes. They showed it what "normal" crashes look like and what "suspicious" crashes look like. The DNN learned to spot the subtle differences, like a detective spotting a fake ID.
• The Result: They used this AI to throw away the boring, normal crashes and keep the ones that looked interesting.

5. The Results: The "Silence" is the News

After sifting through all that data, looking at the energy levels (mass) of the electron and muon pairs, they found nothing unusual.

• The Finding: The number of events they saw matched the Standard Model predictions perfectly. There were no "ghosts," no extra footprints, and no magic handshakes.
• The "Exclusion" Limits: Even though they didn't find new physics, they learned something valuable: New physics cannot exist below a certain energy level.
  • They set a "speed limit" for this new physics. If it exists, it must be heavier than 6.9 to 9.0 TeV (for the first model) or have a specific energy-to-coupling ratio of 2.0 to 2.6 TeV (for the second model).
  • The Analogy: It's like saying, "We didn't find a dragon in this forest, but we can now say for sure that if a dragon does exist, it must be bigger than a house." This pushes the search for new physics to even higher energies.

6. A Side Quest: Checking the Rules (Lepton Universality)

The scientists also checked a fundamental rule called Lepton Flavor Universality.

• The Rule: The universe should treat electrons and muons exactly the same way (like twins).
• The Check: They compared how often electron pairs appeared vs. muon pairs.
• The Result: The twins behaved exactly as expected. No cheating, no favoritism.

Summary

This paper is a report from the front lines of particle physics. The scientists took a massive amount of data, used advanced AI to filter it, and looked for subtle signs of new forces interacting with heavy quarks.

The verdict? The universe is behaving exactly as the current "instruction manual" (Standard Model) predicts. While they didn't find the "new physics" they were hoping for, they successfully raised the bar, proving that if new physics exists, it is hiding at even higher energy levels than we thought possible. It's a "no new physics found today" result, but a very important one that tells us exactly where not to look next.

Technical Summary

1. Problem and Motivation

The Standard Model (SM) of particle physics, while successful, fails to explain phenomena such as dark matter, the matter-antimatter asymmetry, and neutrino masses. Consequently, searches for New Physics (NP) at the multi-TeV scale are a primary focus of the LHC.

This analysis specifically targets nonresonant production of high-mass dilepton pairs ($e^+e^-$ or $\mu^+\mu^-$) in association with b-tagged jets. The motivation stems from:

• Third-Generation Sensitivity: The high mass of the top and bottom quarks makes the third-generation quark doublet particularly sensitive to NP.
• Flavor Anomalies: Experimental hints of Lepton Flavor Universality (LFU) violation in $B$-meson decays (e.g., $R(D^{(*)})$) suggest potential NP contributions involving $b$-quarks.
• Gap in Literature: While resonant searches and inclusive dilepton searches are common, this is the first CMS search specifically for nonresonant dilepton production associated with $b$-jets, probing specific Effective Field Theory (EFT) operators.

2. Methodology

Data and Detector

• Dataset: Proton-proton collision data collected by the CMS experiment during 2016–2018 at $\sqrt{s} = 13$ TeV.
• Integrated Luminosity: $138 \text{ fb}^{-1}$.
• Channels: Two lepton flavor combinations: dielectron ($ee$) and dimuon ($\mu\mu$).
• Event Categories: Events are classified by the number of $b$-tagged jets: 0, 1, or $\geq 2$.

Signal Models (EFT)

The analysis interprets results using two dimension-six EFT models involving four-fermion contact interactions (CI):

1. $bb\ell\ell$ Model: Interaction between two $b$-quarks and a lepton pair ($\ell\ell$). This models tree-level exchanges of heavy neutral gauge bosons ($Z'$) or scalar leptoquarks (LQs). Depending on whether the $b$-quarks are valence or sea quarks, the final state can have 0, 1, or 2 $b$-jets.
   • Interference: Both constructive ($\eta_{ij} = +1$) and destructive ($\eta_{ij} = -1$) interference with the SM Drell-Yan (DY) process are considered.
   • Chirality: All four chirality structures (LL, LR, RL, RR) are tested.
2. $bs\ell\ell$ Model: Flavor-changing neutral current (FCNC) transition $b \to s\ell^+\ell^-$. This is highly suppressed in the SM (loop-level) but could be enhanced by $Z'$ or LQs with flavor-violating couplings.
   • Target: Events with exactly 0 or 1 $b$-tagged jet.

Event Selection and Reconstruction

• Leptons: High-$p_T$ electrons ($p_T > 35$ GeV) and muons ($p_T > 53$ GeV).
  • Charge: Opposite-sign required for $\mu\mu$; no charge requirement for $ee$ due to high charge misidentification rates at high $p_T$.
• Jets: Reconstructed with anti-$k_T$ algorithm ($R=0.4$).
• $b$-tagging: Uses the DeepJet algorithm.
  • Tight working point: Required for the leading $b$-jet (50% efficiency, 0.1% light-jet mistag).
  • Relaxed working point: Used for additional $b$-jets (68% efficiency).
• Background Suppression:
  • In the 1b and 2b categories, the dominant background is $t\bar{t}$ production. A Deep Neural Network (DNN) is trained to distinguish signal from $t\bar{t}$ and other SM backgrounds using kinematic variables (lepton/$b$-jet $p_T$, $\Delta R$, $p_T^{miss}$, etc.). A DNN score cut of $>0.6$ is applied, rejecting 90% of $t\bar{t}$ events.
• Mass Range: Signal regions (SR) defined for $m_{\ell\ell} > 200$ GeV ($bs\ell\ell$) or $>300$ GeV ($bb\ell\ell$).

Background Estimation

• Dominant Backgrounds: Drell-Yan ($Z/\gamma^* + \text{jets}$) and $t\bar{t}$.
• Data-Driven Corrections:
  • $t\bar{t}$ Scale Factors ($SF_{tt}$): Derived from a Control Region (CR) using opposite-sign, different-flavor ($e\mu$) pairs.
  • DY Scale Factors ($SF_{DY}$): Derived from a DY CR ($120 < m_{\ell\ell} < 200$ GeV).
  • Fake Leptons: Estimated using data CRs with relaxed lepton ID (matrix method), contributing <3%.
• Validation: Validation Regions (VRs) are used to verify the modeling of $t\bar{t}$ and DY shapes after applying DNN selections.

Systematic Uncertainties

Key uncertainties include:

• Integrated luminosity (1.6% combined).
• Jet Energy Scale (JES) and Resolution (JER).
• $b$-tagging efficiency and mistag rates.
• Lepton trigger, reconstruction, and identification efficiencies.
• PDF and renormalization/factorization scale variations (significant for $t\bar{t}$ at high mass).
• Trigger prefiring effects (specifically for ECAL in 2016/2017).

3. Key Contributions

1. First Nonresonant $b$-Jet Search: This is the first CMS analysis specifically targeting nonresonant high-mass dileptons in association with $b$-jets, filling a gap between inclusive searches and resonant searches.
2. DNN-Based Background Rejection: The implementation of a DNN classifier specifically trained to suppress the $t\bar{t}$ background in the 1b and 2b channels significantly enhances sensitivity in these high-background regions.
3. Comprehensive EFT Interpretation: The analysis provides limits on both $bb\ell\ell$ (contact interaction) and $bs\ell\ell$ (FCNC) operators, covering various chirality structures and interference scenarios.
4. Lepton Flavor Universality (LFU) Test: A dedicated measurement of the ratio $R_{\mu\mu/ee} = \sigma(\mu\mu)/\sigma(ee)$ is performed, correcting for detector acceptance and resolution effects, to test for deviations from SM unity.

4. Results

• Observation: No significant excess of events over the SM background was observed in any of the signal regions (0b, 1b, 2b) for either $ee$ or $\mu\mu$ channels.
  • A slight excess ($<2\sigma$) was noted in the $ee$ channel for $m_{ee} > 1.9$ TeV in the 0b category, consistent with previous inclusive CMS results.
• $bb\ell\ell$ Model Limits:
  • Lower limits on the energy scale $\Lambda$ were set at 95% Confidence Level (CL).
  • Range: $\Lambda$ limits range from 6.9 TeV to 9.0 TeV, depending on the chirality structure and interference sign.
  • Significance: These results provide the first LHC constraints on the $bb\ell\ell$ model, significantly extending sensitivity beyond previous LEP bounds (which were tighter only for the LL scenario).
• $bs\ell\ell$ Model Limits:
  • Lower limits on the ratio $\Lambda/g^*$ (energy scale over effective coupling) were set.
  • Range: Limits range from 1.9 TeV to 2.6 TeV (with a combined limit of 2.4 TeV).
  • These results are comparable to existing ATLAS limits.
• Lepton Flavor Universality:
  • The ratio of $\mu\mu$ to $ee$ cross-sections was measured.
  • No significant deviation from the SM expectation of unity was observed. The $\chi^2$ test yielded a p-value of 0.05 for the 0b channel and 0.80 for the $\geq 1b$ channel, indicating good agreement with the SM.

5. Significance

• Constraint on BSM Physics: The analysis places stringent constraints on models predicting heavy $Z'$ bosons or leptoquarks that couple to the third generation of quarks and leptons.
• Validation of SM: The lack of deviation in the high-mass tails and the confirmation of Lepton Flavor Universality reinforce the Standard Model's validity up to the multi-TeV scale.
• Methodological Advancement: The successful application of DNNs to suppress $t\bar{t}$ backgrounds in high-mass dilepton searches demonstrates a robust technique for future high-luminosity LHC analyses where background rejection is critical.
• Future Prospects: The results set the stage for Run 3 and High-Luminosity LHC (HL-LHC) searches, where increased luminosity will allow for probing even higher energy scales and more subtle deviations in flavor universality.