Search for a Time-Dependent Z' Resonance in the Dimuon Channel

This paper introduces a novel time-domain search strategy using a two-dimensional unbinned likelihood framework to detect time-varying $Z'$ resonances in CMS dimuon data, demonstrating that incorporating temporal information can enhance sensitivity to signals with periodically modulated masses compared to conventional time-integrated analyses.

The Gist

Imagine you are a detective trying to find a specific type of car in a massive, chaotic parking lot. In a standard investigation, you would take a photo of the entire lot, count every red sports car, and look for a sudden spike in the number of red cars compared to the background noise. This is how particle physicists usually search for new particles: they look for a "bump" or a spike in the data that stands out from the ordinary background.

However, this paper proposes a different kind of detective work. It suggests that some new particles might not be stationary; instead, they might be shapeshifters.

The Shapeshifting Car

The authors are looking for a hypothetical particle called a Z' boson (think of it as a heavy, invisible cousin of the Z boson, which is a known particle). In this specific theory, this Z' boson is being influenced by a mysterious, invisible "wind" called ultralight dark matter.

Imagine the Z' boson is a car that changes its color and size every few hours.

• Standard Search: If you take a photo of the parking lot over 24 hours and count all the "red" cars, you might miss the Z' boson. Why? Because for half the day, it looks red, and for the other half, it looks blue. If you just count "red cars," you dilute the signal. The car blends in with the background because you aren't looking at when it changed color.
• The New Strategy: The authors developed a method to watch the parking lot live. Instead of just counting cars, they track the car's color over time. They realized that if a car is changing color in a predictable, rhythmic pattern (like a heartbeat), you can spot it even if it's hiding in a crowd of normal cars.

The "Time-Dependent" Detective Work

The paper describes a new mathematical tool (a "two-dimensional likelihood") that looks at two things at once:

1. Mass: How heavy the particle is (like the car's size).
2. Time: When the particle was detected.

In a normal experiment, physicists ignore the "time" part and just look at the "mass" part. But this paper argues that if a particle's mass is wiggling back and forth because of the dark matter wind, you need to watch the movie, not just the snapshot.

The Experiment

The team tested this idea using real data from the CMS experiment at the Large Hadron Collider (CERN).

• The Data: They looked at collisions that produced pairs of muons (heavy electrons) from a specific period of data collection (Run G).
• The Model: They simulated a scenario where the Z' boson's mass oscillates (wiggles) up and down over a period of about 5.7 hours.
• The Result: They found that by using their "time-aware" method, they could set stricter rules on where this shapeshifting particle could be hiding. Even though they didn't find the particle (which is expected, as it's hypothetical), their method proved to be more sensitive than the old "snapshot" method.

The Key Takeaway

The paper claims that if new particles exist that change their properties over time due to interactions with dark matter, ignoring the time factor makes you blind to them.

By treating the data as a flowing river rather than a frozen pond, the researchers showed they can spot these "wiggling" signals much better. They demonstrated that for certain types of heavy particles, looking at when they appear is just as important as looking at what they are. This opens a new door for finding physics that would otherwise remain invisible to traditional searches.

In short: If you are looking for a ghost that changes its shape every hour, you can't just take a photo and hope to see it. You have to watch the video. This paper built the video camera and showed that it works better than the old photo camera for this specific type of ghost hunt.

Technical Summary

Technical Summary: Search for a Time-Dependent Z′ Resonance in the Dimuon Channel

Problem Statement
Conventional high-energy collider searches for new physics resonances, such as a $Z'$ boson, rely on time-integrated event samples. These methods assume that signal properties, specifically the invariant mass, are stationary. However, this approach is insensitive to signals where the mediator's mass undergoes periodic temporal modulation. Such phenomena arise in scenarios involving ultralight bosonic dark matter, where a coherently oscillating scalar field induces time-dependent variations in the effective mass of vector bosons. In these cases, the signal does not appear as a stationary peak in the invariant mass distribution but rather traces a nontrivial trajectory in the $(m, t)$ plane. Standard time-integrated analyses dilute such signals, particularly when the modulation amplitude is large, rendering them invisible to conventional "bump hunt" strategies.

Methodology
To address this limitation, the authors developed a time-domain search strategy utilizing a two-dimensional unbinned likelihood framework in the observable space of invariant mass ($m$) and event time ($t$).

• Theoretical Model: The study benchmarks a $Z'$ boson arising from a gauged $U(1)_{B-L}$ extension of the Standard Model. The $Z'$ mass is modulated by an oscillating ultralight scalar dark matter field ($\phi$). The effective mass is parameterized as $m_{Z'}^2(t) = \tilde{m}_0^2 (1 + \kappa \cos^2(m_\phi t))$, where $\kappa$ represents the modulation amplitude. A benchmark scalar mass of $m_\phi \simeq 10^{-19}$ eV is used, corresponding to an oscillation period of $\tau \simeq 20,660$ s (approx. 5.7 hours).
• Signal Modeling: The signal probability density function (PDF) is constructed as a Gaussian in invariant mass with a time-dependent mean $\mu(t)$, explicitly capturing the correlation between mass and time. The signal PDF is multiplied by a time-dependent acceptance function $T(t)$ to account for variations in detector conditions.
• Background Modeling: The background consists of resonant $Z$-boson production (modeled with a double Crystal Ball function) and the non-resonant Drell-Yan continuum (modeled with a log-curved power law). Both components share the same time-dependent acceptance function $T(t)$, derived from the observed $Z$-boson yield in individual data-taking runs, ensuring the background model reflects data-taking conditions without introducing artificial mass-time correlations.
• Statistical Framework: The analysis employs an extended unbinned likelihood ratio test statistic within a modified frequentist approach. Systematic uncertainties are handled via nuisance parameters. Upper limits are derived using the $CL_s$ prescription at the 95% confidence level.

Data and Implementation
The method was implemented using dimuon events from the CMS Open Data corresponding to the "Run G" data-taking period at $\sqrt{s} = 13$ TeV, with an integrated luminosity of $7.54 \text{ fb}^{-1}$. Selection criteria required opposite-sign muon pairs with $p_T > 10$ GeV and $|\eta| < 2.4$, with invariant masses between 50 GeV and 3000 GeV. The use of a single data-taking period (Run G) was chosen to maintain consistent detector and running conditions while preserving the necessary temporal structure for the analysis.

Key Contributions

1. Framework Development: The paper introduces a fully unbinned, two-dimensional likelihood framework that treats invariant mass and time as correlated observables, allowing for the direct reconstruction of signals that evolve in time.
2. Signal Discrimination: The study demonstrates that the time-dependent signal possesses a non-factorizable structure in the $(m, t)$ plane, which is distinct from Standard Model backgrounds that only exhibit time dependence through the data-taking profile (luminosity variations).
3. Sensitivity Enhancement: The authors show that incorporating temporal information acts as an additional discriminating observable, enhancing sensitivity to new physics without necessarily increasing the total number of signal events.

Results
The analysis derived upper limits on the gauge coupling $g'$ as a function of the $Z'$ mass for various modulation amplitudes ($\kappa = 0, 8, 24$).

• Comparison with Standard Searches: The time-dependent analysis ($\kappa > 0$) demonstrated a systematic improvement in sensitivity compared to the conventional time-integrated search ($\kappa = 0$).
• Modulation Amplitude: Larger modulation amplitudes ($\kappa$) resulted in lower excluded values of $g'$, indicating the ability to probe weaker couplings. This improvement is most pronounced in low- and intermediate-mass regions where event statistics are sufficient to resolve the temporal modulation.
• High-Mass Limitations: At higher mediator masses, the improvement in sensitivity diminishes due to the rapidly decreasing signal production cross-section, which limits the number of events available to reconstruct the time-dependent modulation.

Significance and Claims
The paper claims that time-domain analyses provide a complementary and effective strategy for probing new physics scenarios involving time-dependent mediator properties. By exploiting correlations between invariant mass and time, this approach can access regions of parameter space that remain inaccessible to standard time-integrated searches. The authors emphasize that while the current study uses a limited subset of CMS Open Data, the observed trends validate the feasibility of the time-domain approach. They assert that the method offers a systematic gain in sensitivity for scenarios where the mediator mass evolves during data taking, a signature that would otherwise be diluted or missed in conventional resonance searches. The work does not claim the discovery of new physics but rather establishes the methodology and its potential to enhance exclusion limits in specific theoretical contexts.