Overview
This study presents a projected sensitivity analysis for hadronic mono-$Z$ dark matter production. The investigation utilized data from the Compact Muon Solenoid (CMS) experiment's Run 2015D HTMHT open dataset, which corresponds to an integrated luminosity of 2.256382381 $\text{fb}^{-1}$. From this dataset, 1,439,523 events met the criteria for hadronic mono-$Z$ selection. The research employs a machine learning approach, specifically a conditional flow-matching continuous normalizing flow, to model background processes.
Research Context
The core objective of the study involves searching for evidence of dark matter production in the hadronic mono-$Z$ channel. This channel is characterized by the production of a $Z$ boson that subsequently decays hadronically, in conjunction with missing transverse momentum, which is an experimental signature often associated with dark matter particles escaping detection. The analysis relies on publicly available experimental data from the CMS collaboration, specifically the Run 2015D HTMHT open data, which provides a foundation for sensitivity projections.
Approach
The methodology focused on background modeling and signal sensitivity projections:
- Data Selection: The study began with 2.256382381 $\text{fb}^{-1}$ of CMS Run 2015D HTMHT open data. A total of 1,439,523 events were identified as satisfying the hadronic mono-$Z$ selection criteria.
- Background Modeling: Background processes were modeled using a conditional flow-matching continuous normalizing flow. This flow was trained on the selected HTMHT events. Evaluation of the model occurred on a held-out validation split, which was then reweighted to represent the full selected population.
- Mitigation of Artifacts: Several techniques were applied to mitigate potential artifacts and biases. Sentinel imputation was used for undefined angular features, particularly concerning missing-object features. The train/validation split indices were persistently maintained. To avoid in-sample scoring bias, a minimum reported background yield of 20 events was enforced when determining the working point.
- Signal Processing: A signal-side offline trigger proxy was applied to simulated signal events before they underwent scoring.
- Ablation Study: An additional analysis, referred to as an ablation study, was conducted. This study involved removing the detailed extra-jet kinematics from the analysis to assess their contribution to the discriminating power.
Findings
The primary outcomes of the sensitivity study are as follows:
- Expected Significances: The baseline analysis procedure resulted in projected expected significances for three distinct simplified-model benchmarks. These significances were calculated as $2.89\sigma$, $7.62\sigma$, and $7.41\sigma$ respectively.
- Impact of Extra-Jet Kinematics: The ablation study revealed a substantial reduction in expected significance when detailed extra-jet kinematics were removed. This reduction ranged from 53% to 71%, indicating that the extra-jet topology possesses significant discriminating power within the hadronic mono-$Z$ channel.
- Study Status: The reported results represent projected sensitivities, and no unblinding of data was performed in this study.
Why This Matters
The study provides projections for the sensitivity to hadronic mono-$Z$ dark matter production using a specific open dataset and a machine learning technique. Identifying the substantial discriminating power of extra-jet kinematics offers insight into effective feature utilization for such searches.
Key Limitations Mentioned by Researchers
The study explicitly notes that its limitations and reproducibility are discussed in dedicated sections within the full paper, referred to as "limitations" and "reproducibility." The abstract, however, does not detail these specific limitations.