The ATLAS Roma1 group explores physics beyond the Standard Model through complementary strategies: model-agnostic anomaly detection, emerging and semi-visible jet signatures, high-mass resonances in ZZ→4ℓ, and leptoquark pair production.
Anomaly Detection
Anomaly detection searches aim to reveal unexpected phenomena without specifying a signal model. They learn the structure of Standard Model data and flag outliers. Strategies include unsupervised density/autoencoding models (VAEs, normalizing flows) and weakly supervised approaches (e.g. CWoLa/LLP) trained on data sidebands or minimally trusted simulation.
Key ideas: build an anomaly score that is robust to background mismodelling; decorrelate from resonant variables to avoid sculpting (adversaries/planing); interpret high-score clusters with physics-aware features and cross-checks; explore online/trigger-level deployment for rapid sensitivity.
Roma1 contribution: design and validation of ML pipelines, jet/substructure studies, background modelling, decorrelation and systematics.
ATLAS Briefing · Recent AD example
Jets + ETmiss: monojet, semi-visible & emerging jets
The classic monojet signature (one high-pT jet + large ETmiss) generalises to a broader family of jet + ETmiss searches that target dark-sector scenarios beyond the WIMP paradigm. In semi-visible jets, dark-hadronisation yields a mixture of visible and invisible particles inside the jet, so the missing momentum can be aligned with the jet. In emerging jets, long-lived dark hadrons decay with displaced tracks/vertices that “emerge” at finite radii, often with unusual track/cluster timing and substructure. These signatures provide powerful, complementary coverage of BSM models with invisible or long-lived states.
Roma1 contribution: definition of monojet and ISR-based selections, ETmiss/jet calibration, control-region and transfer-factor design, displaced/track-based validation for long-lived scenarios, and statistical interpretation.
Review of collider dark-matter searches · APS/PRD: semi-visible & emerging-jet concepts
Resonance Searches: X → ZZ → 4ℓ
The 4ℓ (“golden”) final state offers outstanding mass resolution and background control. Building on the full Run-2 search, the Run-3 programme at 13.6 TeV updates the strategy for heavy resonances decaying to ZZ, keeping the clean 4ℓ core and extending acceptance to topologies with jets and/or missing transverse momentum when relevant. Reconstruction improvements (e.g. Z-mass–constrained fits and refined treatment of FSR) further sharpen the m4ℓ resolution and the high-mass reach.
Roma1 contribution: electron/muon selection and calibration, ZZ reconstruction with FSR and mass-constraint, background controls, high-mass lineshape modelling and statistical interpretation.
Run-2 baseline · Run-3 performance · 4ℓ with jets/ETmiss (extended topologies)
Leptoquark Searches
The pair-produced scalar/vector leptoquarks connect quark and lepton sectors, yielding multi-lepton + jets (often b-jets) or τ signatures. ATLAS combines channels to set stringent, model-dependent limits and to scan generation-specific couplings and branching ratios, leveraging b/τ tagging and kinematic shapes.
The pair-produced scalar/vector leptoquarks connect quark and lepton sectors, yielding multi-lepton + jets (often b-jets) or τ signatures. ATLAS combines channels to set stringent, model-dependent limits.
Roma1 contribution: b-tagging and τ-ID performance, multi-lepton selections, improved fake-factor modelization, background estimation and combinations.
Run3 Leptoquark search · ATLAS LQ combination (full Run2)
Main sources: anomaly detection (ATLAS Briefing 2023; arXiv:2306.03637; arXiv:2502.09770), monojet 139 fb−1 (arXiv:2102.10874; PRD 103, 112006), X→ZZ (EPJC 81:332, 2021; arXiv:2401.04742), leptoquarks (arXiv:2305.15962; 2306.17642; 2401.11928).