Data-driven Model-independent Searches for Long-lived Particles at the LHC

TL;DR

This paper introduces a data-driven, model-independent approach for long-lived particle searches at the LHC, significantly enhancing sensitivity across a broad range of masses and lifetimes by reducing background prediction uncertainties.

Contribution

It proposes a novel, general strategy for background estimation in LLP searches, enabling more inclusive and sensitive detection methods at the LHC.

Findings

Enhanced sensitivity for LLP proper lifetimes > 10 m.

Extended reach to low-mass glueballs in Neutral Naturalness models.

Applicable to various signal models and detector subsystems.

Abstract

Neutral long-lived particles (LLPs) are highly motivated by many BSM scenarios, such as theories of supersymmetry, baryogenesis, and neutral naturalness, and present both tremendous discovery opportunities and experimental challenges for the LHC. A major bottleneck for current LLP searches is the prediction of SM backgrounds, which are often impossible to simulate accurately. In this paper, we propose a general strategy for obtaining differential, data-driven background estimates in LLP searches, thereby notably extending the range of LLP masses and lifetimes that can be discovered at the LHC. We focus on LLPs decaying in the ATLAS Muon System, where triggers providing both signal and control samples are available at the LHC Run-2. While many existing searches require two displaced decays, a detailed knowledge of backgrounds will allow for very inclusive searches that require just one detected LLP decay. As we demonstrate for the $h \to X X$ signal model of LLP pair production in exotic Higgs decays, this results in dramatic sensitivity improvements for proper lifetimes $\gtrsim 10$ m. In theories of Neutral Naturalness, this extends reach to glueball masses far below the $\bar b b$ threshold. Our strategy readily generalizes to other signal models, and other detector subsystems. This framework therefore lends itself to the development of a systematic, model-independent LLP search program, in analogy to the highly successful simplified-model framework of prompt searches.