Kinematic Edge Detection Using Finite Impulse Response Filters

TL;DR

This paper introduces a robust FIR filter-based method, specifically using the first derivative of a Gaussian, for precise localization of kinematic edges in particle physics data, outperforming previous techniques.

Contribution

The paper presents a novel application of FIR filters, particularly FDOG, for edge detection in binned data, improving robustness and accuracy in particle physics measurements.

Findings

FIR filters outperform previous methods in edge detection accuracy.

FDOG kernel effectively localizes kinematic edges in binned data.

Method demonstrated on supersymmetric scenarios at future colliders.

Abstract

Various physics observables can be determined from the localisation of distinct edge-like features in distributions of measurement values. In this paper, we address the observation that neither differentiating nor fitting the measured distributions is robust against significant fluctuations in the experimental data. We propose the application of Finite Impulse Response (FIR) filters instead. To demonstrate the method, we consider the typical case in particle physics in which the precise localisation of kinematic edges, often blurred by e.g.\ background contributions and detector effects, is crucial for determining particle masses. We show that even for binned data, typical for high energy physics, the optimal FIR filter kernel can be approximated by the {\em first derivative of a Gaussian} (FDOG). We study two highly complementary supersymmetric scenarios that, if realised in nature, could be observed at a future high-energy $e^+e^-$ collider such as the International Linear Collider (ILC) or the Compact Linear Collider (CLIC). The first scenario considers the production of $\tilde{e}^{\pm}-$pairs while the second focuses on the $\tilde{\chi}^{\pm}_{1}$ and $\tilde{\chi}^{0}_{2}$-pair production. We demonstrate that the FIR filter method for edge extraction is superior to previously employed methods in terms of robustness and precision.