Eur. Phys. J. Plus (2022) 137: 870
https://doi.org/10.1140/epjp/s13360-022-03024-8

Regular Article

Learning to increase matching efficiency in identifying additional b-jets in the $\text{t}\overline{\text{t}}\text{b}\overline{\text{b}}$ process

¹ A.I. Institute, Hanyang University, 04763, Seoul, Republic of Korea
² Department of Computer Science, Hanyang University, 04763, Seoul, Republic of Korea
³ Department of Physics, Hanyang University, 04763, Seoul, Republic of Korea
⁴ Korea Institute for Advanced Study, 02455, Seoul, Republic of Korea

^f taekim@hanyang.ac.kr

Received: 6 July 2021
Accepted: 29 June 2022
Published online: 28 July 2022

Abstract

The $\text{t}\overline{\text{t}}\text{H}(\text{b}\overline{\text{b}})$ process is an essential channel in revealing the Higgs boson properties; however, its final state has an irreducible background from the $\text{t}\overline{\text{t}}\text{b}\overline{\text{b}}$ process, which produces a top quark pair in association with a b quark pair. Therefore, understanding the $\text{t}\overline{\text{t}}\text{b}\overline{\text{b}}$ process is crucial for improving the sensitivity of a search for the $\text{t}\overline{\text{t}}\text{H}(\text{b}\overline{\text{b}})$ process. To this end, when measuring the differential cross section of the $\text{t}\overline{\text{t}}\text{b}\overline{\text{b}}$ process, we need to distinguish the b-jets originating from top quark decays and additional b-jets originating from gluon splitting. In this paper, we train deep neural networks that identify the additional b-jets in the $\text{t}\overline{\text{t}}\text{b}\overline{\text{b}}$ events under the supervision of a simulated $\text{t}\overline{\text{t}}\text{b}\overline{\text{b}}$ event data set in which true additional b-jets are indicated. By exploiting the special structure of the $\text{t}\overline{\text{t}}\text{b}\overline{\text{b}}$ event data, several loss functions are proposed and minimized to directly increase matching efficiency, i.e., the accuracy of identifying additional b-jets. We show that, via a proof-of-concept experiment using synthetic data, our method can be more advantageous for improving matching efficiency than the deep learning-based binary classification approach presented in [1]. Based on simulated $\text{t}\overline{\text{t}}\text{b}\overline{\text{b}}$ event data in the lepton+jets channel from pp collision at $\sqrt{s}$ = 13 TeV, we then verify that our method can identify additional b-jets more accurately: compared with the approach in [1], the matching efficiency improves from 62.1$\%$ to 64.5$\%$ and from 59.9$\%$ to 61.7$\%$ for the leading order and the next-to-leading order simulations, respectively.