Probing the triple Higgs boson coupling with machine learning at the LHC

Murat Abdughani,Jin Min Yang,Daohan Wang,Jun Zhao,Lei Wu

doi:10.1103/physrevd.104.056003

Abstract

Measuring the triple Higgs boson coupling is a crucial task in the LHC and future collider experiments. We apply the message passing neural network (MPNN) to the study of nonresonant Higgs pair production process $pp\ensuremath{\rightarrow}hh$ in the final state with $2b+2\ensuremath{\ell}+{E}_{\mathrm{T}}^{\mathrm{miss}}$ at the LHC. Although the MPNN can improve the signal significance, it is still challenging to observe such a process at the LHC. We find that a $2\ensuremath{\sigma}$ upper bound (including a 10% systematic uncertainty) on the production cross section of the Higgs pair is 3.7 times the predicted Standard Model cross section at the LHC with the luminosity of $3000\text{ }\text{ }{\mathrm{fb}}^{\ensuremath{-}1}$, which will limit the triple Higgs boson coupling to the range of $[\ensuremath{-}3,11.5]$.

Highlights

The discovery of a 125 GeV Higgs boson [1,2] is a great leap in the quest to the origin of mass
We explored the discovery potential of Higgs pair production process pp → hh → bbWWÃ → 2b þ 2l þ EmT iss with the message passing neural network at the (HL-)LHC
In the message passing neural network (MPNN), we can represent each collision event as an event graph that consists of the final state objects and use the supervised learning to optimize training parameters

Summary

INTRODUCTION

The discovery of a 125 GeV Higgs boson [1,2] is a great leap in the quest to the origin of mass. [9,10,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68], the potential of measuring the Higgs pair production has been investigated in various decay modes: bbbb , bbτþτ−, bb WWÃ, γγbb , γγWWÃ, and WWÃWWÃ Among these channels, the process of hh → 4b has the largest branching ratio, while the process of hh → bbγγ has a more promising sensitivity because of the low backgrounds. Given the importance of the Higgs pair production, in this work we apply the machine learning method message passing neural network (MPNN) [75] to explore the potential of observing such diHiggs events through the channel pp → hh → bb WWÃ.

EVENT GENERATION AND RECONSTRUCTION

X X XX

EVENT GRAPH AND NETWORK ARCHITECTURE

RESULTS AND DISCUSSIONS

CONCLUSIONS