Sensitivity on Two-Higgs-Doublet Models from Higgs-Pair Production via $b\bar{b}b\bar{b}$ Final State

TL;DR

This paper demonstrates how machine learning enhances the detection of Higgs pair production in two-Higgs-doublet models at the HL-LHC, improving sensitivity to model parameters through analysis of the $b\bar{b}b\bar{b}$ final state.

Contribution

It introduces a three-stream convolutional neural network to improve signal-background discrimination in Higgs pair production within two-Higgs-doublet models.

Findings

Machine learning significantly improves sensitivity coverage.

The process can probe parameter space beyond current limits.

Results are applicable to Types I to IV of the models.

Abstract

Higgs boson pair production is well known to probe the structure of the electroweak symmetry breaking sector. We illustrate using the gluon-fusion process $pp \to H \to h h \to (b\bar b) (b\bar b)$ in the framework of two-Higgs-doublet models and how the machine learning approach (three-stream convolutional neural network) can substantially improve the signal-background discrimination and thus improves the sensitivity coverage of the relevant parameter space. We show that such $gg \to hh \to b \bar b b\bar b$ process can further probe the currently allowed parameter space by HiggsSignals and HiggsBounds at the HL-LHC. The results for Types I to IV are shown.