Speaker
Description
A calibratable experimental strategy is proposed to enhance the $HH(4b)$ search sensitivity via full-particle classification. Inspired by the competitive performance from the boosted-topology $HH$ analysis, which uses state-of-the-art jet neural networks to analyze $o(100)$ particles within large-$R$ jets, this approach aims to extend its strong signal-to-background discrimination power beyond the boosted regime to a broader phase space accessible through conventional $HH(4b)$ triggers.
The approach involves training a universal classifier to distinguish $X\rightarrow Y_1 Y_2 \rightarrow b\overline{b} b\overline{b}$ signals from QCD and $t\overline{t}$ multijet backgrounds across a wide range of $X$ and $Y_{1,2}$ mass values, and simultaneously estimating the $Y_{1,2}$ masses via a multiclass classification technique. Results demonstrate that the background suppression capability matches that of identifying two boosted $X\rightarrow b\overline{b}$ jets, revealing a scaling law governing signal and background yields in both cases. The framework is complemented by a robust signal calibration and validation procedure: event-level classier calibration is performed using an orthogonal dimuon-triggered phase space and an ``event hemisphere mixing’’ technique to construct fake $ZZ(4b)$ events; validation is then conducted using genuine $ZZ(4b)$ data passing the analysis trigger. With combined Run 2 and 3 datasets, the proposed strategy can achieve the first observation of the $ZZ(4b)$ process and deliver a search sensitivity for $HH(4b)$ comparable to HL-LHC projection. This approach holds a great premise to accelerate the pace of HH search at the LHC and advance our understanding of the Higgs self-coupling.
