Polarization fraction measurement in same-sign $WW$ scattering using deep learning

Studying the longitudinally polarized fraction of $W^{\pm }{W}^{\pm }$ scattering at the LHC is crucial to examine the unitarization mechanism of the vector boson scattering amplitude through Higgs and possible new physics. We apply here for the first time a deep neural network classification to extract the longitudinal fraction. Based on fast simulation implemented with the Delphes framework, significant improvement from a deep neural network is found to be achievable and robust over all dijet mass region. A conservative estimation shows that a high significance of four standard deviations can be reached with the High-Luminosity LHC designed luminosity of $3000{\mathrm{fb}}^{-1}$.

Figure 1

Simplified structure of the particle-based DNN model. Inputs are features of each particle, gradually merging to the output layers.

Figure 2

ROC curves with a selection of $m_{\mathrm{jj}}>1500\mathrm{GeV}$ applied to VBS $W^{\pm }{W}^{\pm }$. The X-axis showing signal efficiency of LL component, and Y-axis the rejection rate of $\mathrm{TT}+\mathrm{TL}$ components.

Figure 3

Kinematic distributions for $p_T^{l_1}$ (a), $\mathrm{\Delta }{\varphi }_{\mathrm{jj}}$ (b), and DNN discriminant corresponding to the particle-based model (c).

Figure 4

Fit results for the LL fraction using particle-based DNN in $\mathrm{Mjj}>1500\mathrm{GeV}$ region. The horizontal lines represent the 68%, 90%, and 95% confidence level, from lower to upper. Apart from those, the solid and dotted lines are the log likelihood distributions, with or without systematics included.

Figure 5

Significance dependence on $m_{\mathrm{jj}}$ selection, through fitting $p_T^{l_1}$, $\mathrm{\Delta }{\varphi }_{\mathrm{jj}}$, or the DNN discriminant. Greatly improved performance is found with DNN.