Mathis
Dubau

I am a particle physicist in the ATLAS collaboration. During my Doctorate at the Laboratoire d'Annecy de Physique des Particules (LAPP), I measured the single and joint longitudinal polarisation states of the W and Z bosons in Vector Boson Scattering (VBS), a process directly linked to electroweak symmetry breaking and among the rarest signatures accessible at the LHC.

With only a few tens of $\mathrm{W_0Z_0jj}$-EW events expected, my work focused on Machine Learning techniques to exploit the full information of the Monte Carlo simulations: Deep Neural Networks (DNN) and a binning built from clustering algorithms and Voronoi regions, which led to the highest sensitivity achieved so far for this channel within ATLAS. In parallel, I developed DNN-based identification and calibration of forward electrons for the HL-LHC in the $\mathrm{e}/\gamma$ Combined Performance group, and an active-learning framework to explore high-dimensional parameter spaces with theorists from LAPTh. My research, with the schemes →

Multivariate study of polarisation states in electroweak vector boson scattering WZ events using ATLAS data at the LHC

Université Savoie Mont Blanc, 2026 · supervisor Olivier Arnaez · CERN repository

This thesis presents the first measurement of the longitudinal polarisation states of one or both bosons in $\mathrm{WZjj}\text{-EW}$ processes, using the data recorded by the ATLAS experiment during the full Run-2 of the LHC ($140\,\mathrm{fb}^{-1}$ at $13\,\mathrm{TeV}$) combined with the first part of Run-3 ($164\,\mathrm{fb}^{-1}$ at $13.6\,\mathrm{TeV}$). Deep neural networks are trained to separate $\mathrm{WZjj}\text{-EW}$ events from the background and then to disentangle the polarisation states. Their outputs define a two-dimensional discriminant plane, in which a clustering algorithm builds an optimised Voronoi binning that maximises the sensitivity of a binned maximum-likelihood fit. The fraction in which the $\mathrm{Z}$ boson is longitudinally polarised regardless of the polarisation of the $\mathrm{W}$ boson is measured to be $f_{\mathrm{X0}}= 0.324^{+0.092}_{-0.085}$, for a prediction of $0.275$, with an observed (expected) significance of $4.0$ ($2.9$) standard deviations, providing the first evidence for longitudinally polarised $\mathrm{Z}$-boson states in $\mathrm{WZjj}\text{-EW}$ events. In the context of the High-Luminosity LHC, this thesis also presents a first calibration and identification of electrons in the forward region with the ITk, improving the background rejection at fixed identification efficiency and preparing future analyses with Run-4 data.

The analysis step by step → · Its schemes →

GitHub · LinkedIn · INSPIRE · ORCID · Google Scholar · ResearchGate