Search for anomalies in vector-boson fusion production of the Higgs boson in $H(\rightarrow γγ) jj$ events using 164 fb$^{-1}$ of $pp$ collision data collected at $\sqrt{s}=13.6$ TeV with the ATLAS detector

Abstract

This article details two studies of Higgs boson properties using the vector-boson fusion production mode and the $γγjj$ final state. Both efforts are based on a data sample corresponding to 164 fb$^{-1}$ of $\sqrt{s}=13.6$ TeV proton--proton collisions recorded by the ATLAS experiment at the Large Hadron Collider. The first study employs matrix element-based optimal observables to constrain CP-odd couplings beyond the Standard Model within the Standard Model Effective Field Theory framework, expressed in the Warsaw basis. The second study exploits angular distributions to probe the Higgs boson's couplings to longitudinally and transversely polarised $W$ and $Z$ bosons in the production of the Higgs boson. To maximise the sensitivity, the constraints of the CP-odd couplings are combined with those from a previous analysis performed in $γγjj$ events in a data sample of proton--proton collisions at $\sqrt{s}=13$ TeV, corresponding to an integrated luminosity of 140 fb$^{-1}$. A significant improvement with respect to the previous analysis is achieved through the implementation of a new neural network-based classification algorithm. All measurements are in agreement with the Standard Model prediction of a CP-even Higgs boson with the expected relative coupling strengths to longitudinally and transversely polarised vector bosons.

Figures (10)

• Figure 1: Representative lowest-order Feynman diagrams of vector-boson fusion production of a Higgs boson and subsequent decay into two photons $H\rightarrow \gamma\gamma$ via (a) a top-quark loop and (b) a $W$-boson loop.
• Figure 2: Distributions of (a) the $\mathcal{OO}$ observable shown for various configurations of the Wilson coefficient $c_{H\widetilde{W}}\xspace$, and (b) the $\Delta\Phi_{jj}\,$ observable for various configurations of the $a_\mathrm{L}\xspace$ and $a_\mathrm{T}\xspace$ parameters. These comparisons are performed in VBF events at the reconstruction level using the predictions of the MadGraph5_aMC@NLO + Pythia 8 Alwall:2014hcaSjostrand:2014zea generators.
• Figure 3: Comparison of the NN response score ($D_{NN}$) distributions for the VBF signal, the ggF background, and the non-resonant background, overlaid with sideband data. The non-resonant background includes simulated $\gamma\gamma$ continuum events as well as $\gamma+j$ and $jj$ processes where one or more jets are misidentified as photons; the latter are estimated from control regions in data as detailed in Section \ref{['sec:Modelling']}. The VBF signal and ggF background are scaled by factors of 400 and 200 relative to their SM expectations, respectively. Vertical dashed lines indicate the thresholds used to define the signal categories, while events with a $D_{NN}$ score below the lowest threshold (solid line) are excluded from the analysis.
• Figure 4: Post-fit distributions of (a) the optimal observable and (b) $\Delta\Phi_{jj}$ for events in the combined tight (T), medium (M), and loose (L) analysis regions within the invariant diphoton mass window $m_{\gamma\gamma}\in [120,130]\,\text{Ge V}\xspace$. The signal and background yields are fixed to the best-fit values from the $c_{H\widetilde{W}}$ and $a_\mathrm{L}$ interpretations. Contributions from the three regions are summed, weighted by $\ln(1+S/B)$, where $S$ and $B$ are the best-fit signal and background yields, respectively. The overflow and underflow are included in the outermost bins. The uncertainty band includes all systematic uncertainties. The lower panels show the background-subtracted data compared with the best-fit VBF prediction and two BSM scenarios corresponding to the two POI values excluded at exactly $95\%$ CL by fits using only shape information.
• Figure 5: Weighted post-fit distribution of the data events compared with the signal and background contributions for the $m_{\gamma\gamma}$ spectrum, summed over the tight (T), medium (M), and loose (L) analysis regions and optimal observable intervals. Events are weighted by $\ln(1+S/B)$, where $S$ and $B$ are the best-fit signal and background yields in each analysis region and optimal observable interval. The signal and background contributions are fixed to the values obtained from the best fit at $c_{H\widetilde{W}}=0.24$.