14.5 Systematic uncertainties

We consider several sources of theoretical and experimental systematic uncertainties on the signal and background modelling in the signal regions, which are summarized in Table 14.3. Overall, the dominant source of uncertainty on the $\mathrm{\text{H}}\mathrm{\text{H}}$ signal strength is the statistical uncertainty in the QCD multijet background estimation, driven by the limited signal region sample size. This uncertainty has an impact on the best-fit signal strength of 50% (38%) relative to the overall uncertainty in the nonresonant (resonant) analyses.

The uncertainties on the signal efficiency of $\mathrm{\text{H}}\to \mathrm{\text{b}}\overline{\mathrm{\text{b}}}$ and $\mathrm{\text{H}}\to \mathrm{\text{V}}\mathrm{\text{V}}$ are also significant. The $\mathrm{\text{H}}\to \mathrm{\text{b}}\overline{\mathrm{\text{b}}}$ efficiency scale factors and uncertainties are measured for data versus simulation in a control region dominated by $\mathrm{\text{g}}\to \mathrm{\text{b}}\overline{\mathrm{\text{b}}}$ jets [164] and represent a 10% (15%) relative impact. The $\mathrm{\text{H}}∕\mathrm{\text{Y}}\to \mathrm{\text{V}}\mathrm{\text{V}}$ efficiency scale factor and uncertainty measurements are described in Chapter 13.3 and vary depending on the production mode and different coupling strengths of the nonresonant $\mathrm{\text{H}}\mathrm{\text{H}}$ signal, and the X and Y masses of the $\mathrm{\text{X}}\to \mathrm{\text{H}}\mathrm{\text{Y}}$ signal. Measured scale factors and uncertainties are shown in Table 14.4 and Table 14.5, respectively. Overall, the $\mathrm{\text{H}}∕\mathrm{\text{Y}}\to \mathrm{\text{V}}\mathrm{\text{V}}$ signal efficiencies represent around a 23% (50%) relative impact in the nonresonant (resonant) analyses.

Other significant sources of experimental uncertainty include the scale and resolution of the regressed jet mass and reconstructed jet energy [398] in data versus simulations. Jet mass corrections and uncertainties are measured in a control region enriched in $t\overline{t}$ events, using AK8 jets originating from hadronic W boson decays [399], and with a 7% (15%) impact, while jet energy corrections and uncertainties constitute 2%. Finally, there are large statistical uncertainties related to the MC simulations of the subdominant top quark and vector boson backgrounds, reaching $30\%$ in some bins of the signal regions; these have roughly a 10% impact.

On the theoretical side, there is a large uncertainty related to the nonresonant $\mathrm{\text{H}}\mathrm{\text{H}}$ production cross sections, which has a relative impact of up to 24% on best-fit signal strength. Uncertainties related to the parton showering performed using pythia 8 [400] are propagated to the MC kinematic distributions and have a 15% (5%) impact, while QCD renormalization and factorization scale uncertainties are estimated by considering the envelope of distributions obtained by varying the scales by a factor of $2$ and constitute a 5% impact.

Subdominant sources considered include uncertainties related to parton distribution functions (PDFs), H branching fractions, luminosity [401–403], pileup interactions, and trigger efficiencies, which have sub-percent-level impacts.