Understand how different selections of ZZ candidate, both at gen and reco level, (and ZZ reconstruction) affect acceptance and efficiency values used in the analysis.

[bonanomi]

@AlessandroTarabini presentation - CJLST meeting: https://indico.cern.ch/event/976616/#12-run2-legacy-plans-different

[bonanomi]

ggH

ZH

[bonanomi]

GEN_ variables: Generator ZZ candidate selected following analysis selection.

• Z1 candidate: selected by taking the 2l pair that builds a Z candidate with M(Z1) closest to M (Z) PDG.
• Z2 candidate: all the other 2l pairs are build and the one with largest sum of pT is chosen.

Gen_ variables: Generator ZZ candidate selected using true MC history. Hence in this case only events coming from H->ZZ->4l decay chain are considered. Of course, this has the largest impact on associated productions (WH, ZH, ttH).

Only the former allows to: "In order to minimise the dependence of the measurement on the specific model assumed for Higgs boson production and properties, the fiducial phase space for the H→4l cross section measurements is defined to match as closely as possible the experimental acceptance defined by the reconstruction-level selection."

As we see in GENmassZ2 above, there is a striking difference between ggH and ZH. This striking difference comes from the fact that events building the associated Z can be picked up using the GEN_ approach, but would be discarded using the Gen_ one. Also note that, by definition, the Gen_ approach is a non-complete collection of the events selected using the GEN_ approach.

Looking at GENmassZ2 and GENmassZ1 it seems that some wrong events are being picked (i.e. GENmassZ2 does not have a sharp cutoff at 35 GeV as expected). However this is necessary to evaluate the non-resonant contribution when building the pdf and creating the fiducial volume.

When building response matrices we want to make sure we are selecting GEN-level events following as closely as possible reco-level events selection. This is done correctly only when using the GEN_ approach, which is the only one allowing to:

"Furthermore, an algorithm for a topological selection closely matching the one at the recon-struction level is applied as part of the fiducial phase space definition. At least two SFOSlepton pairs are required, and all SFOS lepton pairs are used to form Z boson candidates.The SFOS pair with invariant mass closest to the nominal Z boson mass (91.188 GeV) is takenas the first Z boson candidate (denoted as Z1). The mass of the Z1candidate must satisfy40<m(Z1)<120 GeV. The remaining set of SFOS pairs are used to form the second Z bosoncandidate (denoted as Z2). In events with more than one Z2candidate, the SFOS pair with thelargest sum of the transverse momenta magnitudes,Σ|pT|, is chosen. The mass of the Z2candi-date must satisfy 12<m(Z2)<120 GeV. Among the four selected leptons, any pair of leptons li and lj must satisfy∆R(lilj)>0.02. Similarly, of the four selected leptons, the invariant massof any opposite-sign lepton pair mustsatisfy m(lilj)>4 GeV. Finally, the invariant mass ofthe Higgs boson candidate must satisfy 105<m4l<140 GeV."

The two sets of plots above also answer the question on why the acceptances, efficiencies and fraction of non fiducial events (i.e. all the coefficients used as input to build pdfs) change so much when using GEN_ instead of Gen_ approach, especially for the associated productions. Events peaking at -1, present in both methods, are not selected in the analysis and only represent the fraction of events for which it was not possible to build the ZZ candidate: it's clear how dramatic this effect is on associated productions.

In addition, using the Gen_ approach instead of the GEN_ one, would enhance the model-dependence of the measurement. As, quoting from [1], using the ZZ candidates from GEN_ allows to:

"The inclusive values of the factor(1+fnonfid)e from Eq. (1) are shown in Table 3 for different signal production modes and different exotic models. The relatively weak dependence of this factor on the exact signal model is a consequence of the particular definition of the fiducialphase space introduced in Section 5, and enables a measurement with a very small dependence on the signal model."

[bonanomi]

Even though the "correct" way of selecting ZZ candidates might be with the use of topological selection, the events at mZ2 >~ 60GeV in the associated productions are: at reco level due to detector resolution and at gen level are coming from the fact that we build all lepton pairs starting from all the gen leptons.

This being said, we are not really interested in these events and one should get rid of them. One way of doing it could be introducing a cut at mZ2 < 60 GeV and check the effect on the analysis.

[bonanomi]

mZ2 in ttH

[bonanomi]

Related to this: #21

[bonanomi]

Distribution of events with mZ2>60 GeV

[bonanomi]