Event Selection

100 %. It can be determined using following equation:

ε^trig_γγ =

"

dp^lead_T dp^sublead_T ε^trig_γ,35(p^lead_T )·ε^trig_γ,25(p^sublead_T )·f(p^lead_T ,p^sublead_T ). (5.2) Here, f(p^lead_T , p^sublead_T ) denotes the two-dimensional p.d.f. of the p_T for the leading and sub-leading photon candidates. The trigger efficiency of the single-photontrigger legof the diphoton trigger with a p_T threshold of 35 GeV (25 GeV) is denoted by ε^trig_γ,35(25). These single-photon trigger efficiencies, defined relative to isolated photon candidates which pass offline identification requirements, were measured using Z →`<sup>+</sup>`⁻γ events [191], similar to the method outlined in Section 4.3.1. The systematic efficiency uncertainty is given by the difference between the measured trigger efficiency and the trigger efficiency as estimated in simulation, and is propagated to the diphoton efficiency determination following Eq. (5.2). The relative systematic diphoton trigger efficiency uncertainty amounts to 0.7 %.

Object-Level Selection

Both photons are required to have a p^γ_T of larger than 25 GeV. The p^γ_T of the leading (sub-leading) photon candidate must also be larger than 35 % (25 %) ofm_γγ. Photon candidates are disregarded if they are located within the barrel-endcap transition region, 1.37<|η|<1.52. In order to ensure that the two considered photon candidates correspond to the objects that fired the diphoton trigger, a matching based on the angular distance between photon candidates and trigger-firing objects is performed. The photon candidates must pass the tightidentification selection, consisting of selection criteria regarding variables that parametrize the shape of the shower in the calorimeter; see Chapter 4 for details on the identification procedure. In addition, requirements on the isolation of the photon candidates are imposed; see Section 3.3.3.

Event-Level Selection

At least one primary vertex must be reconstructed, and all ATLAS detector components at the time of the event recording must have been functional. The invariant mass of the diphoton candidate must lie within 105 GeV<m_γγ<160 GeV. Choosing a range that is significantly larger than the typical range ofm_γγ fromH→γγevents enables a robust subtraction of non-Higgs background.

Selection of Jets and Leptons

Some of the measured differential cross sections describe jet-related properties of Higgs events.

In the measurements of differential cross sections for jet-related variables, jets are required to have a p_T>30 GeV and an absolute rapidity of|y_j|<4.4. In the region|y_j|<2.4,p_T<60 GeV an additional selection based on a multivariant discriminant calledjet vertex tagis applied to jets, reducing the number of jets from pileup interactions [192, 193].

As described in Section 3.3.3, a given shower in the calorimeter can be reconstructed as both photon and electron candidate simultaneously if the inner-detector information does not allow a unique categorization. This ambiguity is resolved at the stage of theH→γγanalysis by removing electron candidates that are close to a photon candidate: if an electron candidate has a distance∆R = p

∆φ²+ ∆y²to a photon candidate of less than 0.4, the electron candidate is disregarded. Similarly, if a jet overlaps with either one of the selected photon candidates (∆R<0.4) or a remaining electron candidate (∆R<0.2), it is removed. The reason for this is that electron candidates and photon candidates are also reconstructed as a jet by the jet reconstruction algorithm. By removing the corresponding jets, this ambiguity is resolved. Remaining electrons that are too close to a remaining jet (∆R<0.4) are removed for reasons of consistency with the electron isolation efficiency measurement, and because it removes electrons from decays of heavy-flavor hadrons. Muons overlapping with either selected photons or jets (∆R<0.4) are removed in the last step of the overlap removal.

Considered electrons must have a pTof larger than 10 GeV and a pseudorapidity within the range of 0<|η|<1.37 or 1.52<|η|<2.47. Selected muons must satisfy|η|<2.7, corresponding to the acceptance of the muon spectrometer, and p^µ_T>10 GeV. In addition, they must be isolated and identified [157].

Ab-jet is a jet originating in a hadron that contains ab-quark. In order to classify jets asb-jets at reconstruction level, ab-taggingalgorithm with an efficiency of 70 % is used [194]. To qualify as ab-jet in this analysis, a jet also must be central, i.e. the rapidity of that jet has to satisfy

|y_j|<2.5.

5.3.2. Particle Level

The fiducial particle-level selection is based on final-state particles with a lifetime larger than 10 ps, calledstable, and which do not result from the GEANT4 simulation of particle-detector interactions. Moreover, photons, electrons and muons are disregarded if they result from a hadron decay.

To a large extent, the particle-level selection mirrors the selection at reconstruction level.

Photons must satisfy p^γ_T>25 GeV. Moreover, the p^γ_Tof the leading (sub-leading) photon must exceed 35 % (25 %) of the invariant mass of the leading and sub-leading photons mγγ. The diphoton invariant-mass is required to be within the range 105 GeV<m_γγ <160 GeV. The pseudorapidity of selected photons must lie within 0<|η|<1.37 or 1.52<|η|<2.37. A particle-level isolation selection is applied to photons; it is defined as the transverse energy of the summed four-momenta of all charged particles within a cone of radius ∆R<0.2 around the photon, considering only particles with a p_Tof at least 1 GeV. If this quantity is larger than 5 % of a photon’s transverse momentum, the photon is disregarded.

The four-momenta of electrons and muons include those of stable photons within a cone with radius∆R<0.1 in order to recover the lepton energy lost by bremsstrahlung processes. Electrons with a pseudorapidity within 1.37<|η|<1.52 or a pTlower than 10 GeV are disregarded. Muons are required to have a pseudorapidity within|η|<2.7 and a p_Tlarger than 10 GeV. The overlap between photons and electrons is removed by disregarding electrons that are closer than∆R<0.4 to one of the two leading photons.

Jets at particle-level are reconstructed based on stable particles, excluding neutrinos and muons, because those do not deposit sizable amounts of energy in the calorimeter, and therefore are effectively not accounted for by reconstruction-level jets. The absolute rapidity of jets is required to be within|y_j|<4.4, and their p_T to be larger than 30 GeV. In order to remove the overlap of jets with photons, a jet is removed if it is closer than∆R<0.4 to a photon with apT

of at least 25 GeV. Similarly, if the distance between a jet and an electron with a p_Tof at least 10 GeV is smaller than∆R<0.2, the jet is disregarded.

A jet that satisfies the requirements above and which contains ab-hadron with a p_T larger than 5 GeV and an angular distance to the jet direction smaller than∆R<0.4 is considered to be ab-jet. Additionally, in order for ab-jet to be considered in the analyses, it must lie within

|y_j|<2.5.

5.3.3. Dalitz Events

Not all photons fromH→γγdecays are stable; some photons from Higgs boson decays have high virtuality, i.e. they do not obey the energy-momentum relation, resulting in a quick decay to a pair of charged particles. The occurrence of such H→γ^∗γ→ ff¯γevents, calledDalitz events, needs to be considered in order to properly correct the extracted number of signal events for a comparison with H→γγ predictions. About 6 % of the simulated events are Dalitz events.

These events are not considered in the computation of theH→γγbranching ratio. As they do not result in a stable diphoton final state, they are not part of the fiducial selection at particle level.

The remaining 94 % are reweighted such that the predictedH→γγproduction cross section is matched again.

BecauseH→γ^∗γ→ ff¯γevents have a low efficiency of passing the diphoton selection as described in Section 5.3.1, Dalitz events constitute only about 0.4 % of the selected diphoton events at reconstruction level according to simulation; these selected Dalitz events are not removed. Therefore, the unfolding of the measured cross sections to particle level removes the expected impact of Dalitz events on the cross section.