In addition to the signal region several orthogonal control and validation regions were defined to help understand and validate different backgrounds. These regions are constructed to isolate a single background with high purity. Generally a control region is used to compare the background prediction to the data, tweaking parameters like normalization factors and in more complicated situations, the shape, so that the control region background models the data well. In the validation region the background is then adjusted in the same way as the control region and compared with the data. If all is well in the validation region, the same tweaks are done to the prediction in the signal region to give a background estimate. The regions defined in this section are used for the analysis in order to construct background estimates for the signal region. There are other control and validation regions used to determine the different data-driven mis-identification rates which are outlined in the sections for electron-to-photon fakes, Section 7.4.1.1, and jet-to-photon fakes, Section 7.4.2.1. A summary of these regions, including the signal region, can be seen in Table 7.4. Table 7.4: Selection criteria for the different control, validation, and signal regions used in the analysis. The ‘(electron)’ in the W Z control region signifies the condition of anMedium electron being required in place of the signal photon. The ‘(pseudoγ)’ in the jet control region signifies the condition of a pseudo photon being required in place of the signal photon.
Cut Control Reg.W Z Control Reg. jets Valid. Reg. jets Control Reg.Zγ Valid. Reg.Zγ Signal Reg.
Pass triggers and vetoes X X X X X X
2 signal leptons X X X X X X
At least 1 signal photon >25GeV(electron) >25GeV(pseudoγ) >25GeV >25GeV >25GeV >25GeV mwin
`` 81−101GeV 81−101GeV 85−120GeV 81−101GeV 81−101GeV 81−101GeV
Emiss
T >95GeV >95GeV >35GeV 20−35GeV 35−70GeV >95GeV
BalpT <0.2 <0.2 - <0.2 <0.2 <0.2
∆φ``,γEmiss
T >2.8 >2.8 <2.2 - - >2.8
∆φ(`, `) <1.4 <1.4 - <2.0 <2.0 <1.4
7. Search for Exotic Higgs Decays 128
except that instead of requiring a signal photon, aMedium electron is required with the same signal region kinematic topology selections. Eventually, the electron-to-photon mis-identification rate is applied to this region in order to estimate the background in the control region. A validation region with similar kinematics is very hard to carve out while having meaningful statistics and low signal
contamination. This is because the W Z → eν`` kinematics match the signal region kinematics.
Thus a validation region, without aZ boson in the event, was defined as a region with two photons and exactly one electron. This region contains a small amount ofW γγ contamination but most of the events come fromZγ events where one of the electrons from theZ is mis-identified as a second photon.
Analogously, there is a control region for the jet-to-photon fakes which is constructed in the same way as the signal region though rather than a signal photon, a pseudo photon is required. The jet- to-photon mis-identification rate is then applied in order to get a signal region background estimate. A validation region is defined using unbalanced events containing a signal photon + two leptons in an asymmetric invariant mass window. This asymmetric window is utilized in order to remove the radiative Zγ contamination. The BalpT and ∆φ(`, `) cuts are dropped, the E
miss
T requirement is
relaxed, and the∆φ``,γEmiss
T cut is inverted such that events are required to have the dilepton system
unbalanced with theγEmiss
T system.
The Zγ background shape is taken from simulation, while the normalization is derived from a
control region. This control region is similar to the signal region, but with relaxing the ∆φ(`, `) requirement, dropping the∆φ``,γEmiss
T criterion, and requiring substantially less missing transverse
momentum in the event. The validation region is then defined in the same way but requiring the missing transverse momentum to be in an intermediate range, greater than the control region but less than the signal region.
Composition and Signal Leakage of Control and Validation Regions
Aside from the mis-identified objects control regions, which will be discussed in depth in Section 7.4, the composition and purity of the three other regions can be seen in Figure 7.7. The purity of targeted background is over50% for all regions, though the validation region for the jet-to-photon fakes has
lower purity than other regions. This is due to the contribution from Zγ with real photons and
because the photon identification of ATLAS detector is excellent at rejecting mis-identified photons. These control and validation regions were also designed to have as little signal leakage as possible in order to avoid prematurely “unblinding” the analysis. The percentage of signal events entering
Figure 7.7: Composition of the control and validation regions mentioned in the text. The purity of the desired background is noted in as a percentage of total events.
each region can be seen in Figure 7.8. The highest percentage of signal leakage is less than2.5% in the Zγ validation region. This is due to that region being closest, kinematically to the signal region. The other regions have less than0.5%of events due to signal processes. This holds true for the mis-identification control regions used for the signal region fake background estimates.