Chapter 5: First evidence of a Higgs boson decay to a pair of muons
5.3 Physics objects
5.3.2 Muons
a straight line, and information from the muon chambers improves the π
T
measurement significantly, because they provide essential information about the track curvature.
One can define additional muon identification (ID) criteria based on several kine- matic variables: the number of hits and the fit quality of the track; the compatibility between the tracker hits and the segments in the muon stations; and the compati- bility between the muon track and the primary vertex. The global track fit π2and a kink-finder π2are used as indicators of the fit quality of the global muon track.
The compatibility between the tracker track and muon station segments is evaluated with a position match π2, and a variable called the segment compatibility. The compatibility between the track and the primary vertex is evaluated with the track impact parameters (closest distance of approach in the x-y plane or z axis). Based on these variables, muons can be further classified into three different types as follows:
β’ Loose muon ID: Muons that are either a tracker or a global muon, with no further requirements.
β’ Medium muon ID: A medium muon satisfies the loose ID criterion, and pos- sesses valid hits on more than 80% of the number of tracker layers that it tra- verses. Additionally, the muon must be classified either as agood global muon or should pass a tight segment compatibility requirement (> 0.451). A good global muon has a global track-fit π2/ndof < 3, the kink-finder π2 <
20, the position match π2< 12, and a segment compatibility > 0.303.
β’ Tight muon ID: A tight muon is a global muon with a global track-fitπ2/ndof
< 10, has at least one hit in the muon chamber and at least six hits in the inner tracker (with > 1 pixel hits). It also satisfies certain track impact parameter (w.r.t. the primary vertex) requirements, which are dπ₯ π¦ < 0.2 cm and dπ§ < 0.5 cm.
Events in theπ» β π π analysis are required to have at least two oppositely charged Medium ID muons with π
T > 20 GeV and |π| <2.4. Furthermore, each of the muons is also required to pass a relative isolation requirement in which the sum of the π
T of charged and neutral hadrons, and photons, lying within a cone of Ξπ =βοΈ
Ξπ2+Ξπ2=0.4 around the muon, is less than 25% of the π
Tof the muon.
If a photon in the isolation cone is associated with the final state radiation (FSR)
of the muon (see Section 5.3.2.2), then it is not included in the isolation sum. A Ξπ½-correction is also applied to the isolation sum which subtracts half of theπ
Tsum of charged hadron candidates lying within the isolation cone but originating from pile-up vertices. This correction helps to remove the contamination from neutral particles produced by pile-up interactions that spuriously enter into the vicinity of the candidate muon.
The sensitivity of this search depends critically on the muon π
Tresolution and also on the resolution of theππ π mass peak. Theπ
Tresolution of muons worsens with increasing muon |π|, the resolution being around 1β2% in the central barrel region of the detector (|π| < 0.9) , and degrading to 2 to 3.5% for muons passing through the endcaps of the muon system (|π|> 1.2), as shown in Fig. 5.2. The next few subsections will discuss ways to correct for any mis-reconstruction of the muon πT originating from software bias, uncertainties in the magnetic field or detector misalignment.
Figure 5.2: Resolution, as a function of π
T, for single, isolated muons in the barrel (|π|< 0.9), transition (0.9 <|π|< 1.4), and endcap regions (1.4 <|π|< 2.5). For each bin, the solid (open) symbols correspond to the half-width for 68% (90%) intervals of the residuals distribution [89].
5.3.2.1 Rochester Corrections
The muon momentum scale and resolution are calibrated in bins of muonπ
T,πand charge, usingπ β π πdecays as a standard candle, following the method described in Ref. [136]. The correction method is briefly summarized as follows
β’ Negative and positive muons are divided separately into differentπandπbins, for both data and simulation. In each bin, the 1/π
Tdistributions for both data and simulation are corrected, so that the mean value (<1/π
T >) becomes the same as that in theπ β π πsimulation (which is assumed to be very precisely modelled).
β’ A smearing is applied to the resolution of the 1/π
Tdistribution in the simula- tion, such that it matches the resolution in data.
β’ After all the above steps, there may still be some residual offset in eachππ π bin, when compared to the π β π π simulation. The last step is to apply the ratio of this offset and the nominal Z mass, as a correction factor to the muon π
T. The correction is applied in an iterative manner, until the offset is minimized.
5.3.2.2 Final state radiation (FSR) recovery
In a small fraction of signal events (9%), a muon in the final state may radiate a photon, thereby losing some of its momentum. This causes a slight degradation in the resolution of the signal ππ π peak. To recover this loss in the resolution, a procedure was developed to look for FSR photons within the isolation cone of the muon and can be summarized as follows
β’ Consider all reconstructed muons with π
T > 20 GeV and |π| < 2.4 as candi- dates for FSR recovery.
β’ For a given muon, consider all photons (πΎ) withΞπ (π, πΎ) < 0.5, π
T > 2 GeV, 0.0< |π| < 1.4442 or 1.566 < |π|< 2.5 as possible FSR candidates.
β’ Ignore photons that are associated with the bremsstrahlung of a reconstructed electron.
β’ In order to strongly suppress the contamination from Hβ ZπΎ β π ππΎdecays, FSR photon candidates are required to have π
πΎ T/π
π
T < 0.4.
β’ Impose a loose isolation requirement on the photon : πΌπΎ/π
T(πΎ) = (Ξ£π πΉ
π ππ
T(Ξπ (πΎ , π) < 0.3))/π
T(πΎ) < 1.8, where π
T(πΎ) is the π
T
of the FSR photon candidate and the indexπrefers to the PF candidates other than the muon within a cone ofπ =0.3 around the photon.
β’ Require photon to be collinear with the muon : Ξπ (π, πΎ)/π
T(πΎ)2 < 0.012
β’ In case of multiple FSR photon candidates, only the one with the smallest value ofΞπ (π, πΎ)/π
T(πΎ)2is considered.
If an FSR photon is associated with a muon, its momentum is added to that of the original muon. This procedure increases the signal efficiency by about 3% and improves theππ πresolution by around 2%.
5.3.2.3 GeoFit Corrections
In CMS, the muon π
Tvalues are primarily computed using the measured radius of curvature (π ) of the reconstructed muon track from hits in the inner tracker. This reconstruction has inherent uncertainties, which affect both the track trajectory and measured π
T. As prompt muons originate directly from the collision vertex, the measured point of closest approach between the muon track and the collision vertex in theπ₯βπ¦plane (known as the track impact parameter: π
0) should be exactly zero for muons coming from W/Z/H decays. However, if the muon track radius estimation is incorrect, theπ
0value will be non-zero, and is related to the mis-measurement in radius of curvature (Ξπ ) as follows (see [137] for derivation):
|π
0| βΌ Ξπ π 2
. (5.1)
Since in homogeneous magnetic fields, π
T βΌ π , we can re-write the above equation as
|π
0| βΌ |πΏ π
T| π2
T
. (5.2)
Therefore, the precision of the muonπ
Tmeasurement can be improved by including the interaction point position as an additional hit of the muon track. The correspond- ing adjustment in the π
T,πΏ π
T, is given by Eqn.5.2. The resulting improvement in the expectedππ πresolution in signal events ranges from 5% to 10%, depending on muon π
T,π, and the data-taking period (see Fig.5.3).