The identification and reconstruction of particle-flow candidates is carried out in each PF block separately. The muons are reconstructed first (tracker and muon chamber tracks linked), followed by electrons (tracks and ECAL clusters linked) and isolated photons (isolated ECAL clusters). After reconstruction of each type of PF candidate, the corresponding tracks and clusters are removed from the PF block. The hadrons and non-isolated photons, arising from fragmentation and hadronization in jets, are reconstructed last. The remaining ECAL+HCAL clusters with (without) a linked track
114 8. Event reconstruction
are identified as charged (neutral) hadrons, and the ECAL clusters without a linked track as nonisolated photons. Outside the tracker coverage, the ECAL+HCAL deposits are classified as "hadronic energy" and the isolated ECAL deposits as "electromagnetic energy", without associating them to any particles. Details on the reconstruction of each type of PF candidate are given in the following.
Muons:
As calorimeters absorb nearly all particles except muons and neutrinos, muons can be reconstructed with high purity based on signals in muon chambers. As discussed in Section 5.5, signals from the muon chambers can be reconstructed as hits with well-defined positions. While the RPC chambers contain only a single layer, the CSC and DT chambers have multiple layers, so hits in their subsequent layers are connected by fitting a straight line through them to form segments. Three types of muons can be reconstructed: standalone muons, global muons and track muons [163]. The joint collection of these different types of muons is referred to as PF muons. According to simulation studies, the muon reconstruction efficiency for PF muons is>99%. The standalone muons rely only on the muon chamber signals. They are built using the Kalman filtering method [164], by choosing the DT and CSC segments as seeds, and using pattern recognition to find associated hits in all muon detectors (CSC, DT and RPC). Then a track fit is performed using all these hits.
The global muons contain a standalone muon track matched to another track, recon- structed in the tracker system as described in Section 8.1.1. The matching is performed by extrapolating both tracks to a common surface. After matching, a global track fit is performed with the Kalman filter method.
The tracker muons are built by extrapolating reconstructed tracks from tracker to muon systems in the transverse plane, and requiring that at least one DT or CSC segment matches to the extrapolated track. The tracker muons complement the other approaches especially in the pT < 10 GeV range, where it is common that a proper
track cannot be reconstructed in muon chambers due to multiple scattering in the magnetic return yoke, and typically only one matching segment in the innermost muon stations is found. The muons need to have a minimum three-momentum of a few GeV to reach even the innermost muon stations. For muon with (almost) transverse trajectories, this corresponds to a few GeV, while for muons with a large longitudinal momentum component the pT threshold is even lower.
8.1. Particle flow algorithm 115
In the PF algorithm, first the isolated global muons are identified in the given PF block. The isolation is defined by requiring the sum of the pT of the tracks and the ETof the
calorimeter energy deposits to be less than 10% of the muon pT. For nonisolated muons
(typically found inside jets), tight selection criteria as described in Section 8.2.1 are required. Additionally, three track segments in muon chambers or lack of significant calorimeter deposits is required to suppress punch-through hadrons misidentified as muons. As the tight selection criteria require successful muon track reconstruction in both the muon chambers and the tracker, also muon candidates with a large number of hits in the tracker alone (tracker muons) or muon chambers alone (standalone muons) are accepted. The details are given in Ref. [155].
The momenta of muons are obtained from the curvatures of the corresponding tracks. The tracker system provides the best momentum resolution up to pT ≈ 200 GeV,
whereas for higher pTvalues the inclusion of muon chamber track curvature improves
the estimate. To optimize the momentum resolution across the momentum range, for each muon candidate above 200 GeV the pTis calculated using several combinations
of tracker and muon information, and the most precise result is chosen. Electrons and isolated photons:
As electrons pass through the tracker, they are likely to radiate bremsstrahlung photons. According to simulation studies, on average 33% (86%) of the electron energy is radiated for |η| ≈ 0 (|η| ≈ 1.4), corresponding to the smallest (largest) amount of tracker material to traverse before reaching ECAL. Both prompt and bremsstrahlung photons, on the other hand, are likely to convert to e+e− pairs, which then radiate
photons. Thus it is convenient to reconstruct electrons and isolated photons using a common approach.
While isolated electrons with large enough momenta are typically associated with ECAL superclusters, this is not always the case for non-isolated electrons inside jets or small-pT electrons with large track bending. Thus the PF electron candidates are
seeded either by an energy deposit in the ECAL (supercluster ET > 4 GeV) or by
reconstructed tracks with pT >2 GeV, linked with an ECAL supercluster. A BDT that
combines several variables describing the properties of calorimeter clusters, as well as the direction and quality of the associated tracks, is used to identify the electrons as described in Section 8.2.2.
Isolated photons are seeded by ECAL superclusters with ET >10 GeV not matched to
116 8. Event reconstruction
tracks and calorimeter clusters. The ratio of the ECAL and HCAL energy deposits is required to be compatible with an electromagnetic shower initiated by a photon. For the candidates passing these requirements, the direction and the corrected energy of the supercluster are taken as the PF photon energy and direction.
Hadrons and non-isolated photons:
Once muons, electrons and isolated photons have been reconstructed and the PF elements associated to them removed from the PF blocks, hadrons produced in frag- mentation and hadronization of quarks and gluons can be reconstructed. As e.g. neutral pions decay to photons, also the nonisolated photons are identified at this point.
All remaining ECAL clusters (with or without a linked HCAL cluster) within the tracker acceptance (|η| < 2.5) but without associated tracks are interpreted as noniso- lated photons. Similarly, all HCAL clusters without associated tracks are interpreted as neutral hadrons.
Outside the tracker acceptance the charged and neutral hadrons cannot be separated, so the classification is based on calorimeter information alone: the ECAL clusters with no linked HCAL clusters are interpreted as photons, while all HCAL clusters (with or without a linked ECAL cluster) are classified as hadrons.
For the calorimeter clusters associated with tracks, the sum of track momenta is compared to the calibrated cluster energy. Each track is interpreted as a charged pion, the momentum of which is recalculated with a combined fit in the associated track and calorimeter deposits. If the clustered calorimetric energy is larger than the track momenta, the additional ECAL energy is associated with photons and HCAL energy with neutral hadrons.
Finally, in cases where calorimetric energy is smaller than the sum of track momenta, additional global muons earlier obscured by the presence of other particles are identi- fied and added to the list of PF muons. Anomalous cases from misreconstructed tracks are solved by masking tracks with large pT uncertainty. Charged-particle tracks associ-
ated with secondary vertices are used to reconstruct the original charged hadrons that were present before nuclear interactions took place. The reconstructed hadrons are associated with the mass of a charged pion and used to replace their decay products in the list of final PF objects.