Performance of the HLT muon reconstruction used in CMS

The Large Hadron Collider (LHC) generates millions of collision per second. These collisions take place inside the detector, which should read out the data and store it for further analysis. However, there is no possible way of storing all the data generated at the LHC and, if it was, not all the events are interesting for further analysis since the majority of them are low-momenta events. Therefore, detectors need a trigger system to select which events are interested for further analysis and store them. In order to achieve that, the CMS trigger system is a two-level system composed by the Level1 and the High Level Trigger (HLT) that reduces the event rate from the collision rate at the LHC (30GHz) to what is stored and analyzed offline (around 1KHz). When a collision takes place, the Level1 trigger performs a fast readout of the detector. From the information of the detector, basic quantities are created and a pure hardware decision is taken on whether the event should be accepted or not. The Level1 reduces the rate from 30GHz to 100KHz. If the event is accepted by the L1 trigger, all the information passes to the High Level Trigger. The HLT is a software implemented system that decides whether the event is interesting for further analysis and, as a consequence, whether it should be stored. To do that, the HLT performs complex reconstructions as similar as possible as the ones executed offline. The focus of this proyect is the muon HLT reconstruction.The muon HLT reconstruction has changed during the last year taking advantage of the installation of the new pixel detector with the aim of improving performance, facilitating the maintenance and making use of an advanced iterative HLT tracking techniques. The new muon reconstruction is called IterativeL3 and it is an iterative method which comprises three consecutive sequences: Outside-In (OI), Inside-Out from L2 (IOL2) and Inside-Out from L1 (IOL1). Figure 1 illustrates how IterativeL3 works. As can be seen on the diagram, the starting point is always a L1 muon which is the muon candidate created from the L1 trigger. By using some of those L1 muons and information from the muon chambers, L2 muons are reconstructed and some of them are then used in the Outside-In sequence. The remaining L2 muon are used as the seed of the Inside-Out from L2 sequence. Finally, the L1 muons not used to reconstruct a L2 muons are the seed of the Inside-Out from L1 sequence. The three collections are then merged and the reconstructed muon must be accepted by the filter. Although the three sequences generate muons, they are conceptually different. The Outside-In sequence starts from the L2 muons generated using L1 muons and information from the muon chambers. From those muons, a trajectory is found, which is propagated to the outermost-tracker layer. A track candidate is then generated by finding compatible hits in the silicon trackers layers propagating inwards from the initial trajectory. Finally, the track candidate is matched with the L2 muons in order to reconstruct the muon. The L2 muons that are not used in the Outside-In sequence are the seed of the Inside-Out form L2 sequence. Around the L2 muon, a η-φ region is build inside which the full tracker reconstruction is run. During the tracker reconstruction, pixel quadruplets and triplets are created to generate pixel tracks, which are used ad seeds to build the track candidates. Then, the track candidates need to pass the high purity filters. From the tracker reconstruction, two track collections are created, which are merged afterwards. As in the case of the outside-in sequence, the track candidates are matched to the L2 muons and, if the match is positive,the muon is reconstructed. The process followed by Inside-Out from L1 is identical as the Inside-Out from L2 with the only difference that the muons used are L1 muons instead of L2 muons. As a consequence, the η-φ region has to be larger due to the poorer resolution of L1 muons.