3.1 The Large Hadron Collider

Transcription

1 3 T H E L H C b E X P E R I M E N T 3.1 The Large Hadron Collider The Large Hadron Collider (LHC) is a particle collider based at CERN, accelerating two beams of hadrons, either protons or lead ions. It has been designed to collide bunches of hadrons at a frequency of 40 MHz at a centre-of-mass energy of s = 14 TeV, for a maximum instantaneous luminosity 1 of cm 2 s 1. Figure 8 shows a scheme of the CERN accelerators complex. The proton beam originates from a hydrogen gas source. The gas is injected into a metal cylinder surrounded by an electrical field and is thus broken down into the constituent protons and electrons. Initially, protons are accelerated up to 750 kev; a first spacing in bunches is achieved by means of a radio frequency pulse. The proton beam enters then the linear accelerator (LINAC2) and accelerates up to 50 MeV. From the Linac protons enter the 157 m circumference of the Proton Synchrotron Booster (PSB), a circular four-rings accelerator. Here they are accelerated to 1.4 GeV and injected into the 628 m long Proton Synchrotron (PS), which accelerates protons to 25 GeV and provides a 25 ns-spaced scheme of bunches for the LHC. The bunches formed in the PS are injected into the 7 km circumference of the Super Proton Synchrotron (SPS) and accelerated to 450 GeV. At this point, the protons are ready to enter the 26.7 km long LHC, both in a clockwise (beam1) and anticlockwise (beam2) direction. In LHC particles are accelerated by means of Radio-Frequency (RF) cavities, which are metallic chambers molded in such a shape and size to contain a resonant oscillating electromagnetic field. To make sure that charged particles passing through the cavity always feel the overall accelerating force of the field, the RF frequency is tuned to an integer multiple of the particle s revolution frequency. The frequency of the LHC RFs is 400 MHz and the protons reach a speed very close to the light speed c. The two LHC beams are made up of cylinder-like bunches, separated by about 7.5 m (corresponding to a time separation of 25 ns), nominally 7.55 cm long and with a transverse size at the LHCb interaction point of σ = 70.9 µm [66]. To reach its design luminosity, the collider operates with a maximal number of bunches, while each bunch contains up to protons [67]. The individual bunches are identified by the RF bucket they occupy; in LHCb, the bunch-crossing identifier (BCID) is used to label the slots (3564 in total) along the beam. Due to injection and operation constraints, the maximum number of filled bunches per beam is Depending on whether a bunch-slot is actually empty or filled with a bunch, the bunch-crossings are distinguished in LHCb as: beam-beam (bb) when both slots of beam1 and beam2 are filled, beam-empty (be) or empty-beam (eb), when the bunch-slot of one of the two is empty, and empty-empty (ee), if both slots are empty. Each category of crossings is selected adopting a specific trigger strategy, as explained in Sec The luminosity is a measurement of the number of collisions per second and per unit of area (cm 2 ), see Sec for more details. 21

2 22 The LHCb experiment Figure 8... Sketch of the CERN accelerator complex adapted from Ref. [65]. To steer the circulating particles around the ring, the LHC uses about 9300 superconducting dipole magnets cooled to a temperature below 2 K and operating at magnetic fields up to 8 T. Quadrupole magnets are used to focus the beam in the x, y directions. In a high luminosity collider, there is a significant probability that a single bunch crossing may produce multiple proton-proton interactions, so-called pile-up events. Figure 9 shows a collision event with multiple primary interactions reconstructed in the LHCb VErtex LOcator (see Sec ). The pile-up of events has introduced a challenge for the event selection and reconstruction and has driven LHCb to develop trigger tools for cleaning the data sample from too crowded events. The probability of having n visible interactions in an LHC bunch crossing can be described by a Poissonian distribution: P(n) = µn n! e µ, (53) where the mean µ represents the average number of pp interactions detected in the experiment. Under the assumption of a 100% efficient L0-trigger, the experiment will record all events that contain at least one visible pp interaction, with probability P(n 1). The pile-up, µ PU, is then the average number of pp interactions in visible events and is defined as the zero-suppressed mean of the distribution in Eq. 53: µ µ PU = 1 P(0) = µ 1 e µ. (54)

3 3.1 The Large Hadron Collider 23 Figure 9... Collision event (from 2010 data taking) with multiple primary interactions reconstructed in the LHCb VErtex LOcator (VELO) and displayed by the LHCb Event Display. The two Pile-Up (PU) system modules installed at the upstream end of the VELO modules are highlighted in the yellow box; see Sec for more details. The first LHC collisions were delivered in November 2009, at s = 900 GeV, corresponding to the injection energy of the SPS proton beams. Soon after the centre-of-mass energy was raised shortly to s = 2.36 TeV, to stabilise then at s = 7 TeV for the 2010 and 2011 runs, and finally to s = 8 TeV for the 2012 run. More information on other LHC parameters for LHCb operation during 2012 operations is given in Tab. 1. Table 1... The LHC parameters for 2012 operations at LHCb. Beam energy 4 TeV Instantaneous luminosity cm 2 s 1 Bunch spacing 50 ns Particles per bunch Bunches per beam 1380 <µ> 1.64 Four large detectors are placed along the LHC circumference at different collision points. ATLAS and CMS are general purpose detectors, LHCb is a dedicated heavy-flavour experiment and ALICE is designed to study heavy ion collisions. Both ATLAS and CMS are

4 24 The LHCb experiment designed to operate at the full LHC design luminosity, providing a coverage at large angles up to pseudorapidity 2 η <2.4. LHCb instead provides a forward pseudorapidity coverage of 1.9<η<4.9, optimal for the detection of boosted B-hadrons. In order to perform precision physics in a moderate background environment and due to the limited hardware trigger rate at 1 MHz, LHCb has been designed to operate in a low pile-up environment: a luminosity of cm 2 s 1. During the first three years of operations, the LHC machine and the LHCb detector have shown excellent performance, allowing LHCb to take data at a factor two above the design rate L = cm 2 s 1, in both 2011 and The experiment accumulated about 1.1 fb 1 pp collisions at s = 7 TeV in 2011, and about 2.1 fb 1 at s = 8 TeV in More information on the LHCb experiment are given in the next sections. 3.2 The LHCb experiment LHCb [68] is a dedicated experiment for the study of heavy flavour physics at the Large Hadron Collider. In particular, the experiment focuses on the study of CP violation and rare decays of beauty and charm particles, to test the Standard Model and to search for evidence of New Physics. The LHCb physics programme is complementary to the flavour physics studies conducted at the B-factories and to the direct searches for new particles performed at ATLAS and CMS. In fact, while the high-p T experiments search for on-shell production of new particles, LHCb can look for their virtual quantum effect in beauty and charm decay processes precisely predicted in the SM. In particular, the LHC produces b-quarks at an unprecedented rate: the b-production cross-section is measured to be about 300 µb at s = 7 TeV, significantly higher than that of the B-factories (1 nb [136]) and Tevatron ( 30 µb [69]). Taking into account the correlated kinematics of the produced b b pair at high energies and their typically small polar angle with respect to the beam axis, the LHCb detector is designed as a single arm spectrometer, with a forward pseudorapidity acceptance corresponding to an angular coverage from approximately 10 mrad to 300 mrad. LHCb is installed at LHC interaction point 8 and the spectrometer is placed following the direction of LHC beam1 (clock-wise beam), as visible in Fig. 10. This direction coincides with the the positive z- axis in the LHCb coordinate system, adopted as global system for the alignment studies presented in Ch. 5. The positive y-axis points upwards and the positive x-axis points out of the figure. The rich flavour physics program of LHCb requires the detector to trigger and reconstruct heavy flavour decay events with hadronic, (semi-)leptonic or radiative final states. This is an ambitious challenge at a hadron collider. The detector setup therefore includes a precision tracking system (see Sec. 3.3), an extensive Particle Identication (PID) system (see Sec. 3.4) and a highly adaptable trigger (see Sec. 3.5). The tracking system serves to reconstruct the trajectories of all charged particles produced in the collision, or in subsequent decays. Moreover, vertices are found from the crossing point of two or more tracks. The first subsystem of the tracking detector consists of a silicon-strip vertex detector (VELO) surrounding the pp interaction region. The second subsystem is a large-area silicon-strip detector (Tracker Turicensis, TT) located upstream [ ( )] 2 The pseudorapidity, η, is commonly used as a spatial coordinate. It is defined as η = ln tan θ2, where θ is the angle between the particle momentum p and the beam axis.

5 3.2 The LHCb experiment 25 of a warm dipole magnet. The third subsystem consists of three stations of silicon-strip detectors (Inner Tracker, IT) close to the beamline, surrounded by three large straw drift tube detectors (Outer Tracker, OT) placed downstream. The magnet dipole, with a bending power of about 4 Tm, bends charged particles and allows to determine their momenta by means of the track deflection measurement. To identify the nature of the reconstructed particles, the information collected from the particle identification (PID) detectors is used. The LHCb PID system comprises two Ring Imaging Cherenkov systems (RICH) to identify charged hadrons, a hadron (HCAL) and electromagnetic (ECAL) calorimeter for photon, electron and hadron separation, and finally a set of muon chambers (MUON), composed of alternating layers of iron and multiwire proportional chambers, to identify muons. The trigger system performs an initial real-time data selection, needed to discard most of the events that are not interesting for physics analysis. This online reduction is required, since the amount of data generated by LHC collisions is too high to be stored offline. The layout of the LHCb detector is depicted in Fig. 10; more details of the various subdetectors are given in the following sections. Figure A drawing showing the longitudinal view of the LHCb detector, including a reconstructed event superimposed on a picture of the experiment in the cavern; LHC beam1 traverse the experiment from the VELO to the muon chambers.

6 26 The LHCb experiment 3.3 Tracking at LHCb Tracks are reconstructed across the spectrometer by combining hits of the VELO detector (see. Sec ), surrounding the collision region, and the main tracking system (see Sec ), upstream and downstream of the LHCb magnet. Track reconstruction results in two main observables: a precise particle momentum measurement, leading to a good invariant mass resolution, and accurate vertexing, needed to distinguish between primary and decay vertices and to resolve the fast B s oscillations. The better the mass and vertex resolution, the better the background suppression in the search for a rare decay process, as e.g. B 0 s(d) µ + µ decays presented in this dissertation The VELO detector The Vertex Locator surrounds the LHCb interaction point; its role is to precisely determine both the position of the primary collision vertex and the position of secondary vertices from the decays of beauty and charm particles. Such hadrons have a lifetime of the order of s and a highly relativistic boost, hence they have an average flight distance of 1 cm. The VELO plays a central role both in the trigger and in the offline reconstruction. It consists of two retractable halves, each containing 21 modules. Each module has two semicircular silicon sensors mounted back-to-back, each half having a different strip orientation to provide measurements of both the radial co-ordinate (R-sensor) and the azimuthal angle (φ-sensor). The strips on the R-sensor are arranged in four sectors, as shown in Fig. 11, where the strip orientations of an R and a φ-sensor are illustrated. A photograph of some VELO sensors during the assembling of the detector is shown in Fig. 12. In addition to the 42 VELO modules, 4 Pile-Up (PU) veto modules containing only an R- sensor are installed at the upstream end of the VELO modules. These detectors only observe tracks produced in the opposite direction ( upstream ) compared to the rest of LHCb. Figure 9 highlights the PU modules in the longitudinal view of the VELO. A considerable part of the work presented in this dissertation is dedicated to commissioning tests and data analyses for the Pile-Up detector; these will be presented in Chaps Each VELO and PU sensor is formed by a 300 µm thick silicon wafer and contains 2048 microstrips, with a pitch varying between 40 µm and µm. During injection and adjustment of the LHC beams, each half of the VELO is retracted by 2.9 cm to avoid potential damages, while during stable beams the halves are brought together, such that the nearest silicon strip is located 8.2 mm from the beam during data acquisition. The modules of one half are slightly offset in position along the beamline with respect to the other, leading to overlapping regions between the two halves when the VELO is fully closed. This overlap plays an important role in determining the relative alignment of the two detector halves and also provides a full azimuthal coverage. The detectors are placed in a secondary vacuum that is separated from the LHC vacuum by a 300 µm thick aluminum foil. The modules are cooled by a CO 2 system, operating at -30 C. The Beetle chip [70] is the front-end chip used to read out the VELO detectors. It was designed to resist high radiation levels and has a sampling rate of 40 MHz. The VELO data signals are transmitted to analogue receiver cards located on the TELL1 boards [71] of the

7 3.3 Tracking at LHCb Figure Schematic view of two sensors of the VELO module, showing the different strip-orientation between a φ-sensor (on the left) and an R-sensor (on the right). Figure Picture of some VELO sensors taken during the assembling phase of the last module, in 2007 (from Ref. [137]). 27

8 28 The LHCb experiment LHCb data acquisition system. In total, 84 TELL1 boards are used to read out the 21 VELO stations and 4 TELL1 s for the 2 Pile-Up stations. The VELO single hit resolution ranges between 4 and 20 µm, depending on the inter-strip pitch and on the track angle (the best hit resolution is measured for the 40 µm strip pitch). The closing mechanism allows to bring the silicon detectors close to the LHCb beams. The vertex resolution performance achieved in the nominally closed position, standard during data taking, is shown in Fig. 13. For a typical primary vertex (PV) producing 25 tracks, the PV resolution is about 13 µm in x and y, and 71 µm in z. Figure VELO primary vertex resolution in the transverse plane (x direction) as a function of the number of tracks; in red, results obtained with 2011 data and in green results obtained with MC simulation. Figure taken from Ref. [72]. Moreover, a high separation power between primary and displaced vertices is guaranteed by a good track impact parameter (IP) resolution, where IP is the closest distance between the primary interaction vertex and the secondary particle trajectory. This is below 40 µm for tracks with p T > 1 GeV/c [72] Trackers: TT, IT and OT To measure the momenta of charged particles, the LHCb tracking system includes a tracking station located upstream of the magnet (Tracker Turicensis, TT) and three stations (IT, OT) downstream of the magnet. The TT consists of four layers of 500 µm thick silicon detectors covering the whole acceptance of the experiment (total area of about 4 2 = 8 m 2 ). The four measurement planes are arranged into two stations separated by 30 cm along the beam (z) axis, with 9.44 cm long strips oriented at 0, +5, -5 and 0 stereo angles with respect to the vertical (y) axis. The strip pitch is 183 µm and the measured hit resolution is 59 µm [73]. The tracking system behind the magnet, in the downstream area, is divided in two parts: a small silicon micro-strip detector at high rapidity (Inner Tracker, IT), placed in a cross-

9 3.3 Tracking at LHCb 29 shaped area surrounding the beam pipe, and a gaseous strawtube detector system (Outer Tracker, OT), covering most of the LHCb acceptance in the outer region. The picture in Fig. 14 shows a view from within the magnet, including one of the IT stations in front of the OT. The choice of different detection technologies follows from the physics of particle production at the LHCb experiment: although the IT covers only 1.2% of the surface of the downstream tracking stations, this region sees the highest particle flux. In fact, about 30% of the charged particles coming from the interaction point traverse the silicon area, due to the forward-peaked multiplicity distribution. Each IT station consists of four independent boxes arranged around the LHC beam pipe, where each box in turn contains four layers of silicon micro-strips with the same stereo orientation as the TT detector. The IT covers a total active area of about 4.2 m 2 ; its sensors have a strip pitch of 198 µm, with a measured hit resolution of 50 µm [73]. Figure View of the OT with one of the IT stations in front, taken from the inside of the LHCb magnet region. Picture from Ref. [137]. Both TT and IT are operated at 0 C and use the same radiation-hard front-end chip (Beetle) and readout electronics as the VELO, operating at a clock frequency of 40 MHz. Clustering and zero-suppression are performed on a common off-detector readout board (TELL1), located in a zone accessible during LHC operation. The Outer Tracker is a gaseous detector surrounding the IT and covering a total active area of 360 m 2. It consists of straw tubes arranged in three tracking stations, each station having four double layers of detection modules in the same stereo arrangement as the IT and TT. Each module consists of 2 staggered layers of 64 straw tubes. During operation, the straw tubes are flushed with a Ar(70%)/CO 2 (28.5%)/O 2 (1.5%) gas mixture and the anode wire is kept at a high voltage of 1550 V. A precise mass resolution requires good momentum resolution, which requires in turn good spacial resolution. In LHCb, mass resolution studies separating B Kπ from B ππ

10 30 The LHCb experiment Figure Sketch of the LHCb detectors used for tracking and the corresponding reconstructed track types. decays and other background studies have lead to a requirement of a momentum resolution of about δp/p<0.5% and an OT hit resolution of 200 µm [74]. Under nominal data taking, a hit resolution of 220 µm has been achieved, which improves to 180 µm after an additional monolayer alignment study [75]. The combined LHCb tracking system results in a momentum resolution, δp/p, that varies from 0.4% at 5 GeV/c to 0.6% at 100 GeV/c. Thanks to the good momentum resolution and the high separation power between primary and displaced vertices, the typical B-decay time resolution at LHCb is about 50 fs, to be compared with the B-lifetime of about 1500 fs Tracking algorithms Depending on the production vertex and three-momentum, charged particles can leave various signatures in the detector. Different types of tracks are reconstructed in LHCb, depending on the subdetector information used. Figure 15 schematically illustrates the LHCb track types. The track reconstruction strategy is based on a pattern recognition procedure and on a subsequent track fitting procedure. Initially, VELO track seeds are determined in the r z projection plane, under the assumption that they originate from the collision region. Secondly, φ-hit information is added to these track seeds. Afterwards, VELO tracks are combined with hits from the downstream tracking stations using two strategies. The first one is based on forward propagation of the VELO track seeds through the magnetic field, adding hits of the T-tracking stations to the track candidates. The second strategy starts instead from track seeds independently found in the downstream tracking stations (standalone) and propagates them upstream to find a match with VELO seeds. As a track may be found by different algorithms, a clone killing stage is also implemented. Tracks leaving hits in all tracking detectors are named long tracks and are the main track sample used in physics analyses. Tracks based on TT and T-station measurements only are called downstream ; such trajectories are often products of long lived particles (as K s decays). VELO tracks with only additional hits in the TT stations are named up-

11 3.4 Identifying particles at LHCb 31 stream tracks and usually have very low momentum. They play an important role in the reconstruction of so-called slow particles, e.g. the pion in the D Dπ decay. The particle trajectories are fitted with a bi-directional Kalman Filter algorithm [76]. It accounts for multiple scattering and energy loss in the detector material and includes misalignment corrections of the individual measurement layers. The tracking system has an average track reconstruction efficiency above 95% and achieves a momentum resolution δp/p = 0.35%-0.5% [77]. In particular, the momentum resolution is a crucial parameter for invariant mass reconstruction; for instance, LHCb achieves a mass resolution of about 24 MeV/c 2 for 2-body B-decays [78], as B 0 s µ + µ. A good invariant mass resolution allows to suppress backgrounds due to combinatorics or incomplete decay reconstruction. 3.4 Identifying particles at LHCb Particle Identification (PID) is used in the identification of exclusive final states in heavy flavour decays. In particular, hadron identification allows to separate pions, kaons and protons. It is applied in the analyses of purely hadronic decays, as well as in that of (semi)- leptonic ones. Hadronic decays are often an important background source for rare decays with a purely leptonic final state, as the B s(d) 0 µ+ µ decay presented in this dissertation. Additionally, kaon identification is used for B-flavour tagging purposes to distinguish between B and B decays. Similarly, lepton identification is used in the analyses of (semi)- leptonic decays, as well as in flavour tagging algorithms. In LHCb, the RICH detectors provide pion-kaon separation (see Sec ), the calorimeters provide electron, photon and hadron identification (see Sec ) and the MUON system muon identification (see Sec ). In addition, the calorimeters and MUON system measure particle energy and momentum, respectively. They both provide information used in the trigger system (see Sec. 3.5), whereas this is not the case for the RICH detector The RICH detector The Ring Imaging Cherenkov technique measures the velocity of charged particles, by determining the angle of the Cherenkov radiation emitted above a given threshold velocity. The RICH system consists of a detector upstream of the magnet (RICH1) and a detector downstream the magnet, in front of the calorimeter. The first RICH identifies kaons and pions at relatively low momentum (2<p<60 GeV/c); it employs two radiators, silica aerogel to detect particles in the momentum range up to 10 GeV/c and C 4 F 10 gas for 10<p<40 GeV/c. The second RICH uses only CF 4 as gas radiator in a limited acceptance of 15 mrad to 120(100) mrad in the horizontal(vertical) plane and extends the momentum coverage up to about 100 GeV/c. The RICH photodetectors are placed outside the LHCb acceptance. A set of spherical and flat mirrors projects the Cherenkov light onto the detection plane. The light is detected by Hybrid Photon Detectors (HPDs), which are custom-built Si pixel chips, able to operate fast and efficiently over a large active surface area.

12 32 The LHCb experiment The two RICH detectors allow charged kaon identication with an efficiency of 96% in the GeV/c momentum range, for a corresponding fake rate from pions as kaons of 7% [79] Calorimeters The LHCb Calorimetry System (CALO) is located downstream of RICH2. It comprises a Scintillating Pad detector (SPD), a pre-shower detector (PS), an electromagnetic calorimeter (ECAL) and a hadron calorimeter (HCAL). The combined SPD-PS system consists of 2.5 radiation lengths of lead sheet, sandwiched between two scintillator plates. Similarly, ECAL is a lead-scintillator calorimeter covering a depth of 25 radiation lengths; it is made of 66 layers of 2 mm thick lead and 4 mm thick scintillation material. Also the HCAL is a sampling calorimeter alternating iron and scintillator, for total 5.6 interaction lengths. The main function of the PS and SPD systems is to provide longitudinal segmentation of the electromagnetic shower and help distinguishing electrons from the π 0 and π ± background. Figure 16 shows a picture taken during the ECAL installation phase. The ECAL energy resolution is σ/e = 9%/ E 0.8% and the HCAL energy resolution is 69%/ E 9% [80]. The average electron identification efficiency is about 90%, with a misidentication rate of about 3-5% for electrons with momentum above 10 GeV/c The MUON system The MUON detector is used to trigger and identify muons. It consists of five stations, the first station (M1) located between the RICH and the calorimeters and the other four (M2-M5) behind the HCAL, interleaved with 80 cm thick iron walls to absorb hadrons. M1 is not used for particle identification purposes, but to provide a better estimate of the muon p T in the muon Level-1 trigger, since the muon impact point on the first station is not affected by the multiple scattering in the calorimeters [81]. The muon stations are equipped with Multi-Wire Proportional chambers, with the exception of the high particle-flux region in the centre of the first station, which uses triple-gem 3 detectors. The muon identification efficiency ranges between %, depending on the track p and p T. The corresponding misidentication rate, an important parameter for analysing B 0 s(d) µ+ µ decays, is approximately 1-2% Particle Identification algorithm Particle identification (PID) in LHCb is accomplished by constructing a relative Delta Log Likelihood (DLL) function, that calculates the difference between the logarithm of the likelihood assigned to two different particle hypotheses. For instance, to identify muons, the likelihoods L(µ) and L(π) are computed by combining information from all different subdetectors, and the muon DLL function is constructed as DLL(µ) = ln[l(µ)] ln[l(π)] = ln[l(µ)/l(π)]. 3 A GEM is a Gas Electron Multiplier.

13 3.4 Identifying particles at LHCb 33 Figure Picture of ECAL photomultipliers, taken in 2006, during the commissioning of the detector (from Ref. [137]).

14 34 The LHCb experiment Particles are thus selected by cutting on the ratio of likelihoods between different hypotheses. Each likelihood function is a composition of likelihoods based on the information of the muon chambers, calorimeter systems and RICH detectors. The MUON detector likelihood is built depending on how the muon chamber hits are aligned with respect to the particle track reconstructed from the tracking system. The CALO likelihood uses the energy deposits in the various calorimeter regions. Since muons are minimum ionising particles (MIPs), they traverse the calorimeters losing a characteristic energy 4 which is used to implement a CALO PID. The RICH likelihood is obtained by comparing the Cherenkov angle of the radiated photon ring for specific particle hypotheses. The RICH allows to assign relative pion, kaon and proton likelihoods and its efficiency versus misidentication depends on the requirement on the correspondent DLL variable (DLL(K π), DLL(p π), etc) implemented in the analysis. Results on the particle identification performance of muons and hadrons for the B 0 s(d) µ+ µ analysis are given in Sec A trigger for heavy flavour physics Although the total visible pp interaction cross-section at LHCb is large (σ TOT 65 mb), only one out of about 200 interactions contains b-quarks. In addition, the B-decays of interest are typically rare decays. Therefore, a selective and efficient trigger is required Trigger system The trigger consists of a hardware stage (L0), based on information from the calorimeters (CALO), the muon detector (MUON) and the Pile-Up system (PU), and a software stage (HLT), running a full event reconstruction on a farm of parallel-processing CPUs. Moreover, Global Event Cuts (GEC) are applied at each trigger level to decrease the HLT processing time. During 2011 and 2012 data acquisition, the L0 has taken events at an input frequency rate of 15 MHz and reduced the rate of crossings with at least one pp interaction to below 1.1 MHz. The HLT has subsequently reduced the event rate to an output rate of around 3 khz [82]. The main purpose of the L0-trigger is to efficiently select c- and b-hadron events while reducing the rate of visible pp interactions to 1 MHz. The L0-trigger also allows to select beam-gas interactions, as explained in Ch. 7. The L0-trigger detectors synchronously send information to the L0 decision unit (L0DU), where specific selection algorithms (or channels ) are run (mainly based on p T cuts on muons and E T cuts on clusters in the calorimeters); some of them are discussed in Sec The logical OR of all L0 channels is calculated for every bunch-crossing and delivered by the L0DU to the LHCb readout supervisor (ODIN), together with the veto of the GEC and the decision of the B1gas and B2gas channels (see Sec. 7.1). The only GEC implemented at the L0 level is on the number of hits in the SPD detector. ODIN uses the L0DU decisions and other inputs (e.g. timing and bunch-crossing type information) to decide whether the full event information should be read-out and transferred to the HLT. 4 A MIP typically deposits in the calorimeters about 400 MeV of total transverse energy.

15 3.6 LHCb software framework and analysis work flow 35 All events that are selected by the L0-trigger go to the Event Filter Farm. Here the HLT1 trigger rejects high multiplicity events by applying a global event cut on the number of hits in the OT. It subsequently partially reconstructs the event tracks and performs lepton identification; its trigger decision is based on a combination of cuts on the p T, IP and invariant mass. If an event meets the HLT1 requirements, track finding and fitting algorithms similar to offline algorithms are applied. HLT2 reconstructs all tracks in the event with p>5 GeV/c and p T >500 MeV/c and selects candidates using as signatures leptonic decays, finite lifetime and invariant mass [82]. The HLT consist of trigger lines, algorithms designed on the base of specific heavy flavour decay signatures and operating independently. The relevant line for B 0 s(d) µ+ µ decays is discussed in Sec A combination of HLT lines, together with a L0 configuration, form an unique trigger configuration encoded in the Trigger Configuration Key (TCK). This is a hexadecimal word identifying the set of trigger decisions, the algorithms to be run and the event cuts to be applied; examples of TCKs are given in Ch. 7. Finally, trigger information is also used to distinguish between the different bunch crossing types: beam-beam and beam-empty crossings are triggered by the calorimeters, since particles are produced in the forward direction; a PU veto is applied in particular for beam-empty crossings (see Sec. 7.1). Alternatively, crossings of type empty-beam are characterised by particles in the backward direction and thus selected requiring a minimum hit multiplicity in the PU system and a maximum calorimeter activity. Nominally empty-empty bunch slots are triggered via the logical OR of the triggers for beam-empty and empty-beam types, to be able to include their background. 3.6 LHCb software framework and analysis work flow The selection and preparation of the acquired data for its subsequent physics analysis is performed by means of C ++ tools and algorithms, grouped in different projects. The package which serves as a framework for all the projects is called Gaudi. The Moore project gathers all the trigger and hit clustering algorithms, while Brunel collects the event reconstruction code. MC simulations are produced with the Gauss project, using a full GEANT4 simulation [83]; this provides simulated data in an identical Raw data format as real data. MC events are reconstructed with the same software used for real events. All reconstructed events are split in subsamples according to their physics content, on the base of inclusively or exclusively reconstructed signatures. In the LHCb framework, the splitting pre-selection procedure is named stripping and each of the selections is embedded in a stripping line. The purpose of implementing a stripping phase is to avoid the different working groups the analysis of the whole triggered sample. The groups of libraries containing the stripping and analysis tools are respectively called DaVinci and Erasmus. The final event selection, running over the stripped datasets, is executed on a distributed system of worldwide spread computer centres, called Grid. Grid computing both ensures that the data is not lost and allows physicists easy access to a huge computing power, to perform reconstruction or simulation tasks. The LHCb framework also contains projects for stand-alone analyses. For instance, Vetra is used to emulate and commission the VELO readout algorithms, while the Alignment

16 36 The LHCb experiment project groups the tools needed to spatially align each subdetector. The Conditions Database (CondDB) provides time-dependent geometry and conditions data for all the mentioned LHCb applications. 3.7 Conclusions The LHCb detector, placed at one of the four interaction points along the LHC beam line, includes a number of tracking and PID subdetectors, which are used for both triggering and offline event reconstruction. Each of the subdetectors has been operating successfully during LHC run 1 ( ). The following chapters are in particular dedicated to the Pile-Up subsystem, giving a detailed overview of the system and of the work performed to commission it.