Muon reconstruction in ATLAS

Transcription

1 Muon reconstruction in ATLAS Niels van Eldik CERN

2 Muons for physics analysis: Four flavors Combined muons: ID+MS hits + full track fit the bulk of all muons Standalone muons track in the MS, no associated ID track signal muons outside eta > 2.5 decays in flight, combinatorial fakes Segment tagged muons ID track + matching segment regions of poor coverage + low pt muons Calorimeter tagged muons ID track + MIP signature in calo 2

3 Performance of the muon identification: Efficiency tag-and-probe method using Z bosons J/psi Overall efficiency of the muon reconstruction is close to 100% except in the central barrel region Very good data/mc agreement differences between data and MC efficiencies smaller than 0.1% for most of the coverage slightly large the central bins seem to fluctuate again possible minimal efficiency gain there? Efficiency Data/MC CB+ST -1 Ldt =20736 pb 2012 data, chain 3 MC ATLAS Preliminary data η Max Goblirsch (MPI) Muon Performance - reprocessed 2012 data / 22 3

4 Correction on Z Mass, CB, Correction All 2012: on Muons Z Mass, (forca Resolution correction factors ATLAS strategy to correct 33 the momentum 10 Internal measurements: correct Combined the tracks MC chain to 3 look ATLAS like Internal the Data Data ss = 8 TeV data -1 Corrected Uncorrected simulation simulation (resolution only) L = 20.7 fb fb 600 Two corrections applied ATLAS s = 8 TeV L = 20.7 fb -1-1 [GeV events/δm µµ µµ ] [GeV events/δm µµ µµ µµ ] Combined tracks chain 3 Data Data Corrected Uncorrected simulation simulation (resolution only) ATLAS Internal ss s = 8 TeV L = 20.7 fb fb fb [GeV] m µµ Data/MC m µµ [GeV] [GeV] m µµ Data/MC m µµ [GeV] [GeV] ass for chain 3Dimuon Combined invariant muons, mass isolated for chain and with 3Dimuon Combined pt>25invariant GeV. muons, The mass isolated plotfor chain and w3 mass for 2012shows data and thepowheg invariantsimulation mass for 2012 of Z shows! data µµ and + the backgrounds. Powheg invariantsimulation mass Nofor 201 of Z on the left plot, correction only smearing is applied correction on the is left applied plot, correction only the smearing plot is applied at correction the on center the is left app pl cale correctionand aresmearing applied toand thescale plot on correction the right. and are The smearing applied corrections toand thehave scale plot on been correctio the rig l 2012 dataset. derived from the full 2012 dataset. derived from the full 2012 datase m µµ 4

5 Correction on Z Mass, CB, All 2012: Muons (for A Resolution correction factors ATLAS strategy to correct 3 3 the momentum 10 Internal measurements: correct Combined the tracks MC chain chain 3to 3 look like ATLASInternal the Data Data s = s = 8 TeV 8 TeV data -1 Corrected Uncorrected simulation simulation (resolution only) L = L = fb fb 600 Two corrections applied ATLAS s = 8 TeV L = 20.7 fb -1-1 [GeV [GeV events/δm events/δm µµ µµ additional smearing ] Internal -1-1 [GeV [GeV events/δm events/δm µµ µµ ] Combined tracks chain chain 3 3 Data Data Corrected simulation (resolution only) ATLAS Internal s = s = 8 TeV 8 TeV -1-1 L = L = fb fb [GeV] m µµ Data/MC Data/MC m µµ µµ [GeV] [GeV] m µµ µµ Data/MC Data/MC m µµ µµ [GeV] [GeV] ass for chain 3 Dimuon Combined invariant muons, mass isolated for chain and with 3 Combined pt>25 GeV. muons, The isolated plot and w mass for 2012 shows data and thepowheg invariant simulation mass for 2012 of Z! data µµ and + backgrounds. Powheg simulation No of Z on the left plot, correction only smearing is applied correction on theisleft applied plot, only the smearing plot at the correction center is app cale correctionand are smearing applied toand the scale plot oncorrection the right. are Theapplied corrections to thehave plot been on the rig l 2012 dataset. derived from the full 2012 dataset. m µµ µµ 5

6 Resolution correction factors ATLAS strategy to correct 3 the momentum 10 measurements: Internal correct Combined the tracks MC chain to 3 look like ATLAS the 700 Data 2012 s = 8 TeV data -1 Corrected simulation (resolution only) L = 20.7 fb Two corrections applied ATLAS s = 8 TeV L = 20.7 fb additional smearing momentum scale correction Good data/mc agreement after the 200 corrections have been applied Bare in mind: m µµ scale correction 1 applied to MC not data: m mass µµ [GeV] 0.95 peak position does not necessarily correspond [GeV] to PDG value m µµ [GeV] ] -1 [GeV events/δm µµ Data/MC Internal ] -1 [GeV events/δm µµ Data/MC Combined tracks chain 3 Data 2012 Corrected simulation ATLAS Internal s = 8 TeV -1 L = 20.7 fb m µµ [GeV] [GeV] ass for Also chain measured 3 Combined but not covered muons, here: isolated scale and with pt>25 GeV. The plot t mass factors for 2012 for the data errors and from Powheg the track simulation fit of Z! µµ + backgrounds. No d on the left plot, only smearing correction is applied on the plot at the center scale correction are applied to the plot on the right. The corrections have been ll 2012 dataset. m µµ 6

7 This talk Focusses on how we get from the raw hits to the muons that enter physics analysis my area of expertise most of you probably not aware of the concepts many features observed with muons can be tracked back to algorithmic choices or the geometry Not covered here: how to use the muons in analysis very well documented on the MCP wiki pages weekly MCP meeting offer the opportunity to go raise specific issues you encounter 7

8 Event Reconstruction in a Nutshell

9 Event Reconstruction in a Nutshell typical HEP detector tracker to measure charged particles e.m. and hadronic calorimeter to measure energy of particles (jets) muon spectrometer

10 Event Reconstruction in a Nutshell photons shower in e.m. calorimeter (ideally) no charged particle seen in tracker neutrons showers in hadronic calorimeter no particle seen in

11 Event Reconstruction in a Nutshell electrons shower in e.m. calorimeter a charged particle seen in tracker protons/pions particle seen in tracker and leave a showers in hadronic calorimeter

12 Event Reconstruction in a Nutshell muons charged particle seen in tracker little energy seen in calorimeters particle seen in muon spectrometer

13 Event Reconstruction in a Nutshell Jet neutrinos leave undetected missing transverse energy jets bundle of showers in calorimeter bundle of charged particles in tracker vertex

14 In Reality?... a bit more complicated ZZ* 4μ candidate Markus Elsing 9

15 The ATLAS muon system 10

16 The precision chambers Monitored Drift Tube (MDT) chambers used as main precision tracking detector 3 cm wide tubes with a wire at the center, average resolution of 80 µm two separate multi-layers to obtain a better angular resolution 700 ns maximum drift time in-plane optical alignment system to monitor chamber deformations Cathode Strip Chambers (CSC) deployed in the forward region exploit cluster shape to obtain average spacial resolution of 60 µm less sensitive to background due to much smaller time window z y a. u. (a) (c) x Cathode tube R min Anode wire mm µ Voltage (V) Drift time (ns) (b) Multilayer In plane alignment Longitudinal beam Cross plate Time (ns) (d) Threshold Time 200 t r Drift time (ns) Radius (mm) 11

17 Trigger chambers TGC wheel Two technologies: RPC/TGC Fast detectors provide the L1 trigger time resolution < 25 ns, can be used to identify the corresponding BC id of a hit Measure second coordinate for the MDT chambers RPC detector design 12

18 The magnet system Very complicated magnetic field large gradients around the coils significant contributions from magnetic masses in the muon system and the calorimeter Field map calculated using measurements from magnetic field sensors using a finite element method required precision: 1x10-4 Tesla Implications for tracking cannot use analytical track models crucial to get the second coordinate right as field varies by a factor two in a chamber interference of barrel and endcap magnetic fields result in two areas without any total field integral and poor momentum resolution 13

19 Track Finding Transition Radiation Tracker New Tracking the task of the track finding identify track candidates in event cope with the combinatorial explosion of possible hit combinations different techniques rough distinction: local/sequential and global/parallel methods local method: generate seeds and complete them to track candidates global method: simultaneous clustering of detector hits into track candidates some local methods track road track following progressive track finding Silicon Track Candidate Silicon Detectors Nominal Interaction Point Space Point some global methods conformal mapping - Hough and Legendre transform adaptive methods Seed Silicon Track - Hopfield network, Elastic net (will not discuss the latter) TRT Extension 14

20 Pattern recognition in the muon system: strategy considerations Studying the detector the global trajectory of muons cannot accurately be described by an analytical model so a high precision numerical approximation is needed (better than 20 µm over the full track length) in the non-bending (xy) plane of the spectrometer the trajectory of the muons is close to a straight line within the layers of the spectrometer, muon trajectories are in first order straight lines the position along the MDT tubes is an important quantity and has to be measured with the trigger chambers the time-windows of the MDT chambers is significantly larger than those of the other technologies making the chambers much more sensitive to background background levels are expected to rise up to 10k hits per event in the MDTs in 2015 Software considerations offline: highest possible efficiency and precision in the HLT trigger: fast execution, sharp turn-on curve Performance requirements efficiency close to geometrical acceptance, maximum resolution 15

21 Pattern recognition Finding the trajectory in the xy-plane y R 0 Use 2D Hough transform using (R0,φ) The Hough transform transforms points in the x,y space into lines in R0,φ straight lines in the xy plane are points in the Hough space the lines of all hits from a given line cross in one point in the Hough space when combined with a histogramming technique the problem reduces to finding the bins with the highest value in the histogram Advantages of the method very good background rejection properties complexity almost linear with number of hits 10 y φ (rad) Figure 3.4: Set of points (left) and their representation in the Hough-space (right) x φ (rad) R R Figure 3.5: Representation points of figure 3.4 in binned Hough space. R x 5 φ (rad) φ (rad) 16

22 Exploiting the precision in the bending plane: CSC segment finding CSC detector: passing particle induces charge over multiple strips exploiting cluster shape to get 60 µm per point cluster splitting to deal with close by particles if all fails use cluster centroid (σ ~2 mm) Clusters are used as input to combinatorial segment finding Handles to remove fake segments fit chi2/ndof number of hits on the segment 17

23 Exploiting the precision in the bending plane: CSC segment finding CSC detector: passing particle induces charge over multiple strips exploiting cluster shape to get 60 µm per point cluster splitting to deal with close by particles if all fails use cluster centroid (σ ~2 mm) Clusters are used as input to combinatorial segment finding Handles to remove fake segments fit chi2/ndof number of hits on the segment 17

24 Exploiting the precision in the bending plane: CSC segment finding CSC detector: passing particle induces charge over multiple strips exploiting cluster shape to get 60 µm per point cluster splitting to deal with close by particles if all fails use cluster centroid (σ ~2 mm) Clusters are used as input to combinatorial segment finding Handles to remove fake segments fit chi2/ndof number of hits on the segment 17

25 Exploiting the precision in the bending plane: CSC segment finding CSC detector: passing particle induces charge over multiple strips exploiting cluster shape to get 60 µm per point cluster splitting to deal with close by particles if all fails use cluster centroid (σ ~2 mm) Clusters are used as input to combinatorial segment finding Handles to remove fake segments fit chi2/ndof number of hits on the segment 17

26 Exploiting the precision in the bending plane: MDT segment finding Local segment finding in the individual MDT stations offers a powerful way of reducing combinatorics the bending of muons above p = 3 GeV is sufficiently small: their trajectory can be approximated by a straight line MDTs provide a very high precision measurement of the trajectory of the muon (80µm average resolution) -> good background rejection trigger confirmation can be used to reduce out-of-time background RPC BML RPC 18

27 Local segment finding in the MDTs basic principles Seed selection calculate tangent lines associate other hits to the seed lines using a fixed road width Segment creation 2D fit to associated hits outlier removal + hit recovery + hole search Segment selection and ambiguity solving exploit 99.5% hit efficiency of the MDTs to reject combinatorial fakes holes + out-of-time hits remove remaining overlaps between segments when possible keep ambiguities of same quality 19

28 Local segment finding in the MDTs basic principles Seed selection calculate tangent lines associate other hits to the seed lines using a fixed road width Segment creation 2D fit to associated hits outlier removal + hit recovery + hole search Segment selection and ambiguity solving exploit 99.5% hit efficiency of the MDTs to reject combinatorial fakes holes + out-of-time hits remove remaining overlaps between segments when possible keep ambiguities of same quality 19

29 Local segment finding in the MDTs basic principles Seed selection calculate tangent lines associate other hits to the seed lines using a fixed road width Segment creation 2D fit to associated hits outlier removal + hit recovery + hole search Segment selection and ambiguity solving exploit 99.5% hit efficiency of the MDTs to reject combinatorial fakes holes + out-of-time hits remove remaining overlaps between segments when possible keep ambiguities of same quality 19

30 Local segment finding in the MDTs basic principles Seed selection calculate tangent lines associate other hits to the seed lines using a fixed road width Segment creation 2D fit to associated hits outlier removal + hit recovery + hole search Segment selection and ambiguity solving exploit 99.5% hit efficiency of the MDTs to reject combinatorial fakes holes + out-of-time hits remove remaining overlaps between segments when possible keep ambiguities of same quality Wrong segment: one empty tubes one out-of-time hit Good segment, no empty tubes 19

31 Local segment finding in the MDTs the reality with background Constraints for the algorithm overall segment finding efficiency higher than 95+% high background rejection to allow track finding to work within CPU budget maximum average processing time 200 ms offline in the trigger: no tails in the processing time beyond 180 s Challenges of the MDT segment finding under background conditions average signal to background ration at hit level of 1:500 (2015) correctly resolving the left/right ambiguities dealing with hit masking due to background dealing with very non-homogeneous coverage 20

32 Local segment finding in the MDTs the reality with background Constraints for the algorithm overall segment finding efficiency higher than 95+% high background rejection to allow track finding to work within CPU budget maximum average processing time 200 ms offline in the trigger: no tails in the processing time beyond 180 s Challenges of the MDT segment finding under background conditions average signal to background ration at hit level of 1:500 (2015) correctly resolving the left/right ambiguities dealing with hit masking due to background dealing with very non-homogeneous coverage 20

33 Local segment finding in the MDTs the reality with background Constraints for the algorithm overall segment finding efficiency higher than 95+% high background rejection to allow track finding to work within CPU budget maximum average processing time 200 ms offline in the trigger: no tails in the processing time beyond 180 s Challenges of the MDT segment finding under background conditions average signal to background ration at hit level of 1:500 (2015) correctly resolving the left/right ambiguities dealing with hit masking due to background dealing with very non-homogeneous coverage 20

34 Local segment finding in the MDTs the reality with background Constraints for the algorithm overall segment finding efficiency higher than 95+% high background rejection to allow track finding to work within CPU budget maximum average processing time 200 ms offline in the trigger: no tails in the processing time beyond 180 s Challenges of the MDT segment finding under background conditions average signal to background ration at hit level of 1:500 (2015) correctly resolving the left/right ambiguities dealing with hit masking due to background dealing with very non-homogeneous coverage a.u. a.u. MoMu Moore Muonboy hits on segment MoMu Moore Muonboy hits on segment 20

35 High pt jet punch through 21

36 Cosmic air showers

37 Pile-up + cavern background

38 Pile-up + cavern background

39 Improving seeding of local segment finding using a Hough transform Project hits on a reference plane in the center of each detector layer one projection surface for each station layer in each sector use both position and angle wrt the direction pointing to the IP avoid fake maxima due to multiple close hits in the same layer using the Layer mode each layer can only contribute one entry per bin in the transform -> if there are six layers (MDT) the maximum value of the transform is limited to 6 selection cuts on the Hough directly correspond to the number of detector layers Advantages of the method locally in each station layer the trajectory straight layer mode ensures clean definition of the maxima in the Hough space making tuning of the cuts easier maxima represent pseudo segments per layer which makes it possible to used them as input for a pseudo track finding easy to apply different weights per technology 25

40 Improving seeding of local segment finding example event L =

41 Improving seeding of local segment finding example event L =

42 Improving seeding of local segment finding example event L =

43 Improving seeding of local segment finding example event L =

44 Output of the Hough transform Separation levels at very good signal to background ratio Separation levels at signal barely visible in the plot consistent with the expected large number of charged particles in the MS bkg signal L = associated maximum value in Hough bkg signal L = associated maximum value in Hough 30

45 Output of the Hough transform Separation levels at very good signal to background ratio Separation levels at signal barely visible in the plot consistent with the expected large number of charged particles in the MS Can still exploit the pointing-ness of the particles to improve the rejection L = bkg signal angle wrt ip (rad) L = bkg signal associated maximum value in Hough 3 bkg signal 1600 bkg 1400 signal L = angle wrt ip (rad) L = associated maximum value in Hough 31

46 Output of the Hough transform: exploiting pointing angle Select only hits in the central, pointing, bin Very good separation at low luminosity, 1:1 signal to background ratios at Background rejection factor of more than a factor 50 with acceptable efficiency losses bkg signal L = associated maximum value in Hough 3 10 bkg 250 signal efficiency bkg signal L = efficiency 0.9 bkg signal L = L = associated maximum value in Hough associated maximum value in Hough associated maximum value in Hough 32

47 Summarizing the local pattern recognition Hough transforms are deployed to separate background hits from signal muons in a fast and efficient way use of trigger hits to only select hits in the current bunch crossing good separation even at high occupancies expected at Local segment finding further reduces the background level by exploiting the high precision and efficiency of the MDT/CSC chambers 33

48 Pattern recognition in the muon system: segment seeded track finding Four stage track reconstruction select high quality seed segment Seed in outer station 34

49 Pattern recognition in the muon system: segment seeded track finding Four stage track reconstruction select high quality seed segment search for second segment in a second layer of the detector Seed in outer station Add second station 35

50 Pattern recognition in the muon system: segment seeded track finding Four stage track reconstruction select high quality seed segment search for second segment in a second layer of the detector continue until all layers are included Seed in outer station Add second station Add third station 36

51 Pattern recognition in the muon system: segment seeded track finding Four stage track reconstruction select high quality seed segment search for second segment in a second layer of the detector continue until all layers are included use hole-search to reject bad combinations Seed in outer station Add second station Add third station 37

52 Pattern recognition in the muon system: segment seeded track finding Four stage track reconstruction select high quality seed segment search for second segment in a second layer of the detector continue until all layers are included use hole-search to reject bad combinations Back-extrapolation to the vertex (correcting for energy loss in calorimeters) Seed in outer station Add second station Add third station 38

53 Recap Four stage muon reconstruction road finding using a Hough transform MDT and CSC based segment finding segment seeded track finding back-extrapolation to the vertex Seed in outer station Add second station The difference stages are needed to keep combinatorics under control pile-up and cavern background punch through jets cosmic showers Add third station Software used offline and in HLT 39

54 Measuring the muon momentum Inverse momentum proportional to the sagitta 8 S 1/p = B L 2 Design requirement for the muon system: resolution 10% at 1 TeV sagitta of a 1 TeV muon is about 1 mm, sagitta resolution should be better than 100 µm L 40

55 Measuring the muon momentum Inverse momentum proportional to the sagitta 8 S 1/p = B L 2 Design requirement for the muon system: resolution 10% at 1 TeV sagitta of a 1 TeV muon is about 1 mm, sagitta resolution should be better than 100 µm Contributions to the sagitta resolution assembly precision of the MDT and CSC chambers hit resolution of the MDT and CSC detectors alignment precision Bare in mind: only true if track crosses three stations, significant fraction of muon tracks only have two layers and much worse momentum resolution the resolution is multiple scattering dominated below 100 GeV 41

56 Combined reconstruction: combined muons Outside-in, inside-out Outside -> in: looks for ID tracks that match with the tracks found in the muon system advantage: low combinatorics as the number of muon tracks is small standalone reconstruction independent of ID reconstruction disadvantage: inefficiencies in regions of poor coverage in the MS Inside -> out: extrapolate ID tracks into the muon system and use them to seed the segment and track finding advantage: works well in regions with poor coverage in the MS disadvantage: CPU intensive at large pile-up due large number of ID tracks small bias due to ID seeding of the track fit Both approaches used to maximize efficiency 42

57 Combined reconstruction: combined muons Outside-in, inside-out Outside -> in: looks for ID tracks that match with the tracks found in the muon system advantage: low combinatorics as the number of muon tracks is small standalone reconstruction independent of ID reconstruction disadvantage: inefficiencies in regions of poor coverage in the MS Inside -> out: extrapolate ID tracks into the muon system and use them to seed the segment and track finding advantage: works well in regions with poor coverage in the MS disadvantage: CPU intensive at large pile-up due large number of ID tracks small bias due to ID seeding of the track fit Both approaches used to maximize efficiency 42

58 Combined reconstruction: combined muons Outside-in, inside-out Outside -> in: looks for ID tracks that match with the tracks found in the muon system advantage: low combinatorics as the number of muon tracks is small standalone reconstruction independent of ID reconstruction disadvantage: inefficiencies in regions of poor coverage in the MS Inside -> out: extrapolate ID tracks into the muon system and use them to seed the segment and track finding advantage: works well in regions with poor coverage in the MS disadvantage: CPU intensive at large pile-up due large number of ID tracks small bias due to ID seeding of the track fit Both approaches used to maximize efficiency 42

59 Combined reconstruction: combined muons resolution Combined fit offers the best momentum resolution below 50 GeV the fit profits from the superior ID momentum measurement and the smaller impact of the energy loss in the calorimeter in the forward region eta > 2 without TRT coverage the improvement is visible even at lower momenta above 50 GeV the muon system significantly improves the momentum measurement 43

60 Combined reconstruction: combined muons decay in flight rejection Pion and kaon decays to muons form a partially irreducible background to direct muons an ID track (with hits from pion and or muon) a track from a muon in the muon system 44

61 Combined reconstruction: combined muons decay in flight rejection Pion and kaon decays to muons form a partially irreducible background to direct muons an ID track (with hits from pion and or muon) a track from a muon in the muon system Handles to identify and reject decays poorly reconstructed ID track kink in the track at decay point curvature before and after the decay differ applying the MCP ID track selection significantly reduces the background large momentum imbalance between ID and MS momentum balance significance offers good discrimination starting from pt = 10 GeV calorimeter and track isolation 10 < pt < 15 GeV Direct muons Decays in flight 45

62 Combined reconstruction: combined muons decay in flight rejection Pion and kaon decays to muons form a partially irreducible background to direct muons an ID track (with hits from pion and or muon) a track from a muon in the muon system Handles to identify and reject decays poorly reconstructed ID track kink in the track at decay point curvature before and after the decay differ applying the MCP ID track selection significantly reduces the background large momentum imbalance between ID and MS momentum balance significance offers good discrimination starting from pt = 10 GeV calorimeter and track isolation Decay reconstruction probability mom bal sign < 4 Integrated Pt (GeV) 0.2% of all pions above 20 GeV are reconstructed as muons probability independent of pile-up 46

63 Combined reconstruction: Segment tagging Extrapolate ID tracks to muon system match extrapolated track with segments in the different layers of the muon system use position and angular residual and pull Advantages of the method very robust as very simple algorithm Drawbacks very few handles to reject decays in flight In practice only hand full of the muons in the third chain are segment tagged, most of which are in the eta =0 hole restricting the phase space allows the fake rate to be controlled 47

64 Combined reconstruction: Segment tagging Extrapolate ID tracks to muon system match extrapolated track with segments in the different layers of the muon system use position and angular residual and pull Advantages of the method very robust as very simple algorithm Drawbacks very few handles to reject decays in flight In practice only hand full of the muons in the third chain are segment tagged, most of which are in the eta =0 hole restricting the phase space allows the fake rate to be controlled 47

65 Combined reconstruction: Calorimeter muons Tight track and calorimeter isolation to ensure that calorimeter measurements only from muon Calculate path length in individual cells crossed by the extrapolated ID track and collect measured energies Calculate probability that the measured signal in the calorimeter is from a minimal ionizing particle and cut on the probability Properties of calorimeter muons lower purity and efficiency than the muons reconstructed in the muon system used to clean-up muon selection for tag-and-probe efficiency calculation used by some analyses which require very high efficiency (H -> 4mu) to fill the hole in the coverage around eta =0 Markus E 48

66 Combined reconstruction: Muon collection building Current strategy in the ATLAS muon reconstruction is to run multiple independent algorithms The output of these algorithms is merged with removal of overlaps as a final step in the reconstruction Ranking in overlap removal Outside-in combined fit Inside-out combined fit Statistical combination MuTag Standalone muon Calorimeter muon (starting from 2015) 49

67 To summarize... The ATLAS muon reconstruction provides highly efficient and performant identification of muons both in the HLT and offline The reconstruction is divided in multiple steps to keep the reconstruction times in check without compromising on the performance The combined reconstruction is the final step in the chain and the one that provides the final muons for analysis more than 98% of the muons in the ID acceptance are combined muons (hits in ID and MS + global refit) combined muons offer most handles to reject decays in flight Several different combined algorithms are deployed to ensure robust muon reconstruction Data/MC a merging stage at the end removes overlap to avoid double counting of muons Efficiency Data/MC Transition Radiation Tracker Silicon Track Candidate Silicon Detectors Nominal Interaction Point Space Point Seed Silicon Track reprocessed TRT Extension -1 Ldt =20736 pb 2012 data, chain 3 MC ATLAS Preliminary data η the central bins seem to fluctuate again possible minimal efficiency gain there? 50

68 51

69 52

70 Selection of muons for your analysis ID track selection guarantees well measured momentum removes poorly reconstructed inner detector tracks (decays in flight, combinatorial fakes) Medium muon selection (third chain) combined muons: mom. bal. sign < 4 number of hole layers < 2 chi2/ndof < 5 for single station tracks standalone muons eta > 2.5 number of precision layers > 2 High pt selection (Z /W analysis) the track has hits in at least three station layers at least one phi measurement exclude all regions with known poor alignment 53

71 Drift Tubes in ATLAS: Inner Detector and Muon Spectrometer classical detection technique for charged particles based on gas ionization and drift time measurement cathode (HV ) anode wire (HV+) nobel gas 54

72 Drift Tubes in ATLAS: Inner Detector and Muon Spectrometer classical detection technique for charged particles based on gas ionization and drift time measurement drift tubes used in muon systems and ATLAS TRT anode wire (HV+) nobel gas cathode (HV ) primary electrons drift towards thin anode wire charge amplification during drift (~10 4 ) in high E-field in vicinity of wire: E(r) ~ U 0 / r signal rises with number of primary e s (de/dx) [signal dominated by ions] macroscopic drift time: v D / c ~10 4 ~30 ns / mm determine v D from difference between signal peaking time and expected particle passage spatial resolution of O(100 µm) TRT: Kapton tubes, = 4 mm MDT: Aluminium tubes, = 30 mm 54

73 Drift Tubes in ATLAS: Inner Detector and Muon Spectrometer classical detection technique for charged particles based on gas ionization and drift time measurement drift tubes used in muon systems and ATLAS TRT ions drift to cathode ionised electrons drifting to wire anode wire (HV+) nobel gas charged particle cathode (HV ) primary electrons drift towards thin anode wire charge amplification during drift (~10 4 ) in high E-field in vicinity of wire: E(r) ~ U 0 / r signal rises with number of primary e s (de/dx) [signal dominated by ions] macroscopic drift time: v D / c ~10 4 ~30 ns / mm determine v D from difference between signal peaking time and expected particle passage spatial resolution of O(100 µm) TRT: Kapton tubes, = 4 mm MDT: Aluminium tubes, = 30 mm 54

74 Drift Tubes in ATLAS: Inner Detector and Muon Spectrometer classical detection technique for charged particles based on gas ionization and drift time measurement ions drift to cathode ionised electrons drifting to wire drift anode circle wire (HV+) nobel gas charged particle cathode (HV ) example: drift tubes segment used in in muon muon drift systems tubes reconstruction and ATLAS TRT from measured drift circles (left-right ambiguity) primary electrons drift towards thin anode wire charge amplification during drift (~10 4 ) in high E-field in vicinity of wire: E(r) ~ U 0 / r signal rises with number of primary e s (de/dx) [signal dominated by ions] macroscopic drift time: v D / c ~10 4 ~30 ns / mm determine v D from difference between signal peaking time and expected particle passage spatial resolution of O(100 µm) TRT: Kapton tubes, = 4 mm MDT: Aluminium tubes, = 30 mm 54

75 The optical alignment system Both the barrel and the endcap are equipped with a system of optical lines connecting chambers within the towers of the muon system The system provides a very accurate measurement of the relative position of the precision chambers in the bending plan The aim is to control the sagitta resolution with a precision better than 30 µm to achieve a 10% resolution at 1 TeV 55

76 Low pt energy scale Small shifts of the combined J/Psi mass wrt the PDG value visible in the endcaps maximum shift 30 MeV (1%) at eta > 2 known issue with the energy loss parametrization at low pt will be fixed in the next reprocessing only affects the low-pt regime as energy loss suppressed by 1/p has no impact on the Z mass data/mc agreement very good shift < 5 MeV, RMS < 3 MeV CB 56

77 Data/MC scale differences for Standalone muons Small Data/MC mass differences observed at the Z scale effects in the order of 100 MeV biggest effects in the transition region between the barrel and endcap toroidal field most likely caused by detector effects like a wrong field map, inadequate description of the material or misalignment very hard to interpret and to fix Muons SA 57

78 Inner tracker reconstruction efficiency Performance of the muon identification: Efficiency Use tag-and-probe method using Z bosons Comb on efficie Muon Spectrometer reconstruction efficiency The muon reconstruction efficiency is measured wit respect to Inn Detector track Tag selection: combined muon with pt>20gev and η <2.4 Probe selection: Nicola Orlando ID track with pt>20gev and η <2.5 on be INFN Sezione selection: di Lecce and Di - Tag-and-probe required to come from a common primary - opposite charge tracks, invariant mass falling in a window of 10GeV around the Z boson mass - both tag and probe are required to be isolated according to a track based algorithm and pass the ID track selection 58

79 03 (40 mrad) stereo strips to measure both coordinates, with one set of strips in each layer parallel to the beam direction, measuring R. They consist of two 6.4 cm long daisy-chained sensors with a strip pitch of 80 µm. In the end-cap region, the detectors have a set of strips running radially and a set of stereo strips at an angle of 40 mrad. The mean pitch of the strips is also approximately 80 µm. The intrinsic accuracies per module in the barrel are 17 µm (R ) and 580 µm (z) and in the disks are 17 µm (R ) and 580 µm (R). The total number of readout channels in the SCT is approximately 6.3 million. A large number of hits (typically 36 per track) is provided by the 4 mm diameter straw tubes of the TRT, which enables track-following up to = 2.0. The TRT only provides R information, for which it has an intrinsic accuracy of 130 µm per straw. In the barrel region, the straws are parallel to the beam axis and are 144 cm long, with their wires divided into two halves, approximately at = 0. In the end-cap region, the 37 cm long straws are arranged radially in wheels. The total number of TRT readout channels is approximately 351,000. J/psi Tag selection: high quality combined muon with pt>4gev and η <2.5, with low impact parameter with respect to the interaction point; η- based geometrical matching of the tag with one of the muon triggering the event; Probe selection: high quality track with p>3gev and η <2.5 (named Inner Detector probes or, briefly, ID probes). Tag-and-probe selection criteria: good tracks vertex fit; ηcuts to avoid near-by tag and probe tracks to avoid residual trigger ID probes Efficiency = 35.5 pb Ldt s = 7 TeV -1 3 < PT P > 3 GeV 4 GeV 0.1 < Unmatched ID probes ATLAS Preliminary CB Gauss+quadratic fit 1000 CB+ST Gauss+quadratic fit Matched ID probes m [GeV] 2500 Efficiency Tag and probe selection with J/ψ Z bosons Counts/0.05 GeV Use tag-and-probe method using Counts/0.05 GeV Performance of the muon identification: Muon reconstruction Efficiency Determ Efficiency efficiency at low pt Calo tagged probes 2000 s = 7 TeV -1 4 GeV 0.1 < Matched CT probes ATLAS Preliminary P > 3 GeV 3 < PT 1500 Ldt = 35.5 pb CB Gauss+quadratic fit CB+ST Gauss+quadratic fit m [GeV] Unmatched CT probes C Simultaneous fit of the invari

80 Performance of the muon identification: Efficiency Use tag-and-probe method using Z bosons J/psi Overall efficiency of the muon reconstruction is close to 100% except in the central barrel region Very good data/mc agreement differences between data and MC efficiencies smaller than 0.1% for most of the coverage small discrepancy the centralin bins the seem most central to fluctuate bins under again investigation possible minimal efficiency gain there? Efficiency Data/MC CB+ST -1 Ldt =20736 pb 2012 data, chain 3 MC ATLAS Preliminary data η Max Goblirsch (MPI) Muon Performance - reprocessed 2012 data / 22 60