Docstoc

Perez

Document Sample
Perez Powered By Docstoc
					Triggering the LHCb
    experiment
    Introduction
    LHCb trigger system
    High Level Trigger with muons:
          Specific selections
          Generic selections

                      Hugo Ruiz




 4a Trobada de Nadal de Física Teòrica a la
         Universitat de Barcelona
                 B physics at LHC
 B production at LHC:
    Large bb cross/section (~ 500 mb)
    All B hadrons produced
       Bu 40%, Bd 40%, Bs 10%, Bc, baryons 10%
    bb pairs bunched either forward or backward

                                            Forward geometry preferred for
                                            a dedicated B experiment

 Examples of interesting B-decays (BR < 10-4):
    CP asymmetries:
       BdJ/y Ks, BdJ/y f with J/y m+m- or J/y e+e-
       Bdp+p-, BdK+K-, BsDs+K-
    Bs oscillations: BsDs+p-                    Needed:
                                                  • good p-K separation
    CP violation in radiative decays: BdK*g
                                                  • hadron trigger
    Rare decays: Bsm m-+

                                Hugo Ruiz                           2
             B physics at LHC
 What do we need to measure to do B physics?
   Typical case: measure time-dependent decay
    asymmetry for Bdp+p-
                                                  not always there!
                        n    l-
                                        K-
                                                  tagging B
                                                     indicates original
                                                    flavor of signal B
                   B0
                                             p-   signal B
                                        p+


            Very good proper time resolution needed!

                            Hugo Ruiz                            3
                      Detector overview
                             RICH’s:                   Muon System
   VELO:
   primary vertex            PID: K,p separation
   impact parameter
   displaced vertex


PileUp
System
Interaction
      point




                                                         Calorimeters:
          Trigger Tracker:    Tracking Stations:         PID: e,g, p0
          p for trigger       p of charged particles
                     Event rates
 LHC rates:
    40 MHz crossing rate
    30 MHz with bunches from both directions at LHCb IP

 Luminosity: 2·1032 cm-2 s-1
    Chosen to maximize number of events with a single interaction
        Easier to identify secondary vertices from B mesons
    10 to 50 times lower than @ ATLAS, CMS
        Well under machine design!


 Visible rates:
  (visible event  at least 2 tracks in acceptance)
    Total rate (minimum bias): 10 MHz (60 mb out of 80 mb)
    bb: 100 KHz
        Whole decay of one B in acceptance: 15 KHz
    cc: 600 KHz
                          Need a trigger organized in levels, starting
                          by simple algorithms on custom electronics
                     Trigger overview
                                                          10 MHz
                                   On custom boards:
Pileup
                                       L0: hight pT + not too busy
system
                                        Synchronous 40 MHz, latency: 4ms


                                   PC farm                 1 MHz
VELO +
                                   ~1800 CPUs
Trigger
tracker                                L1: IP + high pT
                    Calorimeters        <latency>: 1 ms (max 50 ms)
                    + Muon              Buffer: ~ 59k events
                    system
                                                           40 KHz
                                       HLT
                                        Full detector: ~ 40 kb / evt

          LHCb trigger TDR                                200 Hz
          September ‘03                                      recent
                                                             change:
                                                             ~ 2 KHz
                               Level 0
 Fast search for ‘high’ pT particles (calorimeters, muon syst)
    Charged hadrons: HCAL (~ 3 GeV)
    Electrons, photons, p0: ECAL (~ 3 GeV)                ATLAS, CMS:
    Muons: muon system (~ 1 GeV)                          • ~ 6 GeV cut for e, m
                                                           • No hadron trigger

 Cut on global variables:
    Minimum total ET in HCAL (calorimeters)
        Reduces background from halo-muons


    Rejection of multi-PV and busy events (Pileup system, SPD) :
        Fake B signatures (lots of tracks with high IP)
        Busy events spend trigger resources without being more signal-like
             Better throw them early and use bandwidth to relax other cuts
                 Level 0: Calorimeter
                                                 Scintillator Pad Detector (SPD)

 The LHCb calorimeter:
    ECAL: 6000 cells, 8x8 to 24x24 cm2
    HCAL: 1500 cells, 26x26, 52x52 cm2                           ECAL             HCAL

 Trigger strategy: look for high ET
  candidates:                                                    Pre-Shower Detector

    In regions of 2x2 cells
    Particle identification from           SPD-PreShower            ECAL                  HCAL
        ECAL / HCAL energy
        PS and SPD information        FE


                                                                               Validation cards
 Sent to L0 decision unit:
    Highest ET candidate each type                                          Selection crates

    Global variables:
        Total calorimeter energy     SPD mult      e±      g         p0    hadr          ETtot
        SPD multiplicity
    Level 0: Muon system and Pi
 Muon system:
    Strategy:
       Straight line search in M2-M5
       Look for compatible hits in M1
    Sent to L0 decision unit: 2 highest pT
     candidates per quadrant

 Pileup system:
   Concept: 2 silicon planes backwards from
    interaction point, measure R coordinate
                                                                   PV2
                                                             PV1
   Trigger strategy: veto multi-PV evts
       From hits on two planes  produce a histogram of z
        on beam axis
   Sent to L0 decision unit: heights of two highest
    peaks
                   Level 0: Decision
 Thresholds and partial rates:
  (Trigger TDR, Sept 2003)

                          Thresh    Rate        Global
            Type                                                Cut
                          (GeV)    (kHz)       Variable
           Hadron          3.6     705        Tracks in 2nd
                                                                     3
           Electron        2.8     103           vertex

           Photon          2.6     126         Pile-Up              112
                                              multiplicity          hits
           p0 local        4.5     110
                                                 SPD                280
           p0 global       4.0     145        multiplicity          hits
            Muon           1.1     110          Total ET          5 GeV
         Di-muon   SpTm    1.3     145




   Composition after L0:                     bb %         cc %
                                    Visible    1.1            5.6
                                   After L0    3.0           10.6
                  L1-HLT infrastructure
                                                                                                   Front-end Electronics
 L1 & HLT share infrastructure:                  FE   FE    FE   FE      FE   FE   FE     FE    FE      FE   FE     FE    TRM
                                      126 links
     Ethernet network                 44 kHz
                                      5.5 GB/s
     ~ 1800 computing nodes                      Switch Switch Switch                    Switch           Switch
                                                                                                      Multiplexing Layer
                                      64 Links
                                                                                                                L1-
                                                                                                                Decision
 Provides flexibility, scalability                          Readout Network                                          Sorter


                                                                                                               94 Links

 L1 task has top priority
                                                                                                               7.1 GB/s
                                                        SFC       SFC                SFC        SFC           94 SFCs

     HLT & reconstruction run in
                                                                          …
                                                       Switch Switch                Switch Switch                  ~1800
      background                                            CPU     CPU               CPU          CPU
                                                                                                                   CPUs

                                                        CPU        CPU               CPU         CPU                 CPU
                                                       CPU        CPU               CPU         CPU                 Farm
 CPU share: ~ 55% L1, 25%
  HLT, 20% reconstruction
                                                                                Gb Ethernet
                                                                                Level-1 Traffic
                                                                                HLT Traffic
                                                                                Mixed Traffic
                               Level 1
 Trigger strategy:
    Look for high IP tracks (tracking in VELO)
    Confirm track / estimate pT from TT
    Special treatment treatment for trackts pointing
     to calorimeter and muon objects
      ATLAS, CMS cannot use IP at trigger!



 The LHCb VELO:




                                                        f sensor
                                             R sensor
                                                                   Interaction
                                                                     region
    21 stations (~ 100 cm)
    Alternated R-f sensors
    40 μm to 100 μm pitch



    Busy environment:
        ~ 70 tracks/event after L0
        but low occupancy in VELO (~0.5%)
                 Level 1: IP at VELO
 Fast-tracking strategy:
    First in R-Z view (only R sensors)
    Primary vertex σZ ~ 60 mm
    Select tracks with IP in (0.15, 3) mm
     or matching calo or muon candidates
        about 8.5 / event
    3D tracking for those tracks


 pT measurement using TT
    Silicon, 2 layers, 200 mm pitch
    Only 0.15 T.m between VELO and TT
              DpT / pT ~ 20-40%
    Rejects:
        Ghosts
        Low momentum tracks, which can
         fake high IP
                       Level 1 decision
 ‘OR’ of 5 different streams:
    Generic two high pt tracks with IP >
     0.15 mm
    Electron or photon with high pT,
     together with two high PT & IP tracks
    Single muon with high PT & IP
    Dimuons
        High mass, flight significance
        Mass ~ mJ/y, no flight bias

                                              Composition after L1
  Overall L0xL1 efficiency:
      40% for                                             bb %   cc %

         hadronic channels                      Visible   1.1     5.6
         e/γ/π0 channels                       After L0   3.0     10.6
      70-80% for di-muons                      After L1   15.5    18.4
                  HLT flow diagram
  40 KHz (15% bb, 18% cc)
                              Re-reconstruct L1-firing tracks
                             (now using all tracking stations)
                                                                                 Good m
                               and confirm calo/m objects
                                                                                 or calo
                                                     Rest                        object
                               Confirm generic L1 decision
         HLT no                      (p)/p ~ 0.6 %
                                Apply loose pre-selection
10 KHz (45% bb, 20% cc)

                                   Reconstruct all tracks


HLT no                           HLT selection algorithms

              CP channels,         Flight-                  B-generic           D*
              large e,             unbiased J/ys            (single m)          ~300 Hz
              ~200 Hz              ~ 500 Hz                 ~1KHz

     Complete reco                                Storage
                              (more complicated access due to distribution on grid)

          Hot !              New high-rate flows, effect on computing model
                         Status of HLT
 Lot of activity now! Two milestones on summer 2005:
    Trigger-DAQ Challenge: one complete sub-farm running
     continuously on MC data
    Computing TDR


 Status of HLT reconstruction:
    All pieces of code are there
        Calo reconstruction, m id, tracking, L1 confirmation
        Studying feasibility of using pId from RICH at HLT
    Effort is put in improving performance of tracking:
        Efficiency: current losses of 3 to 15% per track depending on channel!
             reverts in > 10% selection inefficiencies!
        Computation of errors in track parameters (needed to compute c2 of
         vertices, significance of flight from PV)
             Use of Kalman filter takes too long
             Alternative method needs to be used           next 2 slides
                 Status of HLT        Tracking
 Obtain track errors from a parameterization on 1/pT:
1. Fit (IPx), (IPx) vs 1/pT         2. Assume cylindrical errors and
                                         insert in covariance matrix




                                                            x
                                                      y   observed




        10 GeV              125 MeV
                                       Use the matrix to compute IP
                                        significances, vertex c2, flight
                                        significances
                  Status of HLT               Tracking
 Advantage of the
  parameterization:                               Flight significance computed
                                                      using parameterized matrix:
    Fast & simple
    Direct control of minimum bias                      Dimuons from offline-
     retention                                            selected B  J/y f
        No need of tuning errors of Kalman
         filter
    Performance proven to be ~ the
     same!


 (Near) future improvement:
                                                           All pairs of pions
    Make use of the fact that main
     component of the error is multiple
     scattering at VELO’s RF foil

                                                                                 FS

                                      Hugo Ruiz                                 18
      Status of HLT              Specific selections
                                                                                Reconstruct all tracks (R

                                                                          HLT selection algorithms
 Status: proved that 10 benchmark
                                                         HLT no

                                                                                       Flight-          B-ge
  channels can be selected (with enough                               CP channels,
                                                                      large e,         unbiased J/ys    (sin
                                                                                       ~ 500 Hz         ~1K
  sidebands) with a rate of some tenths of Hz                         ~200 Hz
                                                             Complete reco                           Storage
     But:                                                        Hot !
         Inefficiency due to tracking
         Some other tenths of channels to be included

 Example: B decays containing a J/y can be very robustly triggered
  by only looking at J/y
    ‘HOT line’:
        Can collect 90Hz with efficiency of ~ 85% on all J/y channels
              50 Hz of true J/y !
       To get ~ 85% on B K*m+m-, use of K* is needed
    BUT:
       Expected ~ 100% (modulus tracking inefficiency)
        for channels with a J/y, via J/y line

                                      Hugo Ruiz                                          19
     Status of HLT           The high rate flows
 Inclusive D*: ~ 300 Hz                                           Reconstruct all tracks (RICH in
    Charm physics
                                                             HLT selection algorithms
    PID calibration: D*+D0( K-p+) p+
      d(mD*-mD0) = 0.5 MeV! (dmD* = 6 MeV)               CP channels,     Flight-           B-generic
                                                         large e,         unbiased J/ys     (single m)
                                                         ~200 Hz          ~ 500 Hz          ~1KHz
                                                 Complete reco                          Storage

 Inclusive J/y mm (+ higher mass                   Hot !
  mm): ~ 600 Hz
    150–200 Hz of J/ y signal (10% from B)
    Pure ‘real track’ sample  detailed
     calibration of tracking resolution
    Unbiased B  J/y X (lifetimes)                                                     m
 Inclusive BmX: ~ 2 kHz                         tagging B
    Lifetime-unbiased b-hadrons with very                                                  X
     good tagging
    B-generic  useful for data mining
    No bias on signal B  study                    signal B
     acceptance and trigger biases
                                     Hugo Ruiz
                                                             BR(B mX) ~ 10%
                                                                         20
      Inclusive J/y: CDF experience
 J/y trigger has been very useful
  for CDF!
 Dimuon trigger: pT m > 1.5 GeV
      ~ 2 million J/y
          80% prompt
          20% from B
   Used to set absolute mass scale
    better than 1 MeV:
      Needed for spectroscopy of B, D and
       quarkonium states


   Understanding of trigger
    acceptance:
      Compare acceptance vs pT, IP
       between MC and high statistics of
       J/y data
         Inclusive J/y: LHCb
 After 0.5 seconds of LHC running:




                                                       +50 MeV
                                   -50 MeV
                                             m = 36 MeV
                                             Offline 9 MeV
 ~ 1.5·109 J/ys per year!
    Expected to make possible study of track errors as
     function of pT
            Inclusive B  mX
 Purity of the triggered sample:
                       For cut in pT
                      of 1,2,3,4 GeV
                                          Can get a purity of
                                           > 50% up to rates of
                                           ~ 2 KHz
                                          Effective number of
                                           generic B’s / year in
                                           LHCb ~ B factories by
                                           2014!

                                                neff = n  etageff

                                            etageff = etag  (1-2w)2
                                                w = wrong tag
                                                   fraction
                     Inclusive B  mX
 Comparison of neff with specific selection (ex: Bd  p+p-, for offline-
  selected events)
               Sample
            triggered by
                            n      etag (%)   w (%)   eeff (%)     neff
             Specific       100     42        35       3.8        3.8
            Single-m       ~4        100      ~ 25     ~ 25        ~1

 As a function of cut on IPS and pT of the m:




                                                                 Very robust against online
                                                                 reco inefficiencies!
                  Conclusions
 L0&L1 of LHCb trigger are mature and ready
    Of course, expect improvements / proposals for upgrades!


 Lot of activity in developing HLT
    Track reconstruction still need some work
    Most of specific selection algorithms are there


 High-rate flows in HLT are promising
    Open doors to new physic analyses
    Allow studying systematics and backgrounds from data
     instead of MC
        Which is safer and…
        ~ 0 economic impact, as less MC needs to be produced!


                            Hugo Ruiz                            25
BACKUP
SLIDES
L1
           Expected event yield
 Taking into account
  efficiency from:
    L0xL1
    Offline selection
Close-up comparison of effective # evts




                 Hugo Ruiz        29
Performance: L0 x L1




                L0 efficiency
                L1 efficiency
                L0L1 efficiency
                Inclusive J/y
 As a function of flight significance:



             Total rate
            True Jpsi
            * True J/y & bb




                              Hugo Ruiz   31
                    MC generation
   PYTHIA 6.2 used
      Minimum bias: hard QCD, single / double diffraction, elastic scattering
      Signal: forcing B-mesons in a minimum bias event to decay into specific
       final state
      Charged particle distributions tuned to data for s < 1.8 TeV
      Predicted cross-sections: inel = 79.2 mb, bb=633 mb
      Pileup (multiple interactions in single bunch crossing) simulated

   GEANT4 for full simulation of all events (minimum bias, signal)

   Additional backgrounds:
      off-beam muons
      low-energy background at muon chambers

   Spillover simulated in detector response
      from two preceding and one following bunch crossings

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:1
posted:2/15/2012
language:
pages:32