Physics Analysis With AliRoot

1
Physics Analysis With AliRoot

• Peter Hristov
• ALICE Collaboration
• NEC2005, September 16, Varna

2
Outline

• The framework AliRoot
• Analysis examples
• Distributed analysis and parallel analysis
  facility
• Summary of the analysis tools in AliRoot
• Conclusions

3
The frameworkAliRoot

• Up-to-date description of Alice detectors
  (geometry, physics processes, detector response,
  etc.)
• Rich set of event generators, easily extensible
• Possibility to use different transport packages
• Robust IO scheme
• User friendly steering classes for simulation and
  reconstruction

• Efficient tracking in the barrel detectors and in
  the MUON arm
• Elaborated reconstruction of the neutral
  particles
• Reconstruction of V0s
• Combined PID based on the Bayesian approach
• ESD classes suitable for further analysis
• Detailed analysis examples
• Possibility to explore wide spectrum of heavy-ion
  and pp physics

4
AliRoot Layout
G3
G4
Fluka
HIJING
PDF
EGEEAliEn
GRID
VZERO
CRT
STRUCT
Virtual MC
PYTHIA6
HLT
START
DPMJET
EVGEN
RAW
Monit
STEER AliSimulation AliReconstruction ESD classes
PMD
ISAJET
Analysis
HBTAN
JETAN
FMD
ITS
TPC
TRD
TOF
PHOS
EMCAL
RICH
MUON
ZDC
ROOT
CINT
HIST
GRAPH
TREES
CONT
IO
MATH

5
AliRoot Execution Flow
Initialization
Event Generation
Particle Transport
Hits
AliSimulation
Event merging
Digits/ Raw digits
Clusters
Summable Digits
AliReconstruction
Tracking
PID
ESD
Analysis
6
Event Model

• RAW data Written once, read up to 3 times.
• Size 1(pp)- 50(PbPb) MB per event.
• Event Summary Data (ESD) Written up to 3 times,
  read many times. Contains all the information
  needed for analysis, fine-tune calibration and
  further processing into AOD. Single version for
  everybody.
• Size 1/10 of raw per event.
• Analysis Object Data (AOD) Written many times,
  read many times. Specific for a set of analysis
  tasks.
• Size 1/10 of ESD per event.
• Tag Written few times, read many times.
• Size 100 B 1 kb per event, exist many per
  event.
• Done for fast event data selection global
  experiment tags, physics working group tags, user
  defined tags.

7
Existing Analysis Examples in AliRoot

• ESD analysis
• V0 cascade reconstruction/analysis
• Open charm
• Specific AOD-based analysis
• HBT
• Jet
• V0
• MUON
• Any user can contribute with additional examples
  and extending the functionalities

8
PT Distributions of ?,K,p

• Experimental PT distributions.
• Selection of tracks from the primary vertex cuts
  on impact parameters
• PID using probabilities as weights, taking the
  most probable type, taking type above a
  probability threshold,
• Corrections
• Contamination comparison of the PID and true MC
  particle type. Depends on the way one obtains the
  exp. PT spectra.
• Combined acceptance Geometrical acc. x Decays x
  Reconstruction efficiency x PID efficiency.
  Depends on the primary vertex position, momentum,
  and occupancy gt Comparison between MC kinematics
  tree and reconstructed particles.
• Resolution comparison between the MC and
  reconstructed momenta.
• gt The framework provides easy access to the
  kinematics tree.

9
Example Histograms PbPb events
10
? Meson Mass, Width, PT Distribution, Yield

• Selection of tracks from the vertex, PID of
  kaons.
• Cuts to increase the significance (5 parameters
  for each pair PT, azimuthal and polar angles,
  azimuthal and polar angles in the rest frame).
• Effective mass distribution for all pairs of kaon
  candidates (global and for different PT
  intervals).
• Effective mass distribution for the
  (combinatorial) background.
• Event mixing positive particles from the same
  event, negative from a different, but similar
  one. Similarity close event impact parameters
  and number of negative particles. Event plane?
  One might rotate all the negative tracks
  accordingly.
• Mass resolution, combined acceptance from MC.
• gt The framework provides easy access to events
  for mixing.

11
? Meson Mass in pp Events
12
V0s and Cascades PT Spectra, Yields

• V0 pair of tracks with opposite charge, selected
  according to the following criteria
• Min. Allowed impact parameter(in XY) for each
  track.
• Max. Allowed DCA between the two tracks.
• Max. Allowed cosine of the V0 pointing angle.
• Min. And max. Radius of fiducial volume.
• Max. ?2 gt needs covariance matrices of the track
  parameters.
• Cascades V0 and bachelor track selected
  according the cuts above and in addition.
• Mass window for V0.
• Effective mass distribution in different PT
  intervals -gt fit (or background subtraction from
  mixed events) -gt experimental PT.
• Corrections gt Need for MC for combined
  acceptance, resolutions,

13
(No Transcript)
14
(No Transcript)
15
D ? D0 ?

• Search for D0 candidate with some loose cuts.
  (See the D0 analysis classes.) combined with pion
  from the primary vertex -gt Distribution of m(D0?)
  - m(D0)
• Secondary vertex close to the primary one. Need
  for precise secondary vertex finding/fitting,
  especially in the case of proton-proton events.
• Estimation of the background distributions gt
  event mixing
• Combined acceptance, resolutions, etc. gt from MC
• gt Classes for precise secondary vertex
  finding/fitting are available within the framework

16
Example Jets

• Leading hadrons correlations
• Jet reconstruction algorithm UA1 cone algorithm
  modified for HI
• Cone size R lt 1
• pt cut
• Background subtraction
• Event by event jet identification
• Jet parameters multiplicity, kt, f(z), g-jet
• Jet spectrum

17
Jets
Pythia
Hijing
18
Tag event

• Information (meta-data) to select and analyze
  events of interest only via user-selected
  conditions
• The framework (inspired from STAR)
• Collects the file identifiers where to find the
  events
• Localizes the files
• Passes the events to the analysis program
• Metadata (gt 70 parameters)
• Run specific information
• Information concerning the state of LHC per ALICE
  run
• Detector information per run
• Information about each event being collected from
  PWG
• Populated during and after ESD creation
• Architecture defined, partially implemented,
  analysis cuts tested

19
Distributed analysis

• Prototype demonstrated to work
• Used routinely by one single user
• Not yet released to general users because of
  major middleware reorganization (Alien2/gLite) ?
  end 2005
• For most of the analyses the files have been
  moved to a single location to be analysed
• Proposal for a Parallel Analysis Facility
• Fast on line processing of data (calibration,
  tuning of algorithms, online physics results)
• Few users (detectors and PWG experts), many x a
  few events, short latency
• Low cost farm of dual core processors

20
Analysis Job Structure
File Catalogue query
User job (many events)
Data set (ESDs, other)
Job output
Job Optimizer
Grouped by SE files location
Sub-job 2
Sub-job n
Job Broker
Submit to CE with closest SE
CE and SE
CE and SE
CE and SE
processing
processing
processing
Output file 2
Output file n
File merging job
21
AliEn2 Development

• User Interface (gShell)
• Universal only one for accessing all the Grids
• Complete Provides all the functionality a user
  might need
• Robust Does not crash or hang up in case of
  misuse
• Simple Does not require expert knowledge to run
  it
• New catalog
• Fast
• Reliable
• Secure data managment encrypted access envelopes

• Job agents
• Response to analysis requests
• Priorities
• Quotas
• Transport protocol xrootd
• Uniformity with Root and PROOF
• API under active development, available now in
  Root
• Installation new installation utility for easy
  installation

22
Parallel Analysis Facility

• Interactive prompt analysis
• Data recorded at CERN T0
• After first reconstruction pass analyze results
  using PROOF on CERN T1
• Check detector functioning and hot channels,
  calibration, alignment, physics
• Prototyping in progress
• Redesign of the existing analysis classes using
  Root selectors

23
Interactive Analysis with PROOF on a Large Cluster
PROOF SLAVE SERVERS
PROOF SLAVE SERVERS
PROOF SLAVE SERVERS
Slave servers access data via xrootd from local
disk pools
PROOF SUB-MASTER SERVER
Proofd Startup
Grid Service Interfaces
PROOF MASTER SERVER
TGrid UI/Queue UI
Guaranteed site access through PROOF Sub-Masters
calling out to Master (agent technology)
Grid Access Control Service
Grid/Root Authentication
Grid File/Metadata Catalogue
USER SESSION
Client retrieves list of logical files (LFN MSN)
24
Summary of Analysis Tools

• The tools work on ESD/AOD.
• Basic kinematics from Root 3-vectors, 4-vectors,
  rotations, boosts, eff. masses, PT calculations
• Geometrical tools propagation of tracks, DCA,
  primary and secondary vertexes.
• Specific V0s, cascades -gt AliV0vertexer,
  AliCascadeVertexer.
• General AliITSVertexerTracks.

25
Summary of Analysis Tools

• Impact parameter of the interaction based on ZDC
  energies and number of participants from ESD.
• Event plane AliFlowAnalysis.
• Typical tasks
• Comparison with MC and access to the kinematics
  tree. Creation of MC AOD
• Calculation of combined acceptance (physics
  efficiency)
• Background studies using the event mixing
  technique.

26
(Trivial) Practical Advises

• Instantiate the objects outside the analysis
  loops and use setters inside. Use references
  instead of new objects.
• Do the pre-selection of tracks and particles in
  single loop before the main analysis. Use indexes
  to access the original objects.
• Use standard tools for sorting, minimization,
  etc. Do not rewrite the existing ROOT tools!
• Load to memory when possible (example
  tree-gtLoadBaskets())
• Foresee checks for the data consistency

27
Conclusions

• The main tools for data analysis are available in
  Root and AliRoot
• The analysis code based on ESD / AODs seems
  adequate, and is evolving according to feedback
• The infrastructure tools for distributed analysis
  are under active development
• Several detailed analysis examples are available
• Analysis infrastructure for distributed data will
  be offered to test users 1Q06