ATLAS SUSY DATA Challenge

Updated June 11 2003




Introduction


The ATLAS susy group will run a detailed study as part of the ongoing Atlas Data Challenge (DC1)

We have chosen to study the following SUGRA model. m0= 100, m(½)=300, A=-300, mu=+,tan beta=6.0.
The case is chosen to have as many interesting features as possible. It is consistent with existing limits.
 The isasugra implementation of the model is used and the particle spectrum and decay modes can be found here.
Herwig will be used for event generation

Preliminary report on the 500 event sample can be found here

Results were presented at the Physics Workshop in Athens. The talks are here and here and here.


The Plan:

  1. Generate and event sample and pass through a simple Atlfast analysis to ensure that the generated events are satisfactory --- Done at LBNL with release 4.3.0
  2. Pass 500 events through atlsim  (DC1 code release) to produce a fully simulated sample, there are 100 events per file, therefore five data files -- Done at LBNL: using simulation in release 3.2.1 and event generation in 4.4.0
  3. Reconstruct the 500 event sample and reproduce the atlfast plots -- Done with release 4.4.0 at BNL, verified that reconstruction works at CERN
  4. Generate, simulate and reconstruct 5000 event sample utilizing sites/persons interested in participating. These include Sheffield (Dan Tovey) and Cambridge (Chris Lester) -- Goal -- Complete before November 22. Tovey has simulated 500 with release 3.2.1 in Sheffield (12 November 2002). Chris has succeeded in Cambridge(18 Nov). Davide resimulated LBL sample (16 Nov). Kaushik to test on US Testbed.
  5. Generate and simulate  50K event sample as part of DC1 phase 2 -- Goal -- Complete by early January 2003 Done
  6. Reconstruct 50K sample using software release 6.0.0 -- Goal -- end of January 2003 -- release 6.0.0 will not be available before february Done
  7. Dedicated samples with prefiltering for specific studies -- A lepton sample will be needed, others to be determined Dropped
  8. Some issues can be addressed in parallel using fast simulation -- To be determined
  9. Addtional 50K events were produced via the US Grid Testbed April 2003
  10. Reconstruction run on full sample of 100K events April-may 2003.
  11. Present results a Physics workshop -- Athens 21-25 May 2003 Done
 

The Aims:


  1.  Evaluation and testing of Software including GRID tools where possible
  2. Slectron/Smoun masses from dilepton end points and tests of  mu/e universality
  3.  Determination of quark masses via constrained fits 
  4.  The use of the entire shape of kinematic distributions rather than just their end points. 
  5. Develop a strategy for using the event rates (all previous studies have ignored them) 
  6.  Exploit hadronic tau decays to measure the properties of stau 
  7.  Use gluino to bottom sbottom to reduce the combinatorics in determination of the gluino mass 
  8. Determine the stop and sbottom masses and their mixings 
  9.  Understand the impact of jet algorithms on the analysis 
  10. Others to be added.
Some of these require full simulation. others can be done with Athena/Atlfast

Logistics

The Herwig input file is here.
Links to the jobOption files for event generation. Use release 4.5.0 or later (Releases 4.0.0 to 4.4.0 will not make more than 1000 events due to a problem with error exception handling in Herwig)

A file containing 10K events can be found at /afs/cern.ch/user/c/costanzo/scratch0/Susy1a.root and the log file of this generation here.  These can be used for testing simulation in the 5K sample.
The Monte Carlo Truth in 4.5.0 is not acceptable in that the mother/daughter GENZ in atlsim is messed up for Herwig events. This has been fixed and in 4.6.0 (A change to HepMC and Generators/Herwig is needed). This file linked from this page generated with the code that will be in 4.6.0. You can generate with earlier releases and the events will be useable (only this Truth is not functional).

Links to the jobOptions files for processing via Atlfast (Use release 4.3.0 or later)

The events can be passed through full simulation using the DC1 (3.2.1) release of atlsim for which a binary RPM is available. Be careful to set the rapidity range to (-6.0 6.0) . Note that it takes approx 12 mins per event on a 1GHZ processor.  The output files have 100 events per file. Note later versions should not  be used at present.

A file containing 50 fully simulated events is  located at /afs/cern.ch/user/c/costanzo/scratch0/Susy.401.zebra. This can be reconstructed with release 4.6.0 or later using Reconstruction/RecExample (comment out the muon line in the joboptions or you may get a core dump)

Event number assignments for the 5K test. These event numbers are from the the file /afs/cern.ch/user/c/costanzo/scratch0/Susy.401.zebra.
Tovey
 5001-5500
Nordic
 8001-8500
De
 7001-7500
LBL
 6001-6500
Cambridge
 1-500



Events from full simulation run. The output from simulation has filenames  dc1.002315.simul.xxxxx.susy.sugra_100_300_6.zebra where xxxx is replaced by the numbers in the table. Each file has 100 events. The table indicates the status of the files and whether reconstruction has been run. The files were redigitized to correct the drift time in the MDT's. redigi replaces simul in the file names. These redigitized files were used for reconstruction.

site
 Event numbers
 in Castor
 In AMI
 in Magda
 reconstructed
 contact person
Sheffield
 01001-01100
 Yes
no
 yes
yes (6.0.0)and later
 Dan Tovey
Cambridge
 01101-01200
 Yes
no
 yes
yes (6.0.0)and later
 Chris Lester
Copenhagen
 01201-01300
 Yes
no
 yes
yes (6.0.0)and later
 Else Lytken
LBL
 01301-01400
 yes
 no
 yes
yes (5.1.0)and later
 Davide Costanzo
Israeli
 01401-01500
 yes
no
 yes
 yes (6.0.0)and later
 Ehud Duchovni
US Testbed grid
 01501-02000
 yes
 no
 yes
 yes 6.0.3
 Kaushik De








The Generated files are dc1.002315.evgen.xxxx.susy.sugra_100_300_6.root where yyyy is 0010 to 0019. They are in castor (~ianh and DC) and in magda both BNL and LBL. There are 10K events per file. The mapping from generated to simulated is 0010 fills 01001-01100, 0011 fills 01101-01200 etc.
There will shortly be a link from this page to the reconstructed Ntuples.

April, 2003: All 50k simulated events with redigitized muons are now in /castor/cern.ch/atlas/replica/dc1/simul/data/002315 with file names dc1.002315.redigi.NNNNN.susy.sugra_100_300_6.zebra. (The *.simul.* files in this Castor directory have incorrect muon digits and should not be used.)

May 2003:  Reconstructed files: Reconstruction was done without calorometer noise, with noise and no cuts, with noise and a cut on the noise.  Ntuples for the reconstructed data are available from this web site

Use the nsls command to list Castor files and the rfcp command to copy Castor files to local disks.

Comments on the physics:


The case that has been chosen is described very briefly here. It is similar to "Point 5" discussed in detail in the physics TDR (Chapter 20). In particular Figures 20-11, 20-12, 20-13, 20-20, 20-2, 20-26,  20-22, 20-28, 20-29, 20-31, 20-32, 20-33, 20-38 and 20-39. Other analyses such as those outlined in Section 20.2.8.2 are also applicable since tan beta for the new case is bigger than for point 5 and the tau tau final states have more rate than the electon and muon states. The decay chi_2 to chi_1 Higgs does not occur as the Higgs mass is larger.


Mailing list of atlas susy group

Ian Hinchliffe   Modified 11 June 2003