Introduction

We are continuing the tracking studies that were started in the ATLAS inner detector TDR . We had particular responsiblity for Chapters 5 and 6. These studies focussed on the efficiency for finding single tracks and then on reconstruction of tracks in b and u jets, using the higgs decay to b and u quarks as a test bench. The efficiency for finding b quarks using a lifetime tag was examined and the corresponding u quark rejection tested.

The work was carried further in ATLAS note 188 where studies of degradation of the detector were conducted to see the effect of removing an entire layer as well as to see the effect of killing some portion of modules and chips.

New studies on this taking into account the effects of pile-up are underway. The remainder of this page concentrates on the techincal details of how to do these studies and where to find the results.

The steps to producing the TDR plots as ps files are:

The control of which tapes to analyze is established by creating a "template" directory structure that ultimately has the tape names and file numbers at the bottom of the tree. Once this is done, the rest of the scripts will automatically search the tree for job sumbmission and analysis.

The organization of the Analysis Code

Alan Poppleton has written the program iPatRec and this is currently being studied. Sets of simulated Higgs data are stored on tapes. The analysis is run on these tapes and ntuples formed. The ntuples are then analyzed by a C++ program and histograms from these are stored. Finally the histograms are manipulated to produce the figures found in the TDR.

Reconstructed Simulated Data from Tape

The first step in the analysis involves running the iPatRec job on the appropriate tapes. In order to understand how this is done, it is usefule to start by looking on atlas.cern.ch in ~rstdenis/public/iPatRec. Here one finds the top directory from which all the analysis is done.

Two maxidisks have been allocated for this work. Links in ~rstdenis/public/iPatRec/maxidisk and maxidisk2 have been set. All files on maxidisk 2 can be accessed by referring to maxidisk: links were set between the two disks.

Of particular interest is the subdirectory reconstruction:

~rstdenis/public/iPatRec/maxidisk/reconstruction

There one finds the various studies done with iPatRec. Under the directory tree headed by iPatRec are the ntuple outputs of the plain reconstruction. Other directories, such as iPatRec_DchipsXXpX head the tree with output where XX.X % of the chips were killed. Similarly for DmodulXXpX.

Under each of these trees, one finds "pile" and "nopile" indicating ntuples with pileup and without pileup. Beneath these in turn are directories "b" and "u" indicating the Higgs decays to b or u quarks were used. Beneath these directories are directories named after the tape and the file on the tape. For example,

~rstdenis/public/iPatRec/maxidisk/reconstruction/iPatRec/nopile/b

contains

T_LT0228_01 T_LT0228_02 T_LT0228_03 T_LT0228_04 T_LT0228_05

Each of these directories contains the ntuple output for the job as well as a script to run the job. There may also be a script called "recon.jobl" that is obsolete. The correct script for tape LT0228 file 01 is

recon_iPatRec_n_b_LT0228_01.jobl

Imbedding the details of the job in the name make it easy to see what jobs were running.

The ntuple output is in

LT0228_01.his.gz

Reconstructing Simulated Data from Tape

In order to do a new reconstruction of the data, it is necessary to copy the template directory structure and rename the top to reflect the new work. For example, if one wishes to try out version 42 of iPatRec, one would go to the maxidisk and copy the "iPatRec_skeleton" to "iPatRec_V42".

cd ~rstdenis/public/iPatRec/maxidisk/reconstruction

cp -r iPatRec_skeleton iPatRec_V42

This assumes write permissions were granted -- the structure could be copied to any disk, although other scripts may need modification. If you have a new disk, you only need to put a link to that disk into ~rstdenis/public/iPatRec/maxidisk/reconstruction. In that case, you would replace the example above with

cd ~myname/public/mymaxidisk/reconstruction

cp -r ~rstdenis/public/iPatRec/maxidisk/reconstruction/iPatRec_skeleton iPatRec_V42

cd ~rstdenis/public/iPatRec/maxidisk/reconstruction

ln -s ~myname/public/mymaxidisk/reconstruction/iPatRec_V42

Once this is done, then the scripts for the jobs must be created. This is done by the script

~rstdenis/public/iPatRec/create_jobs.zsh

No modification should be necessary if the above link to your own space has been established on the maxidisk. For the example above, the command is just

~rstdenis/public/iPatRec/create_jobs.zsh iPatRec_V42

All is now ready for job submission.

Submission and job maintenance

The workhorse of the job control and analysis is a script

~rstdenis/public/iPatRec/job_status_check.zsh

Giving the command

~rstdenis/public/iPatRec/job_status_check.zsh iPatRec_V42 his

gives a list of jobs it would have submitted based on the fact that the ntuples are not there or are too small. In order to do a submission one can type

~rstdenis/public/iPatRec/job_status_check.zsh iPatRec_V42 submit

The submission of which jobs is determined by the directory structure. To modify the tape names and numbers of tapes, the iPatRec_template directory structure should be modified.

Likely failures and Utilities to handle them

Unfortunately it is a bad idea to do the submission as described above. Perhaps there are methods to handle this better, but the fact is that to read the 56 files on one tape, the above submission will cause the tape robot to first mount the tape, get one file, then dismount, get the next, etc. This takes a VERY long time -- a day -- and it seems prone to failure. It is best to use the tape stagein command to stage an entire tape and THEN to submit ONLY for that tape. The arguments to job_status_check continue after submit: one can list tape numbers to EXCLUDE from the submission. The tapes of interest are LT0228, Y20301, Y20302, and Y20308. with 25, 56, 53 and 16 files respectively. Each jobs needs about 1200s (a new version of iPatRec needs 2400s). Hopefully one can stage all four tapes and then use the submission method described above. If there is not enough stage space, then issue the submit command with the tape to be excluded.

The status of jobs can be checked any time by doing the

~rstdenis/public/iPatRec/job_status_check.zsh iPatRec_V42 his

command. This will note if the histograms are there. If all jobs have gone through a queue, then it is possible to resubmit failing jobs by just issuing the command agian with "submit". Jobs with histograms will be skipped and will not be resubmitted. Further information is contained in the comments at the start of the script. One can ask for more details -- such as a printout of the size of each of the histograms -- by using the "hisdeb" argument.

Analyzing the Ntuples

The ntuple analysis is done by copying the ntuple to a directory, gunzipping it, and running the C++ program from Andy Pickford. The script

~rstdenis/public/iPatRec/iPatRecanadev

does all the work. It takes two arguments:

In fact the second argument is a directory in the scratch area under /scratch/zp/. So for rstdenis, the directory /scratch/zp/rstdenis exists and the second argument is rstdenis.

The histogram produced can then be examined with Kumacs.

The Kumacs

The script in ~rstdenis/public/iPatRec/iPatRecPaw takes the same arguments as iPatRecanadev. The results are put into

~rstdenis/public/iPatRec/maxidisk2/ipatrecana/job01

Obviously this script is not as clever and this hardwired feature needs to be modified. The directory above contains directories with names like iPatRec_DChipsXXpX -- and hence somewhere below these are the TDR ps files as well as some useful text files summarizing efficiencies and purities.

The relation of ps file names to figure numbers has been scribbled in my copy of the TDR that now lies in Glasgow. An example of the histograms for iPatRec can be found in

~rstdenis/public/maxidisk2/ipatrecana/job01/iPatRec/nopileup/psplots/

and in

~rstdenis/public/maxidisk2/ipatrecana/job01/iPatRec/nopileup/psbtag/

Text that goes to the PAW screen has been preserved in

~rstdenis/public/maxidisk2/ipatrecana/job01/iPatRec/nopileup/*.output

Here the efficiency and rejection numbers can be read in ascii text. A very handy script is

~rstdenis/public/iPatRec/scan_logfile iPatRec Ru

which scans the output of the condition iPatRec for the string Ru in the "output" file. Issuing the above command gives: 

>scan_logfile iPatRec Ru
------------ Results for iPatRec ---------------------------------------------
========== pileup
  Average Efficiency, Fake Rate, Ru for 50% b-eff
 Ru for 50% b-eff =   77.93536    +-  5.116707
========== nopileup
  Average Efficiency, Fake Rate, Ru for 50% b-eff
 Ru for 50% b-eff =   77.45992    +-  5.031562
-------------------------------------------------------------------------