6. ALICE simulations
6.1 Introduction
This section describes the current (08/2002) status of the ALICE
simulation enviroment. Since this is under heavy development, this section
is probably allready now outdated. But nevertheless, a short description,
since the only usefull documentation is the code itself, most of the stuff
on the webpages is terrible outdated. And I need to remember how I get
this stuff installed...
The ALICE simulation and reconstruction package is called AliROOT and based on ROOT. It can be obtained via anonymous CVS
from the CERN server. A short description how to get it/how I got it:
Look on the AliSoft webpage
what is the latest tagged version, this is available via the push-down
list at the end of the CVS status page
. Now set the CVS server: setenv CVSROOT blah blah
(username cvs) and login with cvs login (password cvs). Now
you can get the latest version with cvs -z3 checkout -r tag
AliRoot , where tag is the latest tag
obtained before. You need to set some enviroment variables specific to
your enviroment, a sample for my laptop is here . Don't
forget to set the variables in the ~/.rootrc file, a sample
is in the AliRoot/ directory. Compiling takes a while.
When it's finished, you can test you installation by calling the
grun.C macro with aliroot -l -b -q grun.C which produces
a test file.
6.2 Status
In the ideal case, with one AliROOT run, one could produce particles
with one or a combination of eventgenerators, simulat the passage of
the particles through the ALICE detectors, simulate the response of the
detectors/electronics and finally reconstruct and match the reconstructed
particles with the MC-particles. But since AliROOT is still under development,
only parts of this are possible - how much depend on the detector...
Each detector has it's own set of ROOT macros, which can be found
in AliRoot/detectorname/*.C. Most of the functionality is
in the detector class AliRoot/detectorname/*.{h,cxx}. Some
things are common for most of the detectors. Starting point are the MC-tracks
produced by the EventGenerator (these are available in ???). After the
detector simulation (GEANT) you get the hits in the detectors
(???). The hits are the starting points for the detector
response simulation. This task is splitted into two steps, the first one
leads to the generation of sdigits (TreeS) and contains all
linear effects. The second one goes from sdigits to
digits (TreeD) and contains the nonlinear effects (e.g. noise).
This approach allows to generate backgound (e.g. HIJING) and signal events
(e.g. PYTHIA) up to the sdigits level and merge them there,
which saves a lot of processing time.
6.2.1 ALICE-USA EMCAL proposal
For the ALICE-USA EMCAL proposal, some jet simulations where run.
These were focussed on the EMCAL/PHOS. The TPC hits were not saved,
the response of the TPC was "simulated" with a gaussian smearing of
the momenta according to dp/p = sqrt(a**2 + (b*p)**2) * 10^-2
with some values for a and b depending on occupancy
of the TPC. The code is in AliROOT/EMCAL/AliEMCALFast.cxx
. The tracking efficiency is taken into account with a random reject
of tracks, the assumption is that the efficiency is 90-95% for tracks
with p> 0.5GeV/c. The loss of tracks due to the sector gaps is accounted
for by the application of a geometrical acceptance factor of 90%. This
might be sufficient for a first look at the EMCAL/PHOS, but clearly are
not sufficient for a analysis with the TPC as main detector...
All simulated data can be found on PDSF, Berkley.
Some words to the jet finder they used: It's based on an "old"
STAR finder (it's still used by the Wayne State STAR-EMC group, however
there are newer versions available which are used by Yale/BNL/LBNL). There
is some C++ wrapper (AliROOT/EMCAL/AliEMCALJetFinder.h )
on the fortran finder ( AliROOT/EMCAL/jet_finder_ua1.F). It's
a seed cluster algorithm as used by UA1, probably this implementation
is from the LUND JETSET program (merged with PYTHIA from PYTHIA version
6 onwards). The C++ wrapper class has some functions for hadronic background
subtraction based on the TPC information and supports the integration
into the AliROOT infrastructure.
There is a second round of ALICE-USA/EMCAL simulations sheduled
for the end of this year which probably includes also TPC + ITS clusterfinding/tracking.
These might become very usefull for the HLT studies, so it make sense
to look for them... Maybe joint with the TPC PPR simulations???
6.2.2 TPC simulation
??? - this needs to be clarified as soon as possible - how "final"
are clusterfinding/tracking/matching... Simply running the macros is no
problem... Maybe there exist some jet simulations for the TPC? If they
are reliable, this would be the perfect starting point!!!
Next round of TPC-simulations for PPR with
clusterfinding/tracking/matching is coming:
Mail from Andres Sandoval Fri, Sep 6:
> There is going to be a production in the frame of the PPR of
jet-jet and
> gamma-jet embeded in Hijing events very soon with at the moment
100 events
> from Jets from 25,50, 75, 100, 200 GeV. This events will
simulate the
> whole setup, create hits and digits and reconstruct the particles
with
> Kalman filter.
So, THIS IS the perfect starting point.
But someone must check, what they are going to produce, what are
the simulation-parameters etc... Decisions at ALICE-week, software-week???
6.2.3 HLT simulation
As far as I know, not tested so far...
Based on the expierence from STAR, the starting point are the
digits on which an offline version of the online clusterfinder/tracking/analysis
algorithms is run. The current STAR implementation allows association
with MC tracks as well as with offline tracks. This allows "eazy" calculations
of tracking efficiency, trigger efficiency etc... For all this, it's
very important to be in contact/understand what offline is doing...
The current HLT implementation uses indeed the digits as starting point,
two different tracking chains are under development:
a) conventional clusterfinding + tracking, currently used for low multiplicities
(up to dN/dy=3000).
b) hough-transform tracking, used for high multiplicities...
Current HLT jet
finding simulations are mickey mouse simulations... I'm still looking
at the jetfinder they used...
For reliable results on the applicability of the jet algorithms
on the HLT such a simulation is needed/crucial, everything here is
only a first approximation... (Constantin, have fun ;-)
6.3 Proposal for simplyfied parameterized "simulation"
for this study
Since this should be only a first look and we don't know the momentum
resolution/tracking efficiency etc of the HLT software, I propose to use
a simple parametrization, however a little bit more advanced than the one
in EMCAL... Let's run AliROOT with simulation of the TPC and save the
TPC hits. This gives the geometrical acceptance and removes one weak point
of the EMCAL "simulation". Then we take some reasonable simulation of the
momentum resoltion and efficiency (the TPC group should have one, the one
from the EMCAL proposal is clearly wrong. Explanation why... ) Probably
it makes sense to vary the parameters of the parametrization to get a feeling
how sensitive one is to efficiency/momentum resolution. And since the
HLT resolution/efficiency is probably a little bit worse than offline it
makes sense anyway. Again, this is only a first approximation, to do
it right one needs offline TPC simulation/clusterfinding/tracking and
the same for the HLT...
Thorsten Kollegger
Constantin Loizides
IKF - University of Frankfurt
Last updated: 10/30/2002 03:08pm EDT