As already underlined this search will proof all CMS features. We wil
therefore work side to side with the working groups in charge of the
high-level analysis tools ( energy reconstruction by Particle Flow,
b-tagging, high pT electrons tagging, etc.). The successful candidate
will focus on the main streams for this analysis, performing a
study of the signal and the background features using the detailed
simulation of CMS. He will also validate on the real data the results
of the simulation, focusing on the means of high pT muon
reconstruction, exploiting  our group consolidated expertise in
dealing with the CMS muon chambers.
Simulation studies will proceed as follows:
- generation of several signal sets (about 10000 events each) in
the two lepton final state for different values of the heavy fermion
masses, sampled in the range 500 GeV to 1 TeV. These events will then
be processed with a detailed simulation of the CMS apparatus
(including the underlying event) ;
- measurement of the dilepton trigger efficiency on these datasets;
- definition of preselection skims, based on very loose cuts on the
leptons pT, on the missing energy and on the jet multiplicity, assuring
high selection efficiency and significant background reduction (as
measured on CMS fast simulation);
- selection of the background events from the detailed simulation of
standard processes, filtered with the above mentioned skims ;
- definition of a set of sequential cuts on the leptons pT, jet
multiplicity, b-tagging and missing energy. The cuts will be optimized
for different values of the heavy fermion masses and for different
luminosity scenarios;
- measurement - in an independent set of simulated events - of the
excess of signal event over background based on the invariant mass
distribution of the surviving candidates, enforcing all possible
kinematic constraints on top and W masses ;
- definition of a data driven means of background determination, using
e.g. events with opposite lepton charge, side-bands, control
samples in which one of the cuts is reversed;
- estimate of the systematic error on the background achievable this
way;

The appliant will meanwhile validate on the real data the muon
identification tools, in detail:
- measuremnt of the hadron fake-muon rate, from either hadrons
punching through the calorimeters and the iron yoke, or decay in
flight of pions and kaons. Punch-through can be reduced to a
negligible level requiring hits in the outer layers of the CMS muon
chambers. Decays in flight are reduced with cuts on the track quality
and rejecting tracks with large impact parameters w.r.t the beam
collision point. Fake rate can be measured in the data using samples
of pions from Ks decays, of Kaons from Phi and of protons and pions
from Lambda. Preliminary studies show that CMS will reconstruct several
thousand resonances (Ks,Phi or Lambda)  per inverse pb delivered by
LHC.
- measurement of the fraction of charge mis-assignment.  As mentioned
previously,  few per mille errors induce too large a background for
this analysis. The large magnetic field of CMS (4T) and its refined
tracking system (inner layers of silicon tracker and an outer system
of drift tubes for muon id) should provide an error as small as some
part in 10^-4. These values are validated using cosmic rays collected
by CMS in its fully operational configuration. Tens of thousand high
momentum tracks traversing all the tracking system are already collected on
tape, another cosmic run will take place next summes, just before LHC
start. Charge confusion rate will be measured by splitting the track
in two halves reconstructed in he top and in the bottom part of the
apparatus, and comparing then charges of the two half tracks.

We underline that for an early stage of the analysis, we do not need
to know the muon reconstruction efficiency with a precision much
larger than 5-10 %. (A much better precision would be needed in case
of observation of a large signal, to perfrom a precise determination
of the heavy fermion production cross section). This precision will be
easily achieved with samples of Z0 and Upsilon decaying to muon pairs,
triggered by one single muon.

We expect a candidate with a firm experience in data analysis, able to
deal with the signals produced by modern particle detectors. She/he
should be at ease with basic statistics applied to HEP. Even if not
strictly necessary, experience with the use of multivariate algorithms
for event classification (neural network, boosted decision trees,
random forests, etc.) is welcome. (These tools might be used instead of
a cut based analysis in a later stage of the search). 
The candidate must be an expert OO c++ programmer, familiar with the
analysis packages (ROOT, MINUIT) commonly used in HEP.
We appreciate - but do not strictly require - knowledge the high level
languages used by CMS (Perl, Python). We require also fair English
knowledge, and easiness in dealing with the coworkers.