Yet another approach to the ToF-based PID at PANDA A.Kiselev, Yu.Naryshkin PANDA Collaboration Meeting Stockholm, 14.06.2010
Layout of the talk Motivation of this work Suggested formalism Monte-Carlo results Discussion
ToF detectors at PANDA Forward Wall Barrel ToF (?) Side Wall Continuous beam , no “start” counters tstart unknown!
Algorithms considered so far Three main procedures: Reconstruction from one identified particle Minimization mass residuals under start time variations Mass hypothesis find time offset The following 3 slides: Stefano Spataro, Sep’2008
Start time reconstruction: the 1-st way identify one particle in the event (assume its mass) calculate its time-of-flight (from momentum and path length) calculate START time of the reaction recalculate tof of the other particles in the event
Start time reconstruction: the 2nd way New algorithm was implemented (Alexander Schmah) variation of start time tstart calculate mass from tstart, momentum, tof, path length find the closest realistic mass construct residual for all the particles Find tstart that minimizes the Q functional
Start time reconstruction: the 3rd way use PID informations from other detectors or mass hypothesis per each particle calculate the time-of-flight calculate the mean offset or not stand-alone Improvements for the global PID …Work in progress
The “4-th way”: goals Quantify PID decision as much as possible Provide statistically significant formulation Have a better interface to tracking package Be flexible in terms of input: do not require ANY external PID info per default be able to use such information, if available Allow flexible output: be able either to provide its own PID decision … … or, alternatively, give input for the “global” PID
The “4-th way”: logic consider all N tracks in a given event at once, that’s clear assume each track can be pion/kaon/proton evaluate all 3N {m1,…,mN} mass configurations separately, one by one obtain their c2 weights (and tS offsets) reject (reweight) some, using external PID input compare various mass configurations obtain the significance estimate select the “most probable” one(s)
Formalism (simplified version) how to evaluate a given {m1,…,mN} mass configuration? Y (tS0) is distributed as c2 with N-1 degrees of freedom for “correct” mass configurations Tends to have higher c2 values for the “wrong” ones
Standalone PID evaluation calculate “weight” for each {m1,…,mN} configuration: define “probability” of j-th track to be say a pion as use Monte-Carlo to check selection efficiency & proton contamination (in this report: PYTHIA @10GeV/c, forward spectrometer tracks only)
Proton-pion separation Pions, <p>=2.15 GeV/c Protons, <p>=3.25 GeV/c NB: fluxes are not accounted here!
Proton-pion separation, cont’d Consider modest (100ps) ToF resolution Pions Protons
Kaon separation <p>=2.7 GeV/c NB: again, fluxes are not accounted here!
Formalism (advanced version) Extra requirements to the tracking package: append {tS} to the “standard” Kalman filter {x,y,sx,sy,1/p} parameter set do track fitting for pion/kaon/proton hypotheses separately provide {p,l,tS}i parameter set with cov.matrix on per-track (and per-mass-hypothesis) basis This allows to extend the suggested formalism in a way, that ToF-based PID 1) is not biased by unknown tS offsets during the 1-st tracking pass, 2) gives the most complete and statistically meaningful set of information to the “global” PANDA PID code
Formalism (advanced, in detail) Kalman filter c2 Track->event matching ToF measurement extended parameter set requires further studies!
Summary The (relatively) new formalism to handle PANDA ToF-based PID problem is suggested It allows to implement a statistically solid interface between tracking code, ToF measurements and the “global” PID