From abid@fnal.gov Mon Apr 17 15:30:45 2006 Date: Fri, 14 Apr 2006 14:11:16 -0500 (CDT) From: Abid Patwa To: Daniela Bauer Cc: Frederique Badaud , Per Jonsson , Florent Lacroix Subject: Re: L3 taus -- for the minutes Hi Daniela, Sure, I can fill in some of the blanks... Frederique, can you please fill in the details for NN-based, CAL-based, TRK-based algos below? For the tau-related info, their are two categories of tau decays: a) taus decaying leptonically (to e or mu) or b) taus decaying hadronically (to jets w/ or w/o EM subclusters). For the L3 tau triggers, we consider mainly taus decaying hadronically. There are e+tau and mu+tau triggers that get defined in the trigger list and again, the tau decays hadronically with electrons or muons, respectively. In turn, we classify these "hadronic" decays of the taus by 3 tau-types. They are characterized by what signature they leave behind: a) tau-type 1: tau decays to charged pion + MET (i.e., pi-like) * yielding CAL cluster and one track b) tau-type 2: tau decays to charged pion + pi0 + MET (i.e., rho-like) * yields CAL cluster + one track + EM subcluster from pi0 c) tau-type 3: this separates into two subclasses (both representing a 3-prong decay) c.1) tau decays to 3 charged pions (pi+/-, pi+/-, pi-/+) + MET c.2) tau decays to 3 charged pions + pi0 + MET *yields a wide CAL cluster plus 2 or more tracks and no EM subcluster for (c.1) and EM subcluster for (c.2) Next, a summary of the variables used in the input to the tau-NN at L3 --- please note that the variables are ratios of energy and tracking quantities, which serves to give a range between 0 and 1 and helps restrict the NN output to 0 and 1. a) rms: a measure of the tau-cluster width. * rms = sqrt(((dphi^2+deta^2)*E_T)/E_T) * can be used for all tau-types. b) fhf: fine hadronic fraction; * or the fraction of ET(tau) in the fine hadronic section of the calorimeter * usually used for tau-types 1 and 2 for best discrimination. c) EM12isof: an isolation fraction in terms of EM energies. * EM12isof = [E_EM(layer1) + E_EM(layer2)]/E(tau) in a R < 0.5 cone * E_EM(layer1), E_EM(layer2) are energies deposited in the first two layers of the EM calorimeter. * usually a parameter used in the NN for tau-type 1 to reject jets with one energetic track and soft pi0 mesons. d) iso: tau isolation based on calorimeter * iso = [ET(R<0.5) - ET(R<0.3)]/ET(R<0.3) * usually used for all tau-types. e) prof: tau energy profile measured by calorimeter towers. * profile = (ET1 + ET2)/ET(tau) * ET1, ET2 are the ET's of the two most energetic calorimeter towers * usually used for all tau types. The "2" labeled in the slides for iso2, profile2, and rms2 indicate a special eta-parameterization that is done on these variables to help better separate signal from background across the full range in eta. These variables without a "2" marking indicates that no such eta-parameterization is applied. That should be most of it... Frederique should provide details of the L3 algorithms. Abid ----------------------------------------- On Wed, 12 Apr 2006, Daniela Bauer wrote: > Hi All, > > I am trying to write the minutes for yesterdays meeting, but I am having > trouble with the taus. I am not sure I understand what the algorithms are > actually doing. Can you tell me in one or two sentences what kind of event > 'signature' you are looking for ? What kind of tau decays are you after > typically ? > > At Level 3 there are 3 different algorithms for tau reconstruction > available: > > NN-based: for each cluster (cal cluster ?? aren't those way too many ? is > there any kind of slection on it ? ) a NN output (of which > variables ?) is calculated (and later cut on, I assume ?). These clusters > are then matched to one or more tracks. The number of tracks is > meaningfull, why ? > > CAL-based: Based on cal clusters with ET > minET (rcp parameter) and a > width > minWidth (trigger list parameter) -- why are you looking for wide > clusters ? - and profile (which is what ?) > minProfile (trigger list > parameter). These clusters are matched to tracks and ??? what do you do > with that information ? > > TRACK-based: what is a "track cluster" ? > This algorithm starts from tracks, and not from cal clusters ? > > > What is type1, type2, type3 (slide 4) ? > > The variables: > rms (of what ?) > fhf: fine hadronic fraction (of the cal cluster ?) > EM12isof: what is E^{EM1} and E^{EM2} > iso: isolation (of the cal cluster ?) > prof: profile (what is that ?) > > > Thanks, > Daniela > > > ------------------------------------------------------------------- > Daniela Bauer > Imperial College E-Mail: dbauer@fnal.gov > MS 352 Fermilab > PO Box 500 FNAL Phone: +1-630-840-3859 > Batavia, IL 60510 FNAL Fax: +1-630-840-8886 > http://www-d0.fnal.gov/~dbauer > ------------------------------------------------------------------- > >