Hi- To follow up on Daniele's posting ... (this is all found on my page on brem recovery linked from ISL->vubnosf->...) As stated in the meeting - i explored 2 different ways of identifying brem photons - the opening angle between the lepton 4-vec and the photon 4-vec, and the difference in theta, phi origin between the lepton and photon. I compared my 2 ways and Daniele's way and here are the pstar and Mx plots: http://www.slac.stanford.edu/BFROOT/www/Physics/Analysis/AWG/InclusiveSL/vub-nosf/ejhill/ana_issues/bremrecov/pstar-comp.eps http://www.slac.stanford.edu/BFROOT/www/Physics/Analysis/AWG/InclusiveSL/vub-nosf/ejhill/ana_issues/bremrecov/mxhadfit-comp.eps The 3 methods are pretty similar here, and agree with what Daniele posted. A much more interesting plot is this one-> I made a plot of the pstar for ONLY the events where brem was recovered - and compared it to those corresponding events before brem recovery. http://www.slac.stanford.edu/BFROOT/www/Physics/Analysis/AWG/InclusiveSL/vub-nosf/ejhill/ana_issues/bremrecov/pstar-comp_bremrecov.eps You can see very clearly in this plot the positive effect of applying a brem recovery. I also checked on the efficiency of making the selection for recovery For a sample of signal (brevubmix)-> **TRUTH is determined if the photon is: 1. a daughter of a B (apparently they are written in the truth block as B->Xlvg, not B->Xlv; l->lg) 2. the photon's mother B is the same mother B of the electron (i.e. - the electron and photon are from the same B) If i cut on opening angle between e, gam: 1084 leptons are selected as having brem photon(s) -->1004 of those are TRUE brem leptons ( 80 are mistakenly selected) 401 TRUE brem leptons are missed by the selection N(selected)/N(True) = 0.77 If i cut on delta theta/phi seperately: 1064 leptons are selected as having brem photons --> 993 of those are TRUE brem leptons (71 are mistakenly selected) 412 TRUE brem leptonss are missed by the selection N(selected)/N(True) = 0.76 Following DDR's cutting scheme: 1072 leptons are selected as having brem photons --> 985 of those are TRUE brem leptons (87 are mistakenly selected) 420 TRUE brem leptons are missed by the selection N(selected)/N(True) = 0.76 Cheers, Ed On Friday 09 April 2004 04:39 pm, Daniele del Re allegedly wrote: > Hi all, > > when studying momentum resolution in lepton ID I realized that > that bremsstrahlung for electrons can introduce large biases in > p* efficiency and resolution. It can also affect the mm2 resolution > and finally it produces large biases in the inclusive Mx reconstruction. > > I the past we always neglected this point. > > A couple of examples. Here is the p* resolution comparison > between e and mu: > > http://www.slac.stanford.edu/~daniele/exclvub/bremstudy/resoelemu.eps > > that shows that the effect is large. And this is mm2 for pi0lnu: > > http://www.slac.stanford.edu/~daniele/exclvub/bremstudy/mm2pi0elemu.eps > > (for the inclusive calculation of mm2 the effect is smaller because the > brem photon is included in the X system and then recoil has it). > > I implemented a brem recovery algorithm that is very similar to the one > in Beta. Code is in VubAnalysis (in util.icc). > It is based on two variables > > - Delta(theta): it is the difference between the theta of the brem photon > and the electron one. They should be the same within resolutions. > > - DD = (Phi(gamma)-Phi(e production))/Phi(e EMC)-Phi(e production)): here > I am requesting that the phi of the gamma is in between the phi of the > electron at the EMC and the phi at the origin. I used this ratio > (instead of a simple cut on Phi(gamma)-Phi(e production) as in Beta) > because numerator and denominator are somehow proportional. > > Here the distributions for electrons > Delta(theta) (for -0.3<DD<0.6) > http://www.slac.stanford.edu/~daniele/exclvub/bremstudy/Deltatheta.eps > > DD (for -0.04<Delta(theta)<0.04) > http://www.slac.stanford.edu/~daniele/exclvub/bremstudy/DD.eps > > I finally select brem photon with these cuts > -0.04<Delta(theta)<0.04 && -0.3<DD<0.6 > and I redefined the electron p4 as p4+p4(brem photon). Moreover the > brem photons are removed from the good photon list (goodPhoton[i] = 0) > > > The following plots are a comparison with and without the correction: > > resolution in p*: > http://www.slac.stanford.edu/~daniele/exclvub/bremstudy/resoelecorr.eps > > mm2 for pi0lnu: > http://www.slac.stanford.edu/~daniele/exclvub/bremstudy/mm2pi0elecorr.eps > > inclusive mm2 > http://www.slac.stanford.edu/~daniele/exclvub/bremstudy/mm2elecorr.eps > (small effect as expected, see above, mostly due to gain in p* cut > efficiency) > > inclusive fitted X mass (mxhadfit) > > http://www.slac.stanford.edu/~daniele/exclvub/bremstudy/mxhadfitelecorr.eps > > > Conclusions: > - brem recovery improves largely the resolution in p* and the efficiency > of p* cut (~+5%) > - it improves mm2 resolution in exclusive calculation of mm2. total mm2 > cut efficiency increase by 10-15% depending on the mode. > - the removal of brem photon changes and improves significantly the > reconstruction of the inclusive X mass (mxhadfit) > > Daniele