Jet issues in Higgs physics

Europe/Zurich
No room. Remote only

No room. Remote only

Bruce Mellado Garcia (University of Wisconsin (US)), Daniele Del Re (Universita e INFN, Roma I (IT)), Frank Tackmann (DESY), Gavin Phillip Salam (CERN)
Description
Underlying event effects and uncertainties in H+2j selection
 
 
Indico of the meeting:
https://indico.cern.ch/conferenceDisplay.py?confId=232206
 
Intro (Gavin Salam):
- Content:
   * intro on UE systematics
   * some histograms at MC level (Pythia) shown to check impact of
     switching on and off UE
   * BOTTOMLINE: switching UE on and off gives a ~10% effect for VBF
     selection cuts (with up to 20%  for 3rd jet selection).  
     Is it a fair estimate?
- Comments and questions:
    * are you using pythia where 2nd jet comes from showering? is it
     because you want to be conservative?  R: the reason for using 
     Pythia (which has the matrix element for the first jet but not 
     the second) is in part that Paolo Nason had mentioned the 
     existence of possible issues in the UE treatment when interfacing
     NLO codes such as MINLO through Les Houches Events (LHE).
     The difference between Pythia showers ("partons, no UE" column
     of table on p.6)  is not to be taken as a UE-related systematic.
     Torbjorn Sjostrand indicated at this point that he wishes to check a
     question of the distribution of collision impact parameter that is
     used with Les Houches events, to establish whether there are
     any potential issues with UE for LHE inputs.
 
 
UE and Higgs to gammagamma dijet analysis (CMS) (Paolo Meridiani):
- Content:
   * a summary on how UE systematics are estimated in CMS for the
     Hgg analysis and why
   * BOTTOMLINE: uncertainty about 30% (comparing delfaut Powheg 
     + Pythia6 Z2* and Powheg Pythia6 ProQ2 Tunes). Switching 
     on and off UE gives <5%  systematics (on Z2* Tune)
- Comments and questions:
   * two issues about tunes in forward region 1) energy flow more
     important than ntracks (which is not a very relevant quantity
     for jets) 2) the largest discrepancy is at low energies but maybe
     low energy is not very important. R: in the paper you can
     find more  plots which show that there is no tune working fine
     everywhere
   * can you redo the UE on and off of proQ20? R: yes, it can be done.
     please keep in mind that this 30% is not a limiting factor for
     Higgs analysis now (for instance for the extraction of Rv
     which has a 50% stat error)
   * how to interpret results in the plots at slide 5 and how to
     compare with ATLAS? R: there are two nuisance parameters: the
     first describes the global uncertainty of the sum of the two
     categories (correlated), which is about 30%, the other describes
     the category migration between the two dijet categories
     (anticorrelated) and it is 15% maximum.
   * did you try to vary the eigentunes of these tunes? R: no
   * how big is the statistical uncertainty in results at page 5? R:
     about 7%
 
Dag Gillberg (UE and Higgs dijet analysis (ATLAS)):
- Content:
   * a summary on how the UE systematics is estimated in ATLAS for the
     Hgg analysis and why
   * BOTTOMLINE: original uncertainty was about 30% (evaluated on 
     Perugia2011 Tune). Switching on and off UE gives 10-20% 
     systematics (20% for inclusive 2-jet selection). Other checks
     done to estimate showering uncertainty from scale variation 
     and UE uncertainty from PDF variation
- Comments and questions:
   * why reweighting based on deltaphi? R: just to evaluate the effect
     on reco quantities, we used the variable which is most affected
   * UE on-off looks a normalization-only effect? R: yes, it seems so,
     it is mainly a normalization effect, except for 3rd jet quantities
     (like deltaphi variable)
   * MC truth right? Do you have also the same numbers for full reco?
     R: we don't have the samples but we can give a try
   * pythia-Herwig difference. can you see the impact of UE turning on
     and off? R: yes, it is easy to do
  * correlation and bin migration, that should reduce the effective
     uncertainty. R: we are planing to have two nuisance parameters
   * need to think carefully about eigentunes to not underestimate
     uncertainties, also there is no tune that fits both central and
     forward region
 
General discussion:
   * different results in CMS and ATLAS when turning on and off deserve
     a deep understanding. maybe they come from the fact that different
     tunes are used as default
   * would be good to also monitor change PDFs for the different tuning
     to see interplay
   * real MPI to be discussed, i.e. when a dijet events overlaps to a
     Higgs events. no sure how big it is but needs to be studied
   * hadronization uncertainties? Part of that should be in Jet Energy
     Corrections. But it is not exactly the same as what gavin showed
     in his slides
   * so what is the prescription for summer conferences at the end?
     today's discussion not conclusive neither varying tunes (which
     tunes?) nor switching on and off seems to be ok.
     switching on and off looks reasonable but tunes should
     model correctly energy flow in the forward region (+-25%?)
     PROPOSAL TO START WITH: take 2 or 3 tuning which are not
     completely unreasonable in the forward region, evaluate difference
     in efficiency turning on and off UE and take the largest difference
There are minutes attached to this event. Show them.
    • 10:00 10:20
      Intro 20m
      Speaker: Gavin Phillip Salam (CERN)
      Slides
    • 10:20 10:40
      UE and Higgs to gammagamma dijet analysis (CMS) 20m
      Speaker: Paolo Meridiani (Universita e INFN, Roma I (IT))
      Slides
    • 10:40 11:00
      UE and Higgs dijet analysis (ATLAS) 20m
      Speaker: Dag Gillberg (CERN)
      Slides
    • 11:00 11:30
      Discussion 30m