Run 13 W Analysis : Final Results : Embedding QA

Embedding QA - SL16g_embed

7/05/17 - updates
Check on Geant level distributions / Particle ID (pid_ID)

***The GEANT level information are not filled in the W trees.  Why ?
Looking at the code => 

How we get geant info form MuDsts..
 

StMcEventmMcEvent = 0;                                                                                                             
mMcEvent = (StMcEvent*) StMaker::GetChain()->GetDataSet("StMcEvent"); 
    StMcVertex *V=mMcEvent->primaryVertex();                                                                             
    mVertex=TVector3(V->position().x(),V->position().y(),V->position().z());

        StMcTrackmcTrack = mMcEvent->tracks()[i];                                                                               
        int pdgId=mcTrack->pdgId();

           if(pdgId==11 || pdgId==-11){ //select e+ and e-
              if(mcTrack->parent() && abs(mcTrack->parent()->pdgId()) == 24 ){   
                 pElectron=mcTrack->momentum();

                 pW=mcTrack->parent()->momentum();

                 eW=mcTrack->parent()->energy();


Once



[We don't have StMcEvent in our MuDsts, what we have is StMuMcEvent. StMuMcEvent is same as StMcEvent. However, StMuMcEvent does not have all the details that are in StMcEvent. So from where we get "pig_id"s. From geant.root ?  ]



7/05/17 - updates
MC W analysis Comparison between the SL16g_embed vs SL16f (Larger sample - 296 runs)

1) Cuts vs Statistics

No enhancement of (~2%) as expected in track reconstruction. In fact a decrees in ~1% can be seen both in primary track reconstruction and in final Ws. 





2) Final W ET / Sign PT distributions






3) Final W Eta Distribution





6/28/17

Jinlong's QA page:  https://drupal.star.bnl.gov/STAR/blog/jlzhang/run13-w-embedding-reproduction 

First, the code was checked to make sure the various parameters (such as PYTHIA parameters, geometry tags, BFC chain options, timestamps, library) are correct. Then the parameters were checked between two codes used for the new production and preliminary productions. Log files were checks to see where proper tracking parameters / status/ calibration  tables were grabbed during the productions. 
Codes / Logs Checking  

Parameter SL16g_embed SL16f same /diff
PYTHIA version*** 6.4.28 6.4.22 diff
PYTHIA (part) library y2013_1c    y2013_1c  same
PYTHIA tune Perugia 0 Perugia 0** **
embedding (part) library (geometry ) ry2013_1c  ry2013_1c same
Vz_sigma 47 cm 42 cm Diff
Track cuts PPV_W PPV_W same
Timestamp  checked checked correct 
BFC chain option  checked (STICA) checked (STICA) same
BEMC gains Run 9 Run 9 same*
 Beamline constraint   used  used same 
       

=>*** PYTHIA version is different (possible changes in the TUNE (pdfs), potentially could impact the event generation process / decay behavior)  ===> Need Investigation!!!!!!!

=> ** Possible changes in the tune due to the updates in the version.

=>* As long as same gains are used for the MC-embedding muDst production and for the muDst reading no impact of calibration gains would occurs.



MC W analysis Comparison between the SL16g_embed vs SL16f (small sample - 33 runs)
33 MC runs that were produced and analyzed using W codes were used. New set =>SL16g_embed, Preliminary results production => SL16f

=>SL16g library STICA is expected to have more reconstructed tracks (~2%) than SL16f.

=>Comparison plots seems all good. For a such small sample the difference in the shape of the ET distribution of final W looks quite large though. New one looks bit weird. I confirmed both samples were analyzed using the same corresponding gains that were used for the MC muDst production.

=>Did Jinlong Use the same W cuts as preliminary version ????????????

Apple to Apple Comparison Plots 


1) Cuts vs Statistics

As expected roughly 1-2% enhancement in tracks can be seen. [not much of a increase in golden W, but this is a very small sample)
Enhancement in track Pt cut expect to continues within other cuts. However the drop in Tr2Cl (Track cluster matching) / or in eta1, and noAway (signPt) is not expected. [We have seen a drop before noNear ( nearCone) cut due to the reduction of more Background)  This is problematic. Need investigation !!!



2) Final W ET / Sign PT distributions

The ET distribution (shape) of the new sample is quite different from the old one. For a small sample I would hardly expect a significant difference between them. And the enhancement at low ET is quite worrisome.  => Causes form changes in the PYTHIA version ????




3) Final W Eta Distribution

Looks OK!





4) Charge-sign Distribution :



Difference Between PYTHIA versions (6.4.22 vs 6.4.28)
PERUGIA Tunes:

* The starting point for all the Perugia tunes, apart from Perugia NOCR, was S0A-Pro, i.e., the original tune “S0”, with the Tune A energy scaling (S0A), revamped to include the Professor tuning of flavour and fragmentation parameters to LEP data (S0A-Pro).


PERUGIA 0  (320) [this is what we use.]

* Uses CTEQ5L parton distributions.
* Uses (Lambda) 
CMW instead of MS(bar) scheme, which results in near-perfect agreement with the Drell-Yan pspectrum, both in the tail and in the peak.

* Has slightly less colour reconnections than S0(A), especially among high-pstring pieces, which improves the agreement both with the hpi distribution and with the high-ptail of charged particle pspectra.
* Slightly more beamremnant breakup than S0(A) (more baryon number transport), mostly in order to explore this possibility than due to any necessity of tuning at this point. Without further changes, these modifications would lead to a greatly increased average multiplicity as well as larger multiplicity fluctuations. To keep the total multiplicity unchanged, relative to S0A-Pro, the changes above were accompanied by an increase in the MPI infrared cutoff, p⊥0, which decreases the overall MPI-associated activity, and by a slightly smoother proton mass profile, which decreases the 
fluctuations.
* The energy scaling is closer to that of Tune A (and S0A) than to the old default scaling that was used for S0.



Descriptions of 6.4.22 vs 6.4.28 in the updates

6.4.22:

6.4.22 : 11 November 2009
  - Bug fixes in PYPTIS for massive quark evolution close to threshold
    in annihilation-type processes, for the special case where both
    incoming massive quarks get so close to threshold that creation
    has to be forced. Previously, the algorithm would then always pick
    the Q on side 1 to start with, thus possibly creating an artificial
    asymmetry. This choice has now been randomized to alternate
    between sides 1 and 2. Also, when performing the creation for the
    first Q, the algorithm did not check explicitly whether enough
    phase space was still left for creation of the second Q. This is
    now ensured as well. Thanks to H. Hoeth and S. Kama for pointing
    to issues that led to the identification of this bug. 
  - Change in PYINPR for MSEL = 39, SUSY production. The manual states 
    that MSEL = 39 switches on all SUSY production processes except Higgs
    production. However, previously MSEL = 39 would switch on the SUSY Higgs
    pair production processes 297-301. The meaning of MSEL = 39 has now been
    changed to correspond better to what is written in the manual, i.e., 
    pair production of states with R-parity = -1, leaving out processes 
    297-301. Thanks to M. Johansen for pointing this out. 
  - Protected PYPTFS from compiler-dependent behavior regarding the
    logical evaluation of an IF statement. Noticed by S. Kama. 
  - Enforced initialization of NJN in PYPTIS and of NCHN in
    PYPTMI. Added SAVE statement for NCHN in PYPTMI. Should prevent
    problems experienced with some compilers, notably
    gfortran. Noticed by S. Kama and E. Ozcan.
  - Bug fix in PYEVNW to address problems with how particles point
    back to their ancestors in the pT-ordered shower when
    MSTP(125)<=1. In the new shower, each time a parton acts as a
    "recoiler", it is saved as a new copy with modified
    momentum. Previously, when using the options MSTP(125)<=1, Pythia
    would attempt to compress this history down to a single parton,
    but this gave rise to some inconsistent mother-daughter pointers
    which in turn resulted in some problems in interpreting the event
    history reported by users of the new framework. The "fix" is to
    give up on saving space and instead retain each of the recoiler
    copies separately, without compressing them down to a single
    parton. For MSTP(125)=2 there should be no changes, since this
    option is equivalent to keeping the entire branching history
    without compression.  
  - Updated PYLOGO, Skands now at CERN.
6.4.28:

6.4.28 : 05 Sep 2013
  - Bug fix in the handling of final-state radiation in decays into a
    gluon, specifically q* -> q g, where an incorrect handling of the
    matrix element matching leads to a strong suppression of radiation
    from the gluon. 
  - Bug fix in PYUPEV to handle certain processes, such as VV->VV
    scattering (MSUB=71-77), where a mother index would become negative.
  - Included four new Perugia 2012 variations, one of which is
    motivated by theory uncertainties on min-bias, two others having
    been requested by the LHC experiments, and the last exploring a
    different set of fragmentation parameters, obtained by the
    Innsbruck group. Also included 9 new complete ee+pp tunes provided 
    by G. Rudolph and N. Firdous (Innsbruck), 3 of which use LO PDFs,
    3 using NLO ones, and 3 LO* ones. Note that we do not recommend
    using NLO or LO* PDFs for serious soft-physics studies, but include 
    them nonetheless to illustrate the results that one obtains with them. 
    MSTP(5) = 
      380 P12-MB2     : same as Perugia 2012, with PARP(87)=0D0    (2013)
      381 P12-ueHi    : Variation with higher UE (lower pT0)       (2013)
      382 P12-ueLo    : Variation with lower UE (higher pT0)       (2013)
      383 P12-IBK     : Perugia 2012 with IBK ee fragmentation parameters
      390 IBK-CTEQ5L  : Innsbruck ee+pp tune with CTEQ5 LO PDFs    (2013)
      391 IBK-CTEQ6LL :       with CTEQ6LL LO PDFs
      392 IBK-MSTW08LO:       with MSTW08 LO PDFS
      393 IBK-CTEQ66NLO:      with CTEQ66 NLO PDFs
      394 IBK-CT10NLO :       with CT10 NLO PDFs
      395 IBK-MSTW08NLO:      with MSTW08 NLO PDFs
      396 IBK-MSTW08LO*:      with MSTW07 LO* PDFs
      397 IBK-MRSTLO**:       with MRSTMCal (LO**) PDFs
      398 IBK-CT09MC2 :       with CTEQ09MC2 PDFs  
    The variation with PARP(87)=0 (see manual) has a slightly improved 
    behaviour at very low minimum-bias multiplicities. The new P12 UE 
    variations are complementary to the existing mpiHi variation (373). 
    Where the latter increases the amount of semi-hard MPI scatterings 
    (i.e., the amount of MPI minijets) by increasing the alphaS associated 
    with MPI, the new variations adjust the soft pT0 scale and hence affect 
    the amount of soft MPI produced, without changing the rate of hard MPI. 
    The energy-scaling of the new variations has also been chosen
    conservatively, so that the activity of the "ueLo" variation scales 
    slower with CM energy (i.e., pT0 scales up faster) than for "ueHi". 
  - Extended PYSLHA to be able to handle more than 25 new particles
    introduced via QNUMBERS blocks in SLHA files and/or LHEF
    headers. By recycling previously unused space in the particle
    arrays, the maximum number of QNUMBERS blocks that can be stored
    is now at 130 new particles, at which point the code will advise
    you to move to PYTHIA 8 instead. Thanks to A. Belyaev for
    pointing to this problem and for help testing the updated PYSHLA.  
  - Modification of PYPTFS for the case of interleaved showers, where
    final state showers occur after initial state showers. In
    the previous version, a soft parton from ISR could form a dipole
    with a final state parton, leading to a suppression of hard final-
    state radiation. In the corrected version, a final-final dipole is 
    selected, using other hard partons in the final state, unless
    the pT-scale is at or below the scale of the original initial-final
    dipole. Thanks to Gavin Salam for noticing a problem in jet
    masses for certain kinematic configurations. See also the study
    reported in section 8.4 of Dasgupta et al., arXiv:1307.0007. 
    Note that this correction slightly increases the average amount of
    FSR in all tunes of the pT-ordered shower. Thus, for instance, the 
    average multiplicity of minimum-bias events at LHC (7 TeV) increases 
    by 1 - 2 per cent in this version relative to previous PYTHIA 6 versions. 
  - Identified a bug in final states including exactly one vector boson
    and one quark. This could be a native Pythia process or an
    external LHE event. In this case, a matrix element correction is
    applied to FSR assuming the decay q -> q V.   The effect appears
    to be minimal, and no solution is implemented at this time.

Documentations 
PYTHIA Tune : PERUGIA Tunes (Click here)
PYTHIA 6.4 updates (Click here)
PARP: (Click here)