Draft: Bug fix for BandQ, B2CC/BnoC dimuon builder: lifetime resolution bias
Test result about Jpsi efficiencies before and after the bug fix
Tested using the MR status in 2024/April/30, 15:01 CEST
Test Sample
Sample 1 : Inclusive Jpsi->MuMu, containing about 80% prompt Jpsi and 20% from-b Jpsi.
sim+std://MC/Dev/Beam6800GeV-expected-2024-MagDown-Nu7.6-25ns-Pythia8/Sim10c/24142001/DIGI
root://gridproxy@lhcbdcache-kit.gridka.de:1094//pnfs/gridka.de/lhcb/LHCb-Disk/lhcb/MC/Dev/DIGI/00212728/0000/00212728_00000099_1.digi
Sample 2: exclusive B->JpsiK sample
sim+std://MC/Dev/Beam6800GeV-expected-2024-MagDown-Nu7.6-25ns-Pythia8/Sim10c/12143001/DIGI
Test result using 2024-patches (note: the full chain of hlt1->hlt2->sprucing is done. For 2024-patches and this-MR test, the same input is used for hlt1 option file. One can just look at the final yield for comparison, do not need to check the efficiency especially for sprucing, as the hlt2 is different for these two sectors.)
Inclusive Jpsi->MuMu:
LAZY_AND: SpruceBandQ_JpsiToMuMuDetached #=212 Sum=110 Eff=|( 51.88679 +- 3.43157 )%|
LAZY_AND: SpruceBandQ_JpsiToMuMuTightPrompt #=212 Sum=116 Eff=|( 54.71698 +- 3.41870 )%|
LAZY_AND: Hlt2_JpsiToMuMuDetachedFull #=720 Sum=125 Eff=|( 17.36111 +- 1.41161 )%|
LAZY_AND: Hlt2_JpsiToMuMu #=720 Sum=322 Eff=|( 44.72222 +- 1.85298 )%|
LAZY_AND: Hlt2_DiMuonJPsiTightFull #=720 Sum=140 Eff=|( 19.44444 +- 1.47496 )%|exclusive B->JpsiK
LAZY_AND: SpruceBandQ_JpsiToMuMuDetached #=547 Sum=450 Eff=|( 82.26691 +- 1.63309 )%|
LAZY_AND: SpruceBandQ_JpsiToMuMuTightPrompt #=547 Sum=234 Eff=|( 42.77879 +- 2.11543 )%|
LAZY_AND: Hlt2_JpsiToMuMuDetachedFull #=914 Sum=529 Eff=|( 57.87746 +- 1.63320 )%|
LAZY_AND: Hlt2_JpsiToMuMu #=914 Sum=578 Eff=|( 63.23851 +- 1.59483 )%|
LAZY_AND: Hlt2_DiMuonJPsiTightFull #=914 Sum=272 Eff=|( 29.75930 +- 1.51228 )%|Test result using this MR's update (note: the full chain of hlt1->hlt2->sprucing is done. For 2024-patches and this-MR test, the same input is used for hlt1 option file. One can just look at the final yield for comparison, do not need to check the efficiency especially for sprucing, as the hlt2 is different for these two sectors.)
Inclusive Jpsi->MuMu:
LAZY_AND: SpruceBandQ_JpsiToMuMuDetached #=301 Sum=110 Eff=|( 36.54485 +- 2.77564 )%|
LAZY_AND: SpruceBandQ_JpsiToMuMuTightPrompt #=301 Sum=191 Eff=|( 63.45515 +- 2.77564 )%|
LAZY_AND: Hlt2_JpsiToMuMuDetachedFull #=720 Sum=125 Eff=|( 17.36111 +- 1.41161 )%|
LAZY_AND: Hlt2_JpsiToMuMu #=720 Sum=517 Eff=|( 71.80556 +- 1.67685 )%|
LAZY_AND: Hlt2_DiMuonJPsiTightFull #=720 Sum=228 Eff=|( 31.66667 +- 1.73361 )%|exclusive B->JpsiK
LAZY_AND: SpruceBandQ_JpsiToMuMuDetached #=556 Sum=450 Eff=|( 80.93525 +- 1.66589 )%|
LAZY_AND: SpruceBandQ_JpsiToMuMuTightPrompt #=556 Sum=242 Eff=|( 43.52518 +- 2.10262 )%|
LAZY_AND: Hlt2_JpsiToMuMuDetachedFull #=914 Sum=529 Eff=|( 57.87746 +- 1.63320 )%|
LAZY_AND: Hlt2_JpsiToMuMu #=914 Sum=596 Eff=|( 65.20788 +- 1.57550 )%|
LAZY_AND: Hlt2_DiMuonJPsiTightFull #=914 Sum=281 Eff=|( 30.74398 +- 1.52629 )%|Historical message
This is a draft MR for discussion. Probably a bug fix.
When investigating the early 2024 MC, I noticed a lifetime asymmetry for BandQ upsilon sample. One can reproduce the result using the output of https://lhcb-analysis-productions.web.cern.ch/productions/?wg=bandq&analysis=quarkoniatomumu2024&ver=v0r0p7182118&dset=inclusive_upsilon1s_mc_24_nu7_6_magdown_veloclosed_root , and do DecayTree->Draw("QQbar_END_VZ - QQbar_BPVZ>>h100(100,-2,2)", "QQbar_TRUEID==553").
But if we do it in RD Upsilon output (Many thanks to @rquaglia for suggestion of including RD line in AP !) the lifetime bias disappear. One can check it in https://lhcb-analysis-productions.web.cern.ch/productions/?wg=bandq&analysis=quarkoniatomumu2024&ver=v0r0p7182118&dset=inclusive_upsilon1s_rd_mc_24_nu7_6_magdown_veloclosed_root , and do DecayTree->Draw("QQbar_END_VZ - QQbar_BPVZ>>h100(100,-2,2)", "QQbar_TRUEID==553").
One can find the result by doing DecayTree->Draw("QQbar_BPVDLS>>h100(100,-5,5)","QQbar_TRUEID==553") for both sample. The BPVDLS negative part is missing. This is due to the cut here .
I confirm BandQ lines are generally influenced, including the early measurement lines. The tz-fit for prompt/from-b separation will be more challenging. c.c. to @lcapriot @zexu .
I suspect that B2CC lines are also influenced. The make_jpsi() at https://gitlab.cern.ch/lhcb/Moore/-/blob/master/Hlt/Hlt2Conf/python/Hlt2Conf/lines/b_to_charmonia/b_to_jpsix.py?ref_type=heads#L17 is taken from our shared directory. It would be appreciated if @oozcelik @yimingli can kindly have a look.
If both WGs are influenced, need a greenlight from conveners from both WGs before we move to the standard RTA/DPA review. From B&Q side, I will do a local test of hlt1->hlt2->sprucing->tuple to see if this MR really fixes the lifetime resolution bias issue. If not we just close it.
If B2CC is also influenced, it would be super appreciated an independent local cross-check can be made. We have a few quarkonia->mumu and b2psiX MC samples available at: /MC/expected-2024.Q1.2/Beam6800GeV-expected-2024.Q1.2-MagDown-Nu7.6-25ns-Pythia8/Sim10c
Not sure which labels to add. I will think about it after finishing the local test tomorrow. Necessary ones to the best of my knowledge are added.