B&Q: adjust rates of prompt D0 anti-D0 B&Q double charm lines

This MR targets the branch bandq_EoY_sprucing to be merged in !3839 (merged) and depends on !3896 (merged) (there was also previously !3878 (closed)). This branch is frequently updated with all changes made in bandq_EoY_sprucing and 2024-patches.

Some issues with some/all of these lines were discussed in #840

• Steps were taken to reduce HLT2 rate for these lines in !3883 (merged)
• !3870 (merged) also helped

Related Analysis Productions:

• Double Charm with both D0 To HH lhcb-datapkg/AnalysisProductions!1771 (merged)
• Double Charm with all other D0 decay combinations lhcb-datapkg/AnalysisProductions!1905 (merged)

Things to understand:

• Vertex fit chi2 cut in HLT2 of pressing concern --- possibly for efficiency calculation --- but may also affect our background distribution, since some of our background may be 2 D0s from different closely spaced PVs, see arxiv:1205.0975
  • In any case, we think that we can just use the main spectroscopy lines to model closely spaced PV background for most of the multi-body D0 lines. Perhaps there is also enough data in AP pipeline test tuples to have a look at this.

To-do List For This MR

• Remove sprucing prescales on D pairs involving at least one multi-body decay, so that we can be aware of our un-prescaled rates. (Cuts in HLT2 have reduced our rates by several factors, so prescales may no longer be needed in the sprucing.
• Except for D0ToHH which does not have a mass cut, reduce upper neutral D pair invariant mass limit to 5000 MeV (from 6000 MeV).
• Determining where are we now, in the landscape of HLT2 rate and Sprucing rate, by running over any available \mu = 5.3 samples
  • For Sprucing (see below)
  • For HLT2
    • A bit of a challenge as cannot configure HLTEfficiencyChecker (https://mattermost.web.cern.ch/lhcb/pl/qhk5hiaqkt8utprhbjj7erkzry)
    • HLT2 rate test instead performed with past samples
• removal of vertexChi2dof cut for D0_To_HH line, increase to <30 (from <8) for all other D0D0 lines
  • cut of <8 removes signal which would pass vertexChi2dof<5 after DTF. There are some stray events above 30 but probably background.
• removal of IPchi2 cut for D0_To_HH line, increase to <30 (from <16) for all other D0D0 lines
  • cut of <16 removes signal which would pass vertexChi2dof<5 after DTF. There are some stray events above 30 but probably background.

Sprucing Retentions

• test with ~223k (mu=5.3) events using https://gitlab.cern.ch/lhcb/Moore/-/blob/c112bcc711a0b28464beaa319b1e3c99232087fe/Hlt/Hlt2Conf/options/examples/bandq/spruce_bandq_example.py (provided by Yajing)
  • The input is HLT2 full stream data. Its rate is 45.1kHz.
  • test with sprucing prescales removed and with HLT2 <5000 MeV charmonia mass cut, vertexchi2dof and IPchi2 both loosened to < 30 on non-D0ToHH lines and without vertex chi^2 and IP chi^2 cut (matching Same/Opposite sign lines) on D0ToHH line
  • The mass cut appears to do something for the non D0ToHH-only lines, which is good. However the lack of effect of the IP chi^2 cut removal on the D0ToHH line (And loosening on the other lines) makes me think that this sample may already have an IP chi^2 cut (& vertex chi2dof cut) applied in HLT2 (would be the case for mu=5.3 data).

     LAZY_AND: SpruceBandQ_DoubleCharmOppositeSignDecisionWithOutput                   #=223233  Sum=226         Eff=|(0.1012395 +- 0.00673094)%|
     LAZY_AND: SpruceBandQ_DoubleCharmSameSignDecisionWithOutput                       #=223233  Sum=338         Eff=|(0.1514113 +- 0.00822945)%|
     LAZY_AND: SpruceBandQ_DoubleCharm_D0ToHHDecisionWithOutput                        #=223233  Sum=134         Eff=|(0.06002697 +- 0.00518398)%|
     LAZY_AND: SpruceBandQ_DoubleCharm_D0ToHHHHDecisionWithOutput                      #=223233  Sum=60          Eff=|(0.02687775 +- 0.00346944)%|
     LAZY_AND: SpruceBandQ_DoubleCharm_D0ToHHHH_D0ToHHDecisionWithOutput               #=223233  Sum=73          Eff=|(0.03270126 +- 0.00382677)%|
     LAZY_AND: SpruceBandQ_DoubleCharm_D0ToHHHH_D0ToKsDDHHDecisionWithOutput           #=223233  Sum=23          Eff=|(0.01030314 +- 0.00214824)%|
     LAZY_AND: SpruceBandQ_DoubleCharm_D0ToHHHH_D0ToKsLLHHDecisionWithOutput           #=223233  Sum=61          Eff=|(0.02732571 +- 0.00349822)%|
     LAZY_AND: SpruceBandQ_DoubleCharm_D0ToKsDDHHDecisionWithOutput                    #=223233  Sum=30          Eff=|(0.01343887 +- 0.00245343)%|
     LAZY_AND: SpruceBandQ_DoubleCharm_D0ToKsDDHH_D0ToHHDecisionWithOutput             #=223233  Sum=70          Eff=|(0.03135737 +- 0.00374734)%|
     LAZY_AND: SpruceBandQ_DoubleCharm_D0ToKsLLHHDecisionWithOutput                    #=223233  Sum=27          Eff=|(0.01209499 +- 0.00232754)%|
     LAZY_AND: SpruceBandQ_DoubleCharm_D0ToKsLLHH_D0ToHHDecisionWithOutput             #=223233  Sum=50          Eff=|(0.02239812 +- 0.00316722)%|
     LAZY_AND: SpruceBandQ_DoubleCharm_D0ToKsLLHH_D0ToKsDDHHDecisionWithOutput         #=223233  Sum=30          Eff=|(0.01343887 +- 0.00245343)%|

Postponed

• Is it OK to remain in FULL/persistreco=True or do we need to or should we move to selective persistence
  • if we did, we would still want any kaon/pion/photon from the PV for possibly full reco of DDbarpi0/gamma and investigation of (cone) isolation variables which may have some use in busy events.
  • I couldn't find a discussion of this in LHCB-ANA-2018-042, but I wonder if we chose not to persist anything at all we could reject situations where instead of D0anti-D0(X) resonance where X is neutral one might have a D0Dst+ resonance decaying to D0(D0pi+), and then by not reconstructing the pi+ if there were some sort of X(3872)+ decaying to D0Dst if we would have a way to control this background.
    • We need to reject an approximately 5% T_{cc}^+ background, by rejecting D^0D^+\gamma D^0D^+\pi^0 and D^0D^0\pi^+ [should be able to achieve a >80% rejection of this background]