Skip to content

Retuning of cut-based inclusive dilepton trigger

Jamie Gooding requested to merge jagoodin-retuning-cutbased-dilepton into master

MR for the retuning of the cut-based inclusive dilepton trigger, as documented in #670 (closed), motivated in the talk in RTA WP3, 13th Nov 2023.

To-do:

  • Move builders to standard builders
  • Reimplement fake PID lines
  • Remove F.SIZE(pvs) > 0 workaround (MooreAnalysis#36 (moved))
  • Remove deprecated (original) lines
  • Update docstrings
  • Tune new lines
  • Redefine prescales

Performance on 2024 MC

The efficiencies and rates of the retuned line are posted below for reference. All results were produced from MC with expected-2024 conditions, applying HLT1 directly as either hlt1_pp_forward_then_matching or hlt1_pp_forward_then_matching_no_ut.

⚠ The numbers below are calculated prior to the rate-reducing commit https://gitlab.cern.ch/lhcb/Moore/-/pipelines/6960839 (discussed in !2747 (comment 7648919)). Performance thus differs slightly, though it is not expected that the efficiencies listed are different by more than a few percent.

Efficiencies

The following efficiencies are intended as a broader illustration of performance, by considering a few channels of interest for which 2024 conditions are available. All efficiencies are given with a HLT1 filtered denominator.

No UT

Channel Cut-based efficiency (master) / % Cut-based efficiency (retuned) / % MVA-based efficiency (master) / %
B+ -> (J/psi -> e+ e-) K+ 16.6 +/- 2.1 36.1 +/- 2.7 55.0 +/- 2.8
B+ -> (J/psi -> mu+ mu-) K+ 17.2 +/- 2.3 26.7 +/- 2.7 44.3 +/- 3.1
Bs0 -> mu+ mu- 30.3 +/- 3.7 45.2 +/- 4.0 61.9 +/- 3.9
Bs0 -> e+ e- 30.0 +/- 3.9 36.4 +/- 4.1 52.1 +/- 4.2
Bs0 -> (J/psi -> mu+ mu-) (Phi -> K+ K-) 12.7 +/- 1.9 32.1 +/- 2.7 52.9 +/- 2.8
Lb0 -> (Jpsi -> mu+ mu-) p+ K- 13.2 +/- 1.9 30.1 +/- 2.6 49.8 +/- 2.8
Incl. J/psi -> mu+ mu- 3.1 +/- 1.0 8.8 +/- 1.7 13.9 +/- 2.0
tau -> mu+ mu- mu+ 12.4 +/- 2.0 43.5 +/- 3.0 56.2 +/- 3.0

With UT

Efficiencies are effectively consistent with "no UT" scenario, table available upon request.

Rates

Rates calculated with HltEfficiencyChecker on 2.5 mn events of 2024 minbias MC. HLT1 was applied directly with the sequences "hlt1_pp_forward_then_matching_no_ut" and "hlt1_pp_forward_then_matching", with rates then calculated based on a 30 MHz input rate. The inclusive rate of the cut-based dilepton trigger in master and in this MR (and relevant triggers in master) are quoted below:

Name HLT2 rate (no UT) / kHz HLT2 rate (with UT) / kHz
Total HLT1 9119 +/- 13 8435 +/- 12
Master cut-based dilepton trigger (real lines) 0.333 +/- 0.092 0.333 +/- 0.092
!2747 (merged) cut-based dilepton trigger (real lines) 4.44 +/- 0.34 4.28 +/- 0.12
!2747 (merged) cut-based dilepton trigger (single fake lines) 0.59 +/- 0.12 0.59 +/- 0.12
!2747 (merged) cut-based dilepton trigger (double fake lines) 0.69 +/- 0.13 0.69 +/- 0.13
!2747 (merged) cut-based dilepton trigger (total) 5.72 +/- 0.38 5.56 +/- 0.38
Master MVA-based dilepton trigger (total) 8.67 +/- 0.47 8.46 +/- 0.47
Master b-topological trigger (total) 27.77 +/- 0.84 26.98 +/- 0.83

Conclusion: the rate sits at ~2/3 that of the MVA-based dilepton trigger.

Important note: the increase in rate with respect to the cut-based dilepton trigger in master is not because this retuning has a higher-than-intended rate, instead it is due to the rate of the cut-based dilepton trigger in master being significantly lower than anticipated upon initial merge (O(1 kHz)).

Edited by Jamie Gooding

Merge request reports