Skip to content

Draft: Charm changes for start of 2025 data-taking

Carlos Eduardo Cocha Toapaxi requested to merge charm_2025_reopt into master

Depends on lhcb-datapkg/ParamFiles!123

Goals

MR collecting all modifications to HLT2 lines aiming to reduce the bandwidth/rate consumption detailed in #876

Current status

Commit Bandwidth [GB/s] Reduction
Master 1.24 -
0cc6d7a2 1.13 8.8%
3f8cec51 1.07 13.7%
7bbbc326 1.06 14.5%
2e443765 1.03 16.9%
e9620f3f 1.03 16.9%
1aa50c92 1.01 18.5%
06457f94 0.91 26.6%
Local test - -
50db8e95 0.83 33.0%
9c148637 0.7867 36.5%

Contributing

If you want to contribute to this MR, please open a new MR and target the charm_2025_reopt branch and explain your goals.

All the additions need to be tested locally, including:

  • Impact on the bandwidth and rate: previous and new rates and bandwidths, respectively. This must be tested using the HLT1 filtered (1.3 MHz) mdfs available at /eos/lhcb/wg/rta/WP3/mdfs_hlt1filtered_307586. To do so, it is recommended to use this snippet.
  • In case of the usage of PersistReco, it has to be well justified.
  • (optional) Impact on the throughput: comparing the time with the execution of master with no extra modifications.
  • (optional) Determine efficiencies: if there is simulation already available, it will be interesting to know the efficiencies so we ensure that there is no bug in the selections.

For lines with exclusive retention rate of 0 would be worth it to check the overlap with other lines.

Do not rebase or force-push to this branch without the consent of the liaisons @lpica @aanelli and @cacochat, since this might lead to losses on the changes

Changes

Existing modules/lines

hexaquarks @ipolyako !4302 (merged)

  • replace PID cuts with PROBNN
  • split all lines into RS and WS, put additional prescale of 0.5 of those WS lines with more than 1 decay mode
  • tighten LTIME and removed prescale in Ps/Hs/Hss lines
  • tighten PROBNN and (sum)PT cuts for PsToPpPimPim line (probably 20% of BW)
  • tighten selection on PT and VCHI2 for Ks and Lambda on top of those in standart_particles
  • tighten cuts on (sum)PT and IPCHI2 on all exclusive Hq lines
  • for detached pp and LamP lines: replace full PersistReco with saving only Long(Downstream)Protos and Long(Downstream)Tracks, relax IPCHI2 < 100 cut given that's its only partially reconstructed thing

d0_to_hh mu-tagged modes (and related D0 builders) @fbetti !4456 (merged)

  • increase PID_K cut of kaons from 0 to 2
  • increase MINIPCHI2CUT cut on D0 children from 9 to 20
  • tighten the mass window of D0 from 1715-2015 to 1780-1950 MeV (and similar in CombinationCut)
  • tighten D_END_VZ - B_END_VZ from -3 to -1.2 mm

ccbaryon_to_hyperon_dh @ziyiw @msaur !4490 (merged)

  • Tighten the PT for proton and D-meson
  • Apply lifetime and VDRHO cut on D-meson
  • Add Xim_END_VZ - Omcc/Xicc_END_VZ cut
  • Enlarge the right mass windows of cc-baryons

hyperonTT @sjaimese @ziyiw !4487 (merged)

  • Tighten the cuts for pion from Xi and lambda_tt
  • Tighten the CHI2DOF and DIRA for xicp_to_ximpipi decays
  • Disable three lines for Xim(TTD) with downstream pions/kaons(Xic0ToXimPip_TTD_D, XicpToXimPipPip_TTD_DD, Xic0ToXimPimPipPip_TTD_DDD)

charm_toh0x module @thadaviz !4362 (merged)

  • Add Hlt1 TIS and TOS filter
  • Add Hlt1 TOS filter
  • Add neutral isolation particles

d0_tokshh module @mamartin !4466 (merged)

  • Add Hlt1 TIS and TOS filter
  • Add KS selection on ∆z between KS and D0 candidates
  • Changed cut on D0 LT from 0.1 ps to 0.2 * 0.410 ps = 0.820 ps for the default lines

cmeson_to_sl_btag @xiaokali @msaur !4250 (merged)

  • Replaced the D-meson and B-meson combiners, optimized the code structure.
  • Add lifetime cut F.OWNPVLTIME > 0.2 * ps for D-meson and B-meson. And add F.OWNPVIPCHI2 < 15 for B-mesons.
  • All BPV... funcotors are replaced by OWNPV... version.
  • Replaced names of lines with Signal_line_WS(RS) and ensure it fully correspond to the decay descriptor of the input particles.
  • Included final mass cut at CompositeCut. Tighten the mass window of D-mesons and B-mesons for semi-leptonic lines.
  • Add F.SDOCA(2, 3) < 0.15 * mm, for B-mesons. Changed MAXSDOCACUT from 2 * mm to 0.15 * mm.
  • Change the DIRA cut from 0.95 to 0.99.
  • Changed the OWNPVVDZ cut from 1.5 * mm to 0.5 * mm.

dst_to_dee @cacochat !4493 (merged)

  • Apply a DOCA cut on the dielectron maker.
  • Add a momentum asymetry cut on the dielectrons tracks to remove clones (?)
  • Add HLT1 TOS filter (under study)
  • Prescale SS lines (under study)

d_to_etah @gtuci @tpajero !4306 (comment 9234613)

  • Rewrite decay descriptors to enable reconstructing the eta(') directly
  • Add HLT1 TOS filter
  • Tighten D(s) momentum cut

d_to_hhh @guloacha !4495 (comment 214786)

  • Switch the DIRA cut from DIRA>cos(14.1) to DIRA>cos(10) in all the lines.
  • Apply (Hlt1TrackMVA_TOS||Hlt1TwoTrackMVA_TOS)|| (Hlt1TrackMVA_TIS||Hlt1TwoTrackMVA_TIS) filter to all lines.
  • The lines Hlt2Charm_DpDspToKmKpPip_NoCuts and Hlt2Charm_DpDspToKmPipPip_NoCuts are removed.

d0_to_hhpi0 @freiss !4636 (merged)

  • remove vertex chi2/ndof cut on neutral-neutral combination, which rejected all events

d0_t0_hhgamma, d_to_hhhgamma, d0_to_hhpi0 @gtuci !4566 (merged)

  • Require D0 to be TOS on Hlt1TrackMVA OR Hlt1TwoTrackMVA

cbaryon_spectroscopy, cbaryon_spectroscopy_mva, ccbaryon_hadronic @cacochat !4545 (merged)

  • Disable persistreco and instead save rawbanks

prod_xsec !4275 (comment 9258475)

  • adding back the double charm lines only

d0_to_ksks @gtuci !4479 (merged)

  • Apply TOS filter on D0
  • Remove "loose" version of lines
  • Tighten the requirement of the vertex chi2 for the UL Ks
  • Remove prescale for DDDD and DDLD lines

cbaryon_to_ph0_btag @yingl @msaur !4576 (comment 9267597)

  • Apply TOS filter on lb
  • Tight the DIRA

cbaryon_to_sl @yingl @msaur !4253 (comment 9278363)

  • Optimzed the code structure. Add the cbaryon and bbaryon combiner.
  • Add lifetime cut F.OWNPVLTIME > 0.2 * ps.
  • All BPV... funcotors are replaced by OWNPV... version.
  • Tighten the mass window of baryon.
  • Change the DIRA cut from 0.95 to 0.99.
  • Unify the selection for DD line
  • Add the F.CHILD(i, F.END_VZ) - F.END_VZ

cbaryon_spectroscopy_mva @shuyu !4382 (merged)

  • Use MVA trained with 2024 data to reduce bandwidth for Lcp, Xicp and Xic0.
  • All BPV functors are changed to OWNPV.

New lines/modules

d0_t0_kstmunu @icelesti !4468 (merged)

  • New HLt2 line collecting the decay chain [D*(2010)+ -> (D0 -> (K*-(892) -> (KS0 -> pi+ pi-) pi-) mu+ nu) pi+]cc, with both long-long (LL) and downstream-downstream (DD) neutral kaons.
  • Physics motivation in this presentation: https://indico.cern.ch/event/1487741/contributions/6335563/attachments/3000600/5287572/Presentation_WGHLT2_22_01_2025.pdf
  • All BPV functors moved to OWNPV
  • Require D*+ to be TOS on (Hlt1TrackMVA OR Hlt1TwoTrackMVA OR Hlt1TrackMuonMVA OR Hlt1KsLines), with Hlt1KsLines being Hlt1TwoTrackKsDecision for the LL line and Hlt1DownstreamTwoTrackKsDecision OR Hlt1DownstreamKsToPiPiDecision for the DD line

lambdac_to_pgammaconv @hartmanm !4544 (merged)

  • New HLt2 line collecting the decay chain Lambda_c+ -> p+ gamma(e- e+) cc, with both long-long (LL) and downstream-downstream (DD) electrons. It is inspired by the Hlt2Charm_LcpToPpEmEp line.
  • Used a new charm particle maker make_rarecharm_lambdacs2pee_looselambdamass which is inspired by make_rarecharm_lambdacs2pel: Wider Lambda_c mass window (+/- 150 MeV) than Hlt2Charm_LcpToPpEmEp line (+/- 50 MeV), to be able to model contributions from the Lambda_c -> p pi0 in data. Tighter cut on bpvdira_min=0.99998 (instead of 0.99996) bpvipchi2_max=10 (instead of 25).
  • Filters on the following HLT1 lines: Hlt1TrackMVADecision, Hlt1TwoTrackMVADecision, Hlt1TrackElectronMVADecision, Hlt1DiElectronLowMass_DisplacedDecision, Hlt1DownstreamGammaToEEDecision.
Edited by Marie Hartmann

Merge request reports