CosmicsAnalyzer

Instructions to run the code on lxplus

cmsrel CMSSW_13_0_10
cd CMSSW_13_0_10/src/
cmsenv
git clone -b main git@github.com:tvami/CosmicsAnalyzer.git
scram b -j
cd CosmicsAnalyzer/EarthAsDMAnalyzer/test/
// just an example file
xrdcp root://cms-xrd-global.cern.ch//store/data/Run2023D/Cosmics/RAW-RECO/CosmicSP-PromptReco-v1/000/369/811/00000/2ad63d9f-234b-4c68-8c18-49d485d42bc7.root .
cmsRun muon_analyzer_cfg.py

Ntuplizer versions

The ntuplizer is tagged in git (git tag). Each version adds branches on top of the previous one; older ntuples remain readable, they simply lack the newer branches.

Version	Date	Main additions
v2	2025-11-11	More general-track info, and the HLT_Random trigger (PR #17)
v3	2025-11-12	Extra info on the muon-to-general-track match (muon_fromGenTrack_*), removed the R-squared variable (PR #18)
v4	2026-02-13	Number of valid tracker hits on the muon object (muon_numberOfValidHits); track_n made uint (PR #19)
v5	2026-06-22	L1 trigger info for the per-hemisphere L1 trigger-efficiency measurement on cosmics: uGMT muon candidates (L1muon_*) and L1 DT Local Trigger primitives (dtTrigPh_*, one per DT chamber, with chamber global position for hemisphere assignment). Also the magnetic field bField [T]: in data read from the online DCS magnet current (so B-off runs show ~0), in MC the nominal 3.8 T (set isData=0 for MC, the DCS current is dummy there)

Per-hemisphere L1 DT trigger efficiency (L1TriggerEfficiency.py)

EarthAsDMAnalyzer/test/L1TriggerEfficiency.py measures the L1 DT Local Trigger efficiency with a tag-and-probe on cosmics that cross both hemispheres, following the CMS cosmic commissioning paper (JINST 5 (2010) T03002, Fig. 12): tag = the DT trigger fired in one half, probe = did it also fire in the other half, as a function of the offline muon pT. It uses the dtTrigPh_* branches (v5+).

1. Produce the ntuples (the dtTrigPh_* branches must be filled)

The DT trigger primitives are not stored in standard files, so the ntuplizer config has to provide them. The relevant config snippets are already set up in test/4N-Data_cfg.py, test/CRAB/4N-Data_cfg.py and test/CRAB/4N-CentralMC_cfg.py:

• Data (RAW-RECO) — re-unpack from the raw FED data:

  process.load('EventFilter.L1TRawToDigi.bmtfDigis_cfi')
  process.muonPhiAnalyzer.dtTrigPhiCollection = cms.InputTag("bmtfDigis")
  process.p = cms.Path(process.bmtfDigis * process.muonPhiAnalyzer)

• MC (GEN-SIM-RECO / with muonDTDigis) — re-emulate from the DT digis:

  process.load('L1Trigger.DTTrigger.dtTriggerPrimitiveDigis_cfi')
  process.dtTriggerPrimitiveDigis.digiTag = cms.InputTag("muonDTDigis")
  process.muonPhiAnalyzer.dtTrigPhiCollection = cms.InputTag("dtTriggerPrimitiveDigis")
  process.p = cms.Path(process.dtTriggerPrimitiveDigis * process.muonPhiAnalyzer)

2. Run the measurement

cd CosmicsAnalyzer/EarthAsDMAnalyzer/test/

# single sample (data or MC): combined + per-hemisphere (upper/lower) curves
python3 L1TriggerEfficiency.py -i Ntuplizer-Data-Run2025A.root -n L1DTTrigEff_Data

# data vs MC overlay of the combined efficiency
python3 L1TriggerEfficiency.py -i Ntuplizer-Data-Run2025A.root \
                               --mc Ntuplizer-MC-CosmicToMu.root \
                               -n L1DTTrigEff_DataMC

# with paper-like fiducial cuts (pT > 5, >= 2 DT stations, away from phi-cracks)
python3 L1TriggerEfficiency.py -i data.root --mc mc.root --acceptance -n L1DTTrigEff_DataMC_acc

Each run prints the inclusive efficiency (Clopper-Pearson interval) and writes <name>_vs_pt.png, <name>_vs_pt.pdf and <name>.root (holding the TEfficiency objects eff_comb, eff_probeUpper, eff_probeLower).

Options: -i input ntuple, --mc second ntuple to overlay, -n output basename, -o output dir, -t tree name (default muonPhiAnalyzer/tree), --qual min DT primitive quality code (default 4), --bx max |bx| (default 1), --dr max dR of the primitive-to-leg match (default 0.4), --minseg min DT segments per hemisphere (default 1), --acceptance apply the fiducial cuts.

Note: on an el8/el9 mismatched host, run inside cmssw-el8. If the input ntuples live on /ceph (not mounted in the container), bind it first, e.g. export SINGULARITY_BINDPATH="/ceph:/pnfs" and refer to the files under /pnfs.

3. BX / quality scan (timing vs intrinsic, L1TriggerEfficiencyBXscan.py)

L1TriggerEfficiencyBXscan.py tabulates the inclusive efficiency and the data/MC scale factor (SF = eff_data / eff_MC) over a grid of BX windows and quality thresholds in a single pass, and writes a plot of efficiency vs BX window:

python3 L1TriggerEfficiencyBXscan.py -d Ntuplizer-Data-Run2025A.root \
                                     -m Ntuplizer-MC-CosmicToMu.root \
                                     -n L1DTTrigEff_BXscan

This disentangles the two sources of any data/MC difference: the BX window probes timing, the quality threshold probes the intrinsic local-trigger quality. For the Run2025A vs CosmicToMu test samples the intrinsic efficiency agrees (SF ~ 1.0 at |bx|<=2, quality >= 4); the apparent gap at the nominal |bx|<=1 is a BX/timing-synchronisation effect (a cosmic crosses the two hemispheres ~1-2 BX apart, which the MC emulation models too tightly). Options: -d data ntuple, -m MC ntuple, --bxlist, --quallist, --plotqual, --dr, --minseg, -n/-o output name/dir.