MC for displaced low mass dimuons (scouting data) UCSB : N. Amin, C. - PowerPoint PPT Presentation

MC for displaced low mass dimuons (scouting data) UCSB : N. Amin, C. Campagnari, U. Sarica UCSD : V. Krutelyov, M. Masciovecchio, A. Yagil Boston : I. Suarez, A. Tsatsos Rutgers : A. Gandrakota, A. Lath, H. Routray Nebraska : F . Golf July 6, 2020

Introduction to analysis ⚫ Look for events with at least 2 displaced muons and an associated displaced vertex (DV) • Datasets /ScoutingCaloMuon/Run201{7,8}*/RAW • HLT: DST_DoubleMu3_noVtx_CaloScouting (muon p T > 3GeV, | 𝜃 | < 2.4) • Consider full mass range (down to 2 m muon ) and displacements between 0 and 11cm ⚫ In addition to model-independent results, provide interpretations for 3 main signal models • Example t with EXO MC): Z d H Z d erpretation . gg → φ B → φ H → Z d Z d 2

Analysis selections ⚫ Select events with 2 good ScoutingCaloMuons and associated displaced ScoutingVertex with selections below 2017 L1 trigger seeds Baseline selections OR of ‣ DoubleMu4 _SQ_OS_dR_Max1p2 At least 2 OS muons and 1 DV in collections, then ‣ DoubleMu0er1p4_SQ_OS_dR_Max1p4 ‣ DV ‣ DoubleMu_15_7 ‣ (x, y, z) errors < (0.05, 0.05, 0.1) cm ‣ chi2/ndof < 5 2018 L1 trigger seeds ‣ 𝜍 < 11 cm OR of ‣ Muon (p T > 3 GeV, | 𝜃 | < 2.4) ‣ DoubleMu4p5 _SQ_OS_dR_Max1p2 ‣ ID ‣ DoubleMu0er1p4_SQ_OS_dR_Max1p4 ‣ Num. tracker layers with meas. > 5 ‣ DoubleMu_15_7 ‣ chi2/ndof < 3 ‣ Isolation Additional selections ‣ Track isolation < 0.1 ‣ 𝛦 R with closest jet > 0.3 ⚫ distance(DV, nearest pixel module plane) > 0.05cm ‣ Dimuon kinematics ⚫ if DV 𝜍 > 3.5cm, expected pixel hits ≤ observed pixel hits, for both muons ‣ | 𝛦𝜚 (muon 1, muon 2)| < 2.8 ⚫ log 10 (| 𝛦𝜃 𝜈𝜈 |/| 𝛦𝜚 𝜈𝜈 |) < 1.25 ‣ | 𝛦𝜚 (dimuon, DV vector)| < 0.02 ⚫ |d xy / 𝜏 dxy | > 2 for both muons ⚫ |d xy |/( 𝜍 m/p T, 𝜈𝜈 ) > 0.1 for both muons 3

Signal extraction ⚫ After making baseline selections, bin events by • DV 𝜍 : [0, 0.2, 1, 2.4, 3.1, 7, 11] cm • dimuon p T : [0, 25, ∞ ] GeV • both muons isolated/one muon non-isolated ⚫ Fit the background distribution to evaluate exclusion limits ⚫ More analysis details in a previous LL EXO presentation 𝜍 <1 cm 𝜍 >1 cm 4 isolated category

Generation details ⚫ gg → 𝜚 • MadGraph gridpack generating 𝜚 with one extra parton • Subsequent Pythia fragment gives 𝜚 lifetime and forces 𝜈𝜈 decay ⚫ H → Z d Z d • One Z d decays to 𝜈𝜈 and the other according to SM Z BRs • Using latest version of Higgs Powheg gridpacks ( /cvmfs/cms.cern.ch/phys_generator/ gridpacks/2017/13TeV/powheg/V2/gg_H_quark-mass-e ff ects_ZZ_NNPDF31_13TeV/gg_H_quark-mass- e ff ects_NNPDF31_13TeV_M125/v4/gg_H_quark-mass-e ff ects_NNPDF31_13TeV_M125.tgz ) with modified JHUGen decay files, making SM Z have LL properties, for a given mass and lifetime • Default pythia fragment used afterward ⚫ B → 𝜚 • Generation starting with Pythia-only, for a given mass and lifetime • Gives rise to soft muons, so this is accompanied by a gen-level filtering module requiring two p T >3GeV, | 𝜃 |<2.4 muons ‣ Filter e ffi ciency ~1% for m 𝜚 =2GeV, c τ =20mm point 5

Generated kinematics 6

Selection e ffi ciencies ⚫ To get an idea of selection e ffi ciencies wrt generation, consider H → Z d Z d (m Zd =8GeV, c τ =10mm) events after requiring exactly 2 muons with p T >3GeV, | 𝜃 | <2.4 at gen. level, then plot e ffi ciencies vs l xy or trailing muon p T ⚫ Due to HLT configuration requiring at least 2 pixel hits, 0 e ffi ciency past l xy ~11cm Gen. l xy (cm) Trailing gen. muon p T (GeV) 7

Production requests ⚫ We would like to request a 2D grid (mass,c τ ) of benchmark samples with 2017 and 2018 conditions • 100k events per point • For each model, this is 3x3x2=18 datasets, 1.8M events ⚫ Scouting collections only stored up to AODSIM , so in principle, we do not need any tiers beyond that ⚫ Links to gridpacks and fragments in the summary table below (subdirectories here) ⚫ If appropriate, will fill out request spreadsheet entries after this meeting ⚫ NOTE: We are still investigating some details with the B → φ model, but wish to proceed with the generation of the other two Pythia mass points c τ points model gridpack conditions fragment (GeV) (mm) MG gridpack modified gg → φ 1, 8, 15 1, 10, 100 2017, 2018 (set mass) (set c τ ) powheg + JHUGen H → Z d Z d gridpack default 1, 8, 15 1, 10, 100 2017, 2018 (set mass, c τ ) modified + gen. B → φ no gridpack filter 1, 2, 4 1, 10, 100 2017, 2018 (set mass, c τ ) 8

Backup 9

Displaced simulation issues ⚫ Wei Shi previously reported issues with LLP not being propagated past beam pipe • They use a patch to CMSSW to specially treat a particular pdgId ⚫ We observed the same issue • Worked around it by using pdgId>4000000 (e.g., 6000211) for the 𝜚 particle (no need to edit CMSSW) • For the Z d particle, which is a modified SM Z (pdgId 23), propagation works well out of the box from Wei Shi 10

⃗ ⃗ Lifetime reweighting p T ) m ct = c ( L xy ⋅ p 2 T exp ( c τ 1 ) w ( t ) = c τ 0 ct − ct c τ 1 c τ 0 ⚫ Make matrix of reweighting between c 𝜐 ={1,10,50,100}mm samples • Plotting reco lxy distributions after analysis selections ⚫ Lower diagonal (reweighting high c 𝜐 to low c 𝜐 ) shows good agreement (compare orange and green) and doesn't su ff er from extreme weights 11

Cutflow e ffi ciencies ⚫ pass L1 OR : OR of 3 seeds ⚫ Cuts are cumulative starting with the denominator ⚫ pass HLT : DST_DoubleMu3_noVtx_CaloScouting (using gen. level quantities) : ⚫ pass object sel : event has at least 1 DV and 2 matching muons, • Exactly 1 Z d with 2 muons (p T > 5 GeV, | 𝜃 | < 2.4) with some baseline cuts on the DV ⚫ pass iso. baseline : isolated baseline selection • 𝜍 < 11cm ⚫ pass extra cuts : • 𝛦 R 𝜈𝜈 <1.2 • distance(DV, nearest pixel module plane) > 0.05cm • distance(DV, nearest pixel module plane) > 0.05cm • if DV 𝜍 > 3.5cm, expected pixel hits ≤ observed pixel hits • log 10 (| 𝛦𝜃 𝜈𝜈 |/| 𝛦𝜚 𝜈𝜈 |) < 1.25 • log 10 (| 𝛦𝜃 𝜈𝜈 |/| 𝛦𝜚 𝜈𝜈 |) < 1.25 ⚫ pass d xy cuts : (|d xy / 𝜏 dxy | > 2) AND (|d xy |/( 𝜍 m/p T, 𝜈𝜈 ) > 0.1) for both muons 12 H → Z d Z d