270 likes | 404 Views
SUSY Roberto Rossin (UCSB). USCMS meeting 8-9 May 2009. mSUGRA vs GMSB vs …. Not really the topic of this talk. Benchmark points have been used extensively to study the physics reach of CMS for various SUSY hypotheses
E N D
SUSY Roberto Rossin (UCSB) USCMS meeting 8-9 May 2009
mSUGRA vs GMSB vs … • Not really the topic of this talk. • Benchmark points have been used extensively to study the physics reach of CMS for various SUSY hypotheses • Good, but not the only possible, nor optimal, approach to design a search • Presented here: few searches based on simple signatures defined by their physics objects (Reference Analyses) • As model independent as possible • As data driven as possible • Lot of work has been done and is ongoing, only a small part of it will be touched in this presentation
Outline • Searches and methods for data driven backgrounds and • efficiencies estimations • most of this talk Simplified models based on topologies Detector commissioning from SUSY perspective
g + jets +MET (1) • GMSB if you like, but it is actually a simple final state. • 2 photons • Jets • MET • Focus on the first and last bullet • Measure photon efficiencies on data: Z->ee • e/g definition as similar as possible • Estimate the MET contributions from: • QCD • EWK • CR and beam halo
g + jets +MET (2) • Efficiencies via Z->ee (tag & probe) • iso variables chosen to be similar for iso electrons and g • tag: pixel match and tight requirement on h cluster width • probe: measure iso and pixel matching efficiencies • Isolation efficiency • check (and correct) for biases (MC vs T&P) • signal will have higher jet activity • MC will be validated on data and account for this
g + jets +MET (3) • Pixel matching (in)efficiency • data again: ee vs eg vs gg • EWK backgrounds • Wg->eng is a background if the ele has no pixel match • Select e+g and weight by the pixel match inefficiency factor fe->g/(1- fe->g) • The estimation is subtracted to the background, leaving only the QCD background contribution
g + jets +MET (4) • QCD backgrounds • assuming no signal at low MET to normalize • two control samples considered for the modeling of the MET spectrum • fake-fake (both EM objects fail at least one iso) • Z->ee • control samples spectra are different from the gg one • reweight to account for different hadronic activity • closure tests successful • no impact on signal • MET prediction
(di)jet +MET with alphaT (1) SUSY: squark-squark production with Mgluino > Msquark Squark decaying to quark+LSP • Final state: di-jet+MET • 2 high pT jets • MHT = - (pTj1+ pTj2) • not aligned w/ jets • lepton veto • third jet veto • Main backgrounds: • QCD di-jet • Z->vv +jets • W+jets, Z->ll and top when leptons are lost
(di)jet +MET with alphaT (2) QCD background: Randall & Tucker-Smith suggest to use a kinematics variable • for QCD di-jets: aT=0.5 (or smaller if mis-measured ET) • exploits that for QCD jets need to be back-to-back and of equal magnitude • for real MET aT can be greater
(di)jet +MET with alphaT (3) • Data driven method to estimate the backgrounds: • Z->nn + jets • W -> nl, Z->ll, top • QCD (again) • ABCD method • 3 out of 4 regions must be signal free • need 2 uncorrelated variables for all backgrounds • aT and h of the leading jet
(di)jet +MET with alphaT (4) Data driven method to estimate the backgrounds: results w/o signal (closure test) w/ signal (LM1) • Extra checks: • Check the background flatness in h on data by relaxing the HT and as a consequence diluting the (potential) signal • Alternative data driven Z->nn+jets estimation from W->nl+jets Update: aT definition extended to multi-jet events. Ongoing.
MET modeling in V + jets (1) If V=Z/g the main source of MET are the jets (If V=W the lepton can parameterize the n ) Idea: model the jet system using QCD events V • Algorithm: • Construct MET templates from the QCD sample based on 2 variables: Nj and HT=SjETj • For each V+jets event pick MET template according to Nj and HT
MET modeling in V(=Z) + jets (2) Z+jets: easy. Blue: measured Red: predicted Nj=2 Nj=3 Nj=4 Measured/Predicted
MET modeling in V(=W) + jets (3) • In lepton+jets+MET the dominant background is SM W(ln)+jets and semileptonic ttbar+jets • To predict MET need to know 2 ingredients: • MET from resolution effects • neutrino spectrum -> use the charged lepton spectrum Nj=4 Nj=2 Nj=3 Blue: measured Red: predicted Prediction is 20% higher for Nj=2 due to W+jets contribution with W polarized in the transverse plane after event selection. Can be corrected
Blue: measured Red: predicted Purple: prediction in presence of LM1 Nj=3 MET modeling in V(=W) + jets (4) What if BSM is there? Q: Can it bias the prediction? A: No Regimes where the method can become biased: • jet inefficiency • bias when very high pT jet is lost. • non Gaussian jet resolution • bias when large fluctuation. • hot cells • tail overestimation. • energy scale offset • bias, coherently adding/removing energy. Solution: calibration <10% or search in the direction ortogonal wrt the V Solution: apply quality selection criteria to remove badly mis-reconstructed events.
Simplified models for search design/optimization (1) ~ constant all lines on-shell Full model calculation Parameterization * spins couplings off-shell states … branching ratios
Simplified models for search design/optimization (2) Model Parameterization “blobs” represent dynamics that are parameterized by one rate and possibly an additional shape parameter on- and off-shell masses several couplings control both kinematics and rates Production contributions: Associated q-g: intermediate q and g Same sign q: g and 4 neutralinos Off-shell particle do not appear, their effects present in the rates ~ ~ ~ gauginos do not appear in the OSET ~
Simplified models for search design/optimization (3) Very simple Monte Carlo scripting language Quantum numbers / mass sup : charge=1 color=1 mass= 800 New particle sup > up LSP Decay mode g g > sup sup : matrix 1 Production mode • Other particles, production processes and decays can be added similarly • No cross section calculation, reweight applied at analysis level
Simplified models for search design/optimization (4) • Example. Simple SUSY-like system with: gluino, sup, LSP Final states: 2-6 jets, MET
Simplified models for search design/optimization (5) • Simple exercise: • Mass grid generated with variable Mgluino, Msup MLSP • Cross sections calculated w/ Prospino • Using QCD only as background • Using MET/HT as discriminator • Punzi significance as figure of merit • Optimizing the leading jet transverse momentum pTj • Grid example for fixed MLSP=100 • When there is direct decay to LSP (and since no systematics are considered) the optimization suggests to cut very hard
Simplified models for search design/optimization (6) • Significance grid for fixed MLSP=100 • suggests which physics parameters are in the reach of the search • Significance curves VS cut for several parameters • suggests whether the optimized cut is valid/stable as function of parameters
Simplified models for search design/optimization (7) • Advantages: • focus on the topologies you are interested on • deal with (few) physics parameters (masses) • very simple scripting language • ideal for (physics) parameter scanning, grid generating tools available • Caveats: • you get what you put in. No Lagrangian has been calculated for you • might oversimplify the model • no “signal” events will be present in control samples with different topologies • the speaker loves it and he is biased...
HCAL noise impact on MET(1) • Impact on MET spectrum (tail) from noisy HCAL channels can be estimated by looking at CRAFT data • Collect events with a MinBias HCAL trigger • Get noise MET spectrum (and cross section) • Mix it with MC QCD events • split noise and QCD in MET bins, then “sum” • Approximate approach • Mixing done at tower level, not at digi-level (DataMixingModule) • No timing: everything considered synchronous
HCAL noise impact on MET(2) CRAFT MET spectrum (metNoHF) MC QCD before mixing • Interested in MET>200GeV • mixing only bins that can contribute above threshold
HCAL noise impact on MET(3) Results (before any cleaning cuts) • Remarks • bin with lowest initial MET is the only relevant because of the steeply falling spectrum • MET and MHT behave very similarly wrt the noise
HCAL noise impact on MET(4) • Analysis preselections • Jet 1,2,3: pT > 50 GeV, |h| < 2.5 • Df ( Jet 1,2,3 ; MET ) > 0.3 • SUSY cleaning cuts • Jet 1,2,3: 0.05 < EMF < 0.95 • Df (MPT ; MET ) < 0.75 + • Looks good. Work is ongoing, still have to: • apply the real mixing at digi level • apply the HCAL clean-up routines
Conclusions • SUSY group is prepared for the data coming this year • developed/under development analysis techniques which rely on data driven approaches • independent/multiple background estimations will provide further confidence in case of signal observation • quasi model independent approaches are evaluated • the group is also working on the detector commissioning • recall: we are picky customers • What has not been presented here: A LOT • There are 8 Reference analysis groups working on different topologies to cover as much parameter/model space as possible • Apologies for all the work which has not been cited • The structure of SUSY in Reference Analyses allows to naturally build cooperation among groups facing similar issues (workshop next week)