11Dan ToveyDan Tovey University of SheffieldUniversity of Sheffield
Physics Commissioning and Initial Background Estimation for SUSY
Dan ToveySUSY Working Group
Physics Commissioning and Initial Background Estimation for SUSY
Dan ToveySUSY Working Group
22Dan ToveyDan Tovey University of SheffieldUniversity of Sheffield
IntroductionIntroduction• Preparations needed to ensure efficient/reliable searches
for/measurements of SUSY particles in timely manner:– Initial calibrations (energy scales, resolutions, efficiencies etc.);– Minimisation of poorly estimated SM backgrounds;– Estimation of remaining SM backgrounds;– Development of useful tools.
• NB This is not the Tevatron (no previous σ measurements at same √s) !• Many issues will be common with other WG, esp:
– Standard Model (W ( lν) + n jet, Z( ll) + n jet) from Z( l+l-) + n jet);– Top (full reconstruction of semi-leptonic ttbar events);– Higgs (Estimation of high ET
miss backgrounds)– Jet/ET
miss (Estimation of fake ETmiss QCD backgrounds, jet energy scale
etc.);– Combined Performance groups (calibration of energy scales, resolutions
and efficiencies).• Should work together to develop common tools and analysis strategies
wherever possible …
33Dan ToveyDan Tovey University of SheffieldUniversity of Sheffield
StrategyStrategy• R-Parity conserving SUSY search
channels:– Large ET
miss;– Large jet multiplicity;– Large ET
sum.• Will need convincing estimates of
backgrounds with as little data as possible.
• Background estimation techniques will change depending on integrated lumi.
• Ditto optimum search channels & cuts.• Aim to use combination of
– Fast/’brisk’-sim;– Full-sim;– Estimations from data.
• Use comparison between different techniques to validate estimates and build confidence in (blind) analysis.
ATLAS
5σ
44Dan ToveyDan Tovey University of SheffieldUniversity of Sheffield
StrategyStrategy• Main backgrounds:
– Z( ll) + n jets– W ( lν) + n jets– ttbar– QCD
• Generic approach :– Select low ET
miss background calibration samples;
– Extrapolate into high ETmiss signal region.
Jets + ETmiss + 0 leptons
ATLAS
10 fb-1
• Used by CDF / D0• Extrapolation non-trivial.
– Must find variables uncorrelated with ET
miss
• Several approaches developed.
QCDW+jetZ+jetttbar
• Also:– Single top– WW/WZ/ZZ
ATLAS
55Dan ToveyDan Tovey University of SheffieldUniversity of Sheffield
W/Z + n JetsW/Z + n Jets• Comes from Z νν + n jets, W lν + n jets, W τν + (n-1) jets (τ fakes jet)• Estimate from Z l+l- + n jets• Tag leptonic Z• EITHER : Discard one or both leptons and use data sample • OR : Use simulation normalised to Z l+l- + >1 jet data (good stats - CDF)• Scale by inclusive N/N+1 jets factor (below) + appropriate σ.BR ratio• Correct for lepton identification efficiencies• Also appropriate for WW/WZ/ZZ
CDF
M. SpiropuluThesis
66Dan ToveyDan Tovey University of SheffieldUniversity of Sheffield
Top BackgroundsTop Backgrounds• Estimation using simulation possible (normalised to data ttbar
selection) - cross-check with data ?• Standard (TDR) semileptonic top cuts look rather like SUSY cuts with
looser ETmiss requirement!
• If harden ETmiss cuts top sample contaminated with SUSY signal …
• Possible approach (probably extremely difficult? - combinatorics):– Select semi-leptonic candidates (standard cuts – what btag available?);– Fully reconstruct top and W momenta;– Replace hadronic W with leptonic decay (appropriate boosted 2-body
decay distribution) high ETmiss events.
– Worry about correlations between selection cuts and ETmiss distribution.
ATLASPhysics TDR
77Dan ToveyDan Tovey University of SheffieldUniversity of Sheffield
QCD and Fake Missing ETQCD and Fake Missing ET
• Caused by lack of detector hermeticity, dead channels, non-gaussiantails to jet energy distributions (high tail from pile-up, low tail from dead material, punch-through etc.)
• Hardest background to estimate.– Simulations require detailed understanding of detector performance (not
easy with little data).– Would require vast full simulation effort.
• Strategy: 1) Initially choose channels which minimise contribution until well
understood (e.g. jets + ETmiss + n leptons).
2) Choose hard cuts which minimise contribution to background.3) Estimate background using data and/or calibrated fast MC.
88Dan ToveyDan Tovey University of SheffieldUniversity of Sheffield
Fake Missing ET MinimisationFake Missing ET Minimisation• Ensure careful pre-calibration of
calorimeters• Inter-calibration precision most important• Reject events where fake ET
miss likely :– Reject beam-gas and machine background
with event ‘cleaning’.– Require primary vertex in central region – Reject events with hot cells– Reject CR muons etc.– Reject events where ET
miss vector points in (opposite) direction of (to) jets (jet fluctuations)
– Reject events with jets pointing at regions of poor response (barrel-extended barrel, barrel-endcap, endcap-FCAL, FCAL (for high pT)).
– Cut on Missing ET Significance
CDF
M. Spiropulu Thesis
R1
R2Dππ
Collinear dijets?δφ1
δφ2δφ1(2) = φ1(2)-φETmissR1(2)=sqrt(δφ2(1)
2+(π-δφ1(2))2)Dππ=sqrt((π-δφ1)2 + (π-δφ2)2)
99Dan ToveyDan Tovey University of SheffieldUniversity of Sheffield
Missing ET SignificanceMissing ET Significance• Used at Tevatron.• Useful variable for identifying ‘real’
ETmiss.
• Several definitions:– ET
miss/√ETsum;
– Likelihood-type quantity calculated with MC smearing of energies, primary VX position etc.
• ATLAS needs similar tools.
D0 Note 3629
W+jets
D0 Note 3629
QCD
ATLAS
1010Dan ToveyDan Tovey University of SheffieldUniversity of Sheffield
Fake Missing ET EstimationFake Missing ET Estimation• Fake ET
miss rejection partially successful. • Estimate remaining contribution from data or MC.• Possible approaches?• Find variables uncorrelated with ET
miss (e.g. Dππ) which reject SUSY and measure background in sidebands.
• Use γ+jet / pT balance in low ETmiss collinear di-jets
to estimate jet energy distributions (inc. tails).– Feed results into dedicated fast simulation / smear
low ETmiss multijets using bootstrap technique.
– May require dedicated calibration run with prescaledlow pT jet trigger (a la CDF etc.)
• Use ‘brisk’ simulation with real geometry.• Set upper limits to background using fullsim?
CDF
CDF
D0D0
Data MC SUSY QCD
1111Dan ToveyDan Tovey University of SheffieldUniversity of Sheffield
Fake Missing ETFake Missing ET• Can poorer (resolution) ET
miss
measures be found for which the QCD tails can be estimated more accurately?
• Example: ETmiss based on
reconstructed physics objects– Easier to use bootstrap/fast
simulation to estimate;– Easier to calculate Missing
ET Significance;– ~40% increase in gaussian
resolution;– Could also add in
unclustered energy .• At what point does it become
worth using these for searches (as opposed to measurements)?
ATLAS
For illustrative purposes only!
1212Dan ToveyDan Tovey University of SheffieldUniversity of Sheffield
Action PlanAction PlanSUSY Group Plans:• Study optimum search strategies for low mass scale models at limit of
statistical sensitivity as function of integrated lumi.• Develop fake ET
miss rejection methods and tools.• Develop background estimation methods (QCD, ttbar, Z/W+n jets).• Determine required pre-scaled trigger thresholds and stats for QCD jet
calibration samples.Requests:• Vital to have detailed plan of expected calibration uncertainties in
energy scales, resolutions, efficiencies etc. of physics objects as a function of integrated luminosity benchmarks (e.g. 0 pb-1, 10 pb-1, 100 pb-1, 1 fb-1, 10 fb-1 or finer), also inter-calibration precision.
• ‘Brisk’ simulation tools (possibly integrated with ATLFAST).• Tools for rejection of hot/dead cells, beam related background, beam-
gas, cosmic rays (inc. inside events), out of time events etc.• Large fully simulated data samples matching pre-scaled trigger run.
1313Dan ToveyDan Tovey University of SheffieldUniversity of Sheffield
‘Brisk’ Simulation‘Brisk’ Simulation• For ET
miss estimation useful to have intermediate (‘brisk’) simulation between fast and full.– Detailed calorimeter model (simplified geometry
database?);– Access to conditions database (dead channels,
calibrations etc.);– Fast 3D shower simulation (a la GFLASH /
energy spotting) – also e.g. Conversions.– Goes beyond FastShower (3D, higher
granularity).• Already have something similar in CMS
(CMSJET Famos).• Build on existing work:
– LAr: Barberio and Straessner (SW June 2003)– Tile: Sutiak, Tokar, Zenis and Kulchitsky (SW
March 2003)
longitudinal profile
d(mm)radial profile
r/rM
ATLAS
ATLAS
1414Dan ToveyDan Tovey University of SheffieldUniversity of Sheffield
Tevatron ExperienceTevatron Experience
CDF
CDF
CDF
1515Dan ToveyDan Tovey University of SheffieldUniversity of Sheffield
Tevatron ExperienceTevatron ExperienceCDF