1. Nov 11, 2007Jim Alexanderp1 /13 MET Significance New Physics Neutralinos (SUSY) Gravitinos (Xtra dim) Heavy photon (Littlest Higgs with T-parity) Unparticles, hidden valley stuff, etc Something else… Old Physics and Detector Issues Jet Mismeasurement (resolution) Underlying event/ unclustered energy Neutrinos in b, c jets Z , W  l Muons Pile up (at hi lumi) Hot/dead cells, cracks, dead regions E T has many possible sources We want to identify New Physics and suppress the rest

2. Nov 11, 2007Jim Alexanderp2 /13 D0 idea: Use Likelihood Method (B. Knuteson et al, D0 Note 3629) Probable MismeasurementPossible MET candidate Example: distinguish measurement error from real MET MET Fluctuations can cause this Fluctuations are unlikely to cause this L ( E T event ) Can we construct a likelihood function to distinguish these cases evt by evt?

3. Nov 11, 2007Jim Alexanderp3 /13 D0 Study (A. Schwartzman, D0 Note ????) Signal: W+jets Bkg: QCD ≥ 2jets. For 50% signal eff (example) bkg eff is 7.5% with ordinary MET cut, Bkg eff is 3.8% with MET signif cut. D0 study MET Significance improves bkg suppression x2 Simple MET cut MET Significance cut Background Efficiency Signal Efficiency Example

4. Nov 11, 2007Jim Alexanderp4 /13 MET Significance for CMS… MET Significance is on this group’s official to-do list First steps taken by Koji Terashi (now gone to ATLAS…) Our plan: –Start with one issue: jet mismeasurement. (Save others for later…) –Construct toy Monte Carlo explore general features of the problem in a simple environment develop structure, code, math –Move to CSA07 / Spring07 Monte Carlo samples: QCD, W  l –Add other features one by one: b, c jets underlying event detector flaws etc –Develop general tool for use by all –Apply to a first physics measurement: W  l Today’s talk

5. Nov 11, 2007Jim Alexanderp5 /13 Jet Mismeasurement Likelihood functions for combining mis-measured vectors. If you add two vectors (think of these as jet vectors, with resolution smearing) where is the resultant likely to end up? This is a 2d likelihood function, L(  x,  y ), which can be used to plot likelihood contours in the  x,  y plane.

6. Nov 11, 2007Jim Alexanderp6 /13 Gaussian case is elegant If the primordial likelihood functions are gaussian, then everything is analytical. –Likelihood shape is simply characterized: V –Gaussians convolve to give more gaussians 2 11  2 If not gaussians, linear combos of gaussians is just as easy to use.

7. Nov 11, 2007Jim Alexanderp7 /13 Definition of the Significance Based on jet measurement uncertainties (black ellipses), calculate the uncertainty of the E T vector…  red ellipses Define S = ln In this illustration, 3.5  S = 6.1 Contours shown: -2lnL = -2lnL min + {1,4,9,…} 1 2 3 4 3.5 L(E T =observed) L(E T =0) ) (

8. Nov 11, 2007Jim Alexanderp8 /13 Angular effects are automatic In the example from p2, what makes the MET on the right look anomalous is largely its direction Error ellipses automatically take this into account. MET  

9. Nov 11, 2007Jim Alexanderp9 /13 Toy Monte Carlo Studies Contents of the toy Monte Carlo: –2-5 “jets” per event, poisson distributed –Jet “E T ” is flat between 10 and 100 –“Measured” Jet E T vector is smeared by  (E T )/ E T = 144%/Sqrt(E T )  (  ) = 2 O –~100 lines of code: VERY SIMPLE Generated E T can be set to 0 or to any desired value. We will call E T =0 samples “QCD”, and E T =40 samples, “W  l ” Warning: Despite the nomenclature, there is no physics in this! This is totally made up…. don’t over- interpret! Purpose is to explore general features and methods.

10. Nov 11, 2007Jim Alexanderp10 /13 Example from Toy Monte Carlo Blue arrows are the “true” underlying jet vectors. Dots are results of 10k smearings of the same true underlying event Contours are calculated values of -2lnL In this case, true MET was fixed at 20. (Green arrow) “Event Display”

11. Nov 11, 2007Jim Alexanderp11 /13 Compare “QCD” and “W  l ” events in the Toy Monte Carlo MET Significance “QCD” events: E T (true)=0 “W  l ” events: E T (true)=40 Significance Generate 10k evts with E T =0 (QCD) and 10k evts with E T =40 (W  l )

12. Nov 11, 2007Jim Alexanderp12 /13 “Signal” Eff and “Background” Eff in Toy Monte Carlo “Signal” = “W” events “Background” = “QCD” events Compare cut on MET significance with simple cut on MET itself. MET Significance cut Simple MET cut D0 Plot, from p3.

13. Nov 11, 2007Jim Alexanderp13 /13 Summary & Plans So far: –Toy Monte Carlo only –Jet mismeasurement is the only effect considered (resolution smearing) –General structure and approach clarified by this exercise Next: –QCD and EWK CMSSW files: expose method to “real” physics –Add more components to the likelihood: semileptonics in b, c jets, underlying event, etc etc. –Eventually: apply to W, W+jets –Lots to do! Participants… Xin Shi * Kyle Story * Deb Mohapatra ** * grad student * post doc