260 likes | 367 Views
QCD Control Sample. Shilei Zang University of Colorado, Boulder. GMSB Meeting, 1 Aug 2008. γγ. HLT; trkIso<9 ; HoE<0.1; not electron ; at least 2 γ ; Pt1>90; Pt2>30; (dE>0). e γ. HLT; trkIso<9 ; HoE<0.1; Electron: haveSeeds && # of track (Pt>1.5; Δ R<0.1)>=1 at least 1e1 γ ;
E N D
QCD Control Sample Shilei Zang University of Colorado, Boulder GMSB Meeting, 1 Aug 2008
γγ • HLT; trkIso<9; HoE<0.1; not electron; at least 2 γ; • Pt1>90; Pt2>30; (dE>0) • eγ • HLT; trkIso<9; HoE<0.1; • Electron: haveSeeds && # of track (Pt>1.5; ΔR<0.1)>=1 • at least 1e1γ; • Pt1>90; Pt2>30; (dE>0) • γγ control • HLT; trkIso<9 or trkIso>12; HoE<0.1; not electron; at least 2 γ; • Pt1>90; Pt2>30; (dE>0); do not satisfy γγ • eγ control • HLT; trkIso<9 or trkIso>12 ; HoE<0.1; at least 1e1γ; • Pt1>90; Pt2>30; (dE>0); do not satisfy eγ
γγ vs. γγ-control γγ vs. γγ-control GMSB Bkg γγ vs. eγ • 1/fb • Blue: γγ • Red: γγ control (or eγ) • γγ : 80 signal; 2698 bkgs • γγ control : 55 signal; 3157 bkgs • eγ : 61 from bkgs Bkg
Jet resolution: • MET resolution: Signal yield: Control sample:
MET Bkg MET GMSB k=5.0 pt1>80, pt2>20 k=5.0 MET vs. for background (left) and for GMSB signal (right). • has powerful separation. • almost no correlation with Pt1, Pt2, trkIso. (but MET has.)
Signal yield: Control sample: • Electron: haveSeeds && # of track (Pt>1.5; ΔR<0.1)>=1 Selected Sample (gumbo & chowder) • HLT; HoE<0.1; not electron; at least 2 γ; • Pt1>80; Pt2>20; Di-photons: Control sample: trkIso <9 >12 trkIso/Pt <0.08 >0.1 Pt1 >120 <90 Pt2 >60 <30 Pt1& Pt2 Pt1>110&&Pt2>50 Pt1<90 or Pt2<30
Pt1 Bkg Pt1 GMSB trkIso1 trkIso1 Pt2 Pt2 Bkg GMSB trkIso2 trkIso2
MET Bkg MET GMSB trkIso1 trkIso1 MET MET Bkg GMSB trkIso2 trkIso2
MET Bkg MET GMSB trkIso1/Pt1 trkIso1/Pt1 MET MET Bkg GMSB trkIso2/Pt2 trkIso2/Pt2
MET Bkg MET GMSB Pt1 Pt1 MET MET Bkg GMSB Pt2 Pt2
MET MET • HoE<0.1; track un-match • Pt1>80; Pt2>20; • Blue: trkIso<9 (4828 evts) • Red: trkIso>12 (550 evts) • HoE<0.1; track un-match • Pt1>80; Pt2>20; • Blue: trkIso/Pt<0.08 (3011 evts) • Red: trkIso/Pt>0.1 (771 evts) trkIso trkIso/Pt
MET/σ(γPt) MET/σ(γPt) Bkg GMSB trkIso1 trkIso1 MET/σ(γPt) MET/σ(γPt) Bkg GMSB trkIso2 trkIso2
MET/σ(γPt) MET/σ(γPt) Bkg GMSB trkIso1/Pt1 trkIso1/Pt1 MET/σ(γPt) MET/σ(γPt) Bkg GMSB trkIso2/Pt2 trkIso2/Pt2
MET/σ(γPt) MET/σ(γPt) Bkg GMSB Pt1 Pt1 MET/σ(γPt) MET/σ(γPt) Bkg GMSB Pt2 Pt2
MET/σ(γPt) Bkg MET/σ(γPt) GMSB Pt1+Pt2 Pt1+Pt2 Pt1 Pt1 Bkg GMSB Pt2 Pt2
MET/σ(γPt) MET/σ(γPt) • Blue: trkIso<9 (4828 evts) • Red: trkIso>12 (550 evts) • Blue: trkIso/Pt<0.08 (3011 evts) • Red: trkIso/Pt>0.1 (771 evts) trkIso trkIso/Pt
MET/σ(γPt) MET/σ(γPt) • Pt1>90; w/o trkIso • Blue: Pt2>60 (3098 evts) • Red: 20<Pt2<30 (2238 evts) • Pt2>30; w/o trkIso • Blue: Pt1>120 (2452 evts) • Red: 80<Pt1<90 (2355 evts) Pt1 Pt2
w/o trkIso • Blue: Pt1>110 && Pt2>50 (2249 evts) • Red: Pt1<90 or Pt2<30 (5410 evts) MET/σ(γPt) Pt1 & Pt2
MET/σ(MET) MET/σ(MET) GMSB Bkg Pt1 Pt1 MET/σ(MET) MET/σ(MET) Bkg GMSB Pt2 Pt2
MET/σ(MET) MET/σ(MET) • Blue: trkIso<9 (4828 evts) • Red: trkIso>12 (550 evts) • Blue: trkIso/Pt<0.08 (3011 evts) • Red: trkIso/Pt>0.1 (771 evts) trkIso trkIso/Pt
MET/σ(MET) MET/σ(MET) • Pt1>90; w/o trkIso • Blue: Pt2>60 (3098 evts) • Red: 20<Pt2<30 (2238 evts) • Pt2>30; w/o trkIso • Blue: Pt1>120 (2452 evts) • Red: 80<Pt1<90 (2355 evts) Pt1 Pt2
MET/σ(γPt) MET Di-photons; 1/fb MET/σ(MET) Separation: MET/σ(γPt) > MET > MET/σ(MET)
The best choice now: MET/σ(γPt); Pt1, Pt2 together for the control sample (we can get enough control events!) MET/σ(γPt) MET/σ(γPt) Pt1 & Pt2 Di-photons; 1/fb
A new technique for mass variable (stopped for the moment) • Momentum of two photons (known) • Momentum of two gravitinos: P1x, P1y, P1z; P2x, P2y, P2z. (unknown) • MET: METx, METy. (known) Generate P1x, P1y, P1z; P2x, P2y, P2z distributions according to GMSB MC truth. (take all GM1b-GM1g GMSB simulated events for this.) sample A. For each event i (not in sample A; already passed all cuts: iPt1>80, iPt2>20, iMET>80), use all events in sample A with |Pt1-iPt1|<ic1, |Pt2-iPt2|<ic2, |MET-iMET|<ic3 to calculate 4 neutrilino-mass variables: Photon1 Photon2 Photon1 Photon2
Require |mass(j)-mass(k)|<mass-cut (j, k=1, 2, 3, 4) • For each event i, calculate the mass likelihood: • Take the maximal point (maximal likelihood) as the mass of event i. • For all events, we get themass distribution. • Between step 3) and step 4), we can also de-convolute the mass(j) distribution to get a narrower mass distribution, this may recover some information and improve the analysis. • Narrow distribution for GMSB signal • Wide distribution for Background • Treat GM1b-GM1g at the same time (parameter independent) • Can maximize the final significance. Good properties:
New Mass (preliminary) GMSB (GM1e) Bkg GMSB GMSB formula mass MET Pt1>90; Pt2>30 Pt1>90; Pt2>30