Performance and physics results for Phase1 upgrade

Pixel Upgrade Workshop Grindelwald, August 29th 2012

PERFORMANCE AND PHYSICS RESULTS FOR PHASE1 UPGRADE

Alessia Tricomi (University and INFN Catania)on behalf of the Tracker Upgrade Simulation Group

Pixel Upgrade Meeting Alessia Tricomi

2

Performance studies The goal for the TDR was to show

improvement in Physics cases and robustness of the new design.

Focus on relative improvement of the upgrade wrt current geometry.

Two complementary approaches: Show improvements in basic building blocks for

physics using a full (Geant) simulation of the upgrade Demonstrate improvements in tracking

efficiency and fake rate Demonstrate improved IP resolution and b-

tagging performance Improvements in relevant physics channels

(under PC responsibility)Grindelwald, 29/08/12


3

Pixel Upgrade dictionary

Grindelwald, 29/08/12

Current Detector: Current pixel detector geometry (3 barrel layers, 2

disks) Current beam pipe Dedicated “SLHC” release CMSSW_4_2_8_SLHC2 to

use Design/Ideal conditions and same configurations/settings for tracking

Phase1 Upgrade (R30F12) geometry: Upgrade geometry with 4 BPIX layers and 3 endcap

disks First barrel layer at R=30 with 12 faces New detailed material description according to PSI drawings New beampipe (Sunanda) implemented

CMSSW_4_2_8_SLHCtk + 520 backporting


4

Upgrade Studies Study at

2×1034 cm-2s-1 at 25ns (50ns),<PU>=50 (100) 1×1034 cm-2s-1 at 25 ns,<PU>=25 zero PU

Dynamic data loss (due to pixel ROC) used in simulations

Tracking steps modified for upgrade geometry and high PU Using CMSSW_4_2_8 but with 5_2_0 tracking backported Dropped detached tracking steps (see backup slides)

Used regular CMS DQM validation packages to get tracking and b-tagging performance plots

Fullsim, 14 TeV, ideal conditions, no pixel CPE templates used ttbar from PYTHIA (10K events) Muon gun (10K events-4 muons/event, generated flat in pT

and eta, 200k events-ten muons/event, generated flat in p and eta)


5

Data loss for Upgrade Studies



Peak luminosity valuesCurrent Detector Radius

(cm)% Data loss at 11034 @25ns

% Data loss at 21034 @25ns


BPIX1 4.4 4.0 16 50

BPIX2 7.3 1.5 5.8 18.2

BPIX3 10.2 0.7 3.0 9.3

FPIX1&2 0.7 3.0 9.3

Phase 1 Detector Radius(cm)




BPIX1 3.0 1.19 2.38 4.76

BPIX2 6.8 0.23 0.46 0.93

BPIX3 10.9 0.09 0.18 0.36

BPIX4 16.0 0.04 0.08 0.17

FPIX1-3 0.09 0.18 0.36

6

Upgrade Iterative Tracking (Stdgeom)



5_2_0 tracking for current pixel geometry (from “2012 tune”) Close to 5_2_0 tracking, use steps 0-2, and 4A (for high eta) Reduce step 4A d0 cut to reduce CPU and memory usageIteration Seeds pT cut

(GeV)d0 cut (cm)

dz cut (cm)

Min hits

0 pixel triplets 0.6 0.02 4.0σbs 3

1 low pT pixel triplets 0.2 0.02 4.0σbs 3

2 pixel pairs with vtx 0.6 0.015 4.0σbs 3

3 detached triplets 0.3 1.5 15.0 3

4A pixel +(TEC(1 ring)) triplets

0.4 0.02 10.0 3

4B BPIX+TIB triplets 0.6 1.5 10.0 3

5 TIB, TID, TEC pairs (fewer)

0.7 2.0 10.0 4

6 TOB, TEC pairs 0.6 6.0 30.0 6

Release CMSSW_4_2_8_SLHCstd2_patch1 Tracking steps

7

Upgrade Iterative Tracking (Phase 1)



5_2_0 tracking for Phase 1 geometry (not optimized) Make close to 5_2_0 tracking, use steps 0-2, and 4A, add step “-1” Step 3 (pixel pairs) to recover efficiency in eta ~1.2–1.4 regionIteration Seeds pT cut

(GeV)d0 cut (cm)

dz cut (cm)

Min hits

0 pixel quadruplets 0.6 0.02 4.0σbs 3




3old detached triplets 0.3 1.5 15.0 3


0.4 0.02 10.0 3



0.7 2.0 10.0 4

6 TOB, TEC pairs 0.6 6.0 30.0 6

Release CMSSW_4_2_8_SLHCtk3_patch1 Tracking steps


8

Pixel Upgrade Material Budget Reduced material even with more layers

“Volumes” Mass (g)Current Design Upgrade

BPIX <2.16 16801 6618FPIX <2.50 8582 7024

Rad. Len. Nucl. Int. Len.

Dots – UpgradeGreen – Curr geom

Pixels Pixels


50% less photon conversion in/before pixel at eta 1.5

9


Transverse and Longitudinal IPPrimary VertexTrackingBtaggingRobustness

Object performance



Impact Parameter Resolutions

10 Transverse: muon sample (10 muons/event), zero

pileup Generated flat in E and eta (plot vs absolute p and in 4

eta regions) Compare current and upgrade detectors (modified MTV)



11


Longitudinal: muon sample (10 muons/event), zero pileup Generated flat in E and eta (plot vs absolute p and in 4 eta regions) Compare current and upgrade detectors



12

Impact Parameter Resolutions Transverse: muon sample (10 muons/event), <PU>=50

Generated flat in E and eta (plot vs absolute p and in 4 eta regions) Compare current and upgrade detectors



13

Impact Parameter Resolutions Longitudinal: muon sample (10 muons/event), <PU>=50

Generated flat in E and eta (plot vs absolute p and in 4 eta regions) Compare current and upgrade detectors



14

Primary Vertex Resolution

ttbar sample, zero PU and <PU>=50



Tracking in ttbar15

ttbar sample, High purity pT > 0.9 GeV/c



16

Tracking with muons

Muon sample High purity pT>1 GeV/c num tracking

layers with hits >= 8



17

Tracking Efficiency/Fake Rate ttbar sample, compare current and upgrade detectors

High purity, pT > 0.9 GeV/c



18

Tracking vs PU Average tracking efficiencies vs PU

ttbar, high purity tracks, pT > 0.9 GeV/c



19

Tracking vs PU Average track fake rates vs PU

ttbar, high purity tracks, pT > 0.9 GeV/c



20

B-tagging Performance ttbar, CSV tagger, compare current and upgrade

ak5PFjets PFnoPU, jet pT > 30 GeV, DUS ,c,b jets

15% absolute gain in b jet efficiency for 1% fake rate



21

B-tagging Performance ttbar, CSV tagger, compare current and upgrade, <PU>=100

ak5PFjets PFnoPU, jet pT > 30 GeV, DUS ,c,b jets

Improvement even more impressive at 100 PU



22

B tagging performance


Upgrade as good or better at high pileup as current at low pileup

Upgrade PU50Current PU0

Upgrade PU50Current PU25


23

B-tagging Performance vs PU

ttbar, CSV tagger, compare current and upgrade, <PU>=50 ak5PFjets PFnoPU, jet pT > 30 GeV, DUS,b jets

Much better handling high Pile-UpGrindelwald, 29/08/12


24

Robustness: BPIX1 Inefficiency Study

Vary inefficiency of BPIX layer 1: 0%, 5%, 10%, 20% All other layers at 100% ttbar, <PU>=50, light quark mis-tag=1% Upgrade detector more robust to BPIX1 inefficiency



25

Robustness: BPIX1 Inefficiency Study

Vary inefficiency of BPIX layer 1: 0%, 5%, 10%, 20% All other layers at 100% ttbar, <PU>=50, high purity tracks, pT>0.9 GeV/c Upgrade detector more robust to BPIX1

inefficiency



26

TIB Inefficiency Study Switch off certain modules (in black below)

List provided by Frank Hartmann

Or consider a uniform 20% inefficiency in TIB1,2 Grindelwald, 29/08/12


27

Robustness to TIB degradation


With Dead Modules

With 20% uniform inefficiency in TIB1,2

Upgrade detector more robust wrt TIB loss


28

Small pixel scenario: tracking with ttbar at 100PU


BPIX Layer1: pixel size 75x100 mm2, 220 mm thicknessROC threshold 1200 e- instead of 2000

Significant improvement at 100 PU wrt Upgrade Phase1 detectorGood news towards Phase2


29

Small pixel scenario: btagging performance


BPIX Layer1: pixel size 75x100 mm2, 220 mm thicknessROC threshold 1200 e- instead of 2000No data loss

Significant improvement at 100 PU wrt Upgrade Phase1 detectorGood news towards Phase2

30


ZHllbbHZZ4lSUSY MT2

SUSY gg+METAll analysis show relative improvement and have not been optimized/retuned for high PU

Physics Performance



31

ZHllbb


Analysis based on:1. triggering on lepton events;2. kinematic reconstruction of Z from isolated dileptons;3. reconstructing invariant mass from two b-tagged jets;4. multivariate final variable

Higher muon/electron ID efficiency helps with (1-2), better b-tagging helps with (3-4)

Compare relative performance of current detector and upgrade at 14 TeV with 50 pileup events


32

ZHmmbb event selection



33

ZHmmbb cut flow


Values greater than 1 show increased efficiency for the Phase1 upgrade and vice versa


34

ZHeebb event selection


Same selection as for muons but Electrons use 95% working-point of VBTF Isolation is relaxed as in the di-muon

analysis Dimuon mass ~2 GeV high, so Z mass

cut increased by +2 GeV like in dimuon analysis

Other criteria same as for di-muon channel


35

ZHeebb cut flow




36

ZHllbb results


ZHmmbb 65% relative gain in signal efficiency for di-muon

channel 75% gain with single muon HLT 175% gain with dimuon HLT (upgrade detector

barely affected by the three pixel hit requirement) ZHeebb

65% relative gain in signal efficiency for di-electron channel

Not enough MC to properly estimate total reductions in backgrounds


37

HZZ4l


Analysis based on: 1. triggering on di-lepton events;2. kinematic reconstruction of 2 Zs from isolated

di-leptons;3. reconstructing invariant mass of Higgs

Higher muon/electron ID efficiency helps with (1-2)

Compare relative performance of current detector and upgrade at 14 TeV with 50 pileup events


38

HZZ4l event selection


Using same cuts as in 2012 analysis (HIG-12-016) PF leptons used Electrons pT > 7 GeV, |η|<2.5; Muons pT > 5 GeV, |η| <

2.4 Isolation relaxed from 0.15 to 5.0 |SIP3D|<4 for each lepton 40 < mZ1<120 GeV; 12 < mZ2<120 GeV; ml+l->4GeV m4l > 100 GeV same HLT estimate as in ZH: 3+ pixel hits on trigger

leptons


39

HZZ4m cut flow



40% gain in 4m channel


40

HZZ4e cut flow



50% gain in 4e channel


41

HZZ2e2m cut flow



48% gain in 2e2m channel


42

SUSY MT2b analysis


Fully hadronic final states with large MET B jets coming from cascade decay of gluino and squark

to third generation sbottom, stop SUSY particles identified through the discovery

parameter MT2 Tail of Supersimmetric Transverse Mass related to parent

sparticle mass (endpoint) Compare relative performance of current detector

and upgrade at 14 TeV with 50 pileup events wrt to btagging improvement Signal: LM9 benchmark point Bkg: ttbar

See AN-2012/275 for details


43

SUSY MT2b event selection


Good primary vertes Veto electrons: pT>10 GeV/c; |η|<2.4; |d0|

<0.04cm; |dz|<1.0cm; missing inner hits<2; PFIso<2.0

Veto muons: pT>10 GeV/c; |η|<2.4; |d0|<0.04cm; |dz|<1.0cm; Trk hits>10; Pix hits>0; PFIso<2.0

Jets: 2 PF jets with pT> 20 GeV/c passing loose JetID; veto events with jets pT>50 GeV/c but failing jet ID

MET: PFMET>30GeV;PFH >750GeV

B-tag: Tight CSV tag > 0.898


44

SUSY MT2b analysis


Before any b-tagging One b-jet required

20% higher signal selection efficiency can be obtained without any real optimization for the new detector and high pile-up.


45

SUSY gg+MET analysis


Di-photon events+MET signature for SUSY

No significant SM bkg, small contribution from Vgg

Main bkg comes from fake MET See AN-2012/269 Same 14 TeV, 50 pileup scenario as

others Main improvement comes from fake rate

reduction


46



Events divided in four classes: gg, ge, ee, fake-fake

ge and ee samples used to estimate fake rate by fitting Z peak


47



Fake rate 7.0% for the current detector and 1.25% for Phase1 pixel detectors

Fake rate with the upgrade detector at 50 PU is comparable with the performance of the current detector in low luminosity run


48

Conclusions


All results show that the new detector at high PU performs as well or even better than the current detector at low luminosity

The results also show that the new detector is fairly robust against possible inefficiency in BPIX1 and TIB1,2

All the results have been approved by Tracking/btagging/Physics group – PAS SUS-12-020

Improvements from the new design are broad and substantial and will have a significant impact our physics program

Still a lot of work to be done to tune algorithms and analysis for high PU scenario Stay tuned!

49


Back-up slides


50

Tracking for Upgrade Studies



Use 4_2_8 but with 5_2_0 tracking, and drop detached trks Fullsim, 14 TeV, ideal conditions, no pixel templates Regular CMS validation package, current and upgrade pixel detIteration Seeds pT cut

(GeV)d0 cut (cm)

dz cut (cm)

Min hits




3 detached triplets 0.3 1.5 15.0 3


0.4 1.5 10.0 3



0.7 2.0 10.0 4

6 TOB, TEC pairs 0.6 6.0 30.0 6

Regular 5_2_0 Tracking steps


51


Transverse: muon sample (10 muons/event), zero pileup Generated flat in E and eta (plot vs absolute p and in 4 eta regions) Compare with/without (50PU) dynamic data loss for current detector



52


Longitudinal: muon sample (10 muons/event), zero pileup Generated flat in E and eta (plot vs absolute p and in 4 eta regions) Compare with/without (50PU) dynamic data loss for current detector



53

Primary Vertex Resolution

ttbar sample, zero PU and <PU>=50 Compare with/without (50PU) dynamic data

loss for current detector



54

Tracking Efficiency and Fake Rates

Average tracking efficiencies, high purity, pT>0.9 GeV

Sample Stdgeom Efficiency (%)

Phase 1 Efficiency (%)

Mu PU0 pt0.9,8hit 97.4 ± 0.1 98.1 ± 0.1Mu PU0 pt0.9,8hit, dloss 93.9 ± 0.1 97.9 ± 0.1Mu PU50 pt0.9,8hit 90.1 ± 0.2 94.9 ± 0.1Mu PU50 pt0.9,8hit, dloss

81.5 ± 0.2 94.4 ± 0.1

ttbar PU0 pt0.9 89.6 ± 0.1 93.5 ± 0.1ttbar PU0 pt0.9, dloss 85.6 ± 0.1 93.2 ± 0.1ttbar PU50 pt0.9 84.9 ± 0.1 92.2 ± 0.1ttbar PU50 pt0.9, dloss 79.7 ± 0.1 92.0 ± 0.1Grindelwald,

29/08/12


Tracking Efficiency & Fake rate55

Using Standard validation packages: MultiTrackValidator

Tracking efficiency = #sim trks assoc. to reco trk #sim trks

(for signal sim tracks only)

Tracking fake rate = #reco trks not assoc. to sim trk#reco trks

(for “all” reco tracks)



56

Tracking Efficiency/Fake Rate ttbar sample, current detector

generalTracks, pT > 0.1 GeV/c



57

Tracking Efficiency/Fake Rate ttbar sample, upgrade pixel detector detector




58

Tracking Efficiency/Fake Rate ttbar sample, compare current and upgrade detectors




59

B-tagging Performance (BPIX1 Study)

ttbar, CSV tagger, current detector, <PU>=50, compare BPIX1 ak5PFjets PFnoPU, jet pT > 30 GeV, DUS,b jets

All other layers are at

100% efficiency



60

B-tagging Performance (BPIX1 Study)

ttbar, CSV tagger, upgrade detector, <PU>=50, compare BPIX1 ak5PFjets PFnoPU, jet pT > 30 GeV, DUS,b jets

All other layers are at

100% efficiency



61

Outer Tracker Inefficiency Study

Uniform 20% inefficiency in TIB1,2: zero PU and <PU>=50 ttbar sample, high purity tracks, pT>0.9 GeV/c



62


Uniform 20% inefficiency in TIB1,2: zero PU and <PU>=50 ttbar sample, high purity tracks, pT>0.9 GeV/c Upgrade detector more robust to Outer Tracker inefficiency



63


Dead Modules in Outer Tracker: zero PU and <PU>=50 ttbar sample, high purity tracks, pT>0.9 GeV/c



64


Dead Modules in Outer Tracker: zero PU and <PU>=50 ttbar sample, high purity tracks, pT>0.9 GeV/c



65

HZZ4l analysis


Isolation


66

HZbbll


Di-lepton requirement


67

HZZ4l analysis


More pixel hits for muon tracks


68

HZZ4l analysis


Better IP


69

HZZ4l analysis


Better Z mass resolution


70

SUSY MT2 analysis


Documents

Performance and physics results for Phase1 upgrade