ZHH channel: software and detector performances for ILC

Michele Faucci GiannelliFabrizio Salvatore

Mike Green, Tao Wu

07/02/2007 Michele Faucci Giannelli 2

OUTLINE

• ZHH Channel: summary of reconstruction and software used

• Generator differences

• Tracking performances

• Particle Flow Algorithm performances

• Detector comparison

• First look at background

• Conclusions

This is an update to the LC note LC-PHSM-2007-003

07/02/2007 Michele Faucci Giannelli 3

ZHH Channel• The e+e-→ZHH channel is an excellent

benchmark for many steps of the simulation:– Different generators give different cross sections– Test physics lists available in detector simulation– Requires high performances from all detectors

• Vertexing• Tracking• Clustering

– Thus can be used to test particle flow algorithms– Finally can be used to compare different detector

models

07/02/2007 Michele Faucci Giannelli 4

Generation and Simulation

• Events have been generated using Pandora-Pythia and Whizard

• The reconstruction was performed by Mokka (V06-03p02):– Some information @ generation level:

• ECM = 500 GeV• M(Higgs) = 120 GeV/c2

• Polarized 80% electron beam• Two detector model (LDC00Sc and

LDC01Sc)• Z→l+l- (muons and electrons)

07/02/2007 Michele Faucci Giannelli 5

Marlin Processors

Marlin 0.9.7 + MarlinReco 0.3– Processors used:

• VTXDigi• FTDDigi• TPCDigi• Tracking Processors• PFA Processors• PairSelector• SatoruJetFinder• BosonSelector• MyROOTProcessor

Tracking Processors:– FullLDC:

• LEPTrackingProcessor• SiliconTracking• FullLDCTrackin

– TrackCheater

PFA Processors:– Wolf:

• TrackWiseClustering• Wolf• ClusterMerge

– PandoraPFA:– TrackBasedPFA:

07/02/2007 Michele Faucci Giannelli 6

ZHH selection• Select and extract the two leptons which better

reconstruct the Z.• Combine all the other particles in 4 jets.

– Reconstruct the two Higgs minimizing the quantity:

• Look at different variables to compare the two available PFA algorithms:– D2 and other combinations of jet-jet inv. Mass

• The mass of the Higgs used in the analysis is 114 GeV instead of 120 GeV to take into account the effect of invisible particles

2h342

h122hh mmmmD

07/02/2007 Michele Faucci Giannelli 7

WOLF

Invisible particles

Black: normal reconstructionRed: with neutrinos

On average, contribution from ‘invisible particles’ ~ 6 GeV

Adding this contribution, reconstructed H mass with Wolf higher than mH value used in the

generation. LDC01Sc

PANDORA

07/02/2007 Michele Faucci Giannelli 8

GeneratorsTree generators:– Pandora Pythia– Pandora Phytia with K, , , NOT decayed– Whizard

No visible

difference noticed

07/02/2007 Michele Faucci Giannelli 9

TrackingTwo Tracking algorithms:– FullLDC Tracking– TrackCheater

No visible

difference noticed

Tracking is good enough

07/02/2007 Michele Faucci Giannelli 10

PFA ComparisonThree PFA available:– PandoraPFA– Wolf– TrackbasedPFA

Z→ee

Problem in TrackbasedPFA with particle identification

Cut on Z mass in the Higgs Plot to select good events

Z→mm

07/02/2007 Michele Faucci Giannelli 11

Higgs Mass

Z→ee

Z→Pandora has a very good RMS for muons, probably too low mean

Wolf reconstructs too high mass Higgs

The problem with muon id affects the Higgs reconstruction in TrackbasedPFA

Pandora has some problem with electron id, high energy tail (bremsstrahlung?)

Wolf reconstructs too high mass Higgs

TrackbasedPFA has a very good performance with electrons! Algorithm works!!

07/02/2007 Michele Faucci Giannelli 12

Higgs discrimination: D plot• D is defined as h3412hh mmmD *2

For muons PandoraPFA is the best algorithm

For electrons Pandora and TrackbasedPFA are comparable

Z→ee

Z→

07/02/2007 Michele Faucci Giannelli 13

Higgs discrimination: D2 plot 2h34

2h12

2hh mmmmD • D2 is defined as

Z→ee

Z→

07/02/2007 Michele Faucci Giannelli 14

Detector comparison

LDC00SCLDC01Sc

No differences in the muon case

LDC01Sc is better than LDC00Sc for electron, less material then less bremsstrahlung?

Z→ee

Z→ PandoraPFA

07/02/2007 Michele Faucci Giannelli 15

First look at Background(I)

Z→ee

ZHHZZHZZZ

TrackbasedPFA

PandoraPFA

Pandora is better not only for a more efficient signal reconstruction but for a smaller contamination too

LDC00Sc

07/02/2007 Michele Faucci Giannelli 16

First look at Background (II)

Z→ee

Z→

ZHHZZHZZZ

PandoraPFA

Good discrimination for muons, a factor 2 better than electrons.

ParticleID has a crucial role, more effort are needed!

LDC00Sc

07/02/2007 Michele Faucci Giannelli 17

First look at Background (III)

ZHHZZHZZZ

PandoraPFALDC00Sc

Same as previous slide, linear scale

07/02/2007 Michele Faucci Giannelli 18

Conclusion• Comparison between generators:

– No visible differences between generators

• Comparison between tracking:– FullLDC is as good as cheater for our analysis

• Comparison between PFA: – Trackbased almost as good as Pandora, both need

a better Particle ID.

• Comparison of LDC00/01Sc using PandoraPFA– Small differences once electron ID is solved, both

detector models can be used to reconstruct this channel

07/02/2007 Michele Faucci Giannelli 19

Conclusion• Studies on SM backgrounds

– ZZH and ZZZ have been simulated and reconstructed:

it is possible to discriminate the signal!!

• Future Plans– Study high cross section channels

• Understand how to apply cut at generation level to reduce the amount of events to simulate in Mokka

– Move to 6 jets analysis• B tagging is necessary• Looking at a new strategy for Z selection

07/02/2007 Michele Faucci Giannelli 20

Backup slides

07/02/2007 Michele Faucci Giannelli 21

Preparation: calibration

• Check calibration for pions and electrons

Black PandoraRed Wolf

LDC01Sc

Pions

Electrons

GeV

GeV