The PFA Approach to ILC CalorimetryThe PFA Approach to ILC Calorimetry

Steve Magill

Argonne National Laboratory

Multi-jet Events at the ILCPFA ApproachPFA Goals

PFA Requirements on ILC Calorimetry

PFA-motivated Detector Models

PFA Results

Detector Optimization with PFAs

Summary

Precision Physics at the Precision Physics at the ILCILC

ZHH500 events

• e+e- : background-free but can be complex multi-jet events

• Includes final states with heavy bosons W, Z, H

• But, statistics limited so must include hadronic decay modes (~80% BR) -> multi-jet events

• In general no kinematic fits -> full event reconstruction

Parton Measurement via Jet Reconstruction

From J. Kvita at CALOR06

Cal Jet -> large correction -> Particle Jet -> small correction -> Parton Jet

The Particle Flow Approach to Jet ReconstructionPFA Goal : 1 to 1 correspondence between measured detector objects and particle 4-vectors -> best jet (parton) reconstruction (energy and momentum of parton)

-> combines tracking and 3-D imaging calorimetry :

good tracking for charged particles (~60% of jet E)-> p (tracking) <<< E for photons or hadrons in CAL

good EM Calorimetry for photon measurement (~25% of jet E)

-> E for photons < E for neutral hadrons-> dense absorber for optimal longitudinal separation of photon/hadron showers

good separation of neutral and charged showers in E/HCAL-> CAL objects == particles-> 1 particle : 1 object -> small CAL cells

adequate E resolution for neutrals in HCAL (~13% of jet E)-> E < minimum mass difference, e.g. MZ – MW

-> still largest contribution to jet E resolution

fluctuations

Jet Energy Resolution – “Perfect” PFA

Dilution factor vs cut :integrated luminosity equivalent

• Want mZ - mW = 3σm (dijets) -> dijet mass resolution of ~3.5 GeV or 3-4% of the mass

• Better resolution is worth almost a factor 2 in useable luminosity – or running cost

/M ~ 3%

Dijet masses in WW and ZZ events

PFA Goal – Particle-by-Particle W, Z ID

/M ~ 6%

102 GeV Higgs?!!

PFA Goal – Jet Energy Resolution Dependence

For a pair of jets have:

Assuming a single jet energy resolution of

+ term due to 12 uncertainty

For a Gauge boson mass resolution of order

(Ej) < 0.03 Ejj(GeV)

~3%

Ejj/GeV (Ej)

91 < 29 %

200 < 42 %

360 < 57 %

500 < 67 %Jet energy resolution requirement depends on energy

tt event at 500 GeV-

PFAs and Detector DesignPFAs and Detector Design

PFA key -> complete separation of charged and neutral hadron showers

-> hadron showers NOT well described analytically, fluctuations dominate -> average approach -> E resolutions dominated by fluctuations-> shower reconstruction algorithms -> sensitive to fluctuations on a shower-by-shower basis -> better E resolution - PFA approach

Calorimeter designed for optimal 3-D hadron shower reconstruction :

-> granularity << shower transverse size (number of "hits")-> segmentation << shower longitudinal size ("hits")-> digital or analog readout?-> dependence on inner R, B-field, etc.

uses optimized PFA to test detector model variations

-> Need a dense calorimeter with optimal separation between the starting depth of EM and Hadronic showers. If I/X0 is large, then the longitudinal separation between starting points of EM and Hadronic showers is large

-> For electromagnetic showers in a dense calorimeter, the transverse size is small

-> small rM (Moliere radius)-> If the transverse segmentation is of size rM or smaller, get optimal transverse separation of electromagnetic clusters.

ECAL Requirements for Particle-FlowECAL Requirements for Particle-Flow

MaterialMaterial II (cm) (cm) XX00 (cm) (cm) II/X/X00

WW 9.599.59 0.350.35 27.4027.40

AuAu 9.749.74 0.340.34 28.6528.65

PtPt 8.848.84 0.3050.305 28.9828.98

PbPb 17.0917.09 0.560.56 30.5230.52

UU 10.5010.50 0.320.32 32.8132.81

MaterialMaterial II (cm) (cm) XX00 (cm) (cm) II/X/X00

Fe (SS)Fe (SS) 16.7616.76 1.761.76 9.529.52

CuCu 15.0615.06 1.431.43 10.5310.53

Dense, Non-magnetic Less Dense, Non-magnetic

. . . use these for ECAL

conconee

mean mean (GeV)(GeV)

rmrmss

//meanmean

22

.025.025 1.921.92 1.41.444

.78.78 9.369.36

.05.05 2.942.94 1.31.399

.41.41 4.294.29

.075.075 3.593.59 1.21.288

.31.31 2.422.42

.10.10 4.014.01 1.21.233

.25.25 2.352.35

.25.25 4.644.64 1.31.300

.23.23 2.702.70

.50.50 4.774.77 1.21.299

.23.23 2.502.50

.75.75 4.794.79 1.21.288

.23.23 2.412.41

1.001.00 4.804.80 1.21.288

.23.23 2.402.40

conconee

mean mean (GeV)(GeV)

rmrmss

//meanmean

22

.025.025 2.072.07 1.61.622

.79.79 10.6110.61

.05.05 2.962.96 1.61.666

.51.51 4.514.51

.075.075 3.633.63 1.51.566

.38.38 2.742.74

.10.10 4.084.08 1.41.488

.31.31 2.562.56

.25.25 4.764.76 1.41.444

.25.25 2.492.49

.50.50 4.854.85 1.41.433

.25.25 2.422.42

.75.75 4.864.86 1.41.422

.25.25 2.252.25

1.001.00 4.874.87 1.41.422

.25.25 2.452.45

SS W

rms

cone

Energy in fixed cone size :-> means ~same for SS/W-> rms ~10% smaller in W

Tighter showers in W

. . . dense HCAL as well?-> 3-D separation of showers

Single 5 GeV Single 5 GeV

HCAL Requirements for Particle-FlowHCAL Requirements for Particle-Flow

Digital HCAL?

Analog linearity Digital linearity

GEANT 4 Simulation of SiD Detector (5 GeV +)-> sum of ECAL and HCAL analog signals - Analog-> number of hits with 1/3 mip threshold in HCAL - Digital

Analog Digital

Landau Tails+ path length

Gaussian

/mean ~22% /mean ~19%

E (GeV) Number of Hits

Occupancy Event DisplayOccupancy Event Display

All hits from all particles

Hits with >1 particle contributing

PFA-Motivated Detector Models

SiD, LDC, GLD in order of Inner Radius size

Common featuresLarge B-field with solenoid outside calorimeterSuppress backgroundsSeparate charged hadronsDense (W), fine-grained (Silicon pixels) ECALHigh granularity and segmented HCAL digital and analog readouts

SiD

Pictures, short descriptions

LDCGLD

SiD Detector ModelSi Strip TrackerW/Si ECAL, IR = 125 cm 4mm X 4mm cellsSS/RPC Digital HCAL 1cm X 1cm cells5 T B field (CAL inside)

PFA Results at Z Pole in SiDPFA Results at Z Pole in SiD

Average confusion contribution = 1.9 GeV <~ Neutral hadron resolution contribution of 2.2 GeV -> closing in on PFA goal

2.61 GeV 86.5 GeV 59% -> 28%/√E

3.20 GeV 87.0 GeV 59% -> 34%/√E

Energy Sum (GeV) Energy Sum (GeV)

PandoraPFA v01-01 LDC00Sc B = 4 T 3x3 cm HCAL (63 layers) Z uds (no ISR) TrackCheater

√s = 91 GeV

√s = 360 GeV √s = 200 GeV

PFA Results on qqbar LDC00Sc

EEJETJET

EE/E = /E = √(E/GeV)√(E/GeV)

|cos|cos|<0.7|<0.7

45 GeV45 GeV 0.2950.295

100 GeV100 GeV 0.3050.305

180 GeV180 GeV 0.4180.418

250 GeV250 GeV 0.5340.534

rms90

PFA Results - Angular Dependence

PandoraPFA v01-01

• Fairly flat until next to beam pipe• Evidence for shower leakage at cos = 0?

RMS = 15.89 GeVRMS90 = 9.632 GeV[66.7%/sqrt(E)]

RMS = 11.44 GeVRMS90 = 8.45 GeV[~59%/sqrt(E)]

Removing eventswith shower leakage

PFA Results : qq at 200 GeV

Barrel Events

RMS = 30.25 GeVRMS90 = 21.4 GeV[~97%/sqrt(E)]

RMS = 43.88 GeVRMS90 = 28.11 GeV[127.%/sqrt(E)]

Removing eventswith shower leakage

• Shower leakage affects PFA performance at high energy• Events with heavy shower leakage could be identified by hits in the muon detectors• Use hits in the muon detectors to estimate shower leakage?

PFA Results : qq at 500 GeV

Z

Z

W,Z Separation in Physics Events with PFA

Detector Optimisation Detector Optimisation Studies Studies

From point of view of detector design – what do we want to know ?

Optimise performance vs. cost

Main questions (the major cost drivers):• Size : performance vs. radius• Granularity (longitudinal/transverse): ECAL and HCAL• B-field : performance vs. B

To answer them use MC simulation + PFA algorithm

!• Need a good MC simulation• Need realistic PFA algorithm (want/need results from multiple algorithms)

These studies are interesting but not clear how seriously they should be taken

how much is due to the detector how much due to imperfect algorithm

Caveat Emptor

4.3 I 5.3 I

HCAL Depth and Transverse HCAL Depth and Transverse segmentationsegmentation Investigated HCAL Depth (interaction lengths)

• Generated Zuds events with a large HCAL (63 layers)• approx 7 I

• In PandoraPFA introduced a configuration variable to truncate the HCAL to arbitrary depth• Takes account of hexadecagonal geometry

NOTE: no attempt to account for leakage – i.e. using muon hits - this is a worse case

HCAL leakage is significant for high energy Argues for ~ 5 I HCAL

1x1 3x3 5x5 10x10

Analogue scintillator tile HCAL : change tile size 1x1 10x10 mm2

“Preliminary Conclusions”

3x3 cm2 cell size No advantage 1x1 cm2

• physics ?• algorithm artefact ?

5x5 cm2 degrades PFA• Does not exclude coarser

granularity deep in HCAL

Radius vs FieldRadius vs Field

100 GeV jets

LDC00Sc

180 GeV jets

Radius more important than B-field

Radius more important at higher energies

B-field studies on the way…

ECAL Transverse GranularityECAL Transverse Granularity• Use Mokka to generate Z uds events @ 200 GeV with different ECAL segmentation: 5x5, 10x10, 20x20 [mm2]

• For 100 GeV jets, not a big gain going from 10x10 5x5mm2

With PandoraPFA

[ for these jet energies the contributions from confusion inside the ECAL is relatively small – need ]

• 20x20 segmentation looks too coarse

100 GeV jets

PFA + Full Simulations -> LC detector design

- unique approach to calorimeter design

- requires generation of e+e- physics processes as well as single particles

- needs good simulation of the entire LC detector

- requires flexible simulation package -> fast variation of many parameters

- huge reliance on correct! simulation of hadron showers

-> importance of timely test beam results!

PFA Reliance on Simulation

PFA Calorimeters in Test Beam – Hadron Shower Model Comparisons

For upgrade we will add the plastic ball detector (GSI, Darmstadt) as a recoil detector. This will help with the tagged neutrons

MIPP Experiment and Upgrade -StatusMIPP Experiment and Upgrade -Status

• MIPP E907 took data in 2005 is busy analyzing 18 MIPP E907 took data in 2005 is busy analyzing 18 million events—Results expected soon.million events—Results expected soon.

• MIPP Upgrade proposal P-960 MIPP Upgrade proposal P-960 – was deferred in October 2006 till MIPP publishes was deferred in October 2006 till MIPP publishes

existing dataexisting data– Obtains new collaborators-(10 additional Obtains new collaborators-(10 additional

institutions have joined)institutions have joined)• MIPP Upgrade will speed up DAQ by a factor of 100 MIPP Upgrade will speed up DAQ by a factor of 100

and will obtain data on 30 nuclei. This will benefit and will obtain data on 30 nuclei. This will benefit hadronic shower simulator programs enormously—hadronic shower simulator programs enormously—See Dennis Wright’s talk at the ILC test beam See Dennis Wright’s talk at the ILC test beam workshop.workshop.

• MIPP Upgrade will provide tagged neutral beams MIPP Upgrade will provide tagged neutral beams for ILC calorimeter usage.for ILC calorimeter usage.

Models (MARS, PHITS, Geant4) Models (MARS, PHITS, Geant4) disagree with each other and data disagree with each other and data when tested on thick Al targetwhen tested on thick Al target

Proton production from thick aluminum target at 67 GeV/c Ratio of calculated and measured cross sections

Energy deposition in cylindrical Energy deposition in cylindrical tungsten thick target-Before and after tungsten thick target-Before and after the Hadronic shower simulation the Hadronic shower simulation workshopworkshop

Before workshop

After workshop

1 GeV protons on 10cm tungsten rod (1cm radius)

10 GeV protons on 10cm tungsten rod (1cm radius)

MIPP Upgrade will acquire 5 MIPP Upgrade will acquire 5 million events/day on the million events/day on the following nucleifollowing nuclei• The A-ListThe A-List

• HH22,D,D22,Li,Be,B,C,N,Li,Be,B,C,N22,O,O22,Mg,Al,Si,P,S,Ar,K,Ca,,Fe,Ni,C,Mg,Al,Si,P,S,Ar,K,Ca,,Fe,Ni,Cu,Zn,Nb,Ag,Sn,W,Pt,Au,Hg,Pb,Bi,Uu,Zn,Nb,Ag,Sn,W,Pt,Au,Hg,Pb,Bi,U

• The B-ListThe B-List

• Na, Ti,V, Cr,Mn,Mo,I, Cd, Cs, BaNa, Ti,V, Cr,Mn,Mo,I, Cd, Cs, Ba

• With this data, for beam energies ranging from With this data, for beam energies ranging from 1 GeV/c-120 GeV/c for 6 beam species, one can 1 GeV/c-120 GeV/c for 6 beam species, one can change the quality of hadronic shower change the quality of hadronic shower simulations enormously—Net conclusion of simulations enormously—Net conclusion of HSSW06.HSSW06.

Tagged neutron and K-long Tagged neutron and K-long beams in MIPP-For ILC Particle beams in MIPP-For ILC Particle flow algorithm studiesflow algorithm studies• MIPP Spectrometer permits MIPP Spectrometer permits

a high statistics neutron and a high statistics neutron and K-long beams generated on K-long beams generated on the LH2 target that can be the LH2 target that can be tagged by constrained tagged by constrained fitting. The neutron and K-fitting. The neutron and K-long momenta can be known long momenta can be known to better than 2%. The to better than 2%. The energy of the neutron (K-energy of the neutron (K-long) can be varied by long) can be varied by changing the incoming changing the incoming proton(Kproton(K++) momentum. The ) momentum. The reactions involved arereactions involved are

pnpp

pKpK

pKpK

pnpp

L

L

0

0

See R.Raja-MIPP Note 130.

Expected tagged neutral rates/day of running

Beam Mom entum Proton Beam K+ beam K- beam Antiproton beam

GeV/c n/day K-Long/day K-Long/day anti-n/day

10 20532 4400 4425 665020 52581 9000 9400 1145030 66511 12375 14175 1350060 47069 15750 14125 1355090 37600

An expression of support from the ILC community will help speed up the approval process

Summary

Much “cleaner” environment for jetsin e+e- collisions than in ppbar - at theLHC, jet E resolution is dominated bycontributions from underlying eventsand final state gluon radiation

Without large UE and FSR contributions, CAL resolution dominates-> An obvious goal – W/Z ID with dijet mass measurement? Current calorimeters - M ~ 9 GeV at Z-Pole PFA potential improvement - M ~ 3 GeV -> in addition to leptonic decay modes, can use >80% of hadronic decays as well

PFAs for LC Detectors?PFAs for LC Detectors?~13% for CMS

PFA vs Compensation? Hardware compensation – for high energy particles (ZEUS 35%/E for > ~3 GeV) but for jets resolution degrades. Software compensation - requires knowledge of the particle type and/or particle energy, + reliance on shower and/or jet models Particle Flow works as long as "mistakes" < E of neutral hadrons (10%)

ppbar -> qqbar -> hadrons + photons -> large calorimeter cells traditional jet measurement

Z -> qqbar -> hadrons + photons = small 3D cal cells

PFA jet measurement

Dijet event in CDF Detector

One jet in Z -> qqbar event in a LC Detector