14/01/2017 IAS@HKUST 1

Status & Updates from CEPCSimulation – Detector optimization

Manqi RUAN

On behavior of the CEPC-SPPC Study group

14/01/2017 IAS@HKUST 2

Higgs program at CEPC

Observables: Higgs mass, CP, σ(ZH), event rates ( σ(ZH, vvH)*Br(H->X) )

Derive: Higgs width, branching ratios & absolute value of coupling constants

14/01/2017 IAS@HKUST 3

CEPC detector: PFA oriented

High Precision VTX close to IP: b, c, tau taggingHigh Precision & light Tracker:

PFA oriented Calorimeter: Tagging, ID, JER, etc

Reconstruct ALL the physics objects(lepton, γ, tau, Jet, MET, ...) with high

efficiency/precision

14/01/2017 IAS@HKUST 4

Detector Design & Optimizations

CEPC_v1, Conceptual detector designed from ILD... Multiple IP - New ideas are always welcome

14/01/2017 IAS@HKUST 5

Highlight 1: Reconstruction

LEPTON PHOTON

JET/MET FLAVOR

Reconstructed by Arbor Particle Flow Algorithm, 1403.4784

14/01/2017 IAS@HKUST 6

Applied to CEPC Higgs analysis

Now: ~50 independent analyses at Full Simulation level

High light in Exotic Full Simulation study – See Weiming's talk

14/01/2017 IAS@HKUST 7

Br(H→WW)H→WW/ZZ: Portal to Higgs width & perfect test bed for detector/reconstruction performance...

● Br(H→WW), Combined accuracy ~ 1.0% from 13 independent full simulation analyses

– 1.45% at llH, H→WW*→ inc channels, 12 independent channels.

– ~ 1.7% at vvH, H→WW*→ 4q channel (Preliminary. ILC extrapolation = 2.3%)

– 2.3% at qqH, H→WW*→ 2qlv channel (extrapolated from ILC full simulation)

– Combined: 1.0%

14/01/2017 IAS@HKUST 8

A preliminary validated CEPC Delphes Card

● https://github.com/delphes/delphes/blob/master/cards/delphes_card_CEPC.tcl

Thanks Michele & Delphes team!

μμH, H->inclsive LEPTONS

qqH, H->invisibleJET/MET

vvH, H->WW* LEPTONS + JET/MET

14/01/2017 IAS@HKUST 9

Highlight 2: Optimization● Collision environment is very different from Linear Collider

● Requirement study

– Needed accuracy for beam energy calibration for Higgs/Z pole?

– Needed accuracy for Luminosity measurement/monitoring?

● Feasibility study

– Can TPC be used for Z pole operation?

– Can calorimeter works without active cooling?

– Can beam energy be monitored to 1e-5?

● Optimization study

– B Field, Tracker Size,

– ECAL/HCAL Layout & option

– Flavor tagging Performance VS VTX layout (inner radius/pixel size)?

14/01/2017 IAS@HKUST 10

Feasibility of active cooling free Calo● Granularity: No power pulsing at CEPC

– Wi Active cooling: + 2mm thick cooper per active layer...

– Wo Active cooling: reduce the granularity by ~ 1 order of magnitude (inconsidering Electronics/Sensor progress...)

– Performance:

● Lepton ID: No impact for isolated charge object with En > 2 GeV● Physics benchmarks:

– Z→di lepton, Higgs to everything

– Z→vv; H→WW→lvqq; – Z→vv; H→ZZ→llqq;

14/01/2017 IAS@HKUST 11

Leptons: essential for Higgs program

ll vv qq Z bosondecay

Final state

qq,gg

ττ

WW, ZZ,Zγ

Higgs

μμ, γγ

● Key objective: Identify the initialleptons

– Leptons generated in Zdecays in ZH events

– Electrons in Z fusions

● Secondary: leptons generated inHiggs decay

– H->WW/ZZ/tautau/μμ

– H->bb, cc->leptonic decay

– Hadron decays

14/01/2017 IAS@HKUST 12

Generic Lepton ID for Calorimeter with Highgranularity (LICH)

DanYu: BDT based

Performance close to physics limit

Stable performance even reduce Granularity bytwo orders of magnitude: only slightly degradedat low energy

10 mm Cell Size 60 mm Cell Size

14/01/2017 IAS@HKUST 13

LICH @ llH events

Original Design (5 mm Cell) and test Geometries: 10 - 20 mm Cell

High energy leptons identification: high efficiency & purity (limited by shower overlap)

More stringent requirement arises from jet leptons and pi0 reconstruction...

14/01/2017 IAS@HKUST 14

Lepton + Jets: Br(H→WW)

Br(H→WW) via vvH, H→WW*→lvqq

No lose in the object level efficiency; JER slightly degraded, ~ 5/10% at 10/20 mm(ill. behaviors: stay to be tuned)

Over all: event reco. efficiency varies ~1%

Simu. Recon. Efficiency

CEPC_v1 2885 2783 96.5%

TG1 2878 2814 97.8%

TG2 2878 2807 97.5%

TG1: E30L_H48L_10mm, TG2: E30L_H48L_20mm

15 – 95 GeV

14/01/2017 15

Separation: the price to pay!!

10 mm (2 cell)Eff = 10%

25 mm (5 cell)Eff ~ 80%

● Significant degrade – affects τ, pi0 reconstruction

● At cell size > Moliere radius: separation distance ~ 2-3 cell size

● Pi0 at 50 GeV: minimal separation between photons ~ 10 mm

– Smaller cell in ECAL first layers...

...Prelim

inary...

14/01/2017 IAS@HKUST 16

Feasibility of TPC

Trajectory of Track & Primary Ion

Trajectory of the Back Flow Ions =

Track Image formed by Back Flow Ion

HV Plane

IP

Endcap

Shift speed of electron: 80 km/s; Shift speed of ion: 5m/s

● Suggested by International Advisory Committee

● Limited by

– Voxel occupancy &

– Charge Distortion Induced by ions: Mainly from Ion back flow

14/01/2017 IAS@HKUST 17

At CEPC Z pole (with 2E34 integrated luminosity):600 quasi 2D-projection of Z->qq events shift

towards HV plane

Trajectory of the Back Flow Ions =

Track Image formed by Back Flow Ion

HV Plane

IP

Endcap

Trajectory of Track & Primary Ion

...

Each slides is corresponding to the projection of one Z->qq events...

18

60 Million Hits; 4 M Hits for 600 Events ~ 1 second of data taking at CEPC Z pole (L = 2E34)Voxel number for 1 second: 1.3E14 (40 MHz * 1.67M Channels * 2 endcaps); Occupancy at inner most layer: 2E-7 @ CEPC Z pole with evenly distributed bunch; 2E-5 @ CEPC with PDR/FCCeeMini-jets event not included yet.

Hit Map & Hit Occupancy

14/01/2017 IAS@HKUST 19

Ion density ~ 100 P.I/cm ... According to the reference travel length (Weighted by 1/sin(theta)) and number of P.I per length...

Charge density

K = IBF * GAIN

14/01/2017 IAS@HKUST 20

Charge Distortion

At 1E35, k = 100 & Velocity = 5m/s

Distortion, 1 order of magnitude higher than intrinsic resolution – unacceptable!!

B

Er Distortion

Distortion ~ (1+k)*(L*Dis*Er)/(B*v*E

Z)

~(1+k)*(L/1034cm-2s-1) μm |R = 400mm

14/01/2017 IAS@HKUST 21

Really Unacceptable??● Mitigating options Reduction factor

– Large HV (20 kV -> 40 kV ?) 4

– Large Ion Mobility (5m/s -> 10m/s) 2 - 3

– Shorter TPC Length (2.3 -> 2.0) 1.5

– Larger TPC inner Radius (400 -> 600) 8

– Increasing B Field 1.2

– Ion back flow control (k = 100 -> 5 ) 20

– Combination of these options reduces the distortions up to 3orders of magnitude

● Distortion is predictable, what matters is only its fluctuation

● TPC is mainly used for Track finding (with Silicon devices)

● If k = IBF*GAIN is limited to < 10, TPC is feasible for CEPC and even for FCCee (L = 2e36 cm-2 s-1)

14/01/2017 IAS@HKUST 22

Based on Compton scatteringFor 1 MeV accuracy (1E-5)Recommend to place in the last Dipole

Photon Energy range OK Angular uncertainty of laser & emittance of beam OKCalibration OKPhoton productivity 1E7/sSeparate from SR background: Pretty Challenge Device stability ?

*

Ge CaloDP

IP

14/01/2017 IAS@HKUST 23

Summary● CEPC full simulation studies

– Well supported by the reconstructions, Delphes card validated

– Higgs coupling analysis – a complete version

– To focus on EW, Higgs differential & Exotic analyses

– To iterated with Detector optimization

● Conclusions achieved for feasibility studies

– Active cooling free Calorimeter? Yes!

● Cell Size ~ 20 mm & Smaller cells in first few layers for pi0 reconstruction

– TPC for Z pole? Require IBF*Gain < 10 Preliminary but Yes!

– 1 MeV Beam energy calibration with Compton scattering? Promising...

● Require detailed error estimation & controls

● Lots of optimization study, lots of fun,

– Tremendous efforts needed to converge to an optimized design

– Analyses + geometry changes + hardware study + reconstruction software

14/01/2017 IAS@HKUST 24

CEPC Workshop

19-21 April 2017Central China Normal

University

http://indico.ihep.ac.cn/event/6433/

8-10 Nov 2017

14/01/2017 IAS@HKUST 25

Backup

14/01/2017 IAS@HKUST 26

Minijet – incl. Xsec ~ 400nb @ 3 TeV CLIC Collision

LCD-2011-020

14/01/2017 IAS@HKUST 27

14/01/2017 IAS@HKUST 28

Key R&D issues● Beam Energy Calibration

● MDI:

– Partial double ring

– L* optimization & beam background

● VTX & Main Tracker

– Large area silicon detector R&D

– TPC feasibility studies:

● Hit occupation; ● Ion feedback-charge distortion; ● dEdx

● Calorimeter & B-Field

– Less demanding in Jet Energy Resolution + no power pulsing

14/01/2017 IAS@HKUST 29

● Foreword: The design of MDI is not finalized yet.

● Scan over inner radius of VTX: 16 mm → 8 – 20 mm

Vary the VTX inner Radius

14/01/2017 IAS@HKUST 30

● With 8 – 20 mm VTX Inner radius, very good b-tagging

– At efficiency ~ 80%: almost reject all the light background & only 8-10% c-jetsmisidentified as b-jets (Purity ~93-96% at Z to qq events).

B-tagging @ different radius

light background c background

1- 1-

Gang Li

14/01/2017 IAS@HKUST 31

C-tagging @ different radius

● At the same purity: C-tagging efficiency could be improved by ~ 10% byreducing the inner radius from 20 mm to 8 mm...

light background b background

1- 1-

14/01/2017 IAS@HKUST 32

Br(H→ZZ)

● Br(H→ZZ), explored at 18 different channels with full simulation (llvvqq, 4lqq, ll4q, 2l4v)

– 8 Channels has individual accuracy better than 25%: Combined accuracy ~ 5.4%

– 8 with accuracy worse than 25 - 50%

– 2 with accuracy worse than 50% (llH, H→ZZ→4q and vvH, H→ZZ→llvv)

– If electron id efficiency ~ muon id: 4.8%

– If tau finder (used for veto) is mature: ??

– TLEP extrapolation: 4.3%

14/01/2017 IAS@HKUST 33

Lepton + Jets: Br(H→ZZ)

Br(H→ZZ) via vvH, H→ZZ*→llqq

Over all event reco. efficiency reduced ~2%

Events Recon. Efficiency

CEPC_v1 4143 3957 95.5%

TG2 808 754 93.3%

14/01/2017 IAS@HKUST 34

Highlight 2: Optimization● Collision environment is very difficult from Linear

Collider

● Lower E: Smaller Detector & B Field:

– mH & σ(ZH) at μμH: accuracy reduced by 20%/3% with

25% smaller Radius

● Without Power pulsing or Active Cooling

– Granularity reduced ~ 1-2 order of magnitudes

– Br(H->WW/ZZ) & Higgs recoil analysis@llH: eventreconstruction efficiency reduced by ~2%

● Different technology options:

– ECAL: Scintillator vs. Silicon Sensor compared

– Tracking: Full Silicon Tracker - TPC

14/01/2017 IAS@HKUST 35

ECAL Saturation/Linear Range Study

50 GeV Photon Cluster at ECAL with 10 mm Cell Size

~o(1k) hits, hottest hit with E ~ 1k MIP. T.Takeshita, ILDDET@KEK

Scintillator: MIP→Photon→P.E

14/01/2017 IAS@HKUST 36

Impact on H→γγ measurement

Mean σ/M

ECAL Linear Ranger: recommended to be >1k/1.8k MIP (for 10/20 mm Cell) 10k pixel SiPM readout is very challenging (If Photon generation > 10 per mip)

Empirical formula on needed ranger of a single photon:

log10(Ranger) = 0.87*x + 0.97*y - 0.24*y2 + 1.26x = log10(E), y = log10(Cell Size/cm) Shuzheng Wang

14/01/2017 IAS@HKUST 37

● Arbor: shower = tree...

– Lepton

– Photon

– Separation & Composed objects

– JET/MET

–

– Application on Higgs analysis & detector optimization

– Key tech. Issues...

14/01/2017 IAS@HKUST 38

CEPC_v1Forward Region & Yoke Thickness

Modified w.r.t ild_o2_v05

Used for CEPC Higgs analysis

Simplified, Defect free geometry... Cylinder like calorimeter layers,

& Silicon tracker (Optional)

Used for Arbor tuning, Calorimeteroptimization & Conceptual SPPC

Detector study...

Geometries

14/01/2017 IAS@HKUST 39

Photon construction at CEPC_v1

Dedicated EM Shower Energy Estimator developed,Correct the geometry effect

14/01/2017 IAS@HKUST 40

At defect free geometry

● Performance looks better atlarge cell size:

– Pattern of Arbor clustering:lose of isolated hits

– Small cell can always bemerged into large ones...

● Performance degrading with

– Geometry defects: 20% aftercorrection

– In-Homogeneity: 13%.induced by 20%inhomogeneity

– Dead zone: 8% by 1 mmDead zone along 45 mm cellsize

Calibration constant not finalized...

14/01/2017 IAS@HKUST 41

Separation

2 cm

10 cm

Tiny inefficiency: bridging effect by fragments

14/01/2017 IAS@HKUST 42

Arbor @ pp collisions

140 PU event reconstructed at ~1 min/evtWith KD-tree algorithm, complexity reduced

From N2 to Nlog(N)

Separation & Speed

14/01/2017 IAS@HKUST 43

Composed object: converted photon

Save ~ 7% of the H->di photon

statistic

14/01/2017 IAS@HKUST 44

Composed object: π0 (Preliminary)

MCTruth Reconstructed

Testing on Higgs to di tau events. Tau inclusive decay (X axis, Energy of Pi0, Y axis, Angle between two photons decayed from pi0)

For Tau with En > 3 GeV && En < 30 GeV, Reconstruction efficiency ~ 65%..Horizontal line corresponding to 30 mm separation at ECAL.

14/01/2017 IAS@HKUST 45

Tau reconstruction

● In no-jet environment: counting number of charged particle – (pions & leptons), photons (pi0s) +restrict impact parameters leads to very high efficiency in Tau finding:

– At inclusive Higgs decay sample: Efficiency ~ 98% for of H→ττ event finding, with llH andvvH final state. The remaining bkgrds are irreducible: H→WW/ZZ→leptonic/tau final state

– In μμH channel: δN/N = 3%

14/01/2017 IAS@HKUST 46

JER/MET

H→gg, Z→inv H→inv, Z→qq

● Digital ECAL mode: Energy Estimated as k*NHit for HCAL Cluster, Calibration Constant (k)optimized for both Pandora & Arbor via Scan

● MET: usually no ambiguity;

● Jet: Highly depending on Jet clustering if #Jet > 2...

14/01/2017 IAS@HKUST 47

To improve the JER...● Key issues

– Splitting

– Hit collection efficiency at low energy for low energyneutral particle

● Missing elements

– Identification of Fragments & Core Clusters

● Aim at 1-1 correspondence between Core Cluster& Incident particle with Energy > Threshold

– Energy Estimation & Software compensation

– Isolated hits absorption

Leading Cluster

All Clusters

All Hits

Neutrons at defect free geometry

NHit(LC)/NHit(All Clusters)

Nhit(All Clusters)/NHit(All Hits)

14/01/2017 IAS@HKUST 48

Spatial

Timecorrelation

...

NeutronKinematic energy = 1 GeV

V = 0.86 c

Kinematic energy = 50 GeV

Bless/Curse?...

14/01/2017 IAS@HKUST 49

Arbor...– Lepton: eff/purity ~ 99% for initial leptons identification in Higgs events 95%

● To be extended to Particle identification (Pi-Kaon, Photon-Neutral hadron)

– Photon: Impact of various effects on γ reconstruction & H→γγ studied 90%

– Separation (pi & photon efficiently separated at Dis > 30 mm) 80%

– Composed object reconstruction 50%

● Tau finding in qqH, H→tautau, etc...

– Jet Energy Resolution 70%

● Dedicated Identification of Fragments & Core Cluster● Energy Estimator/Software Compensation for Hadrons & Isolated hit absorption

– KD algorithm enhanced, speed ~ o(Nlog(N))

– Fully Tested at Higgs analyses

Detector Optimization:

● Active cooling free calorimeter leads to a degrading of event reconstructionefficiency about 1-2% (at Higgs recoil, Br(H→WW & ZZ) measurements), butwould be good to evaluate at Physics analysis with tau in the final states...

●

– Long to do list & Needs lots of efforts...

14/01/2017 IAS@HKUST 50

Reconstructed with Arbor

1 TeV hadronic shower at 20 PU: Perturbation from PU ~ 0.1% level

Chasing: eff, δE(E, PID, #PU, η)