ILD Detector Optimization and Benchmarking
description
Transcript of ILD Detector Optimization and Benchmarking
ILD Detector Optimizationand
BenchmarkingAkiya Miyamoto, KEKat Tsinghua University
12-January-2009
2Seminar at Tsuingha Univ., 12-Jan-2009
ILD Introduction ILD origins in the European and Asian based Large Detector
study.
Akiya Miyamoto, KEK
LDC GLD
ILC Reference Design (RDR) in 2007 GLD Detector Outline Document (DOD) arXiv:physics/0607154 LDC DOC http://www.ilcldc.org/Common feature: Tracker(Pixel & Silicon & Gas) + PFA calorimeter
+ …
At LCWS2007, we agreed to work together for a joint LOI GLD (B=3T, RECAL=2.1m) + LDC(B=4T, RECAL=1.6m ) ILD
3Seminar at Tsuingha Univ., 12-Jan-2009
International Large Detector ( ILD ) LDC and GLD had a common future;
Pixel vertex detector placed very close to the beam pipe. Gaseous tracker, TPC, for highly efficient and precise track
measurements,supplemented by silicon trackers.
EM and HD calorimeters are placed inside a solenoid field and read our by very small sensors to achieve a good energy measurement by Particle Flow Analysis (PFA).
Akiya Miyamoto, KEK
But in detail: B=4 Tesla(LDC) vs 3 Tesla(GLD) ECAL radius: 1.6m(LDC) vs 2.1m(GLD) Sub Detector technologies ….
Simulation studies of physics performances are used to reach agreement of detector parameters.
4Seminar at Tsuingha Univ., 12-Jan-2009
How we optimize Optimization tools
GLD Jupiter/Sattelites, LDC Mokka/MarlinReco intermediate detector models were introduced for comparison
GLDPrim by Jupiter, and LDCPrim by Mokka, both having B=3.5T and RECAL=1.85m.
Akiya Miyamoto, KEK
Performances have been studies as a function of major parameters.
Reached a consensus on the ILD reference detector for LOI benchmark studies at Cambridge (Sep. 2008).
Physics performance studies have been performed based on ILD model.
5Seminar at Tsuingha Univ., 12-Jan-2009
Jupiter/Satellites for Full Simulation Studies : GLD
JUPITERJLC Unified
Particle Interactionand
Tracking EmulatoR
IOInput/Outputmodule set
URANUS
LEDA
Monte-Calro Exact hits ToIntermediate Simulated output
Unified Reconstructionand
ANalysis Utility Set
Library Extention for
Data Analysis
METISSatellites
Geant4 basedSimulator
JSF/ROOT basedFramework
JSF: the analysis flow controller based on ROOT The release includes event generators, Quick Simulator, and simple event display
MC truth generator Event Reconstruction
Tools for simulation Tools For real data
Akiya Miyamoto, KEK
6Seminar at Tsuingha Univ., 12-Jan-2009
Mokka− Mokka is a full simulation using Geant4 and a realistic description of a detector for the future linear collider.
− Home page: http://polzope.in2p3.fr:8081/MOKKA Mokka is now a part of the ilcsoft, http://ilcsoft.desy.de/portal/software_packages/
LDC
Akiya Miyamoto, KEK
ex. ECAL structure
− Detector Geometry: − managed by MySQL data base and
CGA (Common Geometry Access) API.− LDC and other variants are prepared
and used for ILD optimization.− Implementation of detailed detector
model based on engineering studies is in progress.
PandoraPFALCFIVertex
8
GLD + LDC Combined Framework
Akiya Miyamoto, KEK Seminar at Tsuingha Univ., 12-Jan-2009
Whizard Physsim
StdHep
MOKKA Jupiter
LCIO
Marlin Sattelites
LCIO
DST and Analysis
LDC GLDStdHep: Same generator dataLCIO: Common IO format GLDPrim/LDCPrim: Similar detector model
LCIO helps to collaborative works for detector optimization
After the LOI, two frameworkswill be merged to a single framework.
9Seminar at Tsuingha Univ., 12-Jan-2009
Detector Parameters for Opt. studies
GLD/GLDPrim/J4LDC prepared for Jupiter LDC/LDCPrim/LDCGLD prepared for Mokka Physics performance was compared between different
geometries
Akiya Miyamoto, KEK
Jupiter MokkaGLD GLDPri
mJ4LDC LDCGLD LDCPri
mLDC
B(T) 3.0 3.5 4.0 3.0 3.5 4.0VTX Rmin (cm) 1.75 1.6 1.5 1.65 1.50 1.4# VTX layers 3 x double super layers 5 layers# IT layers 4 layers 2 layersTPC Rmin(cm) 43.7 43.5 34.0 37.1ECAL Rmin(cm) 210 185 160 202 182.5 161HCAL Thick. (Int.L)
6.79 6.29 5.67 5.86Geometryiesin Mokka and Jupiter are similar, but there are many small differencesin geometry and assumed detector technologies
10Seminar at Tsuingha Univ., 12-Jan-2009
Pt resolution
Akiya Miyamoto, KEK
LDC : ~5% worse at high Pt Shorter Lever armGLD/GLD’: ~10%worse at low Pt Lower B
Single muon, produced at cosq=0. by Jupiter+Satellites: TPC+IT+VTX fitting
11Seminar at Tsuingha Univ., 12-Jan-2009
GLDPrim - LDCPrim
Akiya Miyamoto, KEK
LDCPrim(Mokka+Pandora) is better than GLDPrim(Jupiter+Sattelites) by 15~30%.
Possible source: srf(IT) 4mm(LDCPrim) 10mm(GLDPrim) Silicon External Tracker in Mokka
3x10-5
Sub-detector technology is more important than geometry
4mm
12Seminar at Tsuingha Univ., 12-Jan-2009
GLDPrim vs LDCPrim (srf(IP))
Akiya Miyamoto, KEK
GLDPrim is better than LDCPrim ;
3 double layers vs 5 layers ?
Fast sim. study by M.Berggren
srf=sZ=2.8mm
13Seminar at Tsuingha Univ., 12-Jan-2009
kaon_0L Energy Resolution
Akiya Miyamoto, KEK
Hadron Model: LCPhysics
HCAL response is not smooth around 13 GeV- LE/HE behaviour
ECAL resolution: same
14Seminar at Tsuingha Univ., 12-Jan-2009
Jet measurement: Particle Flow Analysis
Akiya Miyamoto, KEK
PFA: Charged particles by Tracker Neutral particles by Calorimeter, remove charged particle energies
Performance studies depend on shower simulation; longitudinal, lateral, and tof distribution, neutron response, etc.
15Seminar at Tsuingha Univ., 12-Jan-2009
Jupiter data analyzed by PandoraPFA
Akiya Miyamoto, KEK
Ejet(GeV)
Z pole uds-pair events:GLDPrim
Pandora PFA: Sophisticated algorithm tuned to Geant4 shower shape has achieved the performance goal of ILC, DE/E ~ 30%/√E
Jet Energy Resolution
16Akiya Miyamoto, KEK Seminar at Tsuingha Univ., 12-Jan-2009
Same trend is seen by analysis of Jupiter models, though performance is slightly worse than Mokka model
17Seminar at Tsuingha Univ., 12-Jan-2009
ECAL Seg. and HCAL thickness
ILD: ECAL+HCAL= 6.8 Int. L.(48layers)6.8 Int. L look OK, but worse resolution is seen for 90o jets.
by M. THomson
−Performance is strong function of ECAL seg. size. −2x2cm2 too large, 1x1cm2 would be ok for jets with E < 100 GeV
Akiya Miyamoto, KEK
18Seminar at Tsuingha Univ., 12-Jan-2009
Physics Benchmark Studies ILC goal precise studies of Tera scale physics.
Akiya Miyamoto, KEK
Signal processes: the minimum set. Recoil mass measurement by e+e- ZH e+e-/m+m- + H H c cbar decay in e+e- ZH process e+e- t tbar 6 jets and t (tbar) charge ID for AFB meas. e+e- t+ t- and t pol. measurement. Separate WW and ZZ in Chargino/Neutrino pair production process
According to the request by ILC Research Director (RD) and International Detector Advisory Group(IDAG), simulation studies for LOI should based on a realistic Monte Calro program based on a realistic reconstruction program include backgrounds by physics processes and those caused by
accelerator.
19Seminar at Tsuingha Univ., 12-Jan-2009
Higgs recoil mass meas. e+e - ZH e+e-X / m+m-X , Ecm=250 GeV, 250 fb-1 Analysis.
Select e+e- / m+m - consistent with Z and study recoil mass Precise track meas. is a key for
Akiya Miyamoto, KEK
GLDprim case, with backgound
e+e - Channel m+m- channel
Compare 3 geometries
m+m- X
Differences are small.
by Itoh Kazutoshi
20Seminar at Tsuingha Univ., 12-Jan-2009
Benchmark study: Example
Akiya Miyamoto, KEK
0 01 1
0 0 0 02 2 1 1
e e W W W W
e e ZZ
+ - + - + -
+ -
Using several detector models, performance to separate W/Z in jet mode have been studied using SUSY processes
by Taikan Suehara
21Seminar at Tsuingha Univ., 12-Jan-2009
e+e- t+t-, t rn
Akiya Miyamoto, KEK
t is polarized probe New Phyaics
SMSM+NP
by Taikan Suehara
22Seminar at Tsuingha Univ., 12-Jan-2009
ILD reference detector model At 2nd ILD WS at Cambridge, we agreed to created the new
model,ILD reference design model for LOI, in Mokka:
Model parameters, B=3.5 Tesla Rin ECAL=185cm, TPC: halfZ=230cm VTX three double layers. Silicon trackers: ( SIT, FTD, SET, SOT) Calorimeters (ECAL 22X0, 0.5x0.5cm, HCAL ) ….
Akiya Miyamoto, KEK
for the sake of simulation, some detector technologies are assumed in Mokka. But as ILD, many detector technologies are open and not selected at the time of LOI.
By the time of LOI, we have no time to merge Jupiter/Sattelites and Mokka/Marlin framework. A work to merge two framework for “ILD Software” will come after LOI.
CAD Model
Mokka model
25Seminar at Tsuingha Univ., 12-Jan-2009
ILD_00 MC/DST production ILD performance are expected to be similar to
GLDPrim/LDCPrim. But for consistent and complete study, new MC&DST production has been lunched with an improved software.
Akiya Miyamoto, KEK
Started since Dec. last year, using GRID Goal: 250 fb-1 @ 250 GeV, 500 fb-1@500 GeV, Signal + SM
background StdHep (@SLAC) Sim(Mokka), reconstruction and DST maker. DST contents:
lcio format contains : Tracks, PFOs, [23456]-Jets,
LCFIVertex, MCParticls,.. Production profile:
Typical CPU time: ~0.5 min.(mm) to 4min.(6f) Typical event size ( for uds-pair @ 500 GeV )
Sim. ~950kb, Rec.~1800kb, DST ~ 23kb
26Seminar at Tsuingha Univ., 12-Jan-2009
ILD_00 MC/DST production
Akiya Miyamoto, KEK
Ecm
Signal Events NEvents
L [1/fb]
250 Sig1: ZH, Zee/mm 105k 4624Sig2: ZH, Znn, Hqq 194k 1000Sig3: ZH, Zqq, Hqq 567k 1000
500 Sig4: ee tt 2385k 517Sig5: eettbbqqqq 1012k 3737Sig6: ee+
1+1, 0
202 137k 678
SM (250GeV)
NEvents
L [1/fb]
2f (w/o ee) 1314k 3.84f 10631k 7726f 200k 3753kee 95k 0.0014gg X 0 0egeg 0 0nn+ng 0 0gg+ng 0 0
SM (500GeV)
NEvents
L [1/fb]
2f (w/o ee, tt) 760k 14.34f 2610k 42.46f (w/o bbqqqq)
624k 1221
ee 0 0.0gg X 544k 3560egeg 0 0nn+ng 0 0gg+ng 0 0
@ last weekRough summary: ~ 23M eventsO(50) TB Sim/Rec. files~ 0.5 TB DSTs so far
Production continues
27Seminar at Tsuingha Univ., 12-Jan-2009
GRID for MC production GRID provides
Huge CPU and storage resources A way to communicate world wide
VO ILC is hosted by DESY, based on LHC Computing GRID MC production and production are running on GRID Simulated, Reconstructed, and DST are placed on GRID.
DST: 20~50 MB x O(10k) files or more.
In Japan, Replications of DST to KEK/Tohoku/Kobe U.
sites are in progress in parallel to the production. Resources in KEK will be increased in near
feature.
Akiya Miyamoto, KEK
Grid
KEK
DESYIP2P3
UK
more …
28Seminar at Tsuingha Univ., 12-Jan-2009
Summary ILD has been optimized
ILD MC and DST production is in progress, for performance studies of LOI 3rd Workshop will be held at Seoul in Feb 16-18, LOI is due March 31. Presented at TILC09 ( 17-21, April ) Detector TDR phase will follow.
Many physics channels are yet to be analyzed. Your participations are welcomed.
Akiya Miyamoto, KEK
Backup Slides