Jupiter and Friends: Jupiter and Friends Akiya Miyamoto
for LC Simulation Mini Workshop, DESY
through VRVS presentation Based on works done by ACFA-SIM (-J) members Contents: Contents Introduction – Concepts
Few progresses since 7th ACFA at Taipei. I’m sorry if you heard them already
Figures presented here are preliminary. They are included to indicate status of our software packages. Jupiter/Satellites Concepts: Jupiter/Satellites Concepts JUPITER JLC Unified
Tracking EmulatoR IO Input/Output
module set URANUS LEDA Monte-Calro Exact hits To
Intermediate Simulated output Unified Reconstruction and
Set Library Extention
for Data Analysis METIS Satellites Geant4 based
Simulator JSF/ROOT based
Framework 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 Jupiter: Jupiter Features:
Modular structure for easy update, install/uninstall of sub-detectors
Powerful base classes that provide unified interface to
facilitate easy (un)installation of components by methods such as InstallIn, Assemble, Cabling
Help implementation of detailed hierarchical structures.
This helps to save memory size.
Minimize user-written source code by
Automatic naming system & material management
B-field compositions for accelerators
Input : HEPEVT, CAIN (ASCII) or generators in JSF.
Output class allows external methods. Using this mechanism, it can output ASCII flat file and JSF/ROOT fie. : Geant4 based Full Detector Simulator Status: Status Before LCWS2004, GLC-3T geometry except for the muon detector was implemented in Jupiter and used for studies on
Performance of CDC, T0 resolution, etc.
Beam backgrounds on VTX
Implementation of new detector model, Generic Large Detector, has been started. So far, simple ideal geometries are implemented. ( “GLD”, is a tentative name for this presentation )
Components in Satellites are enriched
For implementation of more complete geometry, a XML interface has been developed.
Geant4 6.1 was used for studies presented here GLC and “GLD” models: GLC and “GLD” models GLC-3T “GLD” version 1 parameter Inner radius of EM Cal is increased for better BR2
Coil radius is same, length icreased for better B field uniformity Calorimeter in Jupiter: Calorimeter in Jupiter An accuracy of simulated results depends on precision of Calorimeter geometry.
In the current version, we implemented an ideal geometry to know ultimate performance
pointing tower structure, no crack between towers
Mini-tower can be defined within the tower to define smaller cell structure. Simulate with a small cell structure and merge cell signals appropriately at later analysis stage to save CPU and storage space.
Tower geometry, materials and granularity can be modified easily. Tower structure EM HD Mini-Tower
Inside Tower Calorimeter Parameters: Calorimeter Parameters Default parameters
EM Cal: 38 layers of 4mm Pb + 1mm scinti.
HD Cal: 130 layers of 8mm Pb + 2mm scinti.
Transverse granuarity: 4cmx4cm(EM), 12cmx12cm(HD)
1cmx1cm EM CAL segmentation was tried, but it is CPU time consuming. Other method such as using parameterized geometry construction of Geant4 are considered.
Standard Geometry of Jupiter: Standard Geometry of Jupiter Solenoid TPC Hadron Calorimeter Muon/Iron Elemag. Calorimeter As of Nov. 4
By H.Ono(Niigata) GLD Performance of Jupiter : Performance of Jupiter With a standard configuration,
~6000sec to generate 100 ZH events at 350 GeV
Memory size: 60~100 MBytes
Output data size exceeds 2G Bytes at ~0.5K events
Need to reduce output data size or JSF update to support multiple-file output
Without output, run more than 5000 events
Problems/To do for future improvements
It takes ~ 3 min for the geometry initialization. Increase significantly when we use smaller segmentation.
Some improvement since Taipei by using Replica.
The standard cut value for the range in Calorimeter is 1mm. It has to be O(1) mm to get reasonable ECAL resolution of ~15%.
Missing detectors, such as forward trackers
For implementation of precise geometry, it would be better to use a XML geometry interface. EM Cal Performance: EM Cal Performance By A.L.C.Sanchez (Niigata U.) Linearity is good, but to get energy resolution similar to beam test results,
Randge cut of O(1) mm is required. It is very small and we don’t know why we need O(1) mm. HD Cal Performance: HD Cal Performance By A.L.C.Sanchez (Niigata U.) If the range cut as small as ~1 mm is used, few signals are generated in
scintillator by hadronic shower, though responces to electrons are OK. The range cut value greater than ~50 mm is necessary to get reasonable respince. Needs to find range cut values which meet EM cal and HD cal requirements. XML geometry interface: XML geometry interface Motivation of XML interface
Reduce detector attributes ( solid, material, size, etc ) hard coded in Jupiter
Reduce mother-daughter relations of geometries hard coded in Jupiter
Avoid “Fatty” Jupiter base class to support a newly customized “G4VSolid-factory-method”, otherwise base class becomes bigger and bigger.
How to achieve
Instead of manually writing C++ code using Jupiter classes, user prepares XML inputs, which will generate geometry structures dynamically using new base classes
XML parser, etc were developed and now trying to prepare a working set for a simple detector structure.
Hoping to use it near future for implementations of detail detector structures By S.Kishimoto (Kobe U.) Slide14: After(XML data) Before(mannualy intalling geometry C++ code) Satellites: IO, Metis, Leda: Satellites: IO, Metis, Leda Satellites are event reconstruction modules for simulated data. IO – Input / Output module set
Convert Jupiter output to ROOT object
Though ASCII file
JSFJ4 ( Run Jupiter in JSF frame work and convert)
METIS – Monte Carlo Exact hit To Intermediate Simulated Output
Module set for simulated data reconstruction
HitMaker : Exact hit to smeared hit
TrackMaker : Track reconstruction
ClusterMaker: Cal. Cluster reconstruction
PartcleFlowObjectMaker : Make Particle Flow Object
Leda – Library Extension for Data Analysis
Library for reconstruction
Kalman filter package, etc.
METIS Leda IO Status of Metis: Status of Metis Current aim is to prepare a minimum set of Metis modules for studies of Particle Flow Algorithm.
Novice users will be able to do physics analysis using information of PFO classes.
As a first step, a cheated track finder and a cluster maker, etc are in preparation in order to know ultimate performance.
Each module is independent, thus shall be easy to implement different reconstruction algorithm according to interests Metis Analysis Flow: make smeared TPC hits from exact hit make tracks from TPC make hybrid tracks ( TPC+IT+VTX) make smeared/merged CAL hits from exact hit make cluster from CAL hits make Particle Flow Objects jet clustering Metis Analysis Flow Cheated PFO analysis: Cheated PFO analysis ZH event at Ecm=500 GeV By K.Fujii(KEK), S.Yamamoto(GUAS), A.Yamaguchi(Tsukuba) - Exact hit points of TPC and CAL
Hits belong to the same PFO are shown with the same color
A framework of event display
in JSF is used. X3D: X3D ROOT’s X3d view of the same event X3D-Jet: X3D-Jet Same event, after a forced 4-jet clustering on PFObjects Metis Status: TPC sp/p - 1: Metis Status: TPC sp/p - 1 TPC exact hit points created by single m were fitted by Kalman filter package TPC parameter for
GLD_V1 configuration :
Bfield = 3 Tesla
Rin = 40 cm
Rout = 206 cm
Half length = 235 cm
# sampling = 200
srf = 150 mm (constant)
sz = 500 mm spt/pt2 (GeV -1) GLD momentum resolution
(TPC alone) By A.Yamaguchi(Tsukuba) Metis Status: TPC sp/p - 2: Metis Status: TPC sp/p - 2 If an effect of diffusion is included, … Without diffusion With diffusion By A.Yamaguchi(Tsukuba) Metis : Calorimeter sE/E: Metis : Calorimeter sE/E E(GeV) E(GeV) Energy resolution of EM/HD clusters attached to the PF objects
EM Cal: Pb(4mm)+Scinti(1mm): 38 layers
HD Cal: Pb(8mm)+Scinti(2mm): 120 layers Resolutions are worse, because the range cut is too large(1mm). New simulations are now in progress….
Entries near E~0 GeV are probably secondary gs and neutrons. They would be removed by cuts on signal TOF … under investigation By S.Yamamoto(GUAS) Jet mass – first try: Jet mass – first try Using PFOs, we clustered jets and plots Higgs mass.
Neither event statistics nor tuning/debugging of Jupiter/Satellites are not sufficient. But we are getting close to do physics studies using these tools
By K.Fujii(KEK) Mh=120GeV Conclusion: Conclusion GLD_V1 geometry has been implemented in to the Jupiter.
Packages in Satellites have been enriched. Now it can create Particle Flow Objects based on cheated algorithm. Our hope is to know the ultimate performance of the detector based on these packages soon.