[PPT] - GEANT4 CMS SI MULATI ON Pedro Arce (CERN/ CI EMAT) (on behalf of PowerPoint Presentation

SLIDE 1

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

1

GEANT4 CMS SI MULATI ON

Pedro Arce (CERN/ CI EMAT) (on behalf of CMS collaboration)

SLIDE 2

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

2

OSCAR functionality
OSCAR/GEANT4 robustness and performance
Physics comparisons

Outline

SLIDE 3

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

3

All CMS det ect ors
Also several Test beams
Current ly moving t o geomet ry in XML dat abase, common t o

Simulat ion/ Reconst ruct ion/ Visualisat ion

All geomet ry convert ed f rom GEANT3 geomet ry
Building of XML geomet ry int o GEANT4 ready
Already can be visualised wit h I GUANA (heavily used f or debugging)
Missing: det ect or numbering schemes wit h new geomet ry
Will need t o review t he use of ‘Set Smart Less’ (3 year s old)
Det ailed checking quit e advanced
Thorough comparison of GEANT3 and GEANT4 geomet ry
All mat er ials, volumes and t ouchables pr oper t ies
Posit ions of t ouchable cent r es
Mat erial budget
CMS Overlap Det ect ion Tool (now in GEANT4)

OSCAR f unctionality: Geometry

SLIDE 4

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

4

XML based Detector Description Language (XML DDL)

Single description inter-

face for the ideal detector

Generic description

. Materials . Geometry . Component approach

Attaching specific

Information to . Components . Subsets of Volumes

Human editable/readable

/expandable

Geometry f or OSCAR

SLIDE 5

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

5

View of CMS muon system Sliced view of CMS barrel detectors View of 180 Higgs event simulated in CMS Tracker detector

SLIDE 6

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

6

OSCAR f unctionality: Primary generator

CMS part icle gun:
Scan in θ, η, energy
Random (f lat/ gaussian) in θ, η, energy
I nt er f ace wit h HEPEVT ASCI I f iles
I nt er f ace wit h PYTHI A nt uples
Long- lif e particles:
I n GEANT3 only t he decay products are simulat ed
I n GEANT4 t he part icle is simulat ed bef ore decaying (and t he decay product s

are kept )

Advant age: you t ake int o account t he possibilit y of int eract ions, ener gy

loss, magnet ic f ield bending, et c. bef ore decaying

The lif e t ime is not kept
The moment um of decay product s is not kept
OSCAR: user can select t o simulat e t he part icles bef ore decaying or simulat e

decay products (keeping t he lif e t ime)

Need deeper st udy t o underst and what will be t he f inal solut ion

SLIDE 7

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

7

OSCAR f unctionality: Persistency

Hits f or all detectors
Primary generator particles
Selected tracks
Use CMS f ramework (COBRA)
Fully compat ible wit h Reconst ruct ion
Using Objectivity, moving t o ROOT (end 2002)

SLIDE 8

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

8

OSCAR f unctionality: Visualisation

I nteractive visualisation is done wit h I GUANA

Visualise any GEANT4 Geomet r y
Select by logical volumes / by physical volumes
P

icking, slices

Volume propert y window
GEANT4 command line
A wizar d t o guide t hr ough OSCAR settings
Adding ot her ext ensions is t r ivial (e.g. GUI s like over lap det ect ion)
Visualise Tracks
Visualise magnet ic f ield
I nt egr at ed wit h over lap det ect ion
Find overlaps, show result det ails in a list
Highlight overlaps in 3D
Hit s, in t he near f ut ure

SLIDE 9

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

9

OSCAR - I GUANA

Correlated Picking Example extension (a trivial wizard) Queried from plug-in database, loaded on request and bound to IGUANA G4 Run Manager IGUANA viewer displaying OpenInventor scene Control of arbitrary GEANT4 tree Overlap detection

SLIDE 10

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

10

OSCAR - I GUANA: magnetic f ield

2D plane t hat can be rot at ed
Field direct ion as an arrow
Field int ensit y as colour

SLIDE 11

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

11

OSCAR f unctionality

Physics:

Elect r omagnet ic pr ocesses needed at LHC
Hadronic pr ocesses needed at LHC

CMS Physics Technical Design Report is planned to be written using OSCAR (due end 2005) OSCAR milestone Spring/ Summer 2003:

“Repr oduce t he physics r esult s of t he GEANT3 simulat ion wit h similar per f or mance”

Physics Cuts:

Production cut s per material
Minimum kinetic energy cut s per material
St ill under invest igat ion if we really need t hem
Time of f light cut

SLIDE 12

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

12

OSCAR f unctionality

UserActions:

Several user act ions of the same type loadable on demand

Framework:

COBRA manages t he main() and t akes car e of persistency
Now OSCAR is being r eengineer ed t o achieve f ull integration

wit h CMS Fr amewor k (see wednesday t alk by M. St avr ianakou)

Documentation:

User’s Guide updat ed f or each r elease
doxygen document at ion
Several Tutorials on t he web
For OSCAR developers
For Physics Reconst ruct ion sof t ware developers
For Summer St udent s

SLIDE 13

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

13

OSCAR f unctionality

Future developments:

GEANT4e as GEANE r eplacement (er r or pr opagat ion f or

t rack f it t ing)

Under pr ogr ess
GFLASH f or f ast simulat ion
RTAG/ G4 discussion

Summary:

OSCAR is ready f or physics validation studies

SLIDE 14

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

14

Several problems f ound:

Part icles loop bet ween t wo volumes

⇒ push t hem by 0.01 mm

Several G4Except ion f ound:
solved by CMS t eam or/ and wit h t he help of GEANT4

t eam ⇒>10000 pp events wit hout crashing

NOTE: new G4Except ion f ound wit h GEANT4.4.1 (under invest igat ion)

OSCAR/ GEANT4 robustness

SLIDE 15

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

15

Full event s: H→ττ→µ /

t t →e / Z→ττ

Cut s in primary part icles: Pt : 1GeV, | η| <

2.4 (3.0)

Full CMS geomet ry
3D TOSCA magnet ic f ield
Product ion and t racking cut s as in CMSI M (GEANT3)
GEANT4 voxel navigat ion
GEANT4.4.0.r ef 02
Pent ium I I I 850 Mhz
Checked t hat t hey are really t he same event s (see next )

GEANT4 vs GEANT3 Time Perf ormance

SLIDE 16

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

16

G3- G4 events detailed comparison

No Tracks MeV

H→ττ→µ in all CMS (10 Event s)

No Tracks mm

I nitial track energy Track length

SLIDE 17

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

17

OSCAR

(427136 tracks)

G3- G4 events: detailed comparison

R (mm)

CMSIM

(442828 tracks) Z (mm) Z (mm) R (mm)

Origin of secondary tracks: R vs Z

SLIDE 18

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

18

Time perf ormance

H→ττ→µ (10 event s)

CMSIM: 439k tracks 64 sec/ evt OSCAR: 400k tracks 117 sec/ evt (1. 83)

t t →e (10 event s)

CMSIM: 1809k tracks 184 sec/ evt OSCAR: 1159k tracks 425 sec/ evt (2. 31)

Z→ττ (10 event s)

CMSIM: 1105k tracks 96 sec/ evt OSCAR: 941k tracks 248 sec/ evt (2. 58)

SLIDE 19

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

19

RecHits

Number of rechits per track vs eta

Cmsim 122 OSCAR 1 3 2 pre 03

100 GeV P

t muons in region | η| <

2.5

resolution vs eta

CMSI M(GEANT3) - OSCAR(GEANT4) comparison

Central Tracker

SLIDE 20

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

20

CMSI M(GEANT3) - OSCAR(GEANT4) comparison:

Muon System

reconst ruct ed – generat or quant it ies 5-200 GeV P

t muons in region | η| <

2.4

SLIDE 21

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

21

Big dif f erence bet ween reconst ruct ed and generat ed t rack

par amet er s was f ound because t r acking pr ecision par amet er s wer e t oo high.

‘deltaI ntersectionAccuracy’

is by def ault 100 µm

CMS t racker precision 20 µm
CMS Muon prcision 200 µm
Has t o be set t o 0.1 µm
Time j ust increased by 3%

I mportance of tuning tracking precision parameters

SLIDE 22

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

22

CMSI M(GEANT3) - OSCAR(GEANT4) comparison:

EM calorimeter

SETUP:
CMS elect romagnet ic calorimet er: lead tungstate crystals

point ing t o int eract ion point

Photon of E = 30 GeV ent ers f ront f ace of cryst al
No magnet ic f ield
What was compared:
E in hot t est cryst al, in 3x3, in 5x5 mat rix
E t ot al
dNhit s/ dt , f or 1 ns t ime slices
GEANT3.21 vs GEANT4.4.0.p02

SLIDE 23

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

23

CMSI M(GEANT3) - OSCAR(GEANT4) comparison:

EM calorimeter

Shower containment:

Fr om GEANT4 galler y: Elect ron 5 Gev in PbWO4

(more detailed)

GEANT3 looks wider t han GEANT4
Fr om old t est beam GEANT3/ GEANT4 compar ison

CMS Note 1997/037 by A.Givernaud: good agreement with G3 for shower profile CMS Note 1998/031 by K.Lassila-Perini: G3 shower is a bit wider than real shower

SLIDE 24

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

24

CMSI M(GEANT3) - OSCAR(GEANT4) comparison:

EM calorimeter

Hit timing:

GEANT4 shower has ~8% less

hit s t han GEANT3 shower I t doesn´ t cont r adict wit h GEANT4 galler y plot below, however direct compar ison is needed:

SLIDE 25

GEANT4 workshop 02 CMS GEANT4 Simulation Pedro Arce (CERN/ CI EMAT)

25

ECAL + HCAL testbeam :

GEANT3 - GEANT4 comparison 1996 Testbeam Setup f or CMS HCAL

A t est module of copper absorber plates wit h scintillator t ile

sampling

28 scint illat or plat es wit h absorber of varying t hickness in bet ween
A pr ot ot ype lead t ungst at e crystal elect r omagnet ic calor imet er
Energies: 225 GeV µ (f or calibr at ion), 10-300 GeV π, 10-300 GeV e-
Magnetic Field: 0, 0.75, 1.5, 3 t esla (dir ect ion par allel t o t he f ace of

t he scint illat or)

Conf iguration: Only HCAL or ECAL + HCAL
∼5000 event s wer e t aken f or each set up
Response of t he Calor imet er was st udied as a f unct ion of :
Magnet ic f ield: ef f ect on scint illat or
Absorber t hickness: opt imisat ion of resolut ion versus cont ainment
Absorber dept h: energy cont ainment
Elect romagnet ic Calorimet er cont ribut ion: e/ π ef f ect s

Some Test Beam compar isons st ar t ed:

– Show reasonable agreement wit h dat a – Slight ly bet t er t han GEANT3 – Ot hers t o be done: ECAL TB ‘99, ECAL TB ‘02, ECAL+HCALTB ‘99, HCAL ‘02, Tracker ´ 02

Next mont hs: syst emat ic compar ison wit h GEANT3 simulat ion