Atlas software

advertisement
Atlas software developments
Stan Bentvelsen
Layout

Atlas software status



Building simulation



Geant4 validation
Detector Description
Reconstruction



Milestones
Frameworks
Event definition
Track reconstruction
Preparations for physics

Analysis plans
Friday, December 14
NIKHEF Annual Meeting
2
Timeline towards start LHC
2001
04
06
12
12
2002
–
–
–
–
2003
2004
2005
2006
Simulation: Geant4 validation
06 – Physics readiness report
Major migration to OO software (no Fortran!)
07 – Full chain real environment
Core software agreements complete
12 – Full database infrastructure
Data Challence 0 (‘continuity’ check)
04
08
12
12
–
–
–
–
DC1 (large scale reconstruction & analysis)
Computing
TDRprocessing
finished farm
01 – 40%
10% 06
processing
farmtunes
– LHC beam
100 TB dBase prototype
04 - LHC pilot run
06 – Decide
(linux?)
06 -OS
100%
processing farm
07 – DC208
(increased
complexity)
– LHC fist
physics run
Friday, December 14
NIKHEF Annual Meeting
3
Framework schism in Atlas
Development of the C++ core software:

Reconstruction/analysis:
Athena







Geant4 based Object
Oriented Folly
Based on Geant 4 ‘event
loop’
Developed further by
American groups (LBL)
Keywords:

Simulation:
FADS/Goofy
Gaudi Framework from
LHCb


Developed by European /
Japanese collaborators
Keywords:
Event store on which
algorithms and services act
Transient/persistency split
Thoughtful design



Extremely lightweight
Efficient
Little default behavior
Objectivity / ASCII
Friday, December 14
NIKHEF Annual Meeting
4
Geant 4 physics validation

Major milestone in ATLAS Simulation.


The goal of the validation is to increase trust in Geant4 simulation
(as much as G3) such that it can be used for data production.
1)
Collaboration projects with the Geant4 team.
2)
ATLAS’ own internal activity.
Status of Geant4 similar to status of Geant3 before LEP
started (around `85)



Geometry workable (one annoying bug)
Electromagnetic physics fine (upto % level)
Hadronic physics less well described
Friday, December 14
NIKHEF Annual Meeting
5
Muon Detector
EM shower production by muons in absorber: extra hits in Muon Drift Tubes
• Transverse distance of extra hits from muon track in Geant-4 broadly
reproduces test beam data
• Detailed agreement better for lighter absorber material
Friday, December 14
NIKHEF Annual Meeting
6
Tile Calorimeter
G4
G3
Geant3 vs data
data
100 GeV
Geant4 vs data
20 GeV
Electrons
Muons
•No perfect match between G4 and test beam data
Friday, December 14
NIKHEF Annual Meeting
7
FADS/Goofy robustness test

Geant4 simulation test (done before Xmas01)

100 job-streams each of which processes 1,000 events
Objectivity Database
HepMC
Objectivity Database
FADS/Goofy
Hits/
Stream #001
Hit-collections
MCTruth
FADS/Goofy
Hits/ Stream #002
Hit-collections
MCTruth
FADS/Goofy
Hits/ Stream #100
Hit-collections
MCTruth
H-> 4 leptons
Pythia 6
1,000 events
HepMC
1,000 events
HepMC
Friday, December 14
1,000 events
NIKHEF Annual Meeting
8
FADS/Goofy robustness test

Expected data size and CPU required
4-vectors database
Hits/Hit-collections
database
CPU
(Pentium III, 800MHz)

per event
~ 50 KB
1,000 events
~ 50 MB
~ 1.5 MB
~ 1.5 GB
~ 60 sec
~ 17 hours
Required resources

PC farm
~ 10 CPUs ( 5 machines with dual processors)

Disk space
~ 155 GB

Process period
~ 1 week
Friday, December 14
NIKHEF Annual Meeting
9
Detector description scope
Geometry
description
Materials
Identifiers
Conditional
dBase
Simulation
Detector
Description
framework
Digitalization
Visualization
Reconstruction
Alignment
provider
Friday, December 14
Fast simulation
client
NIKHEF Annual Meeting
10
DD of ideal static detector
Define geometry description
for ATLAS in ascii format

E.g. geometry dBase contains

Volumes
Use XML





Unambiguous storage of
numbers with their meaning
Customized syntax
Many free software tools









Parsing, validating

Widely accepted standard
Geometry dBase
Logical description
(Identifier)
Material dBase
Friday, December 14
Elementary solids
Boolean solids
Logical groupings of volumes

Iteration
Positioners
AGDD syntax

Generic and descriptive
Description of the instances of
the volumes



NIKHEF Annual Meeting
Absolute or relative, single or
multiple
Define the translation and
rotation matrix wrt origin of
the constituents
Utilize symmetries
11
Example: XML implementation
<?xml version="1.0"?>
<!DOCTYPE AGDD SYSTEM "AGDD.dtd"[
ENTITY Materials SYSTEM “Materials_AGDD.xml”
]>
<AGDD DTD_version=”v4”>
<!—Include the materials 
&Materials;
<!--
AGDD Header, version of DTD
Section: sub-detector + author + version
Workshop example file
**********************
-->
<section name
=
version
=
date
=
author
=
top_volume =
DTD_version=
"WK"
"1.0"
"Wed Oct 27"
"Detector Description group"
"ATLAS"
”v4”
>
<tubs name="WK_tube" material="Aluminum" Rio_Z="0. 15. 1000." />
Section: detector, version, author
Creation solid: tube
<composition name="WK_layer" >
<mposZ volume="WK_tube" ncopy="10" dZ="30." rot="0 90 0" />
</composition>
<composition name="WK_multilayer" >
<posXYZ volume="WK_layer" X_Y_Z="
<posXYZ volume="WK_layer" X_Y_Z="
<posXYZ volume="WK_layer" X_Y_Z="
<posXYZ volume="WK_layer" X_Y_Z="
</composition>
0
0
0
0
0 0" />
30 0" />
60 0" />
90 0" />
Various positioning of volumes
<composition name="ATLAS" >
<posXYZ volume="WK_multilayer" />
</composition>
</section>
</AGDD>
Friday, December 14
NIKHEF Annual Meeting
12
AGDD: parametric description

Problem with explicit description



Many redundant numbers
Allow for parametric description
“Compactification”



AGDD.dtd
Syntax def

Compact AGDD
XML
factory
Specialized XML syntax
Contain parameters as
attributes
Associated C++ algorithms to
unpack
Registration of algorithms to
create transient objects
ASCII file
AGDD_Printer
C++ Algorithms
compact
ASCII file
Generic Model
Transient C++
‘expanded’
Generic AGDD
XML
Client programs
Detector description
Friday, December 14
NIKHEF Annual Meeting
13
The Geant4 client

Build the Geant4 geometry from the information stored in
XML
AGDD
syntax
‘Generic’ translation

Automatic translation; blind
to the specific detector


Need explicit description
G4Builder
‘Detector specific’
translation

Need algorithmic (compact)
description
Close connection to ATLAS
simulation group
Friday, December 14
Syntax
check
Syntax
check
Compact
description
Optimized
code
Expand()
Explicit
description
Automatic
building
Geant4
Geant4
Compare
representation performance representation
NIKHEF Annual Meeting
14
SCT gallery
Friday, December 14
NIKHEF Annual Meeting
15
Muon gallery
Friday, December 14
NIKHEF Annual Meeting
16
Status AGDD in ATLAS
Work ongoing to create AGDD
interface for all sub-detectors in
ATLAS.
To agree among large
community of people is not
trivial!
Friday, December 14
Pixel
SCT
TRT
Presampler
Larg-EM
Larg-FCAL
Larg-HEC
Tile
Muon
NIKHEF Annual Meeting
yes
yes
yes
~20%
yes
yes
yes
yes
yes
~50%
yes
~50%
17

Simulation of the cosmic ray setup (at NIKHEF)
Frank’s description of the geometry
NIKHEF Annual Meeting
Friday, December 14
18
Friday, December 14
NIKHEF Annual Meeting
19
All materials
Cosmic ray material-budget

MDT tubes
Standard tool (in G4Builder)


Aluminum
3d scan in (eta-phi) of
geantinos
Connection hits to digits
available
Radiation
length
Volume
counter
Friday, December 14
NIKHEF Annual Meeting
deg
20
Digitization
D Calvet

Common frame for SCT and
PIXEL

From hits to digits




Surface charge
Charged diodes
Digits
Detector element

Local description of



SCT strips
Pixel diodes
Modular design

‘Plug-in’ parameterizations

Detailed versus fast simulation
Friday, December 14
NIKHEF Annual Meeting
21
Silicon wafer local description
A Fornainii
- Strips/pixels will not be separate volumes in G4
- Separate packages describing readout structure
-segmentation into diodes
-connection to readout electronics
Pixel module
Friday, December 14
Barrel strip module
NIKHEF Annual Meeting
Endcap strip module
22
Event model

Collection of digits


Each collection connected to one
‘detector element’
Each digit connected to one
‘readout channel’
Detector-element description from XML
Comparison to test beam data
Threshold scan well described by the
simulation
Friday, December 14
NIKHEF Annual Meeting
23
Track reconstruction


AMBER: first full OO tracking algorithm (P. Hendriks, 2000)
COBRA: COmBined Reconstruction (W. Lavrijsen)

Global generic track fitting using



Track segment refit inner tracker and muon chambers
Extrapolation and matching from track segments
Global track refit
Friday, December 14
NIKHEF Annual Meeting
24
COBRA comparison

COBRA performs well
COBRA vs IPATREC

COBRA vs MUONBOX
Comparable results for IPATREC
(inner tracker)
Pt
Friday, December 14
Comparable results for MUONBOX
(muon detector)
NIKHEF Annual Meeting
25
Higgs reconstruction
Golden channel at MH<400 GeV benefit from
combined reconstruction
inner
Friday, December 14
muon
comb (%)
130 GeV
1.57
2.55
1.18
170 GeV
1.50
2.36
1.35
200 GeV
1.78
2.45
1.61
400 GeV
20.6
21.8
20.2
NIKHEF Annual Meeting
26
ATLAS physics at NIKHEF

Main themes of interest have been defined


Higgs, SUSY, Top, B-physics
NIKHEF staff will become active players in these fields


PhD students:






Develop NIKHEF toolkit, eg track reconstruction, jet definitions,
front-end analysis.
Higgs -> 4 leptons (Lavrijsen, Peters, Woudstra)
SUSY (Buis)
Observation of Bc, production model (Scholte)
Bs mixing, CP violation (Fornaini)
Single top with polarization, Vtb and mass (Barisonzi)
Students in the group:

G. Lim, Y. Boer, F Jansen, B Dirks, J Dalhuizen
Friday, December 14
NIKHEF Annual Meeting
27
Conclusions
Big challenge to get the Atlas software ready at
startup of LHC
Middle of development




Transition to OO software paradigm
Detector description going
Geant4 toolkit validated. Atlas simulation underway.
Reconstruction/analysis chain getting up steam
Friday, December 14
NIKHEF Annual Meeting
28
Download