Programma lezioni - INFN sezione di Lecce

Download Report

Transcript Programma lezioni - INFN sezione di Lecce

Simulation of HEP experiments
the Geant 4 toolkit
Tommaso Boccali
SNS, INFN Pisa
21/07/2015 04:47
23/5/2004
Outline

Simulation – a step towards physical
understanding

How do we simulate today a HEP experiment?

The Geant4 toolkit
21/07/2015 04:47
23/5/2004
2
Why?

First of all, why do we simulate at
all?


A lot of experiments in the past were
without detailed simulations … and
they worked!
Short answer: today’s theories do
not predict such an overwhelming
hint from data in most interesting
measurements

And, moreover, we want are interested
not only in discovery - see what LEP has
done
21/07/2015 04:47
23/5/2004
J/Psi discovery
3

Detailed tests of theories: when
the precision level required for
validating/discarding a theory is
high, effects due to the “bias”
introduced by the detector are
not negligible




Finite resolution
non-homogeneity
Particle misidentification
such that we cannot simply look
at the data and say “Oh, it is
clear that the answer is X”
21/07/2015 04:47
23/5/2004
“B” events
4
simulation
theory
data
simulation
21/07/2015 04:47
23/5/2004
5

So, the more subtle is the hint we want to look at,
the more detailed the simulation must be


We have to thrust the background distributions to a level
much greater than before
We need a tool capable to simulate (realistically)
particle interactions with matter from the scale


of the TeV: primary particles generated in LHC collisions
of the keV: delta rays from secondary (tertiary…) particles
leaving energy for ionization in our detectors
Correct physics simulation needed
for ~ 9 orders of magnitude
21/07/2015 04:47
23/5/2004
1 TeV
9
 10
1 keV
7
Geant 4: what is it?

Once upon a time … a
Fortran code called Geant
3 was used to simulate
virtually all the HEP
experiments of the ‘80-’90

Very well tested, at the heart
of the LEP Simulation


Agreement in physics up or
better than % level
A product at the end of
lifetime …


Fortran is not suited for a
10x complexity
Fortran is not maintainable
for ~20 years
21/07/2015 04:47
23/5/2004
Visible energy in ALEPH
(calo+tracker), DT vs MC
8
Geant 4

The project started in 94 with the aim to
completely rewrite Geant3

To use new technologies (C++)
To implement all the lessons learned
To use new data and ideas for calibrations

March 25th 2004: Geant 4 6.1 Release


21/07/2015 04:47
23/5/2004
9
HARP
Geant4
Univ. Barcelona
Lebedev
Helsinki Inst. Ph.
21/07/2015 04:47
23/5/2004
PPARC
Collaborators also from nonmember institutions, including
Budker Inst. of Physics
IHEP Protvino
MEPHI Moscow
Pittsburg University
10
What exactly is G4?
It is not a program … there is nothing to run
but examples when you download it
 It a is a toolkit: a set of libraries which know
how to




Handle a complex geometry
Treat decays / energy loss of all the common
particles
Extend its capabilities with user code
New physics inputs
 New features: visualization, data analysis

21/07/2015 04:47
23/5/2004
11
Analogy with a real experiment
Define a Geometry
Construct the detector
Well, usually not
done by us ;)
Define physics
Define primary particles
(the beam)
Build an accelerator
Beam ON!
(N events)
Switch it on
Let it run, and save data
Plot histograms, and eventually
21/07/2015
04:47
23/5/2004
publish
physics papers
Analyze data
12
Geometry
Define a detector
1.
1.
Shapes of each single component (a
screw, a silicon wafer, a power
cable)
1.
2.
3.
2.
3.
4.
Box
Cylinder
…
Hierarchical placement of each
component
Materials used
Define passive and active materials,
instrumented with measurement
capabilities
1.
For these, define format of read
data, as close as possible to the real
data taken with the apparatus
21/07/2015 04:47
23/5/2004
Define a Geometry
Define physics
Define primary particles
(the beam)
Beam ON!
(N events)
Analyze data
13
Complex facilities

Not only easy volumes, but also

Polycones


Useful for HEP detectors,
useful with cylindrical symmetry
Boolean solids
G4UnionSolid
21/07/2015 04:47
G4SubtractionSolid
23/5/2004
G4IntersectionSolid
14
Level of complexity…

CMS Tracker Simulation




900 different components defined
20000 sensitive detector
200000 global objects in the simulation
600 materials and composite materials
21/07/2015 04:47
23/5/2004
15
Materials…

For some kind of “easy” physics, the definition of
materials provides enough informations:
double density = 1.390*g/cm3;
double a = 39.95*g/mole;
G4Material* lAr =
new G4Material("liquidArgon",z=18.,a,density);
a = 1.01*g/mole;
G4Element* elH =
new G4Element("Hydrogen",symbol="H",z=1.,a);
a = 16.00*g/mole;
G4Element* elO =
new G4Element("Oxygen",symbol="O",z=8.,a);
density = 1.000*g/cm3;
G4Material* H2O =
new G4Material("Water",density,ncomp=2);
H2O->AddElement(elH, natoms=2);
H2O->AddElement(elO, natoms=1);
21/07/2015 04:47
23/5/2004
16
…

Materials also define automatically the “material
budget of a detector”


“Quantity of matter” seen by a particle starting from the
interaction point
Two parameters are used:

Electromagnetic interactions: X/X0 is the number of
radiation lengths of the given detector


A radiation length is the distance in which an (high energy)
electron loses all but 1/e of its energy
Hadronic interactions: l/l0 is the same due to hadronic
effects (nuclear interaction length)

The same for nuclear interactions; used in hadronic calorimeters
21/07/2015 04:47
23/5/2004
17
CMS Tracker
21/07/2015 04:47
23/5/2004
18
CMS Hadron Calo
21/07/2015 04:47
23/5/2004
19
Physics…

It is completely unrealistic to
try and develop a physics
model which covers different
particles and energy ranges…


Like: model for LHC physics,
model for TeVatron physics
Please note it now: it is NOT
the LHC physics (qqHqq);
it is only the physics of
particle-matter interaction
21/07/2015 04:47
23/5/2004
Define a Geometry
Define physics
Define primary particles
(the beam)
Beam ON!
(N events)
Analyze data
20
A different approach…

G4 allows more than a single
physics model to contribute to the
simulation of the interactions
Physics processes are defined with



Energy range

a given energy range
a given particle type
By combining these wisely and
paying some attention at the
overlaps, a model for the
interesting field can be
constructed
21/07/2015 04:47
23/5/2004
Hadronic inelastic
interactions
E>75 GeV
E<100 GeV model
model
Particle type
Process type
21
But …

This needs expertise from the user


It is easy to forget processes and so generate
invalid simulations
Often more than 1 physics model is available for
the same range:

You can decide to use the more detailed or the fastest…
at a price!
compt: Total cross sections from a parametrisation.
Good description from 10 KeV to (100/Z) GeV.
21/07/2015 04:47
23/5/2004
23
Concept of “range”

Given a process, the user can define how accurate it has to be


The more accurate = the slower !
Consider Brem photon emission in a material


The charged track loses energy, and a photon is emitted.
In principle, infrared divergence which would lead to infinite # of
photons of E0
21/07/2015 04:47
23/5/2004
24
Range



We have to specify a minimum range for the process:
the charged track always loses correctly energy, but
we can veto the creation of photons with Range
smaller than a threshold
Passive materials: if the track goes inside a block of
iron, there is no point in generating particles which
would not exit from that – set range of the order of
the volume size.
In an active detector, no need to simulate particle
which would give effects below the experimental
resolution.
21/07/2015 04:47
23/5/2004
25
Calorimeters

Classic example, an electron enters the
calorimeters



Where/when to stop the showers?
In principle, G4 could simulate electrons down to
O(eV) – hours needed to simulate an event
What is usually done is to try and simulate events
with increasing cuts, and define a point in which
the price performance ratio is considered
acceptable
21/07/2015 04:47
23/5/2004
26


500 MeV proton in a LAr-Pb
(ATLAS) sampling calorimeter
Range at 1.5 mm



LAr
Pb
LAr
Pb
Corresponds to 450 keV in LAr
Corresponds to 2 MeV in Pb
Silicon detectors in CMS


Strip pitch in ~100 mm
No need to produce delta rays
which would travel less than 10 mm
21/07/2015 04:47
23/5/2004
27
A few examples…
●
●
●
EM physics
“standard” processes valid from ~ 1 keV to ~ PeV
“low-energy”
valid from 250 eV to ~ PeV
optical photons
Weak physics
decay of subatomic particles
radioactive decay of nuclei
Hadronic physics
pure hadronic processes valid from 0 to ~100 TeV
m and e-nuclear
valid from 10 MeV to ~TeV
●
Parameterized or “fast simulation” physics
●
Tracking physics
●
Particle transportation in the field
21/07/2015 04:47
23/5/2004
28
A review…


Standard em processes:
Gamma




Photo-electric effect
Compton scattering
Electro, muon pair
production
Electron




e ionization
e bremsstrahlung
e+e- annihilation
Syncrotron radiation

Muons




Charged hadrons





23/5/2004
Hadron ionization
All charged particles

21/07/2015 04:47
mu ionization
Mu bremsstrahlung
e+e- pair production
Multiple scattering
Transition radiation
Scintillation
Cerenkov radiation
29
A review…


Hadronic processes:
At rest







Different processes for
each hadron
Photo-nuclear, electronuclear, muon-nuclear
Ions (for example, Fe+++)

Neutron capture
Fission

Same processes for all the
hadrons
Inelastic
Capture

Elastic


Stopped mu, p, K, antiproton

Neutron-induced, deexcitation
Most of these are data
driven – eagerly waiting
for declassified data!
Radioactivity!
21/07/2015 04:47
23/5/2004
30
Hadronic models
Process
At rest
In flight
Cross sections
Direct impl.
Models
Direct impl.
Direct implementations
Isotope production
Theory framework
High energy
String parton
Direct impl.
21/07/2015 04:47
Direct impl.
Direct impl.
Spallation framework
Transport utility
String fragmenation util.
Direct impl.
Event biasing
Precompound
Evaporation util.
Frag function intfc
Frag function impl.
23/5/2004
Cascade
Direct impl.
Direct impl.
Direct impl.
31
Processes

Basic distinction:

Continuous processes: ionization, multiple
scattering


Energy loss is continuous and not localized
Discrete: decay, Compton scattering, annihilation

These are the ones limiting the step length
21/07/2015 04:47
23/5/2004
32
How does it work?


When a particle is under study, all of the possible
energy losses mechanisms and decay processes must
be able to “interfere” with the particle
A step is a “time” quantum for a particle traveling in
a material
Start of step point
21/07/2015 04:47
End of step point
23/5/2004
33
The Step



The step length is limited by the processes of the particle: for
example, after a few mm, the particle can decay
Each physics process must propose a step length
 Given a lifetime for a decay, a random number is thrown
and the resulting distance before decaying is returned
The “physical step length” is the minimum of all proposed
lengths
If the minimum between that step and the distance to the
next detector is taken
21/07/2015 04:47
23/5/2004
34
The beam

Well, not really the beam! In
most HEP cases, Geant 4 is not
starting from e+e- (pp), but it is
interfaced via to specialized codes
for the simulation of primary
interactions






Pythia
CompHEP
ALPGEN
…
The beam is a single particle thrown
inside the experimental setup
21/07/2015 04:47
Define physics
Define primary particles
(the beam)
Geant 4 is able to read directly
the output of these programs,
and treat these particles as “the
beam”
Exception: test beam simulations

Define a Geometry
23/5/2004
Beam ON!
(N events)
Analyze data
35
How an event is processed
List of primary
particles
List of particles
to be tracked
Track next
particle
Push secondaries
in the list
Interaction
particle-matter
Out of the world,
decayed or stopped
21/07/2015 04:47
23/5/2004
36

Some volumes are special:
they are “active” in the real
detector.


Data is read from these in real
world
Geant 4 allows to simulate
their response:


Special user code is needed
The output of this code must be
as close as possible to the real
response

Usually experimental
parameterization used
21/07/2015 04:47
23/5/2004
ADC counts
Simulation of the response of a
detector
Energy deposit
in the strip
37
Visualization

With complex geometries, often the best way
to debug it to see …


Holes in the geometry
Overlapping volumes
Geant 4 provides natively (3) visualization
features, most notably OpenGL
 This means geometry, tracks, etc can be
visualized without any user code

21/07/2015 04:47
23/5/2004
38
Visualization
LCD in JAS3
BaBar Offline
BaBar Online
Geant4
GLAST
21/07/2015 04:47
23/5/2004
39
21/07/2015 04:47
23/5/2004
40
Validation
So, most of the HEP community is / will be
using Geant4
 The standard of the Physics simulation must
be very high

Huge effort on validation
21/07/2015 04:47
23/5/2004
41
Compton Scattering
Data, G4 LowE , G4 standard
21/07/2015 04:47
23/5/2004
42
Rayleigh Scattering
Data, G4 LowE
21/07/2015 04:47
23/5/2004
43
Stopping Power
Data, G4 LowE, G4 Standard
21/07/2015 04:47
23/5/2004
44
Pair Production
Data, G4 LowE, G4 Standard
21/07/2015 04:47
23/5/2004
45
Multiple Scattering (1)
21/07/2015 04:47
23/5/2004
46
 Production from 730 MeV p
(LEP Model)
21/07/2015 04:47
23/5/2004
47
 Production from 730 MeV p
(Bertini Model)
21/07/2015 04:47
23/5/2004
48
From the experiments point of
view…

Use test beam data
21/07/2015 04:47
23/5/2004
CMS HCAL
49
Crystal
25
1717
h
8
7
6
5
4
3
2
1
21/07/2015 04:47
f
23/5/2004
50
Just a glimpse …

Geant 4 is aimed / supported / funded mostly by HEP
experiments, but is is becoming a de-facto standard

Medical applications



Space applications




Simulation of radiotherapic apparata
Simulation of dose absorption in human tissues
Detectors
Damages from high energy cosmic rays
Effects of cosmic rays on astronauts
Industrial applications


CCD of digital cameras
Implantation of oxides in silicon wafers
21/07/2015 04:47
23/5/2004
51
Simulation


Simulation of dose deposit by radiotherapy in human
tissues
Very useful to plan therapy cycles…
21/07/2015 04:47
23/5/2004
52
 astrophysics
-ray bursts
GLAST
NASA mission
AGILE
GLAST
GLAST
Typical telescope:
Tracker
Calorimeter
Anticoincidence
 conversion
 electron interactions
 multiple scattering
 d-ray production
 charged particle tracking
21/07/2015 04:47
23/5/2004
53
Conclusions
There are few doubts that Geant 4 will the the
tool used to simulate HEP experiments at least
for the next decade
 It is still under validation, but



Already delivers physics performance better or
equal to the Geant 3 ones
It is almost as fast

Not an easy achievement Fortran vs C++
21/07/2015 04:47
23/5/2004
54
Where to start from?

Geant 4 provides a nice & easy roadmap



Complete class documentation


http://conferences.fnal.gov/g4tutorial/
A complete CD is present, in Linux and WinXP
versions
http://cern.ch/geant4
Packaged Examples:
A number of easy examples there
 If we have time, we can try one…

21/07/2015 04:47
23/5/2004
55